Differential Interactions between GSTM1 and NAT2 Genotypes on Aromatic DNA Adduct Level and HPRT Mutant Frequency in Lung Cancer Patients and Population Controls¹

Most common cancers result from the interaction between genetic and environmental factors (1). Because lung cancer risk is clearly related to tobacco exposure, genetic susceptibility to lung cancer has been investigated in terms of interindividual variations in the ability to activate or detoxify potential carcinogens such as PAHs³and aromatic amines present in cigarette smoke. Among enzymes with common polymorphisms are GSTM1, which catalyzes conjugation of diol-epoxide derivatives of PAHs, and NAT2, which generally detoxifies aromatic amine procarcinogens by N-acetylation (2). Approximately 50% of most Caucasian populations lack GSTM1 activity or are NAT2-slow acetylators.

The GSTM1 deficiency is attributable to a homozygous deletion of the GSTM1 gene (null genotype) and has been associated with increased risk of lung cancer, colorectal cancer, and urothelial cancer (2). The NAT2-slow phenotype is attributable to homo- or heterozygosity of three major slow alleles (3), and has been associated with higher risk for bladder cancer (2). It may also be a risk factor for breast cancer among smoking postmenopausal women (4). Regarding lung cancer risk, recent studies have shown contradictory results. An increased risk for lung cancer has been associated with the NAT2 rapid genotype in a German study (5), but with the NAT2 slow genotype in a Japanese study (6).

The impact of genotypes on lung cancer risk has been supported by their influences on exposure biomarkers of tobacco smoke. A several-times higher urinary mutagenicity has been reported in GSTM1-null smokers as compared with GSTM1-positive smokers (7). Individuals with the GSTM1-null genotype also have a greater likelihood of having detectable PAH-DNA ALs in lung (8). Furthermore, benzo(a)pyrene diol-epoxide DNA adducts were detectable in currently smoking lung cancer patients only if they had the GSTM1 null genotype (9).

Several studies have reported evidence that suggest GSTM1and NAT2 as concurrent modifiers of genetic susceptibility to DNA damage. We have previously found that among Swedish garage workers with the NAT2 slow genotype, the aromatic DNA AL was significantly higher in those with the GSTM1 null genotype compared with those with GSTM1 present (10). In a study of coke oven workers (11), urine mutagenicity was found to occur at a higher frequency among smokers with the GSTM1-null/NAT2-slow combination than among smokers with other genotypes. Similarly, bus drivers with the null/slow combined genotype showed a significantly increased frequency of lymphocytes with chromosomal aberrations as compared with those with the positive/rapid genotype (12). In addition, a significantly increased risk of developing malignant and nonmalignant pulmonary disorders was reported among asbestos-exposed subjects with the null/slow genotype (13, 14).

Recently, in a carefully designed study of Swedish nonsmoking and smoking lung cancer patients and matched population controls, we reported a possible gene-gene interaction (15), with the NAT2 slow and GSTM1 positive genotype conferring particularly high risk among never smokers. Among smokers, rapid acetylators tended to show a steeper PY-related increase in risk compared with slow acetylators. We then studied the aromatic DNA AL and the MF in the HPRT reporter gene in peripheral lymphocytes of both cases and controls (16). We found no difference between cases and controls and no effect of ETS among nonsmokers but a significant effect of smoking on the two biological end points in surrogate tissue. In addition, the increase of AL and MF with age and smoking dose was stronger in patients than in controls, which suggested an interaction between smoking and genetic host factors. In the present study, we have studied the influence of GSTM1 and NAT2 genotypes on the two biological end points in smokers,with special emphases on gene-gene and gene-environment interactions.

The current study population originated from a larger epidemiological study designed to investigate the effect of passive smoking on lung cancer risk (17). Cases were recruited during 1992–1995 at the three major county hospitals in Stockholm County responsible for diagnosis and treatment of lung cancer. Patients were asked to participate in the study when they had received a lung cancer diagnosis. Each never-smoking case was used as an index case for next diagnosed ever-smoking case of the same gender and age (30–49, 50–69,and ≥70 years) in the same hospital. Healthy population controls were extracted from the Stockholm residence files every 6 months and frequency matched to cases in regard to hospital catchment area,gender, and age group, as well as broad smoking categories:“smoker” (current or recent-quit within 2 years), former smoker and never smoker. Detailed exposure data on smoking, ETS (from spouse,work, or other places), and dietary habits, as well as residential and working histories, were collected mainly by personal interview according to a standard questionnaire.

