Abstract
Recently, we reported that among Singapore Chinese, cigarette smoking and alcohol drinking were independent risk factors for colorectal cancer. Both tobacco smoking and alcohol use are plausible colorectal cancer risk factors, partly due to their ability to induce mutations in the colorectal lumen. In the present study, we investigated the role in colorectal cancer of single-nucleotide polymorphisms in five DNA repair genes: XRCC1 (Arg194Trp and Arg399Gln), PARP (Val762Ala, Lys940Arg), XPD (Asp312Asn, Lys751Gln), OGG1 (Ser326Cys), and MGMT (Leu84Phe). We conducted this study within the Singapore Chinese Health Study, a population-based cohort of 63,257 middle-aged and older Singapore Chinese men and women enrolled between 1993 and 1998. Our study included 1,176 controls and 310 cases (180 colon and 130 rectum cancer). We observed a positive association between the PARP codon 940 Lys/Arg and Arg/Arg genotypes and colorectal cancer risk [odds ratio (OR), 1.8; 95% confidence interval (95% CI), 1.1-3.1], and an inverse association between the MGMT codon 84 Leu/Phe or Phe/Phe genotypes and colon cancer risk (OR, 0.6; 95% CI, 0.3-0.9), but not rectal cancer (test of heterogeneity by tumor site, P = 0.027). We observed evidence that XRCC1 may modify the effects of smoking (interaction P = 0.012). The effect of smoking among carriers of the Arg194-Gln399 haplotype was OR = 0.7 (95% CI, 0.4-1.1), whereas, among carriers of the Trp194-Arg399 haplotype, it was OR = 1.6 (95% CI, 1.1-2.5). We also observed a nonstatistically significant modification of XRCC1 on the effects of alcohol (P = 0.245). Whereas alcohol had no effect among carriers of the codon 194 Arg/Arg (OR, 1.0; 95% CI, 0.6-1.7) or Arg/Trp genotypes (OR, 1.1; 95% CI, 0.6-1.9), there was a positive association among carriers of the Trp/Trp genotype (OR, 2.8; 95% CI, 1.0-8.1). Our results support a role for reactive oxygen species as relevant genotoxins that may account for the effects of both smoking and alcohol on colorectal cancer risk. (Cancer Epidemiol Biomarkers Prev 2007;16(11):2363–72)
Introduction
Singapore Chinese are historically at low risk for colorectal cancer (1, 2). However, rates for this common malignancy in the West have steadily increased in newly affluent Asian Chinese societies, including that in Singapore where rates in both sexes have doubled in the past three decades (1). Recently, we reported that, among Singapore Chinese, cigarette smoking and alcohol drinking were independent risk factors for colorectal cancer (2).
Both tobacco smoking and alcohol use are plausible colorectal cancer risk factors. Smoking can lower levels of antioxidants, such as carotenoids (3), induce expression of proteins related to colon tumor progression and invasion, such as 5-lipooxygenase, vascular endothelial growth factor, and metalloproteinases (matrix metalloproteinase-2 and matrix metalloproteinase-9) in colon cancer cells (4), and decrease levels of circulating folate, a protective factor for the formation of colorectal adenomas, precursors of colorectal cancer (5). Furthermore, smoking is a rich source of genotoxic agents, such as reactive oxygen species and chemical carcinogens. Reactive oxygen species can initiate lipid peroxidation, oxidize proteins, and generate single- and double-strand breaks, abasic sites, and base adduct formation, such as thymine glycol, 5-hydrozymethyluracil, and 8-hydroxy-2-deoxyguanosine (6). Chemical carcinogens in cigarette smoke (i.e., polycyclic aromatic hydrocarbons, nitrosamines, arylamines) can induce bulky adducts in crypt cells, which contribute to the formation of mutations in the colon.
Absorption of tobacco carcinogens is facilitated by alcohol (7). Furthermore, once in the colon lumen, alcohol is converted to acetaldehyde. The latter can form DNA adducts, such as N2-ethyl-2′-deoxyguanosine and 1, N2-propanodeoxyguanosine (8), inter crosslinks (8), and induce oxidative DNA damage, such as DNA strand breaks (9). The presence of base adducts, such as N2-ethyl-2′-deoxyguanosine and 1, N2-propanodeoxyguanosine, can induce point mutations in DNA, and 1, N2-propanodeoxyguanosine adducts can also act as topoisomerase II poisons, leading to an increase of unresolved double-strand breaks induced by this enzyme (10). Acetaldehyde can inhibit the action of O6-methylguanine-methyltransferase (MGMT), which is an important enzyme for the removal of alkylating damage such as that induced by cigarette smoking, thus contributing to further accumulation of DNA damage (9). In addition, ethanol has been shown to impair the repair of strand breaks and bulky adducts (11, 12), to induce CYP2E1, which has a high redox potential and can induce oxidative stress and lipid peroxidation in the colon, thus contributing to further DNA damage (9), and to reduce levels of circulating folate, which may contribute to the formation of abasic sites (9).
Given that both smoking and alcohol are important sources of DNA damage and that they likely exert an effect on colorectal carcinogenesis at the expense of inducing mutations in the colorectal lumen, a role for DNA repair gene variants, as susceptibility genes and effect modifiers, is plausible. In the present study, we investigated the role of single-nucleotide polymorphisms (SNP) in genes that participate in the base excision repair pathway (XRCC1, Arg194Trp and Arg399Gln; OGG1, Ser326Cys; and PARP, Val762Ala and Lys940Arg) which repairs smoking- and alcohol-induced oxidative DNA base modifications and single-strand breaks, the nucleotide excision repair pathway (XPD, Asp312Asn and Lys751Gln) which repairs smoking- and acetaldehyde-induced adducts, and the reversal of damage pathway (MGMT, Leu84Phe) which repairs smoking-induced alkylated DNA damage. These SNPs were selected based on putative effect on protein function and/or previous evidence of cancer risk associations. We undertook a multistage approach that allowed us to understand the contribution of each SNP to colorectal cancer, the combined effect of multiple SNPs per gene and multiple genes per pathway on colorectal cancer, and their potential role as effect modifiers of smoking or alcohol.
Materials and Methods
Study Subjects
The design of the Singapore Chinese Health Study has been described (13). Briefly, the cohort was drawn from permanent residents or citizens of Singapore who resided in government-built housing estates (86% of the Singapore population resided in such facilities during the enrollment period). The age eligibility criterion was 45 to 74 years. We restricted recruitment to the two major dialect groups of Chinese in Singapore, the Hokkiens and the Cantonese. Between April 1993 and December 1998, 63,257 subjects (∼85% of eligible subjects) were recruited. At recruitment, information on life-style factors and diet during the last year were obtained through in-person interviews. The study was approved by the Institutional Review Boards of University of Minnesota, University of Southern California, and National University of Singapore.
Incident cancer cases among cohort members were identified through record linkage with the nationwide Singapore Cancer Registry database (14). As of April 2002, when subjects for this study were identified, 592 colorectal cancer cases had been diagnosed among cohort participants. All cases were histologically confirmed, except for three who were ascertained by death records and clinical evidence. Details of the biospecimen collection, processing, and storage procedures from the cases and controls in this study have been previously described (15). Briefly, blood or buccal cells and single-void urine specimens were collected from a random 3% sample of cohort participants between April 1994 and July 1999 for a total of 1,194 subjects. Of these, 13 subjects developed colorectal cancer by April 2002, and five controls with prevalent history of colorectal cancer at recruitment were subsequently excluded from the study. The remaining subjects (n = 1,176) constituted our control group for this study.
