Influence of CYP1A1, GSTM1, GSTT1, and NQO1 Genotypes and Cumulative Smoking Dose on Lung Cancer Risk in a Swedish Population

Lung cancer is the leading cause of cancer death worldwide. Although tobacco smoking is causally implicated in the overwhelming majority of lung cancers, <20% of smokers develop this disease (1). Individual susceptibility to lung cancer may partly be explained by differences in metabolic efficiency in the formation and elimination of carcinogens in tobacco smoke. Functional polymorphisms in genes coding for enzymes involved in this process have been extensively studied in relation to lung cancer. An example of such a gene is the CYP1A1 gene, the enzyme product of which mediates the activation of tobacco carcinogens, such as polycyclic aromatic hydrocarbons, into reactive intermediates capable of damaging DNA. Two closely linked polymorphisms in CYP1A1, the 3801T > C substitution creating a MspI restriction site in the 3′ noncoding region and the 2455A > G substitution resulting in an amino acid change (Ile⁴⁶²Val), were reported in 1990 to be associated with lung cancer in a Japanese population (2, 3). The GSTM1 gene encodes an enzyme involved in detoxification of various reactive intermediates, including those of polycyclic aromatic hydrocarbons. In a study published in 1986, the homozygously deleted GSTM1 gene was indicated as a risk factor for lung cancer (4). Another enzyme belonging to this family, GSTT1, is involved in the metabolism of other tobacco smoke constituents, such as 1,3-butadiene, ethylene oxide, and halogenated alkanes. Consequently, the GSTT1 deletion polymorphism has also been extensively investigated in relation to lung cancer (5-11). The NQO1 gene encodes an enzyme capable of reducing a wide variety of substrates, including quinones derived from polycyclic aromatic hydrocarbons. The functional 609C > T polymorphism causing an amino acid change (Pro¹⁸⁷Ser), resulting in loss of NQO1 activity (12), has also been studied for its role in lung cancer (11, 13-19).

Although many studies have assessed CYP1A1, GSTM1, GSTT1, and NQO1 polymorphisms in relation to lung cancer, the results are conflicting. Besides different study designs, prevalence of genetic polymorphisms and linkage disequilibrium in different ethnic populations are possible explanations for the varying results obtained. Effect modifications by environmental or other genetic risk factors that differ between study populations are alternative causes. To avoid such influences, large studies on homogenous populations are warranted.

Our previous study on the effects of genetic polymorphisms in CYP1A1 and GSTM1 on lung cancer risk included 329 healthy subjects and 296 lung cancer cases of Swedish origin (20). The number of study subjects has been increased and the results presented here, on 530 control subjects and 524 lung cancer cases, also include analyses of the GSTT1 and NQO1 polymorphisms. We evaluated the influence of these genetic polymorphisms on lung cancer risk overall and performed analyses after stratification by histologic subtypes and smoking variables.

Blood samples were obtained from 524 lung cancer patients and 530 control subjects between 1991 and 2000. Part of the study population was included in a previous study of the influence of genetic polymorphisms on lung cancer risk (20). All subjects were of Swedish origin and lived in the southern half of Sweden. The cases were recruited from eight different hospitals. At two of the hospitals, the majority of the patients completed a questionnaire about smoking habits. Information on smoking history and histologic subtypes was available to a lesser extent for the patients from the other six hospitals. The control subjects included laboratory workers, welders, and chimney sweeps from the former study. The group of laboratory staff was extended in the present study. Smoking history from the original and newly recruited laboratory workers was obtained by sending out a brief questionnaire. Smoking information from the sweeps and welders had previously been gathered via interview. In addition, blood samples and smoking history by interview were collected from elderly subjects visiting one of the hospitals for health checkups. The study was approved by the Ethics Committee of Karolinska Institutet (Stockholm, Sweden).

