Genetic Variation in Base Excision Repair Genes and the Prevalence of Advanced Colorectal Adenoma

Base excision repair (BER) corrects DNA damage caused by oxidative stress and low folate intake, which are putative risk factors for colorectal neoplasia. To examine the relationship between genetic variation in BER genes and colorectal adenoma risk, we conducted a case-control study of 767 cases of advanced colorectal adenoma and 773 controls from the baseline screening exam of the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial. Cases included participants diagnosed with advanced left-sided adenoma, and controls were subjects without evidence of a left-sided polyp by sigmoidoscopy, frequency-matched to cases on race and gender. Twenty single nucleotide polymorphisms were genotyped in four BER genes (APEX1, PARP1, POLB, and XRCC1), and conditional logistic regression was used to estimate odds ratios (OR) and 95% confidence intervals (95% CI) for the association with colorectal adenoma. Two variants with possible functional significance were associated with risk. The APEX1 51H variant was associated with a borderline significant decreased risk of colorectal adenoma (OR, 0.66; 95% CI, 0.44–1.00), and the XRCC1 399Q variant was inversely associated with risk among Caucasians (OR, 0.80; 95% CI, 0.64–0.99). Homozygotes at two PARP1 loci (A284A and IVS13+118G>A) were also associated with a decreased risk of colorectal adenoma compared with wild-type carriers (OR, 0.70; 95% CI, 0.49–0.98 for both), which was restricted to advanced adenomas displaying histologically aggressive characteristics (OR, 0.51; 95% CI, 0.33–0.78, P = 0.002 for PARP1 A284A). This study suggests that polymorphisms in APEX1, XRCC1, and PARP1 may be associated with advanced colorectal adenoma. [Cancer Res 2007;67(3):1395–404]

Defects in DNA repair have been implicated in the development of colorectal neoplasia. Reduced DNA repair capacity has been associated with both colorectal adenoma (1) and cancer (2), and germ line mutations in mismatch repair genes have been linked to hereditary nonpolyposis colorectal cancer. In addition, rare variants in the base excision repair gene, MUTYH, have been associated with an autosomal recessive form of adenomatous polyposis (3). However, the role of common genetic variation in DNA repair genes and the risk of colorectal neoplasia is uncertain.

Base excision repair (BER) may be particularly important for the prevention of colorectal adenoma and cancer because it repairs uracil and oxidative DNA damage. Epidemiologic studies suggest that low folate intake is associated with an increased risk of colorectal neoplasia (reviewed in ref. 4). By decreasing the availability of N⁵,N¹⁰-methylenetetrahydrofolate, which is needed for the de novo synthesis of thymine, folate deficiency leads to the misincorporation of uracil into DNA and subsequently single-strand DNA breaks (5). Low intake of fruit and vegetables may contribute to the development of colorectal neoplasia by decreasing the availability of antioxidants, which reduce oxidative stress and damage (6). Tobacco smoke, which increases oxidative damage, has also been consistently associated with an increased risk of colorectal adenoma (reviewed in ref. 7). Differences in the efficiency of BER may alter susceptibility to colorectal neoplasia.

BER consists of two major subpathways: short-patch BER and long-patch BER (8). In both pathways, a DNA glycosylase cleaves the damaged base from the DNA backbone, creating an abasic site. Depending on the type of DNA glycosylase that removed the damaged base, APE1 (encoded by APEX1) either nicks the DNA backbone 5′ of this site or removes the 3′ residue. In short-patch BER, polymerase β (POLB) replaces the missing nucleotide, and the gap is sealed by the DNA ligase III/XRCC1 complex with the help of PARP1 (9). In long-patch BER, an oligonucleotide of two to seven bases is synthesized by either POLB or polymerase δ/ε. A flap-endonuclease then removes the damaged strand, allowing the newly synthesized nucleotides to fill the gap, and DNA ligase I seals the gap.

To examine the relationship between polymorphisms in four important BER genes (APEX1, PARP1, POLB, and XRCC1) and advanced colorectal adenoma risk, we conducted a case-control study within the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial. Because functional variants of most BER genes have not been well characterized, we selected multiple polymorphisms in each gene in order to capture the common haplotypes present among Caucasians. We also explored gene-environment interactions between the BER gene polymorphisms and dietary and lifestyle factors, such as tobacco smoke, fruit and vegetable consumption, folate intake, and alcohol use.

