Multiple exposure to environmental factors and variations in CYP27B1 and the microRNA‐binding site of IL‐13 are associated with breast cancer risk

Abstract Purpose Several molecular epidemiology studies have evidenced an association of environmental factors and genetic polymorphisms with breast cancer (BC) risk. However, most have considered the functions of a single element rather than combined effects. Methods This case‐control study of 693 newly‐diagnosed BC cases and 714 cancer‐free controls evaluated the effect of multiple exposures to environmental factors and polymorphisms in CYP27B1 and IL‐13 on BC risk. Genotypes were detected using TaqMan genotyping. Combinations and interactions were analyzed using cross‐over analysis and multivariate logistic regression. Combining exposure models were assessed using classification and regression tree and multivariate logistic regression analyses. Results No significant independent association was observed for any polymorphism in CYP27B1 or IL‐13 with the risk of BC. However, significant combined effects were noted for ≥1 time/wk physical activity with rs10877012 (adjusted odds ratio [ORadj] = 0.21, 95% confidence interval [CI] = 0.11‐0.39) and rs4646536 (ORadj = 0.21, 95% CI = 0.11‐0.39) in CYP27B1. Furthermore, taking garlic ≥4 times/wk, ≥1 time/wk physical activity, and a psychological index score ≥33 all displayed significant combined effects with three IL‐13 polymorphisms. These relationships remained significant after Bonferroni correction for multiple comparisons. Combining exposure models indicated that compared with consuming garlic ≥4 times/wk, five models (model 5, ORadj = 2.94, 95% CI = 1.07‐8.06; model 6, ORadj = 10.26, 95% CI = 5.81‐18.10; model 7, ORadj = 5.05, 95% CI = 2.78‐9.17; model 8, ORadj = 3.95, 95% CI = 2.79‐5.58; and model 9, ORadj = 7.97, 95% CI = 5.26‐12.07) showed a significant increased risk. Conclusions Our findings suggest that personalized adjustments to diet and behavioral patterns may aid BC prevention in variant carriers of CYP27B1 and IL‐13.


| INTRODUCTION
Breast cancer (BC) is a major malignant disease that threatens women's health. The American Cancer Society estimated that the number of new cases of female BC would reach 2.1 million worldwide in 2018, and that 626 679 deaths from BC would occur. 1 Breast cancer was also reported to be the leading cause of cancer death among women in developing countries. 1 Approximately 268 600 new BC patients and 69 500 BC deaths in female occurred in China in 2015. 2 Breast cancer is known to be predominantly caused by environmental factors, such as diet (the consumption of meat, fruit, and vegetables), excessive drinking and smoking, and a lack of physical exercise, 3,4 especially in sporadic patients. Moreover, some aromatic hydrocarbons in tobacco and increased polycyclic aromatic hydrocarbons in barbecued meat may act directly as carcinogens or indirectly modify the individual risk of BC-associated genetic polymorphisms. 5,6 Most previous epidemiological studies have focused on the modified risk of a single environmental factor, so the challenge is to recognize and interpret combined exposure effects and translate from concept to utility to better delineate the causes and provide personalized prevention for BC. Recently, Wild et al 7 defined the exposome and drew attention to the need for more complete environmental exposure assessment in epidemiological studies. They indicated that information should be gathered to investigate not only the link between one type of exposure and a disease, but also to provide the measurement for combinations of multiple exposures and yield multiple exposure models associated with disease.
Vitamin D, present in the blood as 25-hydroxyvitamin D 3 (25(OH)D 3 ), was shown to be associated with a reduced risk of female BC. 8 CYP27B1 encodes a member of the cytochrome P450 family, which catalyzes the conversion of 25(OH)D 3 to 1-alpha,25-dihydroxyvitamin D3 (1α,25(OH) 2 D3); this then binds to the vitamin D receptor to promote bone resorption and regulate cell proliferation and differentiation. [9][10][11] Interleukin 13 (encoded by IL-13) is a critical immune regulatory cytokine secreted predominantly by activated T-helper 2 lymphocytes. 12 IL-13 plays an important role in lymphocyte proliferation and activation and can induce apoptosis. 13 Additionally, microRNAs (miRNAs), endogenous noncoding nucleotide sequences, function in gene regulation by binding to targeted mRNAs in 3′-untranslated regions (UTRs) to alter mRNA levels and protein expression. 14 Single nucleotide polymorphisms (SNPs) located within miRNA-binding sites could therefore influence cancer risk and recurrence. [15][16][17] Several studies have focused on the association between polymorphisms in CYP27B1 and IL-13 with cancer risk, 18,19 but little is known about polymorphisms in these genes and miRNA-binding sites with respect to BC risk. Therefore, the present study investigated the association of polymorphisms in CYP27B1 and the miRNA-binding site of IL-13 with the risk of BC, and assessed the combination and interaction between these and possible environmental exposure on BC. The effects of multiple exposures to environmental factors and polymorphisms on BC were also examined, together with a correlation between these polymorphisms and clinical characteristics.

