DNA Repair and Cell Cycle Control Genes and the Risk of Young-Onset Lung Cancer

Exposure to tobacco smoke and to mutagenic xenobiotics can cause various types of DNA damage in lung cells, which, if not corrected by DNA repair systems, may lead to deregulation of the cell cycle and, ultimately, to cancer. Genetic variation could thus be an important factor in determining susceptibility to tobacco-induced lung cancer with genetic susceptibility playing a larger role in young-onset cases compared with that in the general population. We have therefore studied 102 single-nucleotide polymorphisms (SNP) in 34 key DNA repair and cell cycle control genes in 299 lung cancer cases diagnosed before the age of 50 years and 317 controls from six countries of Central and Eastern Europe. We have found no association of lung cancer risk with polymorphisms in genes related to cell cycle control, single-strand/double-strand break repair, or base excision repair. Significant associations (P < 0.05) were found with polymorphisms in genes involved in DNA damage sensing (ATM) and, interestingly, in four genes encoding proteins involved in mismatch repair (LIG1, LIG3, MLH1, and MSH6). The strongest associations were observed with heterozygote carriers of LIG1 −7C>T [odds ratio (OR), 1.73; 95% confidence interval (95% CI), 1.13-2.64] and homozygote carriers of LIG3 rs1052536 (OR, 2.05; 95% CI, 1.25-3.38). Consideration of the relatively large number of markers assessed diminishes the significance of these findings; thus, these SNPs should be considered promising candidates for further investigation in other independent populations. (Cancer Res 2006; 66(22): 11062-9)

Exposure to tobacco smoke and to mutagenic xenobiotics causes various types of DNA damage to lung cells (1). Such lesions and the mutations that arise from them can lead to alterations in the normal program of cell cycle and growth, favoring the development of cancer. Cells counteract the deleterious effects of damaged DNA through different repair pathways and strategies, including cell cycle control and apoptosis induction (2). The “base excision repair” (BER) pathway is the primary mechanism to remove incorrect and damaged bases, including those formed by reactive oxygen species, such as 8-hydroxyguanine, and methylating agents (e.g., 7-methylguanine and 3-methyladenine; ref. 3). The mismatch repair (MMR) pathway corrects errors of DNA replication that escape the proofreading functions of DNA polymerases as well as heterologies formed during recombination. It has also been linked to cell cycle checkpoint activation and cell death, and thus, alterations in this process can have wide-ranging biological consequences (4).

The “nucleotide excision repair” (NER; either “global genome” NER or “transcription-coupled” NER) acts on a variety of helix-distorting DNA lesions, such as the pyrimidine dimers caused by UV light, and chemical adducts, such as those caused by benzo(a)pyrene (5). Finally, direct reversal of DNA damage, an unusual form of DNA repair, is also present in human cells. The promutagenic lesion O⁶-methylguanine is repaired by such a process where a specific methyltransferase, encoded by the MGMT gene, transfers the methyl group from the DNA guanine residue to a cysteine residue located at its active site in an error-free process (6).

Cells must also cope with the presence of DNA single-strand and double-strand breaks (SSB and DSB). SSBs can arise directly from damage to the DNA or indirectly as intermediates in the BER pathway. Like many of the other DNA repair pathways in human cells, the processing of such breaks involves a series of coordinated, sequential reactions in which the damage is detected and processed and any gaps generated are filled and religated. Key proteins in this process are XRCC1, poly(ADP-ribose) polymerase 1 (PARP1), and the proliferating cell nuclear antigen (PCNA; ref. 7). The repair of DSBs, which are dangerously cytotoxic lesions formed after exposure to ionizing radiation, occurs through two main mechanisms: nonhomologous end joining (NHEJ) and the homologous recombination repair (HRR). These two processes have different fidelities: HR can be achieved with high fidelity, whereas NHEJ may result in loss of genetic information (8, 9).

A second arm in the maintenance of genome stability in the presence of DNA damage is achieved by the precise regulation of the cell cycle. DNA strand breaks and certain base adducts are very effective in inducing cell cycle blocks. Such damage is detected through different mechanisms involving the protein products of key genes, such as ataxia-telangiectasia mutated (ATM) and ATM and Rad3-related (ATR), which act upstream of p53 and other checkpoint proteins to induce cell cycle arrest (10). In addition to these mechanisms, apoptosis, the program of cell suicide, which is an important escape route to avoid damaged cells progressing to a malignant phenotype, can be activated in response to DNA damage. The ATM and p53 proteins are also involved in this process (11).

