Abstract
Whereas animal and in vitro studies support a role of unsaturated fatty acids in colon carcinogenesis, the epidemiologic evidence is inconclusive. Using a large sigmoidoscopy-based case-control study (753 cases and 799 controls) in Los Angeles County, we investigated possible associations between single-nucleotide polymorphisms in the XRCC1 (codons 194 Arg/Trp and codon 399 Arg/Gln) and XRCC3 (codon 241 Thr/Met) genes and colorectal adenoma risk and their possible role as modifiers of the effect of monounsaturated fatty acid, the ratio of ω-6/ω-3 polyunsaturated fatty acids, and antioxidant intake. We found no evidence of associations between the XRCC1 codon 194 Arg/Trp or Trp/Trp genotypes and the XRCC3 codon 241 Thr/Met or Met/Met genotypes. Subjects with the XRCC1 Gln/Gln genotype were inversely associated with adenoma risk (odds ratio, 0.6; 95% confidence interval, 0.4-0.9; P = 0.01) when compared with subjects with Arg/Arg and Arg/Gln genotypes combined. We found no evidence of gene-dietary fat interactions for the XRCC3 codon 241 polymorphism. However, our data suggest an XRCC1-unsaturated fat interaction. High monounsaturated fatty acid intake was associated with adenoma risk only among subjects with the XRCC1 codon 194 Arg/Arg and codon 399 Gln/Gln combined genotypes (P for interaction = 0.018). High ω-6/ω-3 polyunsaturated fatty acid ratios were associated with adenoma risk among subjects with the XRCC1 codon 194 Arg/Arg and codon 399 Gln/Gln or the codon 194 Arg/Trp or Trp/Trp and codon 399 Arg/Arg or Arg/Gln combined genotypes (P for interaction = 0.026). These interactions were not modified by antioxidant intake. However, low antioxidant intake was associated with an inverse association only among subjects with the XRCC1 codon 194 Arg/Trp or Trp/Trp and codon 399 Arg/Arg or Arg/Gln combined genotypes (P for interaction = 0.022), which was independent of unsaturated fat intake. Our data suggest that the XRCC1 codon 194 and codon 399 single nucleotide polymorphisms may modify the effect of unsaturated fatty acid and antioxidant intake and that this XRCC1 effect modification may explain, in part, previously reported inconsistencies on the role of unsaturated fatty acids and adenoma risk.
Introduction
The role of unsaturated dietary fat in the development of colorectal adenomas, established precursors of colorectal cancer, is inconclusive. An extensive review of the literature by an expert panel has concluded that whereas diets high in total fat and saturated fat may possibly increase the risk of colorectal cancer, the epidemiologic data for monounsaturated fatty acids (MUFA) and polyunsaturated fatty acids (PUFA) is too inconsistent to make final conclusions yet (1).
Nonetheless, animal and in vitro studies support a role for PUFAs in colon carcinogenesis. In particular, high intake of ω-6 PUFAs, such as linoleic and arachidonic acid, and low intake of ω-3 PUFAs, such as docosahexanoic, eicosapentanoic, and linolenic acid, are linked to increased colon cancer risk (2). In rats, docosahexanoic acid can suppress the formation and growth of aberrant crypt colon foci (2). Therefore, the dietary ratio of ω-6/ω-3 PUFAs seems to be critical (2, 3). In support of this, elevated levels of arachidonic acid are found in cell membranes of colon tumors when compared with normal colon mucosa (4). The fatty acids composition of cell membranes, which influences membrane fluidity and permeability, can be determined by dietary intake of fat (5), with high membrane fluidity being associated with elevated rates of cell division in malignant cells (5).
Oxidation of cell membrane unsaturated fatty acids generates lipid hydroperoxides, which can either be reduced by glutathione peroxidases or, after reacting with metals, can become reactive compounds such as malondialdehyde, 4-hydroxynonenal, and peroxyl radicals (6, 7). In particular, lipid hydroperoxides derived from ω-6-linoleic acid have been found to initiate lipid peroxidation in colon cell membranes (8). Malondialdehyde is highly mutagenic and forms DNA adducts that can lead to base pair substitutions. These adducts are repaired by the nucleotide excision repair pathway (9). The epoxide form of 4-hydroxynonenal can form genotoxic etheno DNA adducts that can induce base pair changes. ω-6 PUFA-induced etheno adducts are removed primarily by the base excision repair (BER) pathway, although the nucleotide excision repair could also be involved (7, 10). Peroxyl radicals can induce DNA base modifications and single-strand breaks, typically repaired by BER, and double-strand breaks, which are repaired by homologous recombination repair or nonhomologous end-joining (11). Although both ω-3 and ω-6 PUFAs can serve as substrates for lipid peroxidation, the major type of adducts induced by ω-3 PUFAs do not seem to cause base mispairing and are therefore nonmutagenic. Conversely, ω-6 PUFAs-4-hydroxynonenal–derived adducts are mutagenic and can preferentially bind to mutational hotspots in critical genes such as p53 (10).
Persistent oxidative stress and oxidative damage to the colon epithelium seems to play an important role in colorectal adenoma development. Sources of oxidative stress include infiltrating lymphocytes, which are a rich source of reactive oxygen species, and colonic bacteria that can release superoxides that convert to reactive hydroxyl radicals in the presence of iron. The latter can react with the contents of feces (e.g., fatty acids) converting them into mutagenic agents (e.g., lipid hydroperoxides) that can damage colon stem cells (12). Furthermore, bile acids could promote this process by solubilizing peroxidizable fatty acids and iron, whereas antioxidants (e.g., vitamin E and carotenoids) and dietary fiber would inhibit it. In support of this, subjects that consume a diet high in fat and meat and low in fiber have a significant increase in free radical production in feces when compared with subjects that consume a diet low in fat and high in fiber (12).