DNA was isolated from whole blood using phenol-chloroform extraction or by a modified salting out procedure (21, 22). The genotypes of the CYP1A1 3801T > C (MspI) and Ile⁴⁶²Val polymorphisms were determined using the MspI RFLP-PCR method and an allele-specific PCR method, respectively, essentially as described by Hayashi et al. (2) and Carstensen et al. (23). The neighboring Thr⁴⁶¹Asp polymorphism does not interfere with the allele-specific method, as the primers used do not cover this polymorphic region. The predominant allele in Caucasians, lacking both of the investigated mutations, has been designated CYP1A1*1, while the allele with only the MspI RFLP has been designated CYP1A1*2A. The CYP1A1*2B allele carries both the MspI RFLP and the Ile⁴⁶²Val polymorphism, whereas, in Caucasians, the extremely rare allele CYP1A1*2C only has the Ile⁴⁶²Val polymorphism. Genotyping for the GSTM1 and GSTT1 null polymorphisms were performed using slightly modified published methods (24, 25), and detailed information of the PCR assays for GSTM1 and GSTT1 are given elsewhere (23, 26). The standard nomenclature for the functional GST alleles, GSTM1*1 and GSTT1*1, and the GST null alleles, GSTM1*O and GSTT1*O, was used. The NQO1 Pro¹⁸⁷Ser polymorphism was determined using the HinfI PCR-RFLP method of Traver et al. (12). The common allele among Caucasians and the variant allele coding for the amino acid exchange Pro¹⁸⁷Ser have been designated NQO1*1 and NQO1*2, respectively.

Differences in distribution of age and cumulative smoking dose (pack-years) between cases and controls were tested by Student's t test or for nonnormally distributed variables by Wilcoxon rank sum test. Pearson χ² analysis was used to test the differences between cases and controls in the distributions of gender, smoking status, and genotypes. Population Hardy-Weinberg equilibrium was tested in controls using the χ² test. Odds ratios (ORs) with 95% confidence intervals (95% CI) were estimated by unconditional logistic regression adjusting for age (continuous variable) and gender. In some analyses, the cases and controls were stratified by cumulative smoking dose. Never smokers were defined as individuals who had never smoked or who had smoked 0.2 pack-years or less. Individuals at or below the median pack-years of cases and controls (median pack-years 21) were considered to be light smokers and individuals above median pack-years were considered heavy smokers. Statistical analyses were performed using the JMP Statistics software package (SAS Institute Inc., Cary, North Carolina).

The distributions of age, gender, and smoking variables of the subjects are shown in Table 1. Seventy-nine percent of the cases were diagnosed with one of the three main histologic subtypes (Table 1). The proportions of females and never smokers were higher among those with adenocarcinoma (47% and 17%, respectively) than among those with squamous (22% and 5%) and small cell carcinoma (35% and 1.5%). Data on age and gender were available for all subjects, whereas the information on smoking habits was incomplete in both cases and controls (Table 1).

The distributions of the CYP1A1, GSTM1, GSTT1, and NQO1 genotypes in controls (Table 2) below the median age did not differ significantly from those found in controls above median age. Furthermore, the genotype distributions did not differ significantly between women and men and between different control subpopulations. Overall, the estimated population frequency of the dominant CYP1A1*1 allele was 0.91, whereas the frequencies for the variant alleles CYP1A1*2A, CYP1A1*2B, and CYP1A1*2C were 0.06, 0.03, and >0.001, respectively. The observed genotype frequencies were 82.4% CYP1A1*1/*1, 10.0% CYP1A1*1/*2A, 0.6% CYP1A1*2A/*2A, 6.6% CYP1A1*1/*2B, 0.2% CYP1A1*2A/*2B, and 0.2% CYP1A1*1/*2C. These genotype frequencies matched the prediction by the Hardy-Weinberg theorem based on the allele frequencies (χ² = 2.46, P = 0.86). Deviations from Hardy-Weinberg equilibrium were not tested for the distributions of GSTM1 and GSTT1 genotypes because the PCR assays did not enable discrimination of heterozygous from homozygous carriers. The overall frequencies of the NQO1*1 and NQO1*2 alleles were 0.84 and 0.16, respectively. The genotype distribution of the NQO1 polymorphism was consistent with the Hardy-Weinberg equilibrium (χ² = 2.37, P = 0.14).

The genotype distributions of CYP1A1, GSTM1, GSTT1, and NQO1 showed no significant differences between controls and all lung cancer cases or cases categorized by histologic subtypes (Table 2).

In Table 3, 312 cases and 273 controls were included in the analysis. Due to fewer subjects, variant CYP1A1 genotypes were combined. Likewise, variant NQO1 genotypes were combined into one group.