In addition, we also attempted to collect blood/buccal cells and urine samples from all incident colorectal cancer cases. Of the 592 colorectal cancer cases, 312 (52%) donated blood/buccal cell samples for genetic analyses. Two cases were later excluded due to misclassification of cancer type. These 310 cases with available DNA samples available (colon cancer, 180; rectum cancer, 130) constituted the case group of our study. Compared with colorectal cancer patients who did not donate a blood or buccal sample, those who donated had a similar mean of age at cancer diagnosis (65.4 versus 66.1 years). Male patients were more likely to donate a biospecimen than female patients (57% versus 49%), as did the patients with Cantonese dialect compared with those with Hokkien dialect (45% versus 37%). Patients who did not donate a blood or buccal sample were less educated than those who did (40% versus 31% had no formal education). Among colorectal cases who donated biospecimens, there was lower prevalence of advance stages (3 and 4) compared with cases who did not donate biospecimens (46.8% versus 62%). There was no difference in the percentage of biospecimen availability by level of body mass index, cigarette smoking, alcohol drinking, history of diabetes, or family history of colorectal cancer.
Baseline Exposure Assessment
At recruitment, an in-person interview was conducted in the home of the subject by a trained interviewer using a structured questionnaire. The questionnaire requested information on demographics, lifetime use of tobacco (cigarettes and water pipe), current physical activity, reproductive history (women only), occupational exposure, medical history, and family history of cancer. Information on current diet, including alcohol consumption, was assessed via a 165-item food frequency questionnaire that has been validated against a series of 24-h dietary recall interviews (13) and selected biomarker studies (16, 17) conducted on random subsets of cohort participants. The Singapore Food Composition Table, developed in conjunction with this cohort study, allows for the computation of intake levels of roughly 100 nutritive and nonnutritive compounds per study subject (13).
For each of the four types of alcoholic beverages (beer, wine, Western hard liquor, and Chinese hard liquor), participants were asked to choose from eight frequency categories: never or hardly, once a month, two to three times a month, weekly, two to three times a week, four to six times a week, daily, and two or more times a day. Consumers were then asked to choose from four defined portion sizes. For beer, the portion sizes were one small bottle (375 mL) or less, two small bottles or one large bottle (750 mL), two large bottles, and three large bottles or more. For wine, the portion sizes were one glass (118 mL) or less, two, three, and four glasses or more. For Chinese or Western hard liquor, the portion sizes were one shot (30 mL) or less, two, three, and four shots or more. One drink was defined as 375 mL of beer (13.6 g ethanol), 118 mL of grape wine (11.7 g ethanol), and 30 mL of Western or Chinese hard liquor (10.9 g ethanol).
For cigarette smoking, subjects were first asked if he/she had ever smoked at least one cigarette a day for 1 year or longer. Subjects who answered “no” were classified as nonsmokers. Subjects who answered “yes, but I quit smoking” were classified as “former smokers,” and those who answered “yes, and I currently smoke” were classified as “current smokers.” Ever smokers were then asked about age at smoking initiation (four categories: <15, 15-19, 20-29, and 30 years or older), the number of cigarettes smoked per day (six categories: 6 or less, 7-12, 13-22, 23-32, 33-42, and 43 or more), and duration of smoking (four categories: <10, 10-19, 20-39, 40, or more years). As reported previously (2), subjects who began to smoke cigarettes before 15 years of age and smoked 13 or more cigarettes per day were classified as “heavy” smokers. The remaining ever smokers were grouped as “light” smokers in the present analysis. We defined a “smoking index” by classifying subjects as never, light, or heavy smokers. Physical activity was assessed by asking subjects to estimate the number of hours spent watching TV per day and the numbers of hours per week spent on moderate activities (such as brisk walking, bowling, bicycling on level ground, tai chi, or chi kung), strenuous sports (such as jogging, bicycling on hills, tennis, squash, swimming laps, or aerobics), or vigorous work (such as moving heavy furniture, loading or unloading trucks, or shoveling).
Genotyping
We obtained genotypes on eight SNPs across these five genes XRCC1 Arg194Trp, XRCC1 Arg399Gln, PARP Val762Ala, PARP Lys940Arg, OGG1 Ser326Cys, XPD Asp312Asn, XPD Lys751Gln, and MGMT Leu84Phe using TaqMan assays, which were all read using an ABI 7900HT sequence detection system, and scored using Sequence Detector software (Applied Biosystems, Inc.). Primers, probes, and PCR conditions are available from the authors upon request. For quality control purposes, 5% randomly selected samples were duplicates with unique identification numbers blind to laboratory personnel and uniformly dispersed across all genotyping plates. We observed 100% concordance between all duplicate samples. The calling rate for all assays was 98% to 100%, except for the XRCC1 codon 399 assay for which we had a call rate of 95%.
Data Analysis
Data were analyzed using standard methods for unmatched case-control studies. Although we sampled our controls from the whole cohort, this study is more case-control than case-cohort in design, because the time period of follow-up was comparable between the cases and subcohort, with only 13 subjects in the latter group developing colorectal cancer during the observation period. Nonetheless, in previous analyses using this same dataset, we conducted parallel analyses using standard case-control and case-cohort methods, and results did not materially differ (18). Therefore, in current studies using this dataset, we present results based on case-control analyses. We compared the observed genotypic frequencies to those expected under the Hardy-Weinberg law using χ2 tests. For all SNPs, <2% of subjects (0.4-1.9%) had missing genotype information, with the exception of XRCC1 codon 399 for which data was missing for 4.8% of subjects. Subjects with missing data were excluded from all analyses. We estimated gene main effects [odds ratios (OR) and 95% confidence intervals (95% CI)] using unconditional logistic regression adjusting for age at recruitment (years), gender, dialect group (Hokkien, Cantonese), and year of recruitment (1993-1995, 1996-1998). We also consider as potential confounders the following factors which have been found to be associated with colorectal cancer risk within the Singapore Chinese Health Study: level of education (no formal education, primary school, secondary or higher education), body mass index (<20, 20 to <24, 24 to <28, 28+ kg/m2), cigarette smoking (smoking index: never smokers, light smokers, or heavy smokers), alcohol intake (nondrinkers, <7 drinks per week, or 7+ drinks per week), history of diabetes (no, yes), family history of colorectal cancer (no, yes), and weekly vigorous physical activities (no, yes). Inclusion of these variables did not change the gene effect OR estimates by >10%; therefore, we did not keep these variables in the final gene main effect models.
We determined the gene mode of action for each of the putative functional variants by using model selection according to the Aikake Information Criteria (AIC) by comparing fitted models assuming an additive, dominant, or recessive mode of action (19). After ranking all models, we selected the genetic model with the lowest AIC values which provided the most parsimonious fit to the data. For the three genes for which we had information on two SNPs (XRCC1, PARP, XPD), we also obtained ORs and 95% CIs for each of the four haplotypes for each genes. We use the EM algorithm to estimate the expected haplotype probability for the double heterozygous individuals (20). We then estimated the association between expected haplotype values and colorectal cancer risk and did a global test for each gene by fitting an unconditional logistic regression model with all four haplotypes adjusting for age at recruitment (years), gender, dialect group (Hokkien, Cantonese), and year of recruitment (1993-1995, 1996-1998) and comparing to a model without the haplotype variables (21).
We tested for gene-smoking and gene-alcohol interactions using logistic regression models adjusting for age at recruitment (years), gender, dialect group (Hokkien, Cantonese), year of recruitment (1993-1995, 1996-1998), level of education (no formal education, primary school, secondary or higher education), body mass index (<20, 20 to <24, 24 to <28, 28+ kg/m2), history of diabetes (no, yes), family history of colorectal cancer (no, yes), and weekly vigorous physical activities (no, yes). Alcohol analyses were further adjusted for a three-level smoking index (never, light, or heavy smokers); smoking analyses were further adjusted by a three-level alcohol index (nondrinkers, <7, 7+ drinks per week). We tested for gene-exposure interactions on a multiplicative scale by including in our regression model product terms between the gene and exposure variables in addition to the terms present in the main effect model and used likelihood ratio tests to compare these models to models that assumed no interaction. We tested for heterogeneity of the gene effect ORs by tumor site (colon versus rectum) and G × smoking or G × alcohol heterogeneity by tumor site using polytomous logistic regression models and likelihood ratio tests. For XRCC1 × smoking analyses using model selection with AIC, we imputed missing genotypes using the expected value under the determined mode of action. We corrected gene main effect and G × E analyses for multiple comparisons using the Benjamini Hochberg method for controlling the false discovery rate (22). All tests were two-sided, and all analyses were done using the statistical software STATA version 8 (STATA Corporation).
