Association of Genetic Variants of O⁶-Methylguanine-DNA Methyltransferase with Risk of Lung Cancer in Non-Hispanic Whites

DNA repair maintains the genomic integrity of human cells against damage induced by environmental hazards, including tobacco carcinogens (1, 2). The G:C → A:T transition is a frequent mutation observed in tumors and cancer cell lines, which may result in activation of the K-ras oncogene or dysfunction of the p53 tumor suppressor gene (3, 4). Although spontaneous deamination of methylcytosine is a source of the G:C → A:T mutation (5), the O⁶-methylguanine lesions induced by the nicotine-derived 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone may also lead to the G → A mutation, and persistence of O⁶-methylguanine is one of the critical events during lung tumorigenesis in the A/J mice model (6, 7).

O⁶-methylguanine-DNA methyltransferase (MGMT) is a DNA repair enzyme that specifically repairs damage to the O⁶ position of guanine in DNA through transferring the O⁶-methyl group to an internal cysteine residue of the protein (8). Interestingly, the G:C → A:T mutation in both K-ras and p53 has been observed to be correlated with inactivation of MGMT (9, 10). In MGMT-defective mice, numerous lung adenomas and thymic lymphomas developed after exposure to low-dose methylnitrosourea (11). In contrast, transgenic mice carrying extra copies of the foreign MGMT gene showed decreased susceptibility to carcinogenesis (12). These results suggest a central role of the MGMT gene in protecting genomic integrity through repairing the O⁶-methylguanine induced by tobacco-specific nitrosamines, including 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone.

The MGMT gene is located at chromosome 10q26, and its mRNA contains only five exons. A single 207-amino-acid peptide, coded by only four exons with the translation starting codon ATG in the second exon, has been shown to have basic methyl-transferring activity (13). Twelve single nucleotide polymorphisms (SNPs) in MGMT exons have been identified in different populations, of which only three are non-synonymous SNPs with a variant allele frequency >0.05 in Caucasians. These three SNPs are codon 84C → T (L84F), codon 143A → G (I143V), and codon 178A → G (K178R) (14-16). They seem to have an effect on the function of MGMT, possibly leading to increased cancer risk. Therefore, there have been several case-control studies that evaluated the associations of these three SNPs with risk of bladder, lung, brain, and breast cancers (17-24). Although modest risks were reported in some studies for MGMT variants, there are no conclusive findings regarding the role of these MGMT genetic variants in cancer etiology because most studies included relatively small sample sizes of cases and controls with mixed ethnic backgrounds and cancer types in different populations.

A typical CpG island exists in the MGMT promoter region, and hypermethylation of CpG sites in this region is common in many cancers (25, 26), indicating that epigenetic events through DNA cytosine methylation are one of the important mechanisms regulating MGMT mRNA expression in specific targeted tissues (27). In addition, there are five SNPs found in the MGMT promoter region, with allele frequencies of 135G → T and 485C → A reported as 0.15 and 0.37, respectively (14). It is likely that genetic variation in the promoter region may also play a role in the MGMT-associated susceptibility to lung cancer (23). To test the hypothesis that genetic variants in exons and the promoter of MGMT are associated with increased lung cancer risk, we selected and genotyped two non-synonymous SNPs [i.e., exon 3 codon 84 C → T (L84F) and exon 5 codon 143A → G (I143V) that is in complete linkage disequilibrium with codon 178A → G (K178R) in Caucasian population (20)] and two promoter SNPs (i.e., 135G → T and 485C → A) in a large lung cancer case-control study.

The recruitment of study subjects has been described previously (28). Briefly, the subjects in the final analysis included 1,121 patients with lung cancer and 1,163 cancer-free control subjects recruited consecutively during the period of August 1995 to December 2004. The case subjects were patients with newly diagnosed and histopathologically confirmed lung cancer. There were no restrictions on age, sex, disease stage, or histology, but all of them were non-Hispanic Whites. The controls were selected from a pool of cancer-free individuals recruited through the Kelsey-Seybold multispecialty physician practice with multiple clinics throughout the Houston metropolitan area and were frequency matched with the patients according to age (±5 years), sex, ethnicity, and smoking status (i.e., current, former, and never). The exclusion criteria for both cases and controls, when appropriate, included previous cancer treatment (i.e., radiotherapy or chemotherapy), previous cancer except non-melanoma skin cancer, and recent blood transfusion (within previous 6 months). After providing a written informed consent, each participant in this study was interviewed by an interviewer with a structured questionnaire, and the blood samples were collected. The study was approved by the institutional review boards of the M.D. Anderson Cancer Center and the Kelsey-Seybold Foundation.

