Abstract

O6 -methylguanine DNA methyl-transferase ( MGMT ) is the only known critical gene involved in cellular defense against alkylating agents in the DNA direct reversal repair (DRR) pathway. Three single nucleotide polymorphism (SNP) coding for non-conservative amino acid substitutions have been identified [C250T (Leu84Phe), A427G (Ile143Val) and A533G (Lys178Arg)]. To examine the importance of the DRR pathway in risk for breast cancer and the potential interaction with cigarette smoking and dietary antioxidants, we genotyped for these variants using biospecimens from the Long Island Breast Cancer Study Project. Genotyping was performed by a high throughput assay with fluorescence polarization and included 1067 cases and 1110 controls. Overall, there was no main effect between any variant genotype, haplotype or diplotype and breast cancer risk. Heavy smoking (>31 pack-year) significantly increased breast cancer risk for women with the codon 84 variant T -allele [odds ratio, OR = 3.0, 95% confidence interval (95% CI) = 1.4–6.2]. An inverse association between fruits and vegetables consumption and breast cancer risk was observed among women with the wild-type genotype for codon 84 (OR = 0.8, 95% CI = 0.6–0.9 for ≥35 servings of fruits and vegetables per week and CC genotype versus those with <35 servings per week and CC genotype). The association between fruits and vegetables consumption and reduced breast cancer risk was apparent among women with at least one variant allele for codon 143 (OR = 0.6, 95% CI = 0.5–0.9 for ≥35 servings of fruits and vegetables per week and AG or GG genotype versus those with <35 servings per week and AA genotype). Similar patterns were observed for dietary α-carotene and supplemental β-carotene, but not for supplemental vitamins C and E. These data suggest that polymorphisms in MGMT may modulate the inverse association previously observed between fruits and vegetables consumption, dietary antioxidants and breast cancer risk, and support the importance of fruits and vegetables on breast cancer risk reduction.

Introduction

The DNA repair protein O6 -methylguanine DNA methyl-transferase (MGMT, also known as AGT, AGAT and ATase) is the only known protein involved in the cellular defense against alkylating agents in the DNA direct reversal repair (DRR) pathway in mammals ( 1 , 2 ). Its normal physiological role is protection of cells against the potentially adverse effects of DNA alkylation damage at the O6 -position of guanine generated by endogenous and exogenous alkylating species (including cigarette smoking, environmental contaminants and diet) ( 1 , 35 ). The human MGMT gene located on chromosome band 10q26 has five exons spanning nearly 300 kb of genomic DNA. Thus far, three single nucleotide polymorphisms (SNP) coding for non-conservative amino acid substitutions within the transcribed region have been confirmed [C250T (Leu84Phe), A427G (Ile143Val) and A533G (Lys178Arg)] ( 6 ). Several studies have explored the association between these polymorphisms and risk of lung cancer, melanoma, and glioblastoma ( 613 ). Results were inconsistent, possibly due to low power associated with the relatively small sample sizes or because cases and controls were not drawn from the same underlying source population ( 613 ). There are little data on how these polymorphisms relate to functionality. Some studies indicated that the codon 84 variant has similar activity and physicochemical properties as the wild-type protein ( 14 ) and the Ile143Val polymorphism are located almost adjacent to the alkyl acceptor Cys145 in the active site of the MGMT gene ( 15 ). If the polymorphisms or haplotypes contribute to inter-individual difference in MGMT expression levels or function, they might have a significant influence on ability to remove alkyl groups because MGMT is the only enzyme involved in the DRR pathway ( 1 , 6 , 8 , 16 ).

Nitrosamines are one class of carcinogens found in cigarette smoke and are the major environmental source of exposure to alkylating agents for smokers ( 17 , 18 ). Compounds in food may also be nitrosated to agents that alkylate DNA ( 18 , 19 ). Nitrosamines induce the formation of O6 -alkylguanine adducts, that are mainly repaired by the protein MGMT in a reaction that results in alkyl group transfer to the active site cysteine ( 20 , 21 ). This transfer reaction renders the protein inactive. Therefore, constitutive levels of MGMT determine the initial repair capacity of a cell ( 22 ). Dietary antioxidants, such as vitamin C and β-carotene, that mainly come from consumption of fruits and vegetables, are well-documented inhibitors of nitrosation ( 2325 ). It has been suggested that these nutrients alter the active site of MGMT ( 9 ).

The relationships between cigarette smoking, dietary antioxidants intakes and breast cancer risk are still uncertain. A large number of case–control studies and a limited number of cohort studies have evaluated whether intake of vegetables, fruits and antioxidant micronutrients is associated with reduced overall breast cancer risk. A meta-analysis including 16 case–control studies and three cohort studies indicated a 25% decreased breast cancer risk for high consumption of vegetables and a 6% lower risk for high consumption of fruits ( 26 ). However, a recent prospective study ( 27 ) and pooled analysis of cohort studies ( 28 ) showed no evidence for a protective effect of the intake of total or specific vegetable and fruit for breast cancer. The discrepancies may be partly related to the features of different study designs, such as recall and selection biases in case–control studies, and underestimated in prospective studies because of the combined effects of imprecise dietary measurements, and limited variability of dietary intakes within each cohort ( 29 ). A modest protective role of fruit and vegetable intake could exist for a subgroup of women, such as those who are postmenopausal or with estrogen receptor (ER) positive tumors ( 30 ), although the molecular mechanisms for such a protective effect remains to be clarified.

Similarly, a recently published meta-analysis found that cigarette smoking had little or no independent effect on the risk of developing breast cancer ( 31 ). This is consistent with findings in the Long Island Breast Cancer Study Project (LIBCSP) ( 32 ). However, several studies have suggested that a positive association of breast cancer with smoking may emerge among specific subgroups, i.e. chronic heavy smokers, smoking exposure prior to first-term pregnancy or 30–40 years after commencement ( 3335 ). An increasing number of studies are also indicating that a positive association between smoking and breast cancer risk may be stronger among or limited to women with certain genotypes ( 33 ).

Based on the fact that cigarette smoking can result in mutagenic O6 -alkylguanine DNA lesions that are repaired mainly by MGMT , and that dietary antioxidants, as one of the major inhibitors of nitrosation, decrease alkylation-damage-lessening of the demand on MGMT capacity ( 25 , 36 , 37 ), it is biologically plausible to hypothesize that the potential associations between dietary antioxidants, cigarette smoking and breast cancer risk may be influenced by heterogeneity in MGMT genotype/haplotype status. In the present study, we genotyped three variants of MGMT at codon 84 of exon 3, and codons 143 and 178 of exon 5, and assessed their association with breast cancer risk and the interactions with cigarette smoking and dietary antioxidants in a large population-based case–control study.