Results
We present in Table 1 demographic and descriptive statistics of cases and controls. We analyzed 310 colorectal cancer cases (180 colon and 130 rectal cases) and 1,176 controls. Cases and controls did not significantly differ by dialect group or level of physical activity. Compared with controls, cases had a higher proportion of men (57% in cases versus 43% in controls, P < 0.001), lower proportion of subjects with secondary school or higher education (21% in cases versus 30% in controls, P = 0.004), higher proportion of subjects with a family history of colorectal cancer (4% in cases versus 2% in controls, P = 0.08), and higher prevalence of diabetes (16% in cases versus 9% in controls, P < 0.001).
Demographics and descriptive statistics of cases and controls
. | Controls, n (%) . | Colon, n (%) . | Rectum, n (%) . | All cases, n (%) . | ||||
---|---|---|---|---|---|---|---|---|
Mean age at interview (±SD), y | 56.5 ± 8.1 | 61.6 ± 7.4 | 60.9 ± 7.6 | 61.3 ± 7.6 | ||||
Dialect group | ||||||||
Cantonese | 571 (49) | 83 (46) | 56 (43) | 139 (45) | ||||
Hokkien | 605 (51) | 97 (54) | 74 (57) | 171 (55) | ||||
Gender | ||||||||
Female | 667 (57) | 87 (48) | 46 (35) | 133 (43) | ||||
Male | 509 (43) | 93 (52) | 84 (65) | 177 (57) | ||||
Level of education | ||||||||
No formal education | 317 (27) | 61 (34) | 33 (26) | 94 (30) | ||||
Primary school | 505 (43) | 84 (47) | 68 (52) | 152 (49) | ||||
Secondary or higher education | 354 (30) | 35 (19) | 29 (22) | 64 (21) | ||||
Family history of colorectal cancer | ||||||||
No | 1,149 (98) | 172 (96) | 125 (96) | 297 (96) | ||||
Yes | 27 (2) | 8 (4) | 5 (4) | 13 (4) | ||||
Diabetes | ||||||||
No | 1,067 (91) | 149 (83) | 112 (86) | 261 (84) | ||||
Yes | 109 (9) | 31 (17) | 18 (14) | 49 (16) | ||||
Weekly physical activity | ||||||||
No | 768 (65) | 120 (67) | 88 (68) | 208 (67) | ||||
Yes | 408 (35) | 60 (33) | 42 (32) | 103 (33) | ||||
Smoking index | ||||||||
Never smokers | 850 (72) | 118 (65) | 66 (50) | 184 (60) | ||||
Light smokers | 294 (25) | 48 (27) | 49 (38) | 97 (31) | ||||
Heavy smokers | 32 (3) | 14 (8) | 15 (12) | 29 (9) | ||||
Alcohol intake | ||||||||
Never drinkers | 965 (82) | 149 (83) | 94 (72) | 234 (78) | ||||
<7 drinks/wk | 164 (14) | 17 (9) | 25 (19) | 42 (14) | ||||
7+ drinks/wk | 47 (4) | 14 (8) | 11 (9) | 25 (8) |
. | Controls, n (%) . | Colon, n (%) . | Rectum, n (%) . | All cases, n (%) . | ||||
---|---|---|---|---|---|---|---|---|
Mean age at interview (±SD), y | 56.5 ± 8.1 | 61.6 ± 7.4 | 60.9 ± 7.6 | 61.3 ± 7.6 | ||||
Dialect group | ||||||||
Cantonese | 571 (49) | 83 (46) | 56 (43) | 139 (45) | ||||
Hokkien | 605 (51) | 97 (54) | 74 (57) | 171 (55) | ||||
Gender | ||||||||
Female | 667 (57) | 87 (48) | 46 (35) | 133 (43) | ||||
Male | 509 (43) | 93 (52) | 84 (65) | 177 (57) | ||||
Level of education | ||||||||
No formal education | 317 (27) | 61 (34) | 33 (26) | 94 (30) | ||||
Primary school | 505 (43) | 84 (47) | 68 (52) | 152 (49) | ||||
Secondary or higher education | 354 (30) | 35 (19) | 29 (22) | 64 (21) | ||||
Family history of colorectal cancer | ||||||||
No | 1,149 (98) | 172 (96) | 125 (96) | 297 (96) | ||||
Yes | 27 (2) | 8 (4) | 5 (4) | 13 (4) | ||||
Diabetes | ||||||||
No | 1,067 (91) | 149 (83) | 112 (86) | 261 (84) | ||||
Yes | 109 (9) | 31 (17) | 18 (14) | 49 (16) | ||||
Weekly physical activity | ||||||||
No | 768 (65) | 120 (67) | 88 (68) | 208 (67) | ||||
Yes | 408 (35) | 60 (33) | 42 (32) | 103 (33) | ||||
Smoking index | ||||||||
Never smokers | 850 (72) | 118 (65) | 66 (50) | 184 (60) | ||||
Light smokers | 294 (25) | 48 (27) | 49 (38) | 97 (31) | ||||
Heavy smokers | 32 (3) | 14 (8) | 15 (12) | 29 (9) | ||||
Alcohol intake | ||||||||
Never drinkers | 965 (82) | 149 (83) | 94 (72) | 234 (78) | ||||
<7 drinks/wk | 164 (14) | 17 (9) | 25 (19) | 42 (14) | ||||
7+ drinks/wk | 47 (4) | 14 (8) | 11 (9) | 25 (8) |
Smoking, Alcohol, and Colorectal Cancer Risk
Previously, using the entire cohort database, we reported on the associations between tobacco/alcohol use and colorectal cancer risk (2). Table 1 shows the frequencies of cases and controls by levels of smoking and alcohol consumption within the present nested case-control study set. As expected, our results are consistent with those based on the entire cohort. Compared with never smokers, ORs (95% CIs) for light and heavy smokers were respectively 0.9 (95% CI, 0.6-1.3) and 2.6 (95% CI, 1.4-4.7) for colorectal cancer, 0.7 (95% CI, 0.5-1.1) and 2.1 (95% CI, 1.0-4.4) for colon cancer, and 1.2 (95% CI, 0.8-2.0) and 3.3 (95% CI, 1.6-7.0) for rectal cancer. Similarly, the ORs (95% CIs) for light (<7 drinks per week) and heavy (7+ drinks per week) drinkers were respectively 1.0 (95% CI, 0.7-1.5) and 2.0 (95% CI, 1.1-3.6) for colorectal cancer, 0.7 (95% CI, 0.4-1.2) and 2.1 (95% CI, 1.1-4.3) for colon cancer, and 1.4 (95% CI, 0.8-2.4) and 1.9 (95% CI, 0.9-4.1) for rectal cancer, as relative to nondrinkers.
DNA Repair SNPs and Colorectal Cancer Risk
We genotyped all cases and controls for SNPs in the following genes: XRCC1 (Arg194Trp and Arg399Gln), XPD (Asp312Asn and Lys751Gln), PARP (Val762Ala and Lys940Arg), OGG1 (Ser326Cys), and MGMT (Leu84Phe). The expected genotypic frequencies for all eight SNPs did not significantly deviate from those expected under the Hardy-Weinberg law. Furthermore, allelic frequencies did not significantly differ by dialect group (Hokkien or Cantonese). We indicate allelic frequencies for controls for all variant alleles in Table 2.
