Case-Control Study of Ovarian Cancer and Polymorphisms in Genes Involved in Catecholestrogen Formation and Metabolism¹

The decreased risk of ovarian cancer associated with gravidity, breast-feeding, and oral contraceptive pill use support a role for steroid hormones and gonadotropins in ovarian carcinogenesis. Hypotheses regarding the hormonal etiology of ovarian cancer have focused primarily on estrogenic stimulation of entrapped ovarian epithelium in inclusion cysts and the increased risk of malignancy associated with repeated cellular damage of the surface epithelium during ovulation (1). However, in addition to its influence on growth promotion, estrogen can be metabolized to form catechol intermediates that cause oxidative DNA damage, lipid peroxidation, and (indirectly) DNA adducts (2, 3, 4). Estrogen metabolism occurs through hydroxylation by the CYP1A1,³ CYP1A2, and CYP1B1 enzymes. Cytochrome P450-mediated metabolic activation of C2 estradiol or estrone catalyzes the formation of 2-hydroxylated estrogen that can only weakly bind the estrogen receptor and would be associated with a reduced risk of ovarian cancer under a “receptor-mediated” hypothesis (2). Alternatively, a competing metabolic pathway leads to the 4-hydroxylated intermediate, which is more estrogenic (2, 5), and can be further oxidized to form 3,4-quinones and, subsequently, depurinating DNA adducts that are genotoxic and potentially carcinogenic to the ovary (6). Each of the enzymes catalyzes both pathways, but CYP1A2 and CYP1A1 favor 2-hydroxylation, whereas CYP1B1 favors 4- hydroxylation.

Although estradiol and estrone are extensively 2-hydroxylated, catecholestrogen concentrations are low in systemic circulation because they are conjugated and metabolized rapidly by O-methylation, glucuronidation, and sulfonation (3). COMT converts catecholestrogens into their corresponding methyl esters (2). If the metabolic clearance of the hydroxylated estrogen is slow or incomplete, reactive oxygen species may accumulate and possibly promote the development of ovarian cancer (7). In this pilot study, we examine allelic variation in genes that mediate the formation of catecholestrogens (CYP1A1, CYP1A2, and CYP1B1) and their further conjugation (COMT) to determine the relation of these genes to the risk of ovarian malignancy.

Eligible cases for this population-based, case-control study in Hawaii comprised all of the patients with histologically confirmed, primary, epithelial ovarian cancer diagnosed between July 1, 1993 and June 30, 1999 in any of the major hospital centers on Oahu. Through the rapid reporting system of the Hawaii Tumor Registry, part of the National Cancer Institute-sponsored Surveillance, Epidemiology, and End Results Program (8), we identified ovarian cancer patients and obtained histological information from the pathology logs and admission records of the participating hospitals. Eligible women were 18 to 84 years of age, were residents of Oahu, and belonged to one of the following ethnic groups: Caucasian, Asian (Japanese, Chinese, Filipino, and Korean), and “other” (Native Hawaiian and Samoan). Native Hawaiian was defined as any part-Hawaiian woman, whereas the other ethnicities were defined as having at least three of four grandparents of that ethnicity. Interview information was obtained from 218 (75%) of the 291 ovarian cancer cases eligible for participation in the study. Reasons for nonparticipation included physician refusal (5%), patient refusal (8%), death (6%), and inability to locate (6%).

The control pool consisted of lists of female Oahu residents who were interviewed by the Health Surveillance Program of the Hawaii Department of Health. Annually, the Department of Health identifies a 2% representative sample of all of the households in the state with a sampling procedure modeled after that of the National Health Survey (9). The refusal rate is negligible (<5%) because these surveys are conducted under statutory provisions. Hence, the lists we received for this study were adequately representative of the population. This source was supplemented with women 65 years of age and older who were Health Care Financing Administration participants on Oahu. Potential controls were randomly selected from the pool so that the ethnic (e.g., Chinese) and 5-year age group distribution would match that of the case group with a 1:1 ratio. Additional eligibility criteria for controls included having at least one intact ovary (subjects must have been able to report whether or not they had an oophorectomy and, if so, whether or not both ovaries were removed) and no prior history of ovarian cancer. Four hundred and sixteen women meeting these eligibility criteria were contacted to participate in the study. Interviews were obtained for 284 (68%) of these women, with 132 (32%) eligible women who could not be located, died, or refused to participate.