In general, the effects of genotypes suggested in the total material (Table 2) were more pronounced in the ever smokers (Table 3). Stratification by histologic subtypes indicated an elevated OR for squamous cell carcinoma associated with the combined variant CYP1A1 genotypes. The OR (95% CI) for small cell carcinoma associated with GSTM1*O/*O was significant with a value of 2.72 (1.32-5.90; P = 0.008). However, if adjustment for pack-years (continuous variable) was performed, the significance was lost (OR 1.97, 95% CI 0.72-5.80), but because of missing detailed smoking data, this analysis was based on only 31 of 65 ever smokers with small cell carcinoma. A protective role of the GSTT1*O/*O genotype that seemed to be most pronounced for small cell carcinoma and adenocarcinoma was observed, but the point estimate was below unity also for squamous cell carcinoma (Table 3). The frequency of the combined variant NQO1 genotypes in patients with squamous cell carcinoma was significantly higher than in controls (P = 0.04). The age-adjusted and gender-adjusted OR (95% CI) for squamous cell carcinoma associated with these genotypes was 1.78 (0.95-3.34) and was significant when adjustment for pack-years was also included (OR 2.21, 95% CI 1.03-4.84; analysis based on 81 of the 115 ever smokers with squamous cell carcinoma).

The study subjects were divided into three groups: never smokers, light smokers, and heavy smokers as described in Materials and Methods. The total number of cases and controls included was 227 and 504, respectively (Table 4). In never smokers, no differences in the distributions of the CYP1A1, GSTM1, GSTT1, and NQO1 genotypes between cases and controls were observed. However, in light smokers, the distributions of the CYP1A1, GSTM1, and GSTT1 genotypes in cases appeared to be different from that in controls (Table 4). The frequency of the variant CYP1A1 genotypes was higher in cases than in controls, but the difference was not statistically significant. In addition, the frequency of the GSTM1*O/*O genotype was higher in cases than in controls (P = 0.05) and the adjusted OR for the GSTM1*O/*O genotype showed borderline significance (2.68, 95% CI 1.00-7.94). We also found an increased frequency of the GSTT1*O/*O genotype in cases compared with controls, although this difference was not statistically significant.

In contrast to the observations in light smokers, the genotypes associated with increased risk (variant CYP1A1, GSTM1*O/*O, and GSTT1*O/*O) appeared to be underrepresented in cases classified as heavy smokers. In heavy smokers, the frequency of the variant CYP1A1 genotypes was low in the cases, but as this also was apparent in the controls, the adjusted OR equaled unity (Table 4). The GSTM1*O/*O genotype frequency in cases was lower than in controls, but the difference was not statistically significantly. A lower frequency of the GSTT1*O/*O genotype in cases compared with controls was also found, and after adjustment, the GSTT1*O/*O genotype was significantly associated with a decreased risk in heavy smokers (OR 0.36, 95% CI 0.13-0.99; P = 0.04).

The NQO1 genotype distribution in cases was similar to that found in controls in never, light, and heavy smokers with adjusted ORs close to unity within each smoking group.

When analyzing the case group only, a difference in the genotype frequencies between light and heavy smokers was observed (Table 4). The frequency of the variant CYP1A1 genotypes in heavy smokers (12.6%) was only half of that found in light smokers (23.3%), but the difference was not statistically significant (P = 0.12). The GSTM1*O/*O genotype was significantly less common in heavy smokers (44.0%) compared with light smokers (70.0%; P = 0.009). The prevalence of the GSTT1*O/*O genotype was also significantly lower in the heavy smokers (10.2%) compared with that found in the light smokers (23.3%; P = 0.04).

When the heavy smokers (pack-years >21) were divided into three equally sized groups, the frequencies of the variant CYP1A1, GSTM1, and GSTT1 genotypes were higher in light smokers compared with those found in each subgroup of heavy smokers (Fig. 1), although each comparison was no longer significant due to the lower numbers in these subgroup comparisons. In contrast, the variant NQO1 genotypes appeared to be similarly distributed in all five groups.