DNA repair SNPs and colorectal cancer risk among Singapore Chinese
DNA repair SNPs . | Controls . | Cases . | OR* (95% CI) . | P . | Variant allele frequency† . | |||||
---|---|---|---|---|---|---|---|---|---|---|
XRCC1 codon 194 | ||||||||||
Arg/Arg | 586 | 161 | 1ref | |||||||
Arg/Trp | 467 | 117 | 0.9 (0.7-1.2) | 0.402 | ||||||
Trp/Trp | 109 | 27 | 0.8 (0.5-1.3) | 0.425 | Trp = 0.295 | |||||
Arg/Arg | 586 | 161 | 1ref | |||||||
Arg/Trp and Trp/Trp | 576 | 144 | 0.9 (0.7-1.1) | 0.321 | ||||||
XRCC1 codon 399 | ||||||||||
Arg/Arg | 607 | 167 | 1ref | |||||||
Arg/Gln | 428 | 112 | 1.0 (0.7-1.3) | 0.956 | ||||||
Gln/Gln | 85 | 15 | 0.6 (0.4-1.1) | 0.129 | Gln = 0.267 | |||||
Arg/Arg and Arg/Gln | 1035 | 279 | 1ref | |||||||
Gln/Gln | 85 | 15 | 0.6 (0.4-1.1) | 0.128 | ||||||
PARP codon 762 | ||||||||||
Val/Val | 381 | 93 | 1ref | |||||||
Val/Ala | 564 | 150 | 1.1 (0.8-1.5) | 0.628 | ||||||
Ala/Ala | 228 | 64 | 1.1 (0.8-1.6) | 0.667 | Ala = 0.435 | |||||
Val/Val | 381 | 93 | 1ref | |||||||
Val/Ala and Ala/Ala | 792 | 214 | 1.1 (0.8-1.4) | 0.596 | ||||||
PARP codon 940 | ||||||||||
Lys/Lys | 1114 | 281 | 1ref | |||||||
Lys/Arg | 50 | 21 | 1.7 (1.0-3.0) | 0.062 | ||||||
Arg/Arg | 1 | 2 | 7.0 (0.6-84) | 0.122 | Arg = 0.022 | |||||
Lys/Lys | 1114 | 281 | 1ref | |||||||
Lys/Arg and Arg/Arg | 51 | 23 | 1.8 (1.1-3.1) | 0.029 | ||||||
OGG1 codon 326 | ||||||||||
Ser/Ser | 183 | 35 | 1ref | |||||||
Ser/Cys | 537 | 152 | 1.4 (0.9-2.2) | 0.109 | ||||||
Cys/Cys | 439 | 116 | 1.4 (0.9-2.1) | 0.156 | Cys = 0.610 | |||||
Ser/Ser | 183 | 35 | 1ref | |||||||
Ser/Cys and Cys/Cys | 976 | 268 | 1.4 (0.9-2.1) | 0.107 | ||||||
XPD codon 312‡ | ||||||||||
Asp/Asp | 1074 | 274 | 1ref | |||||||
Asp/Asn and Asn/Asn | 94 | 27 | 1.1 (0.7-1.7) | 0.720 | Asn = 0.042 | |||||
XPD codon 751 | ||||||||||
Lys/Lys | 998 | 251 | 1ref | |||||||
Lys/Gln | 159 | 48 | 1.2 (0.8-1.8) | 0.306 | ||||||
Gln/Gln | 6 | 4 | 2.0 (0.5-7.7) | 0.303 | Gln = 0.074 | |||||
Lys/Lys | 998 | 251 | 1ref | |||||||
Lys/Gln and Gln/Gln | 165 | 52 | 1.2 (0.9-1.8) | 0.225 | ||||||
MGMT codon 84 | ||||||||||
Leu/Leu | 959 | 251 | 1ref | |||||||
Leu/Phe | 194 | 40 | 0.8 (0.5-1.2) | 0.252 | ||||||
Phe/Phe | 13 | 1 | 0.6 (0.1-4.3) | 0.568 | ||||||
Leu/Leu | 959 | 251 | 1ref | |||||||
Leu/Phe and Phe/Phe | 207 | 41 | 0.8 (0.5-1.2) | 0.222 | Phe = 0.094 |
DNA repair SNPs . | Controls . | Cases . | OR* (95% CI) . | P . | Variant allele frequency† . | |||||
---|---|---|---|---|---|---|---|---|---|---|
XRCC1 codon 194 | ||||||||||
Arg/Arg | 586 | 161 | 1ref | |||||||
Arg/Trp | 467 | 117 | 0.9 (0.7-1.2) | 0.402 | ||||||
Trp/Trp | 109 | 27 | 0.8 (0.5-1.3) | 0.425 | Trp = 0.295 | |||||
Arg/Arg | 586 | 161 | 1ref | |||||||
Arg/Trp and Trp/Trp | 576 | 144 | 0.9 (0.7-1.1) | 0.321 | ||||||
XRCC1 codon 399 | ||||||||||
Arg/Arg | 607 | 167 | 1ref | |||||||
Arg/Gln | 428 | 112 | 1.0 (0.7-1.3) | 0.956 | ||||||
Gln/Gln | 85 | 15 | 0.6 (0.4-1.1) | 0.129 | Gln = 0.267 | |||||
Arg/Arg and Arg/Gln | 1035 | 279 | 1ref | |||||||
Gln/Gln | 85 | 15 | 0.6 (0.4-1.1) | 0.128 | ||||||
PARP codon 762 | ||||||||||
Val/Val | 381 | 93 | 1ref | |||||||
Val/Ala | 564 | 150 | 1.1 (0.8-1.5) | 0.628 | ||||||
Ala/Ala | 228 | 64 | 1.1 (0.8-1.6) | 0.667 | Ala = 0.435 | |||||
Val/Val | 381 | 93 | 1ref | |||||||
Val/Ala and Ala/Ala | 792 | 214 | 1.1 (0.8-1.4) | 0.596 | ||||||
PARP codon 940 | ||||||||||
Lys/Lys | 1114 | 281 | 1ref | |||||||
Lys/Arg | 50 | 21 | 1.7 (1.0-3.0) | 0.062 | ||||||
Arg/Arg | 1 | 2 | 7.0 (0.6-84) | 0.122 | Arg = 0.022 | |||||
Lys/Lys | 1114 | 281 | 1ref | |||||||
Lys/Arg and Arg/Arg | 51 | 23 | 1.8 (1.1-3.1) | 0.029 | ||||||
OGG1 codon 326 | ||||||||||
Ser/Ser | 183 | 35 | 1ref | |||||||
Ser/Cys | 537 | 152 | 1.4 (0.9-2.2) | 0.109 | ||||||
Cys/Cys | 439 | 116 | 1.4 (0.9-2.1) | 0.156 | Cys = 0.610 | |||||
Ser/Ser | 183 | 35 | 1ref | |||||||
Ser/Cys and Cys/Cys | 976 | 268 | 1.4 (0.9-2.1) | 0.107 | ||||||
XPD codon 312‡ | ||||||||||
Asp/Asp | 1074 | 274 | 1ref | |||||||
Asp/Asn and Asn/Asn | 94 | 27 | 1.1 (0.7-1.7) | 0.720 | Asn = 0.042 | |||||
XPD codon 751 | ||||||||||
Lys/Lys | 998 | 251 | 1ref | |||||||
Lys/Gln | 159 | 48 | 1.2 (0.8-1.8) | 0.306 | ||||||
Gln/Gln | 6 | 4 | 2.0 (0.5-7.7) | 0.303 | Gln = 0.074 | |||||
Lys/Lys | 998 | 251 | 1ref | |||||||
Lys/Gln and Gln/Gln | 165 | 52 | 1.2 (0.9-1.8) | 0.225 | ||||||
MGMT codon 84 | ||||||||||
Leu/Leu | 959 | 251 | 1ref | |||||||
Leu/Phe | 194 | 40 | 0.8 (0.5-1.2) | 0.252 | ||||||
Phe/Phe | 13 | 1 | 0.6 (0.1-4.3) | 0.568 | ||||||
Leu/Leu | 959 | 251 | 1ref | |||||||
Leu/Phe and Phe/Phe | 207 | 41 | 0.8 (0.5-1.2) | 0.222 | Phe = 0.094 |
Abbreviation: ref, reference group.
Adjusted for dialect group (Hokkien or Cantonese), age at recruitment (y), gender, and year of recruitment (1993-1995, 1996-1998).
Frequencies estimated among controls both dialect groups combined.
There were no cases who carried the Asn/Asn genotype.