We hypothesized that selected DNA repair genes may modify the effect of unsaturated dietary fat in colorectal adenomas, given the role of these types of fat in the generation of DNA damage. If this hypothesis were true, taking into account the genotype of proteins that participate in the fatty acids-DNA damage/repair pathway may help better understand the role of unsaturated fatty acids in colorectal adenoma risk. This is a complex issue with overlapping pathways and multiple genes. In the present study, we report the results of our investigations studying single nucleotide polymorphisms (SNP) in two genes that participate in two pathways postulated to repair damage induced by unsaturated fatty acid (7, 10, 11): XRCC1 (codons 194 and 399), which participates in BER (13, 14) and XRCC3 (codon 241), which participates in homologous recombination repair (15, 16). We report here the main effect of these SNPs in colorectal adenoma risk and their role as modifiers of the effect of unsaturated dietary fat. Furthermore, given that among other factors, the final amount of DNA damage that accumulates in colonic cells will be determined by the balance between unsaturated fatty acid intake and antioxidant intake, we also explored the role of these SNPs in modifying the effect of antioxidant intake, a hypothesis previously proposed by van Gils et al. (17) for prostate cancer.
Materials and Methods
Study Subjects
All subjects for this study were participants in a USC/Kaiser study of risk factors for colorectal adenomas and were examined by flexible sigmoidoscopy from 1991 to 1993 (phase I) and 1993 to 1995 (phase II) at either of two southern California Kaiser Permanente Medical Centers (Bellflower and Sunset). Phase I of this study has been previously described (18). Phase II subjects were recruited using the same criteria as phase I. Briefly, eligible English-speaking subjects were 50 to 74 years old, lived in the Los Angeles metropolitan area, had no history of invasive cancer, inflammatory bowel disease, familial polyposis, previous bowel surgery, or symptoms suggestive of gastrointestinal disease. Cases were those subjects with a first-time diagnosis of a histologically confirmed adenoma, whereas controls were subjects with no current or past polyps, individually matched to cases by gender, age (within 5 years), sigmoidoscopy date (within 3 months), and center. All subjects signed a written informed consent approved by the Institutional Review Board, donated a blood sample, and provided demographic information, smoking, family history of cancer, and physical activity among other factors during an in-person interview. All subjects were interviewed on average 5 months after sigmoidoscopy. All participants completed a semiquantitative food frequency questionnaire that inquired about diet in the year before sigmoidoscopy, as previously described (18, 19).
Genotyping Methods
XRCC1 and XRCC3 genotypes were determined using genomic DNA isolated from peripheral blood lymphocytes and PCR-restriction fragment length polymorphism assays, as previously described (20, 21). Briefly, the XRCC1-194 and XRCC1-399 SNPs were genotyped using a multiplex PCR assay that amplifies fragments surrounding the codon 194 and 399 polymorphisms, which are then digested with MspI, resolved in 3% metaphor agarose gels (FMC Bioproducts, Rockland, ME) and stained with ethidium bromide (20). The XRCC3-241 SNP was genotyped by amplifying a DNA segment surrounding codon 241 which was then digested with NlaIII restriction enzyme (New England Biolabs, Beverly, MA), resolved in 3% NuSieve 3:1 agarose gels (FMC Bioproducts), and stained with ethidium bromide (21).
Statistical Analysis
Frequencies of various demographic variables between cases and controls and between phase I and phase II subjects were compared using contingency tables and χ2 tests. We compared the observed genotypic frequencies to those expected under the Hardy-Weinberg law using χ2 tests. We estimated gene main effects [odds ratio (OR) and 95% confidence interval (95% CI)] using unconditional logistic regression adjusting for race and the matching variables age, sex, clinic, and exam date. We had previously found that using unconditional logistic regression adjusting for matching variables led to the same results as a conditional logistic regression analysis, with the added advantage that it allowed us to use all available genotypes in our study set (22, 23). We tested for heterogeneity of the ORs by race using logistic regression models including interaction variables between genotypes and race.
For the analysis of dietary fat intake, we used energy-adjusted variables using the residual energy adjustment approach (24). Heteroscedasticity is a frequent concern when analyzing dietary data; therefore, we also ran analyses using log-transformed dietary intake variables. Our results did not change from those observed using nontransformed variables; therefore, we present the results of our analyses using nontransformed data. We used the median intake among controls to generate dichotomous variables of dietary fat intake (low/high), and we used these variables in our gene-diet analyses. Analyses using quartiles supported our findings; however, given the imprecision of estimates obtained with these more complex models we only present results for dichotomous variables. To investigate “total antioxidant intake”, we obtained the sum of each of the following nutrient intakes (from foods and supplements) expressed as a dichotomous variable, using the median among controls as the cut point: selenium, zinc, vitamin C, vitamin E, and β-carotene. We then calculated the median “total intake” and used that to create a dichotomous variable of “total antioxidant intake”.
