Glutathione S-Transferase Polymorphisms and Survival in Primary Malignant Glioma

Patients with primary high-grade brain tumors have poor survival (1). Established prognostic factors including age at diagnosis, pathological grading, extent of surgery, and Karnofsky performance score insufficiently predict the outcome in most instances (2, 3, 4, 5, 6, 7). Although surgery, radiotherapy, and chemotherapy have improved patient outcomes, it is impossible to identify in advance which patients will benefit from such treatments. For some drugs, inherited variability in metabolism of therapeutic agents is suggested to be responsible, in part, for individual differences in response to cancer treatment (8).

Polymorphisms in glutathione S-transferase (GST) enzymes produce significant alterations in the metabolism of many substrates, including carcinogens and chemotherapeutic agents (9, 10, 11, 12, 13, 14). As a result, these polymorphisms have been suggested to have a role in susceptibility to cancer and response to chemotherapy (15, 16, 17, 18, 19, 20, 21). GSTs belong to a family of isoenzymes that catalyze the glutathione conjugation of a variety of electrophilic compounds, including carcinogens, mutagens, cytotoxic drugs, and their metabolites, and detoxification products of reactive oxidation (22, 23). Cytosolic GSTs are dimeric proteins divided into eight classes, α through θ, based on amino acid sequence and immunoreactivity, whereas potentially exhibiting overlapping substrate specificities (22, 24). They are highly heterogeneous proteins expressed in virtually all tissues, including brain and blood brain barrier (25, 26, 27). For the GSTM1 subclass, in the most common allele, GSTM1*0, the gene is deleted, and homozygotes (null genotype) that comprise 42–60% of the Caucasian population express no GSTM1 protein (28, 29, 30). GSTT1 is also polymorphic, and 13–26% of the Caucasian population have homozygous deletion, and, thus, lack GSTT1 function (29, 31, 32). GSTP1 polymorphisms are characterized by single nucleotide substitutions at A¹⁴⁰⁴G (exon 5) and C²²⁹⁴T (exon 6), resulting in amino acid changes Ile¹⁰⁵Val and Ala¹¹⁴Val, respectively (9, 33). The combinations of the two single nucleotide exchanges result in four possible alleles, *A, *B, *C, and *D, respectively. Forty-four percent of the Caucasian population carries GSTP1*A/*A (29). Allelic variation causes a steric effect at the substrate-binding site of the enzyme, which leads to a substrate-dependent functional modification (9, 10, 12, 13, 34, 35). Polymorphisms in the GST family of enzymes have been associated recently with survival in patients with leukemia, breast cancer, and metastatic colorectal cancer (36, 37, 38). GSTs catalyze detoxification of many alkylating agents, including nitrosoureas, platinum compounds, and melphalan, that are used in malignant brain tumor chemotherapy (39, 40, 41, 42, 43). They may also detoxify the free radicals formed by chemotherapy drugs, such as procarbazine, and radiation or sequester drugs such as alkylating agents and steroids, by direct binding (44, 45, 46).

In this study we examined the relationship between polymorphisms in the GST genes and survival in 278 patients with malignant glioma. Our main hypothesis was that patients who have GST genotypes that encode for high-activity enzymes toward the cytotoxic chemotherapeutic agents will have poorer overall survival than will patients with genetically determined low or no detoxification activity. Furthermore, we explored whether genetic polymorphisms of the GST family were correlated with a higher occurrence of adverse effects secondary to chemotherapy. We hypothesized that patients with no or low activity GST genotypes (GSTP1*A, GSTM1 null, and GSTT1 null) would have decreased metabolism of chemotherapy agents and, therefore, would have a higher incidence of adverse treatment effects than would individuals with higher activity GST genotypes.