For all SNPs, we estimated the gene mode of action using model selection with AIC. All SNPs were predicted to have a dominant mode of action, with the exception of XRCC1 codon 399 and XPD codon 312, which were predicted to be recessive. Given that the XPD codon 312 Asn allele is rare in the population (4%), we did not consider the recessive genotype due to instability of model fitting and carried out further analysis assuming a dominant mode of action.
We present in Table 2 ORs and 95% CI for all SNPs investigated using both the additive and the best-fit models (see previous paragraph). We observed a statistically significant positive association between the PARP codon 940 Lys/Arg and Arg/Arg genotypes and colorectal cancer risk (OR, 1.8; 95% CI, 1.1-3.1). When we corrected our findings for multiple comparisons, this finding was no longer statistically significant. We also observed a positive association with colorectal cancer risk for the OGG1 codon 326 Ser/Cys and Cys/Cys genotypes (OR, 1.4; 95% CI, 0.9-2.1) and an inverse association with the XRCC1 codon 399 Gln/Gln genotype (OR, 0.6; 95% CI, 0.4-1.1); however, neither association reached statistical significance (Table 2), particularly after correcting for multiple comparisons. For all SNPs, we tested for heterogeneity of the ORs by dialect group (Cantonese or Hokkien) using variables that assumed the SNPs' predicted mode of action. We only observed statistically significant differences for the OGG1 codon 326 SNP (gene × dialect group interaction test, P = 0.01). Among Cantonese, there was no association between the OGG1 codon 326 Ser/Cys or Cys/Cys genotypes (OR, 0.9; 95% CI, 0.5-1.5), whereas among Hokkien, there was a statistically significant positive association (OR, 2.5; 95% CI, 1.3-4.7). We took into account this gene × dialect group interaction in all further analyses involving OGG1 genotype. For all SNPs, we also tested for heterogeneity of the ORs by tumor subsite (colon versus rectum). We only observed significant subsite differences for the MGMT codon 84 SNP (test of heterogeneity, P = 0.027). Whereas there was no association between the Leu/Phe or Phe/Phe genotypes and rectal cancer risk (OR, 1.2; 95% CI, 0.7-1.9), there was a statistically significant inverse association with colon cancer risk (OR, 0.6; 95% CI, 0.3-0.9).
Joint Analyses of Multiple SNPs per Gene and Colorectal Cancer Risk
For three of the genes (XRCC1, XPD, and PARP), we genotyped more than one SNP per gene. To better understand the combined effect of these two SNPs per gene, we conducted haplotype-based analyses (Table 3). Analyses of haplotypes for PARP defined by the codons 762 and 940 alleles showed that the Ala762-Arg940 haplotype is rare in this population; therefore, the association we observed for the codon 940 Arg allele seems to be driven by the Val762-Arg940 haplotype, which is significantly associated with colorectal cancer risk (OR, 1.8; 95% CI, 1.0-3.2; Table 3). Haplotype-based analyses for XRCC1 indicated that the Trp194-Gln399 haplotype is very rare in this population (Table 3). Both haplotypes that carry either the codon 194 variant allele (Trp) or the codon 399 variant allele (Gln) show slight inverse associations with colorectal cancer risk; however, the global test P value was not statistically significant (P = 0.242). Similarly, haplotype-based analyses for XPD suggested that the Asn312-Lys751 may be inversely associated with colorectal cancer risk; however, the global test was not statistically significant (P = 0.252). Furthermore, this haplotype is rare in the population (Table 3).
DNA repair SNPs and colorectal cancer risk among Singapore Chinese
DNA repair SNPs . | Controls . | Cases . | OR* (95% CI) . | P . | ||||
---|---|---|---|---|---|---|---|---|
XRCC1 codons 194-399 | ||||||||
Arg-Arg | 995 | 286 | 1ref | |||||
Arg-Gln | 576 | 139 | 0.8 (0.7-1.1) | 0.134 | ||||
Trp-Arg | 643 | 158 | 0.8 (0.6-1.0) | 0.080 | ||||
Trp-Gln | 16 | 3 | 0.7 (0.2-3.0) | 0.639 | ||||
Pglobal | 0.242 | |||||||
XPD codons 312-751 | ||||||||
Asp-Lys | 2111 | 537 | 1ref | |||||
Asp-Gln | 108 | 32 | 1.1 (0.7-1.7) | 0.705 | ||||
Asn-Lys | 32 | 5 | 0.4 (0.2-1.3) | 0.148 | ||||
Asn-Gln | 63 | 22 | 1.4 (0.8-2.3) | 0.229 | ||||
Pglobal | 0.252 | |||||||
PARP codons 762-940 | ||||||||
Val-Lys | 1266 | 310 | 1ref | |||||
Val-Arg | 50 | 23 | 1.8 (1.0-3.2) | 0.037 | ||||
Ala-Lys | 1010 | 271 | 1.1 (0.9-1.3) | 0.436 | ||||
Ala-Arg | 2 | 2 | 2.9 (0.3-32) | 0.385 | ||||
Pglobal | 0.106 |
DNA repair SNPs . | Controls . | Cases . | OR* (95% CI) . | P . | ||||
---|---|---|---|---|---|---|---|---|
XRCC1 codons 194-399 | ||||||||
Arg-Arg | 995 | 286 | 1ref | |||||
Arg-Gln | 576 | 139 | 0.8 (0.7-1.1) | 0.134 | ||||
Trp-Arg | 643 | 158 | 0.8 (0.6-1.0) | 0.080 | ||||
Trp-Gln | 16 | 3 | 0.7 (0.2-3.0) | 0.639 | ||||
Pglobal | 0.242 | |||||||
XPD codons 312-751 | ||||||||
Asp-Lys | 2111 | 537 | 1ref | |||||
Asp-Gln | 108 | 32 | 1.1 (0.7-1.7) | 0.705 | ||||
Asn-Lys | 32 | 5 | 0.4 (0.2-1.3) | 0.148 | ||||
Asn-Gln | 63 | 22 | 1.4 (0.8-2.3) | 0.229 | ||||
Pglobal | 0.252 | |||||||
PARP codons 762-940 | ||||||||
Val-Lys | 1266 | 310 | 1ref | |||||
Val-Arg | 50 | 23 | 1.8 (1.0-3.2) | 0.037 | ||||
Ala-Lys | 1010 | 271 | 1.1 (0.9-1.3) | 0.436 | ||||
Ala-Arg | 2 | 2 | 2.9 (0.3-32) | 0.385 | ||||
Pglobal | 0.106 |
Adjusted for dialect group (Hokkien or Cantonese), age at recruitment (y), gender, and year of recruitment (1993-1995, 1996-1998).
DNA Repair SNPs, Smoking, and Colorectal Cancer Risk
We observed evidence that the XRCC1 codon 194 SNP modified the effect of smoking (Table 4). Among carriers of the codon 194 Arg/Arg genotype, cigarette smoking was not associated with colorectal cancer risk, whereas among carriers of one or two copies of the Trp allele, we observed a statistically significant positive association. This was true for all different aspects of cigarette smoking exposure, such as smoking status (Pinteraction < 0.004), years of smoking (Pinteraction = 0.002), and smoking index (Pinteraction = 0.012; Table 4). When we corrected our results for multiple comparisons, these findings remained statistically significant with the exception of the XRCC1 codon 194 × smoking index interaction, which had a P value that remained above the threshold. Further adjustment of our models with the codon 399 SNP did not change our results, suggesting that in this dataset the XRCC1 × smoking interaction is driven by the codon 194 SNP and not the codon 399 SNP.
