Glutathione S-Transferases M1, T1, and P1 and Breast Cancer¹

GSTs³are a family of enzymes involved in detoxication of benzo(a)pyrene and other carcinogens found in tobacco smoke,cytotoxic drugs, and chemical solvents (1, 2). Deletions in two GST genes, GSTM1 and GSTT1, occur at frequencies of 15% or greater in human populations (3). Individuals who are deletion homozygotes, classified as GSTM1 null or GSTT1 null, exhibit absence of enzymatic activity and are hypothesized to be at increased risk for the carcinogenic effects of a wide range of environmental exposures. Associations between GSTM1 and GSTT1 null genotypes and cancer of the lung, bladder, and colon have been reported, but results are inconsistent across studies (2, 3, 4, 5). An amino acid substitution variant in a third GST gene, GSTP1 codon 105 Ile→Val,has been identified recently that encodes an enzyme with reduced catalytic activity (6). The GSTP1 Valallele is common in human populations but has not been extensively examined in association with cancer.

Several previous studies investigated GSTM1 and GSTT1 genotypes and breast cancer risk (reviewed in Refs. 4, 7, and 8), and one study examined the role of GSTP1 genotype (9). We examined the relation of GSTM1, GSTT1, and GSTP1 genotypes and breast cancer risk in the CBCS, a large, population-based,case-control study of African-American and white women residents of North Carolina. To address issues raised by previous studies, we estimated main effects for each GST locus; conducted analyses stratified on smoking, family history, and other factors; and determined age at onset according to GST genotype and family history. We investigated joint effects for combinations of GST genotypes, as well as joint effects for GSTgenotypes and the COMT genotype, a gene involved in detoxication of catechol estrogens (10).

The CBCS is a population-based, case-control study of breast cancer conducted in 24 counties of North Carolina (11). Incident cases of invasive breast cancer among women of ages 20–74 were identified in cooperation with the North Carolina Central Cancer Registry, and controls were identified using Division of Motor Vehicles and Medicare beneficiary lists. Randomized recruitment (12) was used to oversample younger women and African-American women. Between 1993 and 1996, 889 cases of primary invasive breast cancer and 841 population-based controls were enrolled. Overall response rates (number of completed interviews/number of eligible women) were 74% for cases and 53% for controls (13). In-person interviews were conducted in participants’ homes. Over 98% of women who were interviewed agreed to provide a blood sample.

Genotyping for GSTs was conducted for the first 688 cases and 663 controls enrolled in the CBCS (278 African-American cases and 271 African-American controls; 410 white cases and 392 white controls). There were no appreciable differences in risk factors for breast cancer or response rates between participants genotyped for GSTsand the remaining participants in the CBCS (data not shown).

DNA was extracted from peripheral blood lymphocytes using standard methods (14). Genotyping for GSTM1, GSTT1, and GSTP1 was conducted using PCR-RFLP methods, as described in Helzlsouer et al. (9), with slight modifications for GSTT1; the forward primer sequence was 5′-gcc ctg gct agt tgc tga ag, the reverse primer was 5′-gca tct gat ttg ggg acc aca, and the annealing temperature was 59°C.

The genotyping assays for GSTM1 and GSTT1classify individuals with one or two copies of the relevant gene as“present” and individuals with homozygous deletions as “null.”The assay for GSTP1 classifies individuals according to the alleles Ile and Val. Assays that were unreadable for each locus are reported as “missing.” Approximately half of the samples listed as missing did not amplify for the relevant locus, and the remaining amplified too poorly to assign genotypes. The proportions of missing values were similar in cases and controls (5% of cases and 4% of controls were missing for GSTM1; 6% of cases and 4%of controls for GSTT1; 11% of cases and 9% of controls for GSTP1). Missing values for genotypes were not related to smoking or other covariates (data not shown). Sensitivity analyses were conducted by replacing missing values with different genotype combinations, and ORs remained within the confidence limits presented.

Positive and negative control samples were included with each batch of samples (one batch, 94 samples). Gels were scored by two different readers, and discordant samples were repeated. Reliability was assessed by selecting a random sample of 10% of samples from each batch. Batches with <95% agreement were rerun.

Methods for genotyping of COMT and associations with breast cancer in the CBCS have been reported previously (15).

Genotype frequencies and 95% CI for GSTM1, GSTT1, and GSTP1 were calculated as the proportion of individuals with a given genotype divided by the total number of participants. For GSTP1, allele frequencies and 95% CI were calculated as the number of alleles divided by the number of chromosomes, and tests for Hardy-Weinberg equilibrium were conducted by comparing observed and expected genotype frequencies using aχ ² test (16).