Study population. The PLCO Cancer Screening Trial is a randomized trial that was designed to evaluate screening methods for the early detection of prostate, lung, colorectal, and ovarian cancer (10, 11). In brief, the study recruited 154,952 men and women ages 55 to 74 years from 10 centers in the U.S. from 1993 to 2001. Participants were randomized to routine care or screening with (a) prostate-specific antigen testing and digital rectal examination for prostate cancer, (b) chest X-ray for lung cancer, (c) 60 cm flexible sigmoidoscopy for colorectal cancer, and (d) CA125 testing and transvaginal ultrasound for ovarian cancer. Participants found to have polyps or other suspicious lesions by sigmoidoscopy were referred to their primary physician for further evaluation, and trained medical record personnel abstracted information regarding the diagnosis from medical and pathologic reports. All participants provided written informed consent, and the study protocol was approved by institutional review boards at the 10 screening centers and the National Cancer Institute.

Identification of cases and controls. Cases and controls for this investigation were selected from among participants who were randomized to the screening arm of the trial between September 1993 and September 1999. Both cases and controls were required to have had a successful baseline sigmoidoscopy examination (defined as insertion to at least 50 cm with ≥90% of the mucosa visible or suspicious lesion found), completed a baseline risk factor questionnaire, provided a blood specimen, and consented to participate in etiologic studies of cancer. Persons with a self-reported history of colorectal polyps, ulcerative colitis, Crohn's disease, familial polyposis, Gardner's syndrome, or cancer (except basal cell or squamous cell skin cancer) were excluded from the study. Cases were randomly selected from among participants diagnosed with left-sided (descending, sigmoid colon, or rectum) advanced adenoma (≥10 mm in size, exhibiting high-grade dysplasia, carcinoma in situ, or villous characteristics) at baseline. Controls were chosen from among subjects with a negative sigmoidoscopy examination (i.e., no polyp or suspicious lesions) and were frequency-matched to cases on gender and self-reported ethnicity.

A total of 772 cases and 777 controls were originally selected. Five cases and four controls either had no DNA or had discrepancies on repeated DNA fingerprint analyses and were excluded from all analyses, leaving 767 cases and 773 controls for this study. Adenomas were detected in the distal (descending or sigmoid) colon in 536 cases, rectum in 140 cases, and both distal colon and rectum in 91 cases. Approximately 63% had at least one adenoma considered to be histologically aggressive (villous characteristics, high-grade dysplasia, or carcinoma in situ).

Genotype analysis. Common single nucleotide polymorphisms (SNP) with minor allele frequencies ≥5% in four BER genes (PARP1, APEX1, POLB, and XRCC1) were selected from public databases (e.g., dbSNP, SNP500Cancer, and GeneSNP). We selected one SNP approximately every 5 to 8 kb for the larger genes (PARP1, POLB, and XRCC1) and every 1 kb for the smaller gene, APEX1, with the goal of capturing the common haplotypes associated with each gene. Whenever possible, we tried to include SNPs within coding regions or the promoter. A total of 20 SNPs [APEX1: −655T>C (rs1760944), Q51H (rs1048945), and D148E (rs3136820); PARP1: D81D (rs1805404), IVS4+12A>G (rs1805403), A284A (rs1805414), K352K (rs1805415), IVS13+118A>G (rs3219090), V762A (rs1136410), and IVS21+358A>G (rs1891107); POLB: IVS1-89C>T (rs3136717), IVS5+1214C>T (rs3136748), IVS7+171G>A (rs2953983), and IVS11-235A>G (rs3136794); XRCC1: −77T>C (rs3213245), IVS2+4850T>C (rs2854510), IVS2-216G>A (rs1001581), R194W (rs1799782), R280H (rs25489), and R399Q (rs25487)] were selected for genotyping.