We tested for gene-fat interaction using logistic regression models adjusting for age (continuous), sex, race (White, Latinos, African American or Asian/Pacific Islander), clinic (Bellflower or Sunset), exam date, total caloric intake (continuous), body mass index (>27.5 or ≤27.5 kg/m2), vigorous leisure time activity (continuous as MET-hr/wk), and smoking (never-past-current). We tested for possible correlations between different energy-adjusted fat intake variables and found that MUFA intake was highly correlated with saturated fat intake (R = 0.836), moderately correlated with PUFA (R = 0.555) and marginally correlated with PUFA ω-6/ω-3 ratio (R = 0.310). We also observed a marginal correlation between saturated fat and PUFA (R = 0.227) and PUFA ω-6/ω-3 ratio (R = 0.190). To account for these correlations we adjusted all our models for saturated fat intake (dichotomous energy adjusted) and saturated fat intake × gene interaction. When analyzing for gene × MUFA intake, we also adjusted for PUFA intake (dichotomous energy adjusted) and PUFA × gene interaction, and when analyzing gene × PUFA ω-6/ω-3 ratio, we also adjusted for MUFA intake and MUFA × gene interaction. Models that tested for gene-antioxidant interaction and gene × fat × antioxidant interactions were further adjusted for alcohol (g/d), dietary fiber (g/d, continuous), nonsteroidal anti-inflammatory drug use (yes/no), and folate (μg/d, continuous). We tested for gene-nutrient interaction on a multiplicative scale using logistic regression models that included product terms between the gene and each nutrient in addition to the terms present in the main effect model and used likelihood ratio tests to compare these models with models that assumed no interaction. Similarly, we tested for gene-fat-antioxidant interaction using logistic regression models that included product terms between the gene and each nutrient, a product term between both nutrients, and a product term between both nutrients and the gene and we used likelihood ratio test to compare these models to models that assumed no three-way interaction. All analyses were done using the statistical software STATA version 8 (STATA Co., College Station, TX).
Results
Demographic characteristic of phase I subjects have been previously reported (18). Among phase II subjects, cases and controls did not differ in median age (60 years for controls and 61 years for cases), gender (60% males for controls and 61% males for cases), or ethnicity (controls: 47% Whites, 24.7% Latinos, 14.8% African Americans, 12.4% Asian/Pacific Islander, and 1.2% other ethnicities; cases: 47.5% Whites, 17.6% Latinos, 19.2% African Americans, 13.8% Asian/Pacific Islander, and 1.9% other ethnicities). The slight differences between phase II cases and controls for Latinos and African Americans was not statistically significant (Pearson χ2 P = 0.165). Nonetheless, analyses are adjusted for race to avoid confounding from ethnicity. Phase II subjects did not differ from phase I subjects in age and gender. In addition, frequencies for all ethnic groups within phase I cases and controls did not differ from those observed among phase II subjects (Pearson χ2 P = 0.177 for controls and P = 0.543 for cases).
XRCC1 and XRCC3 and Colorectal Adenoma Risk
Using subjects from phase I (466 cases and 509 controls) and phase II (287 cases and 290 controls) of the study, we genotyped for the XRCC1 codon 194, XRCC1 codon 399, and XRCC3 codon 241 SNPs, for a total of 753 cases and 799 controls. Genotypic frequencies between phase I and phase II controls (all races combined) were not statistically different; therefore, we analyzed both study phases combined (Pearson χ2 P = 0.07 for XRCC1 codon 194, P = 0.333 for XRCC1 codon 399, and P = 0.154 for XRCC3 codon 241). Furthermore, no significant differences were observed between phase I and phase II controls when comparing genotypic frequencies within each ethnic group. Genotypic counts and allelic frequencies for these three polymorphisms by ethnic group are shown in Table 1. We only observed deviations from the expected values according to the Hardy-Weinberg law for the XRCC3 codon 241 polymorphism among African Americans (P = 0.045).
Genotypic frequencies among controls, phase I and phase II subjects combined
. | Caucasians (n = 412) . | African Americans (n = 125) . | Latinos (n = 163) . | Asian/Pacific Islander (n = 87) . | ||||
---|---|---|---|---|---|---|---|---|
XRCC1 codon 194 (%) | ||||||||
Arg/Arg | 360 (88) | 114 (91) | 127 (78) | 36 (41) | ||||
Arg/Trp | 50 (12) | 11 (9) | 34 (21) | 43 (50) | ||||
Trp/Trp | 0 (0) | 0 (0) | 2 (1) | 8 (9) | ||||
Trp allelic frequency | 0.06 | 0.04 | 0.12 | 0.34 | ||||
HWE test, P | 0.189 | 0.607 | 0.870 | 0.338 | ||||
Missing | 2 | |||||||
XRCC1 codon 399 (%) | ||||||||
Arg/Arg | 159 (39) | 82 (66) | 74 (46) | 48 (56) | ||||
Arg/Gln | 178 (44) | 37 (30) | 77 (48) | 31 (36) | ||||
Gln/Gln | 71 (17) | 5 (4) | 10 (6) | 7 (8) | ||||
Gln allelic frequency | 0.39 | 0.10 | 0.30 | 0.26 | ||||
HWE test, P | 0.096 | 0.749 | 0.084 | 0.534 | ||||
Missing | 4 | 1 | 2 | 1 | ||||
XRCC3 codon 241 (%) | ||||||||
Thr/Thr | 158 (39) | 70 (56) | 85 (52) | 65 (75) | ||||
Thr/Met | 187 (46) | 41 (33) | 67 (41) | 19 (22) | ||||
Met/Met | 64 (16) | 14 (11) | 11 (7) | 3 (4) | ||||
Met allelic frequency | 0.38 | 0.28 | 0.27 | 0.14 | ||||
HWE test, P | 0.484 | 0.045 | 0.650 | 0.294 | ||||
Missing | 6 |
. | Caucasians (n = 412) . | African Americans (n = 125) . | Latinos (n = 163) . | Asian/Pacific Islander (n = 87) . | ||||
---|---|---|---|---|---|---|---|---|
XRCC1 codon 194 (%) | ||||||||
Arg/Arg | 360 (88) | 114 (91) | 127 (78) | 36 (41) | ||||
Arg/Trp | 50 (12) | 11 (9) | 34 (21) | 43 (50) | ||||
Trp/Trp | 0 (0) | 0 (0) | 2 (1) | 8 (9) | ||||
Trp allelic frequency | 0.06 | 0.04 | 0.12 | 0.34 | ||||
HWE test, P | 0.189 | 0.607 | 0.870 | 0.338 | ||||
Missing | 2 | |||||||
XRCC1 codon 399 (%) | ||||||||
Arg/Arg | 159 (39) | 82 (66) | 74 (46) | 48 (56) | ||||
Arg/Gln | 178 (44) | 37 (30) | 77 (48) | 31 (36) | ||||
Gln/Gln | 71 (17) | 5 (4) | 10 (6) | 7 (8) | ||||
Gln allelic frequency | 0.39 | 0.10 | 0.30 | 0.26 | ||||
HWE test, P | 0.096 | 0.749 | 0.084 | 0.534 | ||||
Missing | 4 | 1 | 2 | 1 | ||||
XRCC3 codon 241 (%) | ||||||||
Thr/Thr | 158 (39) | 70 (56) | 85 (52) | 65 (75) | ||||
Thr/Met | 187 (46) | 41 (33) | 67 (41) | 19 (22) | ||||
Met/Met | 64 (16) | 14 (11) | 11 (7) | 3 (4) | ||||
Met allelic frequency | 0.38 | 0.28 | 0.27 | 0.14 | ||||
HWE test, P | 0.484 | 0.045 | 0.650 | 0.294 | ||||
Missing | 6 |
NOTE: Abbreviation: HWE, Hardy-Weinberg equilibrium test.