All these pathways contribute to the genome stability of the cell, and a deficiency in one or more of the relevant genes can lead to deregulated cell growth and, ultimately, to cancer development. The link between alterations in DNA repair genes and increased cancer risk is well documented in certain inherited disorders, such as xeroderma pigmentosum (deficiencies in NER), Bloom syndrome (deficiency in the DNA helicase RECQ), Fanconi anemia (FANCx genes), ataxia-telangiectasia (ATM), human nonpolyposis colorectal cancer (HNPCC, MMR genes), retinoblastoma (RB1), familial cutaneous melanoma (CDKN1A), Li-Fraumeni syndrome (TP53), familial breast/ovary cancers (BRCA1 or BRCA2), and Nijmegen breakage syndrome (NBS1). In addition, the importance of the integrity of DNA repair pathways to prevent lung carcinogenesis has been shown in several animal models where individual DNA repair genes are disrupted. For example, Xpc^−/− mice show high susceptibility to lung cancer following exposure to acetylaminofluorene (12), Mgmt^−/− and Parp1^−/− mice to nitrosamine-induced tumors (13, 14), and Ogg1^−/− mice to the spontaneous development of lung cancers (15).

Polymorphisms in DNA repair and cell cycle control genes, which affect the normal protein activity, may alter the efficiency of these processes and lead to genetic instability and increased cancer risk. Reduced DNA repair capacity and impaired control of the cell cycle are phenotypes found in lung cancer patients, and polymorphisms, which have been reported to be associated with the risk of developing lung cancer, include OGG1 S326C, XRCC1 R194W, MLH1 −93 G>A, PARP1 V762A, ERCC2 Gln751, RAD23B Val249, XPA 5′-untranslated region (UTR) G>A, CCND1 exon 4 A>G, and TP53 codon 72, intron 3, and intron 6 (16–21).

These results support the hypothesis that genetic variations within DNA repair and cell cycle control pathways might be important in determining an individual risk of developing lung cancer. Thus, to investigate the role of gene variants in DNA repair and cell cycle control pathways, we have analyzed 102 single-nucleotide polymorphisms (SNPs) in 51 genes in a multicenter case-control study of early-onset lung cancer. Our rationale for restricting the analysis to subjects diagnosed before age 50 years was that the role of genetic susceptibility would be expected to be more important among subjects with a younger age of diagnosis. Indeed, previous epidemiologic studies have suggested that a large proportion of lung cancers occurring before age 50 years have a genetic component. Risks due to genetic factors are further amplified by cigarette smoking (22, 23).

Due to the quite low incidence of the disease occurring at younger ages, there are very few published studies on the genetic risk factors for early-onset lung cancer. The majority of these investigated the role of xenobiotic metabolism genes (24–26), and none have explored, using a comprehensive panel of variants, the association of variants in DNA repair and cell cycle control genes and lung cancer risk.