All possible combinations of two genotypes for CYP1A1, NQO1, GSTM1, and GSTT1 were examined. No evidence for effects of gene-gene interactions on lung cancer risk overall was found. However, the suggested increased risk for squamous cell carcinoma in ever smokers with the variant CYP1A1 genotypes was only present in subjects also carrying the GSTM1*O/*O genotype. The age-adjusted and gender-adjusted OR (95% CI) associated with this combined genotype was 1.89 (0.63-5.71). In comparison, the OR (95% CI) for the variant CYP1A1 genotype in combination with the presence of at least one GSTM1*1 allele was 1.13 (0.35-3.50) when using the genotype combination CYP1A1*1/*1 and presence of at least one GSTM1*1 allele as reference group. The adjusted OR for squamous cell carcinoma was substantially higher in the group with combined variant genotypes of CYP1A1 and NQO1 (3.54, 95% CI 0.88-14.3) compared with the ORs in the groups with only one of these variant genotypes (1.36, 95% CI 0.46-3.90 for variant CYP1A1 only and 1.69, 95% CI 0.85-3.39 for variant NQO1 only). The adjusted OR (95% CI) for squamous cell carcinoma associated with the combined genotype CYP1A1*1/*1 and GSTT1*O/*O was 0.33 (0.10-0.95) with the combined genotype CYP1A1*1/*1 and presence of at least one GSTT1*1 allele as reference group. This apparent protective effect of the GSTT1*O/*O genotype was not found in combination with the variant CYP1A1 genotype. The OR (95% CI) for squamous cell carcinoma for this combined genotype was 3.89 (0.80-17.6).

Numerous studies have investigated the association between CYP1A1 and GSTM1 polymorphisms and lung cancer with conflicting results (27). In agreement with our result, an early meta-analysis restricted to non-Asians suggested a small increased lung cancer risk among light smokers associated with CYP1A1 variant genotypes (28). Also in accordance with our study, an elevated risk for squamous cell carcinoma associated with these genotypes was found (28). Pooled analyses of individual data of lung cancer cases and controls from the International Collaborative Study on Genetic Susceptibility to Environmental Carcinogens found a significant association between variant CYP1A1 and lung cancer risk in Caucasians, particularly in never and light smokers (29-31). That the effect of polymorphisms in biotransformation enzymes is stronger at comparably lower exposures has been observed in several epidemiologic studies on cancer risk (32), suggesting that, at high exposures, sufficiently high levels of reactive metabolites are formed in all individuals, regardless of the biotransformation genotypes.

Several meta-analyses of GSTM1 polymorphism and lung cancer risk have suggested a modest increased risk for lung cancer associated with the GSTM1*O/*O genotype (28, 33-35). The association appeared to be more pronounced in Asians than in Caucasians and concerned all three main histologic subtypes. In our study, the increased risk associated with GSTM1*O/*O appeared to be limited to small cell carcinoma. The only meta-analysis addressing the effect of smoking suggested, in an analysis restricted to Asians, that the association between the GSTM1*O/*O genotype and lung cancer risk was stronger in heavy smokers than in light smokers (28). However, some studies, including the present one, have found the opposite association with an increased risk for lung cancer associated with the GSTM1*O/*O genotype in light smokers (36-38). In a pooled analysis from Genetic Susceptibility to Environmental Carcinogens, ever smokers with the GSTM1*O/*O genotype had an increased risk for small cell carcinoma, in agreement with our study, although, in that study, the strongest effect was seen at 20 to 39 pack-years. One explanation for the inconsistent results may be a difference in the definition of light and heavy smokers. For example, in our study, light smokers were defined as those who had smoked equal or below the median pack-years of cases and controls (median pack-years 21). By choosing the median pack-years of the cases only (median pack-years 34), the association observed between the genotypes GSTM1*O/*O, GSTT1*O/*O, and variant CYP1A1 and increased risk in light smokers disappeared. Further, as in most studies, a limitation of our exposure categorization is that smokers included in the light smokers category may not actually have been light smokers for a long time but could have been heavy smokers for a short time with the same cumulative dose.

