Association between single nucleotide polymorphisms in the antioxidant genes CAT, GR and SOD1, erythrocyte enzyme activities, dietary and life style factors and breast cancer risk in a Danish, prospective cohort study

Exposure to estrogens and alcohol consumption - the two only well-established risk factors for breast cancer - are capable of causing oxidative stress, which has been linked to progression of breast cancer. Here, five functional polymorphisms in the antioxidant genes SOD1, CAT and GSR were investigated in 703 breast cancer case-control pairs in the Danish, prospective “Diet, Cancer and Health” cohort together with gene-environment interactions between the polymorphisms, enzyme activities and intake of fruits and vegetables, alcohol and smoking in relation to breast cancer risk. Our results showed that genetically determined variations in the antioxidant enzyme activities of SOD1, CAT and GSR were not associated with risk of breast cancer per se. However, intake of alcohol, fruit and vegetables, and smoking status interacted with some of the polymorphisms in relation to breast cancer risk. Four polymorphisms were strongly associated with enzyme activity, but there was no interaction between any of the studied environmental factors and the polymorphisms in relation to enzyme activity. Additionally, single measurement of enzyme activity at entry to the cohort was not associated with risk of breast cancer. Our results therefore suggest that the antioxidant enzyme activities studied here are not major determinants of breast cancer risk.


INTRODUCTION
Breast cancer (BC) is the most common type of cancer in Denmark among women and the second leading cause of cancer death after lung cancer [1]. The incidence has increased during the last decade [1], which warrants further knowledge about the underlying mechanisms. Environmental factors, in interplay with genetics, are known to be important in the etiology of BC [2][3][4][5]. Therefore, examining gene-environment interactions is an effective tool for identifying biological pathways in disease etiology.
The only well-established risk factors for BC are associated with prolonged exposure to estrogens [6,7] and alcohol consumption [8][9][10]. Both exposure to estrogens and alcohol consumption are capable of causing oxidative stress [reviewed in 11], which has also been linked to progression of cancer, including BC [12][13][14][15][16][17][18]. Oxidative stress is caused by an imbalance between the generation of reactive oxygen species (ROS), such as superoxide anions, hydrogen www.impactjournals.com/oncotarget/ Oncotarget, 2017, Vol. 8, (No. 38), pp: 62984-62997 Research Paper peroxide and hydroxyl radicals, and the antioxidant defense, including various enzymes and high and low molecular weight antioxidants (including glutathione (GSH)), leading to DNA damage, lipid peroxidation and protein oxidation. An efficient antioxidant defense will limit oxidative damage by ensuring a reducing environment [19,20]. The enzymatic antioxidant defense system is complex and may be described as an interacting network of several enzymes, including glutathione reductase (GR), superoxide dismutase (SOD) and catalase (CAT). Where GR catalyzes the reduction of GSH, an important antioxidant, superoxide anion radicals are dismutated by SOD to hydrogen peroxide. CAT converts hydrogen peroxide to water and oxygen, thereby preventing formation of the highly reactive hydroxyl radical. BC may partly be caused by oxidative damage combined with a failure of antioxidants to protect the breast tissue [14]. Smoking and alcohol consumption are sources of ROS whereas fruit and vegetables are rich in dietary antioxidants. Furthermore, several studies have indicated that an increased intake of fruits and vegetables may affect the enzymatic antioxidant defense system, leading to increased enzyme activities [21][22][23][24]. These dietary and life style choices may modulate risk of BC by altering the level of oxidative stress, such that alcohol consumption and smoking increase the risk, while intake of fruit and vegetables decreases the risk. However, where metaanalyses have not provided consisting evidence of an inverse association between intake of fruit and vegetables and BC risk on one side [25][26][27][28][29], and positive association between smoking and BC risk on the other side [reviewed in 30], interactions between smoking, alcohol consumption, fruit and vegetable intake, and polymorphisms in antioxidant genes in relation to BC risk, have been reported [30][31][32][33][34]. Thus, genetically determined variations in the activities of enzymes that protect against or generate oxidative stress could modify associations between dietary antioxidants and exogenous sources of ROS, and BC risk, and thus explain some of the inconsistencies in the results from these studies. Additionally, lower antioxidant enzyme activities in BC patients compared to healthy controls have been reported in several studies [35][36][37][38]. Prospective cohort studies may reveal whether these lowered activities are a cause or consequence of the disease in combination with examination of dietary and life style factors and their interaction with the enzymes in relation to risk of disease.
Functional polymorphisms may impact the enzymatic activities leading to altered antioxidant defense and thereby affect disease risk and health. We have previously found associations between functional polymorphisms in GPX1 and GPX4 genes, erythrocyte GPX activity, alcohol intake, hormone replacement therapy (HRT) use and BC risk in the Danish "Diet, Cancer and Health" (DCH) cohort [31,39]. In the present study, we wanted to extend these findings with polymorphisms in the antioxidant genes SOD1 (encoding SOD), CAT (encoding CAT) and GSR (encoding GR) in a study group of 975 postmenopausal women with BC and 975 matched controls nested within the DCH cohort; and search for gene-environment interactions between the polymorphisms, enzyme activities and intake of fruits and vegetables, alcohol and smoking in relation to BC risk.