| Study subjects
This case-control study included 693 newly-diagnosed BC cases and 714 cancer-free controls. Cases were pathologically diagnosed patients collected from the Third Affiliated Hospital of Harbin Medical University during October 2011 to May 2014. Any patient who had undergone radiotherapy or chemotherapy was excluded. Control subjects were selected from the Department of Orthopaedics and Ophthalmology at the First and Second Affiliated Hospitals of Harbin Medical University, HongQi community of Harbin and XiangFang District Centers for Disease Control and Prevention in the same period (187 community-based and 527 hospital-based). Any individual with a history of BC or malignancies and benign neoplasms was excluded. We also excluded postpartum or lactating women and pregnant women from both cases and controls. Approximately 5 mL of fasting peripheral venous blood was obtained from all participants, before surgery for patients or at enrollment for controls.
All procedures performed in studies involving human participants were in accordance with the ethical standards of the Human Research and Ethics Committee of Harbin Medical University and with the 1964 Helsinki declaration and its

| SNP selection and genotyping
We used the dbSMR database to conduct an extensive search on SNPs located in miRNA-binding sites within the 3′-UTRs of CYP27B1 and IL-13 (http://mirac le.igib.res.in/polyr eg/). The RNAhybrid tool (http://bibis erv.te chf ak.uni-biele feld . de/rnahy brid/submi ssion.html) w as app lied to assess th e Gibbs free energy of binding (DG, expressed in kJ/mol) both for wild-type and va riant alleles. DDG w as c alculated as the difference of DG between the two alleles (wild-type allele DG vs variant allele DG). The absolute values of DDG (|DDG|) were used to avoid positive and negative DDG values negating each other. Finally, the |DDG tot| values (sums of all |DDG|s for each SNP) were computed as parameters to predict the biological effects of the variations. The top SNPs in IL-13 were selected for genotyping. No polymorphisms in miRNA target sites in CYP27B1 were found, but two common polym orphisms (r s10877012 and rs4646 536) with a minor allele frequency of more than 5% were selected based on previous studies. 20 Detailed information on these five genetic polymorphisms is provided in Table S1.

