Glutathione S-Transferase M1 and Susceptibility to Nasopharyngeal Carcinoma¹

There is increasing evidence that genetic polymorphisms for enzymes that metabolize carcinogenic constituents of tobacco smoke are important in determining individual susceptibility to smoking-related cancers. One of the most widely studied enzymes is GSTM1.³ GSTM1 is a phase II enzyme that is present in all human tissues and detoxifies active epoxide carcinogens formed from benzo(a)pyrene and other procarcinogens in tobacco smoke (1). Approximately 50% of Caucasians lack the GSTM1 gene and any corresponding activity. Smokers without GSTM1 have been reported to be at 2–3-fold greater risk for lung cancer than smokers with GSTM1 (2, 3, 4, 5). Because the most common histological type of NPC, squamous cell carcinoma, is related to smoking, we hypothesized that GSTM1 could be involved in determining susceptibility to this cancer as well (6). This hypothesis was tested using data from a multicenter study of NPC in the United States.

The epidemiological methods have been described previously (6). Briefly, male and female cases were prospectively identified through five population-based cancer registries included in the National Cancer Institute’s SEER Program. Individuals who were eligible to be cases were diagnosed between 18 and 74 years of age with one of three histological classifications of NPC: International Classification of Disease-Oncology epithelial NOS (801x–804x), undifferentiated (8020, 8021, and 8082) or nonkeratinizing (8072 and 8073), and squamous (805x–808x, except for 8072 and 8073). Three of the registries participated in the blood collection phase of the study and are included in this report. The eligible diagnosis dates were between April 1, 1987, and June 30, 1991, at the Connecticut and metropolitan Detroit registries and between April 1, 1987, and June 30, 1993, at the western Washington registry.

Controls were identified by random digit dialing, using a modified version of the method of Waksberg (7) and were frequency-matched to the cases by age (in 5-year groups), sex, and geographic site of the cancer registry. From a list of all working exchanges (provided by the local telephone companies), an exchange was randomly selected and four random digits were added, forming a telephone number. This number was then called up to nine times, at varying times of the day and week, to determine whether the number was a residence. For numbers at which a residence was identified, the first five digits plus the area code were used to generate additional phone numbers by adding two random digits. Secondary phone numbers generated in this way were then resolved one by one by calling up to nine times at varying times of the day and week until two additional residences were identified.

Experienced interviewers at each registry conducted structured phone interviews to collect information that included demographics and history of tobacco and alcohol use. All questions referred to the time period before the reference date, which was 1 year prior to diagnosis for the cases and 1 year prior to ascertainment for the controls. At the end of the interview, subjects were asked to give blood, and if they agreed, an appointment was made and blood was drawn. Only subjects with blood specimens and direct interviews were included in this study.

We identified 205 eligible cases. The distribution by SEER registry was: western Washington, 91 cases; Detroit, 60 cases; and Connecticut, 54 cases. We successfully interviewed 175 cases (85.4%) or their next-of-kin (usually the spouse). The most frequent reasons for no interview were subject or physician refusal. To maintain comparability with the controls, we excluded eight cases without a telephone at the reference date from further analyses, leaving 167 to be analyzed. From among these respondents, we were able to draw blood from 85 cases (50.9% response). The most common reason for no blood draw was death of the subject. Specimens from two subjects were of insufficient volume to perform laboratory testing, resulting in a total of 83 cases with complete interview and blood data.

We identified 273 eligible controls through the same three SEER registries and completed interviews with 206 controls (75.5%) or their next-of-kin, with subject refusal being the primary reason for nonparticipation. We were able to collect blood from 144 controls (69.9% response), again with refusal being the most common reason for no blood draw. Two specimens were inadequate for testing, resulting in complete data collection for 142 controls.

Fifty ml of venous blood were drawn into vacutainer tubes containing either acid citrate dextrose or heparin anticoagulants. Blood was processed at a central laboratory, usually within 24 h of being drawn. Lymphocytes and granulocytes were isolated by centrifugation over Ficoll, frozen, and stored in liquid nitrogen. Where samples had low cell numbers, lymphoblastoid cell lines were generated by transformation with EBV. DNA was isolated by incubation in 4 m ammonium acetate and precipitation in isopropyl alcohol (8).