We chose those SNPs with allele frequencies >0.05 and those that are likely to have biologically functional relevance (i.e., non-synonymous SNPs and the SNPs in the promoter region; ref. 29). MGMT mRNA only has five exons, and there is no evidence that alternative splicing involves in regulation of MGMT mRNA. Therefore, we did not include intron SNPs. Three non-synonymous SNPs [codon 84C → T (L84F), codon 143A → G (I143V), and codon 178A → G (K178R)] were chosen, but codon 178A → G (K178R) was subsequently excluded because it is in strong linkage disequilibrium with codon 143A → G (I143V) (D′ = 0.96, r² = 0.89, P < 0.01; refs. 20, 22). We also included two SNPs (i.e., 135G → T and 485C → A) in the MGMT promoter region with reported allele frequencies of 0.15 and 0.37, respectively (14).

Genomic DNA was extracted from blood leukocyte cell pellet by using the Qiagen DNA blood mini kit (Qiagen, Valencia, CA) according to the manufacturer's instruction.

The PCR-RFLP method was used for genotyping. Briefly, a 10-μL PCR reaction mixture was assembled with 40 ng genomic DNA and cycled in a PTC-200 DNA Engine (Peltier Thermal Cycler, MJ Research, Inc., Watertown, MA). The information on primers used to detect the SNPs is listed in Table 1. The lengths of PCR fragments for 135G → T, 485C → A, codon 84C → T (L84F), and codon 143A → G (I143V) were 168, 212, 102, and 179 bp, respectively. The restriction enzymes HhaI, BanI, EarI, and BstF5I (New England Biolabs, Beverly, MA) were used to distinguish these SNPs by generating the 144- and 24-bp fragments in the presence of the 135G → T G allele; 139- and 73-bp in the presence of the 485C → A A allele; 68-, 30-, and 4-bp in the presence of codon 84C → T (L84F) T allele; and 157- and 22-bp for codon 143A → G (I143V) A allele. The PCR products were visualized on a 3% agarose gel containing 0.25 μg/mL ethidium bromide and photographed. The genomic structure of the MGMT gene and the PCR-RFLP patterns of four SNPs are shown in Fig. 1. We also randomly selected >10% of all genotyped DNA samples and repeated PCR-RFLP genotyping for validation of results. The repeated data were 100% concordant with the original genotyping results.

The demographic characteristics (e.g., age and sex), smoking status, pack-years smoked, and tumor histology were included in our analysis. Those who had smoked fewer than 100 cigarettes in their lifetimes were defined as never smokers, and the others were defined as ever smokers. Of the ever smokers, those who quit smoking a year before interview were considered former smokers, and the remainder were considered as current smokers. We used the χ² test to detect the differences in the frequency distributions of categorical variables as well as the genotypes and alleles of the selected MGMT SNPs between the cases and controls. In addition, the Student's t test was used to detect the differences in numerical variables, such as age and number of pack-years smoked.

To assess risk of lung cancer for individuals who carried variant alleles of MGMT, we defined the minor alleles of MGMT exon 3 codon 84C → T (L84F) (T allele) and exon 5 codon 143A → G (I143V) (G allele) as the variant alleles based on their higher frequencies in cases than that in controls. Finally, we created a summary variable based on the numbers of variant alleles (0 versus 1-4 variants). Unconditional multivariate analyses were done to calculate adjusted odds ratios and 95% confidence intervals for the defined MGMT variant genotypes with adjustment for age, sex, smoking status, and pack-years in logistic regression models when appropriate. We further stratified frequency distributions of the defined MGMT variant genotypes by age, sex, smoking status, pack-years, and tumor histologies as well as the MGMT promoter SNPs 135G → T and 485C → As and assessed their potential interactions by fitting a product of two interactive variables in a multiplicative model with adjustment for their main effects and other covariates. All tests were two sided with a significance level of <0.05 and done with the Statistical Analysis System Software program (SAS software, version 9.1; SAS Institute, Inc., Cary, NC).

