Polymorphisms in XPC and ERCC2 genes, smoking and breast cancer risk
To evaluate the associations of breast cancer risk with polymorphisms in the XPC and XPD/ERCC2 DNA nucleotide excision repair genes, a case-control study nested within a prospective cohort of 14,274 women was conducted. Genotypes were characterized for 612 incident, invasive breast cancer cases and their 1:1 matched controls. The homozygous variant of a poly(AT) insertion/deletion polymorphism in intron 9 of the XPC gene (XPC-PAT+/+), was associated with breast cancer risk [odds ratio (OR) = 1.45, 95% confidence interval: 1.07–1.97], after adjustment for other breast cancer risk factors. The breast cancer risk associated with XPC-PAT+/+ did not differ by age at diagnosis. There was an indication of an interaction (p = 0.08) between the XPC-PAT+/+ genotype and cigarette smoking. Ever smokers with the XPC-PAT+/+ genotype were at elevated risk of breast cancer (OR = 1.56, CI: 0.95–2.58), but no differences were observed among never smokers. Analyses of the ERCC2 Lys751Gln polymorphism did not show an association with breast cancer risk, either overall or at younger ages. The results suggest that breast cancer risk is related to the XPC haplotype tagged by the XPC-PAT+/+ insertion-deletion polymorphism in intron 9. Further study of the XPC haplotypes and their interactions with smoking in relation to breast cancer risk is needed. © 2008 Wiley-Liss, Inc.
Much more attention has been paid to variation in DNA base excision repair (BER) genes in relation to breast cancer risk than to variation in nucleotide excision repair (NER) genes, although the bulky DNA adducts and DNA interstrand crosslinks1 repaired by the NER pathway may be of greater significance than the simple base changes repaired by the BER pathway, and several NER genes have been implicated in breast cancer risk.2–5 Our study examines the association of 2 polymorphisms in the NER pathway with breast cancer risk in a case–control study nested within a cohort. A poly(AT) insertion/deletion polymorphism in the XPC gene (XPC-PAT+) has not previously been studied in relation to breast cancer risk, while early results regarding a Lys751Gln polymorphism in the ERCC2 gene and breast cancer were promising6–8 and warranted verification.
The XPC (Xeroderma Pigmentosum, Group C) gene, located at chromosome 3p25, is expressed in breast tissue.9 It codes for a protein that performs early detection of bulky DNA damage and triggers other proteins to repair the damage.10XPC knockout mice showed a 30-fold increase in Hprt gene mutations,11 a greater persistence of induced p53 mutations, decreased apoptosis,12 and increased incidence of skin, liver, lung and testicular tumors.13, 14
A poly(AT) insertion/deletion polymorphism (XPC-PAT+; Genbank AF076952) has been reported in intron 9 of the XPC gene. The XPC-PAT+ variant consists of an insertion of 83 A/T bases along with a 5 base deletion at position 1457–1461.15 The PAT+ variant is adjacent to a retrotransposon and may be involved in exon shuffling or abnormal splicing,15 but this has not yet been established. It is in strong linkage disequilibrium with a common C/A polymorphism located at the −5 position of intron 11 which is a splice acceptor site and triggers a deletion that entails the skipping of exon 12 and leads to diminished DNA repair activity in vitro.16, 17 The PAT+/+ genotype has been associated with lymphocyte chromosome aberrations,18 reduced capacity to repair DNA damage induced by UVR19 and increased mutagen sensitivity.20 To our knowledge there have been no studies of the XPC-PAT+ variant and breast cancer risk.
The ERCC2 gene, located at chromosome 19q13.2-q13.3, encodes a DNA helicase that unwinds DNA for purposes of gene transcription or NER. ERCC2 proteins are required for the p53-mediated apoptotic response, and expression levels correlate with DNA repair capacity.21 An ERCC2 polymorphism (A35931C) at codon 751 in exon 23 (rs1052559) changes the charge of the amino acid and occurs in the domain in which the ERCC2 protein interacts with p44 protein, its helicase activator. ERCC2 751Gln has been associated with increased levels of DNA adducts22–24 and reduced DNA repair capacity.1, 25 It has also been associated with cancer at several other sites.26–29
Material and methods
Study population and epidemiologic methods
Between 1985 and 1991, the NYU Women's Health Study enrolled 14,275 healthy women aged 35–65 years at a breast cancer screening center in New York City. Further details concerning study eligibility and sample collection and storage have been reported previously.30 Data on the use of cigarettes and alcohol were collected in early follow-up questionnaires. The cohort has had both active (questionnaire contact every several years) and passive (tumor registries and National Death Index) follow-up,30 and cancer ascertainment is estimated to be 95% complete.31 Only incident cases (i.e., diagnosed at least 6 months after initial blood donation) of invasive breast cancer were included in our study. One control was selected at random among the matched risk set for each case – matched on age, date and menopausal status at time of cohort enrollment, as detailed previously.30 Informed consent was obtained at enrollment into the study when blood was collected, and the study has been approved annually by the IRB of New York University School of Medicine.