Study population. This study was conducted in 15 centers in six Central and Eastern European countries: the Czech Republic (Prague, Olomouc, and Brno), Hungary (Borsod, Heves, Szabolcs, Szolnok, and Budapest), Poland (Warsaw and Lodz), Romania (Bucharest), Russia (Moscow), and Slovakia (Banska Bystrica, Bratislava, and Nitra). Each center followed an identical protocol and was responsible for recruiting a consecutive group of patients who were newly diagnosed with lung cancer and a comparable group of hospital-based control subjects without lung cancer from February 1998 to October 2002. All cancer diagnoses were confirmed histologically or cytologically. Eligible subjects (case patients and control subjects) must have resided in the study area for at least 1 year before recruitment. Lung cancer case patients were identified through an active search of the records of clinical and pathology departments at the participating hospitals. All centers attempted to recruit all eligible patients as soon as possible after the patient had received an initial diagnosis of lung cancer. The maximum time interval between diagnosis and recruitment was 3 months. All study subjects (case patients and control subjects) and their physicians provided written informed consent. This study was approved by the institutions at all study centers, and ethical approval was obtained from the IARC (Lyon, France), the coordinating center. At all centers, except the Warsaw center, control subjects were chosen from among inpatients and outpatients admitted to the same hospitals as the case patients or hospitals serving the same population; case patients and control subjects from each hospital were frequency matched by sex, age (±3 years), center, and referral (or residence) area. Control subjects were eligible for this study if they had been diagnosed with non-tobacco-related diseases or had undergone minor surgical procedures or had benign disorders, common infections, eye conditions (except cataract or diabetic retinopathy), or common orthopedic diseases (except osteoporosis). At the Warsaw center, control subjects were selected by random sampling of the general population using the Electronic List of Residents. Overall, the average participation rate was 91.0% among case patients and 91.2% among control subjects. Case patients and control subjects underwent an identical in-person interview during which they completed a detailed questionnaire and provided blood samples. The questionnaire collected information about demographic variables, such as sex, date of birth, and education level, medical history, family history of cancer, history of tobacco consumption, including frequency, intensity, duration, and status, history of alcohol consumption, diet history (using a general food frequency questionnaire), and occupational history. Blood samples were stored in liquid nitrogen. Of the 2,633 case patients and 2,884 control subjects who agreed to participate in this study, 2,188 (83%) case patients and 2,198 (76%) control subjects provided blood samples during the interview of whom 299 cases and 317 controls were ≤50 years of age at diagnosis and at recruitment, respectively, and form the study population for this analysis.

Selection of polymorphisms. The polymorphisms that were included in the panel to be studied were selected based on the following criteria: (a) having a biological significance shown through functional studies (e.g., OGG1 326Ser has a 7-fold higher activity for repairing 8-oxoG than the 326Cys variant), (b) associated with biological end points in epidemiologic studies [e.g., XRCC1 399Gln allele was associated with increased levels of aflatoxin B1-DNA adduct and increased bleomycin sensitivity (27–29)], or (c) associated with the risk of cancer at any site in epidemiologic studies [e.g., the polymorphism at codon 194 within XRCC1 (30)]. However, for some DNA repair and cell cycle control genes, there were no variants that met these criteria. Therefore, for some genes, we added selected SNPs from dbSNP,¹⁴

which either (a) had a high frequency of the rare allele (to allow the highest statistical power to detect associations) or (b) coded for missense changes. Among the missense variants, we focused, in particular, on changes having the greatest potential for functional relevance, such as involving a proline residue (e.g., PARP P1328T) or the charge/polarity of the lateral group (e.g., CDKN2A A148T). Although the biological effect of many of the polymorphisms selected for study is not known, we expect that these criteria should have maximized the likelihood of having chosen SNPs with functional significance or associated with cancer.

Laboratory techniques. Genomic DNA was extracted from blood samples using Puregene chemistry (Gentra Systems, Minneapolis, MN). DNA concentrations were measured by using PicoGreen dsDNA quantification kits (Molecular Probes, Leiden, the Netherlands). All polymorphisms were analyzed together for a given sample by a microarray technique based on the arrayed primer extension (APEX) principle.

APEX consists of a sequencing reaction primed by an oligonucleotide anchored with its 5′-end to a glass slide and terminating just one nucleotide before the polymorphic site. A DNA polymerase extends the oligonucleotide by adding one fluorescently labeled dideoxynucleotide triphosphate (ddNTP) complementary to the variant base. Reading the incorporated fluorescence identifies the base in the target sequence. This method is suitable not only for SNPs but also for small insertion/deletion polymorphisms. Because both sense and antisense strands are probed, two oligonucleotides were designed for each polymorphism. In general, two 30-mers, one for each strand, complementary to each side of the polymorphism were designed both with their 3′-end pointing toward the polymorphism. The flanking sequences and their related APEX-oligonucleotides have been previously published (31). Five-prime (C-12) amino-linker oligonucleotides were synthesized by Sigma-Genosys (Cambridge, United Kingdom) and spotted onto silanized slides (32).

