Glutathione S-Transferase M1, T1, and P1 Polymorphisms as Risk Factors for Renal Cell Carcinoma: A Case-Control Study¹

Recognized risk factors for RCC³include cigarette smoking, high relative body weight, and high blood pressure (1). A recent study in the United States evaluating population-attributable risks for RCC reported that after accounting for these three exposures, about 51% of RCC cases remained unexplained (2). Research on genetic factors that affect cancer susceptibility has the potential to improve our understanding of RCC etiology.

GST enzymes catalyze conjugation of electrophilic substrates with glutathione, usually resulting in detoxification of reactive intermediates (3). Common polymorphisms occur in three human GSTs and seem to influence cancer risk (4, 5, 6). Polymorphisms in which the gene is deleted and no active enzyme is expressed occur for GSTM1 and GSTT1(4). A single-nucleotide variant in exon 5 of GSTP1 results in an amino acid substitution,Ile¹⁰⁴Val, and altered enzyme activity (7, 8). GSTM1 and GSTP1 are active in the detoxification of activated forms of polycyclic aromatic hydrocarbon compounds. Substrates for GSTT1 include halogenated solvents, such as dichloromethane, and ethylene oxide, formed endogenously from ethene,which is present at high levels in cigarette smoke (3). GSTT1 and GSTM1 enzymes also have catalytic activity toward phospholipid hydroperoxide (9), evidence that GSTs may prevent DNA damage from lipid peroxides formed endogenously as a result of oxidative stress. Not all reactions catalyzed by GST enzymes result in detoxification; reactions of certain halogenated compounds catalyzed by θ-class GSTs produce mutagenic species (3).

The altered GST activity associated with the polymorphisms is expected to affect cancer risk through decreased protection against DNA damage from reactive electrophiles. GSTs are expressed and have significant activity in the kidney (10, 11, 12), but few studies have considered GSTs in susceptibility to RCC. Published reports include a case-control study nested within a cohort of German workers occupationally exposed to TCE (13) and a hospital-based case-control study in France (14). Reduced risk of RCC associated with GSTT1 null genotype was reported in the TCE-exposed cohort. The hospital-based study found no association between GSTM1, GSTT1, or GSTP1polymorphisms and RCC risk. Thus, previous data on associations between these candidate susceptibility genes and RCC are limited. We investigated GST polymorphisms in a population-based case-control study of RCC.

Incident cases of RCC (ICD-O site code 189.0) diagnosed February 1996 through January 1997 in five counties in western Washington state were identified from the Cancer Surveillance System, a population-based cancer registry. Cases, 20–74 yr of age, were eligible. Control subjects had been identified and interviewed as controls for studies of esophageal (15) and lung⁴cancer. The two control groups, one selected from counties that are largely rural and suburban, and the other from counties that contain large urban areas, were combined to provide geographic balance for the RCC cases. Controls were ascertained in 1993–1996 by random digit dialing and from Health Care Finance Administration records. Participants completed a structured telephone or in-person interview regarding exposures to risk factors for RCC. Interviews were conducted for 173 of 214 eligible case subjects (80.8%). The respondent was the index subject for 144 case interviews, with 90% taking place within 12 months of diagnosis. Interviews were conducted with 29 proxies for cases because the subject was deceased or too ill to participate. Reasons for nonresponse among cases were: subject deceased with no proxy (n = 16), unable to contact subject(n = 10), and refusal [index subject(n = 5), proxy (n = 6), and physician(n = 4)]. The control group included 124 subjects from the esophageal cancer study (71.8% response rate) and 490 subjects from the lung cancer study (85.2% response rate). Interviews were conducted with proxy respondents for eight controls.

Paraffin blocks containing fixed tissue from surgery or biopsy were the source of DNA for cases. Blocks containing only normal kidney were used if available. Multiple sections, 10 microns in thickness, were cut from each block and deparaffinized using 1200 ml of xylene, followed by ethanol wash. DNA was extracted using a commercial kit(QIAGEN, Inc.), a method similar to published procedures (16, 17). For nine case subjects with no suitable pathology specimen available, a buccal cell sample was collected for genotyping. Buccal cells were obtained using a cytology brush, which was rubbed on the inside of the subject’s cheek. Samples were kept on ice during transportation, then frozen at −70°C. DNA was extracted by incubation for 5 min at 95°C with 60 ml of 500 mmNaOH, followed by neutralization with 60 ml of 1 M Tris (pH 8). For controls, DNA was obtained from blood samples, which were kept on ice during transportation, then the buffy coats were isolated and frozen at−70°C until processed for DNA extraction.