Materials and methods

Study design

The study methods of the LIBCSP have been described in detail previously ( 38 , 39 ). In brief, cases were adult female residents of Nassau and Suffolk counties on Long Island, NY, who were of any age or race, spoke English, and were newly diagnosed with in situ or invasive breast cancer between August 1, 1996, and July 31, 1997. Controls were frequency matched to the expected age distribution of the cases, and identified through random-digit dialing for women under age 65 years and through the Center for Medicare and Medicaid Services (CMS) rosters for women 65 and over. Eligible controls were women who spoke English, and resided in the same Long Island counties as the cases, but with no personal history of breast cancer. This study was conducted with approval from participating institutional review boards, and in accordance with an assurance filed with and approved by the United States Department of Health and Human Services. In-person interviews were completed for 82.1% of cases ( n = 1508) and 62.8% of controls ( n = 1556). Of those who completed an interview, 73.1% of cases (1102) and 73.3% of controls (1141) donated a blood sample ( 39 ). As previously reported, an increase in breast cancer among women on Long Island was found to be associated with lower parity, late age at first birth, little or no breastfeeding, a family history of breast cancer, and increasing income and education. Results were similar when the analyses were restricted to respondents who donated blood, or for those with DNA available for these analyses ( 39 ). Factors that were found to be associated with a decreased likelihood that a respondent would donate blood include increasing age and past smoking status. Case–control status and fruit and vegetable consumption were not predictors of blood donation.

Laboratory methods

Genomic DNA was extracted by standard RNase/proteinase K and phenol/chloroform treatment and genotyped by a fluorescence polarization (FP) method using a commercial AcycloPrime™-FP SNP Detection Kit obtained from PerkinElmer Life Sciences ( 40 ). The forward and reverse primers were designed according to the human MGMT gene sequence (GeneBank accession number NT_008818) and were, respectively, 5′-CTA AGC CCC TGT TCT CAC TTT T-3′, 5′-ACA CCG CAG ATG GCT TAG TTA C-3′ for codon 84 and 5′-CGT TGT CCA GAT CCC TGA CT-3′, 5′-GTC GCT CAA ACA TCC ATC CT-3′ for both codons 143 and 178. Because codons 143 and 178 are both within exon 5 of MGMT gene, a single PCR amplification was performed for screening these genotypes. The sequences of TDI (Template-directed Dye-terminator Incorporation) probes were designed using Primer 3 software ( http://frodo.wi.mit.edu/cgi-bin/primer3/primer3www.cgi ) as forward 5′-GAG TTC CCC GTG CCG GCT-3′ for codon 84, forward 5′-GTC TTC CAG GTC CCC ATC CTC-3′ for codon 143 and reverse 5′-GGC AGG CTT GGG AGG GAG-3′ for codon 178. The two allele-specific dye terminators used for each of the three SNPs are C/T for codon 84, G/A for codon 143 and C/T for codon 178. Full technical details are available on request. The assay was validated by sequencing subjects with all three genotypes and these known samples were used as positive controls on each plate. The laboratory staff was blind to the case–control status of subjects. Duplicate quality control samples were randomly included in genotyping, and only 0.8% (3/372) of all genotypes being discordant. Genotyping data were available for 1067 (96.8%) cases and 1110 (97.3%) controls who donated blood, which represent 70.8% (1067/1508) of eligible cases and 71.3% (1110/1556) of eligible controls.

Questionnaire data

Cigarette smoking, defined as never or ever active smoking for these analyses, was assessed by questionnaire administered in-person by a trained interviewer ( 32 ). Light, moderate, and heavy smokers were categorized by using the 50th and 75th percentile pack-year [(cigarettes per day/20) × (years smoked)] values of the controls as the cutoff points (i.e. <15, 15–31 and >31 pack-years). To assess diet for the 12 months prior to the reference date, 98% of participants (98.2% of cases and 97.6% of controls) completed a self-administered modified NCI-Block food frequency questionnaire (FFQ). Fruit and vegetable intake was calculated and intake of carotenoids, ascorbic acid, and α-tocopherol were computed from food composition data as described previously ( 30 ). Dichotomous cutpoints were generated by collapsing the lowest three-fifths and highest two-fifths based on the similarity of the odds ratios (OR) for breast cancer risk in those groups ( 30 ). Other factors considered as potential confounders (see below) were also assessed as part of the in-person interview.

Statistical methods

Hardy–Weinberg equilibrium (HWE) was tested to compare the observed and expected genotype frequencies among cases and controls, respectively ( 41 ). Univariate analyses were done to compare distributions of covariates among cases and controls. Unconditional logistic regression with SAS version 9.0 was used to estimate OR and corresponding 95% confidence intervals (CI) adjusting for the frequency matching variable age (as a continuous variable), in addition to other potential confounding factors ( 42 , 43 ). The potential confounding variables were assessed by examining the percent changes (>10%) in the ORs of the main effects of genotyping. Variables found not to confound the associations of interest included: age at menarche, parity, lactation, months of lactation, age at first birth, number of miscarriages, history of fertility problems, alcohol drinking, race, education, religion and marital status (as previously defined) ( 38 , 39 , 44 ). So all models were only adjusted for age at reference (defined as age at diagnosis for cases and age at identification for controls). Haplotype and diplotype frequencies were estimated from genotype data by PHASE version 2.1.1 and THESIAS version 2, which are based on a Bayesian algorithm and the maximum likelihood model linked with the Stochastic-EM algorithm. Haplotypes and diplotypes were selected according to the corresponding occurring probabilities with a higher likelihood (>0.95 as cut-point) ( 4547 ). The distribution of haplotypes in the cases and controls was compared by χ 2 test. The diplotype of the most common haplotype CAA was selected as the reference in the diplotype analysis. The risk of breast cancer was estimated for each diplotype compared with the reference (CAA/CAA) with adjustment for age. Diplotype data were treated as a categorical variable and were incorporated as dummy variables in the logistic regression models. Pairwise linkage disequilibrium between any two alleles of three polymorphic sites was estimated as relative disequilibrium ( D′ ) from the haplotype data by the Fisher's exact test.

We evaluated interactions on an additive scale between MGMT genotypes and alkylating agent exposures (cigarette smoking status, smoking pack-year) and intake of dietary and/or supplemental antioxidants by using indicator terms for those with the genotype only, exposure only, and those with both the genotype and exposure of interest ( 48 , 49 ). The magnitude of an additive interaction effect was determined by estimating the age-adjusted interaction contrast ratio (ICR) with the following formula ( 49 ). ICR = OR eg − OR e − OR g + 1, where OR eg is the OR for exposure with mutant genotype, OR e is the OR for exposure with a wide-type genotype, and OR g is the OR for mutant genotype among non-exposed. The 95% CI of the ICR was obtained from ICR ± 1.96 SE (ICR). Effect modification was further assessed on a multiplicative scale by calculating ORs for MGMT genotypes stratified by indicators of cigarette smoking and intake of antioxidant levels as defined previously ( 30 , 32 , 38 , 44 ), running separate models by including multiplicative interaction terms in the logistic regression model.

Results

Among all control subjects, the variant allele frequencies of MGMT (84 T -, 143 G - and 178 G -allele) were 13.7, 12.2 and 11.0%, respectively, consistent with HWE ( P -values = 0.9, 0.1 and 0.9). Homozygous and heterozygous variant genotypes were combined to estimate age-adjusted ORs because of small numbers ( Table I ). The frequencies in controls of carrying at least one variant allele (84 T -, 143 G - and 178 G -allele) were 25.6, 23.6 and 20.9%, respectively ( Table I ). In general, there was no evidence for the main effect of any of the MGMT variants on breast cancer risk. The ORs of these polymorphisms did not change materially with menopausal status (data not shown).

