Mutagen Sensitivity and Genetic Variants in Nucleotide Excision Repair Pathway: Genotype-Phenotype Correlation

Disruption of genomic integrity contributes to malignant transformation and subsequent cancer development. The repair of DNA damage plays a key role in maintaining genomic integrity. Four major DNA repair pathways exist in mammalian cells: base excision repair, nucleotide excision repair (NER), double-strand break repair, and mismatch repair (1). The NER pathway recognizes and repairs a variety of bulky DNA adducts, UV-induced pyrimidine dimers, cross-links, and oxidative damage. The NER pathway involves sequential reaction steps of, including DNA damage recognition, incision of damaged DNA, repair of the gapped DNA, and DNA ligation (2). DNA damage is first recognized by the RAD23B-XPC complex followed by the binding of XPA/RPA to the lesion. Then, the transcription factor IIH complex, including XPD and XPB (3), is recruited. Next, XPD and XPB unwind the DNA helix around the damaged site. The recruitment of the XPG and XPF-ERCC1 complex is followed by the excision of a 24- to 32-bp segment containing the bulky adduct at the 3′ and 5′ ends of the damaged site. The resultant gap is filled by DNA synthesis and ligation (4). ERCC6 is thought to participate in the NER of oxidative DNA damage by forming complexes with RNA polymerase 1, XPG, and transcription factor IIH (5).

Most NER genes are polymorphic, and many have been studied in terms of their associations with risk of various cancers (6). The rationale behind these gene-cancer risk associations is that these genetic variants may result in alterations in phenotypes (i.e., DNA repair capacity). However, genotype-phenotype correlations have not been explored for the majority of these polymorphisms. Mutagen sensitivity, measured by quantifying the chromatid breaks induced by a mutagen challenge in short-term cultures of peripheral blood lymphocytes, has been used as an indirect measure of DNA repair capacity. Numerous studies have shown that mutagen sensitivity is a risk factor for a variety of cancers (7-12). Moreover, using a classic twin study design, we recently showed that mutagen sensitivity has high heritability (13), suggesting strong genetic determinants of mutagen sensitivity. We hypothesized that if the genetic polymorphisms in NER genes have functional significance, then there may be a correlation between genotypes and mutagen sensitivity, which serves as an intermediate phenotype for cancer risk. We tested this hypothesis by correlating NER genotypes with benzo(a)pyrene diol epoxide (BPDE)-induced mutagen sensitivity. BPDE is a tobacco carcinogen that forms bulky adducts with DNA and elicits NER activity. In this analysis, we used data available from a twin study to correlate mutagen sensitivity and genetic polymorphisms in NER genes.

The current study was carried out using available blood samples collected in a twin study. The parent twin study had the goal of identifying genetic component of mutagen sensitivity, including BPDE sensitivity (13). The procedures for twin recruitment have been described in full detail elsewhere (14, 15). Briefly, potential twins were identified from the SRI Northern California Twin Registry, initiated in 1995. Twin pairs in the registry are between 18 and 65 years of age. Twins were recruited by the Center for Health Sciences, SRI International. Verbal consent was obtained by initial telephone contact with the potential participants. Following verbal consent, study procedures and protocols were administered to participants. The questionnaire was sent in the mail to all participating twin pairs. The questionnaire assessed zygosity, demographics (age, gender, ethnicity, education, marital status, etc.), tobacco and alcohol use, and family history of cancer. Self-reported zygosity was further confirmed in the lab by comparing members of each twin pair on 10 to 12 highly polymorphic microsatellite markers (16). If the completed questionnaire was not returned within 3 weeks of mailing, the subject was sent a second one to complete and return. All methods for recruitment, informed consent, screening, and data collection were reviewed and approved by the Institutional Review Boards of SRI International and the University of California at San Francisco. Siblings of a subset of twins were also recruited. A total of 422 individuals were available for this analysis. This included 138 pairs of monozygotic twins, 51 pairs of dizygotic twins, and 44 siblings.