The majority (>95%) of subjects were interviewed in their homes by trained interviewers. Occasionally, if requested by the participant, the interview was conducted at the individual’s worksite or at the Cancer Research Center of Hawaii. All of the interviews were administered according to a standard protocol, regardless of the location of the interview, and took 1.5 to 2.5 h to complete. Cases were interviewed within 4 months of diagnosis, on average. The questionnaire included: a reproductive and gynecological history; questions concerning the use of oral contraceptives and other hormones; a contraceptive history, including tubal ligation; a medical history; anthropometric measurements; and sections on other lifestyle practices, such as diet and smoking.

We were able to draw blood on 146 (67%) of the interviewed cases and 192 (68%) of the interviewed controls. Blood samples were processed within 1 to 2 h of collection by the laboratory technician. We selected 129 cases with complete questionnaire (e.g., menstrual regularity) and tumor registry (e.g., histology) information to be included in the genotyping analysis. One hundred and forty-four controls were also selected to match the age and ethnicity of the cases.

Laboratory personnel were blinded to the case-control status of the subjects. DNA was purified from peripheral blood leukocytes by SDS/proteinase K treatment and phenol/chloroform extraction. Genotyping for the CYP1A1 (MspI) (T→C transition 264 bp downstream from the polyadenylate signal that creates an MspI restriction site) and exon 7 (Ile462Val) polymorphisms followed PCR-RFLP methods described in (10). We also studied a G1294C (Val432Leu) polymorphism in the CYP1B1 gene that creates a new Eco571 restriction site (11). A mutagenic reverse primer was used to create a new PstI site if there was a “C” at position 1294 (Val). In addition, a forward primer was created. We used primers 5′-TGTCCTGGCCTTCCTTTATG-3′ and 5′-CTGGATCAAAGTTCTCCGGGTTAGGCCACTGCA-3′. The Val allele is identified as the uncut, 187-bp product after PstI digestion, whereas the Leu allele is digested to 158-bp and 29-bp fragments. A C→A transversion in intron 1 of the CYP1A2 gene, located 734 bp downstream of the first transcribed nucleotide, was detected using PCR primers P1f, 5′-CAACCCTGCCAATCTCAAGCAC-3′ and P4r, 5′-AGAAGCTCTGTGGCCGAGAAGG-3′ (12). A G→A transition in exon 4 of the COMT gene, producing a valine-to-methionine substitution, was detected as described by Lavigne et al. (4).

The ORs associated with different levels of the exposure variables were evaluated by unconditional multiple logistic regression modeling case-control status (13). ORs and 95% CIs were computed by exponentiating the coefficients (and the 95% CIs) for the binary indicator variables representing the levels of the exposure (e.g., genotype) of interest. Adjustment variables included age, as a continuous variable, ethnicity by indicator variables (Caucasian, Asian, and other), education (<13 years, 13–14 years, and ≥15 years), parity (ever versus never), oral contraceptive pill use (ever versus never), tubal ligation (yes versus no), and regular menstrual cycles (yes versus no). We also considered other potential risk factors, such as tobacco smoking (ever versus never), pack-years (number of packs of cigarettes smoked a day × number of years smoked), waist/hip ratio (dichotomized at the median), and history of diabetes (ever versus never) as adjustment variables, but these did not materially alter the fit of the base model. The association of genetic polymorphisms with the risk of ovarian cancer was also modeled within various subgroups (e.g., ethnic group) that may influence enzyme activity. We performed a test for linear trend in the logit of risk on genotype by comparing twice the difference in log likelihoods for models with and without a trend variable, based on a χ² distribution with one degree of freedom. Gene dosage effects were modeled by assigning the values 1, 2, or 3 to a genotype trend variable according to the subject’s number of variant alleles (zero, one, and two variant alleles, respectively). Logistic regression was used to explore the joint association between genes and other variables by modeling each level of interaction between the pairs of variables using subjects who had homozygous wild-type alleles (or homozygous variant for CYP1B1) and were “unexposed” as the reference category and comparing this model by the likelihood ratio test to one in which only the main effects were modeled.

