Glutathione-S-Transferase Variants and Adult Glioma

Polymorphic variants in the glutathione-S-transferases (GSTs) μ (GSTM1), 𝛉 (GSTT1), and π (GSTP) have been studied extensively in relation to cancer etiology because these enzymes, part of the phase II detoxification processes, are involved in metabolism of many electrophilic compounds including carcinogens, mutagens, cytotoxic drugs, and their metabolites and detoxification of products of reactive oxidation (1, 2). Complete gene deletions in GSTM1 and GSTT1 and single nucleotide polymorphisms in GSTP may result in a significant change in the function of the enzymes (1, 3, 4). Although there are a few rare single gene alterations known to substantially increase the risk of neuroepithelial tumors and therapeutic radiation is a known strong environmental risk factor, these factors account for a small proportion of cases (5). These observations combined with the temporal and geographic distribution of primary malignant brain tumors suggest that most of these tumors arise from a complex interplay of genes, developmental processes, and exogenous and endogenous environmental stimuli. Five studies have reported on the associations of GSTs and adult-onset primary brain tumors (6–10) with somewhat inconsistent results that will be discussed in more detail below. We report on GST genotypes in population-based adult glioma cases and controls from the San Francisco Bay Area.

Details of case-control ascertainment for these two series of subjects have been presented in detail elsewhere (11–13). We ascertained all adults newly diagnosed with glioma (International Classification of Diseases for Oncology, morphology codes 9380–9481) in six San Francisco Bay Area counties (Alameda, Contra Costa, Marin, San Mateo, San Francisco, and Santa Clara) from August 1991 to April 1994 (series 1) and from May 1997 to August 1999 (series 2). Cases were ascertained within a median of 7 weeks of diagnosis using Northern California Cancer Center's Rapid Case Ascertainment program as described previously (12, 13). Controls ascertained through random-digit dialing using methods described previously (13) were frequency matched to cases by age, race, and gender. We began collecting blood specimens from willing subjects partway through the first series and asked all participants in series 2 to donate either blood and/or buccal specimen. Our previous reports (6, 10) included about 150 white cases and an equal number of controls.

Constitutive DNA was obtained from subjects' buccal or blood specimens. Genotyping for homozygous deletion of GSTM1 and GSTT1 has been described previously (6, 10, 14). For GSTP A114V, the PCR was done with primers CAAGCAGAGGAGAATCTGGG and AATGAAGGTCTTGCCTCCCT. Cycle condition is 94°C for 5 min followed by 36 cycles of 94°C for 30 min, 60°C for 30 min, and 72°C for 30 min. Fifteen microliters of the PCR product were then digested overnight with Cac8I at 37°C. Genotype is determined on 4% agarose gel. For GSTP I105V, the PCR is done with primers GTAGTTTGCCCAAGGTCGAG and AGCCACCTGAGGGGTAAG. Cycle condition is 94°C for 5 min followed by 32–40 cycles of 94°C for 30 min, 58°C for 30 min, and 72°C for 60 min. Fifteen microliters of the PCR product were then digested with BsmA1 at 55°C for at least 2 h. Genotype is determined on 2.5% agarose gel.

As described previously (15), we determined p53 mutation status of astrocytic tumors using single-strand conformation polymorphism followed by sequencing to classify tumors as containing or not containing p53 mutation.

We began by examining genotype frequencies for cases (all, glioblastoma, and nonglioblastoma) and controls by series, by gender within series, and for whites within series. For all ethnicities and whites separately, we also examined genotype frequencies by age group ≤40, 41–60, and >60 for controls, all cases, and cases who either had glioblastoma, other astrocytic tumors (anaplastic astrocytoma, astrocytoma, and astrocytoma not otherwise specified), tumors with an oligodendroglial component (oligodendroglioma or oligoastrocytoma), or other (ependymoma, medulloblastoma, juvenile pilocytic astrocytoma, ganglioblastoma, and other). We also compared genotype frequencies by p53 mutation status of the tumors.

Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated using logistic regression. Models were run separately for each type of gene adjusting for matching factors, age, gender, and ethnicity. Final models for whites included adjustment for age, gender, series, and all four genes (GSTM1 and GSTT1 null versus present, GSTP I105V Val/Val versus Ile/Val or Ile/Ile, and GSTP A114V Ala/Val or Val/Val versus Ala/Ala). SAS software (16) was used for data management and PROC LOGISTIC was used to estimate all ORs and 95% CIs. We also examined cases, histological subtypes, and controls for distributions for various combinations of presence or absence GST variants. We also summed the number of GST variants that each individual carried, counting one for deletion of GSTM1 or GSTT1 and one for each variant allele of GSTP I105V and A114V.

As described in detail elsewhere (11–13), there were 1129 eligible cases from both series 1 and 2. Of these, full interview was obtained for 81% (n = 896) and 2% (n = 23) completed only a telephone interview. Only 11% (n = 128) refused to enter the study, 1% (n = 22) were too ill or had a language problem, 4% (n = 47) could not be located, and 1% (n = 13) did not get a doctor's consent to contact. Of those cases with completed interview, 2% (n = 17) had to be dropped from the study because a neuropathology review indicated that the subject either did not have glioma or medulloblastoma (n = 4), permission for review was not obtained (n = 4), tumor specimens were unavailable (n = 4), or tumor specimens were insufficient for diagnosis (n = 5). Of the 879 case participants, 461 (52%) provided blood or buccal specimen, and GST genotyping was conducted for 458 subjects (52% of total) with 187 of 476 (39%) from series 1 and 271 of 403 (67%) from series 2.

Of the 15,894 phone numbers in both series contacted through random-digit dialing, 8% (n = 1316) yielded eligible controls. Of the total phone numbers contacted, 19% (n = 3086) were not in service, 14% (n = 2242) were businesses, 7% (n = 1047) were faxes or modems, 12% (n = 1964) had no response after 10 calls, 12% (n = 1824) were refusals, 5% (n = 827) had language or health problems, 20% (n = 3158) were eligible but the quota for their age/race/gender had been filled, and 3% (n = 430) were either too young, had multiple lines, lived out of the area, or closed out. Of the eligible controls, 66% (n = 864) completed a full interview, 13% (n = 169) completed an abbreviated telephone interview only, 14% (n = 184) refused to enter the study, 1% (n = 17) either had a language problem or were too ill to interview, 3% (n = 43) were not located, 3% (n = 39) were either related to cases, out of the area, good matches but the study closed, or otherwise could not be used. Of the 864 controls from series 1 and 2, 558 (65%) provided blood or buccal specimens and 487 (56%) were genotyped for GSTs with 152 of 462 (33%) from series 1 and 335 of 402 (83%) from series 2.

For all participants and those subjects genotyped, Table 1 shows the numbers of subjects, basic demographic characteristics (age, gender, and ethnicity), and histological distribution of cases. There were some differences between the series; most notably, a higher proportion of subjects were genotyped from series 2 than series 1 because funding for collecting specimens was first obtained partway through series 1. The percentage of ethnicities categorized as “other,” which consists of all ethnicities other than white or African American, increased from 11% to 15% of study participants. Series 2 participants were on average, slightly older than series 1. The mean age was 54 years for series 1 cases and 56 years for series 2. These changes may reflect the changing demographics of the Bay Area over the period of 5 years between the studies. In addition, note that although the ages of cases and controls were matched for the overall group, genotyped cases are on average younger than genotyped controls, primarily because case survival (and therefore ability to obtain constitutive material for genotyping) decreases with age. The histological distributions of cases in the two series were similar, except that series 2 had a higher percentage of cases with oligodendroglioma and lower percentage with oligoastrocytoma. Because the proportion of oligodendroglioma and oligoastrocytoma changed substantially between series 1 and 2, we combined these two histologies in subsequent analyses as tumors with an oligodendroglial component.