DNA was extracted from whole blood or buffy coat samples using QIAamp DNA Blood Midi or Maxi Kits. All genotyping was done using TaqMan or MGB Eclipse assays at the National Cancer Institute Core Genotyping Facility (12). Laboratory personnel were blinded to case-control status, and replicate quality control samples (46 individuals assayed two to six times per polymorphism) were interspersed in the plates. Replicate samples displayed 100% concordance for all SNPs, except POLB IVS1-89C>T (% agreement = 99.2% and κ = 0.98). Depending on the batch, 1% to 8% of the subjects had insufficient DNA for genotyping, and a small percentage of individuals (<1%) were found to have discrepancies on repeated DNA fingerprint analyses and were excluded. Of those remaining, genotyping was successfully completed for 93% to 99% of the participants, depending on the genotype.

The genotype frequencies among controls by race were consistent with Hardy-Weinberg proportions (P > 0.05) for all SNPs, except PARP1 IVS4+12A>G (P = 0.03 for Caucasians) and PARP1 A284A (P = 0.05 for Caucasians). All of the PARP1 SNPs were in strong linkage disequilibrium, and a closer inspection of the data revealed that there were fewer heterozygotes than expected for all of the PARP1 SNPs, suggesting that the small differences between the observed and expected genotype frequencies under Hardy-Weinberg equilibrium for PARP1 IVS4+12A>G and A284A were due to chance and not laboratory error.

Dietary data. Information regarding usual dietary intake over the past year was assessed using food frequency questionnaire (FFQ).⁶

The FFQ was adapted from the Willett and Block questionnaires, both of which have been previously validated (13, 14), and included 137 items regarding individual foods, 14 questions about vitamin and mineral supplement use, and 10 questions regarding meat cooking practices. Participants were asked to provide information on frequency and duration of supplement use, and information on portion size was collected for 77 food items. Daily intake of each nutrient was estimated by multiplying the daily frequency of the food item by its gender-specific portion size and the nutrient content as reported in the U.S. Department of Agriculture nutrient database (15).

A FFQ was considered to be invalid if more than seven of the food items were left blank or the total energy intake reported was in the lowest 1% or highest 1% for the entire cohort. Participants with invalid FFQs (n = 80) were eliminated from dietary analyses. Participants who completed the FFQ after the sigmoidoscopy (n = 203) were also excluded to prevent recall bias that might occur if participants changed their diet after receiving the results of the sigmoidoscopy, leaving a total of 610 cases and 647 controls for the dietary analyses. The distribution of intake in the controls was used to determine cutpoints of consumption for each dietary item examined. The data was examined using multiple different cutpoints (e.g., medians, thirds, and fourths) for the nutrients. Folate intake was estimated from both diet and supplement use, and all analyses with dietary factors were adjusted for energy.

Statistical analysis. Conditional logistic regression was used to estimate the odds ratios (OR) and 95% confidence intervals (95% CI) for the association between each polymorphism and colorectal adenoma, adjusting for age at screening (55–59, 60–64, 65–69, or 70+). Test for trend for each polymorphism was assessed by including a single variable for genotype, coded as the number of variant alleles, in the regression model. Further adjustment for smoking status (never, former, current, or cigar/pipe smoker), body mass index (continuous), education (<12 years, high school, some college, college), and aspirin or ibuprofen use did not materially affect the results (data not shown).

For all polymorphisms/haplotypes shown to be associated with adenoma risk, stratified analyses were conducted to compare the associations across exposure categories, and interactions between the polymorphisms and other covariates were assessed by including the cross-product terms as well as the main effect terms in logistic regression models. For continuous exposures, categories were determined based on predetermined cutpoints (e.g., median for folate intake), and the categorical variables were used in evaluating possible interactions. The statistical significance of the interaction was evaluated by comparing nested models with and without the cross-product terms using a likelihood ratio test. Polytomous logistic regression was used to evaluate the association between the polymorphisms and adenoma subtypes, and heterogeneity between subtypes was assessed by comparing ORs using a test of homogeneity.

Because linkage disequilibrium can vary between ethnic groups, measures of linkage disequilibrium and haplotype analyses were conducted among Caucasians only (the sample size was too small for other groups). For SNPs located within the same gene, pairwise linkage disequilibrium measures (D′ and r²) were examined using the program Haploview.⁷

Haplotypes were estimated using an expectation-maximization algorithm, and overall differences between the cases and controls were assessed using the global score test (16) implemented in HaploStats,⁸ adjusted for age and sex. Risks for individual haplotypes were estimated assuming an additive model using the generalized linear model implemented in HaploStats (16, 17). The haplotype containing the wild-type allele at each locus was used as the referent. All statistical analyses, except for the linkage disequilibrium and haplotype analyses, were conducted using STATA 7.0 (College Station, TX).