The mean age of cases (53.2 years) and controls (52.7 years) was similar by design as was the ethnic composition of the two groups (Table 1). The majority of cases were Asian (45%), followed by Caucasian (28%), and “other” (27%). The “other” ethnic category consisted mainly of Native Hawaiian cases (27/35, 77%) and controls (28/33, 85%). We found a reduction in the risk of ovarian cancer associated with higher education, pregnancy, oral contraceptive pill use, tubal ligation, and irregular menstrual periods. Neither tobacco smoking nor other variables (not shown), such as age at menarche, age at menopause, months of menstruation, estrogen/hormone replacement therapy (premarin), and alcohol use, were clearly associated with risk.

The frequency and association of each genetic polymorphism with the risk of ovarian cancer are shown in Table 2, after adjustment for covariates. The genotype distributions within each ethnic group were found to be in Hardy-Weinberg equilibrium for the genes under investigation. With the exception of CYP1B1, we found no significant differences in the frequency of polymorphisms between cases and controls. The frequency of the CYP1B1 Val allele was greater among ovarian cancer cases (29%) than among controls (19%), and risk increased monotonically with heterozygous Leu/Val and homozygous Val/Val genotypes (P = 0.005). The Val/Leu genotype for CYP1B1 was associated with an OR of 1.8 (95% CI, 1.0–3.3) and the Val/Val genotype with an OR of 3.8 (95% CI, 1.2–11.4) compared with the Leu/Leu genotype. Adjustment for oral contraceptive pill use had an important influence on the relative risks for CYP1B1. For example, the OR for the presence of at least one CYP1B1 Val allele was 1.9 (95% CI, 1.1–3.4) after adjustment for age, ethnicity, education, pregnancy history, history of tubal ligation, and regularity of menstrual cycles and 2.1 (95% CI, 1.2–3.7) after additional adjustment for oral contraceptive pill use (data not shown).

We examined the genotype frequencies and the consistency of the associations between the genetic polymorphisms and the risk of ovarian cancer by ethnic group and other factors shown in Table 1. Although the number of cases and controls was small within strata, there was general homogeneity in the stratum-specific ORs for each gene of interest (data not shown). For example, the CYP1B1 Val allele was more frequent among non-Asian women (27% Caucasian and other) than in Asian women (12%), but the positive association of the CYP1B1 Val allele with the risk of ovarian cancer was observed in both ethnic groups, increasing monotonically with the number of Val alleles. Caucasian women with at least one Val allele had an OR of 3.3 (95% CI, 1.3–8.8), and Asian women had an OR of 2.3 (95% CI, 1.0–4.9), compared to women with the Leu/Leu genotype. No notable differences in risk associated with CYP1B1 Val or the other polymorphisms under study were found by menopausal status, clinical stage, or histological type.

Interaction analyses were performed between polymorphisms and smoking history to determine whether the association of the genetic polymorphisms with the risk of ovarian cancer was similar in smokers and nonsmokers (Table 3). Cigarette smokers with a least one variant CYP1A1 (MspI) m2 allele were at significantly increased risk of ovarian cancer compared to never-smoking women with m1/m1 genotypes (OR, 2.6; 95% CI, 1.2–6.0). A significantly increased risk was also found among smokers carrying at least one CYP1B1 Val allele (OR, 3.8; 95% CI, 1.6–9.8), one COMT Met allele (OR, 2.2; 95% CI, 1.0–4.7), or CYP1A2 A/A alleles (OR, 2.6; 95% CI, 1.1–5.9) compared with the reference group (never-smokers with Leu/Leu genotypes for CYP1B1; never-smokers with Val/Val genotypes for COMT; and never-smokers with A/A genotypes for CYP1A2). The interaction was statistically significant (P = 0.03) for CYP1A1 (MspI) where the risk was not elevated among never-smokers with the variant m2 allele.