No noteworthy or significant age, gender, and ethnicity adjusted case-control differences are apparent for the various GST genotypes in either series (Table 2); in addition, there were no notable differences in genotype frequencies for glioblastoma or cases with other histologies. Table 3 shows that there were, however, some very large differences in genotype frequencies by ethnic group within both cases and controls, which were highly statistically significant among cases. Because we lack sufficient numbers of subjects among the non-white ethnicities to fully explore the basis for these differences, we limited our subsequent analyses to whites.

As shown in Table 4, among whites, cases and controls had very similar genotype frequencies for the four GST genotypes studied; multivariate adjustment for all four genes as well as age, gender, and series supported this similarity. There were no marked or significant differences in genotype frequencies among cases with different histological types (Table 5); however, GSTT1 deletion was somewhat less common among patients with astrocytoma or anaplastic astrocytoma and GSTP A114V Ala/Val or Val/Val was more common among patients categorized with “other” histologies.

There appeared to be an increasing trend with age for the association of GSTT1 null genotype and oligodendroglial tumors, with 10%, 25%, or 40% of cases aged ≤40, 41–60, and >60 having this genotype while 20%, 18%, and 23% of controls in these ages had this genotype; however, the interaction of GSTT1 null genotype and age was not significant (P = 0.27 for the test of equal slopes of age with case-control status for GSTT1 null and GSTT1 present genotypes).

The following significant or nearly significant differences were noted in analyses that stratified subjects by histological type and/or age group. Among subjects over age 60, GSTP I105V Val/Val was less common in cases (6.4% of 108) than controls (15% of 176; OR 0.38; 95% CI 0.15–0.93; P = 0.03); a similar result obtained for glioblastoma cases versus controls. Over all age groups and among those under age 40, GSTP A114V Ala/Val or Val/Val was more common in cases with “other” histologies than controls [28% of 32 other histology cases and 16% of 428 controls had these genotypes (OR 2.5; 95% CI 1.0–6.0; P = 0.05); among those under age 40, 35% of 17 other histology cases and 16% of 69 controls had these genotypes (OR 3.2; 95% CI 0.9–12.1; P = 0.08)].

We found no notable or significant differences among histological subgroups, cases, or controls in the distributions of combinations or numbers of variants in GSTM1, GSTT1, GSTP I105V, or GSTP A114V.

Among white glioblastoma multiforme (GBM) cases, 13% of 116 subjects with tumors lacking p53 mutation and 30% of 23 subjects whose tumors had p53 mutation had GSTT1 null genotype (OR 2.8; 95% CI 0.93–8.4; P = 0.07). For the other GST variants, differences in genotype frequencies by tumor p53 status did not approach statistical significance.

Five previous studies (two from the first series of subjects included here) reported on associations of GSTs with adult glioma (6–10).

In a hospital-based study of 109 glioma cases and 577 controls, Caucasians from North Staffordshire, United Kingdom, Elexpuru-Camiruaga (7) et al. first showed that GSTT1 null genotypes were significantly associated with astrocytoma with 32% of cases versus 18% of controls having this genotype. About 60% of the cases and 54% of controls were null for GSTM1 gene, a nonsignificant difference. Respective proportions in the healthy Caucasians have been reported to be 42–60% for GSTM1 homozygous deletion and 13–26% for GSTT1 homozygous deletion (17). In a hospital-based study of 90 malignant glioma cases (49 with GBM) and 90 blood donor healthy controls, Trizna et al. (9) found no statistically significant associations between GSTM1 and GSTT1 polymorphisms and risk of gliomas in adults; 52% of cases and 48% of controls were GSTM1 null and 28% of cases and 30% of controls were GSTT1 null. Recently, De Roos et al. (8) reported on deletions in GSTM1 and GSTT1 in 422 adults with glioma and 604 controls who were part of the National Cancer Institute's hospital-based study of brain tumors in three hospitals in Phoenix, AZ, Pittsburgh, PA, and Boston, MA. They found no significant difference between cases overall and controls for GSTM1 or GSTT1 null genotypes; 53% of cases and 56% of controls were GSTM1 null and 20% of cases and 18% of controls were GSTT1 null. Among cases of different histological types (glioblastoma, anaplastic astrocytoma, other astrocytoma, oligodendroglioma, and oligoastrocytoma), significant heterogeneity in genotype frequencies was noted for GSTM1 (P = 0.03); subjects with oligodendrogliomas had lower than expected percentage of GSTM1 null genotype (33%). In this current study, we found no case-control differences overall for GSTM1 or GSTT1 deleted versus nondeleted genotypes and no indication of differences by histological types. In the combined series, we noted a nonsignifcant deficit of GSTT1 deleted cases who had glioblastoma and a significant excess of cases who were GSTT1 deleted whose glioblastoma tumors contained p53 mutation compared with those whose glioblastoma tumors did not contain p53 mutation.