GSTM1 genotypes were identified with a PCR-based method (9), using primers 5′-GAACTCCCTGAAAAGCTAAAGC-3′ and 5′-GTTGGGCTCAAATATACGGTGG-3′. To make certain that a null genotype was due to the absence of GSTM1 alleles rather than a failure in the PCR analysis, we coamplified human β-globin using primers 5′-AACTTCATCCACGTTCACC-3′ and 5′-GAAGAGCCAAGGACAGGTAC-3′. All assayed subjects were positive for β-globin.

A modified hot-start PCR was carried out in an MJ Research Thermal Cycler PTC-100 or a Perkin-Elmer 4800 thermal cycler (Perkin-Elmer, Norwalk, CT). The 50-μl reaction mixture contained: 100 ng of genomic DNA; 200 μm each dGTP, dATP, dTTP, and dCTP (Boehringer Mannheim, Indianapolis, IN); 300 ng each of forward and reverse primers for GSTM1; 250 ng each of forward and reverse primers for β-globin; 20 mm Tris-HCI (pH 8.4); 50 mm KCl; 4 mm MgC1₂; 0.01% gelatin; and 1.5 units of Taq polymerase (Life Technologies, Inc., Gaithersburg, MD). Cycling conditions consisted of 5 min at 94°C for loading and initial denaturation followed by 40 cycles of 94°C for 10 s, 60°C for 20 s, and 72°C for 45 s and a final extension at 72°C for 5 min. For negative controls, we included in the amplification a reaction mixture that contained all components except the DNA template. The PCR products were resolved on a ethidium bromide-stained 2.0% agarose gel. φX 174/HaeIII DNA digest was used as a DNA size marker. Amplification with β-globin primers produced a 268-bp band, and that with the primers for GSTM1 homozygote and heterozygote produced a 215-bp band.

Unconditional logistic regression models were used to obtain maximum likelihood estimates for ORs and 95% CIs for measurement of the association between case-control status and genotype (10). The same techniques were also used to evaluate and adjust for the effects of potential confounders on our measures of association. Previous findings from this study suggested etiological heterogeneity by histological type of NPC, with squamous cell carcinomas being most strongly associated with cigarette and alcohol use (6). Therefore, we analyzed this histological subgroup separately as well as combined with all carcinomas. There were too few cases of the other histological types to produce stable OR estimates. The following variables were evaluated as potential confounders: age; sex; SEER site; race; level of education completed; history of cigarette use (including current smoking status, pack-years, cigarettes per day, and duration); and history of alcohol use (including duration and drinks per day). Potential modification of the effect of GSTM1 genotype on NPC risk was assessed for the above factors by the addition of interaction terms in the logistic model and by separate analyses of subgroups of subjects determined by these factors.

Disease and demographic characteristics of participating and nonparticipating cases and participating controls are compared in Table 1. The distributions of histological types among cases who participated and those who did not participate in the interview or blood draw were similar. Participating cases tended to be at a slightly earlier stage of disease, were younger, and had a slightly different distribution of minority ethnic subjects than did cases who did not participate. Participating cases and controls were similar with respect to gender, but participating cases were younger and more racially diverse and had fewer years of completed education compared to controls.

ORs for NPC associated with GSTM1 genotypes are presented in Table 2 and were adjusted for age, sex, and history of cigarette use (using a variable composed of current cigarette smoking status and pack-years of use). Overall, persons with the GSTM1 homozygous null genotype were at a 90% increase in NPC risk (95% CI = 1.0–3.3) compared to persons with one or two copies of GSTM1 genes. Females who were homozygous for the null allele were at higher risk than males, although this difference could be attributed to chance (test for interaction, P = 0.18). Similarly, higher risk was observed among those 50 years of age or older, current and former cigarette smokers, and persons drinking two or more alcoholic drinks per day. Using interaction terms in our models, we found that none of these differences were statistically significant.

When cases were restricted to those with squamous cell carcinoma, the general pattern of results was similar, although the OR associated with GSTM1 null genotype was lower among current and former smokers, and the OR among those drinking two or more alcoholic drinks per day was higher (test for interaction by alcohol intake, P = 0.06).

We examined in more detail the patterns of NPC risk associated with GSTM1 null genotype by various measures of cigarette use, including duration, intensity, and pack-years of smoking. However, we found no clear and consistent evidence of increasing risk by increasing exposure to cigarette smoke either among all histologies or among squamous cell cases. On the other hand, the higher risk we observed among heavier drinkers was also evident among those who had regularly drunk alcohol for a long duration; the ORs were 1.3 (95% CI = 0.6–2.7) and 3.1 (95% CI = 1.2–8.3) for those who had regularly drunk alcohol for ≤25 years and >25 years, respectively (tests for interaction not significant). Increased risk among heavier drinkers was consistently noted when we used different cutoff points of drinks per day.

