Genetic Polymorphisms in GSTM1, GSTP1, and GSTT1 and the Risk for Breast Cancer: Results from the Shanghai Breast Cancer Study and Meta-Analysis

Glutathione S-transferase (GST) is a family of genes with a critical function in the protection against electrophiles and the products of oxidative stress (1). The GSTs are involved in the metabolism of many xenobiotics, including an array of environmental carcinogens and chemotherapeutic agents and endogenously derived reactive oxygen species. GSTs are widely distributed in nature and are found in essentially all eukaryotic species. The four major families of GSTs, distinguished on the basis of primary structure, are designated as α, μ, π, and θ and are encoded by the GSTA, GSTM, GSTP, and GSTT genes, respectively. Of these, class π and μ predominate in the breast (2, 3, 4). Humans possess a single functional class π GST gene, whereas human class α, μ, and θ families contain multiple distinct genes, sharing ∼55, 65, and 50% identity, respectively. Mechanisms for biochemical protection by GSTs involve both conjugation of electrophilic compounds with glutathione facilitating their transport from the cell and reduction of organic hydroperoxides that contribute to oxidative stress (1). Certain epoxides formed from polycyclic aromatic hydrocarbons in cigarette smoke are substrates for class μ and π GST. The activated metabolites of the heterocyclic amines, carcinogens formed by cooking meat at high temperatures, are detoxified through conjugation by α and θ transferases. In addition to providing protection against exogenous chemicals, the GSTs are also involved in the protection of cells from oxidative damage, including free radicals generated through the metabolic redox cycle of catechol estrogens (5).

At least 20 isoenzymatic forms of GST have been identified, and many of them show genetically based individual variability of enzyme activity. The GSTM1 and GSTT1 genes both exhibit deletion polymorphisms (6, 7). Homozygous deletions of these genes, referred to as GSTM1–0 and GSTT1–0, respectively, result in complete absence of enzyme activity. An A→G polymorphism at nucleotide 313 in the GSTP1 gene leads to an amino acid change (Ile¹⁰⁵Val). The polymorphism resides at the substrate binding site and has been associated with reduced activity of the enzyme in vitro (8).

The deletion mutations in GSTM1 and GSTT1 and the I¹⁰⁵V variant in GSTP1 have been investigated for associations with breast cancer in a large number of studies. Several studies reported about a 20–50% elevated risk associated with the null GSTM1 genotype (9, 10, 11), and a stronger association was reported in two other studies (12, 13). However, the majority of studies reported no relation or even a possible inverse association of the null GSTM1 variant with breast cancer (14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26). The GSTT1 polymorphism has also been investigated in many studies, also with conflicting results (10, 11, 12, 13, 15, 17, 18, 19, 20, 21, 22, 23, 24, 26). A positive relationship of the GSTP1 Val/Val genotype with breast cancer was reported initially (12, 27), although this finding was not replicated in most subsequent studies (10, 18, 19, 21, 22, 28, 29). Only a few studies had statistical power to examine interactions among GST polymorphisms and the potentially modifying influences of environmental and endogenous substrates for GST. To further explore these relationships, we analyzed data from the large Shanghai Breast Cancer Study, a population-based case-control study of breast cancer conducted in China. To place the current findings in context of the large literature on GST polymorphisms in relation to breast cancer risk, we also performed a meta-analysis.

Case and control women in this study were participants of the Shanghai Breast Cancer Study, a population-based case-control study (30). The study protocol was approved by committees of relevant institutions for the use of human subjects in research. All case patients and control subjects were permanent residents of urban Shanghai between the age of 25 and 64 years; controls had no history of breast cancer. Through a rapid case-ascertainment system, supplemented by the population-based Shanghai Tumor Registry, 1602 eligible case patients with incident breast cancer, diagnosed between August 1996 and March 1998, were identified, and in-person interviews were completed for 1459 (91%) of them. The major reasons for nonparticipation were refusal (109 case patients, 6.8%), death before the interview (17 case patients, 1.1%), and the inability to locate (17 case patients, 1.1%). Cancer diagnoses for all patients were confirmed by two senior study pathologists through a review of tumor slides.