Table I.

MGMT genotype, haplotype and risk for breast cancer among pre- and post-menopausal women, LIBCSP, 1996–1997

MGMT genotype or haplotype status   Cases
 
  Controls
 
  Odds ratio (95% CI) a 

 
N = 1067
 
%
 
N = 1110
 
%
 

 
Codon 84 genotyes      
    CC 778 73.1 824 74.4 1.0 (Reference) 
    CT or TT 286 26.9 283 25.6 1.1 (0.9–1.3) 
    CT 265 24.9 263 23.8 1.1 (0.9–1.3) 
    TT 21 2.0 20 1.8 1.1 (0.6–2.0) 
Codon 143 genotpyes      
    AA 829 77.9 843 76.4 1.0 (Reference) 
    AG or GG 235 22.1 260 23.6 0.9 (0.8–1.1) 
    AG 224 21.1 250 22.7 0.9 (0.7–1.1) 
    GG 11 1.0 10 0.9 1.1 (0.5–2.6) 
Codon 178 genotpyes      
    AA 859 80.5 878 79.1 1.0 (Reference) 
    AG or GG 208 19.5 232 20.9 0.9 (0.7–1.1) 
    AG 197 18.5 219 19.7 0.9 (0.7–1.1) 
    GG 11 1.0 13 1.2 0.9 (0.4–1.9) 
Haplotype b      
    CAA  1600 c 75.4  1658 c 75.2 1.0 (Reference) 
    TAA 272 12.8 266 12.1 1.0 (0.9–1.3) 
    CGG 180 8.5 204 9.2 0.9 (0.7–1.1) 
    TGG 32 1.5 32 1.4 1.1 (0.6–2.1) 
    CGA 30 1.5 32 1.4 1.0 (0.6–1.8) 
Diplotypes (haplotype pairs)      
    CAA CAA 596 55.9 621 56.0 1.0 (Reference) 
    CAA TAA 210 19.7 198 17.8 1.1 (0.9–1.4) 
    CGG CAA 142 13.3 155 14.0 0.9 (0.7–1.2) 
    CAA TGG 43 4.0 55 5.0 0.8 (0.5–1.2) 
    CGA CAA 26 2.4 27 2.4 1.0 (0.6–1.7) 
Other haplotype pairs 50 4.7 54 4.9 0.9 (0.6–1.4) 
MGMT genotype or haplotype status   Cases
 
  Controls
 
  Odds ratio (95% CI) a 

 
N = 1067
 
%
 
N = 1110
 
%
 

 
Codon 84 genotyes      
    CC 778 73.1 824 74.4 1.0 (Reference) 
    CT or TT 286 26.9 283 25.6 1.1 (0.9–1.3) 
    CT 265 24.9 263 23.8 1.1 (0.9–1.3) 
    TT 21 2.0 20 1.8 1.1 (0.6–2.0) 
Codon 143 genotpyes      
    AA 829 77.9 843 76.4 1.0 (Reference) 
    AG or GG 235 22.1 260 23.6 0.9 (0.8–1.1) 
    AG 224 21.1 250 22.7 0.9 (0.7–1.1) 
    GG 11 1.0 10 0.9 1.1 (0.5–2.6) 
Codon 178 genotpyes      
    AA 859 80.5 878 79.1 1.0 (Reference) 
    AG or GG 208 19.5 232 20.9 0.9 (0.7–1.1) 
    AG 197 18.5 219 19.7 0.9 (0.7–1.1) 
    GG 11 1.0 13 1.2 0.9 (0.4–1.9) 
Haplotype b      
    CAA  1600 c 75.4  1658 c 75.2 1.0 (Reference) 
    TAA 272 12.8 266 12.1 1.0 (0.9–1.3) 
    CGG 180 8.5 204 9.2 0.9 (0.7–1.1) 
    TGG 32 1.5 32 1.4 1.1 (0.6–2.1) 
    CGA 30 1.5 32 1.4 1.0 (0.6–1.8) 
Diplotypes (haplotype pairs)      
    CAA CAA 596 55.9 621 56.0 1.0 (Reference) 
    CAA TAA 210 19.7 198 17.8 1.1 (0.9–1.4) 
    CGG CAA 142 13.3 155 14.0 0.9 (0.7–1.2) 
    CAA TGG 43 4.0 55 5.0 0.8 (0.5–1.2) 
    CGA CAA 26 2.4 27 2.4 1.0 (0.6–1.7) 
Other haplotype pairs 50 4.7 54 4.9 0.9 (0.6–1.4) 
a

Adjusted for age at reference (continuous).

b

MGMT haplotypes composed of three polymorphic sites: C250T, A427G and A533G.

c

No. of chromosomes.

In the haplotype and diplotype analyses, PHASE and THESIAS programs showed essentially the same results; those obtained with PHASE software are shown in Table I . Haplotype analysis indicated that there were eight haplotypes in MGMT with three common haplotypes (CAA, TAA and CGG) accounting for >95% of the alleles. The frequencies of haplotype CAA were 75.4% in cases and 75.2% in controls. TAA and CGG haplotype frequencies were also not significantly different between cases and controls. From the diplotype construction analysis, a total of sixteen diplotypes were observed in the population. The three most frequent diplotypes were comprised of CAA/CAA, CAA/TAA and CGG/CAA. No significant associations were found between these diplotypes and breast cancer risk ( Table I ). Ninety-six percent of subjects were of the same genotypes for the codon 143 and 178 ( Table II ), suggesting that the two variant alleles are in strong linkage disequilibrium (pairwise D ′ estimates was 0.96, r2 = 0.89, P < 0.01). Therefore, for all effect modification analyses, we used only the genotypes at codon 84 and 143 to evaluate relationships with breast cancer risk.

Table II.

Linkage between SNPs at MGMT codons 143 and 178, LIBCSP, 1996–1997

Genotype in codon 143  No. (%) of genotype in codon 178
 
  

 
AA
 
AG
 
GG
 
AA 1659 (99.2) 13 (0.8) 0 (0.0) 
AG 68 (14.4) 403 (85.0) 3 (0.6) 
GG 0 (0.0) 0 (0.0) 21 (100.0) 
Genotype in codon 143  No. (%) of genotype in codon 178
 
  

 
AA
 
AG
 
GG
 
AA 1659 (99.2) 13 (0.8) 0 (0.0) 
AG 68 (14.4) 403 (85.0) 3 (0.6) 
GG 0 (0.0) 0 (0.0) 21 (100.0) 

Pairwise D′ = 0.96, r2 = 0.89, P < 0.01.

No differences were observed for the association between cigarette smoking status and breast cancer risk across genotypes described by the MGMT codon 84 T -allele and 143 G -allele ( Table III ). However, when the associations were further evaluated within strata of pack-years (<15, 15–31, and >31), a significant increased risk between the codon 84 variant T -allele and heavy smoking (>31 pack-years) was observed (OR = 3.0, 95% CI = 1.4–6.2). The dose–response trend was also statistically significance ( Ptrend < 0.01). Compared with non-smokers with codon 143 wild-type AA genotype, light, moderate and heavy smokers with AA genotype had significantly higher risks of breast cancer with ORs 1.6 (95% CI = 1.2–2.3), 1.8 (95% CI = 1.2–2.8), and 2.0 (95% CI = 1.3–3.1), respectively. No similar pattern was found in smokers with AG or GG genotypes. There was also no dose–response observed between the codon 143 variant G -allele and pack-years.

