A Prospective Study of XRCC1 Haplotypes and Their Interaction with Plasma Carotenoids on Breast Cancer Risk

It has been suggested that reduced DNA repair capacity may be a susceptibility factor predisposing women to breast cancer (1, 2, 3, 4, 5, 6). Genetic variations in DNA repair genes may lead to interindividual variation in DNA repair capacity and modify the associations between exogenous and endogenous carcinogens and breast cancer risk.

The XRCC1 protein is involved in the BER⁶ pathway. BER is responsible for repair of a wide variety of nonbulky exogenous and endogenous base damage and single strand breaks (7, 8). Although XRCC1 has no known enzymatic activity, there are three distinct domains that are sites of interaction with DNA polymerase β (amino acids 1–183; Refs. 9, 10), poly(ADP-ribose) polymerase (BRCT-I, amino acids 384–476), and DNA ligase III (BRCT-II, amino acids 573–592; Refs. 11, 12, 13). More recently, XRCC1 was also reported to interact with polynucleotide kinase (14) and APE1 (15). This suggests that XRCC1 may act as a nucleating factor in BER by bringing different components together at the site of action to promote the efficiency of the repair machinery. A number of SNPs in XRCC1 have been identified (16, 17). These polymorphisms may alter BER proficiency and, in turn, confer genetic predisposition to breast cancer. We prospectively assessed whether candidate polymorphisms and haplotypes in the XRCC1 gene are associated with breast cancer risk in a nested case-control study within the Nurses’ Health Study.

Reactive oxygen species and free radicals can cause oxidative DNA damage, which is mainly repaired by BER. Carotenoids can neutralize reactive oxygen species and reduce oxidative DNA damage because of their antioxidant properties (18, 19). Among carotenoids, higher plasma α-carotene or β-carotene level was significantly associated with a decreased risk of breast cancer in this nested case-control study.⁷ We further investigated the gene-environment interaction hypotheses that XRCC1 genetic variations may modify the associations of plasma carotenoid levels with breast cancer risk.

The Nurses’ Health Study was established in 1976 when 121,700 female registered nurses between the ages of 30 and 55 completed a self-administered questionnaire on their medical histories and baseline health-related exposures. Updated information has been obtained by questionnaires every 2 years. Incident breast cancers were identified by self-report and confirmed by medical record review. Between 1989 and 1990, blood samples were collected from 32,826 of the cohort members. Subsequent follow-up has been greater than 98% for this subcohort.

Eligible cases in this study consisted of women with pathologically confirmed incident breast cancer from the subcohort who gave a blood specimen. Cases with a diagnosis anytime after blood collection up to June 1, 1998, with no previously diagnosed cancer except for nonmelanoma skin cancer were included. One or two controls were randomly selected among women who gave a blood sample and were free of diagnosed cancer (excluding nonmelanoma skin cancer) up to and including the questionnaire cycle in which the case was diagnosed. Controls were matched to cases on year of birth, menopausal status, postmenopausal hormone use at blood collection, month of blood return, time of day of blood collection, and fasting status at blood draw; menopause was defined as described previously (20). The nested case-control study consists of 1004 incident breast cancer cases and 1385 matched controls. The study protocol was approved by the Committee on Use of Human Subjects of the Brigham and Women’s Hospital (Boston, MA).

Information regarding breast cancer risk factors was obtained from the 1976 baseline questionnaire, subsequent biennial questionnaires, and a questionnaire completed at the time of blood sampling. Menopausal status and use of postmenopausal hormones were assessed at blood draw and updated until date of diagnosis for cases and the equivalent date for matched controls. First-degree family history of breast cancer was asked in 1982 and updated in subsequent questionnaires. Information regarding cigarette smoking was asked in the 1976 questionnaire and updated in subsequent questionnaires (21).

