Polychlorinated Biphenyls, Cytochrome P4501A1 Polymorphism, and Postmenopausal Breast Cancer Risk¹

It has been proposed that higher body burden of PCBs³ may increase the risk of breast cancer because of their hypothesized estrogenic (1), tumor-promoting (2), immune-modulating (3), and enzyme-inducing properties (4). To date, however, findings from epidemiological investigations of this relationship have been inconclusive, with several studies showing no association with risk, and others showing risk only among specific subgroups of individuals (5, 6, 7, 8, 9, 10, 11, 12, 13).

In laboratory studies, PCBs are potent inducers of CYP1A1, a drug-metabolizing gene, involved in the activation of potentially genotoxic endogenous and exogenous substances (14, 15). There is wide interindividual variation in CYP1A1 activity, and several genetic polymorphisms are present. Approximately 10–15% of Caucasians carry a CYP1A1 valine for isoleucine substitution allele (16). Although no difference in enzymatic activity of the variant type compared with the wild type has been demonstrated (17), there is evidence that the CYP1A1 variant genotype results in CYP1A1 activity that is more inducible in lymphocytes (18). Greater activity may lead to enhanced carcinogen activation and steroid hormone metabolism and may, therefore, be potentially related to risk for breast cancer. The CYP1A1 valine for isoleucine substitution has been associated with greater risk of lung cancer in Japanese (19). For breast cancer, there is no evidence for an overall association between this genetic polymorphism and breast cancer risk (20, 21, 22, 23, 24), although two studies reported subgroup effects among light smokers (22) or among women who had commenced smoking before age 18 (24).

In this case-control study, we sought to examine whether body burden of PCBs modified the association between CYP1A1 genotype and breast cancer risk.

This research used a subset of data collected as part of a case-control study (1986–1991) of 933 postmenopausal Caucasian women in Western New York; detailed methods have been reported previously (25). The protocol for the study was reviewed by the Institutional Review Board of the State University of New York at Buffalo and of participating hospitals. Informed consent was received from all participants. Women diagnosed with incident, primary, histologically confirmed breast cancer (n = 439) were frequency-matched by age and county of residence with controls (n = 494), randomly selected from the New York State Motor Vehicle lists (<65 years) and the Health Care Finance Administration rolls (≥65 years). Interview data included medical, reproductive, and lifestyle histories. Of women who provided usable interviews, ≈63% agreed to donate a blood sample. Blood was drawn for 262 postmenopausal breast cancer cases and 319 controls. Controls providing a blood sample did not differ from those who did not with respect to age, years of education, age at menarche, prevalence of benign breast disease or family history of breast cancer, and fruit, vegetable, or meat consumption, although they tended to have fewer years of smoking, consume fewer alcoholic beverages, and undergo menopause at an older age.

A subset of stored blood specimens was used for toxicological and genetic analyses: 154 women with postmenopausal breast cancer and 191 controls. Cases were only included in the final study sample if their blood was drawn before chemotherapy or radiation and within 3 months of surgery. Controls were frequency matched to cases by date of blood draw (±3 months) and age (±3 years). Controls who provided a blood sample but were not selected for these analyses were more likely to consume red meat and alcoholic beverages than those who were selected, but the two groups differed little with regard to demographic, reproductive, or lifestyle characteristics.

PCB concentrations were determined by the method of Greizerstein et al.(26). Total PCBs were calculated from summing the detected levels of 56 congener peaks, measured in the serum (based on 73 individual congeners). Participants with nondetectable levels for individual PCBs were assigned zero for these compounds. The procedures include standardized extraction, clean up, and quantification by high-resolution gas chromatography and comprehensive quality assurance program to minimize systematic and random errors. The sample was mixed with solutions containing IUPAC isomers 46 and 142 (surrogate standards). Methanol was added to precipitate the proteins, and the resulting mixture was extracted with hexane. The extract was concentrated and then cleaned by passing through a Florisil column. The eluate was then evaporated to a small volume, and isomers 30 and 204 were added as internal standards. An aliquot of the mixture was injected into the gas chromatographic system equipped with an electron capture detector. Quantification was based on calibration standards and response factors calculated using purchased reference materials. Quality control activities consisted of analyses of samples in batches of 6–10 simultaneously with quality control samples. The coefficients of variation of the quality control samples for individual PCB congeners ranged from 2.5% (IUPAC 180) to 6.2% (IUPAC 52). The coefficient of variation for total PCBs was 1.8%. The quality assurance program checked that the procedures were under control by the use of control charts and set criteria for data acceptability. The limit of detection for each analyte was determined as the mean of background noise plus three SDs in five reagent blank samples.