We tested for heterogeneity of the ORs by phase I and phase II status using likelihood ratio tests and found no evidence that OR estimates differ between these two study groups (all likelihood ratio test Ps > 0.1). Similarly, we tested for heterogeneity of the ORs by race and found no statistically significant differences, suggesting that associations between the examined genotypes and adenoma risk do not differ between the different ethnic groups within our study sample. Therefore, we calculated ORs and 95% CI for both phases and all ethnicities combined (Table 2). Very few subjects had the XRCC1 codon 194 Trp/Trp genotype (6 cases and 10 controls), therefore, we combined them with the Arg/Trp genotype. For XRCC3 codon 241 we combined the Thr/Met and the Met/Met genotypes that both showed a slight inverse association with adenoma risk. Subjects with the XRCC1 codon 399 Gln/Gln genotype were inversely associated with adenoma risk (OR, 0.7; 95% CI, 0.5-1.0; P = 0.03) when compared with Arg/Arg subjects or Arg/Arg and Arg/Gln subjects combined (OR, 0.6; 95% CI, 0.4-0.9; P = 0.01). The other investigated genotypes did not show evidence of associations with adenoma risk (Table 2).
Genotypic frequencies among cases and controls, phase I and phase II subjects combined
DNA repair genes . | Controls (%) . | Cases (%) . | OR* (95% CI) . | P . | ||||
---|---|---|---|---|---|---|---|---|
XRCC1 codon 194 | ||||||||
Arg/Arg | 639 (81) | 592 (80) | 1 (reference) | |||||
Arg/Trp | 139 (18) | 140 (19) | 1.1 (0.8-1.5) | 0.48 | ||||
Trp/Trp | 10 (1) | 6 (1) | 0.7 (0.2-1.9) | 0.46 | ||||
Arg/Arg | 639 (81) | 592 (80) | 1 (reference) | |||||
Arg/Trp + Trp/Trp | 149 (19) | 146 (20) | 1.1 (0.8-1.4) | 0.57 | ||||
XRCC1 codon 399 | ||||||||
Arg/Arg | 365 (47) | 344 (47) | 1 (reference) | |||||
Arg/Gln | 323 (41) | 332 (45) | 1.1 (0.9-1.3) | 0.48 | ||||
Gln/Gln | 94 (12) | 58 (8) | 0.7 (0.5-1.0) | 0.03 | ||||
Arg/Arg + Arg/Gln | 688 (88) | 676 (92) | 1 (reference) | |||||
Gln/Gln | 94 (12) | 58 (8) | 0.6 (0.4-0.9) | 0.01 | ||||
XRCC3 codon 241 | ||||||||
Thr/Thr | 380 (48) | 375 (51) | 1 (reference) | |||||
Thr/Met | 315 (40) | 285 (39) | 0.9 (0.7-1.1) | 0.40 | ||||
Met/Met | 92 (12) | 77 (10) | 0.8 (0.6-1.1) | 0.23 | ||||
Thr/Thr | 380 (48) | 375 (51) | 1 (reference) | |||||
Thr/Met + Met/Met | 407 (52) | 362 (49) | 0.9 (0.7-1.1) | 0.26 | ||||
Thr/Thr + Met/Met | 695 (88) | 660 (90) | 1 (reference) | |||||
Met/Met | 92 (12) | 77 (10) | 0.8 (0.6-1.2) | 0.32 | ||||
XRCC1 codon 194 and codon 399 combined | ||||||||
194-Arg/Arg and 399-Arg/Arg or Arg/Gln | 541 (69) | 530 (72) | 1 (reference) | |||||
194-Arg/Trp or Trp/Trp and 399-Arg/Arg or Arg/Gln | 147 (19) | 146 (20) | 1.0 (0.8-1.4) | 0.78 | ||||
194-Arg/Arg and 399-Gln/Gln | 93 (12) | 58 (8) | 0.6 (0.6-0.9) | 0.02 | ||||
194-Arg/Trp or Trp/Trp and 399-Gln/Gln | 1 (0) | 0 | — | — |
DNA repair genes . | Controls (%) . | Cases (%) . | OR* (95% CI) . | P . | ||||
---|---|---|---|---|---|---|---|---|
XRCC1 codon 194 | ||||||||
Arg/Arg | 639 (81) | 592 (80) | 1 (reference) | |||||
Arg/Trp | 139 (18) | 140 (19) | 1.1 (0.8-1.5) | 0.48 | ||||
Trp/Trp | 10 (1) | 6 (1) | 0.7 (0.2-1.9) | 0.46 | ||||
Arg/Arg | 639 (81) | 592 (80) | 1 (reference) | |||||
Arg/Trp + Trp/Trp | 149 (19) | 146 (20) | 1.1 (0.8-1.4) | 0.57 | ||||
XRCC1 codon 399 | ||||||||
Arg/Arg | 365 (47) | 344 (47) | 1 (reference) | |||||
Arg/Gln | 323 (41) | 332 (45) | 1.1 (0.9-1.3) | 0.48 | ||||
Gln/Gln | 94 (12) | 58 (8) | 0.7 (0.5-1.0) | 0.03 | ||||
Arg/Arg + Arg/Gln | 688 (88) | 676 (92) | 1 (reference) | |||||