The population for this study is a subset (n = 1594) of primary malignant glioma patients consecutively diagnosed and treated at The University of Texas M. D. Anderson Cancer Center who participated in a family study of malignant gliomas (RO1CA70917) between 1994 and 2000. We selected all patients younger than age 65 at diagnosis who had complete medical records, an available DNA sample isolated from peripheral blood mononuclear cells, who were diagnosed <1 year before the registration date, and had consented to participate in the study, which was approved by the institutional human subjects review committee. Patients who had recurrence or progression at the time of registration and who were <20 years of age at diagnosis were excluded. We also excluded nonwhite patients, because only 1 Hispanic patient matched our inclusion criteria out of the 136 Hispanic and African American participants in the above mentioned family study. Our total study population comprised 278 patients. Patients were divided into three arbitrary groups based on the malignancy of their tumor, high, medium, and low (Table 1). We reviewed the medical records and abstracted the following information, patient characteristics at diagnosis; treatment (extent of surgery; irradiation dose, field, and duration; and type, dose, and duration of chemotherapy); adverse effects secondary to chemotherapy (grade III or IV myelosuppression, neurotoxicity, and skin reactions that resulted in dose reduction or cessation of therapy); and date of death or last follow-up. Among surviving patients, the date of last contact was November 30, 2001. We abstracted all of the treatment information before recurrence or progression, if it occurred.

A multiplex PCR technique was used to amplify both GSTM1 and GSTT1 simultaneously in a single PCR reaction (30). Briefly, isolated DNA was amplified using the following GSTM1 primers: 5′-GAA CTC CCT GAA AAG CTA AAG C-3′ and 5′-GTT GGG CTC AAA TAT ACG GTG G-3, and the following GSTT1 primers corresponding to the 3′ coding region of the human cDNA, 5′-TTC CTT ACT GGT CCT CAC ATC TC-3 and 5′-TCA CCG GAT CAT GGC CAG CA-3. As an internal control, dihydrofolate reductase gene was coamplified using the primers 5′-GCA TGT CTT TGG GAT GTG GA-3′ and 5′-GGA ATG GAG AAC CAG GTC TT-3′. The PCR conditions consisted of an initial melting temperature of 95°C (5 min) followed by 35 cycles of melting (95°C, 30 s), annealing (58°C, 45 s), and extension (72°C, 1 min). The PCR products from coamplification of GSTT1 (480 bp), dihydrofolate reductase (280 bp), and GSTM1 (215 bp) were then viewed by an ethidium bromide-stained 2% agarose gel for the presence or absence of GSTM1 and GSTT1 genes. In 10% of the samples PCR was repeated. In univariate and multivariate analyses null patients with no enzyme activity were compared with the non-null patients.

We used a similar method described by Harries et al. to determine the GSTP1 variant at codon 105 (47). Briefly, genomic DNA was amplified with the following primers, 5′-AAT ACC ATC CTG CGT CAC CT- 3′ and 5′-TGA GGG CAC AAG AAG CCC CTT- 3′ in PCR buffer [10 mm Tris-HCl, 50 mm KCl, and 2 mm MgCl₂ (pH 8.3)] with 200 μmol deoxynucleoside triphosphate and 5 units of Taq polymerase. The PCR conditions consisted of an initial melting temperature of 95°C (5 min) followed by 32 cycles of 95°C (30 s), 60°C (45 s), and 72°C (45 s) in 25 μl volume. A final polymerization step of 72°C for 10 min was carried out to complete the elongation process and yield a 568-bp fragment. The PCR product (8 μl) was then digested with 5 units of BsmAI (New England Biolabs, Beverly, MA) for 16 h at 58°C. The samples were analyzed by electrophoresis on an ethidium bromide-stained 1.5% agarose gel. The presence of the Ile/Ile allele was revealed by 305-, 135-, and 128-bp fragments, whereas the Val/Val allele was revealed by 222-, 135-, 128-, and 83-bp fragments. The heterozygote Ile/Val allele was characterized by five fragments consisting of 305, 222, 135, 128, and 83 bp. The fragments 135 bp and 128 bp cannot be separated.

