Glutathione S-Transferase M1, M3, P1, and T1 Genetic Polymorphisms and Susceptibility to Breast Cancer¹

Breast cancer is both the prevailing malignancy and the most common cause of cancer death among women in Western countries (1, 2). The major risk factors for breast cancer are mainly related to reproductive events that influence lifetime levels of hormones (3, 4). The carcinogenicity of estrogen has been linked not only to its mitotic activity but also to the role of catechol estrogens as carcinogenic metabolites (5). Quinones, the further oxidized metabolites, are the ultimate reactive electrophiles capable of DNA binding if not inactivated by glutathione conjugation (6). There is also substantial evidence on the role of oxidative stress in relation to breast cancer risk (7, 8, 9). Reactive oxygen species may be generated through a number of mechanisms, including the redox cycling of quinones and semiquinones in the metabolism of estradiol (10).

A large proportion of breast cancer cases cannot, however, be explained by the above mentioned risk factors. Identification of susceptibility factors that predispose individuals to breast cancer if they are exposed to particular environmental agents might give further insight into the etiology of this malignancy. It has been suggested that up to 80% of human cancers arise as a consequence of environmental exposure (11). The first line of defense is provided by the ability to metabolize and detoxify exogenous toxins (12). Therefore, inherited capacity for these metabolic activation and/or detoxification reactions may regulate individual susceptibility to environmentally induced diseases like cancer.

GSTs³ are a superfamily of enzymes that are potentially important in regulating susceptibility to cancer because of their ability to metabolize reactive electrophilic intermediates to usually less reactive and more water soluble glutathione conjugates (13). To date, four polymorphic families of cytosolic soluble GSTs (α, μ, π, and θ) of potential effect in this context have been identified in humans (13, 14).

The absence of GSTM1 and GSTT1 enzyme activities in about 50% and 10–25% of Caucasians, respectively, is caused by homozygous deletion (null genotypes) of the corresponding genes (15).

In GSTM3 gene, the GSTM3*A wild type and GSTM3*B variant allele differ from each other by a deletion of three bp in intron 6 resulting in the generation of a recognition sequence for the YY1 transcription factor in the latter. The functional consequences of this are still unclear, but both negative and positive regulatory effects have been suggested (16, 17). Relatively little is known about the role of GSTM3 in the metabolism of harmful agents, except having overlapping substrate specificity with GSTM1 (13).

For GSTP1 gene, two variant alleles, GSTP1*B and GSTP1*C, have been detected in addition to the wild-type allele GSTP1*A (18). In both variant alleles, a point mutation at nucleotide 313 results in a single amino acid change from isoleucine (Ile) to valine (Val) at codon 105. This residue lies in close proximity to the hydrophobic binding site for electrophilic substrates (19), and the Val¹⁰⁵ variant allele has been demonstrated to exhibit altered specific activity and affinity for electrophilic substrates (20).

The GSTM1 genotype has been related to the individual breast cancer risk in several recent studies (21), some of which suggested an association between GSTM1 null genotype and breast cancer risk in postmenopausal women (22, 23), whereas others found no association (24, 25, 26, 27, 28, 29).

In contrast to GSTM1, there is little data on the potential role of GSTP1 and GSTT1 genotypes in breast cancer risk, and no studies have yet been reported on GSTM3 and breast cancer. Two recent studies (29, 30) revealed no significant association between the GSTP1 genotypes and breast cancer proneness, although one study (23) suggested a trend for increasing risk with higher numbers of GSTP1 Val¹⁰⁵ alleles. Similarly, three recent studies found no association between the GSTT1 null genotype and the breast cancer risk (23, 26, 29), but one study (31) suggested a remarkably lower risk for premenopausal women lacking the GSTT1 gene.

There are several potential reasons for the inconsistencies in the outcomes of the above studies; they may arise from an inadequate number of study subjects, unknown menopausal status, or the lack of information or population differences on the other risk factors known to confer breast cancer risk. Moreover, because GSTs are known to have overlapping substrate specificities, deficiencies of GST isoenzymes may be compensated by other isoforms. Simultaneous determination of all of the relevant genotypes for a given exposure may, therefore, be a prerequisite for reliable interpretation of the results. We investigated the potential role of all of the four polymorphic GST genes in susceptibility to breast cancer in a Finnish Caucasian study population consisting of 483 incident breast cancer patients and 482 population controls.