Control subjects were randomly selected from the female general population and were frequency matched to case patients by age (5-year intervals). The number of control subjects in each age-specific stratum was determined in advance according to the most recent data on the age distributions of the breast cancer patients available from the Shanghai Tumor Registry. The Shanghai Resident Registry, which keeps registry cards for all adult residents in urban Shanghai, was used to randomly select control subjects. In-person interviews were completed for 1556 (90%) of the 1724 eligible control subjects identified. Excluded from the study were 168 potential control subjects because of refusal (n = 166; 9.6%) or death or a prior cancer diagnosis (n = 2; 0.1%).

A structured questionnaire was used to elicit detailed information on demographic factors, menstrual and reproductive histories, hormone use, dietary habits, prior disease history, physical activity, tobacco and alcohol use, and family history of cancer. All participants were also measured for their current weight, circumferences of the waist and hip, and heights while sitting and standing. Blood samples were obtained from 1193 (82%) case patients and 1310 (84%) control subjects who completed the in-person interviews (31, 32). Of these, genotyping for one or more of the GST polymorphisms was successfully completed for 1143 (96%) case and 1221 (93%) control subjects.

Genomic DNA was extracted from buffy coat fractions using the Puregene DNA isolation Kit (Gentra Systems, Minneapolis, MN) following the manufacturer’s protocol. DNA concentration was measured by PicoGreen dsDNA Quantitation kit (Molecular Probes, Eugene, OR). Five to 10 ng of genomic DNA were used for each PCR. The laboratory staff was blind to the identity of the subject. Quality control samples (water, CEPH 1347-02 DNA, as well as blinded and unblinded DNA samples) were included in genotyping assays.

A multiplex PCR protocol was used to analyze simultaneously for the presence or absence of GSTM1 and GSTT1 genes (33). The Albumin gene was used as an internal control. The internal control amplified Albumin fragment was 350 bp in length, whereas presence of the GSTM1 and GSTT1 genes were identified by 215 and 480 bp fragments, respectively. Although these assays did not distinguish between heterozygote and homozygote positive genotypes, they conclusively identify the null genotypes. The GSTP1 A³¹³G polymorphism was determined by PCR-RFLP method reported previously (28). The PCR products were digested by BsmAI restriction endonuclease. The A to G substitution at nucleotide position 313 creates a BsmAI restriction site. The PCR product with G allele was digested to two fragments (148 and 41 bp), whereas the PCR product with A allele remained undigested (189 bp). Blinded duplicates were included in the genotyping: concordance rates were 94.5% for GSTM1 (55 replicates); 98.2% for GSTP1 (55 replicates); and 95.2% for GSTT1 (57 replicates).

Odds ratios (ORs) and 95% confidence intervals (CIs) were obtained by logistic regression; multivariate models included terms for age, benign breast disease, body mass index, waist-hip ratio, and number of menstrual cycles (estimated by subtracting age at menarche and time spent pregnant from age at menopause or current age, if premenopausal). Few women reported ever drinking alcohol (<4%). The few ever-active smokers (2%) were excluded. Analyses stratifying by indicators of endogenous estrogen exposure or environmental exposure to GST substrates were conducted to evaluate the potential modifying effects of these variables on associations between GST genotypes and breast cancer risk. Multiplicative interactions were examined by introducing cross-product terms for (dichotomized) risk factors and GST genotype (AA versus AG/GG for GSTP1).

To summarize the literature on GST polymorphisms in relation to breast cancer, we performed a meta-analysis. The meta-analysis was limited to published data. To identify relevant literature, we searched Medline through June 2003 using broad combinations of key words. We cross-referenced literature cited in research articles and checked review articles of breast cancer genetics, GST polymorphisms, and cancer susceptibility genes for studies not otherwise identified. To derive summary ORs and 95% CIs, we used routines available in STATA (34). Results presented are based on the method of Mantel and Haenszel (35), which assumes fixed or invariant effects of the genotype on cancer risk across populations. Meta-analysis ORs were estimated for all women combined and by menopausal status. We tested for publication bias using the method of Begg and Mazumdar (36), which evaluates whether there is correlation between effect estimates (ORs) and study variances in the published literature.

