Vitamin D-Related Genes, Blood Vitamin D Levels and Colorectal Cancer Risk in Western European Populations

Higher circulating 25-hydroxyvitamin D levels (25(OH)D) have been found to be associated with lower risk for colorectal cancer (CRC) in prospective studies. Whether this association is modified by genetic variation in genes related to vitamin D metabolism and action has not been well studied in humans. We investigated 1307 functional and tagging single-nucleotide polymorphisms (SNPs; individually, and by gene/pathway) in 86 vitamin D-related genes in 1420 incident CRC cases matched to controls from the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort. We also evaluated the association between these SNPs and circulating 25(OH)D in a subset of controls. We confirmed previously reported CRC risk associations between SNPs in the VDR, GC, and CYP27B1 genes. We also identified additional associations with 25(OH)D, as well as CRC risk, and several potentially novel SNPs in genes related to vitamin D transport and action (LRP2, CUBN, NCOA7, and HDAC9). However, none of these SNPs were statistically significant after Benjamini–Hochberg (BH) multiple testing correction. When assessed by a priori defined functional pathways, tumor growth factor β (TGFβ) signaling was associated with CRC risk (P ≤ 0.001), with most statistically significant genes being SMAD7 (PBH = 0.008) and SMAD3 (PBH = 0.008), and 18 SNPs in the vitamin D receptor (VDR) binding sites (P = 0.036). The 25(OH)D-gene pathway analysis suggested that genetic variants in the genes related to VDR complex formation and transcriptional activity are associated with CRC depending on 25(OH)D levels (interaction P = 0.041). Additional studies in large populations and consortia, especially with measured circulating 25(OH)D, are needed to confirm our findings.


Introduction
Colorectal cancer (CRC) is the second most common cancer in men and women combined, with approximately 1.4 million new cases diagnosed in 2012 worldwide [1]. There is compelling observational evidence that low circulating vitamin D concentrations are associated with increased risk of incident CRC [2,3]. However, other human evidence is less convincing. A few Mendelian randomization (MR) studies did not support an association between vitamin D genetic score and CRC risk, but the genetic contribution to 25(OH)D is relatively small (7.5% as estimated based on genome-wide association studies (GWAS) on common SNPs [4]), possibly explaining the null findings [5,6]. Also, the relatively few randomized clinical trials (RCTs) of vitamin D supplementation and colorectal neoplasms have not shown statistically significant effects, but sample size, duration and timing of supplementation, issues with compliance and choice of study population, and the limited range of vitamin D exposures assessed may have contributed to the null results [7][8][9]. Finally, the benefits from vitamin D supplementation for the prevention of colorectal neoplasms may vary according to genetic variation in the vitamin D-related genes (e.g., vitamin D receptor (VDR) [10]).
Anti-neoplastic effects of vitamin D on colorectal tissue are also supported by the fact that the normal colorectal epithelium expresses the vitamin D receptor (VDR) and vitamin D metabolizing enzymes (CYP27B1 and CYP24A1) and, therefore, can locally produce and degrade the active form of vitamin D, 1,25-dihydroxyvitamin D (1,25(OH) 2 D), from 25-hydroxyvitamin D (25(OH)D) [11][12][13]. In the colorectum, the active metabolite of vitamin D, 1,25(OH) 2 D, exerts its anti-neoplastic effects by genomic (mediated by the VDR) and non-genomic mechanisms [14], including the regulation of over 200 vitamin D-responsive genes and rapid activation of intracellular signaling pathways, resulting in modulation of the cell cycle, bile acid degradation, immune response, growth factor signaling, and anti-inflammation [15].
Observational and RCT data suggest a potential vitamin D-colorectal neoplasms risk association is modified by polymorphisms in the vitamin D receptor (VDR) [10,16,17] and the vitamin D-binding protein gene (GC) [18]; however, only a few single nucleotide polymorphisms (SNPs) and a limited number of related pathways were considered. Novel evidence highlights a wide array of VDR binding sites across the human genome [19], and multiple pathways related to vitamin D effects [20]. Thus, it is plausible that the vitamin D-CRC risk association may be modulated by variation in a broad array of genes related to vitamin D metabolism (e.g. absorption, endogenous synthesis, transport, activation, and deactivation) and action (including transcriptional activity/post-transcriptional effects). All of these genes are polymorphic, but no studies to date have comprehensively investigated their individual and collective associations with CRC risk or circulating vitamin D levels. In consideration of these points, we investigated whether variation in genes related to vitamin D metabolism and transcriptional activity is related to circulating blood vitamin D levels, and whether genetic variation at the SNP, pathway and gene level, alone and in combination with circulating vitamin D levels, is associated with CRC risk in a large Western European prospective cohort study.