This study is an extension of Kuopio Breast Cancer Study, a prospective study that follows the protocol of the International Collaborative Study of Breast and Colorectal Cancer coordinated by the European Institute of Oncology in Milan. Women with a suspect breast lump and living in the study catchment area between 1990 and 1995 were invited to Kuopio University Hospital for additional examinations and final diagnosis. They were asked to participate in the study at the first hospital examination and were interviewed by a trained study nurse before any diagnostic procedures. The recruitment protocol missed 51 women later diagnosed with breast cancer, all of who were private patients who did not enter the hospital by the standard procedure. Furthermore, 11 cases were missed during the nurses’ one-month strike in 1995. According to comparison with the Finnish Cancer Registry, only 26 breast cancer cases were treated elsewhere. Five hundred and sixteen of the women who agreed to participate in the study and 12 of the women who had refused to participate in the study were finally diagnosed with breast cancer. Thus, the participation rate of the cases was 98%.

Healthy controls were drawn from the Finnish National Population Register covering the catchment area of the cases. They were initially contacted by a letter explaining the study protocol and later called up by a research nurse. In all, 514 controls were interviewed in parallel with the cases. The participation rate for population controls was 72%. Detailed data on socioeconomic background, reproduction history, medical history, family history of breast cancer, current alcohol intake, smoking, and body-size indicators (height, weight, waist, and hip circumferences) were recorded (32).

All of the blood samples were collected before diagnosis and stored at −20°C before DNA extraction. For this study, DNA sample was available for 486 cases and 492 controls. Four population controls were excluded from the study because they had earlier breast cancer diagnosis, and two were excluded because of their non-Finnish origin. In addition, three cases and four controls were excluded because genotype data could not be obtained for them. Thus the final study population consisted of 483 histologically confirmed breast cancer cases and 482 population controls; all of them were Finnish Caucasians.

Genomic DNA (100 ng), extracted from lymphocytes by standard techniques, was used as template in the genotyping analyses performed blinded to the case-control status, essentially as described earlier (33, 34, 35). Briefly, GSTM1- and GSTT1-specific primer pairs were used together with a third primer pair for β-globin in a multiplex PCR analysis. The absence of the GSTM1- and/or GSTT1-specific PCR-product indicated the corresponding null genotype, whereas a β-globin-specific fragment confirmed proper functioning of the reaction (33, 34). In the GSTM3-genotyping analysis, after a PCR reaction a restriction enzyme digestion with MnlI was performed. Presence of a digestion site revealed the GSTM3*B variant allele (16). Similarly, in the GSTP1-genotyping analysis, the variant alleles containing a base substitution at the nucleotide 313 (GSTP1*B and GSTP1*C) resulting in Ile¹⁰⁵Val amino acid change were differentiated from the wild-type allele (GSTP1*A) by SnaBI restriction enzyme digestion subsequent to a PCR amplification (35). Because the method used did not differentiate between GSTP1*B and GSTP1*C alleles, the Val¹⁰⁵ alleles were designated as GSTP1 Val.

To ensure laboratory quality control, two independent readers interpreted the results. Any sample with ambiguous results was retested, and a random selection of 10% of all of the samples was repeated. No discrepancies were discovered upon replicate testing.

ORs and 95% CIs associated with the GST genotypes were calculated by unconditional logistic regression. Because of low numbers, the GSTM3*A/*B and *B/*B genotypes were combined in all of the analyses. The gene present genotypes of GSTM1 and GSTT1 and the homozygous GSTP1 Ile/Ile and GSTM3*A/*A genotypes served as the respective reference categories for all of the separate analyses for these genes.

Breast cancer risks are presented according to menopausal status. Women who reported natural menopause or had undergone bilateral oophorectomy were classified postmenopausal. Hysterectomized women with an intact ovary or ovaries (40 cases and 41 controls) and women for whom the details of the operation were unknown (6 cases and 2 controls) were also classified postmenopausal if they were no longer menstruating and were older than 51 years (median for menopause in Finnish women). All of the others were classified premenopausal. WHR and BMI (kg/m²) were dichotomized based on the median values for population controls.

Known or suspected risk factors for breast cancer were used as adjusting variables in two separate multivariate logistic models. The first model included age (5-year intervals), age at menarche (≤12, 13–14, ≥15 years), age at first full-term pregnancy (nulliparous, <25, 25–30, >30 years), number of full-term pregnancies (continuous), first-degree (mother, sister, daughter) family history of breast cancer (no/yes), and history of benign breast diseases (no/yes). The further-adjusted model also included the use of oral contraceptives, use of estrogen, WHR, BMI, education, current alcohol intake, and smoking. Because further adjustment did not substantially change the outcomes, only results obtained with the first model are presented. Subjects with missing values in any of the variables in a regression model were excluded from the analysis.