We selected haplotype-tagging SNPs for the XRCC1 gene, using data derived from the resequencing of the exons and adjacent intronic and noncoding regions of the gene in a multiple ethnicity group of 90 samples from the NIH DNA Polymorphism Discovery Resource available from the Coriell Institute for Medical Research (22). The PCR products for direct sequencing included the splice sites and the 5′ and 3′ regions of the gene (16, 17). The 90 samples are from United States residents representing the ethnic diversity of the United States population, but the ethnicity of individual samples is unknown. The 90 individuals in this subset are from the following ethnic groups: 23 European Americans; 23 African Americans; 11 Mexican Americans; 11 Native Americans; and 22 Asian Americans. We performed haplotype estimation using the Partition-Ligation EM Algorithm of Qin et al. (23). On the basis of seventeen common (>1% allele frequency) SNPs found in the XRCC1 gene (Table 1), five common haplotypes were inferred with frequency above a threshold of 1.3% (0.05 × 23/90 = 0.013), which was set to ascertain potential common (>5%) Caucasian-specific haplotypes from the mixed population. The five common haplotypes accounted for ∼87% of the alleles of this mixed population group. Only 4 of the 17 SNPs are necessary to reconstruct the five common haplotypes (Table 1). We thus genotyped four haplotype-tagging SNPs [C26304T(Arg¹⁹⁴Trp), C26602T, G28152A(Arg³⁹⁹Gln), and G36189A(Gln⁶³²Gln)] in the present case-control study of mostly Caucasian women.

To maximize power, we included all available plasma samples, including samples left unpaired because of either subsequent exclusions (of either the case or controls) or laboratory issues. Plasma samples of 994 cases and 995 controls were assessed using reversed-phase high-performance liquid chromatography methods described by El-Sohemy et al. (24) in the Micronutrient Analysis Laboratory in the Department of Nutrition at the Harvard School of Public Health to determine the concentrations of α-carotene, β-carotene, β-cryptoxanthin, lycopene, lutein/zeaxanthin, α-tocopherol, and γ-tocopherol. All case control pairs were assayed together; the samples were ordered randomly and labeled within each pair. Plasma quality control samples were interspersed to assess laboratory precision. Coefficient of variation for each antioxidant, weighted by the proportion of samples on a batch- and pool-specific basis, was <10%. Laboratory personnel were blinded to case control status and identity of replicate samples.⁷

Genotyping assays were performed by Pyrosequencing (Pyrosequencing, Inc., Uppsala, Sweden; Ref. 25), RFLP, and the 5′ nuclease assay (TaqMan) using the ABI Prism 7900HT Sequence Detection System (Applied Biosystems, Foster City, CA). The primers, probes, enzymes, and conditions for genotyping assays are available upon request. Genotyping was performed by laboratory personnel blinded to case-control status, and blinded quality control samples were inserted to validate genotyping procedures; concordance for the blinded samples was 100%.

We used a χ² test to assess whether the XRCC1 genotypes were in Hardy-Weinberg equilibrium and to determine the P for difference in haplotype frequencies between cases and controls. Conditional logistic regression was used to calculate ORs and 95% CIs to assess the risk of breast cancer for XRCC1 genotypes among all women. Unconditional logistic regression was used in analyses stratified by genotype. Plasma levels of antioxidants were categorized into quartiles based on the batch-specific cut points of control subjects. Weighted median values for each quartile were based on batch-specific medians of controls weighted by the proportion of subjects in each batch. In multivariate analysis, in addition to the matching variables, we adjusted for the following breast cancer risk factors: BMI (kg/m²) at age 18 years, weight gain since age 18, age at menarche, age at menopause, parity/age at first birth, first-degree family history of breast cancer, personal history of benign breast disease, and duration of postmenopausal hormone use. Alcohol consumption was based on the 1990 dietary questionnaire; the 1986 questionnaire was used for individuals who did not provide this information on the 1990 questionnaire.

Tests for trend were conducted by assigning the weighted median values for quartiles of plasma antioxidant levels and the median values for strata of smoking duration among controls to both cases and controls as continuous variables. To test statistical significance of interactions between the environmental exposures and the XRCC1 genotype, we first used a LRT to compare nested models that included terms for all combinations of the XRCC1 genotype and levels of environmental exposure to the models with indicator variables for the main effects only (nominal LRT). We also modeled XRCC1 genotypes as ordinal variables and environmental exposures as continuous variables as described for trend tests to assess the statistical significance of interactions by LRT test of a single interaction term (ordinal LRT). All Ps were two-sided.