Blood clots and/or cell aggregates (sediment remaining after centrifugation to obtain serum specimens) were available for 48% of the women in the study. For the remaining women, only serum specimens were available, but it was possible to isolate enough DNA from the serum specimens to evaluate polymorphisms in nearly all the subjects. DNA was isolated from 200 μl of all samples (clots, cell aggregates or sera) using Qiagen QIAamp Blood Mini Kits. DNA from clots and cell aggregates was diluted 1:10 but sera were used undiluted in PCR amplifications.
The XPC-PAT+ variant consists of an insertion in intron 9 of 83 A/T bases (ATA AATTTTTTAT AATAATTTAT AATATTTATA AATTATATAT ATTTATATAT ATAAATAAAT TTATAATATT TATAAATATT) along with a 5 base deletion (GTAAC) at position 1457–1461 (Genbank Accession No. AF076952).15 For genotyping the XPC-PAT variant, the forward primer was 5′-AAGGAAGCATTAGAGGTG-3′ and reverse was 5′-GAGATGGATGGTGGTGAT-3′. The size of the common allele was 155 bp while the variant amplicon was 233 bp.
The forward and reverse primers for the ERCC2 polymorphism in exon 23 (position 35931 A > C; rs13181) were as follows: 5′-ATCCTGTCCCTACTGGCCAT-3′ and 5′-TGTGGACGTGACAGTGAGAA-3′, respectively. PCR amplicons were digested with the restriction enzyme PstI (New England Biolabs). Because the reliability of the ERCC2 genotyping was marginal for the serum samples using RFLP analysis, we reanalyzed these samples using Pyrosequencer™ technology which produced 99% agreement between serum and clot samples. For Pyrosequencing, PCR primers were forward 5′-CTCAGCCTGGAGCAGCTAGAAT-3′ and reverse 5′-ACAGAGAAGCCAGAAGTAT-3′.
The chi-square goodness of fit test was used to assess deviation from Hardy–Weinberg equilibrium. In the absence of a priori knowledge of the appropriate genetic model, we examined 3 possible models, i.e., dominant (indicator variable scores for wild-type, heterozygous and homozygous variant of 0, 1, 1, respectively), additive (gene-dose-dependent – scores of 0, 0.5, 1), and recessive (scores of 0, 0, 1). Because of the matched-pair study design, conditional logistic regression analyses were conducted using SAS. All analyses were adjusted for race/ethnicity (Jewish, other Caucasian (except Hispanic), African American, Hispanic and other minorities, unknown). Hispanic and other minorities were grouped together because of their small numbers. Analyses adjusting simultaneously for (ln) body mass index (BMI), (ln) height, parous vs. nulliparous, age at first pregnancy, age at menarche, first degree family history of breast cancer and bilateral oophorectomy are also presented. The potential confounding effects of history of oral contraceptive and hormone replacement use were also examined but these variables were not included in the multivariate model because they had little effect on the odds ratios (OR), which are reported with 95% confidence intervals (CI).
A priori planned analyses of interactions were performed between the DNA repair genes and 2 exposures that produce DNA damage, smoking and alcohol use. The p-value for interaction corresponds to the likelihood ratio test comparing the model including an interaction term (in addition to the main effect terms for the DNA repair gene and for ever/never smoking (or regular alcohol use) to the model excluding it.
There were 612 matched case–control pairs. The cases averaged 59.5 ± 9.3 years of age at diagnosis; 205 (34%) cases were under age 55. Table I summarizes their demographic, reproductive and lifestyle characteristics. The groups were comparable with respect to age, smoking, alcohol use, oral contraceptive use, hormonal replacement use and height. There were nonsignificant differences in the expected direction with respect to age at menarche, nulliparity, age at first full-term pregnancy, BMI and family history of breast cancer. There were small differences between the 2 groups with respect to ethnicity (African Americans and Hispanics/Asians had a lower risk of breast cancer) and postmenopausal BMI (higher in breast cancer cases than in controls).