Possible interactions between genotypes and use of oral contraceptives, use of estrogen, WHR, BMI, history of alcohol drinking, and smoking history were examined using stratified analyses. The interactive effects were assessed by the likelihood ratio tests to compare goodness of fit of the models with and without the interaction term, taking into account the above-mentioned adjusting variables.

All of the reported P values are two-sided.

The mean age was 58.9 years (SD, 14 years; range, 23.3–91.6 years) for the cases and 53.5 years (SD, 11 years; range, 26.2–77.2 years) for the controls, contributing to higher proportion of cases (319/483; 66%) than controls (278/482; 58%) being postmenopausal. The risk of breast cancer was lower for women who had had at least one full-term pregnancy (OR, 0.53; 95% CI, 0.37–0.77) or who had ever used oral contraceptives (OR, 0.66; 95% CI, 0.48–0.91), whereas the risk was increased in women with first-degree relatives with breast cancer (OR, 2.53; 95% CI, 1.48–4.31), WHR above the median (0.91; OR, 1.38; 95% CI, 1.06–1.81), or history of benign breast diseases (OR, 1.33; 95% CI, 1.01–1.75; Table 1).

The distribution of GSTM1 genotypes and multivariate adjusted ORs stratified by menopausal status are presented in Table 2. Forty six percent of the cases and 42% of the population controls lacked the GSTM1 gene. The null genotype was significantly (P = 0.02) more frequent among premenopausal controls (48%) compared with postmenopausal controls (38%), whereas the respective frequencies among cases were almost identical (46%). The GSTM1 null genotype-related risk of breast cancer was significantly increased in postmenopausal subgroup (OR, 1.49; 95% CI, 1.03–2.15), whereas in premenopausal women the respective OR was close to 1 (OR, 0.95; 95% CI, 0.61–1.46).

The prevalence of the GSTM3 variant allele-containing genotypes was similar in cases (24 and 1.5%) and controls (24 and 1.7%) when pre- and postmenopausal women were considered together (Table 3). Moreover, although the frequency of the individuals with these genotypes was somewhat lower among premenopausal (21%) than postmenopausal controls (28%), the difference was not statistically significant (P = 0.09). The respective frequencies in cases were 26 and 24% giving adjusted ORs of 1.48 (95% CI, 0.89–2.47) for premenopausal and 0.79 (95% CI, 0.59–1.20) for postmenopausal breast cancer.

Around 7% of controls and 5% of cases carried the homozygous Val/Val genotype, and 38% of controls and 37% of cases carried the Ile/Val genotype when pre- and postmenopausal women were considered together. The respective adjusted ORs were 0.88 (95% CI, 0.66–1.18) and 0.57 (95% CI, 0.31–1.04; Table 4).

Although the frequency of the GSTT1 null genotype was slightly higher among premenopausal cases compared with controls (15% versus 12%), the association did not reach statistical significance (OR, 1.50; 95% CI, 0.79–2.84; Table 5).

After stratification by selected characteristics, postmenopausal women who had ever used estrogen and lacked the GSTT1 gene were found to have almost 3-fold risk of breast cancer (OR, 2.91; 95% CI, 1.11–7.64) compared with those who had the gene. Moreover, among women who had ever used estrogen, the risk paralleled the number of GSTP1 Val alleles; OR was 0.44 (95% CI, 0.23–0.86) for heterozygotes and 0.17 (95% CI, 0.04–0.77) for homozygotes, when compared with the women with GSTP1 Ile/Ile genotype. A protective effect of GSTM3*B allele-containing genotypes was observed among postmenopausal women who reportedly had never used alcohol (OR, 0.55; 95% CI, 0.32–0.94). No significant interactions were found for the use of oral contraceptives, WHR, BMI, or history of smoking (Table 6).

The possible combined effects of GST genotypes in breast cancer proneness were examined by calculating multivariate adjusted ORs for all of the combinations of two and three of the at-risk genotypes. The reference group consisted of individuals with the putatively most advantageous combinations of the genotypes. Because the outcomes from the analyses were similar for GSTP1 Ile/Val and Val/Val genotypes when the genes were studied separately (Table 4), these genotypes were grouped together in the combination analyses. Also, based on the above analyses for separate genes, the GSTM3*A/*A genotype was used as a reference category for the premenopausal women but the GSTM3*B allele-containing genotypes (*A/*B and *B/*B) for the postmenopausal women.