Adjusted OR for breast cancer and 95% CI were calculated from unconditional logistic regression models using SAS (version 6.11; SAS Institute, Cary NC). PROC GENMOD was used to incorporate offset terms derived from the sampling probabilities used to identify eligible participants (12) and to adjust for race (as two-level categorical variable) and age (as an 11-level ordinal variable reflecting 5-year age categories). Race was defined according to self-report.

Multivariable logistic regression models were used to adjust for potential confounding factors. Covariates included menopausal status, a composite of parity and age at first full-term pregnancy,breastfeeding, family history of breast cancer, history of breast biopsy, smoking, and alcohol consumption, as described previously (17). Family history was defined as having one or more first-degree relatives with breast cancer and was not verified by contacting relatives or reviewing pathology reports from relatives. Women were classified as exposed to ETS if they reported living with a smoker when they were 18 years or older and unexposed if they did not live with a smoker. Our assessment of ETS exposure did not include information on occupational, leisure, or recreational exposure. With the exception of ORs for smoking variables, ORs did not change after adjustment for additional covariates and therefore are adjusted only for sampling fractions, age, and race (where appropriate) in this report.

ORs for GST genotypes and breast cancer were calculated after stratifying on menopausal status, family history of breast cancer, use of hormone replacement therapy, alcohol consumption, and BMI. We conducted stratified analyses in this manner to compare our results with Helzlsouer et al. (9). For BMI, we stratified on the median among controls in our study (27.7 kg/m²), as well as the cutpoint used by Helzlsouer et al. (Ref. 9; 24.7 kg/m²). ORs for smoking and breast cancer were stratified on GST genotype to compare our results with our previous study of N-acetyl transferase genotypes and breast cancer (17). Thus, the method of stratification differed across tables but was necessary to compare results across studies. Interpretation of results did not differ when a uniform method of stratification based on a single common referent group was used(results not shown).

Joint effects of GST genotypes were estimated using the a priori low-risk genotype combination (GSTM1present, GSTT1 present, and GSTP1 Ile/Ile) as a common referent group. Joint effects of GST genotype and COMT were assessed using the low-risk GSTgenotype in combination with COMT HH or HLgenotypes as a common referent group. The COMT L allele encodes a thermolabile form of the enzyme that displays reduced ability to inactivate catechol estrogens through O-methylation (10).

Tests for interaction on multiplicative and additive scales were performed by comparing ORs for the joint effects of genotype and environmental factors (or joint effects for genotypes) using a common referent group (18). No evidence of departure from additive or multiplicative effects was seen (results not shown).

χ² tests were used to compare the prevalence of GST genotypes across stages of breast cancer among cases,and t tests were used to compare mean age at diagnosis among cases according to GST genotypes. Trend tests were conducted by calculating Ps for the β coefficient in a logistic regression model with the exposure coded as an ordinal variable. All Ps are two-sided.

Characteristics of participants in the CBCS have been described previously (17). Briefly, onset of menarche prior to age 12, nulliparity, first full-term pregnancy at age 26 or older, breast cancer in a first-degree relative, and smoking cigarettes for >20 years were positively associated with breast cancer, whereas breastfeeding showed an inverse association. Genotype frequencies for GSTM1, GSTT1, and GSTP1 in the CBCS are presented in Table 1. Estimates among controls are similar to previous studies, including a higher prevalence of GSTM1 null genotype among white compared with African-American controls (2, 4). Allele frequencies for the GSTP1 Val allele were 0.49 (95% CI 0.45–0.53) in African-American cases and 0.51 (0.46–0.55) in African-American controls (P = 0.6), and 0.31(0.27–0.34) in white cases and 0.35 (0.32–0.39) in white controls(P = 0.1). We did not observe significant departures from Hardy-Weinberg equilibrium for GSTP1 genotypes among African-American cases (P = 0.8), African-American controls (P = 0.5), white cases (P =0.9), or white controls (P = 0.8).

We did not observe significant case-control differences in genotype frequencies for GSTM1, GSTT1, or GSTP1 among African-American or white women (Table 1). ORs for GSTM1,GSTT1, and GSTP1 genotypes and breast cancer among African-American and white women are presented in Table 1. Results were close to the null for each GST locus. Similar results were found among pre- and postmenopausal women (Table 2). Results of subsequent analyses were similar among African-American and white women; therefore, results are presented for both races combined to increase precision.

