Potentially Functional Single Nucleotide Polymorphisms in the Core Nucleotide Excision Repair Genes and Risk of Squamous Cell Carcinoma of the Head and Neck

Susceptibility to cancer has been associated with DNA repair capacity, a global reflection of all functional variants, most of which are relatively rare. Among the 1,098 single nucleotide polymorphisms (SNP) identified in the eight core nucleotide excision repair genes, only a few are common nonsynonymous or regulatory SNPs that are potentially functional. We tested the hypothesis that seven selected common nonsynonymous and regulatory variants in the nucleotide excision repair core genes are associated with risk of squamous cell carcinoma of the head and neck (SCCHN) in a hospital-based, case-control study of 829 SCCHN cases and 854 cancer-free controls. Assuming a recessive genetic model, we found that only carriers of the XPC 499Val/Val genotype had a significantly increased SCCHN risk (adjusted odds ratio, 1.65; 95% confidence interval, 1.16-2.36). In analysis of the joint effects, the number of observed risk genotypes was associated with SCCHN risk in a dose-response manner (P for trend = 0.017) and those who carried four or more risk genotypes exhibited a borderline significant 1.23-fold increased SCCHN risk (adjusted odds ratio, 1.23; 95% confidence interval, 0.99-1.53). In the stratified analysis, the dichotomized combined effect of the seven SNPs was slightly more evident among older subjects, women, and laryngeal cancer. These findings suggest that these potentially functional SNPs may collectively contribute to susceptibility to SCCHN. These findings need to be validated in larger, independent studies. (Cancer Epidemiol Biomarkers Prev 2007;16(8):1633–8)

Squamous cell carcinoma of the head and neck (SCCHN; including cancers of the oral cavity, pharynx, and larynx) is relatively common worldwide (1). Although smoking and alcohol use play major roles in the etiology of SCCHN (2, 3), only a small fraction of cigarette smokers/alcohol users develops SCCHN, which suggests a differential susceptibility to the disease in the general population. Tobacco smoke contains many kinds of carcinogens that can cause DNA damage, and variations in repair of tobacco carcinogen–induced DNA damage may contribute to the variation in susceptibility to cancer (4). Nucleotide excision repair (NER) is one of the major repair pathways for removing DNA damage caused by tobacco carcinogens as well as other helix-distorting lesions that interfere with base pairing and obstruct replication and transcription (5, 6).

Several critical genes participate in the NER process and have functions central to the ability of the cell to cope with different types of DNA damage and to maintain genomic integrity (5). In NER, the XPC-HR23B complex detects a damage site in DNA and then recruits the transcription factor IIH (TFIIH, including two helicases: XPD and XPB) to open the DNA strands around the site of the lesion (7, 8). XPA, in conjunction with the single-strand DNA-binding protein replication protein A, constitutes a regulatory factor that monitors DNA binding and unwinding to verify the damage-specific localization of repair complexes or to assure the correct three-dimensional assembly (9). Subsequently, the open DNA complex creates the substrate for cleavage by two structure-specific endonucleases ERCC1-XPF (at 5′ of the lesion) and XPG (at 3′ of the lesion; refs. 10, 11).

Germ-line mutations in the core NER repair genes (i.e., XPA, XPB, XPC, XPD, XPE, XPF, and XPG) that severely alter their protein functions cause the xeroderma pigmentosum (XP) syndrome. XP fits a recessive genetic model, in which only mutant homozygotes manifest the disease phenotype (12). The defective DNA repair capacity (DRC) phenotype in XP represents the extreme low end of the repair spectrum that is associated with a >1,000-fold increased risk of sunlight-induced skin cancer (13). Moreover, there is an ∼5-fold variation in DRC in the general population (14, 15) that may result from the effects of genetic variants in the NER genes (16, 17). Previously, we reported that an intronic poly (AT) variant in XPC was associated with both the DRC phenotype (17) and SCCHN risk (18). We also reported a nonsignificantly increased SCCHN risk associated with the ERCC1 C8092A and XPD Asp³¹²Asn and Lys⁷⁵¹Gln polymorphisms in two small case-control studies (19, 20). It is known that cancer etiology is polygenic, and a single genetic variant is usually insufficient to predict risk of cancer that has a complex disease phenotype.