For each participant, two 10-mL tubes of blood samples were collected at the Clinical Research Center at The University of California at San Francisco at baseline and sent to the Department of Epidemiology, The University of Texas M. D. Anderson Cancer Center by Federal Express mail within 24 h of blood drawing for lab assays. The mutagen sensitivity assay has been described previously (17). Briefly, 1-mL fresh whole blood was cultured in 9 mL of RPMI 1640 tissue culture medium with 10% FCS and 1% phytohemagglutinin at 37°C for 72 h. Then, 2 μmol/L BPDE was added to each culture and incubated for 24 h. After blocking with colcemid, the cells were fixed, washed, and air dried on clean slides. Slides of each culture were stained with 4% Giemsa to visualize the chromosomes. The number of breaks in 50 metaphases per sample was counted and expressed as the average number of breaks per cell (b/c). Each simple chromatid break was scored as one break, whereas each isochromatid break set and each exchange event (including interstitial deletion) were considered as two breaks.

Genomic DNA was extracted from peripheral blood by proteinase K digestion followed by isopropanol extraction and ethanol precipitation. Except for XPA A23G, XPD Asp321Asn, and XPC-PAT polymorphism, which were genotyped using PCR-restriction fragment length polymorphism (18), genotyping was done using the Taqman method with a 7900 HT sequence detector system (Applied Biosystems) for the following NER single nucleotide polymorphisms (SNP): ERCC1 (G/T at 3′-untranslated region, rs3212986), XPD Lys751Gln, XPG Asp1104His, XPC Lys939Gln, XPC Ala499Val, RAD23B Ala249Val, CCNH Val270Ala, ERCC6 Met1097Val, and ERCC6 Arg1230Pro. Amplification mixes (5 μL) contained sample DNA (5 ng), 1× Taqman buffer A, deoxynucleotide triphosphates (200 μmol/L), MgCl₂ (5 mmol/L), AmpliTaq Gold (0.65 units), each primer (900 nmol/L), and 200 nmol/L of each probe. The thermal cycling conditions consisted of 1 cycle for 10 min at 95°C, 40 cycles for 15 s at 95°C, and 40 cycles for 1 min at 60°C. SDS version 2.1 software (Applied Biosystems) was used to analyze end point fluorescence according to the allelic discrimination technique. All these SNPs selected have been reported in the literature to have potential functional significance. In our laboratory, strict quality control procedures are implemented to ensure high genotyping accuracy. Water control, internal control, and previously genotyped samples were included in each plate to ensure accuracy of the genotyping. Five percent of the samples are randomly selected and run in duplicates with 100% concordance. The call rates were >95% for all the evaluated polymorphisms.

Descriptive statistics (mean, median, range, and SD) of mutagen sensitivity were computed for all the participants and for subgroups by age, gender, ethnicity, and smoking status. Hardy-Weinberg equilibrium was tested by a goodness-of-fit χ² test. Mutagen sensitivity by genotype and genotype combinations were compared by Student's t test or ANOVA. Significance of pair-wise comparisons was tested using ANOVA followed by Bonferroni adjustment for multiple comparisons with family-wise significance of 0.05. The effect of genotype combination was evaluated by summing the number of adverse alleles across a panel of genotypes, and comparison was made between mutagen sensitivity as determined by the number of adverse alleles. Mutagen sensitivity was also dichotomized at the median to calculate odds ratios (OR) associated with each genotype or genotype combinations: b/c values greater than the median were considered mutagen sensitive, whereas values less than the median were considered as not sensitive. Other cutoff points, such as 75% percentile, were also used to dichotomize the data. The results were similar to what we reported based on the median dichotomization. In all analyses, the generalized estimating equations were used to account for dependence among family members (19, 20). In our analyses, twins and siblings within the same family are not independent observations; thus, traditional ANOVA should be corrected by generalized estimating equations to account for data dependency. The way generalized estimating equations works to account for dependence is through specification of the covariance structure (or within-group correlation structure) of the repeated measurements, in this case, within-group correlation structure of measurements from members within the same family. We applied the exchangeable structure, one of the several covariance structures of generalized estimating equations, in which the correlation between any two members (e.g., i and j) of the same family (e.g., k) is some constant value, that is, Corr (Y_ki, Y_kj) = ρ, for family k, if i is not equal to j, and (Y_ki, Y_kj) = 1, for family k, if i equals to j. Note that the exchangeable correlation structure has been applied also to pedigree data with general phenotype as well.