Gln/Gln | 94 (12) | 58 (8) | 0.6 (0.4-0.9) | 0.01 | ||||
XRCC3 codon 241 | ||||||||
Thr/Thr | 380 (48) | 375 (51) | 1 (reference) | |||||
Thr/Met | 315 (40) | 285 (39) | 0.9 (0.7-1.1) | 0.40 | ||||
Met/Met | 92 (12) | 77 (10) | 0.8 (0.6-1.1) | 0.23 | ||||
Thr/Thr | 380 (48) | 375 (51) | 1 (reference) | |||||
Thr/Met + Met/Met | 407 (52) | 362 (49) | 0.9 (0.7-1.1) | 0.26 | ||||
Thr/Thr + Met/Met | 695 (88) | 660 (90) | 1 (reference) | |||||
Met/Met | 92 (12) | 77 (10) | 0.8 (0.6-1.2) | 0.32 | ||||
XRCC1 codon 194 and codon 399 combined | ||||||||
194-Arg/Arg and 399-Arg/Arg or Arg/Gln | 541 (69) | 530 (72) | 1 (reference) | |||||
194-Arg/Trp or Trp/Trp and 399-Arg/Arg or Arg/Gln | 147 (19) | 146 (20) | 1.0 (0.8-1.4) | 0.78 | ||||
194-Arg/Arg and 399-Gln/Gln | 93 (12) | 58 (8) | 0.6 (0.6-0.9) | 0.02 | ||||
194-Arg/Trp or Trp/Trp and 399-Gln/Gln | 1 (0) | 0 | — | — |
Adjusted for age at diagnosis, sex, race (Caucasian, Latinos, African American, or Asian/Pacific Islander), clinic, and exam date.
Both XRCC1 codon 194 and codon 399 were in linkage disequilibrium (P < 0.001); therefore, we analyzed possible associations between adenoma risk and the different “combined genotypes” that result from the presence of these two SNPs. For these codon 194 to 399 combined genotypes, we considered dichotomous codon 194 and codon 399 genotypes as follows: due to small numbers for the codon 194 to Trp/Trp genotype, we combined Arg/Trp with Trp/Trp subjects; for codon 399, we combined Arg/Arg with Arg/Gln given that the Arg/Gln showed no evidence of being associated with adenoma risk (Table 2). There were no subjects in our study population that carried both the codon 194 Trp/Trp and codon 399 Gln/Gln genotype, and only one subject (a control) carried the codon 194 Arg/Trp and codon 399 Gln/Gln genotypes, suggesting that the codon 194 Trp/codon 399 Gln haplotype is very rare in our study population. Using as reference group the more prevalent combined genotype (codon 194 Arg/Arg and codon 399 Arg/Arg or Arg/Gln), we observed that subjects that carried the codon 194 Arg/Arg genotype in combination with the codon 399 Gln/Gln genotype had a significant inverse association with risk (OR, 0.6; 95% CI, 0.6-0.9; Table 2).
We carried out further analyses of gene-fat and gene-antioxidant interaction using these combined XRCC1 codon 194 and codon 399 genotypes, using as reference low intake of fat, or high antioxidant intake, and the most common combined genotype (codon 194 Arg/Arg and codon 399 Arg/Arg or Arg/Gln). Due to availability of nutrient data, we were only able to carry out such analyses among phase I subjects (466 cases and 509 controls). We analyzed total MUFA, total PUFA, the ratio between ω-6/ω-3 PUFAs and total antioxidant intake, as described below.
XRCC1, XRCC3, and MUFA Intake
For XRCC1 codons 194 and 399 combined genotypes we observed evidence of modification of these SNPs on the effect of MUFA on adenoma risk (P for interaction = 0.018; Table 3). We estimated genotype-specific effects for MUFA by calculating the ratio of the combined OR in the MUFA high-exposure group to the OR in the low-exposure group (Table 3) and we found that high MUFA intake showed a positive association only among subjects with the codon 194 Arg/Arg and codon 399 Gln/Gln combined genotype (OR, 5.5; 95% CI, 1.4-22). We observed nonsignificant inverse associations among the two other combined genotypes. We found no evidence that XRCC3 codon 241 modified the effect of MUFA intake (data not shown).