The two-way interactions of the genetic polymorphisms and pregnancy history (never versus ever) and oral contraceptive pill use (never versus ever) with the risk of ovarian cancer were also modeled (Table 4). We found no significant interactions between the variant alleles and either of these risk factors. Both in women who had never been pregnant and among those who had been pregnant, the CYP1B1 Val allele was associated with an increased risk of ovarian cancer. The effect was stronger in the nulliparous women (OR, 3.8; 95% CI, 1.0–14.6) than it was in parous women (OR, 1.9; 95% CI, 0.7–5.3), but the interaction was not significant. We found little relation between genotype and risk among women who had used oral contraceptives; e.g., there was no association of the CYP1A1 (MspI) polymorphism with risk among women who used oral contraceptives. The association of the CYP1B1 Val polymorphism with ovarian cancer risk was much stronger among women who had never used oral contraceptives than among ever-users.

We modeled the interaction between various combinations of two polymorphic genes with inconclusive results. The interactions of polymorphic genes involved in estrogen hydroxylation (CYP1A1, CYP1A2, and CYP1B1) and genes involved in estrogen conjugation and inactivation (COMT) with the risk of ovarian cancer are shown in Table 5. We found no relation of polymorphisms in any of the estrogen-hydroxylating genes with the risk of ovarian cancer among women with the high-activity COMT (Val/Val), although a joint association between CYP1B1 and COMT on risk was suggested. The OR for the CYP1B1 Val allele was 2.8 (2.6/0.9) in the COMT Met group and 1.3 in the COMT Val/Val group. None of the gene-gene interactions was statistically significant.

The proliferation of ovarian tissue with estrogenic stimulation and the possible increased risk of ovarian cancer with estrogen/hormone replacement therapy, at least for serous and endometrioid histological types (14), suggest that steroid hormones are related to ovarian cancer. However, Ness and Cottreau (15) have recently hypothesized that inflammation may work in conjunction with ovulation and steroid hormones in mediating ovarian carcinogenesis. The process of ovulation is associated with inflammation at the local level, including the elevation of cytokines and prostaglandins, cell damage, and oxidative stress. Ovarian epithelial cells undergo rapid cycles of division during repair of ovulation trauma and, in addition to hormones, may be exposed to reactive oxygen species produced via catecholestrogen-mediated redox cycling of quinones and semi-quinones (2). Oxygen radicals can oxidize DNA bases, form mutagenic lesions and chromosome aberrations, and activate chemical carcinogens into highly reactive compounds (16, 17).

The metabolism of steroidal estrogens is complex and involves oxidation at the C2 and C4 positions to form catechols, C16 to form 16α-hydroxy estrone, and C17 to form estrone (2). CYP1B1 is expressed in many tissues, including the ovary and breast (5, 6). The Val432Leu polymorphism results in an amino acid change in exon 3 of the CYP1B1 gene that encodes the heme-binding domain, a region critical to the catalytic function of CYP1B1 (11). This polymorphism in the CYP1B1 gene may alter the catalytic activity toward several procarcinogens, such as benzo(a)pyrene, and steroid hormones. Recent pharmacogenetic studies have shown distinct functional differences between wild-type CYP1B1 and polymorphic variants in estrogen hydroxylation (18). Shimada et al. (18) studied the catalytic activities of the gene products of several CYP1B1 polymorphisms and found that the ratio of 4-hydroxyestradiol (and 4-hydroxyestrone) formation:2-hydroxyestradiol (and 2-hydroxyestrone) formation was higher for the Val432 variant of CYP1B1 than the Leu432 variant of the enzyme. These differences in CYP1B1 polymorphic variants in catalytic function toward procarcinogens and steroid hormones may contribute to ovarian cancer susceptibility in humans. In our investigation, the “high-activity” Val432 variant for CYP1B1 was positively related to the risk for ovarian cancer. This association was statistically independent of that of other risk factors identified in our study population, although interactions with tobacco smoking, oral contraceptive pill use, and COMT genotype are suggested.