When combinations of two at-risk genotypes were examined (Table 7), the GSTM3*B allele posed an ≈2-fold risk of postmenopausal breast cancer in combination with the GSTP1 Ile/Ile genotype (OR, 2.07; 95% CI, 1.02–4.18). In contrast, no statistically significant combined effects were seen for postmenopausal women simultaneously carrying two at-risk genotypes. The only significant gene-gene interactions were found between GSTP1 and GSTM3 genes, both in premenopausal (P for interaction = 0.035) and postmenopausal (P for interaction = 0.012) women.

When three of the putative at-risk genotypes were combined, no significant increased risks were seen among the postmenopausal women (Table 8). In contrast, clear combined effects were observed among the premenopausal women. The most remarkable risk was seen for premenopausal women concurrently carrying the GSTM3*B allele-containing genotype, the GSTT1 null genotype, and the GSTP1 Ile/Ile genotype, who had a 10-fold risk (OR, 9.93; 95% CI, 1.10–90.0) of breast cancer compared with the putatively most advantageous combination of these genotypes. Substantially increased risk of borderline significance was also seen for the GSTM1 null women carrying the GSTP1 Ile/Ile genotype together with the GSTT1 null genotype (OR, 3.96; 95% CI, 0.99–15.8).

We were restricted from evaluating the potential combined effects of all of the four GST genes among the premenopausal women because none of the controls and only two of the cases carried all of the four at-risk genotypes simultaneously. Among the postmenopausal women, on the other hand, no significant increased risks were seen (data not shown).

The present study supports earlier suggestions that lack of GSTM1 gene poses an increased risk for breast cancer. The association originates from the significantly lower (38%) frequency of GSTM1 null genotype in postmenopausal controls compared with the cases (46%) and from the respective frequencies observed in premenopausal controls (48%) and cases (46%). A similar decreasing frequency of the GSTM1 null genotype among older healthy controls can be found in the studies by Charrier et al. (22) and Helzlsouer et al. (23), both reporting increased risk of breast cancer among postmenopausal women with the GSTM1 null genotype. Furthermore, in the recent study by Millikan et al. (29), in agreement with our findings, the frequency of GSTM1 null individuals was higher in premenopausal than in postmenopausal controls. On the other hand, although they observed no association in the overall study population, a positive association was seen between the GSTM1 null genotype and breast cancer risk for women with positive family history of breast cancer. Our data do not support this finding, possibly because rather few women with positive family history were included in our study (54 cases and 22 controls; data not shown). Nor was the age at diagnosis younger for the GSTM1 null genotype-carrying patients (data not shown) as in the study by Millikan et al. (29).

A recent study (31) of comparable size to ours (466 age-matched case-control pairs) found neither increased risk for GSTM1 null genotype nor unequal distribution of the genotype among controls. The difference cannot be explained by the fact that our controls were slightly younger and, therefore, the proportion of premenopausal women higher than in case patients. This would cause a bias toward null rather than a false positive finding. Moreover, our findings are supported by a recent meta-analysis (21) concluding that GSTM1 null genotype poses an increased risk of postmenopausal breast cancer (OR, 1.33; 95% CI, 1.01–1.76).

Rather unexpectedly, the GSTP1 Val allele appeared as protective rather than a risk factor for breast cancer, contrasting the conclusions of the above referred meta-analysis (21), where the carriers of the Val allele were suggested to have a significantly increased risk of developing breast cancer (OR, 1.60; 95% CI, 1.08–2.39). One explanation for the discrepancy may be that only two studies (23, 30) with relatively small sample sizes were included in this part of the meta-analysis. Consequently, the number of cases was only about one-third and the number of controls about one-half of that in our study. Moreover, no data on menopausal status or other potential confounding factors were available in one of the studies included in the meta-analysis (30). In the study by Helzlsouer et al. (23), a similar tendency of decreased risk of premenopausal breast cancer could be observed for the women with the Val/Val genotype (OR, 0.54; 95% CI, 0.04–6.67). In contrast, in the recent study by Millikan et al. (29), the GSTP1 Ile/Val and Val/Val genotypes also posed a tendency of decreased risk among white women (OR, 0.7; 95% CI, 0.5–1.0; and OR, 0.7; 95% CI, 0.4–1.2, respectively). This was not seen for African-American women. Furthermore, they found an increased risk among former smokers with GSTP1 Ile/Ile genotype. This is supported by recent observations that the Ile variant may be less active toward carcinogenic diol epoxides and benzo(a)pyrenes compared with the Val allele (36). Therefore, the GSTP1 variants may have different conjugation abilities toward different substrates. The present findings on the potential modifying role of GSTP1 genotypes in individual breast cancer proneness remain to be confirmed.