The distributions of selected demographic characteristics and major risk factors for breast cancer are shown in Table 1. Cases and controls were comparable in age (median: 47 years in both groups) and education (43% completed high school in both groups). Population differences between cases and controls reflected typical risk factors for breast cancer. Approximately one-third of the women were postmenopausal.

As shown in Table 2, allele frequencies for the individual GSTs were similar to those reported previously in other Asian populations. Approximately one-half of the women had the null genotype for GSTM1 or GSTT1 in both cases and controls. Case-control comparisons revealed no differences in genotype prevalence for GSTM1 or GSTT1: age-adjusted ORs were close to one, and CIs excluded large influences of these genotypes on breast cancer risk. For GSTP1, allele frequencies for the A to G substitution at nucleotide position 313 were 21% in cases and 18% in controls. Heterozygotes for the variant G allele had a modestly elevated relative risk (OR 1.08, 95% CI: 0.91–1.29), whereas the less common homozygotes had nearly twice the risk for breast cancer (OR 1.92, 95% CI: 1.21–3.04) when compared with those without the substitution. Results were similar regardless of stage at breast cancer diagnosis (data not shown). All results were essentially unchanged after adjustment for breast cancer risk factors (data not shown).

Pairwise analyses indicated no significant increase in risk for any combination of putative at-risk genotypes (Table 3). Furthermore, no dose response was observed according to these variants; women with all three at-risk genotypes had no increase in risk, combining AG and GG genotypes for GSTP1 (too few women were homozygous to be separated in analysis), when compared with women with three low-risk genotypes (Table 3). All results were similar in pre- and postmenopausal women (data not shown).

Table 4 shows age-adjusted ORs stratified by potential risk modifiers. Overall, ORs were not materially affected by GST genotype for various indices of estrogen exposure, including menopausal status, number of menstrual cycles, and body mass (Table 4); no significant interaction was observed for body mass index in pre- or postmenopausal women, when considered separately (data not shown). ORs for oral contraceptive use and waist-to-hip ratio were similarly unchanged stratifying by these genotypes (data not shown). The one statistically significant interaction for menstrual cycles and GSTM1 suggested a potentially abrogating influence of the GSTM1 protein on breast cancer associated with ovarian estrogen exposure. Interactions with sources of environmental exposure to polycyclic aromatic hydrocarbon and heterocyclic amines also were considered (Table 4): exposure to passive tobacco smoke in the workplace or from the husband smoking had little influence on breast cancer, regardless of GST genotype. Results for meat cooking preference were similarly null, although there was evidence for a modest potential interaction involving GSTT1 (P for interaction = 0.03). No significant interactions were noted for waist-to-hip ratio or for oral contraceptive use (data not shown).

Summary ORs for the individual GST genotypes are shown in Table 5. A total of 5950 breast cancer cases and 6601 controls was genotyped for the GSTM1 null variant in 19 published studies. The summary results from the meta-analysis indicated no relationship of the GSTM1 null variant with breast cancer risk overall (OR 1.05, 95% CI: 0.98–1.13) or in premenopausal women (OR 0.95). In postmenopausal women, the OR was minimally, although significantly elevated (OR 1.14). ORs for GSTT1 were also modestly elevated overall (OR 1.11) and in pre- (OR 1.13) and postmenopausal (OR 1.10) women; a significant test for heterogeneity in the unstratified analysis (P = 0.03) indicated substantial variability in the published results for the GSTT1 deletion. The meta-analysis for the homozygous I¹⁰⁵V variant in GSTP1 indicated no overall association of this variant with breast cancer (OR 1.04), although there was highly significant heterogeneity in the contributing studies (P = 0.009). Excluding the current data, the overall OR for GSTP1 was slightly inverse (OR 0.92, 95% CI: 0.75–1.13), and heterogeneity was attenuated (P = 0.08).