Baseline characteristics of the cases and controls are shown in Table 1. Cases were slightly older than controls at the time of screening with a mean age of 63.2 ± 5.2 years compared with 62.0 ± 5.2 years for controls. After adjustment for age, cases were also less likely to have graduated from college (OR, 0.60; 95% CI, 0.39–0.90) and more likely to be obese (body mass index ≥30: OR, 1.28; 95% CI, 0.97–1.70), have ever smoked (OR, 1.39; 95% CI, 1.12–1.73), and have a first-degree relative with colorectal cancer (OR, 1.36; 95% CI, 0.98–1.89). Folate and fruit intake was also higher among controls than cases (P = 0.05 and P = 0.01, respectively).

When examining the polymorphisms in BER genes individually, two APEX1 variants (51H and 148E) displayed a borderline significant association with advanced colorectal adenoma among all participants (OR, 0.66; 95% CI, 0.44–1.00 and OR, 1.27; 95% CI, 1.01–1.60 for 51H and 148E, respectively), which was slightly attenuated when the analysis was restricted to Caucasians (Table 2). The two APEX1 polymorphisms were in strong linkage disequilibrium in Caucasians (D′ = 0.89) but weakly correlated (r² = 0.04). A haplotype analysis of the APEX1 polymorphisms revealed that the increased risk of colorectal adenoma observed with the APEX1 148E variant was likely driven by its inverse association with the 51H variant, because only the haplotype containing the 51H allele was associated with colorectal adenoma (Table 3) and no associations were observed with two haplotypes containing the APEX1 148E allele. Because adenoma subtypes differ in their propensity to undergo malignant transformation, we examined APEX1 associations according to adenoma subtype, but no substantial differences were observed.

Among Caucasians, an inverse association was observed between colorectal adenoma risk and variants at two PARP1 loci (A284A and IVS13+118G>A) with an OR of 0.70 (95% CI, 0.49–0.98) for homozygotes at either loci compared with the wild-type and heterozygous genotypes (Table 2). Other polymorphisms in PARP1 were not associated with risk. The PARP1 A284A and IVS13+118G>A polymorphisms were in strong linkage disequilibrium (D′ = 1.0) and nearly completely correlated (r² = 0.99) among Caucasians, so that it was impossible to differentiate the effects of one polymorphism from the other statistically. Both polymorphisms were in strong linkage disequilibrium (D′ = 1.0 for all pairwise comparisons) and moderately correlated (r²: 0.16–0.40) with the other PARP1 polymorphisms genotyped in this study. Adjustment for the other PARP1 polymorphisms did not significantly alter the ORs observed for PARP1 A284A and IVS13+118G>A. In the haplotype analysis, the two haplotypes containing the variant alleles at A284A and IVS13+118G>A only displayed weak nonsignificant associations (Table 3); however, this is not surprising because the haplotype analysis was conducted under an additive genetic model for each haplotype and the associations observed for PARP1 A284A and IVS13+118G>A were for a recessive model.

The inverse association between homozygotes at PARP1 A284A and IVS13+118G>A and colorectal adenoma was restricted to participants with histologically aggressive adenoma (OR, 0.50; 95% CI, 0.32–0.78, P = 0.002 for CC homozygotes at A284A among Caucasians with similar results for IVS13+118A>G). No association was observed among Caucasians without histologically aggressive adenoma (OR, 1.05; 95% CI, 0.68–1.60 for CC homozygotes at A284A with similar results for IVS13+118A>G), and the heterogeneity between histologically aggressive and nonaggressive phenotypes was P = 0.005 for both PARP1 loci. Similarly, the association between colorectal adenoma and homozygotes at PARP1 A284A and IVS13+118G>A was limited to adenoma with tubular/villous or villous characteristics (P_{heterogeneity} = 0.02 for both variants compared with tubular adenoma; Table 4). When the results were further stratified, the inverse association with the CC homozygotes at PARP1 A284A seemed to be stronger for villous adenoma (OR, 0.17; 95% CI, 0.03–0.54, P = 0.0003) compared with tubular/villous adenoma (OR, 0.66; 95% CI, 0.41–1.06, P = 0.09) or tubular adenoma (OR, 0.98; 95% CI, 0.60–1.58, P = 0.92). Similar results were observed for AA homozygotes at PARP1 IVS13+118G>A (Table 4).