For the codon 114 polymorphism, 170-bp genomic DNA was amplified with the following primers, 5′-ACA GGA TTT GGT ACT AGC CT-3′ and 5′-AGT GCC TTC ACA TAG TCA ATC CTT G-3′ by PCR conditions as described above with an annealing temperature of 56°C (48). The PCR product (6 μl) was then digested with 6 units of AciI (New England Biolabs, Beverly, MA) for 16 h at 37°C. The samples were analyzed by electrophoresis on an ethidium bromide-stained 2.5% Nusieve3:agarose gel. The presence of the Ala/Ala allele was revealed by a completely digested single 144-bp fragment, whereas the Val/Val allele was revealed by an indigestible fragment of 170 bp. The heterozygote Ala/Val allele was characterized by two fragments consisting of 170 and 144 bp. We repeated the analyses in 10% of all of the samples, and in all of the samples with Ala/Val and Val/Val polymorphisms. We analyzed GSTP1 results by comparing patients with the *A/*A genotype (Ile at exon 5 and Ala at exon 6) with the rest of the group, who had at least one variant allele. We hypothesized that patients with the *A/*A genotype would have lower activity against the nitrosourea agents used primarily in our patient population.

We computed basic descriptive statistics for age at diagnosis, gender, histology, extent of surgery, radiation therapy, chemotherapy, and frequencies of GSTM1, GSTT1, and GSTP1 variants. We first evaluated the data from all of the glioma patients together, and then by the histology group, high, medium, and low separately. We did not conduct multivariate survival analysis for low-group patients, because there were too few events.

We applied the Kaplan-Meier procedure to estimate survival (49). Survival time was calculated from the date of M. D. Anderson registration, date of death, or date of last follow-up. Age at diagnosis was included as a continuous variable and also dichotomized into two groups based on the mean age (for patients in each histology group). We computed Cox proportional hazard regression method to estimate the effect of GST polymorphisms on survival in malignant glioma patients in the presence of other known prognostic factors (50). We identified an interaction between tumor histology and GST polymorphisms, and therefore created second- and third-order interaction terms. We calculated hazard ratios (HR) and their corresponding 95% confidence intervals (CI). When exploring the importance of three-way interaction among histology, GSTP1, and GSTM1, to calculate HRs for the genotype combinations we used the main model with all of the patients. Age at diagnosis and the variable for difference of time (months) between diagnosis and registration at M. D. Anderson were entered in the model on a continuous scale. All of the other variables are categorical as shown in Tables 2 and 3. Unconditional logistic regression analysis was used to estimate odds ratios and 95% CIs for the effect of GST genotypes on the occurrence of adverse events secondary to chemotherapy, adjusting for age at diagnosis, gender, histology, and radiotherapy. For all of the analyses we used SAS (Version 8.02; Statistical Analysis System, Inc., Cary, NC) or SPSS (Version 10.1.3; SPSS Inc., Chicago, IL) software packages. All tests of statistical significance were two sided.

The mean and median age for the overall group was 46 and 45 years, respectively (range, 20–64 years). One hundred and sixty two (58.3%) patients were males. All were Caucasian. By histology, 156 (56%) patients were classified as high, and 78 (28%) were classified as medium (Table 1). The distribution of the demographic characteristics, clinical variables, and genotypes studied are shown in Table 2. The distribution of the polymorphisms was very similar to what has been reported for Caucasians in the literature, for the whole group, and for the three histology groups.

The median survival for the whole group was 21.2 months (95% CI, 17.4–25.0 months), and overall survival was 45% (95% CI, 39–51%) at 2 years and 36% (95% CI, 29–42%) at 3 years. In Table 3, we show median survival time in months for each variable studied and the interaction variables created for the GST polymorphisms. Females and patients who had low histology group (Fig. 1), who had undergone gross total resection, who were younger than age 45 at diagnosis, who were treated with chemotherapy, and who had radiation therapy experienced significantly longer survival (P < 0.05). Figs. 2 and 3 show Kaplan-Meier survival curves for the GSTP1 and GSTM1 combination stratified for histology. In the medium group, patients with GSTP1*A/*A-M1 null genotype survived longer than did the rest of the group (median survival was not reached and 41 months, respectively; P = 0.06). Of note, this outcome observed in patients with medium histology group who had GSTP1 *A/*A-GSTM1 null combination genotype was very similar to all of the patients with low histology tumors (see Figs. 1 and 2). For the anaplastic astrocytoma and anaplastic oligodendroglioma subgroups in the medium histology group patients with GSTP1 *A/*A-GSTM1 null genotype had longer survival (data not shown). However, the number of patients were very small in the GSTP1 *A/*A-GSTM1 null groups to conduct a meaningful statistical analysis (n = 6 with 2 deaths in anaplastic astrocytoma and n = 8 with 1 death in anaplastic oligodendroglioma). In the high histology group, there was no difference between the comparison groups (median survival, 14.0 months and 13.3 months, respectively; P = 0.85).