RESULTS
Baseline characteristics of BC cases and matched controls are presented in Table 1. Findings regarding the included risk factors have been reported previously for the whole DCH cohort, for a subset of the present study, and for the present study group [31,[40][41][42][43][44][45][46][47]. Among controls, the genotype distribution of the polymorphisms did not deviate from Hardy-Weinberg equilibrium (results not shown). None of the polymorphisms were significantly associated with risk of BC per se (Table 2); however, a tendency towards increased risk among variant carriers of the SOD1/ rs202445 polymorphism, was observed (P=0.06). Only GSR/rs1002149 interacted weakly with alcohol intake in relation to risk of BC such that variant T-carriers had a 24% increased risk of BC per 10 g alcohol/day (95% CI: 1.09-1.42), whereas wild-type GG-carriers did not display alcohol-related modification of BC risk (P-value for interaction (P int )= 0.048) ( Table 3). We also found interaction between SOD1/rs202445 and intake of fruits and vegetables in relation to risk of BC (P int =0.016) ( Table 4). Carriage of the variant G-allele of SOD1/ rs202445 was associated with a 13% increased risk of BC (95% CI: 1.03-1.25), whereas wild-type AA-carriers showed no modified risk of BC per 100 g fruits and vegetables per day (Table 4). Smoking status interacted with the CAT/rs1001179 polymorphism in relation to risk of BC (P int =0.0015) ( Table 5). Variant A-carriers of the polymorphism, who did not smoke, had a 41% increased risk of BC (95% CI: 1.07-1.87) compared to non-smoking wild-type GG-carriers ( Table 5).
Activities of CAT, SOD and GR enzymes were measured in a subset of the cohort at the time of entry into the DCH cohort (n=434 cases and 434 controls). In linear analysis (Table 6), increment in enzyme activity of the three enzymes was associated with statistically non-significant decreased risk of BC (IRR CAT =0.89, 95% CI: 0.75-1.06; IRR SOD =0.94, 95% CI: 0.79-1.12; IRR GR =0.83, 95% CI: 0.68-1.02). However, in the tertile analyses of the enzyme activity (Table 6), only increasing CAT activity exhibited a dose-dependent decrease in BC risk. In Table 7, the correlation between intake of fruit and vegetables, alcohol and smoking, and enzyme activity was investigated. A negative correlation between alcohol intake and GR activity was found (P=0.043) ( Table 7). Smoking and intake of alcohol were positively correlated with SOD activity with marginally statistical significance (P alcohol =0.050 and P present smokers =0.054) ( Table 7). None of the investigated lifestyle factors were associated with CAT activity. In Table 8 , the relation between the studied genotypes and enzyme activities was examined. CAT/rs1001179, CAT/ rs12270780 and GSR/rs1002149 polymorphisms were strongly associated with enzyme activities (P≤0.0001 for all three single nucleotide polymorphisms (SNPs)) among both cases and controls, so that enzyme activities increased for each variant allele for CAT/ rs12270780 and GSR/rs1002149, whereas enzyme activities decreased for CAT/rs1001179 for each variant allele (Table 8). CAT/rs769217 polymorphism was also associated with enzyme activity (P all =0.0007) with enzyme activities decreasing for each variant allele, but only among controls (P controls =0.0007 and P cases =0.25) ( Table 8). These associations were not significantly modified by intake of alcohol, fruits and vegetables, and smoking status (P int between 0.26-0.92 (Supplemental Table 1 )). We have previously found association between HRT use and GPX activity [39], but in the present study, we did not find indication of association between CAT, SOD or GR enzyme activities and HRT use (Supplemental Table 2 ). Haplotype analysis of CAT polymorphisms revealed that haplotypes encompassing variant alleles of CAT/rs1001179 and CAT/rs769217, respectively, were functional, and that haplotypes encompassing the variant allele of CAT/rs1001179 had the strongest effect on CAT enzyme activity (Table 9).
We found an increased risk of BC among women carrying both the low activity A-allele of CAT/rs1001179 and high activity G-allele of SOD1/rs202445 (IRR: 1.43; 95% CI: 1.01-2.01) ( Table 10). The interaction was, however, not significant on a multiplicative scale (P int =0.27), but could be considered additive. We also investigated other combinations of polymorphisms in relation to risk of BC, but did not find anything further (results not shown). Values are expressed as medians (5th and 95th percentiles) or as fractions (%). IRR: incidence rate ratio. a The risk estimates for BC are mutually adjusted. b The risk is estimated per additional 2 kg/m 2 . c The risk is estimated for nulliparous versus one birth at age 35. d The risk is estimated per additional 5 years. e Among ever users of HRT. f The risk is estimated per additional 5-year of HRT use. g The risk for abstainers compared to the increment of 10 g alcohol per day. h Among drinkers, risk estimate is estimated for the increment of 10 g alcohol per day. i The risk is estimated per additional 100 g intake fruit and vegetables per day. www.impactjournals.com/oncotarget