The mean age at blood draw of cases was 57.3 years and that of controls was 57.8 years. There were 20.1% premenopausal, 70.7% postmenopausal women, and 9.2% women with uncertain menopausal status at blood draw. Cases and controls had similar mean BMI at blood draw (25.4 versus 25.5 kg/m²) and weight gain since age 18 years (11.8 versus 11.4 kg). Compared with controls, cases had similar ages at menarche (12.5 versus 12.6 years), first birth (24.8 versus 24.8 years), and menopause (48.2 versus 48.0 years). The proportion of women with a first-degree family history of breast cancer was higher among the cases (20.9 versus 14.8%). Cases were more likely to have a personal history of benign breast disease (65.0 versus 49.6%), a longer duration of postmenopausal hormone use (38.0 versus 26.9% current users for ≥5 years), and a longer duration of smoking (13.1 versus 12.4 years). Cases and controls had similar median values of plasma α-carotene (0.10 versus 0.11 μmol/liters) and β-carotene (0.44 versus 0.45 μmol/liters) concentrations.

The five common haplotypes inferred from the four haplotype-tagging SNPs accounted for 99% of the chromosomes in the present study population of mostly Caucasian women (Table 2; self-reported major ethnicity was similar between cases and controls in this study population; cases versus controls: Southern European, 17.5 versus 17.8%; Scandinavian, 8.6 versus 7.6%; and other Caucasian, 63.1 versus 61.3%; Asians, Hispanics, and African Americans comprised ∼1% of the total cases and controls). Five rare haplotypes were also inferred. The haplotype harboring the variant allele ¹⁹⁴Trp was more common in controls than in cases (6.6 versus 5.3%, P = 0.07). There was no notable difference in frequency distribution in cases and controls for other haplotypes. Among these five common haplotypes, one haplotype consisted of the common allele at each polymorphic site; each of the other four haplotypes carried only one variant allele of the four sites exclusively. In other words, none of these four SNPs were in linkage disequilibrium. Individuals with the variant allele of one polymorphism carried the haplotype that only harbored this variant but not others. Therefore, the carriage of a polymorphism can be viewed as the carriage of the corresponding haplotype in the evaluation of the main effect of polymorphic sites and gene-environment interactions. Additionally, it is noteworthy that there is another nonsynonymous polymorphism Arg²⁸⁰His, which is in 100% genotype concordance with C26602T among the 90 individuals in the NIH DNA Polymorphism Discovery Resource, according to the XRCC1 resequencing data of National Institute of Environmental Health Sciences Environmental Genome Project at the University of Washington.⁸

The distribution of all XRCC1 genotypes among the controls was in Hardy-Weinberg equilibrium. We observed a marginally significant reduction in the risk of breast cancer among ¹⁹⁴Trp carriers. As compared with noncarriers, women with at least one ¹⁹⁴Trp allele had a decreased risk of breast cancer (multivariate OR, 0.79; 95% CI, 0.60–1.04; Table 3). No overall associations between C26602T, Arg³⁹⁹Gln, and Gln⁶³²Gln and breast cancer risk were found. For the two nonsynonymous polymorphisms Arg¹⁹⁴Trp and Arg³⁹⁹Gln, carriage of either variant allele was not associated with altered breast cancer risk compared with being homozygous wild-type for both alleles (data not shown).

We assessed whether the associations between the plasma carotenoid levels and breast cancer risk differed by Arg¹⁹⁴Trp. The ORs for breast cancer risk by Arg¹⁹⁴Trp and plasma α-carotene and β-carotene levels are shown in Table 4. We observed a significant 60% reduction in risk (multivariate OR, 0.40; 95% CI, 0.21–0.77) for ¹⁹⁴Trp carriers in the highest α-carotene quartile versus noncarriers in the lowest quartile. The ¹⁹⁴Trp carriers in the third quartile also had a significantly decreased risk compared with noncarriers in lowest quartile (OR, 0.53; 95% CI, 0.30–0.97). The inverse association between plasma α-carotene level and breast cancer risk was significantly different between ¹⁹⁴Trp carriers and noncarriers (P, test for ordinal interaction = 0.02). Incorporating the cross-classified interaction terms also significantly increased the goodness fit of the model (P, test for nominal interaction = 0.03). This pattern was similar for plasma β-carotene level, with a 68% reduction in risk (multivariate OR, 0.32; 95% CI, 0.16–0.61) in the top quartile among ¹⁹⁴Trp carriers compared with noncarriers in the bottom quartile. Compared with noncarriers, the inverse association between β-carotene level and breast cancer risk was significantly stronger among ¹⁹⁴Trp carriers (P, test for trend = 0.003 versus 0.07; P, test for ordinal interaction = 0.003). The cross-classified interaction between plasma β-carotene level and XRCC1 194 genotype on breast cancer risk was statistically significant (P, test for nominal interaction = 0.02). Of note, in these analyses of interactions between XRCC1 Arg¹⁹⁴Trp and plasma α- or β-carotene level, the multivariate ORs did not change materially, and the tests of interaction remained significant after controlling for plasma α-carotene (in β-carotene analyses), β-carotene (in α-carotene analyses), β-cryptoxanthin, lycopene, α-tocopherol, γ-tocopherol, lutein/zeaxanthin, and folate, one at a time or all simultaneously. No statistically significant interactions were observed between the Arg¹⁹⁴Trp genotype and other plasma antioxidants such as β-cryptoxanthin, lycopene, α-tocopherol, γ-tocopherol, and lutein/zeaxanthin on breast cancer risk (data not shown).

