Estrogen Metabolism–Related Genes and Breast Cancer Risk: The Multiethnic Cohort Study

The major susceptibility genes that have been identified for breast cancer, including BRCA1 and BRCA2, have been most helpful in determining who is at risk in multicase families. However, mutations in these genes are rare and only explain a small percent of all breast tumors. Another area of research is focusing on common polymorphisms in genes that affect the synthesis or metabolism of estrogens because these hormones increase cell proliferation in the breast and because their circulating levels have been associated with breast cancer incidence (1-4). The risk associated with each of these polymorphisms is expected to be low. However, because they are common, they have the potential of explaining a large proportion of cases in the population. Past studies of such polymorphisms and breast cancer have been quite inconsistent (1). This may be due to the fact that these studies have typically focused on single genes, whereas multiple competing enzymes are involved. In addition, for the most part, these past studies have not considered modifying effects. It is possible that these associations are particularly strong in the subgroups of women who, due to their obesity, reproductive history, and/or hormone replacement therapy use, are exposed to high levels of estrogens.

We investigated the associations between breast cancer and functional polymorphisms in several genes in the estradiol/estrone metabolism pathway (CYP1A1, CYP1A2, CYP1B1, CYP3A4, COMT, and SULT1A1), as well as in AHR (a transcription factor for CYP1A1, CYP1A2, and CYP1B1), in a large case-control study nested in the Multiethnic Cohort Study.

The design and baseline characteristics of the Multiethnic Cohort Study were described in detail elsewhere (4). In short, participants are Hawaii and California (primarily Los Angeles) residents who entered the cohort from 1993 to 1996 by completing a 26-page mail questionnaire including demographic and lifestyle factors (such as diet and smoking), medical history, medication use, family history of common cancers and, for women, reproductive history and hormone use. The cohort included 118,441 women and 96,810 men ages 45 to 75 years at cohort creation in 1993. Among the women, 25% were Japanese American, 22% White, 21% Latino, 19% African American, 7% Hawaiian, and 6% of other ethnic/racial origin. Breast cancer cases were identified through the Rapid Reporting System of the Hawaii Tumor Registry and through quarterly linkages to the Los Angeles County Cancer Surveillance Program, two cancer registries that are members of the Surveillance, Epidemiology and End Results program of the National Cancer Institute. This was complemented by annual linkages to the State of California's cancer registry. A nested case-control study of breast cancer was initiated in 1995 among women of the five main ethnic groups (5). A sample of cohort participants was randomly selected to serve as controls; the selection was stratified by sex and ethnicity. Incident breast cancer cases occurring since January 1995, and controls were contacted for donation of a blood sample. Samples were collected at the subjects' homes, processed within 8 hours and stored at 80°C. The participation rate among cases was 74% and varied from 70% in African Americans to 81% in Latinos. The corresponding rate for controls was 66% and varied from 60% in African Americans to 71% in Whites.

DNA was purified from whole blood using Midi Kits (Qiagen, Valencia, CA). The DNA samples were analyzed by PCR/RFLP for the COMT Val¹⁵⁸Met, CYP1A1*2A, CYP1A2*1F (A-154C), and CYP1B1 Val⁴³²Leu polymorphisms, as reported previously (6, 7). Samples were also genotyped for CYP3A4*1B using the following allele-specific primers: CYP3A4FC (wild type 5′-ATGAGGACAGCCATAGAGACAAGGGCTA-3′), CYP3A4FV (variant 5′-ATGAGGACAGCCATAGAGACAAGGGCTG-3′), and CYP3A4R (5′-CTTTCCTGCCCTGCACAG-3′). The PCR conditions are as follows: 94°C for 4 minutes, 94°C for 30 seconds, 60°C for 10 seconds, 72°C for 1 minute × 30 cycles, and 72°C for 7 minutes. The reaction product was subjected to electrophoresis and the 338-bp fragment identified on a 2% SeaKem (FMC BioProducts, Rockland, ME) agarose gel.

Genotyping for the SULT1A1 G638A (Arg²¹³His) polymorphism was done using a modification of the method of Coughtrie et al. (8) using primers SULT-F 5′-GTTGGCTCTGCAGGGTTTCTAGGA-3′ and SULT-R 5′-CCCAAACCCCCTGCTGGCCAGCACCC-3′. Amplification consists of a 4-minute denaturation at 94°C followed by 40 cycles at 94°C for 30 seconds, 63°C for 30 seconds, and 72°C for 30 seconds. An incubation step at 72°C is added at the end of the reaction for 7 minutes. PCR products were digested with 5 units HaeII at 37°C for 3 hours and separated on a 3.0% SeaKem ME in 1× TAE buffer. The PCR product is 333 bp in length and is digested by HaeII to168- and 165-bp fragments for the wild-type allele and is undigested for the variant allele.