In an early pilot study, we showed that reduced DRC seemed to contribute to an individual's susceptibility to SCCHN (21). However, if the DRC phenotype is genetically determined, as shown in XP patients, it should reflect the genetic effects of all possible functional variants, most of which may be relatively rare. Indeed, among the 1,098 single nucleotide polymorphisms (SNP) identified to date in the eight core genes (i.e., ERCC1, XPA, XPB, XPC, XPD, XPE, XPF, and XPG) of the NER pathway (Table 1),⁴

there are a total of 40 nonsynonymous SNPs (nsSNP) but only 5 are confirmed as common (i.e., minor allele frequency > 0.05) nsSNPs [i.e., XPC Ala⁴⁹⁹Val (rs2228000) and Lys⁹³⁹Gln (rs2228001), XPD Asp³¹²Asn (rs1799793) and Lys⁷⁵¹Gln (rs13181), and XPG His¹¹⁰⁴Asp (rs17655)] in non-Hispanic whites. Other than these nsSNPs, two common regulatory SNPs located at the 3′-untranslated region of ERCC1 (C8092A, rs3212986) and 5′-untranslated region of XPA (G23A, rs1800975) were well documented and suggested a correlation with the DRC phenotype (22). Compared with other SNPs, these common, potentially functional SNPs are likely to collectively have an effect on the DRC phenotype in the general population. To assess the role of these seven common potentially functional SNPs in the etiology of SCCHN, we expanded our previous work to a large SCCHN study of 829 SCCHN cases and 854 cancer-free controls to further test the hypothesis that common, potentially functional SNPs in the NER core genes are associated with risk of SCCHN.

The recruitment of subjects for the ongoing SCCHN study has been described previously (23). Briefly, all patients had newly diagnosed, untreated SCCHN that was histopathologically confirmed at The University of Texas M. D. Anderson Cancer Center between May 1995 and March 2005. Patients with second SCCHN primary tumors, primary tumors of the nasopharynx or sinonasal tract, primary tumors outside the upper aerodigestive tract, cervical metastases of unknown origin, or any histopathologic diagnosis other than SCCHN were excluded. Because genotype frequencies can vary between ethnic groups and few minority patients were recruited, we included only non-Hispanic whites in this analysis. Of the eligible cases we approached for participation, the response rate was ∼93%. Consequently, this study included 829 non-Hispanic white subjects with primary tumors of the oral cavity (n = 253; 30.5%), pharynx (n = 424; 51.2%, including 386 oropharynx and 38 hypopharynx), or larynx (n = 152; 18.3%).

Cancer-free control subjects were recruited from hospital visitors, genetically unrelated to the enrolled case subjects or each other, who accompanied patients to the clinics but were not seeking medical care. We first surveyed potential control subjects at the clinics by using a short questionnaire to determine their willingness to participate in research studies and to obtain demographic information for frequency matching to the cases by age (±5 years) and sex. Of the eligible controls, the response rate was ∼85%. Having obtained informed consent, we interviewed each eligible subject to collect additional information about risk factors, such as tobacco smoking and alcohol use. Those who had smoked <100 cigarettes in their lifetime were considered “never smokers”; all others were considered “ever smokers.” Among ever smokers, those who had quit and had not smoked for >1 year were considered “former smokers” and the others were considered “current smokers.” Similarly, subjects who had drunk alcoholic beverages at least once weekly for >1 year were considered “ever drinkers” and all others were considered “nondrinkers.” Among ever drinkers, those who had quit drinking and had not had an alcoholic drink for >1 year were considered “former drinkers” and the others were considered “current drinkers.” Each subject provided 30 mL of blood for biomarker tests. The research protocol was approved by the M. D. Anderson Cancer Center institutional review board.

The primers, PCR annealing time, and restriction enzyme (New England Biolabs) conditions for XPA G23A (24), XPC Ala⁴⁹⁹Val and Lys⁹³⁹Gln (25), XPD Asp³¹²Asn (19) and Lys⁷⁵¹Gln (20), and XPG His¹¹⁰⁴Asp (26) have been described previously. For the ERCC1 C8092A variant (rs3212986), the primers were newly designed: forward, 5′-TACACAGGCTGCTGCTGCAGCT-3′ (22 bp; 16,307-16,328) and reverse, 5′-GCCAGAGACAGTGCCCCAAGAG-3 (22 bp; 16,402-16,423). These primers generated a PCR product of 117 bp and digested by PvuII (New England Biolabs) into 97 and 20 bp for CC; 117, 97, and 20 bp for CA; and the uncut 117 bp for AA. The genotyping assays for 10% of the samples were repeated, and the results were 100% concordant.