| Statistical analysis
The Hardy-Weinberg equilibrium (HWE) was adopted to calculate genotype distributions in the controls. Chi-square test for categorical variables and two-sample t test for continuous variables were selected, respectively. The Akaike information Marriage status: single is the person who has not married. Married is the person who has married or cohabitated, including widowed (not remarried), separated (due to discord or long-distance), and divorced (not remarried). criterion (AIC) statistic was used to evaluate the goodness of model fitting. The model with the lowest AIC value was considered the best one and was analyzed in interaction and combined analyses. We used the Bonferroni-corrected P-value for multiple corrections. Haplotypes were evaluated by SHEsis software (http://analy sis.bio-x.cn/myAna lysis.php). Crude and adjusted odds ratios (ORs) and 95% confidence intervals (95% CIs) were calculated using univariate and multivariate logistic regression analyses. The exposure combination models were generated using classifi cation and regressi on tree (CART) with pruning using the statistical software package R Studio. Logistic regression analyses were adopted to assess the relationships of these models and BC risk. Other statistical analyses were carried out using sas software version 9.2 (SAS Institute, Cary, NC). All reported P values were two-sided, and P < 0.05 was considered significant. Psychological index was evaluated using psychosocial stress survey for groups (PSSG). Table 1 shows the demographic characteristics of 693 cases and 714 controls. The mean (±SD) age of BC patients and controls was 52.20 ± 9.61 and 53.20 ± 10.65 years, respectively (P = 0.038). We also observed significant differences between cases and controls regarding education level (P < 0.001), marriage status (P = 0.034), and family history of cancer (P < 0.001). Therefore, we chose these four characteristics as adjusted factors for multivariate analyses. No significant differences were found with regard to BMI (body mass index). and near the 5′ region of CYP27B1, and three (rs847, rs848, and rs1295685) were located in miRNA-binding sites of IL-13. Table S1 also lists the sum of |ΔΔG| values for each SNP. Table S2 shows that the genotype distributions of all five polymorphisms in the controls were in accordance with HWE (P > 0.05). The genotype distribution of the five polymorphisms and their adjusted ORs and 95% CIs for the risk of BC are shown in Table 2

IL-13 and BC risk
We calculated haplotypes and their frequencies among cases and controls for the five SNPs (Table S3). Haplotypes with frequencies <3% were excluded. G-C of CYP27B1 accounted for the largest proportion in cases (65.30%) and controls (62.40%), while A-G-C of IL-13 made up a relatively high proportion (67.00% in cases and 66.80% in controls). However, no noteworthy associations of haplotypes and BC risk were detected.

| Combined and interactive effects of polymorphisms and environmental factors on the risk of BC
We analyzed the associations between all environmental factors in our questionnaire and the risk of BC. Table   S4 shows the results from univariate and multivariate logistic regression for environmental factors. Among those factors, taking >200 g/wk cereal (OR adj = 0.33, 95% CI = 0.20-0.56), ≥4 times/wk garlic (OR adj = 0.23, 95% CI = 0.15-0.36), and ≥1 time/wk physical activity (OR adj = 0.43, 95% CI = 0.32-0.60) reduced BC risk. Higher consumption of poultry, milk, and soybean (all P < 0.05) was also associated with a decreased risk of BC. By contrast, higher consumption of pork, >3 times/wk overnight food (OR adj = 1.64, 95% CI = 1.08-2.51), and a higher psychological index score (≥33, OR adj = 2.39, 95% CI = 1.48-3.86) were significantly associated with an increased risk of BC.
Only those environmental factors identified as significant by multivariate analyses were examined for their combined and interactive effects with polymorphisms in relation to BC (Tables 3 and 4). In a recessive model of CYP27B1 rs10877012, we observed valuable combined effects for the TT genotype and physical activity (OR adj = 0.21, 95% CI = 0.11-0.39). This effect was also found to have statistical significance for the TT genotype in CYP27B1 rs4646536 (recessive model) and ≥1 time/wk physical activity F I G U R E 1 Classification and regression tree analysis for the risk of breast cancer in different exposure models. Several variables (including garlic; psychological index; cereal; milk; physical activity; rs847; soybean; overnight food) were filtered as nodes due to their significant associations with breast cancer risk. Complexity parameter of the tree is 0.01 (OR adj = 0.21, 95% CI = 0.11-0.39). This indicated that performing physical activity ≥1 time/wk could have more beneficial effects on reducing BC risk in carriers of homozygous variants in these two CYP27B1 polymorphisms. No significant interactive effect was observed between CYP27B1 polymorphisms and environmental factors on BC risk. Using a dominant model for IL-13 polymorphisms rs847, rs848, and rs1295685, taking garlic ≥4 times/wk (OR adj = 0.27, 95% CI = 0.17-0.42), physical activity ≥ 1 time/wk (OR adj = 0.44, 95% CI = 0.31-0.63), and a psychological index score ≥33 (OR adj = 2.48, 95% CI = 1.64-3.74) all displayed significant combined effects. This indicated that garlic intake and physical activity played protective effects against BC in the carriers of variants in these three IL-13 SNPs. Additionally, compared with individuals with a psychological index score <33, those with a score ≥33 had more than twofold risk of BC. However, variants in the three IL-13 SNPs were associated with a reduced increase in BC risk in individuals with a psychological index score ≥33. No significant interactive effect was found between IL-13 polymorphisms and environmental factors regarding BC risk.