The Lys⁵⁵⁴Arg polymorphism (G1721A) in exon 10 of the AHR gene (9) was genotyped by allele-specific PCR. The forward primers were 5′-AAACCTAGGCATTGATTTTGAAGACATCCG-3′ and 5′-AAACCTAGGCATTGATTTTGAAGACATCCA-3′ and the common reverse primer was 5′-ACGAATTGGTTAGAGTTCCAATTTTTAAAC-3′. The PCR conditions consisted of an initial denaturation at 94°C for 4 minutes followed by 25 cycles of 94°C for 30 seconds, 65°C for 30 seconds, and 72°C for 30 seconds followed by 28 cycles of 94°C for 30 seconds, 60°C 30 seconds, and 72°C for 30 seconds with a final extension at 72°C for 10 minutes. The reaction product was subjected to electrophoresis and the 348-bp fragment identified on a 2% SeaKem agarose gel.

Out of the 1,425 cases and 1,452 controls genotyped, 168 subjects were excluded because of missing information on covariates. The remaining 1,339 cases and 1,370 controls were available for data analysis. A race-adjusted comparison of these subjects with those who did not provide genotype data because of not donating blood or inconclusive results, revealed no difference in age, age at menarche, age at first birth, and parity. The statistical analysis used unconditional logistic regression to compute odds ratios and 95% confidence intervals for exposures of interest (10). Genotypes were modeled as two dummy variables representing the three levels, or as a gene dosage effect variable assigned a value of 1, 2, or 3 according to the number of variant alleles (zero, one, and two variant alleles, respectively). The final models were adjusted for age at blood draw, race/ethnicity, age at menarche, parity, and age at first birth. The likelihood ratio test was used to determine the significance of the interaction among certain variables with respect to breast cancer. The test compares a main effects, no interaction model with a fully parameterized model containing all possible interaction terms for the variables of interest. A Wald test from a polytomous logistic regression model was used to compare genetic trends for CYP1A2*1F between estrogen receptor (ER)/progesterone receptor (PR)–positive tumors and ER/PR-negative tumors. Genotype frequencies were tested for deviation from the Hardy Weinberg equilibrium with the χ² test.

The baseline characteristics of the cases and controls have been described previously (5). Eighty-seven percent of cases and 82% of controls were postmenopausal at baseline. The mean age at blood draw was 66 years for cases and 63 years for controls. On the average, cases tended to be older at baseline, to have had their first child at an older age, and to have had fewer children. They also more often had a positive family history of breast cancer and used hormone replacement therapy. Cases tended to consume less folate but drank somewhat more alcohol.

The distributions of the genotypes were consistent with Hardy-Weinberg equilibrium within each ethnic/racial group for all genes, except for AHR in Caucasians (P = 0.006) and Latinos (P = 0.002). The allele frequencies (Table 1) showed much variation across ethnic/racial groups and were similar to those reported in past studies (6-9). Table 2 shows the odds ratios and 95% confidence intervals for each genotype, after adjusting for age at blood draw, ethnicity/race, age at menarche, parity, and age at first birth. No association was observed, except for CYP1A2*1F that was inversely associated with breast cancer. The odds ratio (95% confidence interval) for the AA, AC, and CC genotype was 1.0, 0.9 (0.7-1.0), and 0.7 (0.5-1.0), respectively (P for gene dosage effect = 0.03). This association was suggested for three of the five ethnic/racial groups studied (Table 3), although some of the groups (in particular the Hawaiians) were quite small. Exclusion of the Caucasians and Latinos from the AHR model in Table 2 did not change the results.

No effect modification was found for CYP1A2*1F by tertile of age at menarche, body mass index (BMI), HRT use, or pack-years at baseline (Table 4), or by age at first birth or meat doneness preference (data not shown). Similarly, no difference in effect for CYP1A2*1F was found by CYP1A1, CYP1B1, AHR, CYP3A4, COMT, or SULT1A1 genotype (Table 5) or by stage at diagnosis. The effects of CYP1A1, CYP1A2, CYP1B1, and CYP3A4 did not vary by AHR, COMT, or SULT1A1 genotype (data not shown).

Information on ER and PR status of the tumor was available from our two cancer registries on 68% and 65% of the breast cancer cases, respectively. Cases with and without ER and PR information did not differ by geographic site (Hawaii or California), race/ethnicity, age at diagnosis, stage, year of diagnosis, years of education, or CYP1A2 genotype. Table 6 presents the association of CYP1A2*1F with, separately, ER-negative, ER-positive, PR-negative, PR-positive, ER/PR-negative, and ER/PR-positive breast cancers. The inverse association with CYP1A2*1F seemed stronger for ER/PR-negative breast cancer; however, the P value for comparison of gene dosage effects between the ER/PR-negative and ER/PR-positive groups was not statistically significant (P = 0.13).

