Polymorphism in the DNA Repair Gene XPD, Polycyclic Aromatic Hydrocarbon-DNA Adducts, Cigarette Smoking, and Breast Cancer Risk

DNA repair systems are essential to the body's response to endogenous and exogenous carcinogens and mutagens. Functioning DNA repair systems are important to reduce the risk of all cancers (1), including breast cancer. Women with suboptimal DNA repair have been reported to be at increased risk of breast cancer by as much as 5-fold (2, 3). Genetic differences in DNA repair genes can lead to differences in cancer susceptibility and may modify the impact of environmental exposures on cancer risk.

We undertook a study of gene-environment interactions involving nucleotide excision repair, one pathway of DNA repair involved in the removal of bulky DNA adducts. There are several important genes in this category, including XPD (also called ERCC2). There are multiple single nucleotide polymorphisms in XPD including two in exon 8, one in exon 10, and one at position 751 of exon 23, all of which result in amino acid changes. The first two are quite rare (∼0.04%; ref. 4) in most populations, the first three are also nonconservative replacements, whereas the one at position 751 of exon 23 (a Lys/Gln substitution) is a conservative replacement (4).

For this study, we limited our assessment of interactions to one environmental exposure and one environmental exposure marker that may affect bulky adduct formation: cigarette smoking and a direct measure of polycyclic aromatic hydrocarbon (PAH)-DNA adducts, respectively. The Long Island Breast Cancer Study Project was undertaken specifically to investigate the role of PAH-DNA adducts in breast cancer and recently reported an overall association of 1.32 [95% confidence interval (95% CI), 1.00-1.74] for detectable versus nondetectable adducts (5). We examined exon 23 of XPD to characterize subjects into the following genotype categories: Lys/Lys, Lys/Gln, and Gln/Gln. The Gln allele has been associated with suboptimal DNA repair (6-8). We hypothesized that those with suboptimal DNA repair, as characterized by genotype, would face a greater risk from PAH-DNA adducts and cigarette smoking than those with efficient repair mechanisms.

Subjects of the study are from a population-based case-control study conducted on Long Island, NY. Breast cancer cases were composed of women who were residents of Nassau and Suffolk counties, spoke English, and were newly diagnosed with in situ or invasive breast cancer between August 1, 1996 and July 31, 1997. There were no age or race restrictions. For full details of case ascertainment, see the description of the parent study (5, 9). Population-based controls were identified by random digit dialing for those ages <65 years and Health Care Financing Administration rosters for those ages ≥65 years. In-person interviews were completed for 82.1% of cases (n = 1,508) and 62.8% of controls (n = 1,556). Of those who completed an interview, 73.1% of cases and 73.3% of controls donated a blood sample. Of those who donated a blood sample, we were unable to genotype 4.4% of cases and 3.4% of controls mainly due to lack of sufficient DNA to complete the assay. Thus, our final sample size was 1,053 cases and 1,102 controls.

Those with available data on XPD genotype were more likely to be younger, White, and alcohol consumers than those subjects without available XPD genotype data (data not shown). These same factors were reported to differ between those who donated blood and those who did not for the overall study sample (9). There were no differences in XPD data availability by case-control status. Those with data on PAH-DNA adducts did not differ from those without PAH-DNA adducts in the distribution of XPD genotypes (data not shown).

DNA was isolated from blood cells and genotyped with the use of template-directed primer extension with detection of incorporated nucleotides by fluorescence polarization in a 96-microwell format, essentially as described (10). Master DNA 96-well plates containing 10 ng/μL were used to make replica plates containing 25 ng DNA per well. For PCR amplification, the primers (forward 5′-CCCTCTCCCTTTCCTCTGTT-3′ and reverse 5′-GGCAAGACTCAGGAGTCACC-3′) gave a 171-bp product. Conditions for amplification were 0.2 μL (8 pmol/μL) forward and reverse primers, 0.4 μL of 25 mmol/L MgCl₂, 1 μL of 10× PCR buffer, 0.1 μL (5 units/mL) Taq polymerase (Roche Molecular Biochemicals, Indianapolis, IN), 0.25 μL (10 mmol/L) deoxynucleotide triphosphates (Roche Molecular Biochemicals), and 5.35 μL water. Denaturation at 94°C for 5.5 minutes was followed by 34 cycles of 94°C for 30 seconds, 60°C for 45 seconds, and 72°C for 1 minute followed by 4 minutes at 72°C. Primers and deoxynucleotide triphosphates were digested with 1 unit of shrimp alkaline phosphatase (1 unit/μL, Roche Molecular Biochemicals) after the addition of 1 μL of 10× buffer and 1 unit Escherichia coli exonuclease I (10 units/μL, U.S. Biochemical, Cleveland, OH) and 7.9 μL of water for 45 minutes at 37°C followed by heating at 95°C for 15 minutes. The reverse extension primer was 3′-CTGAGCAATCTGCTCTATCCTCT-5′. Acycloprime FP SNP Detection kit G/T contained the dideoxynucleotide triphosphates labeled either with R110 or TAMRA (Perkin-Elmer Life Sciences, Boston, MA). To 7 μL of reaction mixture were added 0.05 μL acycloprimer enzyme, 1 μL G/T Terminator mix, 2 μL of 10× reaction buffer, 0.5 μL extension primer (10 pmol/μL), and 9.45 μL water. Extension was carried out by heating at 95°C for 2 minutes followed by 30 cycles of 95°C for 15 seconds and 55°C for 30 seconds. Plates were read on a Perkin-Elmer Victor instrument. In addition to assay-specific quality control samples, 10% of samples were reassayed after relabeling to keep laboratory personnel blinded to identity. Of the 115 duplicate samples, 97.3% were concordant for genotype result (three pairs were discordant). For these three discordant pairs, we used the genotyping result from the first run, but the overall analyses did not differ by whether these pairs were included or excluded from the analyses (data not shown).