Genomic DNAs were amplified to enrich the fragments carrying the SNPs by using specific primer pairs. Then, PCR products were pooled, purified, concentrated using Millipore (Bedford, MA) Microcon MY30 columns, and fragmented. For single-base extension reaction, fragmented PCR products were incubated onto the slides together with the fluorescently labeled ddNTPs (4 × 50 pmol), 10× buffer, and 4 units ThermoSequenase (GE Healthcare, Little Chalfont, United Kingdom). All the details of the experimental protocol, including primer and probe sequences, were reported in previous articles (31, 33).

Slides were imaged by a Genorama-003 four-color detector equipped with Genorama image analysis software (Asper Biotech, Tartu, Estonia). Fluorescence intensities at each position were converted automatically into base calls by the software under the supervision of trained personnel. In case of more than one signal present on a given position, only the main signal was considered if the intensity of the weaker signal was <10% of the main signal. APEX gives a very high concordance compared with the standard genotyping or sequencing methods as assessed in previous studies (31, 33).

To ensure quality control, we followed several strategies: DNA samples from case patients and control subjects were randomly distributed, and all genotyping was conducted by personnel who were blinded to the case-control status of the subjects; each APEX oligonucleotide was spotted in replicate; each SNP was analyzed independently by genotyping both the sense and the antisense strands of the DNA (in case of disagreement, the base call was discarded); internal positive controls allowed to verify that the intensities of the four channels (A, C, T, and G) were equilibrated; base calls were scored by three independent trained operators and discordant results were rechecked and, in case of disagreement, discarded; DNA samples from individuals of known genotypes were added to check periodically the validity of the genotyping; we randomly selected 10% of the study subjects (i.e., both case patients and control subjects) and regenotyped their DNA samples for each polymorphism.

Statistical analysis. The frequency distributions of demographic variables and putative risk factors for lung cancer, including country of residence, age at recruitment (which for case patients was a proxy for age at diagnosis), sex, highest education level, and smoking status, were examined for case patients and control subjects. Former smokers were defined as smokers who stopped smoking at least 2 years before the interview. Tobacco consumption included smoking of cigarettes, pipes, and cigars. Cumulative tobacco consumption was calculated by multiplying smoking duration (in years) by smoking intensity (in the equivalent of cigarette packs) and expressed as pack-years. We categorized the subjects as light (≤14 pack-years), moderate (>14-38.26 pack-years), or heavy (>38.26 pack-years) smokers based on the tertiles of cumulative tobacco consumption among the control group. We tested the Hardy-Weinberg equilibrium of genotype distributions separately among case patients and control subjects. The minimum detectable odds ratio (OR) was calculated for each sequence variant based on its genotype frequency, our study sample size, and a statistical power of 80% as described previously (34). Our study had an 80% power to detect a minimum OR of 2.5 for relatively rare variants (5%) and a minimum OR of 1.6 for common variants (≥30%).

We used unconditional multivariate logistic regression analysis to examine associations between genetic polymorphisms and lung cancer risk by estimating ORs and 95% confidence intervals (95% CI). Genotypes were categorized into three groups (major allele homozygous, heterozygous, and homozygous variant) when the allele frequencies allowed or into two groups (major allele homozygous and minor allele carriers) for rare polymorphisms. Age, sex, country, and tobacco pack-year were included in all analyses as covariates.

We computed false-positive response probabilities (FPRP; ref. 35) for the nominally significant associations we have observed between SNPs and lung cancer risk. Prior probability is likely to be influenced by the biological knowledge of the gene, the functional significance of the variants, and the available epidemiologic evidence. It remains a subjective measure that may vary from one investigator to another based on the importance they assign to the different pieces of evidence. For this reason, we have calculated FPRP for a range of prior probabilities from 50% to 0.1%. Following Wacholder et al., we used a threshold of noteworthiness of FPRP ≤0.2. All statistical analyses were conducted using STATA software version 8.0 (Stata Corp. LP, College Station, TX). All statistical tests were two sided.