In our previous reports that included most of the same subjects included here in series 1, we also reported no significant relationship of either the GSTT1 or the GSTM1 null genotypes with occurrence of glioma overall in about 160 cases and 160 controls; 53% of cases and 50% of controls were GSTM1 null and 19% of cases and 20% of controls were GSTT1 null (6, 10). We reported that 44% of 16 patients with oligodendroglioma were GSTT1 null genotype yielding a 3-fold OR of 3.2 (95% CI 1.1–9.2) (6). There were some differences in the histopathology categorization in the two earlier reports compared with this current paper because the uniform neuropathology review had not yet been completed. In the finalized uniform review used in this paper, 41% (7 of 17) of white patients from series 1 with oligodendroglioma had GSTT1 null genotype; the equivalent number for series 2 is 19% (8 of 43) patients (i.e., the second series of subjects failed to replicate the finding in the first series). As mentioned above, there was a rather substantial difference in subjects categorized as oligodendroglioma alone versus oligoastrocytoma in series 1 and 2, so we combined these two diagnoses for the present paper; however, to try to understand whether the differences in GSTT1 null genotypes among oligodendroglioma cases between series 1 and 2 might be due to differential classification, we also examined the frequency of GSTT1 null genotype in whites with oligoastrocytoma; these were 11% (3 of 27) in series 1 and 22% (2 of 9) in series 2. In the earlier report on GSTM1 using series 1 cases, we found that GSTM1 null genotype was related to earlier onset of glioma among the female cases but not in male cases (10); in that report, among white women with glioma, the mean age at diagnosis was 43.9 years for 32 GSTM1 deleted women and 52.4 years for 29 GSTM1 nondeleted women; the means were very similar among the women with glioma from series 1 included in this current paper: 34 GSTM1 deleted women (43.4) and 29 GSTM1 nondeleted women (52.4). However, in series 2, there was no similar difference in the age at diagnosis by GSTM1 genotype (numbers of women and mean age at diagnosis for those with and without GSTM1 deletion were 46 women with average age of 55.5 years and 51 women with average age of 52.9 years, respectively).

This is the first study to evaluate GST genotype frequencies by tumor type categorized according to molecular alteration. The finding of a higher rate of GSTT1 deletion among subjects with glioblastoma tumors with p53 mutation might be expected based on the idea that such tumors might be initiated by mutations caused by endogenous or exogenous substances, the dose of which might be influenced by the presence or absence of GSTT1. It is also possible that the finding was due to chance, given the many comparisons performed in these analyses.

De Roos et al. (8) also examined two variants in GSTP, I105V and A114V. They found no case-control differences in GSTP A114V variants; 15% of cases and 14% of controls were Ala/Val or Val/Val. However, for GSTP I105V, 17% of cases versus 10% of controls were Val/Val yielding an age, gender, race, and other factor adjusted OR of 1.8 (95% CI 1.2–2.7). In addition, notable was that an even higher percentage of cases age 60 and younger had Val/Val genotype than those over age 60 (20% versus 11%), but no similar trend with age was seen in controls. We found no overall case-control differences in percentages with GSTP I105V Val/Val genotypes (9% of cases and 13% of controls). However, there were some case-control differences with age in the present study (i.e., cases over age 60 were significantly less likely than controls to be Val/Val genotype). De Roos et al. (8) reported nearly significant heterogeneity in GSTP A114V genotype frequencies (P = 0.06) with higher than expected GSTP A114V genotypes of Ala/Val or Val/Val (25%) among subjects with oligodendroglioma. We did not observe this heterogeneity but did find a nearly significant higher occurrence of these genotypes among those with “other” histologies (a category that includes a very heterogeneous collection of tumors including juvenile pilocytic astrocytoma, medulloblastoma, ependymoma, ganglioglioma, and other).