Exposure information comes from two sources: the parent study questionnaire that was given by trained interviewers in the subject's home (http://epi.grants.cancer.gov/LIBCSP/projects/Questionnaire. html) and laboratory analyses using blood samples to measure PAH-DNA adducts (5). Respondents were asked about their pregnancy, occupational, and residential history; their use of pesticides in their home or on a farm; electrical appliance use; lifetime history of consumption of smoked or grilled foods; medical history; family history of cancer; body size changes by decade; recreational physical activities; cigarette smoking; alcohol use; menstrual history; use of exogenous hormones; and demographic characteristics (9).

We limited our assessment of interactions to an environmental exposure and exposure marker that may affect bulky adduct formation: cigarette smoking and a direct measure of PAH-DNA adducts, respectively. Cigarette smoking was assessed by questionnaire (11). In addition to never/former/current cigarette smoking status, we examined interactions with duration of cigarette smoking (<10, 11-20, and >20 years) as well as with passive and active cigarette smoking. PAH-DNA adduct levels were assessed using a competitive ELISA with a polyclonal antiserum generated against benzo(a) pyrene diol epoxide modified DNA but which recognized the diol epoxide adducts of several other PAHs, as described previously (12). We analyzed the data on PAH-DNA adducts by first assessing differences between those with detectable versus nondetectable adducts (26). Among those with detectable adducts, we categorized subjects into two categories based on the median value (26).

We first compared differences between genotypes in each exposure category using χ² test for categorical variables and ANOVA for continuous variables. Logistic regression was used to estimate odds ratio (OR) and 95% CI adjusting for potential confounding variables (13). We examined potential confounding by the following variables: age, menopausal status, family history of breast cancer, race, religion, and education. Confounders were included in the final model if their inclusion changed the estimate on exposure by >10%. Effect modification was first examined through use of stratified analysis, running separate models for each exposure category, and by including multiplicative interaction terms in the logistic regression model (13). We also further evaluated interaction (both additive and multiplicative) by using indicator terms for those with the genotype only, exposure only, and those with both genotype and exposure of interest (14). If the relative risk, as approximated by OR, for both genotype and exposure was greater than the relative risk of either factor alone added together minus 1, we concluded that there was positive additive interaction.

Genotype frequencies for the exon 23 polymorphism of XPD are reported in Table 1. There were no statistically significant differences between breast cancer cases and controls by allele frequency (Lys: 61.1% for cases and 63.7 for controls; Gln: 38.9% for cases and 36.3% for controls; P = 0.7). The genotypes are in Hardy-Weinberg equilibrium (χ² = 0.61 with 1 df). Presence of at least one variant allele (Lys/Gln or Gln/Gln) was associated with a 20% increase in risk of breast cancer (OR, 1.21; 95% CI, 1.01-1.44). Restricting these analyses to the subset of subjects with PAH-DNA adduct data did not change these results (data not shown).

Table 2 summarizes associations between menopausal status, first-degree family history of breast cancer, and race and XPD status. Cases were older than controls in each genotype category, but there was no association between genotype frequencies and age in either the cases or the controls (data not shown). The relationship between genotype and case-control status varied little by menopausal status or family history. There were very few non-Whites to evaluate, but among these the percentage homozygous for the variant allele Gln/Gln was much lower than among Whites (5.1% versus 15% for cases and 4.8% versus 14% for controls for non-Whites versus Whites, respectively). Genotype frequencies were unrelated to the level of PAH-DNA adducts in the control group (data not shown). Genotype frequencies were also unrelated to the cigarette smoking in the control group (data not shown). Furthermore, there was no association between PAH-DNA adducts in the control group and genotype frequency separately stratified by smoking status (data not shown).