The positive association of duration of smoking with breast cancer risk was stronger among ¹⁹⁴Trp noncarriers than carriers (P, test for trend = 0.06 versus 0.40; Table 5). Among noncarriers, the multivariate OR was 1.31 (95% CI, 0.98–1.74) for women who smoked ≥ 35 years compared with never-smokers. However, there was no evidence of a statistically significant interaction between duration of smoking and the Arg¹⁹⁴Trp on breast cancer risk (P, test for nominal interaction = 0.76; test for ordinal interaction = 0.90).

We did not observe significant interactions between the other three SNPs (C26602T, Arg³⁹⁹Gln, and Gln⁶³²Gln) and the above exposures on breast cancer risk in this study.

In this prospective nested case-control study, we genotyped four haplotype-tagging SNPs in the XRCC1 gene. We observed that XRCC1 ¹⁹⁴Trp carriers were at marginally significant decreased risk of breast cancer; the inferred haplotype harboring the ¹⁹⁴Trp allele was nonsignificantly less frequent in cases than in controls. We observed that the Arg¹⁹⁴Trp modified the associations of plasma α- and β-carotene levels with breast cancer risk. The prospective design blood sample collection before case diagnosis, the relatively large number of incident cases, and the high follow-up rates strengthen the validity of this study.

We inferred five common haplotypes from seventeen SNPs (>1% allele frequency; Table 1) based on data from the resequencing of the exons and adjacent intronic and noncoding regions of the XRCC1 gene on 90 samples from the NIH DNA Polymorphism Discovery Resource (17). We used a threshold of 1.3% (0.05 × 23/90 = 0.013) to ascertain alleles that occurred at 5% prevalence among the 23 Caucasians in the 90 individuals in the sample set used for resequencing. This is a conservative procedure because (a) it is likely that there are few common SNPs that are unique to Caucasians and (b) because of the population admixture it is estimated that ∼40% of the chromosomes in these 90 people are of Caucasian origin (22). Furthermore, two recent studies of haplotype variation in different ethnic groups (26, 27) reported substantial conservation of haplotypes among ethnicities with fewest population-specific common haplotypes in Caucasians. This suggests that use of the NIH DNA Polymorphism Discovery Resource to infer haplotype in our study of mainly Caucasian individuals is unlikely to have missed defining common haplotypes in Caucasians. We genotyped four haplotype-tagging SNPs that were adequate to reconstruct the five common haplotypes. Only the same five haplotypes inferred from the genotyping of these four SNPs in this study were common (>5%).

One previously published study assessed two coding region polymorphisms of XRCC1 and breast cancer risk in whites. Duell et al. (28) reported data on the associations of Arg¹⁹⁴Trp and Arg³⁹⁹Gln with breast cancer risk in a population-based case-control study in North Carolina. These authors observed that ¹⁹⁴Trp carriers in whites were at a nonsignificantly decreased risk of breast cancer (OR, 0.7; 95% CI, 0.4–1.3; 251 cases and 234 controls), and ³⁹⁹Gln carriers in whites had unaltered breast cancer risk (OR, 1.0; 95% CI, 0.8–1.4; 386 cases and 381 controls). These authors also observed no notable difference in the effect of smoking duration according to XRCC1 Arg³⁹⁹Gln or Arg¹⁹⁴Trp genotype in white women. Our data, with a much larger sample size, support these findings, with a marginally significant inverse association of Arg¹⁹⁴Trp with breast cancer risk (multivariate OR, 0.79; 95% CI, 0.60–1.04), and no significant interaction with cigarette smoking. Our study is also consistent with most of the published studies reporting inverse associations between the XRCC1 ¹⁹⁴Trp allele and cancer risk at other sites (29).