As described previously (22), DNA was extracted from blood clots obtained after centrifugation and removal of serum, which had been stored at −70°C. The CYP1A1 isoleucine-to-valine substitution in exon 7 was determined simultaneously with that of glutathione S-transferase (GSTM1), as detailed previously (22). Briefly, using primers specific for GSTM1 and for CYP1A1, where a base was substituted to introduce an NcoI restriction site, genes were amplified using PCR. The amplified PCR product was subjected to restriction enzyme analysis with NcoI (New England Biolabs, Beverly, MA), according to the manufacturer’s instructions. A simultaneous restriction enzyme digestion was also conducted with HinfI (New England Biolabs), which cleaved the GSTM1 fragment so that this band did not overlap the cleaved CYP1A1 fragment. Analysis by gel electrophoresis [4.0% agarose; 3:1 NuSieve (FMC Bioproducts, Rockland ME); agarose (Life Technologies, Inc., Gaithersburg, MD)] revealed 232- and 31-bp fragments for wild-type alleles (isoleucine) or a single 263-bp fragment when the mutation (valine) was present. The assay was validated by confirming polymorphic Mendelian inheritance patterns in seven human family cell lines (n = 134 family members), encompassing three generations (data not shown; samples were obtained from the National Institute General Medical Scientist Human Genetic Mutant Cell Repository, Coriell Institute, Camden, NJ). Genotyping results were read by two independent investigators, and genotyping for 20% of the subjects was repeated for quality control. The investigators were blinded to each other’s interpretations and to case-control status.

Descriptive analyses included Student t tests of means for cases and controls for lifestyle, reproductive, and dietary variables and χ² tests for categorical variables. We carefully examined cases and controls with respect to age and date of blood draw and attempted to individually match study participants on these variables. However, the width of the frequency matching strata did not allow for individual matching. Therefore, we used unconditional logistic regression to calculate ORs and 95% CIs. ORs were adjusted for potential confounders, including age, education, family history of breast cancer, parity, quetelet index, duration of lactation, age at first birth, serum lipids, and smoking status (nonsmokers; light smokers, <29 pack-years; and heavy smokers, 29 or more pack-years). Covariates were only included in the final regression model if they were established risk factors in these data or changed the observed risk estimate by at least 15%. Lipid adjustment was modeled after the method described by Phillips et al.(27), where serum triglycerides and total cholesterol levels were used to estimate total serum lipid content. CYP1A1 was investigated by collapsing the categories for women who were homozygous (Val:Val) with those who were heterozygous (Ile:Val) for the valine allele because of the small number of women in the first group (n = 6). Participants were designated as having either high or low PCB body burden on the basis of a level above or below the median for the controls. Levels ranged from 0.75 to 3.72 ng/g of serum in the low PCB group and from 3.73 to 19.04 ng/g of serum in the high PCB group.

The effects of PCBs on risk (13) and of CYP1A1(22) have been examined in detailed elsewhere. Shown here are those associations in this subsample with both CYP1A1 and PCB determinations. Women with higher body burden of PCBs were not at greater risk of breast cancer than women with lower levels, and women with at least one valine CYP1A1 allele (Ile:Val or Val:Val) were at slightly elevated risk compared with women who were homozygous for the isoleucine allele (Ile:Ile; Table 1). The test for statistical interaction between PCB body burden and CYP1A1 genotype approached statistical significance (P = 0.13). In Table 2, the combined effect of of PCB body burden and CYP1A1 genotype is shown. Women with lower serum PCB levels and CYP1A1 Ile:Ile served as the reference category. Compared with this group, women with lower PCB levels and CYP1A1 valine genotype were not at greater risk for breast cancer (adjusted OR, 0.88; 95% CI, 0.29–2.70), nor were women with elevated PCB levels and CYP1A1 Ile:Ile (OR, 1.08; 95% CI, 0.62–1.89). In contrast, there was evidence of increased risk for women with high PCB body burden and at least one valine allele when compared with women with lower serum PCB levels and who were homozygous for the isoleucine allele (OR, 2.96; 95% CI, 1.18–7.45). We repeated these analyses in two PCB congener groups, defined by CYP1A1 induction activity (IUPAC nos. 60, 105, 118, 128, 138, 153, 180, and 203) and by potential estrogenic activity (IUPAC nos. 18, 47, 52, 77, 136, and 153). The ORs for women with elevated PCB body burden and at least one valine allele were 2.87 (95% CI, 1.19–7.30) and 2.75 (95% CI, 1.09–7.21), respectively (data not shown). This correspondence in the observed risk estimates is likely to be an effect of the high correlations between total PCB levels and measured levels of these specific congener (Pearson r > 0.90).