XRCC1 codon 194 and codon 399 and unsaturated fat or antioxidant intake, phase I subjects
codon 194 codon 399 . | Arg/Arg Arg/Arg or Arg/Gln . | . | Arg/Trp or Trp/Trp Arg/Arg or Arg/Gln . | . | Arg/Arg Gln/Gln . | . | P for interaction . | |||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Controls/cases . | OR* (95% CI) . | Controls/cases . | OR* (95% CI) . | Controls/cases . | OR* (95% CI) . | . | |||||||
MUFA† | ||||||||||||||
Low | 165/128 | 1 (reference) | 37/48 | 1.9 (1.1-3.4) | 34/11 | 0.5 (0.2-1.2) | 0.018 | |||||||
High | 183/191 | 0.8 (0.5-1.3) | 42/40 | 1.2 (0.4-3.8) | 27/31 | 2.9 (0.6-14) | ||||||||
Test for trend | 0.434 | 0.378 | 0.016 | |||||||||||
Genotype-specific OR | 0.6 (0.2-1.8) | 5.5 (1.4-22) | ||||||||||||
ω-6/ω-3 PUFA† | ||||||||||||||
Low | 157/137 | 1 (reference) | 48/39 | 1.6 (0.9-2.9) | 34/13 | 0.4 (0.1-0.9) | 0.026 | |||||||
High | 191/182 | 1.0 (0.7-1.4) | 31/48 | 3.6 (1.7-7.4) | 27/29 | 0.9 (0.4-2.2) | ||||||||
Test for trend | 0.985 | 0.017 | 0.042 | |||||||||||
Genotype-specific OR | 2.3 (1.2-4.5) | 2.5 (1.0-6.0) | ||||||||||||
Total Antioxidants‡ | ||||||||||||||
High | 166/139 | 1 (reference) | 31/51 | 3.0 (1.6-5.5) | 32/17 | 0.6 (0.3-1.4) | 0.022 | |||||||
Low | 182/180 | 0.9 (0.6-1.3) | 48/37 | 1.1 (0.6-2.2) | 29/25 | 0.8 (0.4-1.9) | ||||||||
Test for trend | 0.609 | 0.004 | 0.457 | |||||||||||
Genotype-specific OR | 0.4 (0.2-0.7) | 1.4 (0.6-3.1) |
codon 194 codon 399 . | Arg/Arg Arg/Arg or Arg/Gln . | . | Arg/Trp or Trp/Trp Arg/Arg or Arg/Gln . | . | Arg/Arg Gln/Gln . | . | P for interaction . | |||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Controls/cases . | OR* (95% CI) . | Controls/cases . | OR* (95% CI) . | Controls/cases . | OR* (95% CI) . | . | |||||||
MUFA† | ||||||||||||||
Low | 165/128 | 1 (reference) | 37/48 | 1.9 (1.1-3.4) | 34/11 | 0.5 (0.2-1.2) | 0.018 | |||||||
High | 183/191 | 0.8 (0.5-1.3) | 42/40 | 1.2 (0.4-3.8) | 27/31 | 2.9 (0.6-14) | ||||||||
Test for trend | 0.434 | 0.378 | 0.016 | |||||||||||
Genotype-specific OR | 0.6 (0.2-1.8) | 5.5 (1.4-22) | ||||||||||||
ω-6/ω-3 PUFA† | ||||||||||||||
Low | 157/137 | 1 (reference) | 48/39 | 1.6 (0.9-2.9) | 34/13 | 0.4 (0.1-0.9) | 0.026 | |||||||
High | 191/182 | 1.0 (0.7-1.4) | 31/48 | 3.6 (1.7-7.4) | 27/29 | 0.9 (0.4-2.2) | ||||||||
Test for trend | 0.985 | 0.017 | 0.042 | |||||||||||
Genotype-specific OR | 2.3 (1.2-4.5) | 2.5 (1.0-6.0) | ||||||||||||
Total Antioxidants‡ | ||||||||||||||
High | 166/139 | 1 (reference) | 31/51 | 3.0 (1.6-5.5) | 32/17 | 0.6 (0.3-1.4) | 0.022 | |||||||
Low | 182/180 | 0.9 (0.6-1.3) | 48/37 | 1.1 (0.6-2.2) | 29/25 | 0.8 (0.4-1.9) | ||||||||
Test for trend | 0.609 | 0.004 | 0.457 | |||||||||||
Genotype-specific OR | 0.4 (0.2-0.7) | 1.4 (0.6-3.1) |
Adjusted for age at diagnosis, sex, race (Caucasian, Hispanic, African American, Asian/Pacific Islander), clinic, exam date, total caloric intake (continuous), body mass index (continuous), exercise (continuous), smoking status (never, quit, current), saturated fat intake (dichotomous energy adjusted), gene × saturated fat intake. Models to estimate gene × MUFA interaction were further adjusted by PUFA intake (dichotomous energy adjusted) and gene × PUFA. Models that estimate gene × ω-6/ω-3 ratio PUFA × gene were further adjusted by MUFA intake (dichtomous energy adjusted) and gene × MUFA. Models that estimate gene × antioxidant intake were further adjusted by alcohol (continuous), fiber (continuous), folate (continuous), and use of nonsteroidal anti-inflammatory drugs (yes/no).
Fat intake values were adjusted for energy intake using the regression method of Willett and Stampfer (24); values were dichotomized into low and high using the median among controls.
Sum of Zinc, selenium, vitamin C, vitamin E, and β-carotene intake values expressed as low/high using median among controls as cut point. Antioxidant intake values were adjusted for energy intake using the regression method of Willett and Stampfer (24). Values were dichotomized into low and high using the median among controls.
XRCC1, XRCC3, and PUFA Intake
We found no evidence of an association between total PUFA intake and adenoma risk among either three XRCC1 combined genotypes and no evidence of effect modification (data not shown). Similarly, the lack of association between total PUFA intake and adenoma risk did not change when we took into account the XRCC3 codon 241 polymorphism (data not shown).
We calculated the ratio between ω-6 PUFAs (linoleic acid and arachidonic acid) and ω-3 PUFAs (eicosapentaenoic acid, docosahexanoic acid, linolenic acid, and docosapentanoic acid), as an indicator of the balance between these two PUFAs reported to have putative opposite effects. We observed that the effect of high ω-6/ω-3 PUFAs ratios was modified by the XRCC1 genotypes (P for interaction = 0.026; Table 3). We estimated genotype-specific effects for ω-6/ω-3 PUFAs ratios and found that a positive association with adenoma risk was only observable among subjects with the XRCC1 codon 194 Arg/Arg and codon 399 Gln/Gln combined genotype (OR, 2.5; 95% CI, 1.0-6.0) or XRCC1 codon 194 Arg/Trp or Trp/Trp and codon 399 Arg/Arg or Gln/Gln combined genotypes (OR, 2.3; 95% CI, 1.2-4.5). We found no evidence of an XRCC3 codon 241 effect modification on the effect of high ω-6/ω-3 PUFAs intake ratios (data not shown).