Because of its location in the region of the BRCT-I interaction domain of XRCC1 with poly(ADP-ribose) polymerase, the Arg³⁹⁹Gln polymorphism has recently drawn considerable attention. Associations have been reported between the ³⁹⁹Gln allele and higher DNA adduct levels (30, 31, 32), higher sister chromatid exchange frequency (30), and radiation-induced G₂-phase delay (33). However, in an in vitro transfection experiment, the 399 wild-type and variant alleles equally complemented both the single-strand break repair defect and the sensitivity to methyl methanesulfonate in XRCC1-deficient EM9 cells, suggesting that the ³⁹⁹Gln variant retained a substantial level of function (34). The marginally significant inverse association of the XRCC1 ¹⁹⁴Trp with breast cancer risk in this study may suggest a protective role for Arg¹⁹⁴Trp in the development of breast cancer, potentially by increasing DNA repair capacity. The transition from a positively charged Arg to a hydrophobic Trp within a mammalian-conserved region may alter XRCC1 function. However, no association has been found between Arg¹⁹⁴Trp and altered levels of DNA damage biomarkers (32, 33). Our data showed that the two XRCC1 polymorphisms ¹⁹⁴Trp and ³⁹⁹Gln were not on the same common haplotype. Given the potential functional relevance of the two nonsynonymous SNPs, we compared the association of carriage of either variant allele to those who were homozygous wild-type at both sites. No notable opposite or synergistic effects of these two variants were observed.

Carotenoids can prevent oxidative DNA damage (35, 36, 37, 38). Oxidative DNA damage caused by low plasma levels of carotenoids may lead to increased dependency upon BER activity for maintaining genomic integrity. Thus, variation of BER capacity caused by XRCC1 genetic polymorphisms may modify the associations of plasma carotene levels with breast cancer risk. Our data showed that high plasma levels of carotene were particularly beneficial among ¹⁹⁴Trp carriers. The lowest risk was seen among ¹⁹⁴Trp carriers in the highest α- or β-carotene quartile. Furthermore, the inverse association of the ¹⁹⁴Trp allele and breast cancer risk was abolished in the setting of low carotene levels. This suggests that different DNA repair capacity caused by the genetic polymorphism Arg¹⁹⁴Trp may be more apparent in the situation of low DNA damage level and overwhelmed by excessive DNA damage. Among carotenoids, higher plasma α-carotene or β-carotene level was significantly associated with a decreased risk of breast cancer in this nested case-control study.⁷ The finding of effect modification of XRCC1 genotype on the relation between α-carotene and β-carotene and breast cancer risk suggests that the protective effect of carotenes on breast cancer risk is at least partially because of prevention of oxidative DNA damage. In addition, in the analysis of the interaction between the XRCC1 Arg¹⁹⁴Trp polymorphism and plasma α-carotene or β-carotene, the multivariate ORs did not change materially and the interactions remained significant after controlling for other plasma antioxidants one at a time or all simultaneously, suggesting that α-carotene and β-carotene may be the most relevant carotenoids in prevention of oxidative DNA damage in breast tissues.

Oxidative stress caused by cigarette smoking can induce oxidative DNA damage. However, we did not observe evidence of statistically significant interaction between smoking duration and any polymorphism. We lacked power to detect a subtle effect of smoking on oxidative DNA damage in breast tissues repaired by the BER, and smoking-related DNA adducts repaired by nucleotide excision repair may be more relevant to breast carcinogenesis (39). In addition, alcohol consumption is also believed to contribute to oxidative stress. In this nested case-control study, we did not observe the interactions between XRCC1 genotypes and alcohol consumption on breast cancer risk.

In summary, we provided evidence that women with the XRCC1 ¹⁹⁴Trp allele and the associated haplotype had a marginally significantly decreased risk of breast cancer. The data also indicate an interaction between the Arg¹⁹⁴Trp and plasma carotene level on breast cancer risk, which suggests the potential protective role of carotenes in breast carcinogenesis by preventing oxidative DNA damage. Characterization of this potential functional polymorphism and others that may be in linkage disequilibrium on the same haplotype is warranted to explore the biological significance of this variant.

We thank Dr. Hardeep Ranu, Robert O’Brien, Pati Soule, and Alicia Whittington for their laboratory assistance, Rong Chen for her programming support, and Dr. Hannia Campos for plasma antioxidant assays. We also thank the participants in the Nurses’ Health Study for their dedication and commitment.