XRCC1codon 194, smoking, and colorectal cancer risk among Singapore Chinese
. | XRCC1 codon 194 . | . | . | . | Pinteraction . | |||||
---|---|---|---|---|---|---|---|---|---|---|
. | Arg/Arg . | . | Arg/Trp or Trp/Trp . | . | . | |||||
. | Controls/cases . | OR* (95% CI) . | Controls/cases . | OR* (95% CI) . | . | |||||
Smoking status | ||||||||||
Never smokers | 418/103 | 1ref | 424/78 | 0.6 (0.5-0.9) | ||||||
Ever smokers | 168/58 | 0.7 (0.5-1.1) | 152/66 | 1.0 (0.7-1.6) | ||||||
Genotype-specific OR† | Ever | 0.7 (0.5-1.1) | 1.6 (1.0-2.5) | 0.005 | ||||||
Never smokers | 418/103 | 1ref | 414/78 | 0.6 (0.4-0.9) | ||||||
Past smokers | 66/26 | 0.7 (0.4-1.2) | 64/25 | 0.8 (0.4-1.4) | ||||||
Current smokers | 102/32 | 0.7 (0.4-1.2) | 88/41 | 1.2 (0.7-2.0) | ||||||
Test for trend | 0.199 | 0.012 | 0.004 | |||||||
Genotype-specific OR† | Past | 0.7 (0.4-1.2) | 1.2 (0.7-2.2) | |||||||
Current | 0.7 (0.4-1.2) | 1.9 (1.1-3.1) | ||||||||
Years of smoking cigarette | ||||||||||
Never smokers | 418/103 | 1ref | 424/78 | 0.6 (0.4-0.9) | ||||||
<40 y | 98/33 | 0.9 (0.5-1.5) | 94/30 | 0.9 (0.5-1.5) | ||||||
≥40 y | 70/25 | 0.6 (0.3-1.0) | 58/36 | 1.2 (0.7-2.0 | ||||||
Test for trend | 0.067 | 0.026 | 0.002 | |||||||
Genotype-specific OR† | <40 y | 0.9 (0.5-1.5) | 1.4 (0.8-2.4) | |||||||
≥40 y | 0.6 (0.3-1.0) | 1.8 (1.1-3.1) | ||||||||
Smoking index | ||||||||||
Never smokers | 418/103 | 1ref | 424/78 | 0.6 (0.4-0.9) | ||||||
Light smokers | 153/46 | 0.6 (0.4-1.0) | 137/51 | 0.9 (0.6-1.4) | ||||||
Heavy smokers | 15/12 | 2.0 (0.8-4.7) | 15/15 | 2.2 (1.0-5.0) | ||||||
Test for trend | 0.730 | 0.003 | 0.012 | |||||||
Genotype-specific OR† | Light smokers | 0.6 (0.4-1.0) | 1.4 (0.9-2.2) | |||||||
Heavy smokers | 2.0 (0.8-4.7) | 3.4 (1.5-7.9) |
. | XRCC1 codon 194 . | . | . | . | Pinteraction . | |||||
---|---|---|---|---|---|---|---|---|---|---|
. | Arg/Arg . | . | Arg/Trp or Trp/Trp . | . | . | |||||
. | Controls/cases . | OR* (95% CI) . | Controls/cases . | OR* (95% CI) . | . | |||||
Smoking status | ||||||||||
Never smokers | 418/103 | 1ref | 424/78 | 0.6 (0.5-0.9) | ||||||
Ever smokers | 168/58 | 0.7 (0.5-1.1) | 152/66 | 1.0 (0.7-1.6) | ||||||
Genotype-specific OR† | Ever | 0.7 (0.5-1.1) | 1.6 (1.0-2.5) | 0.005 | ||||||
Never smokers | 418/103 | 1ref | 414/78 | 0.6 (0.4-0.9) | ||||||
Past smokers | 66/26 | 0.7 (0.4-1.2) | 64/25 | 0.8 (0.4-1.4) | ||||||
Current smokers | 102/32 | 0.7 (0.4-1.2) | 88/41 | 1.2 (0.7-2.0) | ||||||
Test for trend | 0.199 | 0.012 | 0.004 | |||||||
Genotype-specific OR† | Past | 0.7 (0.4-1.2) | 1.2 (0.7-2.2) | |||||||
Current | 0.7 (0.4-1.2) | 1.9 (1.1-3.1) | ||||||||
Years of smoking cigarette | ||||||||||
Never smokers | 418/103 | 1ref | 424/78 | 0.6 (0.4-0.9) | ||||||
<40 y | 98/33 | 0.9 (0.5-1.5) | 94/30 | 0.9 (0.5-1.5) | ||||||
≥40 y | 70/25 | 0.6 (0.3-1.0) | 58/36 | 1.2 (0.7-2.0 | ||||||
Test for trend | 0.067 | 0.026 | 0.002 | |||||||
Genotype-specific OR† | <40 y | 0.9 (0.5-1.5) | 1.4 (0.8-2.4) | |||||||
≥40 y | 0.6 (0.3-1.0) | 1.8 (1.1-3.1) | ||||||||
Smoking index | ||||||||||
Never smokers | 418/103 | 1ref | 424/78 | 0.6 (0.4-0.9) | ||||||
Light smokers | 153/46 | 0.6 (0.4-1.0) | 137/51 | 0.9 (0.6-1.4) | ||||||
Heavy smokers | 15/12 | 2.0 (0.8-4.7) | 15/15 | 2.2 (1.0-5.0) | ||||||
Test for trend | 0.730 | 0.003 | 0.012 | |||||||
Genotype-specific OR† | Light smokers | 0.6 (0.4-1.0) | 1.4 (0.9-2.2) | |||||||
Heavy smokers | 2.0 (0.8-4.7) | 3.4 (1.5-7.9) |
Adjusted for age at recruitment (y), gender, dialect group (Hokkien, Cantonese), year of recruitment (1993-1995, 1996-1998), level of education (no formal education, primary school, secondary or higher education), history of diabetes (yes/no), body mass index (<20, 20 to <24, 24 to <28, 28+ kg/m2), family history of colorectal cancer (yes/no), weekly physical activities (yes/no), and alcohol intake (never, <7 drinks/wk, >7 drinks/wk).
Genotype-specific OR is the OR for each level of exposure compared with never smokers among carriers of each of the two genotypes, as derived from models fitted with an interaction term between genotype and exposure.
To confirm a role for the codon 194 SNP, unaffected by the codon 399 SNP, we also conducted a haplotype-based XRCC1 × smoking analysis. Of the four possible haplotypes, there were very few subjects who carried the Trp194-Gln399 haplotype (nine controls and one case); therefore, we allow for these subjects to be included in the reference haplotype group (Arg194-Arg399). The global test of XRCC1 × smoking interaction (smoking status, never/ever) was statistically significant (P = 0.012). Whereas the effect of smoking among carriers of the Arg194-Gln399 haplotype was OR of 0.7 (95% CI, 0.4-1.1), among carriers of the Trp194-Arg399 haplotype, OR was 1.6 (95% CI, 1.1-2.5). These findings suggest that the XRCC1 × smoking interaction is driven by the haplotype that carries the codon 194 variant allele (Trp allele) and not by the haplotype that carries the codon 399 variant allele (Gln allele). Analyses done using model selection with AIC and fitting models that included either or both SNPs, with or without interaction terms between each SNP and smoking, also supported an XRCC1 × smoking interaction driven by the codon 194 SNP only, probably unaffected by the codon 399 SNP (data not shown).
The XRCC1 codon 194 × smoking interaction did not differ by tumor subsite (heterogeneity test of interaction ORs by tumor subsite: P = 0.188 for smoking status, P = 0.288 for years of smoking, and P = 0.304 for the overall smoking index). There is no evidence that the XRCC1 × smoking interaction was modified by alcohol intake. We found no evidence that the candidate SNPs we studied in the XPD, PARP, OGG1, and MGMT genes modified the effects of smoking and found no evidence of heterogeneity of these interaction ORs by tumor subsite. Finally, we found no evidence that dialect group modified the combined effects of OGG1 and smoking.