GSTM1 null and GSTT1 null genotypes were positively associated with breast cancer among women with a family history of breast cancer (Table 3). ORs for GSTP1 did not differ according to family history. There were no differences in ORs for any of the GST loci when we stratified by use of hormone replacement therapy, alcohol consumption, or BMI (Table 3). Results for BMI were unchanged when we used strata for BMI as defined by Helzlsouer et al. (Ref. 9; data not shown). We also did not observe differences in ORs for GST genotypes when we stratified by consumption of fruits and vegetables (data not shown).

ORs for smoking and breast cancer, stratified by GSTgenotype, are presented in Table 4. In a previous paper that estimated main effects for smoking variables (17), we reported weak positive associations with breast cancer among former smokers, women who smoked for >20 years, and former smokers who had quit smoking within the last 3 years. There was little evidence of modification of these associations by GSTgenotype. The ORs for smoking longer than 20 years was slightly higher among women with GSTM1 null genotype compared with those with GSTM1 present. ORs for smoking more than one pack per day and exposure to ETS were higher among women with GSTT1null genotype compared with those with GSTT1 present. There was an inverse association for current smoking among women with GSTT1 null genotype. ORs for former smokers and former smokers who had quit within the past 3 years were higher for women with GSTP1 Ile/Ile genotype compared with women with one or more copy of the GSTP1 Val allele. However, for each of these results, associations remained weak, and confidence intervals were wide, leading to considerable overlap for the groups being compared. Results were similar for pre- and postmenopausal women (data not shown).

ORs for combinations GST genotypes and breast cancer are presented in Table 5. Compared with the putative lowest risk group (GSTM1present, GSTT1 present, and GSTP1 Ile/Ile), ORs were close to 1.0 for all combinations of GST genotypes except the putative highest risk group (GSTM1 null, GSTT1 null, and GSTP1 Ile/Val or Val/Val), for which we observed an inverse association with breast cancer risk. ORs were close to the null for all combinations of GST and COMT genotypes (Table 6).

Mean age at diagnosis of breast cancer according to family history and GST genotypes is presented in Table 7. For women with a family history of breast cancer, mean age at breast cancer diagnosis was significantly earlier (49.1 years) among women with GSTM1 null genotype compared with women with GSTM1 present (56.3 years). Less extreme but similar results were obtained for women with a family history of breast cancer and one or two GSTP1 Val alleles (52.7 years versus 56.8 years for GSTP1 Ile/Ile genotype). There was virtually no difference in mean ages at diagnosis by GST genotype among women who did not report a family history of breast cancer.

Genotype frequencies among breast cancer cases did not differ according to stage at diagnosis (P = 0.9 for GSTM1,P = 0.9 for GSTT1 and P =0.8 for GSTP1). Additionally, ORs for GSTM1,GSTT1 and GSTP1 and breast cancer did not differ when we stratified cases by stage at diagnosis (data not shown).

We examined the relation of GSTM1, GSTT1, and GSTP1 genotypes and breast cancer risk in a population-based, case-control study of African-American and white women in North Carolina. GSTM1, GSTT1, and GSTP1genotypes were not associated with breast cancer risk in African-American or white women, or among pre- or postmenopausal women. Two previous studies reported no overall association between GSTM1 genotype and breast cancer risk (19, 20). Helzlsouer et al. (9) and Charrier et al. (21) reported positive associations for the GSTM1 null genotype among postmenopausal but not premenopausal women, whereas Ambrosone et al.(22) reported a positive association for GSTM1null genotype among younger postmenopausal women. In contrast,Garcia-Closas et al. (23) reported no association for GSTM1 null genotype in pre- or postmenopausal women. Helzlsouer et al. (9)reported no association for GSTT1 null genotype and breast cancer in pre- or postmenopausal women, whereas Garcia-Closas et al. (23) observed an inverse association for GSTT1 null genotype among premenopausal women. All of the aforementioned studies were conducted primarily among white women. In the single previous study to include substantial numbers of African-American women, Bailey et al. (24)observed no association between GSTM1 null or GSTT1 null genotypes and breast cancer among African-American or white women. Only one previous study investigated GSTP1 and breast cancer (9). The authors reported a positive association for GSTP1 Val/Val genotype in postmenopausal women. It is likely that the differences in results across studies are attributable to chance, because they are based upon small subgroups of women.