Table III.

Polymorphisms of MGMT codons 84 and 143 and breast cancer susceptibility by cigarette smoking status, LIBCSP, 1996–1997

Smoking status MGMT codon 84
 
   MGMT codon 143
 
   
  CC
 
  CT or TT
 
  AA
 
  AG or GG
 
 

 
Case/control
 
OR a (95% CI)
 
Case/control
 
OR a (95% CI)
 
Case/control
 
OR a (95% CI)
 
Case/control
 
OR a (95% CI)
 
Non-smoker 364/379 1.0 (Reference) 124/122 1.1 (0.8–1.4) 383/398 1.0 (Reference) 105/101 1.1 (0.8–1.5) 
Smoker 414/444 1.0 (0.8–1.2) 162/161  1.0 (0.8–1.4) b 446/444 1.1 (0.9–1.3) 130/159  0.9 (0.7–1.1) c 
Non-smoker 364/379 1.0 (Reference) 124/122 1.1 (0.8–1.4) 383/398 1.0 (Reference) 105/101 1.1 (0.8–1.5) 
<15 pack-years 101/77 1.4 (1.0–2.0) 40/30 1.4 (0.9–2.3) 118/79 1.6 (1.2–2.3) 24/28 0.8 (0.5–2.1) 
15–31 pack-years 56/41 1.6 (1.0–2.4) 21/15 1.5 (0.8–3.1) 64/40 1.8 (1.2–2.8) 13/16 0.9 (0.4–1.9) 
>31 pack-years 64/44 1.4 (0.9–2.2) 31/10 3.0 (1.4–6.2) 72/35 2.0 (1.3–3.1) 23/19 1.2 (0.6–2.3) 
P for trend     <0.01    0.4 
Smoking status MGMT codon 84
 
   MGMT codon 143
 
   
  CC
 
  CT or TT
 
  AA
 
  AG or GG
 
 

 
Case/control
 
OR a (95% CI)
 
Case/control
 
OR a (95% CI)
 
Case/control
 
OR a (95% CI)
 
Case/control
 
OR a (95% CI)
 
Non-smoker 364/379 1.0 (Reference) 124/122 1.1 (0.8–1.4) 383/398 1.0 (Reference) 105/101 1.1 (0.8–1.5) 
Smoker 414/444 1.0 (0.8–1.2) 162/161  1.0 (0.8–1.4) b 446/444 1.1 (0.9–1.3) 130/159  0.9 (0.7–1.1) c 
Non-smoker 364/379 1.0 (Reference) 124/122 1.1 (0.8–1.4) 383/398 1.0 (Reference) 105/101 1.1 (0.8–1.5) 
<15 pack-years 101/77 1.4 (1.0–2.0) 40/30 1.4 (0.9–2.3) 118/79 1.6 (1.2–2.3) 24/28 0.8 (0.5–2.1) 
15–31 pack-years 56/41 1.6 (1.0–2.4) 21/15 1.5 (0.8–3.1) 64/40 1.8 (1.2–2.8) 13/16 0.9 (0.4–1.9) 
>31 pack-years 64/44 1.4 (0.9–2.2) 31/10 3.0 (1.4–6.2) 72/35 2.0 (1.3–3.1) 23/19 1.2 (0.6–2.3) 
P for trend     <0.01    0.4 
a

Adjusted for age at reference (continuous).

b

Pinteraction = 0.93.

c

Pinteraction = 0.12.

The association between dietary fruits and vegetables and dietary antioxidants and breast cancer risk differed by MGMT genotype ( Table IV ). The inverse relation between fruits and vegetables consumption and breast cancer risk was apparent among women with the wild type, but not the variant, genotype for codon 84 (OR = 0.8, 95% CI = 0.6–0.9 for ≥35 servings of fruits and vegetables per week and CC genotype versus those with <35 servings per week and CC genotype). There were consistent inverse associations observed between high intakes of α-carotene, β11-carotene, vitamin C or E and breast cancer risk among subjects who had the wild-type CC genotype, although none were statistically significant. The association between fruits and vegetables consumption and breast cancer risk was apparent among women with at least one variant G -allele for codon 143 (OR = 0.6, 95% CI = 0.5–0.9 for ≥35 servings of fruits and vegetables per week and AG or GG genotype versus those with <35 servings per week and AA genotype), but not with the wild-type genotype ( Table IV ). Similar patterns were observed for dietary α-carotene and supplemental β-carotene (OR = 0.6, 95% CI = 0.4–0.8 and OR = 0.7, 95% CI = 0.5–1.0, respectively), but not for supplemental vitamin C and E (OR = 0.8, 95% CI = 0.6–1.1 and OR = 0.9, 95% CI = 0.7–1.2). These ORs remained significant after controlling for other antioxidants one at a time or all simultaneously (data not shown). The age-adjusted ICR also suggested deviations from the additive scale for the 143 G -allele combined with high intake of fruits and vegetables (ICR = −0.5, 95% CI: −0.9, −0.1), and high intake of dietary α-carotene (ICR = −0.6, 95% CI: −0.2, −1.1). For carriers of the 143 G -allele with 3817.3 µg/day or higher dietary and supplemental β-carotene, the additive interaction was not statistically significant (ICR = −0.1, 95% CI: −0.5, 0.2), indicating no interaction between dietary and supplemental β-carotene and codon 143. No significant dose–response effect with the 143 G -allele was observed when we categorized total fruits and vegetables consumption based on quintiles of controls (data not shown).

Table IV.

Combined effects of MGMT genotypes and dietary and/or supplemental antioxidants intake on breast cancer susceptibility, LIBCSP, 1996–1997

Antioxidants intake a MGMT codon 84
 
   MGMT codon 143
 
   
  CC
 
  CT or TT
 
  AA
 
  AG or GG
 
 

 
Case/control
 
OR b (95% CI)
 
Case/control
 
OR b (95% CI)
 
Case/control
 
OR b (95% CI)
 
Case/control
 
OR b (95% CI)
 