In this large case-control study of estrogen metabolism-related genes in a multiethnic population of predominantly postmenopausal women, we found no association with the COMT Val⁵⁸Met, CYP1A1*2A, CYP3A4*1B, CYP1B1 Leu⁴³²Val, SULT1A1 Arg²¹³His, and AHR Arg⁵⁵⁴Lys polymorphisms and breast cancer. However, we observed an inverse association with the presence of the CYP1A2*1F allele which was not modified by the classic risk factors for breast cancer or the other genes under study and seemed somewhat stronger for ER/PR-negative tumors.

Multiple enzymes are involved in the metabolism of estradiol and estrone, the most biologically active forms of estrogen. A number of inherited polymorphisms in genes controlling these enzymes are being studied because they are common in the population and have been associated with a change in the gene product or enzyme activity. They may serve as surrogate markers for long-term exposure to altered hormonal tissue levels and thus as biomarkers of individual breast cancer risk, especially when considered as part of a multigene model that take into account metabolic pathways as they become better characterized.

The hydroxylation of estradiol and estrone to water-soluble metabolites (hydroxyestrones, HE) is an important elimination step. Oxidation occurs via two main competing and mutually exclusive pathways resulting in either 16α-HE or formation of catechol estrogens, 2-HE or 4-HE (11, 12). Whereas 2-HE binds only weakly to the ER, 4-HE and 16α-HE retain a potent estrogenic activity. Catechol estrogens are inactivated by O-methylation catalyzed by catechol-O-methyltransferase (COMT). If this inactivation process is incomplete, reactive species, such as 3,4-semiquinones, can react with DNA to form adducts which, if not repaired, may cause mutations (12-14). Increased formation of 16α-HE and 4-HE has been associated with an elevated breast cancer risk (15-17).

Although the enzymes responsible for the 16α-hydroxylation have not been well elucidated, CYP1A2, CYP1A1, CYP3A4, and CYP1B1 have been shown to catalyze the formation of catechol metabolites (11, 18). Polymorphisms in the genes coding for these enzymes are suspected to result in altered estrogen metabolism. Two linked polymorphisms in CYP1A1, one at the 3′ end of the gene giving rise to an MspI restriction site (CYP1A1*2A), the other in exon 7 resulting in a valine for leucine substitution (CYP1A1*2B), have been associated with an elevated CYP1A1 activity (19, 20), and with much inconsistency among studies, an increased breast cancer risk (21). A recent case-control study conducted in North Carolina (698 cases, 702 controls) found a 2.1-fold increased breast cancer risk among the subgroup of women with the MspI variant who smoked for >20 years (22). Our data did not show a similar modifying effect of pack-years and are consistent with most past results in showing no association of CYP1A1 with breast cancer.

In an in vitro study that tested 15 cytochrome P450 isoforms (including CYP1A2, CYP1A1, CYP1B1, and CYP3A4), CYP1A2 had the highest activity for the 2-hydroxylation of both 17β-estradiol and estrone and it also had a considerable activity for their 4-hydroxylation (23). The activity of this P450 enzyme has been shown to vary considerably among individuals, due to inducers (e.g., coffee, smoking, and contaminants) and presumably genetics (24). However, no common polymorphism has been found in the coding region of the CYP1A2 gene that could be shown to contribute to this intraindividual variability in activity (25). Nevertheless, a common C → A polymorphism in intron 1 (CYP1A2*1F) has been associated with a lower CYP1A2 inducibility (26). Although our observation of an inverse association between this variant and breast cancer may be due to chance, it is also consistent with some recent findings. CYP1A2*1F was found to be associated with a lower mean mammographic density among 267 premenopausal women in Hawaii (27). In another study of 220 premenopausal women sampled on the fifth day after ovulation an average of 4.4 times over a 2-year period, the CYP1A2*1F TT genotype was also associated with a 26% lower mean circulating estradiol level (7). Based on the model proposed by Pike et al. (28), a 20% difference in circulating estrogens may result in a halving in lifetime breast cancer risk, an estimate which is similar to what we observed in this study. However, it is not clear in our data why the association with CYP1A2*1F seemed somewhat stronger for ER/PR-negative tumors, and not for ER/PR-positive tumors, as one may expect if the underlying mechanism were through an ER-related pathway. However, this mechanism cannot be excluded because breast tumors that do not express hormone receptors may have initially been ER/PR positive and subsequently transformed into ER/PR-negative tumors via epigenetic or genetic events (29). If reproduced, our findings might suggest that the CYP1A2 variant would have a role early in the natural history of the disease, perhaps through lower circulating estrogens and/or protection against initiation by catechol estrogens. Another possibility is that ER-negative tumors (and especially those that are PR negative) express the newly discovered ERβ, making them differently responsive to the proliferative effect of estrogens (30).