BC risk
The results of CART suggested that eight factors, including psychological index, physical activity, rs847, and the consumption of garlic, cereal, milk, soybean, and overnight food, were significantly associated with a modified risk of BC. Thus, we calculated the adjusted ORs and 95% T A B L E 5 Multiple exposure models analyses on the risk of breast cancer CIs for the nine combining exposure models. Including all these factors into multivariate logistic regression analysis, significant results were observed for five types of combining exposure models. As shown in Figure 1 and Table 5

| Association between polymorphisms and clinical characteristics
Next, we calculated the associations of polymorphisms in CYP27B1 and IL-13 with clinical characteristics in BC patients (Table S5-1 and Table S5-2). Tumor location, maximum diameter, molecular classification, ER status, PR status, HER2 status, TNM stage, and CA15-3 were analyzed for this calculation and TNM stage was classified according to the eighth edition of American Joint Committee on Cancer TNM staging system. 21 We obtained significant correlations between IL-13 rs847 and TNM stage (P = 0.041). TNM stage was also significantly associated with rs848 (P = 0.018) and rs1295685 (P = 0.004) in IL-13. No significant differences were discovered with regard to tumor location, maximum diameter, molecular classification, ER status, PR status, HER2 status, or CA15-3.

| DISCUSSION
In this study, we analyzed multiple exposures of environmental factors and polymorphisms in CYP27B1 and IL-13 with the risk of BC. Noteworthy findings are the significant combined effects between environmental factors (such as garlic intake, physical activity, and psychological index) and polymorphisms in CYP27B1 (rs10877012 and rs4646536) and IL-13 (rs847, rs848, and rs1295685) on BC. Five combining exposure models were suggested to have a significant increased risk of BC. Several environmental factors were observed to have significant associations with BC risk, including physical activity, psychological index, and the consumption of cereal, garlic, poultry, pork, milk, soybean, and overnight food. This could be explained by a number of reasons. For instance, plant lignans in cereal were previously shown to stimulate the production of enterolactone and butyric acid, which has antitumor activity. 22 Mathiasen et al 23 suggested that calcium in milk regulated apoptosis triggered by vitamin D compounds in BC cells, while a cell line study indicated that soybean reduced the growth of BC lines MCF-7 and MDA-MB-231 by silencing JMJD5. 24 27-hydroxycholesterol as the metabolite of cholesterol contained in pork was found to stimulate the xenograft growth of MCF-7 cells in mice and increase the risk of ER + BC. 25 Serum cholesterol is also closely related to biosynthesis, which increases the risk of BC when elevated. 26 Moreover, improper storage of overnight food can lead to the formation of nitrites, which interact with folic acid to increase the risk of BC. 27 Major organosulfur compounds found in garlic oil, such as diallyl disulfide and diallyl trisulfide, were proposed to suppress BC by halting DNA formation, inhibiting the production of reactive oxygen species, regulating cell cycle arrest, and inducing apoptosis. 28,29 Garlic was also shown to play a critical role in maintaining homeostasis of the immune system. 30 Hodge et al 31 confirmed that fresh garlic extract played a role in suppressing Th1 and inflammatory cytokine production, as well as upregulating the production of IL-10. Moreover, aged garlic extract and its product N-α-(1-deoxy-D-fructos-1-yl)-L-arginine alleviated neuroinflammatory responses in BV-2 microglia by regulating gene expression. 32 Additionally, animal work indicated that protein compositions extracted from fresh garlic bulbs restrained tumor growth by elevating CD8 + T cell infiltration into the BC site. 33 In keeping with these findings, our study indicated that garlic is an independent protective factor of BC, with an intake of ≥4 times/wk significantly decreasing the risk of BC (OR adj = 0.23, 95% CI = 0.15-0.36). Moreover, garlic also decreased BC risk when combined with all three IL-13 SNPs, suggesting that individuals with IL-13 variants could benefit from further protection against BC by consuming garlic.