The A allele at XRCC1 IVS2-216G>A and the Q allele at XRCC1 R399Q were also inversely associated with the risk of colorectal adenoma among Caucasians (P_trend = 0.03 for both; Table 2). The XRCC1 IVS2-216G>A and XRCC1 R399Q polymorphisms were in strong linkage disequilibrium (D′ = 0.99) and highly correlated (r² = 0.85) among Caucasians, making it impossible to separate the effects of one polymorphism from the other statistically. Both polymorphisms were in strong linkage disequilibrium (D′ = 0.86–1.0) but weakly to moderately correlated (r² = 0.03–0.48) with the other polymorphisms genotyped in XRCC1, and adjustment for the other XRCC1 polymorphisms had little effect on the ORs for XRCC1 IVS2-216G>A and XRCC1 R399Q. The haplotype containing the variant allele at R399Q was inversely associated with colorectal adenoma (Table 3), but the relationship only reached borderline statistical significance (P = 0.07). No differences were observed by adenoma subtype for the XRCC1 SNPs. No associations were observed between genetic variation in POLB and colorectal adenoma risk (Tables 2 and 3) and no significant heterogeneity was seen between adenoma subtypes.

Because the BER pathway repairs oxidative damage, we hypothesized that genetic variation in APEX1, PARP1, or XRCC1 might modify the association between colorectal adenoma and tobacco use or fruit and vegetable intake; however, no significant interactions were observed between APEX Q51H, PARP1 A284A, PARP1 IVS13+118G>A, or XRCC1 R399Q and these factors. Similarly, we thought that the polymorphisms in BER genes might modify the association of folate intake or alcohol consumption with colorectal adenoma because BER enzymes repair uracil misincorporation and single-strand breaks caused by folate deficiency. No significant interactions were observed between APEX Q51H, PARP1 A284A, PARP1 IVS13+118G>A, or XRCC1 R399Q and folate intake, alcohol consumption, or a combination of folate and alcohol intake. However, the XRCC1 194W variant did modify the association between folate intake and colorectal adenoma risk (P_interaction = 0.02). Among participants with folate intake below the median (534 μg/d), individuals with at least one copy of the XRCC1 194W allele displayed an increased risk of colorectal adenoma compared with the wild-type (OR, 1.66; 95% CI, 1.02–2.69), whereas no association was observed among individuals with folate intake above the median (OR, 0.74; 95% CI, 0.46–1.20). No significant interactions were observed between folate intake and XRCC1 R280H (P = 0.76) or the XRCC1 haplotypes (P = 0.17).

No substantial differences were observed between APEX Q51H, PARP1 A284A, PARP1 IVS13+118G>A, or XRCC1 R399Q and colorectal adenoma risk by age, dietary fat, body mass index, or adenoma location (distal versus rectal).

BER is the primary mechanism for the repair of endogenous DNA damage resulting from reactive oxygen species, hydroxylation, and other cellular processes (18). BER may be particularly important for preventing mutations and genetic instability in the large intestine, because it repairs uracil damage from folate deficiency, a known risk factor for colorectal neoplasia. Genetic variants that increased repair efficiency may protect against colorectal neoplasia by reducing the formation of somatic alterations. In this study, we found that genetic variants in three BER genes, APEX1, PARP1, and XRCC1, were associated with the risk of advanced colorectal adenoma.

APEX1 encodes for an apurinic/apyrimidinic endonuclease (APE1), which initiates the BER process by hydrolyzing the DNA phosphodiester backbone 5′ of the abasic site or removing the 3′ phosphate residue left by a bifunctional glycosylase (8). In addition to its role in BER, APE1 also functions as a reduction-oxidation (redox) activator of transcription factors. It has been shown to stimulate DNA binding of several transcription factors thought to be important in carcinogenesis, including nuclear factor-κB and p53 (19, 20). The DNA repair and redox activities of the protein are encoded by two distinct domains (21).