Differences in the selected demographic variables and smoking and drinking status between the cases and controls were evaluated by using the χ² test. The associations between genotypes of the selected polymorphisms and SCCHN risk were estimated by computing the odds ratios (OR) and 95% confidence intervals (95% CI) from both univariate and multivariate unconditional logistic regression analyses with adjustment for age, sex, smoking status, and drinking status. To assess the joint or interactive effects of two potentially interactive variables, a new indicator variable combining the two was created and used in tabulation and logistical regression modeling. A more-than-multiplicative interaction was suggested when OR₁₁ > OR₀₁ × OR₁₀, in which OR₁₁ = the OR when both factors were present, OR₀₁ = the OR when only factor 1 was present, and OR₁₀ = the OR when only factor 2 was present. To assess the combined effect of all genotypes, the number of “at-risk” genotypes was added, assuming the risk associated with each of these genotypes was simply additive. The significance was established at P < 0.05 with a two-side test. All data were analyzed by using Statistical Analysis System software (version 9.1.3; SAS Institute).

All case patients and control subjects were non-Hispanic whites adequately matched by age and sex; however, compared with the controls, cases had more smokers (current smokers: 34.6% versus 15.8%) and more drinkers (current drinkers: 50.9% versus 40.4%; P < 0.001 for both smoking status and drinking status; Table 2). To control for possible confounding effects of these variables, they were further adjusted in later multivariate logistic regression analyses of the main effects of the genotypes.

Table 3 shows the genotype distributions of the selected polymorphisms for cases and controls and their associations with risk of SCCHN. The genotype distribution of each nsSNP in the control subjects was consistent with those expected from the Hardy-Weinberg equilibrium (data not shown). We further tested the hypothesis that variant homozygous genotypes are associated with risk of SCCHN, assuming a recessive genetic model for the effect of a variant allele (i.e., only considering the variant homozygous genotype as the risk genotype). As shown in Table 3, only the XPC 499Val/Val genotype was associated with a significantly greater risk of SCCHN (adjusted OR, 1.65; 95% CI, 1.16-2.36) in the single locus analysis (i.e., Val/Val versus Ala/Ala + Ala/Val) with adjustment for age, sex, smoking, and alcohol use.

In the analyses of combined genotypes, we categorized all putative risk (ORs > 1.0) genotypes (assuming a recessive genetic model) from each SNP into a new variable according to the number of risk genotypes (for the protective genotype, we reversed the reference group). As shown in Table 4, the number of observed risk genotypes was associated with SCCHN risk in a dose-response manner (adjusted OR, 1.18; 95% CI, 0.88-1.57 for three risk genotypes; adjusted OR, 1.38; 95% CI, 0.99-1.93 for four risk genotypes; adjusted OR, 1.23; 95% CI, 0.77-1.97 for five risk genotypes; and adjusted OR, 4.10; 95% CI, 1.41-11.98 for six risk genotypes; P for trend = 0.017). When we dichotomized these combined effects, subjects carrying four to six risk genotypes exhibited a 1.23-fold increased risk for SCCHN (95% CI, 0.99-1.53). Further stratification analysis showed that this risk was slightly more evident among older subjects (adjusted OR, 1.42; 95% CI, 1.04-1.94), women (adjusted OR, 1.54; 95% CI, 0.97-2.44), former smokers (adjusted OR, 1.45; 1.02-2.05), and current drinkers (adjusted OR, 1.49; 1.06-2.10) and laryngeal cancer (adjusted OR, 1.49; 95% CI, 1.01-2.21; Table 5). Furthermore, there was no evidence of any interaction between the risk genotypes and these variables on risk of SCCHN (data not shown).

The present study investigated the associations between seven potentially functional SNPs of the five core NER genes (ERCC1, XPA, XPC, XPD, and XPG) and SCCHN risk. When we evaluated each polymorphism separately, a significant increased risk of SCCHN was only found to be associated with the XPC 499Val allele in a recessive model. In the combined analysis, the number of observed risk genotypes (assuming a recessive model) was associated with SCCHN risk in a dose-response manner and the dichotomized combined effect of these SNPs was borderline significantly associated with SCCHN risk. To the best of our knowledge, this is the largest case-control study of the association between these seven potentially functional NER SNPs and the risk of SCCHN.