Table I. Characteristics of the Breast Cancer Cases and Controls
|Ethnicity2|| || |
| Jewish||254 (41)||232 (38)|
| Other non-Hispanic Caucasians||222 (36)||202 (33)|
| African–American||50 (8)||59 (10)|
| Hispanic, Asian and other||33 (6)||54 (8)|
| Unknown||53 (9)||65 (11)|
|Age at breast cancer diagnosis (or the corresponding index age for controls) (years)|
| <45||38 (6)||38 (6)|
| 45–54||167 (27)||167 (27)|
| 55–64||205 (33)||205 (33)|
| 65–79||202 (33)||202 (33)|
|Age first full-term pregnancy/ nulliparity (years)|| || |
| Age <25||142 (23)||183 (30)|
| Age 25–29||168 (27)||180 (29)|
| Age ≥30||101 (17)||69 (11)|
| Nulliparous||210 (33)||180 (29)|
|Age at menarche|| || |
| Age <13||309 (51)||286 (47)|
| Age 13+||300 (49)||317 (53)|
|First-degree family history of breast cancer|| || |
| None||468 (76)||475 (78)|
| One affected relative, age 45+||76 (12)||85 (14)|
| >1 affected relative, or 1 age <45||68 (11)||52 (8)|
|Bilateral oophorectomy||84 (14)||74 (12)|
| History of oral contraceptive use3||198 (37)||194 (35)|
| History of hormone replacement therapy3||121 (23)||126 (23)|
|Smoking3|| || |
| Never||253 (49)||253 (48)|
| Past||171 (33)||171 (32)|
| Current||94 (18)||107 (20)|
|Daily alcohol consumption3||68 (13)||76 (14)|
|Height (cm)—Mean ± SD4||162.4 ± 6.6||162.2 ± 6.3|
|Weight (kg)—Mean ± SD4||66.5 ± 12.5||65.4 ± 12.0|
|BMI—Mean ± SD4|| || |
| Age <52||24.2 ± 4.6||24.2 ± 4.5|
| Age ≥522||26.1 ± 4.1||25.5 ± 4.4|
It was possible to obtain an adequate reading on 97.5% of the samples for the XPC genotype and 99.9% for the ERCC2 genotype. As a result, 583 matched pairs could be analyzed for XPC and 611 for ERCC2. A blinded reliability study of 85 paired serum and clot/cell-aggregate biospecimens showed 99% agreement for the ERCC2 locus and 96% for the XPC-PAT+ polymorphism. The overall control group allele frequencies of 35% for the ERCC2 751 A allele and 40% for the XPC-PAT+ allele were similar to other studies.17, 27, 32–35 The alleles were in Hardy-Weinberg equilibrium for both genes (Chi-square goodness of fit, p = 0.79 and p = 0.97, respectively). African Americans had a significantly lower frequency of the XPC-PAT+ variant allele (31%) and the ERCC2 751 A allele (22%) than Caucasians (44 and 37%, respectively), which emphasizes the importance of our having controlled race/ethnicity in the analyses.
Analyses of XPC-PAT+
There was no indication of excess breast cancer risk according to genotype when the dominant or additive genetic models were assumed (Table II). However, the recessive model was statistically significant. With adjustment for race/ethnicity and other potential confounder variables (details in Table II, footnote 2) there was a 45% excess breast cancer risk (XPC-PAT+/+ OR = 1.45, CI: 1.07–1.97) among those homozygous for the variant allele as compared to others. The coefficients for the race/ethnicity indicator variables did not approach statistical significance, which suggested that important confounding by population admixture was unlikely.
Table II. Genotype Frequencies and ORs (95% CIs) for Dominant, Additive and Recessive Genetic Models
|XPC-PAT −/−||211 (36%)||215 (37%)|
| +/−||244 (42%)||271 (46%)|
| +/+||128 (22%)||97 (17%)|
|ERCC2 CC||251 (41%)||259 (42%)|
| AC||292 (48%)||282 (46%)|
| AA||68 (11%)||70 (11%)|
| ||Odds ratios (95% CI)|
|Ethnic adjusted1||Ethnic + Other2|
|XPC-PAT|| || |
| Dominant model||1.01 (0.79–1.29)||0.99 (0.77–1.27)|
| Additive model||1.23 (0.89–1.69)||1.20 (0.86–1.67)|
| Recessive model||1.41 (1.04–1.91)||1.45 (1.07–1.97)|
|ERCC2|| || |
| Dominant model||0.90 (0.72–1.14)||0.89 (0.70–1.13)|
| Additive model||0.90 (0.63–1.28)||0.87 (0.61–1.24)|
| Recessive model||1.02 (0.70–1.48)||0.96 (0.65–1.41)|
The interaction of age and XPC-PAT+/+ with respect to breast cancer risk was not statistically significant (p = 0.54). The OR was 1.24 (CI: 0.73–2.10) among breast cancer cases diagnosed before age 55 years and 1.53 (CI: 1.06–2.21) among cases diagnosed at ages 55+.