Table 4 shows the results of multivariate analyses for all 278 patients. In concert with what has been reported previously, among the entire group, younger age, low histology group, gross total tumor resection, radiation treatment, and chemotherapy were independently associated with longer survival. In the high histology group, same variables were significantly related to better outcome (data not shown). In the medium histology group, although the direction of the HRs was identical, chemotherapy was the only variable that remained significantly associated with longer survival. Compared with patients who did not receive chemotherapy, patients with medium histology tumors who were treated with chemotherapy had 80% lower risk of death (HR = 0.2; 95% CI, 0.08–0.51) compared with the rest of the group. None of the genotypes that we studied were related to survival individually.

When we created interaction variables between the main clinical variables and the genotypes, we observed a three-way interaction among histology, GSTP1, and GSTM1 polymorphisms (P = 0.017). Therefore, we calculated the HRs for each of the four possible genotype combination between GSTP1 and M1 in high and medium histology groups (Table 5). In both histology groups, 95% CIs for the HRs were including 1, and the results were statistically insignificant. In the medium group, patients with the GSTP1 *A/*A and GSTM1 null genotype had the least HR (likelihood ratio test, P = 0.13). R² values for the model with only main demographic and therapy effects; for the model with main effects and main genotypes; and for main effects, main genotypes, and the three-way interaction variable were 12.4%, 12.7%, and 13.7%, respectively.

One hundred and fifty-nine patients were treated with 34 separate regimens of chemotherapy before a recurrence. One hundred and thirty (82%) were treated with an alkylating agent-based regimen; a nitrosourea drug was used in 106 (67%). We had complete toxicity information for 154 patients, and for 150 of these we also had genotyping results for the selected GSTs. Ninety-eight patients (64%) experienced adverse events that prompted dose reduction or cessation of their chemotherapy.

No deaths resulted from any of the adverse events. About 70% of the female patients (45 events in 63 patients) suffered toxicity due to chemotherapy compared with 58% of the male patients (53 events in 91 patients). We conducted logistic regression analyses for all of the patients who were treated with any chemotherapy and for those who were treated with a nitrosourea compound only (Table 6). In the model that included histology, age at diagnosis, gender, radiation therapy, and the genotypes, gender was the only variable related to toxicity significantly in the nitrosourea analysis. When we explored possible interactions between genotypes, we again discovered an interaction between GSTP1 and GSTM1 polymorphisms. Patients with the combined GSTP1*A/*A and GSTM1 null genotype were 2.0 times (95% CI, 0.5–8.1) more likely to suffer from toxicity compared with the rest of the patients (Table 7). In the nitrosourea group analysis, patients with the combined GSTP1*A/*A and GSTM1 null genotype were 5.7 times (95% CI, 0.9–37.4) more likely to experience an adverse event secondary to chemotherapy, compared with the others.

This is the first report that suggests genetic polymorphisms in the GST family are related to survival and to toxicity secondary to chemotherapy in a large group of patients with primary malignant glioma. Others have also attempted to examine a potential association between GST polymorphisms and cancer outcome. Two studies published by the Berlin-Frankfurt-Munster group suggested a possible relation among the GSTM1 and GSTT1 null genotype, non-GSTP1*A/*A genotype, and superior outcome in childhood lymphoblastic leukemia (19, 21). A larger study conducted by the Children’s Oncology Group in >600 children found no association between GSTM1 or GSTT1 polymorphisms and survival in childhood acute lymphoblastic leukemia (51). In children with acute myeloid leukemia, two different groups reported a significant relationship between increased risk for death due to toxicity and GSTT1 null genotype (52, 53). In a study of >200 women with primary breast cancer, GSTT1 and GSTM1 null, and GSTP1 *B or *C variants were related to superior outcome (36, 38). More recently, in 107 patients with metastatic colorectal carcinoma who were uniformly treated with 5-fluorouracil and oxaliplatin, those who had GSTP1 *B or *C variants had a significantly longer survival (37). All of these studies, including ours, are hampered by at least one of the factors, such as inadequate epidemiological design and analysis that does not account for all of the known prognostic variables related to outcome; inability to control for nonuniform utilization of multiple chemotherapy agents without definitive evidence for many to be metabolized by the GST enzyme family; lack of knowledge regarding the differential metabolism of the agents by the different GST alleles; and finally, failure to include the level of tumor GST protein expression.