Linkage disequilibrium index (D′) and haplotype and diplotype frequencies were analyzed using the HelixTree Genetics Analysis Software (version 4.1.0; Golden Helix). All other analyses were done with STATA version 8 (STATA Corp.). All analyses were two sided with a significance level of 0.05.

Data from 422 individuals were available for analyses. Among them, there were 138 pairs of monozygotic twins, 51 pairs of dizygotic twins, and 44 siblings of the twins. There were 324 (76.8%) Caucasians, 12 (2.8%) African-Americans, 49 (11.6%) Mexican-Americans, 16 Asians (3.8%), and 21 (5.0%) of other ethnicities. The majority of participants were never smokers (257, 60.9%). The number of former and current smokers was 93 (22.0%) and 71 (16.8%), respectively (Table 1).

Because twins and siblings from same families shared 100% (in case of monozygotic twins) or 50% (in case of dizygotic twins or twins and siblings) of their genes, their genotypes were not independent. Thus, we tested the Hardy-Weinberg equilibrium by randomly selecting one individual from each family to ensure the independence of genotype in the test population. The genotype distributions of all the SNPs agreed with Hardy-Weinberg equilibrium in all ethnicities (all P > 0.05; data not shown). Consistent with the literature, the three XPC loci were in strong linkage disequilibrium (XPC Ala499Val versus XPC Lys939Gln: D′ = 0.863, P < 0.001; XPC Ala499Val versus XPC-PAT: D′ = 0.833, P < 0.001; XPC Lys939Gln versus XPC-PAT: D′ = 0.968, P < 0.001), so were the two ERCC6 loci (D′ = 0.99, P < 0.001) and the two XPD loci (D′ = 0.76, P < 0.001).

BPDE sensitivity was not affected by ethnicity, smoking status, sex, or age (data not shown). Table 2 summarizes BPDE-induced b/c by genotypes and genotype combinations. We observed several significant genotype-phenotype correlations in individual analysis. Specifically, for XPA (A23G at 5′-untranslated region) SNP, compared with the AA genotype, BPDE-induced b/c decreased with the increasing copy of the G allele. Compared with individuals with the AA genotype, those with at least one G allele exhibited significantly lower b/c (0.66 versus 0.80; P = 0.045). Individuals with the homozygous XPC-PAT−/− genotype (0.63) and the heterozygous XPC-PAT+/− genotype (0.67) exhibited lower BPDE-induced b/c than individuals with the variant (XPC-PAT+/+) genotype (0.79). The difference was significant between the +/+ and the −/− genotypes (0.79 versus 0.63, P = 0.004; Table 2), and there was a significant increasing trend in b/c with the increasing copy of the variant allele (P for trend = 0.005). The XPC Lys939Gln wild-type (AA), heterozygous variant (AC), and homozygous variant (CC) had b/c values of 0.70, 0.63, and 0.79, respectively. Compared with the AA and AC combined, BPDE sensitivity was significantly higher in the CC genotype (0.66 versus 0.79, P = 0.01; Table 2). The homozygous variant RAD23B TT genotype exhibited significantly higher BPDE sensitivity than the wild-type CC genotype (1.00 versus 0.65, P = 0.007), and the trend of increasing b/c with increasing copy of variant allele was borderline significant (P for tend = 0.08). There was no significant genotype-phenotype correlation for other SNPs involved in the same NER pathway (Table 2). Except for the XPA SNP, the significant genotype-phenotype correlations described above were statistically significant after adjustment for multiple comparisons (Table 2).