A total of 185 cases and 164 controls supplied blood for genotyping (15) and measurement of aromatic DNA ALs and HPRT MFs (16). More than 70% of the subjects were women. The age distribution was very similar in the groups of patients (median, 69; range, 30–92 years) and controls (median, 68;range, 30–89 years) because of matching. The patients had, however, a significantly higher dose, duration, or PYs of smoking and longer passive smoke exposures than the controls (16). The majority of cases had adenocarcinoma (51%) or squamous cell carcinoma(22%). Ever smokers had a higher proportion of squamous cell carcinoma than never smokers (37% versus 7%), whereas adenocarcinomas were more frequent among never smokers (62%) than among ever smokers (40%).

All except 10 blood samples were taken before radio- or chemotherapy. All of the cases were, however, considered as eligible as the treatment was assumed not to affect the aromatic DNA AL nor the HPRTMF because the time between initiation of treatment and blood sampling was too short for expression of any induced mutants. Lymphocytes and granulocytes were isolated after density separation in Polymorphprep(Pharmacia, Sweden). The former were frozen (−135°C) in aliquots for subsequent mutant selection and adduct measurement, and the latter were freshly used for DNA isolation (using saturated NaCl) and genotyping.

The principles and details of the methods for GSTM1 and NAT2 genotyping have been described previously (10, 15). The presence or absence of the GSTM1 gene was detected by genomic PCR amplification of a short internal GSTM1 gene segment (177 bp) together with a NAT2segment (284 bp) as an internal PCR control. Individuals with one or two copies of the GSTM1 allele were designated GSTM1 positive. Individuals with homozygous deletion of the GSTM1 allele were designated GSTM1 null.

Identification of the slow NAT2 alleles was performed by restriction analysis of a 578-bp genomic PCR product covering a large part of the intronless NAT2 coding region. The predominating NAT2*5A/B (341C, 481T) and NAT2*6A/B (590A)alleles (18) were identified by loss of a restriction site for KpnI and TaqI, respectively. Individuals in whom no, or only one, slow allele could be identified by these enzymes were further analyzed by BamHI and DdeI digestions for the identification of the NAT2*7A/B (857A)and NAT2*5C (341C, 803G) alleles, respectively. Individuals with at least one wild-type allele were classified as rapid, and those with two slow alleles as slow.

The ³²P TLC assay of aromatic DNA adducts was carried out as described previously (19). In brief, DNA was extracted from the crude nuclei, using organic solvents after degrading RNAs and proteins, and was digested by micrococcal nuclease and spleen phosphodiesterase to 3′ nucleotides. Adducts were then enriched by nuclease P1 treatment. A postlabeling reaction was carried out and applied on a TLC plate for adduct separation in three dimensions. After autoradiography, the adduct spots were excised from the TLC plate for the counting of radioactivity. Two to five assays were carried out for each sample.

Four matched lymphocyte samples (one from each of the four study groups) were analyzed concurrently to minimize the influence of possible methodological variation over time. The culture media used and the T-cell cloning procedure have been described in detail previously (20). Briefly, all of the cells were stimulated for 44 h with 0.3% phytohemagglutinin (PHA, Difco) in RPMI 1640-based medium containing 5% FCS and 5% human serum, and then were subjected to cloning in a T-cell growth factor-enriched medium with or without 6-thioguanine.

The MF was obtained by dividing the cloning efficiency in the presence of 6-thioguanine by that in the absence of 6-thioguanine. A 95%CI was calculated for each MF from the variance of lnMF. Seven MF values with large CIs (>5 × MF, related to low cloning efficiency or too few positive or seeded wells in selection plates)were excluded from statistical analysis (16).

The Wilcoxon rank-sum test was used to test differences in AL or MF between groups. The distribution of AL or MF was normalized by ln-transformations, and the influence of various factors on lnAL or lnMF was studied by multiple linear regression. Factors studied include age, gender, case status, smoking category, average or last daily cigarette dose and PYs of smoking. Strongly correlated variables,such as age and PY, were not used in the same model to assure the independence of explanatory variables. The relationship between lnAL and lnMF was quantified using Pearson’s correlation coefficient.