Analyses were also carried out with race controlled for after excluding the 17 Asian cases. Our overall results remained the same, and in most cases, stratum-specific point estimates were little changed. For example, the adjusted ORs among all cases and squamous cell carcinomas were 2.1 (95% CI = 1.1–1.4) and 2.0 (95% CI = 0.9–4.2), respectively. When analyses were limited to Caucasians (i.e., also excluding African-Americans and Native Americans), the OR for all cases was 2.0 (95% CI = 1.0–3.9).

This is the first reported study of the GSTM1 polymorphism and NPC. Our results suggest that the GSTM1 null genotype is associated with an almost 2-fold increase in risk for NPC. This finding is consistent with numerous studies that have reported a relationship between GSTM1 genotype and risk for smoking related cancer at another site, the lung.

We found no clear modification of the GSTM1 and NPC relationship by history of smoking. Some previous studies have reported that the association of GSTM1 with lung cancer risk is stronger in smokers than in nonsmokers (11) or in heavier smokers than in light smokers (2, 5), whereas other studies have found no modification by smoking status (4). Some studies of lung cancer have found the strongest role for GSTM1 in the squamous cell histological type (4, 12), whereas others have not (3, 5, 13). Our results did not suggest that GSTM1 played the strongest role in squamous cell carcinoma relative to other types, although there were too few cases with other histologies to allow a definitive conclusion.

In our data, GSTM1 was a stronger predictor of risk in older subjects. This finding is consistent with our previously reported finding that, although cigarette smoking is an important risk factor for NPC in persons 50 years of age or older, it is not important for persons younger than 50 years, suggesting a different etiology among younger cases (6).

Previous studies of lung cancer suggest that gender can modify the risk associated with GSTM1. For example, a recent study reported that GSTM1 null Japanese women were at 7.2-fold risk for squamous cell lung cancer, whereas GSTM1 null Japanese men did not have an elevated risk (OR = 1.1; Ref. 5). We also found suggestive evidence that sex modifies the relationship between GSTM1 and risk for NPC. There are reports of sex differences in the metabolism of various drugs (14, 15, 16, 17), including alcohol (14, 18), and it is possible that tobacco smoke is also metabolized differently in men and women. The sex differences reported here could possibly result from differences in expression or regulation of GSTM1 (15, 16) or Phase I enzymes (16, 19) involved in the metabolism of tobacco and alcohol carcinogens.

Previously reported data from this study showed a strong relationship between use of alcohol and risk for NPC (6). We are aware of no evidence that ethanol is a substrate for GSTM1. However, it has been shown that ethanol increases the rate of gene transcription of at least one of the cytochrome P450s involved in tobacco metabolism, CYP2E1 (1), and CYP2E1 has been reported in some studies to be related to NPC (20) and lung cancer risk (21, 22, 23). We hypothesize that induction of CYP2E1 might lead to increased levels of activated tobacco carcinogens and a larger role for GSTM1 in detoxifying them. Our results are consistent with that hypothesis because we found that the relationship between the GSTM1 null genotype and NPC risk was stronger among subjects with more years of alcohol use and those with greater intensity of use after adjustment for cigarette history. We consider these alcohol results to be exploratory in nature and deserving of further study.

Some limitations should be considered. Although our participation rates are low, it is unlikely that refusals among cases or controls would be related to GSTM1 genotype. On the other hand, most of the nonparticipating cases could not be included because they had died between the time of the phone interview and the blood draw. In one study, GSTM1 null genotype was related to increased survival after treatment among breast cancer patients (24), possibly because patients with GSTM1 were able to detoxify chemotherapeutic agents and thereby confer resistance to chemotherapy, whereas GSTM1 null patients were not. If this were occurring among those NPC patients treated with chemotherapy, it would bias the observed association away from the null. This is speculative, however, and we do not consider this to be an important limitation. The racial distributions of participating and nonparticipating cases were somewhat different, with participating cases being more racially diverse than controls. However, when we limited our analyses to Caucasians (the only group large enough to analyze separately) our relative risk estimates were not changed substantially. Further, our study size was small, and some of the results we present could be attributed to chance.

In summary, there is growing evidence that GSTM1 is important in several tobacco-related cancers, including those of the lung, bladder (25, 26), and head and neck (27). Our results indicate that persons who are homozygous for the GSTM1 null allele may be at moderately increased risk for NPC as well.