XRCC1, XRCC3, and Antioxidant Intake
An overall test of interaction suggested an interaction between total antioxidant intake and the XRCC1 codon 194 and codon 399 combined genotypes (P for interaction = 0.022; Table 3). Genotype-specific ORs indicated that whereas low antioxidant intake had a nonsignificant positive association with adenoma risk among subjects with the codon 194 Arg/Arg and codon 399 Gln/Gln genotypes (OR, 1.4; 95% CI, 0.6-3.1), there was a significant inverse association among subjects with the codon 194 Arg/Trp or Trp/Trp and codon 399 Arg/Arg or Arg/Gln genotypes (OR, 0.4; 95% CI, 0.2-0.7; Table 3). We found no evidence of gene-total antioxidant intake interaction for the XRCC3 codon 241 polymorphisms (data not shown).
XRCC1, Total Antioxidant Intake, and Fat Intake Combined
We next investigated whether the previously described interactions between unsaturated fat and the XRCC1 combined genotypes could be further modified by antioxidant intake and vice versa. We found no evidence of a three-way interaction among XRCC1 combined genotypes, MUFA, and total antioxidant intake (likelihood ratio test χ22df = 0.70; P = 0.704). The previously observed inverse association between total antioxidant intake and adenoma risk among subjects with the codon 194 Arg/Trp or Trp/Trp and codon 399 Arg/Arg or Arg/Gln combined genotypes did not differ by MUFA intake. In addition, the level of antioxidant intake did not modify the effect of high MUFA intake within the codon 194 Arg/Arg and codon 399 Gln/Gln genotypes (data not shown). Furthermore, we found no evidence of a MUFA-antioxidant interaction (likelihood ratio test χ21df = 1.30; P = 0.255).
Similarly, we found no evidence of a three-way interaction between XRCC1 combined genotypes, ω-6/ω-3 PUFAs intake ratios, and total antioxidant intake (likelihood ratio test χ22df = 1.04; P = 0.594). The previously observed association between high ω-6/ω-3 PUFA ratios and adenoma risk among subjects with the codon 194 Arg/Arg and codon 399 Gln/Gln genotypes or the codon 194 Arg/Trp or Trp/Trp and codon 399 Arg/Arg or Arg/Gln combined genotypes was unchanged by total antioxidant intake. We also did not find a ω-6/ω-3 PUFA ratio-total antioxidant interaction (likelihood ratio test χ21df = 0.21; P = 0.644).
Discussion
The results of our study suggest that XRCC1 seems to play a relevant role in the repair of damage induced by unsaturated fats, and that SNPs in this gene may modify the effect of unsaturated fats and antioxidant intake in adenoma development. This effect modification may partially explain the inconclusive and controversial data on unsaturated dietary fat and colorectal adenomas risk. The three SNPs we studied have been previously found to be involved in cancer risk for a variety of cancers, as reviewed by Goode et al. (25), including two studies in colorectal cancer on XRCC1 codon 399 (26) and XRCC3 codon 241 (27). One study in colorectal adenomas examined the XRCC3 codon 241 SNP and found no association of this allele with cancer risk (28). To our knowledge, there are no previous studies that have investigated the XRCC1 SNPs we report here in colorectal adenoma risk.
Our analyses suggest that increasing levels of MUFA seem to increase risk for colorectal adenomas only among subjects with the XRCC1 codon 194 Arg/Arg and codon 399 Gln/Gln combined genotypes. Our data suggest that the risk effect of MUFA within these subjects is independent of antioxidant intake. Increasing levels of ω-6/ω-3 PUFA ratios seem to be a risk factor only among subjects with the XRCC1 codon 194 Arg/Arg and codon 399 Gln/Gln or the codon 194 Arg/Trp or Trp/Trp and codon 399 Arg/Arg or Arg/Gln combined genotypes, regardless of antioxidant intake. These findings suggest that whereas most subjects with the XRCC1 codon 194 Arg/Arg and codon 399 Gln/Gln combined genotypes have a lower chance of developing colorectal adenomas, these subjects tend to be more susceptible to diets high in unsaturated fatty acids. This could be explained if at high levels of MUFA or ω-6 PUFA intake, or low levels of ω-3 PUFA intake, the protective effect of the XRCC1 codon 194 Arg/Arg genotype in combination with the codon 399 Gln/Gln genotype were overwhelmed by the increased DNA damage induced by these unsaturated fatty acids. Subjects with the XRCC1 codon 194 Arg/Trp or Trp/Trp and codon 399 Arg/Arg or Arg/Gln seem to have a higher risk of colorectal adenomas at low levels of unsaturated fat, and this risk is even higher when ω-6/ω-3 PUFA ratios are high.