To allow calculation of ORs with 95% CIs, with adjustment for potential confounding factors, multiple logistic regression was carried out after assigning each AL (or MF) value high or low using the overall median level as cutoff point. Joint effects between two genotypes were studied by creating dummy variables, each representing the combination of two genotypes, with the putative low-risk combination as reference category. Departures from a multiplicative interaction model (as implicitly assumed by logistic regression) were evaluated by adding interaction terms to the main effect model. To plot interaction, a lnOR value was predicted for each individual using theβ-coefficients for the two main effects and the interaction term.

Table 1 shows the ALs measured in a total of 171 cases and 146 controls. Smokers (current or recent) showed significantly higher ALs than nonsmokers (former or never). This difference was most clear among controls (median, 4.1 versus 3.4 per 10⁸ nucleotides; Wilcoxon P =0.005) although the cigarette dose in smoking controls was significantly lower than in smoking cases (median, 11.6 versus 14.2 cigarettes/day; Wilcoxon P =0.02). There was no apparent genotype effect in any of the subgroups,but controls with the NAT2 slow genotype showed consistently higher AL than controls with the rapid genotype.

Table 2 shows the distribution of high/low (above/below median) AL in relation to different combinations of GSTM1 and NAT2genotypes within each subgroup. The ORs were calculated using the GSTM1 positive and NAT2 rapid combination (wild type) as reference category, adjusted for age (considered as a susceptibility factor for carcinogen exposure because of reduced DNA repair efficiency; see Ref. 16), and gender, and, for smokers, also cigarette dose. Pooled analysis using either the GSTM1 positive or the NAT2 rapid genotype as reference categories are also shown. Neither the GSTM1 null nor the NAT2 slow genotype appeared to be a predictor of high AL among cases; rather the tendency was the opposite. Smoking cases showed even an antagonistic interaction between the null and slow genotypes. These genotypes exhibited, however, the a prioriexpected effects among controls, especially among smoking controls. The NAT2 slow genotype appeared to be a particularly strong predictor of high AL among currently smoking controls (20/9 versus 4/9 above/below the median of 4.0; OR, 7.5; 95% CI,1.5–38.4). An OR increased 19-fold was seen among currently smoking controls with the null/slow combination compared with those with normal genotypes (10/3 versus 2/4; OR, 19.3; 95% CI, 1.1–338.6). Although current smokers may be most informative in regard to the effect of genotype, the numbers of subjects were however greatly reduced, and the results should be interpreted with caution. The autoradiograms were also reviewed for differences in adduct patterns by genotype. No differences were found, but subtle changes could not be ruled out because there was no possibility to quantitatively assay small specific faction on thin-layer sheets.

On the basis of the previous finding of an age-dependent increase of AL in currently smoking cases (16), we next investigated whether this could be modified by genotypes. A clear influence of both age and daily cigarette dose on lnAL was seen among cases with the GSTM1 null or NAT2 slow genotype but not among cases with the GSTM1 positive or NAT2 rapid genotype (Table 3). PY of smoking affected AL only among cases with the combined GSTM1 null and NAT2 rapid genotype (3.0%increase per PY; P = 0.07, adjusted for gender). Stronger effects were, however, shown by age (6.7% per year; 95% CI,3.7–9.7%) and daily dose (4.7% per cigarette/day; 95% CI,1.3–8.1%), which together with gender (β = −0.56; P = 0.04) explained 84% (adjusted R²) of the variation in AL among the GSTM1 null and NAT2 rapid cases. There were no significant effects of age and smoking dose on AL among currently smoking controls with different genotypes (data not shown).

Because a positive correlation between AL and MF was previously found in high-dose, currently smoking cases only (16), we finally tested whether such correlation could be detected in smoking cases with susceptible genotype attributable to high biologically active or effective dose. An almost significant correlation between lnAL and lnMF was seen in currently smoking cases with the GSTM1 null genotype (r, 0.54; P =0.07; Fig. 1). Most of this correlation was attributable to cases with the GSTM1 null and NAT2 rapid combination(r, 0.64; P = 0.1). There was no correlation between the two biomarkers in currently smoking cases with the GSTM1 positive genotype (P = 0.5), NAT2 rapid (P = 0.9), or NAT2slow (P = 0.6) genotype.

MF data from 158 cases and 154 controls are shown in Table 4. Ever smokers had significantly higher MFs than never smokers among both controls and cases, especially among NAT2-slow cases(median, 18.6 versus 13.6 per 10⁶cells). In regard to genotype effect, only controls with the NAT2 slow genotype showed a significantly higher MF compared with controls with the NAT2 rapid genotype (ever smokers and never smokers pooled, Wilcoxon P = 0.035).