Multivariate-adjusted estimates stratified by exposure category are presented in Tables 3 and 4. For these analyses, we have kept the heterozygotes distinct from the homozygotes, because we did not have any a priori reason to treat them as one group. The increase in risk for homozygosity of the variant allele (Gln/Gln) seemed limited to those with PAH-DNA adducts levels above the median for the detectables (OR, 1.61; 95% CI, 0.99-2.63 for adducts above the median; OR, 1.05; 95% CI, 0.64-1.74 for adducts below the median), although the multiplicative interaction term was not statistically significant (P = 0.48; Table 3). We further examined whether the association was restricted to those in the highest quartile of PAH-DNA adduct levels. Associations were only significant for the fourth quartile: OR, 1.51; 95% CI, 0.93-2.47 for those with AC genotype and OR, 2.35; 95% CI, 1.17-4.72 for those with the CC genotype. There was no association for the first, second, or third quartile among those with detectable PAH-DNA adduct levels.

The increase in risk for homozygosity of the variant allele (Gln/Gln) was only seen among current smokers (OR, 1.97; 95% CI, 1.02-3.81 for current smokers; OR, 1.16; 95% CI, 0.75-1.81 for former smokers; and OR, 0.87; 95%CI, 0.57-1.32 for never smokers); the multiplicative interaction was statistically significant (P = 0.05; Table 4). There were no clear patterns with duration of smoking, but the numbers were very small to formally evaluate interactions (data not shown). The association between XPD genotype and breast cancer risk seemed limited to those subjects who were exposed to active cigarette smoke (OR, 3.44; 95% CI, 1.25-9.48 for homozygosity of the variant allele and exposure to active smoke versus OR, 1.01; 95% CI, 0.62-1.66 for homozygosity of the variant allele and exposure to passive smoke; data not shown).

We also assessed whether there was additive interaction between genotype and exposure by creating indicator variables and using a common reference group (data not shown). The joint effect of both Gln/Gln genotype and adducts above the median was 1.9 (OR, 1.90; 95% CI, 1.15-3.15) versus those with nondetectable adducts and the Lys/Lys genotype. The separate effects of Gln/Gln genotype and adducts above the median relative to those with nondetectable adducts and the Lys/Lys genotype were 0.9 (OR, 0.94; 95% CI, 0.53-1.65) and 1.2 (OR, 1.20; 95% CI, 0.8-1.78), respectively. Thus, the joint additive effect was 1.9 versus the effects or either alone minus 1 (0.9 + 1.2 − 1 = 1.1). Similarly, the joint effect of both Gln/Gln genotype and current cigarette smoking was 1.6 (OR, 1.62; 95% CI, 0.89-2.93) versus those nonsmokers with the Lys/Lys genotype. The separate effects of Gln/Gln genotype and current smoking relative to nonsmokers with the Lys/Lys genotype were 0.9 (OR, 0.88; 95% CI, 0.58-1.34) and 0.9 (OR, 0.86; 95% CI, 0.58-1.28), respectively. Thus, the joint additive effect was 1.6 versus the effects of either alone minus 1 (0.9 + 0.9 − 1 = 0.8). These models supported the presence of an additive interaction between XPD genotype and both PAH-DNA adducts and cigarette smoking.

Overall, we found a modest, statistically significant association between those subjects with at least one variant Gln allele at exon 23 of the XPD gene and breast cancer risk (OR, 1.21; 95% CI, 1.01-1.44). The frequency of the Gln allele was 36.3% in controls; the frequency of the three genotypes in controls was 41.1%, 45.2%, and 13.7% for Lys/Lys, Lys/Gln, and Gln/Gln, respectively. Others have reported similar percentages, with the frequency of the Gln allele ranging from 29% to 42% (4, 6, 15, 16). Several studies have supported an association between Gln/Gln genotype and cancer risk in melanoma (17) and upper aerodigestive tract cancer (18). Studies of bladder cancer (16), lung cancer (19, 20), and basal cell carcinoma (21, 22) have not supported higher risk for the Gln/Gln genotype (for review, see ref. 23).

Several studies have examined the functional significance of the XPD exon 23 polymorphism. A recent study supports an association between XPD Gln allele and defects in nuclear excision repair (8). Other studies have also found associations with the variant Gln allele and suboptimal DNA repair (6, 7). Not all studies agree, however, and one study found no correlation between XPD exon 23 polymorphism and two markers of DNA damage: sister chromatid exchange and polyaromatic DNA adducts (15). Another study supported an association with suboptimal DNA repair and Lys/Lys homozygosity (24). Inconsistency in the existing studies may be attributed to the small sample size of the studies leading to greater statistical uncertainty, differences in selection methods of subjects, and differences in measurements and characterizations of DNA repair capacity. Large, well-designed studies that can clarify the association between genotype and phenotype are needed.