Because significant genotype-phenotype correlation was observed in four SNPs, XPC-PAT, XPC Lys939Gln, XPA A23G, and RAD23B Val249Ala, we next assessed mutagen sensitivity as modified by the combined effects of these four SNPs. Because the A allele of XPA, the + allele of XPC-PAT, the C allele of XPC Lys939Gln, and the T allele of RAD23B Ala249Val conferred higher BPDE sensitivity in single SNP analysis (Table 2), these alleles were defined as adverse alleles. We observed that the b/c values for those carrying zero to two, three to five, and six or more adverse alleles were 0.64, 0.68, and 1.06, respectively (P for trend = 0.008; Table 3). Compared with subjects carrying zero to two adverse alleles, subjects carrying six or more adverse alleles exhibited significantly elevated BPDE sensitivity (1.06 versus 0.64, P < 0.001) and the result remained significant after adjustment for multiple comparisons (Table 3).

The risk of increased BPDE-induced mutagen sensitivity also estimated. BPDE sensitivity was dichotomized at the median b/c value: greater than the median b/c value was considered mutagen sensitive, whereas less than the median was considered as not sensitive. Compared with individuals with the XPC-PAT−/− genotypes, the risks of being BPDE sensitive for those with the XPC-PAT−/+ and XPC-PAT+/+ genotypes were 1.03 [95% confidence interval (95% CI), 0.65-1.61] and 1.92 (95% CI, 1.00-3.67), respectively (data not shown). Risk of BPDE sensitivity did not reach statistical significance in the analysis of other single SNPs. However, using individuals carrying zero to two variant alleles as the reference group, we observed that the risk of increased BPDE sensitivity for those carrying three to five adverse alleles was 1.05 (95% CI, 0.68-1.64) and elevated to 4.48-fold (95% CI, 1.21-16.61) in subjects carrying six or more adverse alleles (P for trend = 0.008; Table 3).

We also explored the joint effects of genotype combinations on mutagen sensitivity. Because there is no compelling evidence in the literature on whether a particular genotype is a risk type, we determined whether the genotype is adverse or favorable based on our own BPDE sensitivity data in Table 2. The genotypes that conferred significantly or borderline significantly higher BPDE sensitivity were classified as adverse genotypes. Specifically, the adverse genotypes were the AA genotype of XPA, the PAT+/+ and PAT−/+ genotypes of XPC intron 9 polymorphism, the CC genotype of XPC Lys939Gln, and the CT and TT genotypes of RAD23B Ala249Val. Individuals were grouped into three categories: those containing zero to one adverse genotypes, those containing two adverse genotypes, and those carrying three or more adverse genotypes. BPDE-induced b/c values were 0.64, 0.70, and 1.00 for the three categories, respectively (P for trend < 0.001). The results remained significant after adjusting for multiple comparisons. In risk analysis, compared with individuals carrying zero to one adverse genotypes, those carrying two adverse genotypes showed a 1.59-fold (95% CI, 0.96-2.64) increased risk of being BPDE sensitive, whereas the risk increased to significant 3.28-fold (95% CI, 1.20-8.95) for those carrying three or more adverse genotypes (P for trend = 0.006; Table 3).

Finally, we explored BPDE sensitivity in association with haplotype and diplotypes of XPC, XPD, and ERCC6. The four common haplotypes (accounted for 99.1% of all haplotypes) based on the three XPC polymorphisms are C-A, C+C, T-A, and C-C. The most adverse haplotype, C+C, contained all three adverse alleles (the C allele of the Ala499Val, the PAT+ allele, and the C allele of the Lys939Gln) and the haplotypes C-C and C-A were intermediate, containing two and one adverse alleles, respectively. The haplotype T-A was the most favorable type containing all three favorable alleles. BPDE sensitivity by XPC haplotype is shown in Table 4. Consistent with genotype analysis, using the haplotype containing all three adverse alleles, C+C, as the reference group, the two intermediate haplotypes (C-A and C-C) and the favorable haplotype (T-A) exhibited lower BPDE sensitivity. However, the difference reached significance in the intermediate C-C haplotype (0.73 versus 0.49, P = 0.002; Table 4). The C-C haplotype was also associated with a significantly reduced risk of being BPDE sensitive with an adjusted OR of 0.48 (95% CI, 0.25-0.92; Table 4).