Because there is no consistent overall association of the GST variants studied with adult glioma in the two series of this current study or in the other reported studies, we conclude that there are no strong effects of GST variants on glioma development. However, results of all the studies published thus far include too few subjects in various subsets of glioma to reach definitive conclusions about the associations of GST variants with the non-GBM histologies or molecularly defined subsets of glioma. Our study showed that there can be very substantial heterogeneity of these variants among the cases within series (e.g., GSTM1 genotypes in GBM cases was 45% in series 1 and 54% in series 2). This suggests that GST genotypes might influence the survival rate for glioma. Evidence both for and against associations of GST genotypes with prognosis for several cancer sites has been reported (for recent examples, see 18–26). As noted previously, specimen collection began partway through series 1 and overall specimen collection rates and percentages of subjects' genotypes are substantially higher for series 2 than series 1. Consequently, selection bias in specimen collection (and genotyping results) related to survival is more likely in series 1 than in series 2. A detailed analysis of GST genotypes related to survival is beyond the scope of this paper, as survival follow-up and collection of pertinent treatment data are not yet complete. However, such an association is possible given that GSTM1 null genotype in GBM cases was associated with days between diagnosis and specimen collection (OR 1.007; 95% CI 1.003–1.012 for series 1; OR 1.003; 95% CI 1.00–1.01 for series 2) controlling for age, gender, and ethnic group. In addition, there was significant heterogeneity in some GST genotypes by ethnic group that was stronger in cases than controls (see Table 3). This points both to the need for very careful control of case-control differences in ethnicity in evaluating possible associations between genotypes and disease status and to consider the effects of these genotypes on survival. It also suggests the benefits of different study groups employing different case-control designs in trying to understand whether true relationships exist between genotypes and disease. For example, the De Roos et al. (8) study had the advantage that hospital-based cases would minimize the loss of subjects due to gene survival associations; however, their control group was also hospital based, potentially obscuring associations if gene variants were associated with the other diseases. This present study has the advantage that both cases and controls were population based but has the disadvantage that blood or buccal specimens could not be obtained from cases with poorest survival. Similarities of genotype frequencies in the two different types of studies lend support to validity of consistent findings.

Although there are no overall main effects of these GST variants with glioma etiology, these genes might be found to be important in conjunction with appropriate environmental exposures. Unfortunately, very few strong environmental risk factors have been identified for adult glioma. Perhaps, combining environmental and genetic data from these relatively large series of subjects from several study sources, along with uniform histological and molecular categorization of tumors, will eventually lead to a better understanding of risk factors for adult glioma.

We thank Dr. Richard Davis for pathology review of series 1 cases; the pathology departments of Alexian Hospital, Alta Bates Medical Center, Brookside, California Pacific Med Center, DR Pinole, Eden Hospital, El Camino Hospital, Good Samaritan, Highland Hospital, John Muir, Kaiser Redwood City, Kaiser San Francisco, Kaiser Santa Teresa, Los Gatos Hospital, Los Medano Hospital, Marin General, Merrithew, Mills Peninsula Hospital, Mt. Diablo Hospital, Mt. Zion Medical Center, Naval Hospital, O'Connor Hospital, Ralph K. Davies Medical Center, Saint Louise, San Francisco General, San Jose, San Leandro, San Mateo County, San Ramon Valley, Santa Clara Valley, Sequoia, Seton Medical Center, St. Francis, St. Luke's, St. Rose, Stanford, Summit, UC San Francisco, Valley Livermore, Veterans Palo Alto, Veterans SF, and Washington Hospital for providing tumor specimens for review and molecular analyses; and Joe Patoka and Pengchin Chen for p53 mutation analyses of tumors.