Table 1 shows the frequency distribution of demographic characteristics among the 299 cases and 317 controls. As expected, the proportion of current smokers was far higher among the cases than controls (86.6% versus 53.6%; P < 0.001). The sex and age distributions were similar among cases and controls, although there was a tendency for cases to have a lower education level than the controls (P = 0.01). With respect to the histologic types of the tumors in the cases, the numbers of adenocarcinoma, squamous, and small cell carcinomas were similar. We tested departure from Hardy-Weinberg equilibrium in the controls by a χ² test using P = 0.01 as threshold. This threshold was chosen based on anticonservativeness of this test as noted by Wigginton et al. (36). All SNPs (except for MSH3 235G>A, MSH6 G39E, ERCC1 354T>C, ERCC2 D312N, and XRCC3 17893A>G) were in Hardy-Weinberg equilibrium in this population. The results of the initial and the repeat genotyping analyses were at least 99% concordant. Unfortunately, because of the nature of the method of genotyping where all the SNPs are analyzed on one chip, genotyping that failed for any individual SNP could not be repeated. Genotyping success rates for individual polymorphisms averaged 92%.

Table 2 shows ORs for polymorphisms in genes involved in detecting DNA breaks as well as in the negative modulation of cell cycle (including apoptosis). Only one SNP (IVS48+238 C>G within ATM) was found associated with a decreased risk of lung cancer (P_trend = 0.03), with the other 34 SNPs (19 genes) showing no association. Table 3 shows similar results for SNPs within genes involved in SSB/DSB repair. These genes can be considered as downstream with respect to the first set of genes, being actually involved in the NHEJ, HRR, and SSB repair (SSBR). Among these 26 SNPs (12 genes), we found no overall association among the homozygote variants. Table 4 shows the results for SNPs within genes in the NER, BER, and MMR pathways and the MGMT gene. We analyzed 40 SNPs (20 genes) and found a borderline association for carriers of Gln751 in ERCC2 (OR, 0.7; 95% CI, 0.49-1.00), an increased risk for carriers of a SNP in the 3′-UTR of LIG3 (OR, 1.7; 95% CI, 1.13-2.56), a reduced risk for carriers of Val219 in MLH1 (OR, 0.69; 95% CI, 0.48-0.98), a borderline significant association for the homozygote carriers of the 540T allele in MSH6 (OR, 1.95; 95% CI, 1-3.78), and an increased risk for two SNPs in LIG1 (−7C>T and IVS9-21), with the strongest OR of 1.89 (95% CI, 1.24-2.87) for carriers of −7 T allele. Among these associations, it is interesting to note that LIG1, LIG3, MLH1, and MSH6 are all involved in the MMR. Calculation of FPRP showed that none of the above associations remained noteworthy (FPRP ≤ 0.2) when a prior probability of association of ≤1% was considered, and only the association of LIG1 −7C>T remained noteworthy assuming a prior probability of 10% (FPRP = 0.154).

We have undertaken an exploratory investigation of genes involved in cell cycle and DNA repair among a population of lung cancer cases and controls on the basis that they are likely to be enriched for risk-associated genetic variants. The validity of this assumption is supported by the increased report of family history of lung and other tobacco-related cancers among the young-onset cases as opposed to those who developed lung cancer after the age of 50 years. Among the young-onset cases, there was over a 2-fold risk of lung cancer in first-degree family members (OR, 2.06; 95% CI, 1.16-3.65) and over a 4-fold risk of other tobacco-related cancers (OR, 4.94; 95% CI, 1.47-16.62). Conversely, among those ages >50 years, the risk of lung cancer among family members was much reduced (OR, 1.45; 95% CI, 1.15-1.81) and the risk of other tobacco-related cancers was not apparent (OR, 0.86; 95% CI, 0.58-1.29). Although we found several interesting statistically significant associations with modest increases in risk, none were significant when a prior probability of association of 1% was applied. When interpreting such results, it should be taken into account that, among the polymorphisms investigated in the present study, information on their effect on expression and protein function is available for only few of the SNPs studied. Although haplotype-tagging SNPs are available for all of the gene studies, this information was not available when SNPs were being selected for the current study. We therefore focused on variants with relatively high prevalence and those coding for missense changes, as these were more likely to affect the function of the encoded protein.