Few investigations of CYP1B1 polymorphisms and cancer risk have been conducted. Results from two case-control studies among women with cancer of the breast, which shares common risk factors with ovarian cancer such as pregnancy and breast-feeding, have been conflicting. In a case-control study of 164 Caucasian and 54 African-American breast cancer cases and 328 Caucasian and 118 African-American controls, Bailey et al. (11) reported no association of breast cancer with the CYP1B1 genotype. In a second case-control study of 186 breast cancer cases and 200 age-matched controls, Zheng et al. (19) found that women with the Leu/Leu genotype were at 2.3 times the risk of breast cancer compared to women with the Val/Val genotype. If confirmed, the direct association of the Leu allele with breast cancer and the inverse association with ovarian cancer suggest different mechanisms for CYP1B1 polymorphisms in mammary and ovarian carcinogenesis. The ratio of 4-hydroxyestradiol:2-hydroxyestradiol is higher for the Val432 variant than for the Leu432 variant, but the Leu allele is more active than the Val allele in catalyzing diverse mammary procarcinogens (17, 19). The discrepant findings for CYP1B1 highlight the need to determine the substrates and kinetics of each newly identified polymorphic enzyme that may influence susceptibility to cancer.

An association of the cytochrome P450 1A2 with the risk of cancer of the ovary is undoubtedly complex. A C→A transversion in intron 1 at position 734 downstream of the first transcribed nucleotide has recently been associated with increased CYP1A2 inducibility (12), although the extent to which this polymorphism explains the interindividual variability in CYP1A2 activity remains unclear. CYP1A2 could increase ovarian cancer risk through the activation of many drugs, dietary constituents, and tobacco procarcinogens (12, 20). Alternatively, CYP1A2 could decrease risk by shifting estrogen metabolism away from the 4-hydroxylation pathway to the 2-hydroxylation pathway (5, 21), leading to more stable DNA adduct formation (2,3-quinones; Ref. 6). We had hypothesized that women with the CYP1A2 A allele would be at reduced risk of ovarian cancer because the rate of 2-hydroxylation catalyzed by the CYP1A2 enzyme greatly exceeds that of 4-hydroxylation (2, 5). However, we found a decreased risk, although not significant, among women with the CYP1A2 C allele. We also found a suggestion of an interaction with smoking, because the association of the CYP1A2 A/A genotype appeared stronger among smokers. CYP1A2, CYP1A1, and CYP1B1 activities are all induced by the binding of aromatic hydrocarbons to the Ah-receptor, and we found a suggestion for an interaction between these polymorphisms and tobacco smoking on the risk of ovarian cancer.

Tobacco smoking alone did not significantly increase the risk of ovarian cancer in our study, but the use of tobacco products strengthened the positive or negative relation of several genetic polymorphisms to ovarian cancer risk. The OR among smokers who were carriers of at least one variant CYP1A1 (MspI) m2 allele were elevated compared to smokers with the wild-type m1/m1 genotype, but no such relation was found among never-smokers. The CYP1A1 gene codes for aryl hydrocarbon hydroxylase, a Phase I enzyme that is involved in metabolic activation of polyaromatic hydrocarbons and the hydroxylation of estradiol at C2 (and other positions) to conjugated estrogens (3). CYP1A1 is expressed in the ovary and a variety of other extrahepatic tissues (22). Although studies of CYP1A1 polymorphisms and breast cancer have found little evidence for a relationship to either the MspI or the Ile/Val polymorphisms, long-term smokers with the variant allele have been shown to be at greater risk of breast cancer than nonsmokers (23, 24).

Catechol estrogens are rapidly O-methylated by COMT, forming 2-methoxyestradiol and 4-methoxyestradiol (or the comparable conjugated estrone; Refs. 4, 5). Although O-methylation is primarily thought to be a detoxification pathway, 2-methoxyestradiol may also inhibit angiogenesis and the proliferation of human breast cancer cell lines. Allelic variation in COMT is thought to be directly related to enzymatic activity, including high-activity (Val/Val), intermediate-activity (Val/Met), and low-activity (Met/Met) variants. We found little association of the COMT genotype with ovarian cancer, which is in agreement with an investigation of 108 ovarian cancer cases and 106 hospital controls in Germany (25). Although not significant, there was a suggestion of an interaction between the COMT and CYP1B1 polymorphisms. The risk of ovarian cancer associated with the CYP1B1 Val allele was stronger among women with the COMT Met variant than among women with the homozygous COMT Val/Val genotype. It is possible that these women had reduced ability to prevent the cytotoxic and genotoxic damage caused by products of 4-hydroxylated catechol estrogen.