DISCUSSION
In the present study, we found that genetically determined variations in the antioxidant enzyme activities of CAT, GSR and SOD1 were not associated with risk of BC per se in this relatively large cohort. However, intake of alcohol, fruit and vegetables, and smoking status interacted with some of the polymorphisms in relation to BC risk. All polymorphisms, except for SOD1/rs202445, were associated with enzyme activity, but there was no interaction between any of the studied environmental factors and polymorphisms in relation to enzyme activity. Haplotype analysis of CAT polymorphisms revealed that CAT/rs1001179 had the strongest effect on CAT enzyme activity. Additionally, single measurement of enzyme activity at entry to the cohort was not associated with risk of BC.
Both in vivo [48][49][50] and in vitro [49] studies have shown that alcohol is capable of decreasing GR activity and protein levels in rat breast [48], liver [49] and brain tissue [50]. The present study supports these findings, since intake of alcohol is associated with decreased overall GR enzyme activity and increased risk of BC among carriers of the high activity T-variant allele of the GSR/rs1002149 polymorphism. Carriage of the variant allele of this polymorphism is associated with statistically non-significant decreased risk of BC per se, indicating that alcohol may inhibit GR activity in the variant enzyme. However, alcohol did not interact with GSR/rs1002149 in relation to enzyme activity in the present study. It is well-known that alcohol consumption increases oxidative stress [51,52] and in vitro studies have shown that acetaldehyde (the product of the first step in alcohol oxidation) can interact with GSH spontaneously [53] (forming a GSH-acetaldehyde conjugate). Although IRR: incidence rate ratio. a Crude. b Adjusted for parous/nulliparous, number of births, age at first birth, length of school education (low, medium, high), duration of HRT use (years), BMI (kg/m 2 ) , previous benign breast disease and alcohol intake (10 g/day). c P-value for the adjusted risk estimates.  this may explain only a minor proportion of observed GSH reduction after acute ethanol intoxication, it is not clear whether such effects are seen at moderate alcohol intake and to what extent this affects GSH availability in breast tissue. A possible inhibitory action of ethanol and/ or acetaldehyde on enzymes involved in GSH synthesis has also been hypothesized [48,53]. GSH is critical in maintaining a reduced cellular environment and plays an important role in detoxification processes, including removal of peroxides. Potential GSH depletion combined with impaired GR enzyme activity may impact antioxidant defense, leading to increased carcinogenic potential in the breast tissue. We observed previously that alcohol intake also influences glutathione peroxidase by increasing the activity only in wild-type carriers of the GPX1 gene and that the GPX1 Pro198Leu polymorphism may increase BC risk [31]. GSH availability seems therefore potentially important for BC in accordance with the current findings regarding GR. We were not able to identify other studies examining the relationship between SOD1 polymorphism, BC risk and potential effect modification by intake of fruit and vegetables. Results from studies investigating the effect of increased intake of fruit and vegetables on SOD activity are inconsistent and the reported effects probably depend both on the presence of specific compounds in the investigated fruit and vegetables as well as the study population. Both negative [54], positive [55,56] and no [23] effects of intake of specific (or combinations of) fruits and vegetables on SOD enzyme activity have been reported. Our results indicate that fruit and vegetables consumption may affect BC risk differently in carriers of the variant allele of SOD1 polymorphism compared to wild-type carriers. Per 100 g increment in fruit and vegetable intake, risk of BC increased significantly among variant allele carriers only. The reason for this is not clear, but if fruit and vegetables increase SOD activity, this could potentially lead to increased formation of the highly reactive hydroxyl radicals, leading to oxidative damage and increased risk of BC if not metabolized to less reactive compounds. The increased risk of BC observed in carriers of both the high-activity SOD1 and low-activity CAT (rs1001179) allele could support this explanation. However, whether this is a plausible explanation awaits further studies in larger populations with higher intakes of fruit and vegetables.
Our finding regarding the association between CAT/ rs1001179 and decreased CAT enzyme activity is consistent with two other studies [33,57]. Compared to the other studied CAT polymorphisms (CAT/rs12270780 and CAT/ rs769217), CAT/rs1001179 is associated with the largest decline in enzyme activity among 743 postmenopausal women in the present study. This is also shown in the haplotype analysis (Table 9) where the variant allele of CAT/rs1001179 is only present on the AGC haplotype which is associated with the largest decrease in CAT We found interaction between CAT/rs1001179 and smoking status in relation to BC. Among nonsmokers, genetically determined lower CAT activity was associated with increased risk of BC, whereas among smokers, homozygous wild-type allele carriers were at non-significantly increased risk. Smoking was though Values are expressed as medians with 5th and 95th percentiles. 59 case-control pairs were excluded because one or both in the matched set had missing data on one or more of the potential confounding variables and/or enzyme activities. IRR, incidence rate ratio. a Crude estimates. b The risk estimates for BC are adjusted for parous/nulliparous, number of births, age at first birth, length of school education (low, medium, high), duration of HRT use (years), BMI (kg/m 2 ) , previous benign breast disease and alcohol intake (10 g/day). c Estimates are per increment in activity of 2 U/g Hb. d Estimates are per increment in activity of 100 U/g Hb. e Estimates are per increment in activity of 2 U/g Hb. f Median (5-95%) levels of enzyme activity (U/g Hb) for cases and controls. g Categories are based on tertiles among both cases and controls. not associated with CAT activity. Taken together, the results may suggest that CAT genetically determined high enzyme activity is protective of BC especially among non-smokers. While smoking had no statistically significant effect on CAT enzyme activity, smoking may change the cellular oxidant/antioxidant environment due to the presence of other compounds in the smoke, thereby masking the effect of the polymorphism we report among non-smokers. Whether the inconsistent effects observed among smokers and non-smokers are linked to   In this analysis, only individuals with missing values on genotypes and potentials confounders were excluded without regard to the match set. a P-value for trend. the possible anti-estrogenic effect of smoking reported in some studies [59,60] is not clear but needs further investigation.
We used a prospective, nested case-control design, which together with complete follow-up minimizes selection bias. In addition, information on life style factors was collected at enrolment, which reduces the risk for differential misclassification between cases and controls. The study is fairly large to study main effects, it is homogenous and alcohol consumption is relatively high in the DCH cohort [61] making it suitable for studying gene-environment interactions. However, Danes including the Danish women in the present study have a relatively low intake of fruit and vegetables (Table 1), and therefore, the present study may have limited statistical power to detect effects of fruits and vegetables consumption on antioxidant enzyme activities. The genes were carefully selected based on known and predicted functionality. Nevertheless, we are aware that our study may not be large enough for some of the gene environment interactions, although we have previously found gene-environment interactions for PPARG and alcohol intake and ADH and alcohol intake in relation to BC in the present cohort [40,41].
Neither enzyme activities measured once at study entry nor genetically determined differences in enzyme activities were associated with risk of BC. There are two possible interpretations; either that antioxidant activity is not involved in breast carcinogenesis. Alternatively, genetically determined variations are small compared to the influence of lifestyle factors. However, smoking, alcohol intake and fruit and vegetable intake only correlated with the measured enzyme activities to a very limited extent. Thus, the genetically determined changes in the enzyme activities were larger than the effects of the lifestyle factors studied here. Our results therefore suggest that the antioxidant enzyme activities included here are not major determinants of BC risk.