It is interesting to note that most of the genes with a significant association (i.e., LIG1, LIG3, MLH1, and MSH6) belong to the MMR and, to a minor extent, BER pathways. Although lung cancer is not considered part of the HNPCC spectrum (37), a significant excess of lung, liver, and brain tumors has been detected within HNPCC families (38). Moreover, an altered expression, including a complete lack of protein, of MLH1, in lung cancer, in conjunction with the presence of microsatellite instability in tumor tissue or allelic imbalance of MSH2, has been reported (39, 40). MLH1 was found inactivated in lung cancers following promoter methylation, indicating a role of this gene in lung cancer (27). It has also been suggested that the MLH1 and MSH3 genes could be involved in lung tumorigenesis through a gene dosage effect in those tumors that are hemizygotes and retaining the wild-type copies of MLH1 and MSH3 and not showing microsatellite instability (41). In a previous study from Korea, the variation −93 G>A within MLH1 was found associated with increased risk of squamous cell lung carcinoma (42). Moreover, the exposure to chromate seems to promote lung cancer also through the inhibition of hMLH1 (43). Finally, engineered mice deficient in the MMR gene Mutyh are prone to develop lung cancer (44). Our findings are consistent with the importance of MMR in protecting against the development of lung cancer and deserve further investigation.

One of the variants that was found to be associated with a protective effect was the IVS48+238 C>G variant in the ATM gene. This variant, when present in the homozygous state, has been previously found to be associated with an increased risk of breast cancer and, when present in the heterozygous state, with a decreased risk of therapeutic radiation sensitivity in breast cancer patients (45). Exposure to X-rays through occupational examinations was associated with an increased risk of lung cancer in a larger study population from which the cases and controls investigated here were selected (46). Clearly, further investigation of the role of the variants in the ATM gene and lung cancer risk is warranted.

The present study has some potential limitations. Potential selection bias can occur when subjects who agree to participate in the study have characteristics that differ from those of subjects who are eligible for the study. The major reason for nonparticipation in this study was the refusal of some eligible subjects, which might represent a form of self-selection. Bias from self-selection may affect estimates of exposure to environmental factors; however, it is unlikely that self-selection would be related to a subject's genotype. Results of a simulation study (47) suggested that selection factors that are related only to environmental factors might still lead to a biased estimate of genetic main effects if the environmental factors modify the genetic effects. Nevertheless, bias in the estimate of the genetic main effect due to the selection factor of environmental exposures can be adjusted for in the analyses by treating the environmental factors as potential confounders as we have done (47). Another limitation is the missing data generated in the process of genotyping. These missing data are likely to be independent from demographic variables and smoking status, and their consequence would be to reduce the statistical power of our study. However, the missing data were associated with country as the quality of the DNA samples varied from one country to another. Nevertheless, the missing data were not likely to bias the estimates to any meaningful extent because the call rate was high (with an average of 92%) and independent from the case-control status. In addition, any variation in the call rate by country was controlled for in the multivariable models.

Another potential limitation is related to the number of cases and controls, which limits the power of the study to detect significant associations particularly for rare alleles responsible for weak associations. For example, in our study, the XRCC1 polymorphism R194W is associated with a reduced risk of cancer, in agreement with the results reported in literature for this polymorphism and a recent meta-analysis (30), but the association was not strong enough to reach statistical significance. We also did not observe the association with the OGG1 S326C SNP that we found when we analyzed the complete study population, regardless of age (48). We think that the reason for this is that the OGG1 SNP confers risk only to subjects who are homozygous for the minor allele, a condition too rare to be associated with a detectable risk with the sample size of the present study. Despite these limitations, the present study is one of the largest investigations, on the role of DNA repair and cell cycle control genes in relation to the risk of lung cancer in people with a young age. The findings of this study shed a new light on early-onset lung cancer; however, confirmation in a second independent study is required. Nonetheless, further studies on the role of ligase LIG1 and LIG3 seem promising.

Grant support: European Commission (DG-XII; contract IC15-CT96-0313), National Cancer Institute R01 grant (contract CA 092039-01A2), Association for International Cancer Research grant (contract 03-281), “Marie-Curie Reintegration Grant” (contract MERG-CT-2004-506373), and Associazione Italiana per la Ricerca sul Cancro principal investigator grant 2005. The Warsaw part of the study was supported by a local grant from The Polish State Committee for Scientific Research (grant SPUB-M-COPERNICUS/P-05/DZ-30/99/2000). F. Gemignani is a recipient of a fellowship from the International Association for the Study on Lung Cancer.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.