Any fruits, fruit juices, and vegetables c         
    0–34 498/482 1.0 (Reference) 164/181 0.9 (0.7–1.1) 507/521 1.0 (Reference) 155/140 1.2 (0.9–1.5) 
    ≥35 268/329 0.8 (0.6–0.9) 117/99 1.0 (0.8–1.4) 307/309 1.0 (0.8–1.2) 78/117  0.6 (0.5–0.9) d 
Dietary α-carotene (µg/day)         
    0–267.7 477/470 1.0 (Reference) 176/168 1.0 (0.8–1.3) 499/506 1.0 (Reference) 156/130 1.2 (1.0–1.6) 
    ≥267.8 281/332 0.8 (0.7–1.0) 103/108 0.9 (0.7–1.2) 306/313 1.0 (0.8–1.2) 76/126  0.6 (0.4–0.8) e 
Dietary + supplemental β-carotene (µg/day)         
    0–3817.2 476/456 1.0 (Reference) 170/168 1.0 (0.8–1.2) 503/480 1.0 (Reference) 143/142 1.0 (0.8–1.3) 
    ≥3817.3 282/346 0.8 (0.6–1.0) 109/108 0.9 (0.7–1.3) 302/339 0.9 (0.7–1.1) 89/114  0.7 (0.5–1.0) f 
Dietary + supplemental vitamin C (mg/day)         
    0–131.0 453/467 1.0 (Reference) 164/158 1.1 (0.8–1.4) 480/481 1.0 (Reference) 138/142 1.0 (0.8–1.3) 
    ≥131.1 305/335 0.9 (0.8–1.1) 115/118 1.0 (0.7–1.3) 325/338 1.0 (0.8–1.2) 94/114 0.8 (0.6–1.1) 
Dietary + supplemental vitamin E (a-te/day)         
    0–29.0 449/467 1.0 (Reference) 153/167 0.9 (0.7–1.2) 470/491 1.0 (Reference) 133/143 1.0 (0.8–1.3) 
    ≥29.1 309/335 1.0 (0.8–1.2) 126/109 1.2 (0.9–1.6) 335/328 1.1 (0.9–1.3) 99/113 0.9 (0.7–1.2) 
Antioxidants intake a MGMT codon 84
 
   MGMT codon 143
 
   
  CC
 
  CT or TT
 
  AA
 
  AG or GG
 
 

 
Case/control
 
OR b (95% CI)
 
Case/control
 
OR b (95% CI)
 
Case/control
 
OR b (95% CI)
 
Case/control
 
OR b (95% CI)
 
Any fruits, fruit juices, and vegetables c         
    0–34 498/482 1.0 (Reference) 164/181 0.9 (0.7–1.1) 507/521 1.0 (Reference) 155/140 1.2 (0.9–1.5) 
    ≥35 268/329 0.8 (0.6–0.9) 117/99 1.0 (0.8–1.4) 307/309 1.0 (0.8–1.2) 78/117  0.6 (0.5–0.9) d 
Dietary α-carotene (µg/day)         
    0–267.7 477/470 1.0 (Reference) 176/168 1.0 (0.8–1.3) 499/506 1.0 (Reference) 156/130 1.2 (1.0–1.6) 
    ≥267.8 281/332 0.8 (0.7–1.0) 103/108 0.9 (0.7–1.2) 306/313 1.0 (0.8–1.2) 76/126  0.6 (0.4–0.8) e 
Dietary + supplemental β-carotene (µg/day)         
    0–3817.2 476/456 1.0 (Reference) 170/168 1.0 (0.8–1.2) 503/480 1.0 (Reference) 143/142 1.0 (0.8–1.3) 
    ≥3817.3 282/346 0.8 (0.6–1.0) 109/108 0.9 (0.7–1.3) 302/339 0.9 (0.7–1.1) 89/114  0.7 (0.5–1.0) f 
Dietary + supplemental vitamin C (mg/day)         
    0–131.0 453/467 1.0 (Reference) 164/158 1.1 (0.8–1.4) 480/481 1.0 (Reference) 138/142 1.0 (0.8–1.3) 
    ≥131.1 305/335 0.9 (0.8–1.1) 115/118 1.0 (0.7–1.3) 325/338 1.0 (0.8–1.2) 94/114 0.8 (0.6–1.1) 
Dietary + supplemental vitamin E (a-te/day)         
    0–29.0 449/467 1.0 (Reference) 153/167 0.9 (0.7–1.2) 470/491 1.0 (Reference) 133/143 1.0 (0.8–1.3) 
    ≥29.1 309/335 1.0 (0.8–1.2) 126/109 1.2 (0.9–1.6) 335/328 1.1 (0.9–1.3) 99/113 0.9 (0.7–1.2) 
a

Variables were derived from a two-step process. Quintiles of the control distribution were used to form fifths. ORs of the last two-fifths and the first three-fifths were similar and therefore collapsed to form these binary variables.

b

Adjusted for age at reference (continuous).

c

In 1/2 cup servings per week.

d

Pinteraction = 0.01.

e

Pinteraction < 0.01.

f

Pinteraction = 0.49.

When stratified by menopausal status, significant inverse associations between dietary α-carotene intake ≥267.8 (µg/day) and breast cancer risk were still observed for women with at least one variant G -allele of codon 143. The relevant ORs were 0.5 (95% CI = 0.3–0.9) for premenopausal women and 0.6 (95% CI = 0.4–0.9) for postmenopausal women. The inverse association between high intake of fruits and vegetables (≥35 servings per week) and breast cancer risk was also observed in postmenopausal women with the variant G -allele of codon 143 (OR = 0.7, 95% CI = 0.5–0.9) and in premenopausal women (OR = 0.6, 95% CI = 0.3–1.1), but was only statistically significant among postmenopausal women.

Discussion

MGMT plays an important role in an individual's ability to remove damage induced by alkylating mutagens because it is the only enzyme involved in the DRR pathway ( 1 , 6 , 8 , 16 ). However, there are few studies exploring the association between MGMT polymorphisms and cancer risk ( 50 ). Our study is a relatively large population-based study that describes polymorphisms in the MGMT gene and their potential relationship to breast cancer risk. This kind of design ensured that cases and controls arose from the same source population. In addition, our study is the first to infer haplotypes and diplotypes in MGMT , and compare their distribution frequencies in breast cancer cases and controls.

Before interpretation of the observed associations, potential bias from subject's selection or exposure recall should be recognized. Response rates to blood collection did vary by age and smoking status ( 39 ). Although older or past smoking women had lower response rates, the mean age and smoking status of the participants did not vary significantly by MGMT genotype status (data not shown). In addition, the genotypes are in HWE, and genotype frequencies are similar to most previous studies ( 6 , 814 ). Thus, the age difference or past smoking status should not affect the conclusions regarding genotype and breast cancer risk. Another potential limitation of this study is inaccurate recall of past potential modifiers, such as smoking and diet. This misclassification would not be likely to differ by genotype status, however. We also observed that the distribution of intake of total energy and micronutrients did not vary substantially by chemotherapy status at the time of interview or by number of days from diagnosis to interview ( 30 ). It is unlikely that misclassification or selection bias related to genotyping detection because our genotype had very high reliability (99.2%) and a failure rate of <0.2% (16 of 6531). The error in measuring genotype may be independent of case–control status, although this non-differential exposure misclassification does not guarantee the bias is towards the null ( 51 ). It should be recognized that non-differential misclassification might give rise to bias and an overestimate, although the results are more likely an underestimate ( 51 , 52 ). The relatively large sample size and extensive information on potential confounding factors allowed us to assess confounding. The fact that observed results from dietary antioxidants, MGMT genotyping and breast cancer association changed ORs <10% when considering potential confounding factors individually or in multivariate models indicated the effects of confounding are limited. Age at reference as a continuous variable included in logistic models minimized any possible residual confounding effect of age.

Our results suggest that heavy smoking (>31 pack-years) significantly increases breast cancer risk for codon 84 variant T -allele carriers. There was also a statistically significant dose–response with increasing amount of smoking (<15, 15–31 and >31 pack-years) ( Table III ). A previous report indicated an association between the codon 143 variant G -allele and lung cancer risk for non-smokers exposed to second-hand smoke ( 9 ). Because the phenotypic consequences of the MGMT polymorphisms are unknown, this difference between cancers of the breast and lung may be due to their different etiologies and molecular background, which needs to be investigated further.