BPDE sensitivity by XPC diplotypes is shown in Table 5. The diplotypes were listed in order according to the number of adverse, intermediate, and favorable haplotypes contained. Compared with the diplotype C+C/C+C, which contains two adverse haplotypes, all other diplotypes exhibited lower BPDE sensitivity (Table 5). However, the difference was significant in the T-A/C-C diplotype (0.82 versus 0.43, P = 0.004), which contained a favorable haplotype and an intermediate haplotype, and in the C-A/C-C diplotype (0.82 versus 0.53, P = 0.02), which is composed of two intermediate haplotypes. The T-A/C-C diplotype also conferred a significantly decreased risk of being BPDE sensitive with an adjusted OR of 0.18 (95% CI, 0.05-0.60; Table 5). BPDE sensitivity was not significantly modified by either the XPD or the ERCC6 haplotypes/diplotypes (data not shown).

In this study, we showed that polymorphisms in XPC, RAD23B, and XPA genes modify mutagen sensitivity, individually and jointly. BPDE interacts with DNA and forms bulky adducts that require NER pathway. The genotype-phenotype correlation observed in this study supports the notion that variations in NER genes may modulate DNA repair capacity and thus contribute to cancer risk in the general population. We previously did genetic analyses on mutagen sensitivity using the twin data (13). Our results showed that mutagen sensitivity, including BPDE sensitivity, has high heritability, strongly supporting a genetic component of mutagen sensitivity (13).

In the current genotype-phenotype correlation study, individuals with the XPC-PAT homozygous variant +/+ genotype exhibited significantly higher numbers of b/c compared with the wild-type genotype and had a 1.92-fold increased risk of being BPDE sensitive. In a previous genotype-phenotype correlation study, the XPC-PAT+/+ genotype was found to have suboptimal DNA repair capacity as measured by the host cell reactivation assay (21). The agreement between the two genotype-phenotype correlation studies suggests that this polymorphism may have functional significance that affects host DNA repair capacity. Indeed, both the + allele of XPC-PAT and the variant C allele of the XPC 939Gln/Gln have been reported to be associated with increased cancer risk. For example, Casson et al. (22) found the XPC-PAT homozygous variant genotype (+/+) was associated with a 3.82-fold (95% CI, 1.05-13.93) increased risk of esophageal adenocarcinoma. Shen et al. (23) reported that XPC-PAT+/− genotype and XPC-PAT+/+ genotype conferred a 1.44-fold (95% CI, 1.01-2.05) and 1.85-fold (95% CI, 1.12-3.05) increased risk of developing squamous cell carcinoma of head and neck, respectively. In a study of cutaneous melanoma, Blankenberg et al. (24) reported that XPC-PAT homozygous +/+ genotype was associated with an increased risk of melanoma (OR, 1.87; 95% CI, 1.10-3.19). Sanyal et al. (25) observed a significantly increased risk of bladder cancer for subjects carrying homozygous variant XPC 939Gln/Gln genotype (OR, 1.97; 95% CI, 1.10-3.57). However, there were also null reports of XPC polymorphisms and cancer risk (26).

We have previously reported that the presence of one or two copies of the G allele of the XPA A23G SNP was associated with a reduced lung cancer risk in Caucasians and in Mexican-Americans (18). Similarly, in a Korean study, the homozygous GG genotype was associated with a significant 44% reduction in risk for lung cancer (27). We previously used a host cell reactivation assay to quantify DNA repair capacity and observed significantly more efficient DNA repair capacity with at least one copy of the G allele compared with those with the AA genotype (9.53 versus 8.29, P = 0.03) in control subjects (18). In this current study, carriers of one or two copies of the G allele exhibited lower mutagen sensitivity compared with the AA genotype (0.66 versus 0.80, P = 0.045), consistent with a higher DNA repair capacity conferred by the GG genotype. Therefore, two different phenotypic assays in different populations suggest that the presence of a G allele results in a higher DNA repair capacity, which may further confer a reduced cancer risk. However, the result should be interpreted with caution given that the difference in mutagen sensitivity is only borderline significant.