Table 5 shows the genotype-related distribution of high/low (above/below median) MF within each subgroup, with ORs calculated using the GSTM1 positive or/and NAT2 rapid genotype as reference category (adjusted for age, gender, and, for smokers, also average cigarette dose). Although never-smoking cases showed no effect of the GSTM1 null genotype and a possibly protective effect of the NAT2 slow genotype, ever-smoking cases showed a positive effect of the GSTM1 null genotype, which was,however, attributable to the NAT2 rapid cases only (OR, 4.0;95% CI, 0.9–17.7). Among controls, both the individual genotype effects and the null-slow interaction were positive, irrespective of smoking status. Overall, cases differed from controls in GSTM1-NAT2 interaction among never smokers in a similar manner as among ever smokers. Adjusted for smoking category (never,former, recent, current), age, and gender, the interaction between the GSTM1 null and NAT2 slow genotype was less than multiplicative among cases (OR, 0.2; 95% CI, 0.05–0.85 for the interaction term) but was more than multiplicative among controls (OR,4.6; 95% CI, 1.1–18.8).

Because cases had much higher PYs of smoking than controls (median,29.4 versus 19.3 PYs), we next stratified all of the ever smokers according to the overall median PY of 23 (cases and controls pooled; Table 6). A similar difference in interaction pattern was seen between higher-and lower-PY smokers as between cases and controls. Whereas no clear genotype effect was seen among higher-PY smokers, the GSTM1null genotype appeared to be associated with high MF among lower-PY smokers (OR, 3.3; 95% CI, 1.1–9.7), in particular among those with the NAT2 slow genotype. Light smokers with both genes defective showed a 5-fold (95% CI, 1.2–22.4) increased risk for high MF compared with those with both genes normal. Compared with those with at least one gene normal, the risk was increased by 4-fold (15/7 versus 22/35; OR, 3.7; 95% CI, 1.3–10.7).

On the basis of the above findings, we finally tested whether the GSTM1 null or/and NAT2 slow genotype interacted with PYs of smoking (as continuous variable). Only cases showed clear interactions, between PY and the GSTM1 null genotype(β = −0.15; P = 0.04), the NAT2 slow genotype (β = −0.085; P = 0.006) or the combined null/slow genotype (β = −0.15; P =0.04). The main effect of genotype was positive in all of the models,but PYs was a clear predictor of high MF in the NAT2 model only (OR, 12.3; 95% CI, 1.6–96.6 for the NAT2 slow genotype; and OR, 1.0;, 95% CI, 1.00–1.09 for each PY increase). The overall effect of NAT2 genotype and PY on lnOR for high MF(adjusted for gender) is illustrated in Fig. 2. The slow genotype was associated with high MF among cases diagnosed at lower PY, whereas the rapid genotype was associated with high MF among cases diagnosed at higher PY.

To our knowledge, the present study is the first case-control study that combines genotype analysis with both adduct and mutation measurements in normal somatic cells. We have previously (16) found age, rather than PY, to be a significant predictor of high AL among currently smoking cases, which suggests an age- and case-related increase in susceptibility to carcinogen exposure. This could indeed be partly explained by the involvement of genotypes in determining cancer susceptibility. The GSTM1null and NAT2 rapid genotypes appeared to be associated with higher cancer risks among high-PY current smokers (15). If these genotypes constitute the susceptible genotypes at high dose level, then the smoking case group would also exhibit a high AL because of the overrepresentation of these genotypes relative to controls. The role of these genotypes at high exposure level is further supported by our present findings among smoking cases that had a significantly higher PY level than controls. Firstly, currently smoking cases with the combined GSTM1 null and NAT2 rapid genotype showed a particularly clear increase of AL with both age and daily cigarette dose. Secondly, AL correlated particularly well with MF in cases with the null/rapid genotype. Thirdly, the combined null/rapid genotype was associated with high MF among ever-smoking cases. These results suggest that lung cancer patients with the GSTM1null and NAT2 rapid genotype may represent a group that is particularly susceptible to DNA damage from PAH-type mutagens in tobacco smoke.