Similar to Helzlsouer et al. (23), we also found a tendency of increased risk of breast cancer among premenopausal GSTT1 null women. However, in the former study, a similar association was also seen for the postmenopausal women. In contrast, Bailey et al. (26) and Millikan et al. (29) found no association between the breast cancer risk and the GSTT1 genotype, and García-Closas et al. (31) reported a decreased risk for this malignancy for premenopausal women lacking the GSTT1 gene.

As can clearly be seen from the review of Dunning et al. (21), the differences in the outcomes of the studies conducted on this topic may at least partly be because of differences in the populations studied and in their exposures to the agents relevant to the development of breast cancer. In this respect, one strength of our study may be the inclusion of relatively large groups of cases and controls who were all confirmed to be of genetically homogenous Finnish origin. However, confounding caused by dietary factors, especially isothiocyanates in broccoli conjugated by the GSTs, has also been suggested recently to contribute to the inconsistencies seen in studies on the role of these enzymes in colorectal adenoma (37) and lung cancer (38). If this was the case, a potential weakness of our study could be that the dietary factors possibly related to GSTs could not be evaluated.

As for the environmental exposures, consistent with most of the earlier studies (23, 27, 31), smoking history did not significantly modify the effect of GST genotypes in breast cancer risk. Instead, we observed interaction between the GSTM3 genotypes and alcohol consumption and between postmenopausal use of estrogen and the GSTP1 or GSTT1 genotypes. Although Millikan et al. (29) did not see a similar effect between use of hormone replacement therapy and the GSTP1 or GSTT1 genotypes and although the possibility of a chance finding cannot be totally excluded because of the multiple comparisons performed, there are a priori hypotheses supporting these observations. Alcohol and estrogen are known risk factors for breast cancer, and GSTs may detoxify reactive intermediates produced by other polymorphic genes involved in the steroid metabolism or conjugate the lipid peroxidation products, cytotoxic compounds, and free radicals generated by alcohol consumption. However, it should be noted that all of these findings are based on substantially low numbers of subjects in stratified analysis.

Supporting the findings of Helzlsouer et al. (23), we observed a clear tendency of increased risk together with increased number of the putative at-risk genotypes. However, Helzlsouer et al. (23) reported statistically significantly increased risk for women with GSTM1 null and GSTT1 null genotypes together with GSTP1 Val/Val genotype (OR, 3.77; 95% CI, 1.10–12.88), whereas we did not observe any associations for this genotype combination (data not shown). Moreover, Millikan et al. (29) found the lowest risk for subjects simultaneously carrying the GSTM1 null, GSTT1 null, and GSTP1 Val allele-containing genotypes (OR, 0.5; 95% CI, 0.3–1.0) when compared with subjects carrying the GSTM1 present, GSTT1 present, and GSTP1 Ile/Ile genotypes. In the present study, the highest risk was seen for premenopausal women concurrently carrying the GSTM3*B allele-containing genotypes, the GSTT1 null genotype, and the GSTP1 Ile/Ile genotype; they had a 10-fold risk of breast cancer compared with the putatively most advantageous combination of these genotypes. Substantially increased risk of breast cancer was also seen for the premenopausal women lacking the GSTM1 gene and carrying the GSTP1 Ile/Ile genotype together with the GSTT1 null genotype. Although these findings are based on small numbers of the combinations of the three at-risk genotypes, together with the results from the analyses on combinations of two at-risk genotypes, they suggest that GSTM3 genotype may be an important modifier of individual susceptibility to breast cancer among premenopausal women. However, as multiple comparisons were performed, the possibility of chance findings cannot be ruled out.

To conclude, our findings support the view that the etiology of breast cancer may differ by menopausal status. They also support the previous suggestions that the GST genotypes may modify individual breast cancer risk, especially in certain combinations.

We thank our colleagues at the Kuopio Cancer Research Center (Kuopio, Finland) and A. K. Lyytinen, R. N., for data collection.