A more important observation in this study is the significant correlation between increasing number of adverse alleles and increasing mutagen sensitivity. Indeed, the correlation between mutagen sensitivity and any individual SNP was modest. For example, the risk of being mutagen sensitive was not significant when considering XPC Lys939Gln, RAD23B, or XPA A23G, separately. However, when genotypes were combined and the effects were assessed by total number of adverse alleles across a panel of genes, a consistent trend emerged. That is, there is an increasing trend in mutagen sensitivity with increasing number of adverse alleles. Matullo et al. (28) reported a dose-response relationship between the number of adverse alleles in DNA repair SNPs and levels of DNA adducts and found that individuals with at least three variant alleles had a significant increased risk for having high levels of adducts. We recently used two functional assays (mutagen sensitivity assay and Comet assay) to determine genotype-phenotype correlations in a study that used a pathway-based multigenic approach to assess genetic predisposition to bladder cancer (29). A significant dose-response trend was observed between mutagen sensitivity and low-, medium-, and high-risk groups as defined by a comprehensive panel of DNA repair and cell cycle SNPs (29). These observations underscore the importance of assessing combined effects of a panel of polymorphisms that act in the same pathway in association studies.

In haplotype analysis, we found that the haplotype C+C of XPC exhibited the highest mutagen sensitivity (b/c = 0.73). According to the genotype-phenotype correlation presented in Table 2, the C+C haplotype contains all three adverse alleles: the C allele of the Ala499Val, the PAT+ allele, and the C allele of the Lys939Gln. Thus, the results obtained from the haplotype analysis and genotype analysis are consistent. We also noted that the C-C haplotype exhibited the lowest BPDE sensitivity, suggesting proficient DNA repair capacity of this haplotype. Consistent with this finding, we recently reported that the C-C haplotype conferred a significantly reduced risk of bladder cancer (30). In diplotype analysis, BPDE sensitivity was lowest for the diplotype T-A/C-C, which contains an intermediate haplotype, C-C, and a favorable haplotype, T-A. The T-A haplotype, which includes all favorable alleles, may confer reduced mutagen sensitivity and higher DNA repair capacity. A significantly decreased risk of endometrial cancer associated with the variant T allele of the XPC Ala499Val and wild-type A allele of the XPC Lys939Gln was reported (31). The variant PAT+ allele and the C allele of the Lys939Gln were found to be associated with suboptimal DNA repair capacity in two previous genotype-phenotype correlation studies (21, 32). The intermediate haplotype C-C, as shown in Table 5, conferred the lowest BPDE sensitivity and is associated with reduced risk of bladder cancer (30).

A limitation of this study is that only selected major genes in the NER pathway are included and only the potential functioning SNPs identified in these genes are included. In conclusion, our data suggest that individual polymorphisms in XPC and XPA genes may modulate mutagen sensitivity and that specific haplotypes/diplotypes of XPC may determine favorable mutagen sensitivity status, supporting the protective role of these haplotypes/diplotypes reported in previous association studies of cancer. The most important finding of this study is the correlation between mutagen sensitivity phenotype and the combination of multiple SNPs in the same DNA repair pathway. Recent advances in association studies have offered strong evidence of the advantages of taking a multigenic approach in studying complex diseases, such as cancer (29). Many studies have taken the initiative and generated promising results in an attempt to unveil the association between multiple genetic combinations and cancer risks, which would otherwise be undetectable in single SNP analysis. As risk assessment for cancer risk is moving beyond analysis of single polymorphisms, future genotype-phenotype correlation studies should also investigate the combined effects of multiple genetic variants.

We thank Mary R. McElroy, Ruth E. Krasnow, and Jill Rubin for their efforts with subject recruitment and data collection.