Although we found some ethnic heterogeneity in the frequency of the variant alleles, the ORs for ovarian cancer associated with the genetic polymorphisms under examination were generally consistent between Caucasians and Asians. Variation in the frequency of cancer “susceptibility” alleles among ethnic groups may provide a basis for some of the international differences in ovarian cancer incidence. In our study population, for example, the CYP1B1 Val allele was more common among Caucasian controls (27%), a high-risk group, than among Asian controls (12%), a low-risk group. The frequency of the CYP1B1 Val allele among Caucasian controls in our investigation was similar to that reported by Stoilov et al. (26) among 100 randomly selected British and Turkish controls (28%) but somewhat less than that reported by Zheng et al. (19) among 200 randomly selected Shanghai women (54%) or by Bailey et al. (11) among 164 Caucasian (51%) and African-American (69%) hospital controls. The CYP1A2 A allele frequency in our control population (70%) was very similar to that reported by Sachse et al. (Ref. 12; 68%) and consistent in our Caucasian (69%) and Asian controls (66%). The high-activity COMT Val allele frequency in Caucasian controls (56%) was similar to that reported by Lavigne et al. (Ref. 4; 48%) and Thompson et al. (Ref. 27; 51%) but differed from the frequency among our Asian controls (69%). Some of the ethnic variation in our study undoubtedly resulted from the limited size of the sample, and it is possible that the frequency of the various polymorphisms might not accurately reflect the true frequency of the polymorphism in the population. Although we think that the case and control groups were broadly representative, there remains a large sampling error. Such inadequacies of this investigation notwithstanding, the differences in allelic frequencies between studies support the need for larger, systematic population-based samples to evaluate the geographic distribution of each newly identified polymorphism.

Limitations of this study include the small number of subjects and the low participation in the blood sample collection (146/291, 50% of cases; 192/435, 44% of controls). This research was conceived to be exploratory in nature, so we did not commit our entire pool of subjects to the genetic analysis. Low participation rates may have biased our findings if the genetic polymorphisms under study were a determinant of participation. The ORs associated with pregnancy, oral contraceptive pill use, and other confounders were similar between subjects answering the interview information only and those participating in blood sample collection. Misclassification of exposure or disease is an important source of concern, and we have spent considerable time validating our assessment methods. Many of the exposures evaluated in our questionnaire, such as smoking and oral contraceptive use, were probably measured without much error. The possibility of such bias was minimized through the use of a structured questionnaire with well-defined probing methods and thorough training of the interviewers. To avoid disease misclassification, differentiation of primary ovarian cancer from metastatic colorectal cancer and from primary peritoneal carcinoma was achieved by close interaction of the pathologists with the clinicians and correlation of the clinical record with pathological review. Women who did not have definite primary ovarian cancer of the epithelial type by clinical or microscopic evaluation were excluded from the study. Loss of cases because of illness or death (6% of eligible subjects) was minimized through the rapid reporting system of the Hawaii Tumor Registry.

Our finding of a positive association of the CYP1B1 Val allele with the risk of ovarian cancer provides evidence for the hypothesized role of estrogen catabolism in ovarian carcinogenesis. The results also suggest that tobacco smoking may modify the association of ovarian cancer with several polymorphic genes known to activate or detoxify polycyclic hydrocarbons and other (pro)carcinogens found in tobacco smoke. The identification of metabolic genes that influence the risk of ovarian cancer can point to important enzymatic pathways in the carcinogenic process. This knowledge may help with the design of preventive interventions aimed at down- or up-regulating these pathways. Future studies with large samples and detailed surveys should focus on the potential interaction of CYP1B1 with other polymorphic genes and established risk factors.

We thank the physicians, administrators, and cancer registrars at the following Honolulu, Hawaii institutions for their support of this study: Castle Memorial Hospital, Kaiser Foundation Hospital, Kapiolani Medical Center for Women and Children, Kuakini Medical Center, Queen’s Medical Center, Straub Clinic and Hospital, St. Francis Hospital, Tripler Army Hospital, and Wahiawa General Hospital. The findings and conclusions of this study do not necessarily represent the views of these physicians and institutions.