GSTM1 and GSTT1 genotypes were determined by multiplexed PCR using three sets of primers to amplify a 215-bp sequence of the GSTM1 gene (18), a 268-bp sequence of the β-globin gene (18), and a 480-bp segment of the GSTT1 gene (19). Presence or absence of the β-globin band was used to determine failed PCR. The A/G polymorphism of the GSTP1 gene (20) responsible for the Ile¹⁰⁴Val substitution was detected using an oligonucleotide ligation assay, a modification of previously described methods (21). GSTP1 DNA sequences were amplified by PCR, followed by allele-specific ligation reaction and colorimetric detection. The PCR primers were: forward primer,5′-GACTGTGTGTTGATCAGGCG-3′, and reverse primer 5′-TGCACCCTGACCCAAGAAGG-3′. The oligomer probes for the A/G polymorphism were: wild-type probe,5′-biotin-AGGACCTCCGCTGCAAATACA-3′; mutant probe,5′-biotin-AGGACCTCCGCTGCAAATACG-3′; and common reporter probe,5′-phosphorylated- TCTCCCTCATCTACACCAACT-digoxigenin-3′. The GSTP1 oligonucleotide ligation assay had been verified in the same laboratory by DNA sequencing of PCR products from samples from 60 subjects, with 100% concordance of results. For all genotyping assays, several quality control samples (DNA from volunteers with known genotype) were included with each batch of study samples.

Of 173 case subjects interviewed, tissue was obtained for 132 (85 from tissue blocks containing normal kidney preserved from surgery, 23 from tissue blocks from surgery containing all or part tumor tissue, 15 from tissue blocks from biopsy, and 9 from buccal cell samples). No DNA samples were obtained for 18 interviewed cases who did not return consent forms for access to tissue and for 23 cases for whom tissue was unavailable from the hospitals. The number of cases with interpretable genotype results was 126 for GSTM1, 126 for GSTT1, and 130 for GSTP1. DNA from blood samples was obtained for 505 control subjects, with genotype results available for 505 for GSTM1, 504 for GSTT1, and 491 for GSTP1.

ORs were calculated using unconditional logistic regression. Effect modification among GST genotypes and between genotypes and age, sex, cigarette smoking, and BMI was evaluated using logistic models including interaction terms between categories of the variables of interest, with statistical significance determined by likelihood ratio tests comparing models with and without the interaction term(s). Stata software (Stata Corporation, College Station, TX) was used for statistical analysis.

Exposures associated with increased RCC risk included cigarette smoking, high BMI, and high blood pressure (Table 1). There was no increased risk associated with cigar or pipe smoking or residential second-hand smoke exposure (data not shown). Among cases,current smoking at reference date was reported more frequently by proxy respondents (34.5%) than by index subjects (22.4%). However, ORs for pack-years of smoking, BMI, high blood pressure, and other exposures were essentially unchanged when information obtained from proxy respondents was excluded. Control subjects from the study of esophageal cancer were older than those from the study of lung cancer (48.9%older than 64 versus 35.5% older than 64), but the two groups had similar distributions of other characteristics such as sex,race, smoking history, and BMI.

Demographic characteristics and exposure histories of cases and controls with DNA available were similar to all interviewed subjects. However, cases with DNA were more likely to have been diagnosed with local or regional stage disease than all interviewed cases. Among cases with DNA available, the percentages with local, regional, distant, and unstaged disease were 55.4%, 25.4%, 17.7%, and 1.5%, respectively,whereas the corresponding percentages among all interviewed cases were 49.1%, 23.1%, 23.7%, and 4.1%. (The percentages among all 214 eligible cases were 48.6%, 19.2%, 26.6%, and 5.6%.). Most cases with distant disease at diagnosis were not treated by surgery, and,therefore, pathological material was unavailable for these cases.