Few studies have examined interactions between XPD genotypes and exposures. One exception is the study conducted by Matullo et al. (16) that supported a stronger association between smoking and cancer among those with the Gln/Gln genotype in bladder cancer. We also found evidence for a gene-environment interaction (both multiplicative and additive) between current cigarette smoking and those with the Gln/Gln genotype. However, there were no clear patterns with duration. We also did not see a clear pattern by allele “dose” for former smokers.

This is the first study to examine the interaction between XPD genotype and PAH-DNA adducts on breast cancer risk. We observed a stronger association between genotype and breast cancer risk among women with higher levels of PAH-DNA adducts (above the median) than among women with nondetectable levels of PAH-DNA adducts. The multiplicative interaction term was not statistically significant, but there was evidence of an additive interaction. A smaller study also supports an association between the Gln allele and PAH-DNA adduct levels in breast tumor tissue of 76 women but not among adjacent nontumor tissue (25). We did not observe an association between PAH-DNA adducts and XPD codon 751 genotype. Reasons for the difference may be due to the use of mononuclear cells versus breast tumor tissue but are also very likely to be due to the vastly different sample sizes used in these two studies. Our much larger study may be less prone to spurious false-positive results (26).

Several alternative explanations for our findings should be considered. In any case-control study, recall bias may be an explanation for reported associations. In this study, it is possible that exposure information on cigarette use, but not PAH-DNA adducts because it was not based on self-report, may have been differentially recalled. If cases had a tendency to overreport exposure and if controls had a tendency to underreport, such misclassification could result in a bias away from the null for putative risk factors. However, it is unlikely that such misclassification would differ by genotype status. Thus, differences between associations by genotype status cannot be explained by recall bias. We were also able to assess confounding by several variables included in the main study questionnaire. For an unmeasured confounder to explain the interaction results, we found it would have to be differentially distributed in the separate genotype exposure strata. It is unlikely that this occurred. There was very high (97%) reliability in the measurement of genotype. The measurement error in measuring genotype was independent of case-control status and thus, if anything, would have led to a bias toward the null. Measurement error in exposure classification may have created the appearance of interaction, although our main findings were with ever/former/current smoking, which is less likely to have error than measures such as duration.

Whereas our study has many advantages including its large size, which provides excellent statistical power to examine interactions, our study is limited because we only examined one polymorphism in one gene involved in DNA repair. Our group started looking at genotypes that have been shown previously to have a phenotypic effect or to be associated with risk for any cancer in epidemiologic studies. We selected the XPD gene, particularly because of its role in the nucleotide excision repair pathway, because PAH-DNA adducts are repaired primarily by the nucleotide excision repair pathway. Therefore, we have started with the XPD gene because of our a priori hypothesis that gene-environment interactions with this gene would be most important to assessing potential differences by PAH-DNA adducts on breast cancer risk. In addition to other genes, there are other polymorphisms on the XPD gene such as the polymorphism Asp³¹²Asn on exon 10. We are currently working on the analysis of the codon 312 variant and also plan to examine haplotypes for DNA repair genes. It is of interest that a recently published article on XRCC1 found that the candidate single nucleotide polymorphisms were all in linkage disequilibrium (27). In this report, analysis of candidate single nucleotide polymorphisms was as powerful as haplotyping.

Overall, this study suggests some support for the hypothesis that polymorphisms in the XPD gene may interact with environmental exposures and markers such as PAH-DNA adducts and cigarette smoking and breast cancer risk. However, as this is the first study to examine interactions between XPD in relation to PAH-DNA adducts and because the multiplicative interaction was not statistically significant, these results need to be replicated in other studies. We also found evidence of both additive and multiplicative interaction among cigarette smoking, XPD genotype, and breast cancer risk. If replicated, this suggests that only a subgroup of women who may be more genetically susceptible will face an increased risk of breast cancer from cigarette smoking.

For their valuable contributions to the Long Island Breast Cancer Study Project, we thank the members of the Long Island Breast Cancer Network and the 31 participating institutions on Long Island and New York City, NY. We thank other collaborators who assisted with various aspects of our data collection efforts, including Julie A. Britton, Ph.D.; Mary Wolff, Ph.D.; Geoffrey Kabat, Ph.D.; Steve Stellman, Ph.D.; Maureen Hatch, Ph.D.; Gail Garbowski, M.P.H.; H. Leon Bradlow, Ph.D.; David Camann, B.S.; Martin Trent, B.S.; Ruby Senie, Ph.D.; Carla Maffeo, Ph.D.; Pat Montalvan; Gertrud Berkowitz, Ph.D.; Margaret Kemeny, M.D.; Mark Citron, M.D.; Freya Schnabel, M.D.; Allen Schuss, M.D.; Steven Hajdu, M.D.; and Vincent Vinceguerra, M.D.