It has been shown that the XPC-HR23B complex can interact with the transcription factor IIH (TFIIH) both in vivo and in vitro, and this interaction is essential for basal transcription and initiation of NER (8). All of the four nsSNPs in XPC and XPD have been well characterized in both functional and epidemiologic studies. In the analysis of correlation with the DNA repair phenotype, we have shown that the XPD 312Asn, 751Gln variant alleles were associated with a suboptimal DRC phenotype (16, 17), but the XPC Lys⁹³⁹Gln variant does not influence the DRC phenotype, as detected by using an allele-specific complementation assay (27), although when post-UV host cell reactivation was used, it altered base excision repair activity in radiation-specific DNA repair (28).

Several case-control studies have been reported in which the associations between the XPC Ala⁴⁹⁹Val polymorphism and risk of cancers were assessed (25, 29–31). In a study of the XPC Ala⁴⁹⁹Val polymorphism in 320 patients with lung cancer and 322 cancer-free controls in a Chinese population, the 499Val variant allele was found to be associated with an increased risk of lung cancer (25), but this was not confirmed in a Korean study of 432 patients with lung cancer and 432 healthy controls (29). Others found that the XPC Ala⁴⁹⁹Val variant had little effect on bladder cancer susceptibility (30) but contributed only to risk of advanced colorectal adenoma by modifying the effects of smoking (31). However, association studies have suggested that the XPC 939Gln allele is associated with risk of bladder cancer (32), breast cancer (33), melanoma (34), and colorectal adenoma (31). Such conflicting results have also been reported by others for lung cancer (25, 29), bladder cancer (30, 35), and melanoma (36).

Several case-control studies with different ethnic populations have also investigated the associations between XPD polymorphisms and cancer risk, particularly lung cancer; however, the results from these molecular epidemiologic studies are also conflicting (37). In two recently published meta-analyses, elevated risk of lung cancer associated with the variant alleles of XPD 312Asn and 751Gln was confined to the fixed combination model but not the random effect model (38, 39). As for the XPG His¹¹⁰⁴Asp variant, a significant protective effect of the variant 1104Asp/Asp genotype on risk of bladder cancer has been reported in a European Caucasian population (32). Such an effect has also been reported in a recent population-based case-control study of lung and squamous cell carcinomas of the oropharynx, larynx, and esophagus (40). In a recently published study on associations of DNA repair and cell cycle genes with cancers of the upper aerodigestive tract, the XPA G23A variant genotypes were suggested to have a protective effect on head and neck cancer (41); however, the results of other studies on lung cancer are mixed (42–44).

Taken together, previous studies and our current observations suggest that, compared with the intermediate phenotype such as DRC, risk of cancer as an end point is determined by many other unknown competing risk factors in addition to the variant genotypes under investigation. Therefore, a single genetic variant is insufficient to predict risk of the complex disease phenotype under a polygenic model that is often applied to cancer etiology. Apart from potentially functional SNPs, other SNPs in these genes, such as those in introns or promoters that may be regulatory in gene expression, should be comprehensively studied for their associations with cancer etiology. Therefore, functional studies and consideration of linkage disequilibrium in the selection of candidate SNPs are necessary to cover those untyped SNPs, particularly those that are rare but truly disease-causing SNPs.

Strengths of this study are the inclusion of all known common potentially functional SNPs in the NER core genes and a relatively large number of patients with SCCHN, a cancer that is rare compared with lung cancer. This study has inherent biases that may have led to spurious findings. Because this was a hospital-based study, our control subjects may not be representative of the general population. However, we recruited a relative large study population, applied a rigorous epidemiologic design in selecting the study subjects, and used further statistical adjustments to minimize potential biases. Because we had missing data on pack-years smoked from a substantial portion of the controls, we did not do rigorous analyses of gene-environment interactions. Although the low penetrance of these common nsSNPs in SCCHN susceptibility needs further validation with larger, population-based studies with additional SNPs representing a much greater coverage of the genetic variation in the eight core NER genes, the findings in this study suggest that these SNPs may be biomarkers of susceptibility to SCCHN. The use of these biomarkers may be further validated by collaborative data pooling efforts, such as the International Head and Neck Cancer Epidemiology Consortium (INHANCE).

We thank Margaret Lung, Kathryn Patterson, and Leanel Fairly for their assistance in recruiting the subjects; Yawei Qiao for technical assistance; Jianzhong He and Kejin Xu for their laboratory assistance; Joanne Sider for manuscript preparation; and Susan Eastwood for scientific editing.