In addition to the current report, seven studies (10, 12, 13, 17, 18, 19, 21) considered combinations of GSTM1 and GSTT1, and four of these (12, 18, 19, 21) presented results for all three genotypes. The combined OR for deletion mutations in both GSTM1 and GSTT1, compared with none (based on 522 exposed cases; 523 exposed controls), was 1.08 (95% CI: 0.92–1.27). The combined OR for having three variant alleles (combining AG or GG for GSTP1) compared with none (based on 177 exposed cases; 185 exposed controls) was 0.99 (95% CI: 0.77–1.28; individual studies shown in Fig. 1).

The test for publication bias was marginally significant for GSTT1 overall (P = 0.048) and in studies of stratified results for postmenopausal women (P = 0.061). There was no similar evidence of publication bias for GSTM1 (P = 0.243) or GSTP1 (P = 0.938). Fig. 2 displays a funnel plot for the 15 studies that examined the GSTT1 polymorphism and breast cancer risk included in the meta-analysis. Asymmetry around the horizontal line for the meta-analysis OR, toward the right of the graph, indicates a pattern of more elevated ORs among the smaller, less precise published studies.

The current case-control study of women in urban Shanghai is the largest and most comprehensive examination of GST polymorphisms in relation to breast cancer risk. Results indicate no overall relationship of the GSTM1 or GSTT1 deletion variants with breast cancer risk in premenopausal or in postmenopausal women. In contrast, women with homozygous GSTP1¹⁰⁵Val substitution were at significantly increased risk for breast cancer when compared with those with two wild-type alleles; heterozygotes for this variant had no excess risk. Results suggest that these genotypes have no additive or multiplicative relationship with breast cancer risk: no association was demonstrated for any combination of putative at-risk genotypes (although too few women were homozygous for the GSTP1 substitution for a full evaluation of relationships with this genotype). Furthermore, we observed no consistent interactions of these genotypes with known or potential GST substrates, including exposures to carcinogens in sidestream smoke or heterocyclic amines in cooked meat (although a potential interaction of the GSTT1 deletion with meat cooking was suggested in the data). A statistically significant interaction with cyclic estrogen exposure was observed only for GSTM1. Taken together, results for the GST deletion variants are in line with the meta-analysis and suggest no overall relationship of these genotypes with breast cancer risk; in contrast, this is the first study to show a statistically significant positive association of breast cancer with the functional GSTP1 substitution variant.

A large number of studies have considered GST polymorphism in relation to breast cancer risk, with conflicting results. Rebbeck (37) has reviewed the early literature on the GST deletion variants in relation to cancer risk and noted a number of design issues that limit the interpretation of many of these studies. The earliest literature often relied on undefined convenience samples for case and control selection, with the possibility for selection bias on age, ethnicity, and other factors. Selection bias is also possible in hospital-based studies because the GSTs are likely to be related to the risk for other cancers and chronic diseases linked to smoking. The importance of matching for age was also noted: Rebbeck (37) cited studies showing increased frequency of the GSTM1-0 in hospital controls versus postmortem or geriatric series, suggesting negative selection of this genotype with advancing age (38, 39). An additional potential source of bias in studies of GST and cancer risk relates to their possible role in therapeutic response. The GSTs are involved in the metabolism of front-line chemotherapeutic agents, including nitrogen mustards and alkylating agents among others (1). Response to therapy may be governed at least in part by the genes driving the metabolism of and therefore exposure to sustained therapeutic levels of these drugs. A number of studies have suggested an improved prognosis in breast cancer associated with reduced expression of the GSTs (30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44). Inclusion of prevalent breast cancers in etiologic studies could lead to biased results by selecting for case women with prognostically favorable genotypes. Finally, the majority of studies were small and many lacked statistical power to detect moderate associations of GST with breast cancer risk or interactions with GST substrates (45). These design issues would have contributed to heterogeneous results in the literature regarding the contribution of these genes to breast cancer risk.