In this study of associations between serum levels of PCBs, CYP1A1 genotype, and breast cancer risk, we found that among women with elevated PCB body burden, the CYP1A1 polymorphism statistically significantly increased risk. No effect of CYP1A1 was noted among women with lower serum levels of PCBs or women with the Ile:Ile genotype. To our knowledge, this is the first study to examine the effect of CYP1A1 in relation to PCB exposure on breast cancer risk. A number of previous investigations found no main effect between CYP1A1 and breast cancer among Caucasian women (20, 21, 22, 23, 24), despite laboratory evidence for a role of PAHs, which are activated by CYP1A1, in mammary carcinogenesis. CYP1A1 is also known to be involved in the metabolism of estradiol to the possibly mutagenic catechol estrogens (28), and increased metabolism related to the polymorphism could increase breast cancer risk through that mechanism. It is possible that such an association was masked by variability in CYP1A1 induction, due to heterogeneity in exposures to PCBs and other CYP1A1 inducers.

As indicated above, results of investigations of PCBs and breast cancer have been mixed; a number of studies did not observe an increase in breast cancer risk associated with elevated PCB levels (6, 7, 11, 12). Interestingly, in one of those studies (12), there was suggestion of an association between exposure and risk among African-American women. Investigators have identified novel CYP1A1 alleles in African-Americans, and the MspI allele, which is thought to be in linkage disequilibrium with the exon 7 substitution, is three times as common among African-Americans as among Caucasians (16) and has been found to be associated with increased breast cancer risk (21). It is possible that risk associated with PCB exposure among African-American women may be related to the increased prevalence of mutant alleles within this population.

The role of PCBs as P450 inducers in experimental animals has been so widely acknowledged that the commercial form is often used in animal studies to ensure and hasten neoplasms initiated by known carcinogens. Repeated studies, using a number of different bioassays and target organs, have all confirmed the role of PCBs as promoters of carcinogenesis (4). Caucasian postmenopausal women with elevated PCB body burden and the CYP1A1 isoleucine-to-valine substitution are possibly at greater risk of breast cancer because of the PCB-mediated enhanced induction of polymorphic CYP1A1, leading to increased activation of environmental carcinogens and subsequently resulting in the production of reactive intermediates and DNA damage. Thus, by inducing CYP1A1, PCBs and other inducers can trigger the activation of xenobiotics, such as those found in tobacco, into mutagenic compounds. This notion is supported by our observation that risk was significantly increased among women with elevated PCB body burden and the CYP1A1 polymorphism who had ever smoked cigarettes (OR, 7.74; 95% CI, 1.12–53.90) but not among women who had never smoked (OR, 1.43; 95% CI, 0.53–3.87; data not shown). It should be pointed out, however, that these risk estimates are based on very small numbers, as indicated by the width of the corresponding confidence intervals, and should be interpreted with caution.

Alternatively, there is some evidence that P450-mediated metabolism of lower chlorinated PCBs, which have been associated with greater toxicological activity, may lead to further metabolism by peroxidases to DNA-reactive metabolites (29). It is possible that women with both elevated PCB body burden and the potentially more inducible CYP1A1 variant genotype are at greater risk for DNA damage and subsequently for breast cancer. In fact, in our previous analyses, we observed a modest increase in risk (OR, 1.66) for women with detectable levels of lower chlorinated PCBs (13). We attempted to examine the effect of lower chlorinated PCBs in this research, but only a small proportion of participants had detectable levels for these congeners, resulting in very small numbers, which did not permit further stratification by CYP1A1 genotype.

Sample size was a major limitation in this study, as in many molecular epidemiological studies. Our findings were based on small numbers, resulting in possibly unstable risk estimates. It is possible that the modifying effect of PCB body burden may be a result of sampling variation. Furthermore, it was not possible to examine the effect of PCB body burden on the association between CYP1A1 and risk in more detail, restricting the analysis to crude stratification into low and high PCB body groups. A larger sample size would allow for an examination of a potential threshold effect at which PCB body burden affects the association between CYP1A1 and risk. Finally, due to sample size restrictions, it was not feasible to explore the interactions identified previously of PCBs with lactation status (13) and CYP1A1 genotype and smoking status (22) in this study of both PCBs and CYP1A1 genotype.

Overall, our results indicate that the CYP1A1 isoleucine-to-valine substitution may be a risk factor for breast cancer among women with elevated PCB body burden. If these findings are confirmed, the lack of an association between PCBs and breast cancer in some studies may, in part, be explained by the fact that only a small proportion of the study population was susceptible to the adverse effects of PCB exposure, i.e., those with the CYP1A1 polymorphism.