Physical activity has been shown to decrease the production of proinflammatory cytokines in adipose tissue by reducing obesity. 34 Steensberg et al 35 reported that increasing exercise contributed to the upregulation of IL-6, which enhanced the level of antiinflammatory cytokines IL-1α and IL-10. Furthermore, IL-6 inhibited the expression of tumor necrosis factor-α. Mammographic density is considered a risk element for BC, being associated with tumor size, lymph node status, and lymphatic or vascular invasion, 36 while participating in physical activity was found to modify the mammographic density by reducing the level of serum estrogens. 37,38 Multivariate analysis in the present study showed that physical activity ≥1 time/wk was associated with a reduced risk of BC (OR adj = 0.43, 95% CI = 0.32-0.60), which is in accordance with previous work. 39,40 Moreover, the results of combined effects for physical activity and SNPs in CYP27B1 and IL-13 showed that genetic variants and taking more exercise collectively decreased susceptibility to BC. Thus, we recommend that carriers of these genetic variants participate in physical exercise to minimize their BC risk.
Analysis of the psychological index and BC risk has hinted at a positive correlation. 41 Psychological stress increases the level of glucocorticoids, and reduces the ability to repair DNA damage by increasing the production of reactive oxygen species/reactive nitrogen species in BC. 42,43 Irwin et al reported that psychological changes related to life events reduced the activity of NK cells. 44 These cells not only mediate the growth and metastasis of BC cells but also play a critical role in antiviral immunity as innate immune effector lymphocytes. 45,46 Similarly, Levy et al 47 showed that the levels of NK cells were decreased in BC patients with psychological stress. Our population-based case-control study observed that a higher psychological index score (≥33) increased the risk of BC, while significant combined effects were detected between the psychological index and the three IL-13 SNPs. Together, these results suggest that maintaining a positive mental state is important for individuals with variant genotypes (AG + GG carriers in rs847, GG + GT carriers in rs848, and CC + CT carriers in rs1295685) to help prevent BC.
Modification of the carcinogenic process occurs both through endogenous biological changes and exogenous exposures, while genetic variation may change an individual's vulnerability to subsequent exposures. Therefore, all exposures should be considered as a whole rather than as individual effects from independent factors. For this reason, we introduced CART and combined it with logistic regression analysis to assess the risk of several possible exposure combination patterns. The classification and regression tree is a recursive method for dividing data sets into two disjointed parts by minimizing the heterogeneity within each outcome. 48 We analyzed exposure combination models in the risk of BC based on the result of the binary tree with strong generalization ability produced by CART. This provided risk assessments for different dietary and lifestyle combinations. For example, compared with model 1 (consuming garlic ≥ 4 times/wk), individuals with exposure model 6 showed the highest risk of BC. Thus, we recommend that women should avoid exposure patterns associated with high risk.
Research into the link between clinical parameters and the important immunomodulatory cytokine IL-13 is limited. Yang et al 49 reported that Th17 cells, effector CD4(+) T cells involved in inflammation, were related to a low TNM stage so may play an anticancer role in BC. We observed significant associations between TNM stage and the three IL-13 polymorphisms (all P values < 0.05), suggesting that they could be potential biomarkers of TNM stage in BC.
The present study has a number of limitations which should be taken into account when drawing any conclusions. First, recall bias may have been inevitable when collecting information about environmental factors, although we tried to minimize this. Second, the method of counting food and lifestyle frequencies may have limited the food intake or the ability to perform precise calculations. Third, the sample size may have been insufficient for the analysis of some multiple exposure models. Finally, the results of this retrospective study are limited to Chinese women, so further prospective evidence is needed to validate them in a larger population.

| CONCLUSION
In summary, we report significant combined effects of polymorphisms in CYP27B1 and IL-13 with environmental factors in the risk of BC although these polymorphisms had no significant impact alone. CART analysis suggested important changes that could be made to dietary and behavioral patterns to reduce the risk of BC development.