A C → G transversion at position 1666 in exon 3 of the CYP1B1 gene, which results in an amino acid substitution of Leu⁴³² to Val, has been shown using human lung microsomes to result in a high enzyme activity and thus may contribute to the interindividual differences observed in CYP1B1 activity (31). Past data on this polymorphism and breast cancer have been inconsistent, with two studies finding an increased risk for the Val/Val genotype (32, 33) and one study showing no association (34). We failed to find an association between this polymorphism and breast cancer in our sample of predominantly postmenopausal women.

The expression of CYP1A1, CYP1A2, and CYP1B1 is regulated by the aryl hydroxylase receptor (Ahr), a ligand-activated transcription factor (35). A G1721A polymorphism within the coding region of the AHR gene, which results in replacement of Arg⁵⁵⁴ by Lys, has been observed in a recent study using ethoxyresofurin O-deethylase to be associated with a 3-fold increase in CYP1A1 inducibility (36). We are not aware of any other published report on the AHR Arg⁵⁵⁴Lys polymorphism and breast cancer to which our null results could be compared.

Finally, a variant allele containing an A → G mutation in a 5′ regulatory element of CYP3A4 has recently been identified (37). This polymorphism was found to be associated with a lower risk of chemotherapy-induced leukemia observed after treatment with drugs metabolized by CYP3A4 (38), suggesting that the variant allele results in a decreased activity. Only one other study (39) has examined CYP3A4*1B and breast cancer. In agreement with our data, no association was found, although Kadlubar et al. (40) reported that girls with the variant allele had earlier menarche.

Conjugation of parent estrogens and catechol estrogens, involving methylation or sulfation, are important detoxification pathways. A G → A transition at codon 158 of the COMT gene, which leads to a substitution of methionine for valine, has been linked to a reduced COMT activity (41). At least eight studies (reviewed in ref. 21) have examined this variant in relation to breast cancer risk. Associations were found in subgroups based on age or other risk factors. However, there is no consistency across studies as to the subgroup in which the association was found. We failed to find any association with breast cancer, overall or in any of the subgroups examined in previous studies.

A common G-to-A transition at nucleotide 638 of the SULT1A1 gene, resulting in an arginine-to-histidine substitution at codon 213, has been associated with a decreased sulfotransferase activity, as measured in platelets (42). Zheng et al. (43) recently found in a case-control study in Iowa that breast cancer risk increased with the number of His alleles among all women and with red meat doneness level among women with the Arg/Arg genotype. The latter association was not observed in women with the His/His genotype. A second study also found a direct association between breast cancer and the His/His genotype that was of borderline significance (44). In the present study, we failed to find any association between SULT1A1 and breast cancer, overall or in subgroups defined by doneness level or smoking.

Compared with previous studies, the present investigation presents a number of important strengths, including its large size, information on hormone receptors, and the prediagnostic assessment of breast cancer risk factors. Its approach was also more comprehensive since past studies almost uniformly examined single genes. Although not all cohort members were genotyped due to the case-control study design, participation was relatively high and no major difference was found between respondents and nonrespondents. In addition, only <8% of eligible patients died before contact and no modifying effect of stage at diagnosis was observed, making a survival bias unlikely. A number of subgroup analyses were conducted on the data, increasing the risk of chance findings, but these analyses addressed hypotheses that were formulated a priori. Limitations to this study include that information on ER/PR status was missing for a third of the cases and that the current sample size does not yet allow for an adequate test of heterogeneity of effects across ethnic groups.

In conclusion, the present study suggests that COMT Val⁵⁸Met, CYP1A1*2A, CYP3A4*1B, CYP1B1 Leu⁴³²Val, SULT1A1 Arg²¹³His, and AHR Arg⁵⁵⁴Lys are unlikely to play a significant role in breast cancer etiology. It also provides evidence for an inverse association between CYP1A2*1F and breast cancer, which may be related to the observation of lower circulating estradiol levels in premenopausal women with the TT genotype. Confirmation of these findings and a systematic characterization of the association between breast cancer and single nucleotide polymorphisms and haplotypes recently identified in CYP1A2 seem warranted.

We thank Jana Koerte for assistance with data analysis and Wendy Chang, Annette Lum-Jones, and Ann Seifried for their help with genotyping.