Subjects
The subjects were selected from the ongoing Danish DCH cohort study. The present study group has been described previously [40,41]. In short, 79,729 women aged 50-64 years, born in Denmark, living in the Copenhagen or Aarhus areas and having no previous cancers at the time of invitation were invited to participate in the study between December 1993 and May 1997. A total of 29,875 women accepted the invitation, corresponding to 37% of the invited women.
Study participants were followed up for diagnosis of BC from date of entry until either the date of diagnosis of cancer using record linkage to the Danish Cancer Registry until 2003 and afterwards by linkage to the Danish Pathology Databank, date of death, date of emigration, or April 27th, 2006, whichever came first. A total of 975 women were diagnosed with BC during the follow-up period. For each case, one matched control was selected [40,41]. The control was cancer-free at the exact age at diagnosis of the case and was further matched on age at inclusion into the cohort (half-year intervals), use of HRT (current/former/never) and on certainty of postmenopausal status (known/probably postmenopausal) upon inclusion into the cohort [40,41]. 208 individuals were excluded because of failed genotyping or no buffy coat was available. Additionally 82 individuals were excluded because of missing information about one or more of the potential confounding variables. 254 individuals were excluded because of a missing partner in the case-control pair, due to the above mentioned exclusions leaving 703 pairs for data analyses.