The current case-control study of women in urban Shanghai addresses many of the limitations in previous studies of GST and breast cancer risk. High enrollment rates, the population-based study design, and careful frequency matching on age would have substantially reduced the possibility for selection bias. Moreover, Chinese women living in Shanghai are relatively homogeneous in ethnic background (>98% are classified into a single ethnic group, the Han Chinese). Therefore, any potential confounding by ethnicity (46) would not be a major concern in these data. Finally, large numbers of women were enrolled in the study, and the extensive information available on reproductive and lifestyle risk factors allowed a relatively powerful stratified analysis on several potential effect modifiers. These latter analyses provide little support for interactions expected a priori; isolated finding of interactions between GSTM1 and menstrual cycles and GSTT1 and meat cooking could be chance observations given the many associations evaluated.

The meta-analysis should provide a reasonable global view of published research on these genotypes in relation to breast cancer. Modestly elevated summary ORs for GSTT1 (all women) and GSTM1 (postmenopausal only) were statistically significant because of the large numbers of women included in the analysis. The analysis suggests that at least some of the excess risk for GSTT1 may be an artifact of publication bias. The meta-analysis also revealed significant heterogeneity in the published results for GSTT1 and GSTP1. This heterogeneity could reflect methodological differences across studies, as noted above. However, it might also indicate population- specific differences in the contribution of these genotypes to breast cancer risk, and if so, summary ORs for these genotypes could be misleading. The outlying results from the current Shanghai study made a substantial contribution to the heterogeneity for GSTP1.

The meta-analysis excludes a moderate association of the GSTP1 Val/Val genotype with breast cancer risk (upper confidence limit: 1.4); the 2-fold excess in risk for this genotype from the Shanghai study could be due to chance or differences in study design or the ethnic make-up of the study populations. GSTP1 is the predominant GST isoform in the breast (2, 3, 4), and overexpression of the protein in breast tumors has been linked to a poorer outcome (42, 43, 44). The Val/Val genotype is uncommon (∼5% in Caucasians (47)), and fewer studies have examined its relationship to breast cancer risk when compared with the other GST variants; it is possible that design issues noted above could have obscured associations in some previous studies. Most research on the relationship of GST polymorphisms and breast cancer risk has been based on Caucasian women. The finding of an excess risk associated with this genotype in Asian women, if not due to chance, could indicate an influence of GSTP1 on cancer risk that depends on the genetic background or possibly environmental cofactors. In this regard, estrogen levels are substantially lower in Asian women;³ it may be speculated that environmental carcinogens, including those metabolized specifically by GSTP1, make a larger contribution to breast cancer in populations where estrogen exposure is low. Additional studies in Asian populations may help to clarify these observations.

Because of the importance of the GST enzyme system in defense against potential mammary carcinogens, additional studies should be undertaken to fully elucidate its contribution to breast cancer. Functional polymorphisms in other GSTs should also be studied as these come to light: a recently identified polymorphism in the proximal promoter of GSTA1, associated with reduced enzyme expression (48), has been linked to risk for colorectal cancer, particularly among consumers of well-done meat (49). A number of other potentially functional variants in GSTP1 have been curated,⁴ and some may be relevant to breast cancer risk (22, 29); a haplotype-based analysis that encompasses all of the candidate GSTP1 variants could provide additional insights. Finally, large studies will be needed to gain a clear picture of the contribution of these genes to breast cancer: these studies should consider interactions with genes encoding Phase I enzymes that produce reactive GST substrates (cytochrome monooxygenases) and, potentially, other detoxifying enzyme systems (i.e., glucuronosyl transferases and sulfotransferases) that may work cooperatively with the GSTs to guard the genome from chemical damage.

We thank the participants in the Shanghai Breast Cancer Study. We also thank all of the study interviewers and Jia-Rong Cheng for coordinating sample processing and storage in Shanghai. Finally, we thank Shoba Reddy-Holdcraft and Adam Morgan for technical assistance in the preparation of this manuscript.