In our study, we found that the APEX1 148E variant was associated with an increased risk of colorectal adenoma, and that the APEX 51H variant was inversely associated with adenoma risk. The APEX1 D148E polymorphism is located within the endonuclease domain of the protein (21), but the residue is not well conserved across species and the polymorphism encodes a conservative amino acid change. Although one study found that the 148E homozygous genotype was associated with increased mitotic delay after exposure to ionizing radiation (22), the variant had no effect on DNA binding or endonuclease activity in biochemical assays (23), indicating that the polymorphism is unlikely to affect the DNA repair capacity of the protein. Consistent with this, when we conducted a haplotype analysis of the genotyped polymorphisms in APEX1, we found that the haplotypes containing the APEX1 148E variant were not associated with adenoma risk and the increased risk observed with the 148E variant in the single locus analysis was likely due to its inverse association with the 51H variant. Unlike the D148E polymorphism, the APEX1 51Q allele is conserved in the mouse genome, and the 51H variant encodes a nonconservative amino acid substitution. The functional significance of the variant has not been directly evaluated, but the polymorphism is located within the redox domain (21) and may affect the ability of APE1 to stimulate p53 transactivation (20) or other transcription factors (19).

PARP1 is an enzyme that catalyzes the addition of ADP-ribose units onto nuclear proteins in response to DNA damage and may play a role in promoting DNA ligation in short-patch BER (9), recruiting repair factors for single-strand breaks (24), and preventing recombination at single-strand breaks (25). PARP1 has also been implicated in other cellular functions, such as cell differentiation (26), modulation of chromatin structure (27), transcription (28), and telomere maintenance (29). A small case-control study showed that PARP1 levels in peripheral blood lymphocytes were significantly reduced in patients with colorectal adenomatous polyps (30), suggesting that PARP1 deficiency may play a role in the development of colorectal adenoma. Consistent with this hypothesis, PARP1-deficient mice show an elevated incidence of chemically induced colon tumors compared with PARP1-proficient mice (31).

In our study, homozygotes at two PARP1 loci (A284A and IVS13+118G>A) in strong linkage disequilibrium were associated with a decreased risk of advanced colorectal adenoma, particularly histologically aggressive adenoma. Neither variant is likely to influence the activity of the protein itself: the C allele at A284A does not affect the amino acid sequence of the protein, and the A allele at IVS13+118G>A is not located within a known regulatory or consensus site. Thus, we suspect that the association that we observed with colorectal adenoma may be due to linkage disequilibrium with another variant in the region, possibly in the PARP1 promoter. Because PARP1 activity is highly regulated by its promoter, variants within transcription binding sites in the promoter may influence its expression. Although one study suggested that the PARP1 762A variant may be associated with lower enzymatic activity (32), another study found no association between the variant and poly(ADP-ribosyl)ation capacity (33). Similarly, we found no association between the PARP1 762A variant and colorectal adenoma risk, suggesting that our observed associations at A284A and IVS13+118G>A were due to linkage disequilibrium with a different variant.

XRCC1 acts as scaffold protein, coordinating the actions of both POLB and DNA ligase III in short-patch BER. It also stimulates the repair of single-strand breaks by activating polynucleotide kinase (24), and is essential for efficient single-strand break repair and genomic stability (34). We found that variants at both XRCC1 IVS2-216G>A and R399Q were associated with a reduced risk of colorectal adenoma. The XRCC1 IVS2-216G>A is an intronic polymorphism and is unlikely to be functional. The XRCC1 R399Q polymorphism is located within the XRCC1 BRCA1 carboxyl-terminal domain (BRCT I; ref. 35) and is hypothesized to have functional significance because it is located within a well conserved region and encodes a nonconservative amino acid change. However, studies examining its association with markers of DNA damage or DNA repair function have yielded mixed results with some studies showing a positive association (36, 37) and others observing no association with the variant (38–41). Transfection of the XRCC1 R399Q polymorphism into XRCC1-deficient EM9 cells showed no difference between the variant and wild-type in single-strand break repair or survival after DNA alkylation treatment (35), suggesting that the variant itself may not appreciably alter the function of the protein. However, there may be another variant in linkage disequilibrium with both polymorphisms (such as XRCC1 N576T) that is responsible for the observed associations. Although evidence indicates that the XRCC1 280H variant reduces DNA repair capacity (42), it is unlikely to be responsible for the association observed with R399Q. The XRCC1 280H variant is not present on the same haplotype as the 399Q variant in our population, and we found no association with the 280H variant.