Also the antagonistic interaction between the GSTM1 null and NAT2 slow genotypes on MF among cases, as opposed to the synergistic interaction among controls, may be related to the high PY of smoking in the case group. This is supported by the similar difference in interaction pattern among smokers stratified by PY level instead of case status. Indeed, all but 1 of the 11 ever-smoking cases with the GSTM1 null/NAT2 rapid combination and high MF had PY above 23. Thus, the NAT2 rapid genotype may be associated with high MF among high-PY smokers with the GSTM1 null genotype. The NAT2 slow genotype may,however, be associated with high MF at lower PY in combination with the GSTM1 null genotype.

The impact of smoking dose on the influence of NAT2 genotype on MF was further demonstrated by the significant interaction of NAT2 genotype with PY of smoking on the odds for high MF among ever-smoking cases. Whereas the NAT2 rapid genotype appeared to be associated with high MF at higher PY, the slow genotype was associated with high MF at lower PY. A similar interaction was observed between the slow genotype and age on lnMF in currently smoking cases (data not shown). Among cases diagnosed at younger age, slow acetylators tended to show higher MF than rapid acetylators, whereas the known positive association with age (16) was evident only for rapid acetylators. The rapid genotype appears, thus, to be associated with an accumulation of mutants as age and cumulative dose reaches higher levels. The rapid genotype was also previously shown to confer greater lung cancer risk at these higher PY levels than did the slow genotype (15). Homozygous rapid NAT2genotype has been associated with an increased lung cancer risk in a German study (5). Notable is that the 17 cases (of 155)with the homozygous rapid genotype were relatively old but did not have a exceedingly high cigarette consumption. In a recent study of Norwegian nonoperable lung cancer patients (21), a clear overrepresentation of NAT2 slow genotypes was seen among those who were younger (≤63 years), as well as among younger light smokers (PY ≤ 30). Consistent with above findings, the NAT2 slow genotype was more markedly associated with an increased risk for adenocarcinoma in a Japanese study (6),when the analysis was confined to patients under the age of 65. The incidence of p53 mutations among the younger adenocarcinoma patients who had the slow genotype was also higher than that among patients with the rapid genotype. These results suggest that individuals with the NAT2 slow genotype may represent a group that is more prone to acquire gene mutations and lung cancer at younger age or lower PY of smoking. Conversely, with the rapid genotype, high MF and lung cancer risk results only with high age or cumulative smoking dose because most somatic mutations are known to be persistent over time (16).

Considering the detoxification reactions catalyzed by NAT2and the method used for adduct measurement, the observed association between the NAT2-slow genotype and high AL in controls is perhaps surprisingly strong. The enrichment of DNA adducts in the present study was carried out by digestion with nuclease P1, which may have caused arylamines bound to the C8 position of guanine to be lost more extensively than if the butanol extraction method has been used (22). However, extensive or complete recovery should be obtained from aromatic amines and PAHs bound to the exocyclic positions of guanine or adenine in the present work. The bottom line is that the identity of adducts detected by the present postlabeling method is unknown, as is the proportion of PAH-DNA adducts (19, 22).

Further supporting our AL data are the low-dose dependency of the NAT2 effect and its interaction with the GSTM1null genotype, which is in agreement with what we found on MF. The NAT2 effect is also consistent with the study of Vineis et al. (23), who found a higher level of 4-aminobiphenyl-hemoglobin adducts in slow acetylators compared with rapid acetylators, but only at low or zero nicotine-cotinine levels. Our data on genotype-related AL and MF in nonsmokers do not support the previous finding of an increased lung cancer risk among never smokers with the GSTM1 positive and NAT2 slow genotype (15). Rather, a positive interaction between the GSTM1 null and the NAT2 slow genotype on MF was seen among never-smoking controls, similar to that seen among ever-smoking controls. This is consistent with the fact that the vast majority of never smokers (if not all) were ever exposed to ETS, and controls had an overall much lower cumulative smoking and ETS dose compared with cases. These results suggest that the NAT2slow genotype, in particular when combined with the GSTM1null genotype, may confer increased susceptibility to tobacco mutagens when exposure is low. It is plausible that at higher exposures, the effect of NAT2 genotype regulating carcinogen detoxification may be overwhelmed, and other metabolic pathways or mechanisms such as DNA repair activity or efficiency may play a more important role for individual susceptibility.

Our present study confirms the important role of gene-gene and gene-environment interaction in the etiology of common cancer such as lung cancer (1). Additional studies in regard to the effect of dietary factors and other genetic polymorphisms, especially those involved in DNA repair, will be necessary to further improve understanding of individual susceptibility to DNA damage, gene mutation, and the development of lung cancer.