Frequencies of GSTM1, GSTT1, and GSTP1genotypes were similar between the two different study groups that formed the control group (data not shown). Among controls, genotype frequencies were similar by gender and across age categories (Table 2). Hispanic controls had a higher frequency of GSTP1 AG or GG genotypes than non-Hispanic controls (P = 0.08). No other important differences were detected by race, ethnicity, or country of birth. The numbers of non-white and Hispanic subjects were small. The distribution of GSTT1 genotypes varied by county of residence (P = 0.01) among controls. When control subjects who were non-white, Hispanic, or born outside the United States were excluded from the comparison, the differences in genotypes across counties persisted. Among cases, the frequency of GSTM1, GSTT1, and GSTP1 genotypes did not differ significantly by source of DNA (normal tissue block, tumor tissue block, biopsy, or buccal cells) or by stage at diagnosis.

An excess of the GSTT1 null genotype was observed in cases(Table 3). The frequencies of GSTM1 and GSTP1 genotypes in cases were similar to controls. Among non-Hispanic white subjects,results for GSTT1 (OR, 1.8; 95% CI, 1.0–3.2), GSTM1 (OR, 0.9; 95% CI, 0.6–1.6), and GSTP1(for AG and GG versus AA genotypes, OR, 1.0; 95% CI,0.6–1.7) were similar to the race-adjusted results for all subjects.

Table 4 shows numbers of cases and controls according to combined genotypes for GSTT1, GSTM1, and GSTP1, and ORs using subjects with GSTT1 present, GSTM1 present, and GSTP1 AA as the reference category. Although CIs were large due to the small number of subjects in some categories, there was increased risk associated with GSTT1 null genotype in all categories of GSTM1 and GSTP1 genotypes. There seems to be a slightly elevated risk associated with GSTM1null or GSTP1 AG or GG among GSTT1 present subjects, but when interaction of each genotype with GSTT1was tested in the logistic model, there was no statistical evidence of interaction between GSTT1 and GSTM1(P = 0.41), nor of interaction between GSTT1and GSTP1 (P = 0.78), nor between GSTT1 and combined categories GSTM1 null and GSTP1 AA or AG (P = 0.56).

An excess of the GSTT1 null genotype in cases was observed for never smokers as well as for smokers (Table 5). There was no indication of a trend in OR for GSTT1 null with pack-years of smoking. ORs for the association between heavy smoking (highest tertile of pack-years compared with never smokers) and RCC were 1.6 (95% CI, 0.7–3.4) and 2.0 (95% CI, 0.5–7.6) among GSTT1 present and GSTT1 null subjects,respectively. ORs for GSTM1 null and for GSTP1 AG and GG genotypes were near 1.0 for never smokers and for smokers, and across categories of pack-years of smoking.

The association between GSTT1 genotype and RCC varied across categories of BMI (Table 5). The elevated risk associated with GSTT1 null genotype occurred primarily among subjects in the lowest BMI category. A similar pattern was observed for GSTM1, with increased risk associated with GSTM1null genotype only within the lowest tertile of BMI. GSTP1genotype frequencies did not vary by BMI. ORs for the association between high BMI (highest tertile versus lowest tertile) and RCC by genotype were 2.3 (95% CI, 1.1–4.6) and 0.5 (95% CI,0.2–1.7) among GSTT1 present and GSTT1 null subjects, respectively, and 2.5 (95% CI, 1.0–6.2) and 0.8 (95% CI,0.4–1.9) among GSTM1 present and GSTM1 null subjects. The excess of GSTT1 genotypes was present both among cases with (17 of 57) and without (19 of 68) a history of high blood pressure.

We observed an excess of the GSTT1 null genotype in RCC cases compared with controls. There was no overall association between GSTM1 or GSTP1 genotypes and RCC and no effect modification of one GST genotype by the other genotypes.

Associations between recognized risk factors for RCC, smoking, high blood pressure, and BMI in this population-based case-control study were consistent with what has been reported in previous literature (1, 22, 23). Adjustment for these exposures did not affect the relationship between GSTT1 null genotype and RCC. The major selection factor that differed between participating and nonparticipating cases in our study was stage at diagnosis, which was unrelated to GSTT1 genotype among participating cases and,thus, seems unlikely to bias the observed association between RCC and GSTT1 null. For GSTM1 and GSTP1, the distribution of genotypes among the distant stage cases who were genotyped were similar enough to local and regional stage cases that bias, if any, related to lower participation by distant stage cases would be small. Although association between BMI and prognosis has been reported (24) in RCC cases, BMI was not associated with stage at diagnosis among cases in the present study, so stage bias is unlikely to have influenced the relationships among GSTM1, BMI, and case-control status.