Selected characteristics of the study population are shown in Table 2. All subjects were non-Hispanic Whites. The cases and controls were well matched on age with a mean age of 61.4 years in the cases and 61.1 years in the controls. Similarly, the cases and controls were well matched on sex. Although the cases and controls were frequency matched on smoking status, there were more current smokers among the cases than the controls (41.0% versus 35.3%, P = 0.017). Cases were also more likely to be heavier smokers, and their mean number of pack-years smoked was significantly higher (44.7 ± 35.3) than that of the controls (35.8 ± 31.1; P < 0.001). For any residual effect of the frequency matching, we further adjusted for these variables in the multivariate logistic regression analysis.

We evaluated the observed genotypes in the cancer-free controls by performing the Hardy-Weinberg equilibrium (p² + 2pq + q² = 1) test, and the Ps in controls were 0.708 for the promoter SNP 135G → T, 0.850 for the promoter SNP 485C → A, 0.674 for codon 84C → T (L84F), and 0.916 for codon 143A → G (I143V). We then compared the overall genotype frequency distribution of these four MGMT SNPs between the cases and controls (Table 3). Although the differences were not statistically significant for any single SNP tested, the adjusted odds ratios showed that homozygotes for the promoter SNP 135T or 485A minor allele had slightly reduced risk of lung cancer, compared with their homozygotes of the common allele, whereas homozygotes for either the 84T or 143G allele were associated with a moderately increased risk, compared with their homozygotes of the common allele (Table 3). By combining two non-synonymous SNPs [codon 84C → T (L84F) and codon 143A → G (I143V)], we created a summary variable of variant alleles (i.e., T in codon 84C → T and G in codon 143A → G) and categorized the codon 84CC and codon 143AA as the reference group (0 variants), and the others as 1-4 variants (Table 4). The overall adjusted odds ratio associated with the 1-4 MGMT variants was 1.19 (95% confidence interval, 1.01-1.41; P = 0.040) compared with 0 variants.

We further examined the effects of the MGMT variants on lung cancer risk in stratification analyses by selected variables (Table 5). Significantly increased risk of lung cancer was more evident for those carrying 1-4 variants who were women, current smokers, and non–small cell lung cancer. There was a borderline significant trend that risk associated with 1-4 MGMT variants increased as the age decreased (P = 0.055).

Finally, we assessed the effects of the two promoter SNPs on the MGMT variants defined from the two non-synonymous SNPs (Table 5). Among those in the 1-4 MGMT variant subgroup, there was an increased risk of lung cancer in the subgroup who carried either the T allele of promoter SNP 135G → T or C allele of 485C → A with borderline statistical significance (P = 0.060 and 0.067, respectively). However, we did not detect any interaction between the MGMT variants of the two non-synonymous SNPs and age, smoking status, or presence of the promoter SNPs (data not shown).

As one of the most important events during tumorigenesis, mutations in DNA need to be repaired to maintain genomic integrity of the cells. MGMT is of particular interest for its role in sporadic lung carcinogenesis because inactivation of MGMT correlates with mutations in K-ras and p53 (9, 10). In this study, we investigated the association between genetic variants of MGMT and risk of lung cancer. We found that the allele frequency distributions of the codon 84T, codon 143G, and promoter SNP 135T variant genotypes were associated with a non-statistically significantly increased risk of lung cancer. By creating a new variable of MGMT variant genotype based on two non-synonymous SNPs (codon 84 and codon 143), we found that 1-4 variants were associated with increased risk of lung cancer compared with the homozygous wild-type of 0 variants (84CC-143AA). The risk was more pronounced in women, current smokers, and non–small cell lung cancer. These findings support our a priori hypothesis that genetic variants of MGMT are associated with increased lung cancer risk. It also compliments our previous report that epigenetic DNA cytosine methylation involves in the etiology of lung cancer, as shown by the finding that a functional C46359 C → T SNP in a novel DNMT3B promoter was associated with lung cancer risk (27, 30).