DNA Repair SNPs, Alcohol, and Colorectal Cancer Risk
We observed weak evidence that the XRCC1 codon 399 modified the effects of alcohol intake (Pinteraction = 0.055 for XRCC1 × never/ever alcohol intake). The data suggested that alcohol intake increased risk of colorectal cancer among carriers of the Arg/Arg genotype (OR, 1.3; 95% CI, 0.9-1.9). However, the number of codon 399 Gln homozygous carriers who were ever drinkers was too small to make meaningful interpretations and estimates (18 controls and 1 case). Conversely, our analyses of the codon 194 SNP indicated that, whereas alcohol had no effect among carriers of the Arg/Arg genotype (OR, 1.0; 95% CI, 0.6-1.7), alcohol had an effect among carriers of the Arg/Trp or Trp/Trp genotypes (OR, 1.3; 95% CI, 0.8-2.1), although a test of interaction was not statistically significant (P = 0.483). To better understand the role of XRCC1 and alcohol, we next carried out haplotype-based analyses of XRCC1 × alcohol interaction by considering carriers of one or two copies of the haplotypes that include the codons 194 and 399 SNPs. The global test of XRCC1 × alcohol interaction (alcohol intake never/ever) was borderline statistically significant (P = 0.06). Whereas the effect of ever alcohol intake among carriers of the Arg194-Gln399 haplotype was OR of 0.8 (95% CI, 0.4-1.3), among carriers of the Trp194-Arg399 haplotype, OR was 1.7 (95% CI, 1.0-2.6). These findings suggest that XRCC1 may modify the effects of alcohol and that such effect modification may be driven by the codon 194 variant allele (Trp). Prompted by this finding, we carried out XRCC1 × alcohol analyses considering the three genotype levels for the codon 194 SNP to better understand if the effect of alcohol differs between carriers of one or two copies of the Trp allele (Table 5). Our analyses show that whereas alcohol has no effect on colorectal cancer risk among carriers of the codon 194 Arg/Arg (OR, 1.0; 95% CI, 0.6-1.7) or Arg/Trp genotypes (OR, 1.1; 95% CI, 0.6-1.9), there was a statistically significant effect among carriers of the Trp/Trp genotype (OR, 2.8; 95% CI, 1.0-8.1). A test of interaction did not reach statistical significance (P = 0.245) likely due to small sample size.
XRCC1 codon 194 and alcohol intake interaction
. | Arg/Arg . | . | Arg/Trp . | . | Trp/Trp . | . | P . | |||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Controls/cases . | OR* (95% CI) . | Controls/cases . | OR* (95% CI) . | Controls/cases . | OR* (95% CI) . | . | |||||||
Alcohol intake status | ||||||||||||||
Never | 479/130 | 1ref | 378/91 | 0.9 (0.6-1.2) | 94/19 | 0.7 (0.4-1.2) | ||||||||
Ever | 107/31 | 1.0 (0.6-1.7) | 89/26 | 0.9 (0.6-1.6) | 15/8 | 1.9 (0.7-4.9) | ||||||||
Test for trend | 0.952 | 0.776 | 0.062 | 0.245 | ||||||||||
Genotype-specific OR† | 1.0 (0.6-1.7) | 1.1 (0.6-1.9) | 2.8 (1.0-8.1) |
. | Arg/Arg . | . | Arg/Trp . | . | Trp/Trp . | . | P . | |||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Controls/cases . | OR* (95% CI) . | Controls/cases . | OR* (95% CI) . | Controls/cases . | OR* (95% CI) . | . | |||||||
Alcohol intake status | ||||||||||||||
Never | 479/130 | 1ref | 378/91 | 0.9 (0.6-1.2) | 94/19 | 0.7 (0.4-1.2) | ||||||||
Ever | 107/31 | 1.0 (0.6-1.7) | 89/26 | 0.9 (0.6-1.6) | 15/8 | 1.9 (0.7-4.9) | ||||||||
Test for trend | 0.952 | 0.776 | 0.062 | 0.245 | ||||||||||
Genotype-specific OR† | 1.0 (0.6-1.7) | 1.1 (0.6-1.9) | 2.8 (1.0-8.1) |
Adjusted for age at recruitment (y), gender, dialect group (Hokkien, Cantonese), year of recruitment (1993-1995, 1996-1998), level of education (no formal education, primary school, secondary or higher education), history of diabetes (yes/no), body mass index (<20, 20 to <24, 24 to <28, 28+ kg/m2), family history of colorectal cancer (yes/no), weekly physical activities, and smoking index.
Genotype-specific ORs test for trend P = 0.244.
The effect of the XRCC1 codon 194 SNP on alcohol and colorectal cancer risk differed by tumor subsite (colon versus rectum; heterogeneity test of interaction ORs by tumor subsite: P = 0.020 for never/ever alcohol intake, P = 0.011 for alcohol frequency, P = 0.014 for number of drinks per week). Among carriers of the Arg/Arg genotype, the OR of alcohol (never/ever) on colon cancer risk was 0.6 (95% CI, 0.3-1.2), whereas the corresponding OR among carriers of the Arg/Trp or Trp/Trp genotype was 1.4 (95% CI, 0.8-2.6), with Pinteraction = 0.04. Among rectal cancer cases, the respective ORs were 1.8 (95% CI, 1.0-3.2) and 0.9 (95% CI, 0.6-1.5), with Pinteraction = 0.358.
We found no evidence that the candidate SNPs we studied in the XPD, PARP, OGG1, and MGMT genes modified the effects of alcohol on colorectal cancer risk or that these interaction ORs differed by tumor subsite. There is no evidence that the combined effects of OGG1 genotype and alcohol intake on risk differed by dialect group.
Discussion
Recent analyses within the Singapore Chinese Health Study have shown that smoking is only a risk factor for rectal cancer (2). We now report that among carriers of the XRCC1 Trp194-Arg399 haplotype, smoking is also a risk factor for colon cancer, in addition to rectal cancer. Furthermore, results of our study suggest that among Singapore Chinese, the PARP Lys940Arg polymorphism is associated with risk of colorectal cancer. Compared with subjects who did not carry this SNP, carriers of two copies of the variant allele had an 80% increase in risk. We note that this finding does not remain statistically significant if we correct our results for multiple comparisons.
To the best of our knowledge, this is the first report of an association between a SNP in the PARP gene and colorectal cancer risk. Three other studies have reported positive associations between the PARP codon 762 SNP and lung cancer among Han Chinese (23), esophagus cancer among Han Chinese (24), and prostate cancer risk among Caucasians (25). This latter study also reported an association between the codon 762 variant allele (Ala) and lower ADP rybosyl transferase activity when cells were challenged with hydrogen peroxide. A recent study has reported that the codon 762 SNP reduces the enzymatic activity of the PARP protein, thus providing direct evidence of a functional effect of this SNP on PARP function (26). We did not find an association with the codon 762 SNP, but instead report an association with the codon 940 SNP. We had 80% power to detect a smallest detectable OR of 1.5 and 2.1 for the codon 762 and codon 940 SNPs, respectively. In our study population, there were only two subjects who carried the haplotype formed by both the 762 and 940 variant alleles (Ala762-Arg940). Whereas carriers of the Val762-Arg940 haplotype had an OR of 1.8 (95% CI, 1.0-3.2), carriers of the Ala762-Lys940 had an OR of 1.1 (95% CI, 0.9-1.3). Thus, among Singapore Chinese the codon 940 SNP seems to be a more relevant risk factor than the codon 762. The effect of the codon 940 SNP on PARP function is still unknown. It is plausible that differences in LD patterns between these different study populations may explain these discrepancies. One should note that none of the three studies mentioned above consider the codon 940 SNP; therefore, it is unknown whether this SNP is or is not associated with cancer risk in other populations. The PARP protein plays an important role in maintaining genomic stability and regulating transcription (27, 28). PARP can detect single-strand breaks and calls upon the action of the base excision repair pathway through its interaction with the XRCC1 protein. More recently, PARP was found to act as a coactivator of the β-catenin–TCF-4 complex, which aberrantly regulates growth and differentiation of intestinal epithelial cells (29). Furthermore, PARP protein has been reported to be overexpressed in colorectal tumors and to correlate with tumor size, histopathology, and β-catenin, c-myc, cyclin D1, and matrix metalloproteinase-7 expression (30). Thus, a direct role for PARP genotype on colorectal cancer risk is biologically plausible. However, we cannot rule out that our finding may be due to small numbers.