Two DNA damaging agents—smoking and alcohol—were examined in relation to the XPC-PAT genotypes to determine if there were joint gene-environmental effects upon breast cancer risk. Information on alcohol usage and smoking was unavailable for 11 and 14% of the subjects, respectively. The test of interaction between alcohol consumption and the XPC-PAT+/+ was not statistically significant (not shown). However, there was evidence for an interaction between the XPC polymorphism and smoking (p = 0.09), implying that XPC-PAT+/+ may heighten breast cancer risk about 80% more among smokers than nonsmokers. Table III summarizes this effect modification: using XPC wild-type nonsmokers as the reference group, there was no increased risk in XPC wild-type smokers or in XPC-PAT+/+ non-smokers but a borderline elevation was seen among the XPC-PAT+/+ who were ever-smokers (OR = 1.56, CI: 0.95–2.58; p = 0.08). The XPC-PAT+/+ risks were higher among current smokers of 10 or more cigarettes per day, but the CI were wide due to the smaller numbers (OR = 2.1, CI: 0.6–7.8).
Table III. Interaction of XPC-PAT+/+ and Smoking on Breast Cancer Risk1
|XPC-PAT wild-type or heterozygous||189/1952||193/231|
|1.0 (Referent)||0.85 (0.62–1.16)|
|XPC-PAT homozygous variant||48/48||62/37|
|1.02 (0.62–1.67)||1.56 (0.95–2.58)|
Analyses of ERCC2 Lys751Gln (C/A)
For none of the models—dominant, recessive, or additive—was there any indication of elevated breast cancer risk associated with the A allele (Table II). Separate analyses for breast cancer cases diagnosed before or after age 55 did not show significant effects either. There were also no indications of any interaction with smoking or alcohol consumption with respect to breast cancer risk.
The strengths of the present study include the prospective nature of the cohort and high case ascertainment rate. This nested case–control study within a cohort minimizes selection biases that may be present in other types of case–control studies. To help control for population stratification, analyses were adjusted for ethnic group. The ethnic indicator variables were far from statistical significance in the main analyses, suggesting that ethnicity is unlikely to be an important source of confounding. Moreover, when we restricted the analysis to non-Hispanic Caucasians the results were very similar (not shown). Nevertheless, the smaller minor allele frequencies for African Americans as compared to Caucasians for both the XPC-PAT+ and ERCC2 codon 751Gln alleles are interesting differences that merit further investigation in studies with larger numbers of African-Americans.
Although there have been no reports regarding the XPC-PAT+ variant and breast cancer risk, several studies have examined breast cancer and the Lys939Gln variant in exon 15 of the XPC gene, which is in partial linkage disequilibrium with the PAT+ allele, with mixed results.3, 36–38 Two studies reported no association between breast cancer risk and an Ala499Val variant in exon 8 of the XPC gene.2, 38
Since the XPC-PAT+ polymorphism occurs in an intronic sequence, its association with breast cancer risk may be because of linkage disequilibrium with another XPC polymorphism. The C/A polymorphism located at the −5 position of intron 11 is a possible candidate since it is at a splice acceptor site which causes a deletion and skipping of exon 12 and is in strong linkage disequilibrium with the XPC-PAT+ polymorphism.16, 17, 39
The finding that the homozygous XPC-PAT+/+ genotype shows decreased DNA NER capacity compared to the other genotypes19 is consistent with finding an interaction with smoking, since the polycyclic aromatic hydrocarbons and nitrosamines in cigarette smoke cause bulky DNA adducts and crosslinks are removed primarily by NER,40 which is triggered by the XPC gene. Also of relevance are the associations of the XPC-PAT+ variant with other smoking related cancers: head and neck,33 lung17 and bladder.35, 41
Seven studies of the ERCC2 codon 751Gln allele and breast cancer have provided mixed evidence regarding an association, ranging from marginally positive to null to inverse.6–8, 23, 42–44 Our study is in accordance with the conclusion that the association of breast cancer with the ERCC2 751Gln allele is null or of small magnitude.
In summary, the findings of special interest are the association between the XPC-PAT+/+ genotype and breast cancer risk and the suggestion that the association occurs primarily because of an interaction with cigarette smoking. Further investigation is needed to confirm the association of breast cancer risk with variation in XPC associated with the PAT+, 939Gln and intron 11 A allele haplotype and its interaction with smoking.
The assistance of Ms. Daniela Masciangelo, Ms. Denise Heimowitz, Ms. Lynne Quinones, Ms. Hiroko Meserve and Ms. LaVerne Yee in management of the samples, documentation of the breast cancer cases, literature review and clerical help is gratefully acknowledged.