ORs for GSTM1 null and GSTT1 null genotypes were elevated slightly among women with family history of breast cancer, and age at diagnosis was lower among women with a family history and GSTM1 null genotype. Helzlsouer et al.(9) reported 2-fold elevated ORs for all three GST genes among women with a family history of breast cancer, whereas Kelsey et al. (20) reported no modification of ORs for GSTM1 by family history. The positive associations for GSTM1 and GSTT1genotypes among women with a family history could be attributable to unmeasured genetic or environmental factors that interact with GST genes to increase risk of breast cancer and/or age at onset. Family-based studies that incorporate genotyping and environmental exposure assessment are the ideal study design to test such a hypothesis (25). We did not estimate joint effects for GST genotypes and BRCA1 or BRCA2status because of the small number of BRCA carriers in the CBCS. In fact, the majority of CBCS cases with a family history did not carry mutations in BRCA1 (26). Our results for family history could be biased, because we did not verify family history information using medical records.

We did not observe modification of ORs for GST genotype by use of hormone replacement therapy, alcohol consumption, or body mass index. Our results contrast with Helzlsouer et al.(9), who reported a strong positive association for GSTM1 null genotype among postmenopausal women with BMI>24.47 kg/m² and a positive association for GSTT1 null genotype among women who consumed alcoholic beverages. We did not observe modification of ORs for GSTgenotype by consumption of fruits and vegetables, in agreement with Ambrosone et al. (27). Smoking effects were modified only slightly by GST genotypes, as in previous studies (9, 20, 22, 23). We only partially addressed exposure to ETS, because the CBCS questionnaire did not include exposure during work or leisure time. Our results suggest that although GST enzymes are expressed in breast tissue (28),polymorphisms in GST genes do not play a strong role in modifying the association of smoking and breast cancer.

Helzlsouer et al. (9), in a study of 100 cases and 115 controls, reported an OR for breast cancer of 3.77 (95% CI,1.10–12.88) for the combination of GSTM1 null +GSTT1 null + GSTP1 Ile/Val or Val/Valgenotypes, compared with GSTM1 present +GSTT1 present + GSTP1 Ile/Ile. In contrast, we observed an OR of 0.5 (95% CI, 0.3–1.0) for the same comparison of GST genotypes. Garcia-Closas et al.(23), in a study of 466 cases and 466 controls, reported ORs close to the null for all combinations of GSTM1 and GSTT1 genotypes. Only our study and that of Garcia-Closas et al. (23) had 80% power to detect joint effects for GST genotypes, and the positive finding in the aforementioned study (9) may represent a chance finding because of the small number of participants (29). Lavigne et al. (10), using data from the same study as Helzlsouer et al. (9), reported an OR of 4.10(95% CI 1.17–14.27) for the combination of COMT LL +GSTM1 null genotypes, and an OR of 3.40 (95% CI,1.17–12.33) for COMT LL + GSTP1 Ile/Val or Val/Valgenotypes, whereas we did not observe positive associations for any combination of COMT or GST genotypes.

We did not observe an association between GSTM1, GSTT1, or GSTP1 genotype and stage at diagnosis of breast cancer. Our results are in agreement with Shea et al. (30)but contradict Kristensen et al. (31) who reported an association between GSTM1 null and GSTT1 null genotype status and more advanced stage at diagnosis in breast cancer patients. Kristensen et al.(31) reported that patients with GSTM1 null genotype had shorter overall survival, whereas Kelsey et al.(20) reported increased survival for patients with GSTM1 null genotype. We were unable to examine the relation of GST genotype and survival because we do not have information on long-term survival for women in our study.

Our results suggest that GSTM1, GSTT1, and GSTP1 genotypes do not play a strong role in susceptibility to breast cancer, in agreement with most previous studies. However,inability to detect effects for GSTs could result from failure to include relevant environmental exposures or genes that interact with GSTs. The presence of positive associations for GSTs in women with a family history suggests that unknown genetic or environmental exposures may modify the effects of GST genes, a hypothesis that could be investigated further in family-based association studies. Unmeasured genetic or environmental factors that interact with GSTs could also contribute to differences in results across epidemiological studies. A potential role for GST genotypes in breast cancer prognosis and response to treatment, as well as the possibility that GSTM1 status might modify age at onset for breast cancer,also merit further investigation.

We gratefully acknowledge the contributions of the nurse interviewers for the Carolina Breast Cancer Study: Martha Beach,Carolyn Dunmore, Dianne Mattingly, Theresa Nalevaiko, Patricia Plummer,Georgette Regan, and Cheryl Robinson. We also acknowledge the invaluable assistance of Susan Jackson and Patricia Moorman as Project Managers of the CBCS and Daynise Skeen for processing of blood samples. We thank three anonymous reviewers for helpful comments on the manuscript.