We have described previously a reduction in postmenopausal breast cancer risk associated with fruit and vegetable intake ( 34 ). The present study provides the first evidence that the combination of the MGMT codon 143 G -allele and high intake of fruits, vegetables, and dietary α-carotene may be associated with a significant decrease in breast cancer risk. There were statistically significant gene–nutrient interactions (both multiplicative and additive). This is inconsistent with previous reports that the MGMT 143 G -allele might be a putative high-risk allele for the occurrence of cancers ( 8 , 9 ). There are two possibilities for the molecular mechanism of the inconsistent results. First, the DNA adducts ( O6 -methyl or ethylguanine) produced by alkylating agents can be repaired by cooperation between MGMT and the nucleotide excision repair pathway (NER) in human cells ( 53 ). In contrast to the protective effects against O6 -alkylating agents, MGMT has been shown to increase the toxic and mutagenic effects of dihaloalkanes such as dibromoethane (DBE) by direct interaction with DBE metabolites ( 1 , 54 , 55 ). Second, MGMT is constitutively present at active transcription sites binding to the estrogen receptor and inhibiting its function in cell proliferation. This suggests that MGMT may have a role in regulation of transcription as well as DNA repair ( 56 ). However, functional analysis using in vitro assays was not able to demonstrate that the presence of the codon 143 variant correlated with a significant change in enzyme activity ( 12 , 57 ). At the same time, vitamin C and other nitrosation inhibitors, whose main source is fruits and vegetables, are well-documented inhibitors of nitrosation ( 58 ). High intake of fruits and vegetables increases nitrosation inhibitors and might reduce individual susceptibility to DNA alkylation damage. This is a plausible explanation for the observed interaction between the MGMT 143 G -allele and high intake of fruits and vegetables in circumstances of low-level exposure to environmental carcinogens ( 59 ). This result should be interpreted cautiously for two main reasons. First, an enhanced protective interaction between the MGMT 143 G -allele and high intake of fruits and vegetables, and dietary plus supplement antioxidant intake was not a primary hypothesis and should be regarded as an exploratory analysis, with a hypothesis-generating rather than hypothesis-testing character. Second, methylation is the main epigenetic modification in mammals. The MGMT gene can be inactivated by hyper-methylation of the CpG islands located in the promoter region and leads to transcriptional silencing. Due to the unknown status of MGMT CpG island hyper-methylation, we cannot exclude the possibility that there is a relationship between 143 G -allele and hyper-methylation. It is possible that the observed protective effect of the MGMT 143 G -allele will be stronger after stratification for MGMT promoter methylation status. Studies involved in both functional polymorphisms, enzyme activity and epigenetic features of MGMT are warranted.

In conclusion, our study found no evidence that any of the three MGMT polymorphisms, haplotypes or diplotypes investigated were independently associated with breast cancer. Previously, we reported significant inverse associations between high intake of fruits, vegetables, dietary α-carotene and postmenopausal breast cancer ( 34 ); in this report we find that this protection is partially dependent upon MGMT 143 G -allele status. These findings can provide clues toward understanding the role of MGMT -related DNA repair capacity and DNA alkylation damage during breast tumorigenesis. Further studies in the functional evaluation of polymorphisms of MGMT in the DRR pathway, the examination of genetic variation in NER, DNA mismatch repair pathways and epigenetic factors in the DRR pathway will be of interest.

For their valuable contributions to the LIBCSP, the authors thank members of the Long Island Breast Cancer Network; the 31 participating institutions on Long Island and in New York City, NY; our National Institutes of Health collaborators, Gwen Colman, PhD, National Institutes of Environmental Health Sciences; G. Iris Obrams, MD, PhD formerly of the National Cancer Institute; members of the External Advisory Committee to the population-based case–control study: Leslie Bernstein, PhD (Committee chair); Gerald Akland, MS, Barbara Balaban, MSW, Blake Cady, MD, Dale Sandler, PhD, Roy Shore, PhD, Gerald Wogan, PhD, as well as other collaborators who assisted with various aspects of our data collection efforts including Gail Garbowski, MPH, Julie Britton, PhD, Mary S.Wolff, PhD, Steve Stellman, PhD, Maureen Hatch, PhD, Geoff Kabat, PhD, Bruce Levin, PhD, H. Leon Bradlow, PhD, David Camann, BS, Martin Trent, BS, Ruby Senie, PhD, Carla Maffeo, PhD, Pat Montalvan, Gertrud Berkowitz, PhD, Margaret Kemeny, MD, Mark Citron, MD, Freya Schnabel, MD, Allen Schuss, MD, Steven Hajdu, MD, and Vincent Vinceguerra, MD. The PAH–DNA adducts (detected by Lianwen Wang, Qiao Wang) and genotyping were completed in Dr Santella's lab. Funding in part was by grants U01 CA/ES66572, P30ES09089, P30ES10126, and K07CA90685 from the National Cancer Institute and the National Institute of Environmental Health Sciences, an award from the Breast Cancer Research Foundation and gifts from private citizens. Conflict of Interest Statement : None declared.