Several other studies have examined the association between the XRCC1 R399Q polymorphism and colorectal neoplasia (43–49). Consistent with our study, a Norwegian case-control study found the 399Q variant to be inversely associated with adenoma risk, particularly in advanced adenoma (49), and a case-control study in Los Angeles County reported a decreased risk of adenoma with the 399Q homozygous genotype (43). A large study in Taiwan also found a similar association with colorectal cancer (48), which was statistically significant among younger cases (OR, 0.68; 95% CI, 0.50–0.94, P = 0.02; ref. 50). However, smaller studies have either observed a positive association (44, 45) or no association (47, 49) with colorectal cancer risk.

We also observed a significant interaction between the XRCC1 194W variant and folate intake on colorectal adenoma risk. Among individuals with folate intake below the median, the 194W variant was associated with an increased risk of colorectal adenoma. The XRCC1 194W variant is a nonconservative amino acid substitution, but it is unclear whether this substitution affects the function of the protein. Although one epidemiologic study reported a positive association between the XRCC1 194W variant and chromosome damage (41), suggesting that the variant may reduce DNA repair capacity and increase cancer risk, other studies have either shown no association (22, 36–38) or an inverse association (51) with markers of DNA damage. Upon transfection into XRCC1-deficient EM9 cells, no difference was observed between the variant and wild-type in methyl methanesulfonate sensitivity (42), suggesting that the variant may not affect DNA repair capacity. Although the effect modification we observed in our study between the XRCC1 R194W polymorphism and folate intake is intriguing, caution should be exercised in interpreting this subgroup finding because the functional effects of the polymorphism are uncertain and we did not observe an association between the variant and colorectal adenoma risk overall.

Our study had several strengths and limitations. By selecting cases and controls from a large population-based screening trial, the potential for selection biases that occur with hospital- or clinic-based controls was reduced. Information on diet was collected prior to or at the time of sigmoidoscopy, so that any recall bias was minimized. However, because our population was well nourished, it is possible that some gene-diet interactions could not be detected due to the lack of subjects with extremely low intakes of folate or fruits and vegetables. By examining multiple polymorphisms per gene, we were able to comprehensively evaluate the association between genetic variation in APEX1, PARP1, XRCC1, and POLB and colorectal adenoma risk. However, some genetic variations may have been missed because we did not sequence the genes to select htSNPs. Our study had the advantage of being limited to cases with advanced colorectal adenoma, which have an increased likelihood of malignant transformation, and thus, the results may be more indicative of colorectal cancer.

Because our study was limited to distal colorectal adenoma, our results may not be generalizable to colorectal adenoma in the proximal colon. It is also possible that some of the controls for our study had proximal adenomas that were not detected by sigmoidoscopy. If the genetic variants evaluated in this study are associated with proximal adenoma, our observed results may underestimate the risk of adenoma in general due to disease misclassification among the controls. For example, if the prevalence of proximal adenoma among persons without a distal polyp is 3% to 28% (52), the actual OR for APEX1 51H may be between 0.66 (95% CI, 0.42–1.00) and 0.59 (95% CI, 0.38–0.90) in our study. In addition, because we conducted several subgroup analyses and multiple statistical tests, it is possible that some of our findings were false-positives. However, the strength of the relationship that we observed with the PARP1 variants increased with villous histology, and the consistency of our XRCC1 R399Q results with findings from other large studies suggests that these observations are less likely due to chance.

In conclusion, we found that genetic variants in three BER genes (APEX1, PARP1, and XRCC1) were associated with the risk of advanced colorectal adenoma. Additional studies are needed to confirm these findings and to further characterize the specific genetic variants that alter adenoma risk.

Grant support: Intramural Research Program of the National Cancer Institute, NIH.

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.

We thank the participants of the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial for making this study possible. In addition, we thank Dr. Kenneth Kinzler for his expertise and Drs. Christine Berg and Philip Prorok as well as the investigators and staff at the screening centers for their dedication to this trial.