Several studies have investigated the role of MGMT/AGT variants in the etiology of lung cancer (19, 20, 23, 24). However, most of the studies had included relatively small numbers of subjects of different ethnic background. One early study found that the 178A → G SNP was associated with risk of lung cancer in Caucasians (20), but this was not substantiated by other two later studies (19, 24). The most recent and largest study of a Korean population with 432 cases and 432 controls (23) found that among the SNPs (i.e., 485C → A, Leu53Leu, and Leu84Phe) tested, only the promoter 485C>A was associated with risk of lung cancer. However, we did not find any evidence for an association of the promoter 135G → T and 485C → A SNPs with risk of lung cancer in the present study of 1,121 cases and 1,163 controls, the largest study to the best of our knowledge.

MGMT is one of the most conserved genes that protect genomic integrity against carcinogenic damage. To date, a total of 662 SNPs in the human MGMT gene have been reported in the dbSNP database of the National Center for Biotechnology Information. Most of the SNPs are located in introns or non-coding regions, and there are only 12 exonic SNPs, 8 non-synonymous, and 4 synonymous. In general, those SNPs altering the conserved amino acids are more likely to be associated with cancer susceptibility (29), and in the two common MGMT non-synonymous SNPs, codon 178A → G (K178R) is in strong linkage disequilibrium with codon 143A → G (I143V) (20, 22). Therefore, the codon 84C → T and codon 143A → G of MGMT are two relevant candidate SNPs in our study. This choice was also supported by a finding that the cells from individuals carrying either the codon 84T (84F) or 143G (143V) alleles were found to be associated with significantly increased frequencies of chromosome aberrations induced by the tobacco carcinogen 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (31).

Two functional domains of human MGMT have been interpreted by its crystal structure and enzymatic-kinetic study (32, 33). The NH₂-terminal residues 1 to 91 of MGMT contains a domain that can bind with zinc, and zinc binding will greatly enhance the methyl transferring activity of the intact MGMT protein. Meanwhile, the COOH-terminal residues 92 to 207 contain all residues thus far implicated in methyl-transferring activity, including a cysteine acceptor site (Cys¹⁴⁵), an O⁶-methylguanine–binding pocket, and a DNA-binding domain (33). It is interesting that codon 84C → T (L84F) is located at the NH₂-terminal domain, and codon 143A → G (I143V) is located at the COOH-terminal domain. Therefore, it is likely that codon 84T (84F) and codon 143G (143V) variants may present different zinc-binding and methyl-transferring activities, respectively. Loss of such functionality in the variants is consistent with our finding that these variants were associated with increased risk of sporadic lung cancer in Caucasians. However, validation studies are needed in the populations with well-defined ethnic backgrounds because studies have shown that there was no codon 143A → G (I143V) variant in Asians (18) and Koreans (23).

The observed increased risk in women is of interest, but this kind of association study cannot provide a clear underlying mechanism. However, studies have shown that women seem to be more sensitive to tobacco-induced DNA damage (34). This may lead to increased O⁶-methylguanine, the repair of which may be impaired in the presence of certain MGMT variants. Moreover, it seems that women tend to have a suboptimal DNA repair capacity to remove tobacco-induced DNA adducts (28). However, more evidence is needed to interpret the functionality of MGMT variants in maintaining the genomic integrity in humans, particularly in women whose DNA repair capacity may be under the influence of altered hormone status (35).

The strength of this study is that we only selected those SNPs that are most likely to be functional. Although we also included two MGMT promoter SNPs in this study, we did not find any risk associated with these variant genotypes, which may imply that epigenetic modification may play a more important role than sequence variation in the MGMT promoter in risk of lung cancer. This result is consistent with our previous finding that epigenetic modulation through de novo DNA-cytosine methylation was associated with increased risk of lung cancer (35).

In summary, we tested our hypothesis that genetic variants of MGMT are associated with increased lung cancer risk. We found that in non-Hispanic Whites, individuals having 1-4 MGMT variants showed a moderately increased risk of sporadic lung cancer compared with those having 0 variants. Stratification analysis further showed that this increased risk was more pronounced in women, ever smokers, and non–small cell lung cancer. Our findings suggest that non-synonymous SNPs in MGMT are associated with modestly increased risk of lung cancer in Caucasians, but these findings need to be further validated.

We thank Susan Honn and the team for subject recruitment and database maintenance; Li-E Wang, Zhaozheng Guo, Yawei Qiao, Jianzhong He, and Kejing Xu for technical assistance; Joanne Sider for article preparation; and Donald R. Norwood for article editing.