Our finding of a positive association between high intake of MUFAs and adenoma risk among subjects with the XRCC1 codon 194 Arg/Arg and codon 399 Gln/Gln combined genotypes is puzzling. Whereas MUFAs can be susceptible to free radical–initiated damage, studies of MUFA-induced DNA damage are limited and those available failed to find associations between MUFA intake and malondialdehyde-DNA adducts (M1-dG) in blood cells (29) or colorectal mucosa (30). Nonetheless, an experimental study by Calder et al. reported that a diet rich in MUFAs promoted human colon tumor growth (31). In addition, in colorectal cancer cells, MUFAs were found to promote proliferation and to increase lipid peroxidation at high concentrations (32). The epidemiologic evidence for the role of MUFA in colorectal adenomas and/or carcinomas is inconsistent, as reviewed in ref. (1). A previous report using the same subjects in this study found a nonsignificant (P for trend = 0.08) positive association between increased levels of MUFA intake and adenoma risk (18). A recent study by Mathew et al. reported a 2-fold increase for colorectal adenomas among subjects with high intake of MUFA (33). The main MUFA in human diets is oleic acid (ω-9 MUFA), which is present in high-oleate sunflower and safflower oil (80%), olive oil (70%), canola (60%), and avocados (60%), among others (34). The experimental evidence suggests that high intake of olive oil–derived oleic acid can reduce cancer risk, as reviewed by Bartsch et al. (2); therefore, the finding of a positive association between high MUFA intake and adenoma risk is counterintuitive. However, there are several primary foods that contribute to oleic acid intake other than olive oil, such as lard (50%), tallow (45%), and butterfat (30%), among others, which may not have a beneficial effect on cancer risk. Recently, Mathew et al. (33) has suggested that in some populations, high oleic acid intake may therefore be an indicator of an “unhealthy diet”, given the prevalence of many MUFAs in unhealthy dietary items, and this may explain the positive association with cancer risk. Of interest, olive oil is also a rich source of antioxidants present in its phenolic fraction (34); therefore, as previously suggested by Nano et al. (32), the beneficial effect observed with high intake of olive oil may be due to these phenolic compounds and not necessarily to the oleic acid content. Studies that distinguish the proportion of MUFA derived from the various dietary factors mentioned above (e.g., olive oil versus butterfat or lard) may help tease out this issue. An alternative explanation might be that ω-9 MUFAs may have other roles in the metabolism of fatty acids and carcinogenesis that are not currently known, and that perhaps after free radical peroxidation they are able to induce other types of DNA damage besides the ones that have been studied thus far (malondialdehyde-DNA adducts and 8-oxo-G).
We observed an unexpected but strong inverse association between low antioxidant intake and adenoma risk, only among subjects with the codon 194 Arg/Trp or Trp/Trp and codon 399 Arg/Arg or Arg/Gln combined genotypes, an effect that was independent of fat intake. This inverse association could be explained by the previously reported proapoptotic effects of low antioxidant intake in colon epithelium (35). As proposed by Salganik et al. (36), among certain reactive oxygen species levels high intake of antioxidant may be harmful if it prevented the elimination of highly mutated cells through phagocytosis and apoptosis. In support of this, it was recently found that in colon cancer cells vitamin C can inhibit apoptosis after exposure to various carcinogens by inhibiting the down-regulation of bcl-XL that needs to occur to trigger apoptosis (37). Our data suggest that among subjects with the codon 194 Arg/Trp or Trp/Trp and codon 399 Arg/Arg or Arg/Gln combined genotypes, the reactive oxygen species levels that accumulate in the colon may be such that low intake of antioxidants might be highly beneficial.
The data on the functional consequences of the XRCC1 polymorphisms is scarce; therefore, it is difficult to provide a mechanistic explanation for our findings. However, phenotype association studies suggest that the XRCC1 codon 194 Arg allele may be associated with a lower repair proficiency than the codon 194 Trp allele (38), and that the codon 399 Gln allele may be associated with a lower repair proficiency than the codon 399 Arg allele (38-41). In contrast with these findings, a report by Taylor et al. using Chinese hamster ovary cells transfected with expression constructs carrying the codon 399 Arg or the Gln isoforms suggests that these two alleles do not differ in their sensitivity to methyl methanesulfonate (42). Given that the effects of the Gln allele on XRCC1 protein function may be more subtle than those detectable with the survival assay used by Taylor et al., clearly, more functional studies are needed to clarify this issue and reconcile the different pieces of evidence.
Our findings for the XRCC1 gene suggest a protective effect for the codon 194 Arg and codon 399 Gln alleles combined, only present at low levels of unsaturated fat intake. These results are in apparent contradiction with the phenotypic findings mentioned above. As we have previously suggested (20), a possible interpretation is that perhaps lower BER repair proficiency might have a protective effect in the development of certain cancers due to a higher chance of highly damaged cells undergoing apoptosis instead of BER, which can be an error prone pathway leading to the accumulation of further mutations if extensively used (43, 44). Under this hypothesis, the XRCC1 codon 194 Arg allele and the codon 399 Gln allele would be expected to be less frequent among cancer cases, which is what we observed in our study. The possibility that these SNPs are just markers of another neighboring putative susceptibility gene cannot be discarded.
A potential bias in our study could be introduced by underreporting of dietary intake, particularly if such underreporting varied with total caloric intake. As reported by Kipnis et al., the use of energy-adjusted variables, such as we have done in our study, can ameliorate the effects of this type of measurement error (45). Another issue of concern is recall bias. In our study, subjects were informed of the results of their sigmoidoscopy before interview; therefore, we cannot discard the possibility that subjects with confirmed adenomas remembered their consumption of fat and antioxidants differently than those subjects without adenomas.
Our study provides evidence that in trying to understand the role of unsaturated fat in adenoma risk it may be important to take into account variants in genes that participate in pathways involved in processing the effects of these fats. This is a first step in trying to understand the complex unsaturated fat-lipid peroxidation-DNA damage pathway. Clearly, many more genes other than the ones studied here play a role, such as those that participate in fatty acid metabolism (e.g., lipooxygenases and cyclooxygenases) and oxidative damage detoxification (e.g., glutathione S-transferases). Modeling strategies that allow a simultaneous analysis of all relevant factors in a complex pathway will be useful in better understanding the role of fat and adenoma risk.
Grant support: American Cancer Society grant IRG-58-007-44 and National Cancer Institute grant 1RO1CA096830.
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 Dr. Joan Levine, Jesse Lin, and Terry Kolb for their assistance with data cleaning and management; Anh Diep for her assistance with biospecimen management; and all the participants of this study for their collaboration.