Study Population
We used a case-control design nested within the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort, a large prospective study with over 520,000 men and women aged 35-70 years enrolled from 23 centers in 10 Western European countries (Denmark, France, Greece, Germany, Italy, the Netherlands, Norway, Spain, Sweden, and United Kingdom). The methods of the EPIC study have been described in detail elsewhere [21,22]. Individuals who were eligible for the study were selected from the general population of a specific geographical area, town, or province. Exceptions included the French sub-cohort, which is based on members of the health insurance system or state-school employees, and the Utrecht (Netherlands) sub-cohort, which is based on women who underwent screening for breast cancer. Between 1992 and 1998, standardized lifestyle and personal history information, anthropometrics, and blood samples were collected from most participants at recruitment. Diet over the previous 1 year was measured at baseline by validated country-specific dietary questionnaires developed to ensure high compliance and better measures of local dietary habits [21]. Blood samples were stored at the International Agency for Research on Cancer (Lyon, France; −196 • C, in liquid nitrogen) for all countries except Denmark (−150 • C, in nitrogen vapor) and Sweden (in −80 • C freezers). The EPIC study was approved by the Ethical Review Board of the International Agency for Research on Cancer (IARC) and the Institutional Review Board of each participating EPIC center. Written consent was obtained from all EPIC participants at enrolment into the study.

Cancer Incidence and Vital Status Follow-Up
Cancer incidence was determined through record linkages with regional cancer registries (Denmark/Italy/the Netherlands/Norway/Spain/Sweden/United Kingdom; complete up to December 2006) or via a combination of methods, including the use of health insurance records, contacts with cancer and pathology registries, and active follow-up through study subjects and their next-of-kin (France/Germany/Naples/Greece; complete up to June 2010).
Vital status follow-up (98.5% complete) was collected by record linkage with regional and/or national mortality registries in all countries except France, Germany, and Greece, where data are collected through an active follow-up. Censoring dates for complete follow-up were between June 2005 and June 2009 in Denmark, the Netherlands, Spain, the United Kingdom, Sweden, Norway, and Italy. In Germany, Greece, and France follow-up was based on a combination of methods, including health insurance records, cancer and pathology registries, and active follow-up through study subjects and their next-of-kin. In these centers, the end of follow-up was defined as the last known date of contact, or the date of death whichever came first. The last update of endpoint information occurred between December 2007 and December 2009.

Case Ascertainment and Selection
CRC cases were selected among participants who developed colon (C18.0-C18.7, according to the ICD-10), rectum (C19-C20), and overlapping/unspecified origin tumors (C18.8 and C18.9). Cancers of the anus were excluded. CRC is defined as the combination of the colon and rectal cancers.
A total of 1420 first-time previously cancer-free colorectal cancer cases (colon cancer = 900; rectal cancer = 520) were identified. Cases were not selected from Norway (blood samples only recently collected; few colorectal cancers diagnosed after blood donation) and the Malmö center of Sweden. The number of cases for gene-environment analyses was 1176 because of missing, previously collected 25(OH)D measurements [23] (France = 6, Italy = 49, Spain = 30, UK = 27, The Netherlands = 8, Greece = 18, Germany = 21, and Sweden = 16).

Control Selection
Controls were selected (1:1) by incidence density sampling from all cohort members alive and not having a reported cancer at the time of diagnosis of the cases and were matched by age (±6 months at recruitment), sex, study center, time of the day at blood collection, and fasting status at the time of blood collection (less than three hours, three to six hours, and more than six hours). Women were further matched by menopausal status (pre-/post-/peri-menopausal, and unknown) and for pre-menopausal women, phase of menstrual cycle at time of blood collection and usage of postmenopausal hormone therapy at time of blood collection (yes/no, regardless of menopausal status). The additional matching criteria for women were required for other studies that were being carried out using the same matched case-control sets. One control sample failed the genotyping and was not included in the analysis, resulting in a total of 1419 controls. The number of controls for analyses involving 25(OH)D was 764 because of missing or unobtainable, previously collected 25(OH)D measurements [23] (France = 18, Italy = 69, Spain = 48, UK = 62, The Netherlands = 41, Greece = 23, Germany = 60, Sweden = 49, and Denmark = 328).