Data on covariates
From food frequency and lifestyle questionnaires, we obtained information on duration of school education, smoking status, HRT use, birth pattern (number of births and age at first birth), alcohol intake and diet. Body mass index (BMI) was computed based on measurements of height and weight at enrolment. Intake of alcohol was inferred from the food-frequency questionnaire and lifestyle questionnaire as described in details in [40,41]. Abstainers were defined as those who reported no intake of alcohol on the food-frequency questionnaire and no drinking occasions on the lifestyle questionnaire. Findings on all known risk factors have been reported previously for both the entire DCH cohort, for a subset of the present study, and for the present study group [40,[42][43][44][45][46].

Blood sampling and storage
From non-fasting participants a total of 30 ml blood was collected in citrated (2 x 10 ml) and plain (1 x 10 ml) Venojects from each participant. Plasma, serum, lymphocytes and erythrocytes were isolated and frozen at -20°C within 2 hours. At the end of the day of collection, all samples were stored in liquid nitrogen, at -150°C.

Erythrocyte enzyme activities
In a subset of the present cohort comprising 434 case-control pairs, SOD, CAT and GR activities were determined spectrophotometrically in erythrocyte lysates on a Cobas Mira analyzer (F. Hoffmann-La Roche Ltd, Basel, Switzerland) according to Wheeler et al. [69]. This subset has been described in details elsewhere [31]. Intraand interday CVs were 2,8% and 4,7% for GR, 6,8% and 9,7% for SOD and 4,6% and 9% for CAT.

Statistical analyses
Deviation from Hardy-Weinberg equilibrium was assessed using a Chi square test.
Due to the study design using incidence density sampling of controls with match on age at diagnosis, matched logistic regression analyses leads to estimation of BC incidence rate ratio (IRR) [70] corresponding to a Cox proportional hazard model for the full cohort with age as the time axis. The associations between genotype, enzyme activity and BC are presented as crude IRR as well as adjusted for potential BC risk factors. Twosided 95% confidence intervals (CI) for the IRR were calculated based on Wald's test of the Cox regression parameter, i.e. on the log RR scale. All models were adjusted for baseline values of risk factors for BC such as parity (entered as two variables; parous/nulliparous and number of births), age at first birth, length of school education (low, medium and high), duration of HRT use, alcohol intake and BMI. IRR was calculated separately for heterozygous and homozygous variant allele carriers. For all the SNPs, variant allele carriers were subsequently grouped for interaction analyses to improve the statistical power since no recessive effects were observed. For the different genetic variations, we investigated interactions with alcohol intake, smoking, and fruit and vegetables, using the likelihood ratio test.
A covariance model was used to determine the effects of fruits and vegetables, intake of alcohol and smoking on enzyme activity according to genotype. Enzyme activity was found to be normally distributed among the participants, and hence, the activity was entered untransformed as the dependent variable in the covariate models.
The procedures PHREG (Proportional Hazard Regression) and GLM (General Linear Model) in SAS, release 9.3; SAS Institute, (Cary, NC) were used for the matched logistic regression analyses and the covariance analyses, respectively.

Ethics statement
All participants gave verbal and written informed consent before enrolment to the study. The DCH study was approved by the regional Ethical Committees on www.impactjournals.com/oncotarget Human Studies in Copenhagen and Aarhus (jr.nr.(KF) 11-037/01), and by the Danish Data Protection Agency.

Author contributions
TIK participated in the design of the study, performed statistical analyses and drafted the manuscript. UV participated in the design of the study and participated in drafting the manuscript. LOD conceived the study. AT designed the case-control study and contributed with materials and analysis tools. GRH participated in the design of the study, carried out analyses of erythrocyte enzyme activities and participated in drafting the manuscript. All authors read and approved the final manuscript.