One study in Spain has reported a positive association between the OGG1 codon 326 Cys allele and colorectal cancer (31), whereas a Norwegian study reported a null finding (32). In the present study, we observed a statistically nonsignificant association between the OGG1 codon 326 Cys allele and increased colorectal cancer risk (OR, 1.4; 95% CI, 0.9-2.1). We had 80% power to detect a smallest detectable OR of 1.8. Previous studies have reported that the Cys allele diminishes the activity of the OGG1 protein (33, 34); therefore, an association between the Cys allele and increased cancer risk is plausible, as lack of OGG1 activity could increase the accumulation of 8-hydroxy-2-deoxyguanosine base damage and thus contribute to mutagenesis in colorectal cells.
In contrast with our finding of a nonstatistically significant inverse association between the XRCC1 codon 399 SNP and colorectal cancer risk, a previous study among Asians reported a positive association for this SNP (35), whereas another study among Asians (36) and two studies among Caucasians (31, 37) reported a null association. A study conducted in Egypt reported a positive association between both the codon 194 and 399 SNPs and colorectal cancer, but it was based on very small numbers (38). We found no evidence of an association with the XRCC1 codon 194 SNP. However, our results suggest that these two XRCC1 SNPs may modify the effect of smoking and alcohol, which only had an effect on colorectal cancer risk among carriers of the XRCC1 Trp194-Arg399 haplotype. The XRCC1 protein plays a key role in the base excision repair pathway, by coordinating all the steps serving as a scaffold via its interaction with other key base excision repair proteins, such as DNA polymerase β, DNA ligase III, polynucleotide kinase, PARP, APE1, and OGG1 (39). To date, only one other study, also among Asians, has investigated a potential XRCC1 × smoking interaction in colorectal cancer (36) and reported a null finding. In contrast, a recent review of the literature concluded that the effect of smoking on risk of cancer might be modified by the XRCC1 genotype (40). Recently, we reported an XRCC1 genotype × smoking interaction effect on risk of colorectal adenomas in a Los Angeles sigmoidoscopy-based study; the smoking effect was noted only among carriers of the Arg194-Arg399 haplotype (41). Phenotype association studies suggest that both the XRCC1 codon 194 Trp and codon 399 Arg alleles are associated with higher repair proficiency (40), although functional studies have failed to find an effect for either SNP (42, 43). Interestingly, a functional effect was reported for the codon 280 SNP (42), which is in LD with the codons 194 and 399 SNPs. Therefore, the results observed in this study might be driven by this SNP, which was not investigated. A causal role of the codon 280 SNP and differences in LD patterns between Singapore Chinese and U.S. subjects also could explain the discrepant findings between the present study and our previous adenoma study in Los Angeles (41). An interaction effect between XRCC1 genotype and smoking or alcohol on colorectal cancer risk is biologically plausible. Both exposures generate reactive oxygen species–induced DNA damage that elicits repair by the base excision repair pathway in which the XRCC1 protein plays an essential role.
Regarding the XPD gene, our finding of a null association with the XPD codon 312 SNP and weak evidence of an association with the codon 751 SNP is in agreement with previous studies (31, 36, 37, 44). Several reports suggest that the codon 751 Gln allele might be associated with lower repair proficiency of damage induced by UV and chemical carcinogens, although other studies suggest this allele might confer higher repair proficiency of damage induced by ionizing radiation (45), suggesting a differential effect for this variant on XPD protein function under different exposures. Previously, we reported some evidence for an XPD × alcohol interaction and no evidence of an XPD × smoking interaction among colorectal adenomas, precursors of colorectal cancer, among a multiethnic sigmoidoscopy study in Los Angeles county (41). This is in contrast with another study that reported evidence of an XPD × smoking interaction among hyperplastic and adenomatous polyps, when considering the codon 312 and codon 751 polymorphisms combined but no evidence of an XPD × alcohol interaction (46), and a study that reported no evidence of an XPD × smoking interaction (47) also among colorectal adenomas. Altogether, the available evidence suggests that variation in this gene might be an effect modifier for the formation of colorectal adenomas but not for carcinoma development. Given that not all genetic variation in this gene has been considered and that results for previous studies considering the codon 312 and codon 751 SNPs are conflicting, more comprehensive analyses of this gene might help sort this issue.
Two reports are in agreement with our finding of a null association between the MGMT codon 84 SNP and colorectal cancer risk (31, 48). However, in Tranah et al. (48), the codon 84 SNP was found to modify the effects of alcohol on risk among women, and this SNP was inversely associated with risk among women with low alcohol intake (48). In our study, we observed an inverse association between the codon 84 SNP and colon cancer only, not rectal cancer. Small numbers precluded us from properly assessing a potential MGMT × alcohol interaction among colon cancer cases. As reported and discussed in a recent study, whereas this SNP does not seem to affect many of the aspects of MGMT function, it is in linkage disequilibrium with 12 other polymorphisms; therefore, a role for variation in this gene on colorectal cancer risk cannot be discarded and deserves further comprehensive analyses (49).
To gain insight into the joint contribution of the studied SNPs, we examined models with all multivariate combinations of the eight DNA repair SNP ranked by AIC (50). Because model selection by AIC may lead to inflated type I errors and to account for the uncertainty in the determination of a single best model, we used a Bayes model averaging procedure via stochastic variable selection to further evaluate the combinations of studied SNPs and to qualitatively compare our conclusions (44). This approach gives posterior probabilities of all 256 models and posterior probabilities of inclusion of any SNP across all models (data not shown). We used Bayes factors for valid statistical testing of hypotheses (51). The overall results of our statistical analyses confirmed that the PARP codon 940 SNP was a risk factor for colorectal cancer and that such association is independent of the other SNPs considered.
The strengths of this study are its population-based design and the availability of prospective data obtained via a face-to-face interview. One potential concern arises from the fact that biospecimens to conduct genetic analyses were only available for 52% of all colorectal cancer cases diagnosed within the cohort by April 2002. An examination of all major demographic variables and selected nutritional and life-style variables indicates that subjects (both controls and cases) who consented to donate blood or buccal cells do not differ from subjects who refused biospecimen donations. However, cases who donated biospecimens had a lower prevalence of advanced tumors (46.8% had stages 3 and 4) than cases who did not donate biospecimens (62% had stages 3 and 4). This is likely due to the lag time between case diagnosis and case notification, with more advanced cases being less likely to be alive or to consent to donate specimens due to illness. The prevalence of the genotypes we investigated did not differ across stages. Therefore, whereas we cannot completely discard the potential for survival bias, it does not seem likely in the current study. Another limitation is the relatively modest number of cancer cases (n = 310) in this study. This is particularly a limitation for our G × E analyses by tumor subsite. We report here statistically significant heterogeneity of XRCC1 × alcohol interaction between colon and rectal cancer cases. This difference seems driven by statistically significant XRCC1 × alcohol interaction among colon cancer cases; however, this interaction seems driven by opposite effects of alcohol among Arg/Arg than among Arg/Trp or Trp/Trp carriers. Therefore, follow-up studies within the Singapore Chinese Health Study and other independent cohorts are necessary to either confirm or refute the present results.
In summary, our results provide support to the hypothesis that selected variants in DNA repair genes may contribute to colorectal cancer risk and may modify the effects of relevant life-style risk factors that have been inconsistently associated with this disease. Our findings of an overall effect of PARP codon 940 SNP on colorectal cancer risk and XRCC1 SNPs as modifiers of the effects of smoking and alcohol on risk highlight the role of the base excision repair pathway in colorectal carcinogenesis and point to reactive oxygen species as relevant genotoxins that may account for the effects of both smoking and alcohol on colorectal cancer risk.
Grant support: National Cancer Institute grants R01 CA55069, R35 CA53890, R01 CA80205, and R01 CA98497 and National Institute of Environmental Health Sciences grant 5P30 ES07048 (M.C. Stern).
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.
Acknowledgments
We thank Siew-Hong Low of National University of Singapore for supervising the fieldwork of Singapore Chinese Health Study, Singapore Cancer Registry for assistance with the identification of cancer outcomes, Kazuko Arakawa of University of Southern California for the development and management of the cohort study database, and Amit Joshi and Ms Tammy Wan of University of Southern California for assistance with genotypic data collection and data analyses.