In agreement with most previous studies, no significant association between GSTT1 polymorphism and lung cancer overall was found. In contrast, in an analysis of smokers only, we found the GSTT1*O/*O genotype to be underrepresented in all three main histologic subtypes. A decreased risk associated with the GSTT1*O/*O genotype has also been observed in other studies (8, 10). This effect by the GSTT1*O/*O genotype was found to be restricted to heavy smokers in our study. Surprisingly, the GSTT1*O/*O genotype was suggested to be a risk factor for lung cancer in light smokers. In agreement with our result, an increased risk for head and neck cancer associated with the GSTT1*O/*O genotype was only evident in light smokers (39). On the other hand, in another Swedish study, an opposite result was observed [i.e., the GSTT1*O/*O genotype was associated with increased risk for lung cancer in heavy smokers (9)]. That study had population-based matched controls, which should contribute to higher study validity, but was much smaller than the present study. In addition, it included only a limited number of heavy smokers, as mainly elderly women were studied. An explanation for the fact that the GSTT1 protein seems to act both as a risk factor and as a factor protecting against lung cancer may be that it is involved in both detoxification and activation of carcinogens (5). It is also known to metabolize protective dietary antioxidants (6, 7).

In studies performed in Asians, the NQO1*1/*1 genotype was found to be associated with lung cancer, particularly adenocarcinoma (11, 13-15), whereas the variant NQO1 allele (NQO1*2) has been suggested to be a risk factor for lung cancer in Caucasians (16-19). The only previous study on Caucasians with sufficient number of cases that performed an analysis of histologic subtypes was in agreement with our study that the variant NQO1 genotypes were overrepresented in squamous cell carcinoma (16).

The gene-gene interactions between CYP1A1 and GSTM1 and between CYP1A1 and NQO1, observed in the present study, support the plausibility of a combined role of these enzymes in lung cancer susceptibility. Thus, the presence of the variant CYP1A1 protein may result in an increased formation of carcinogenic metabolites, and the detoxification of these reactive metabolites are restrained by the absence of functional GSTM1 and NQO1 enzymes.

The present study benefited from consisting of an ethnically homogenous population. This is of importance as ethnic-specific effects on lung cancer susceptibility related to polymorphism in genes of xenobiotic-metabolizing enzymes have been shown to exist. Further, the relatively large sample size of our study made it possible to divide the material by histologic subtypes and smoking status, which revealed some associations not seen in the total material. The selection of cases in the study may not have been optimal, but it is, on the other hand, almost never possible to ascertain all cases from a study base. We consider it unlikely that the procedures for recruiting cases are differential with regard to the genetic polymorphisms studied. The control selection occurred separate from the case selection, but as the controls came from the same ethnic population and the same geographic areas as the cases, they should be genetically representative of the study base. The controls may, however, be unrepresentative of the study base for nongenetic factors (e.g., age, gender, and smoking), so that estimation of main effects of these variables or a full set of interaction parameters in the case-control analysis was not possible. Adjusting or stratifying for age and gender and (where appropriate) for smoking ensures that any comparison of genotypes between cases and controls occurs within levels of these variables.

The case series included a few (n = 9) cases of mesothelioma, which has been associated more with asbestos exposure than with smoking. All mesothelioma cases lacked smoking information and were consequently not included in statistical analyses where this information was considered. Among the cases that lacked information of histologic subtype, it cannot be excluded that one or two actually had mesothelioma. The possible inclusion of a few mesotheliomas in the study would not affect results appreciably.

In conclusion, the variant CYP1A1 and NQO1 genotypes were suggested to confer increased risk for squamous cell carcinoma, whereas the GSTM1*O/*O genotype was associated with small cell carcinoma risk. Stratification by cumulative smoking dose revealed that the risks noted for the variant CYP1A1 genotypes and GSTM1*O/*O genotype seemed to be restricted to light smokers. The GSTT1*O/*O genotype also appeared to be a possible risk factor in light smokers, whereas, in heavy smokers, this genotype decreased the risk for lung cancer overall. One important observation, which is corroborated by findings of DNA adducts (40), is that the variant CYP1A1 genotypes only seem to be risk factors when present in individuals with the GSTM1*O/*O genotype. Due to the many comparisons made in the study and the small number of light smokers, caution in interpretation of the results is required.

We thank Gösta Axelsson, Lars Hagmar, Ragnar Rylander, Janeric Seidegård, and Göran Tornling for kindly providing blood samples from lung cancer patients and some control subjects and for collecting valuable information via questionnaires from these subjects.