The frequencies of the GSTM1, GSTT1, and GSTP1 genotypes among controls were comparable with what has previously been reported for study populations drawn from Caucasians in North America and Europe (4, 20, 25, 26). Evidence from other studies indicates that frequencies of the GSTM1 null and possibly the GSTT1 null genotypes vary by race (4). In our multivariate model, race-adjusted ORs for GST genotypes were similar to crude ORs, and the associations based on all subjects were qualitatively similar to results of models limited to non-Hispanic white subjects. The variation in the frequencies of GSTT1 genotypes among controls by county of residence was not explained by race, and is most likely attributable to chance.

The use of different tissue sources to obtain DNA for cases and controls is a potential limitation of these data. Pathological material was the source of DNA for most cases, whereas blood was used for controls. However, it is unlikely that the source of DNA introduced bias in measurement of genotypes. Tissue blocks containing only normal tissue were used for the majority of case subjects. In a study comparing sources of DNA for PCR-based genotyping, genotypes for GSTM1 and several other genes determined from fixed, paraffin-embedded tissue were 97–100% concordant with genotypes determined from fresh frozen tissue from the same subject (27).

Our finding of an increased frequency of GSTT1 null genotypes among cases suggests that activity of the GSTT1 enzyme protects against development of RCC. Because GST enzymes are active in detoxifying chemicals from cigarette smoke, researchers have hypothesized that the role of the polymorphisms in human cancer would be to modulate the effect of exposures to cigarette smoke. However, in this study the association between GSTT1 and RCC was similar for smokers and nonsmokers. GST enzymes catalyze conjugation of glutathione with a broad range of substrates (3),including lipid peroxides formed endogenously as a result of oxidative stress (9). Kidney tissue has high metabolic activity and oxygen demand, a situation in which enhanced endogenous formation of reactive oxidants is possible; a high level of somatic mutations in kidney compared with other tissues has been reported (28). The association between GSTT1 null genotype and increased RCC risk may be due to reduced protection against endogenous reactive oxidants.

The association between GSTT1 null genotype and RCC risk seemed to vary by BMI, with the strongest association among subjects in the lowest tertile of BMI who are otherwise at low risk of RCC. The association between GSTT1 and RCC among low-BMI subjects may indicate that different etiological pathways are involved in development of RCC in low-BMI versus high-BMI individuals. The mechanism for the association between high BMI and RCC is not well understood. If this mechanism involves reactive oxidants, a possible interpretation of our results is that protection by GSTT1 is important among individuals with low levels of exposure, but becomes overwhelmed and does not affect risk among the highly exposed. We assessed potential interaction between GSTT1 genotype and several exposure variables, so it is possible that the GSTT1-BMI interaction is a chance finding. Statistical power to detect the GSTT1-BMI interaction in the present study was 70% atα = 0.05.

The present study is the only report of increased risk of RCC associated with GSTT1 null. The inconsistent results compared with other studies may reflect chance variation, but it is possible that the role of GSTT1 is different in the presence of different patterns of exposures to environmental risk factors for RCC. The case-control study in France (14), which reported no association with GSTT1, provided no information on the subjects’ exposures to risk factors for RCC. In the study of solvent-exposed workers in Germany (13), GSTT1null genoype was associated with reduced risk of RCC (OR, 0.2; 95% CI,0.1–0.9), based on a small number of cases. The authors note that their result is consistent with a model of TCE toxicity to the kidney through GST-catalyzed formation of glutathione conjugates, which are further metabolized in the kidney to toxic compounds, a mechanism that may not be relevant in individuals without occupational solvent exposure. The decreased risk of RCC associated with GSTT1null genotype among TCE-exposed workers contrasts markedly with the increased risk observed in the present population-based study, in which we would expect that few subjects would have been exposed to TCE. The contrasting results may imply different effects of GSTT1 enzymes according to the presence of specific chemical exposures.

We thank the staff of the epidemiology research unit at Fred Hutchinson Cancer Research Center for its expertise in study management and data collection, Dr. Valle Nazar-Stewart for providing genotype data for the control subjects, and Joe Walker and Cassie Keener(Department of Environmental Health, University of Washington) for conducting the DNA extraction and genotyping.