References

1.
Margison,G.P., Povey,A.C., Kaina,B. and Santibanez Koref,M.F. (
2003
) Variability and regulation of O 6 -alkylguanine-DNA alkyltransferase.
Carcinogenesis
  ,
24
,
625
–635.
2.
Mohrenweiser,H.W., Wilson,D.M.,III and Jones,I.M. (
2003
) Challenges and complexities in estimating both the functional impact and the disease risk associated with the extensive genetic variation in human DNA repair genes.
Mutat. Res.
  ,
526
,
93
–125.
3.
Bartsch,H. and Montesano,R. (
1984
) Relevance of nitrosamines to human cancer.
Carcinogenesis
  ,
5
,
1381
–1393.
4.
Margison,G.P. and Santibanez-Koref,M.F. (
2002
) O 6 -alkylguanine-DNA alkyltransferase: role in carcinogenesis and chemotherapy.
Bioessays
  ,
24
,
255
–266.
5.
Sedgwick,B. and Lindahl,T. (
2002
) Recent progress on the Ada response for inducible repair of DNA alkylation damage.
Oncogene
  ,
21
,
8886
–8894.
6.
Egyhazi,S., Ma,S., Smoczynski,K., Hansson,J., Platz,A. and Ringborg,U. (
2002
) Novel O 6 -methylguanine-DNA methyltransferase SNPs: a frequency comparison of patients with familial melanoma and healthy individuals in Sweden.
Hum. Mutat.
  ,
20
,
408
–409.
7.
Krzesniak,M., Butkiewicz,D., Samojedny,A., Chorzy,M. and Rusin,M. (
2004
) Polymorphisms in TDG and MGMT genes—epidemiological and functional study in lung cancer patients from Poland.
Ann. Hum. Genet.
  ,
68
,
300
–312.
8.
Kaur,T.B., Travaline,J.M., Gaughan,J.P., Richie,J.P.J., Stellman,S.D. and Lazarus,P. (
2000
) Role of polymorphisms in codons 143 and 160 of the O 6 -alkylguanine DNA alkyltransferase gene in lung cancer risk.
Cancer Epidemiol. Biomarkers Prev.
  ,
9
,
339
–342.
9.
Cohet,C., Borel,S., Nyberg,F., Mukeria,A., Bruske-Hohlfeld,I., Constantinescu,V., Benhamou,S., Brennan,P., Hall,J. and Boffetta,P. (
2004
) Exon 5 polymorphisms in the O 6 -alkylguanine DNA alkyltransferase gene and lung cancer risk in non-smokers exposed to second-hand smoke.
Cancer Epidemiol. Biomarkers Prev.
  ,
13
,
320
–323.
10.
Ford,B.N., Ruttan,C.C., Kyle,V.L., Brackley,M.E. and Glickman,B.W. (
2000
) Identification of single nucleotide polymorphisms in human DNA repair genes.
Carcinogenesis
  ,
21
,
1977
–1981.
11.
Inoue,R., Isono,M., Abe,M., Abe,T. and Kobayashi,H. (
2003
) A genotype of the polymorphic DNA repair gene MGMT is associated with de novo glioblastoma.
Neurol. Res.
  ,
25
,
875
–879.
12.
Ma,S., Egyhazi,S., Ueno,T., Lindholm,C., Kreklau,E.L., Stierner,U., Ringborg,U. and Hansson,J. (
2003
) O 6 -methylguanine-DNA-methyltransferase expression and gene polymorphisms in relation to chemotherapeutic response in metastatic melanoma.
Br. J. Cancer
  ,
89
,
1517
–1523.
13.
Yang,M., Coles,B.F., Caporaso,N.E., Choi,Y., Lang,N.P. and Kadlubar,F.F. (
2004
) Lack of association between Caucasian lung cancer risk and O 6 -methylguanine-DNA methyltransferase-codon 178 genetic polymorphism.
Lung Cancer
  ,
44
,
281
–286.
14.
Inoue,R., Abe,M., Nakabeppu,Y., Sekiguchi,M., Mori,T. and Suzuki,T. (
2000
) Characterization of human polymorphic DNA repair methyltransferase.
Pharmacogenetics
  ,
10
,
59
–66.
15.
Chueh,L.L., Nakamura,T., Nakatsu,Y., Sakumi,K., Hayakawa,H. and Sekiguchi,M. (
1992
) Specific amino acid sequences required for O 6 -methylguanine-DNA methyltransferase activity: analyses of three residues at or near the methyl acceptor site.
Carcinogenesis
  ,
13
,
837
–843.
16.
Rudiger,H.W., Schwartz,U., Serrand,E., Stief,M., Krause,T., Nowak,D., Doerjer,G. and Lehnert,G. (
1989
) Reduced O 6 -methylguanine repair in fibroblast cultures from patients with lung cancer.
Cancer Res.
  ,
49
,
5623
–5626.
17.
Wogan,G.N., Hecht,S.S., Felton,J.S., Conney,A.H. and Loeb,L.A. (
2004
) Environmental and chemical carcinogenesis.
Semin. Cancer Biol.
  ,
14
,
473
–486.
18.
Drablos,F., Feyzi,E., Aas,P.A., Vaagbo,C.B., Kavli,B., Bratlie,M.S., Pena-Diaz,J., Otterlei,M., Slupphaug,G. and Krokan,H.E. (
2004
) Alkylation damage in DNA and RNA—repair mechanisms and medical significance.
DNA Repair (Amst)
  ,
3
,
1389
–1407.
19.
Goldman,R. and Shields,P.G. (
2003
) Food mutagens.
J Nutr.
  ,
133
(suppl. 3),
965S
–973S.
20.
Mijal,R.S., Thomson,N.M., Fleischer,N.L., Pauly,G.T., Moschel,R.C., Kanugula,S., Fang,Q., Pegg,A.E. and Peterson,L.A. (
2004
) The repair of the tobacco specific nitrosamine derived adduct O 6 -[4-Oxo-4-(3-pyridyl)butyl]guanine by O 6 -alkylguanine-DNA alkyltransferase variants.
Chem. Res Toxicol.
  ,
17
,
424
–434.
21.
Chaney,S.G. and Sancar,A. (
1996
) DNA repair: enzymatic mechanisms and relevance to drug response.
J. Natl Cancer Inst.
  ,
88
,
1346
–1360.
22.
Pegg,A.E. (
2000
) Repair of O(6)-alkylguanine by alkyltransferases.
Mutat. Res.
  ,
462
,
83
–100.
23.
Coss,A., Cantor,K.P., Reif,J.S., Lynch,C.F. and Ward,M.H. (
2004
) Pancreatic cancer and drinking water and dietary sources of nitrate and nitrite.
Am. J. Epidemiol.
  ,
159
,
693
–701.
24.
Ramirez-Victoria,P., Guzman-Rincon,J., Espinosa-Aguirre,J.J. and Murillo-Romero,S. (
2001
) Antimutagenic effect of one variety of green pepper (Capsicum spp.) and its possible interference with the nitrosation process.
Mutat. Res
  . ,
496
,
39
–45.
25.
Garland,W.A., Kuenzing,W., Rubio,F., Kornychuk,H., Norkus,E.P. and Conney,A.H. (
1986
) Urinary excretion of nitrosodimethylamine and nitrosoproline in humans: Interindividual differences and the effect of aministered ascorbic acid and tocopherol.
Cancer Res.
  ,
46
,
5392
–5400.
26.
Gandini,S., Merzenich,H., Robertson,C. and Boyle,P. (
2000
) Meta-analysis of studies on breast cancer risk and diet: the role of fruit and vegetable consumption and the intake of associated micronutrients.
Eur. J. Cancer
  ,
36
,
636
–646.
27.
van Gils,C.H., Peeters,P.H., Bueno-de-Mesquita,H.B. et al . (
2005
) Consumption of vegetables and fruits and risk of breast cancer.
JAMA
  ,
293
,
183
–193.
28.
Smith-Warner,S.A., Spiegelman,D., Yaun,S.S. et al . (
2001
) Intake of fruits and vegetables and risk of breast cancer: a pooled analysis of cohort studies.
JAMA
  ,
285
,
769
–776.
29.
Riboli,E. and Norat,T. (
2003
) Epidemiologic evidence of the protective effect of fruit and vegetables on cancer risk.
Am. J. Clin. Nutr.
  ,
78
(suppl. 3),
559S
–569S.
30.