Our case-only study was not designed to investigate the relationship between GST polymorphisms and the risk of developing malignant glioma. However, the proportions of the different allelic variants in the three histology groups were very similar to what has been published previously (29). This suggests that GSTM1, GSTT1, and GSTP1 polymorphisms probably are not potential risk factors for the development of malignant glioma in adults. It remains to be established whether this is true for every type of glial tumor, represented in the multiple subgroups of our three main histology groups. A recent report showed that the presence of GSTM1 and GSTP1 Val114Val alleles was significantly related to the risk of developing pediatric astrocytomas (54).

Concerning survival outcome, the three-way interaction among histology, GSTP1, and GSTM1 polymorphisms that we observed post-hoc should be interpreted cautiously and needs additional validation. Our study hypotheses were based on the premise that chemotherapy dose at the tumor site will be modulated by the differential metabolism of the agents that are conjugated by the GST enzyme family. In line with our primary hypothesis, patients with anaplastic astrocytoma, anaplastic oligodendroglioma, and anaplastic oligoastrocytoma, who had the GSTP1*A/*A and GSTM1 null combination genotype, had the best outcome. In GSTP1 variants, the corresponding amino acid transitions cause a steric change at the substrate-binding site of the enzyme, without affecting the glutathione-binding site (9). Therefore, enzyme activity toward electrophilic compounds is significantly different for the particular alleles of different substrates. For 1-chloro-2,4-dinitrobenzene, thiotepa, and chlorambucil, the GSTP1*A allele has significantly higher activity compared with GSTP1*B and GSTP1*C alleles (9, 10, 12, 13, 14, 24). However, in glutathione conjugation of cisplatin, carboplatin, 4-hydroxyifosfamide, and epoxides of benzo(a)pyrene, the GSTP1*B and GSTP1*C alleles have increased activity compared with GSTP1*A allele (10, 11, 24, 34, 35, 55). To our knowledge, no published study has examined the differential activity of the GSTP1 alleles for nitrosourea compounds. The relation between the combined GSTP1*A/*A and GSTM1 null genotype, and longer survival that we observed in the medium histology group is additionally supported by the increased risk of adverse events due to chemotherapy of that group.

Our study was limited in a number of ways. We are well aware that the relationships between the combined GSTP1*A/*A and GSTM1 null genotype and survival in the medium histology group and toxicity outcome are not statistically significant. We did not have sufficient power to reach the 0.05 significance level, because we did not have enough patients in this particular and many other analyses. Increased rate ratios are supported with our a priori hypotheses, and the literature related to the field. In 56 (20%) patients Karnofsky Performance Scores after primary surgery were missing. Another important variable that we were unable to include in our analyses was level and localization of GSTP1 protein expression in the tumors. Both were associated previously with survival in patients with malignant glioma (2, 56). Although our study goals and hypotheses were indirectly related to treatment factors and decisions, the results may have been affected by treatment heterogeneity and selection bias for choosing chemotherapy.

In conclusion, we present our preliminary findings that polymorphisms in GST genes affect outcome in primary malignant gliomas. In patients with malignant gliomas for which there is previous evidence for sensitivity to chemotherapy and radiation, the combination of germ-line GSTP1*A/*A and GSTM1 null genotype confers a survival advantage. Patients with the same genotype also have an increased risk of adverse events secondary to chemotherapy, primarily comprised of nitrosourea alkylating agents. To confirm that this genotype is an independent predictor of outcome, in a larger study with sufficient power we will address whether the observed relationship can be validated.

We thank Marianne Doran from the Department of Scientific Publications for assistance in editing the manuscript.