Gaudet,M.M., Britton,J.A., Kabat,G.C., Steck-Scott,S., Eng,S.M., Teitelbaum,S.L., Terry,M.B., Neugut,A.I. and Gammon,M.D. (
2004
) Fruits, vegetables, and micronutrients in relation to breast cancer modified by menopause and hormone receptor status.
Cancer Epidemiol. Biomarkers Prev.
  ,
13
,
1485
–1494.
31.
Hamajima,N., Hirose,K., Tajima,K. et al . (
2002
) Alcohol, tobacco and breast cancer—collaborative reanalysis of individual data from 53 epidemiological studies, including 58,515 women with breast cancer and 95,067 women without the disease.
Br. J. Cancer
  ,
87
,
1234
–1245.
32.
Gammon,M.D., Eng,S.M., Teitelbaum,S.L., Britton,J.A., Kabat,G.C., Hatch,M., Paykin,A.B., Neugut,A.I. and Santella,R.M. (
2004
) Environmental tobacco smoke and breast cancer incidence.
Environ. Res.
  ,
96
,
176
–185.
33.
Terry,P.D. and Rohan,T.E. (
2002
) Cigarette smoking and the risk of breast cancer in women: a review of the literature.
Cancer Epidemiol. Biomarkers Prev.
  ,
11
,
953
–971.
34.
Band,P.R., Le,N.D., Fang,R. and Deschamps,M. (
2002
) Carcinogenic and endocrine disrupting effects of cigarette smoke and risk of breast cancer.
Lancet
  ,
360
,
1044
–1049.
35.
Russo,J., Tay,L.K. and Russo,I.H. (
1982
) Differentiation of the mammary gland and susceptibility to carcinogenesis.
Breast Cancer Res. Treat.
  ,
2
,
5
–73.
36.
Coss,A., Cantor,K.P., Reif,J.S., Lynch,C.F. and Ward,M.H. (
2004
) Pancreatic cancer and drinking water and dietary sources of nitrate and nitrite.
Am. J. Epidemiol.
  ,
159
,
693
–701.
37.
Ramirez-Victoria,P., Guzman-Rincon,J., Espinosa-Aguirre,J.J. and Murillo-Romero,S. (
2001
) Antimutagenic effect of one variety of green pepper (Capsicum spp.) and its possible interference with the nitrosation process.
Mutat. Res.
  ,
496
,
39
–45.
38.
Gammon,M.D., Santella,R.M., Neugut,A.I. et al . (
2002
) Environmental toxins and breast cancer on Long Island. I. Polycyclic aromatic hydrocarbon DNA adducts.
Cancer Epidemiol. Biomarkers Prev
  .,
11
,
677
–685.
39.
Gammon,M.D., Neugut,A.I., Santella,R.M. et al . (
2002
) The Long Island breast cancer study project: description of a multi-institutional collaboration to indentify environmental risk factors for breast cancer.
Breast Cancer Res. Treat.
  ,
74
,
235
–254.
40.
Chen,X., Levine,L. and Kwok,P.Y. (
1999
) Fluorescence polarization in homogeneous nucleic acid analysis.
Genome Res.
  ,
9
,
492
–498.
41.
Cox,D.G. and Canzian,F. (
2001
) Genotype transposer: automated genotype manipulation for linkage disequilibrium analysis.
Bioinformatics
  ,
17
,
738
–739.
42.
Rothman,K.J. and Greenland,S. (
1998
) Modern Epidemiology . Lippcott-Raven, New York.
43.
Hosmer,D.W. (
1989
) Applied Logistic Regression . John Wiley & Sons, New York.
44.
Gammon,M.D., Sagiv,S.K., Eng,S.M. et al . (
2005
) Polycylic aromatic hydrocarbon (PAH)–DNA adducts and breast cancer: a pooled analysis.
Arch. Envir. Health
  , in press.
45.
Tregouet,D.A., Escolano,S., Tiret,L., Mallet,A. and Golmard,J.L. (
2004
) A new algorithm for haplotype-based association analysis: the Stochastic-EM algorithm.
Ann. Hum. Genet.
  ,
68
,
165
–177.
46.
Stephens,M. and Donnelly,P. (
2003
) A comparison of bayesian methods for haplotype reconstruction from population genotype data.
Am. J. Hum. Genet.
  ,
73
,
1162
–1169.
47.
Stephens,M., Smith,N.J. and Donnelly,P. (
2001
) A new statistical method for haplotype reconstruction from population data.
Am. J. Hum. Genet.
  ,
68
,
978
–989.
48.
Khoury,M.J. and James,L.M. (
1993
) Population and familial relative risks of disease associated with environmental factors in the presence of gene–environment interaction.
Am. J. Epidemiol.
  ,
137
,
1241
–1250.
49.
Lundberg,M., Fredlund,P., Hallqvist,J. and Diderichsen,F. (
1996
) A SAS program calculating three measures of interaction with confidence intervals.
Epidemiology
  ,
7
,
655
–656.
50.
Esteller,M. and Herman,J.G. (
2004
) Generating mutations but providing chemosensitivity: the role of O 6 -methylguanine DNA methyltransferase in human cancer.
Oncogene
  ,
23
,
1
–8.
51.
Jurek,A.M., Greenland,S., Maldonado,G. and Church,T.R. (
2005
) Proper interpretation of non-differential misclassification effects: expectations vs observations.
Int. J. Epidemiol.
  ,
34
,
680
–687.
52.
Greenland,S. (
1980
) The effect of misclassification in the presence of covariates.
Am. J. Epidemiol.
  ,
112
,
564
–569.
53.
Bronstein,S.M., Skopek,T.R. and Swenberg,J.A. (
1992
) Efficient repair of O 6 -ethylguanine, but not O 4 -ethylthymine or O 2 -ethylthymine, is dependent upon O 6 -alkylguanine-DNA alkyltransferase and nucleotide excision repair activities in human cells.
Cancer Res.
  ,
52
,
2008
–2011.
54.
Liu,H., Xu-Welliver,M. and Pegg,A.E. (
2000
) The role of human O(6)-alkylguanine-DNA alkyltransferase in promoting 1,2-dibromoethane-induced genotoxicity in Escherichia coli .
Mutat. Res.
  ,
452
,
1
–10.
55.
Liu,L., Pegg,A.E., Williams,K.M. and Guengerich,F.P. (
2002
) Paradoxical enhancement of the toxicity of 1,2-dibromoethane by O 6 -alkylguanine-DNA alkyltransferase.
J. Biol. Chem.
  ,
277
,
37920
–37928.
56.
Teo,A.K., Oh,H.K., Ali,R.B. and Li,B.F. (
2001
) The modified human DNA repair enzyme O(6)-methylguanine-DNA methyltransferase is a negative regulator of estrogen receptor-mediated transcription upon alkylation DNA damage.
Mol. Cell. Biol.
  ,
21
,
7105
–7114.
57.
Boffetta,P., Nyberg,F., Mukeria,A. et al . (
2002
) O 6 -Alkylguanine-DNA-alkyltransferase activity in peripheral leukocytes, smoking and risk of lung cancer.
Cancer Lett.
  ,
180
,
33
–39.
58.
Bartsch,H. and Frank,N. (
1996
) Blocking the endogenous formation of N-nitroso compounds and related carcinogens.
IARC Sci. Publ.
  ,
1996
,
189
–201.
59.
Vineis,P., Bartsch,H., Caporaso,N. et al . (
1994
) Genetically based N-acetyltransferase metabolic polymorphism and low-level environmental exposure to carcinogens.
Nature
  ,
369
,
154
–156.

Author notes

1Department of Environmental Health Sciences and 2Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, NY 10032, USA, 3Department of Epidemiology, School of Public Health, University of North Carolina, Chapel Hill, NC 27599, USA and 4Department of Community and Preventive Medicine, Mt Sinai School of Medicine, New York, NY 10029, USA