Blood 25-(OH)-Vitamin D Assessment
We previously measured blood concentrations of 25(OH)D using a commercially available enzyme immunoassay kit (OCTEIA 25-(OH)D Kit, Immuno Diagnostic Systems, Boldon, UK) at the Laboratory for Health Protection Research, National Institute for Public Health and the Environment, the Netherlands [23]. The kit is specific for 100% of 25-(OH)-vitamin D 3 form and 75% of 25-(OH)-vitamin D 2 form. The inter-assay coefficient of variation as determined with two kit control samples was minimal (5.9% at the level of 20.3 nmol/L and 5.4% at the level of 77.4 nmol/L). No significant between-day drift, time shifts, or other trends were observed and the percentage of variance attributable to batch-to-batch differences was 4.5%. For all analyses, laboratory technicians were blinded to the case-control status of the samples.

SNP Selection, Genotyping, and Quality Control
Genomic DNA was extracted from whole blood samples using conventional methods. We used the custom GoldenGate Universal-32 3072-plex assay kit (Illumina, CA, USA) to genotype 1716 genetic variants within the genes known and proposed to be involved in (1) vitamin D metabolism (DHCR7, GC, CYP3A4, CYP2R1, CYP24A1, CUBN, and LRP2), (2) mineral homeostasis and endocrine regulations of 1,25(OH) 2 D synthesis (CASR, PTH, TRPV5, and TRPV6), (3) vitamin D genomic effects (VDR, RXRA, RXRB, and RXRG), (4) formation of the VDR complex (co-activators and co-regulators ACTL6A, ARID1A, BAZ1B, CARM1, CHAF1A, CREBBP, EP300, HDAC9, MED1, NCOA1, NCOA2, NCOA3, NCOA7, NCOR1, NCOR2, PCAF/KAT2B, PRMT1, SMARCA2, SMARCA4, SMARCC1, SMARCD1, SMARCE1, SNW1, SUPT16H, TOP2B, and TSC2), and (5) vitamin D post-transcriptional response (tumor growth factor β (TGFβ)-signaling, inflammation, oxidative stress, insulin growth factor (IGF) signaling, cell cycle, and VDR binding sites; please see Supplementary Table S1 for a complete list of genes and SNPs). The custom GoldenGate assay was designed using the Illumina online Assay Design Tool in May 2012. SNP genotype dataset for CEU population (Utah residents with Northern and Western European ancestry; HapMap Data Rel 28 Phase II + III, August 10, on NCBI B36 assembly, dbSNP b126) were loaded in the Haploview program (Broad Institute, MIT and Harvard, Cambridge, MA, USA) and SNPs with minor allele frequencies (MAFs) greater than 5% and the r 2 linkage disequilibrium (LD) statistic of 0.8 were selected as tagging SNPs (tagSNPs). Additionally, we searched published literature for previously reported functional and regulatory SNPs in the genes of interest and included them in genotyping irrespective of MAFs or r 2 with other SNPs. Genotyping was performed by the Genetics Laboratory at Imperial College London. After excluding 409 SNPs [247 (14.4%) that failed genotyping, 54 (3.1%) that failed to satisfy the Hardy-Weinberg criterion (Supplementary Table S1), 98 (5.7%) missing in more than 20% of genotyped samples, and 10 (0.6%) that were monomorphic], a total of 1307 SNPs were included in the analysis. All genotyping underwent standard quality control including concordance checks for blinded duplicates and examination of sample and SNP call rates. The lowest reproducibility frequency across 62 replicate samples was 0.98. The call rate was 95% for all samples and 95% for all SNPs.

Statistical Analysis
The season adjustment of 25(OH)D was carried out by the week of blood draw using the sine curve method [24]. The associations between season-adjusted 25(OH)D concentrations and genetic variants (coded as 0, 1, 2 corresponding to the number of minor alleles) were assessed among controls using linear regression models adjusted for age, sex, and center. Further adjustment for BMI, smoking status, and physical activity did not change the results substantially. We used unconditional logistic regression analysis to assess the association of individual SNPs with CRC risk, adjusting for age (continuous), sex, and study center. Results were similar when we used conditional logistic regression on 1331 complete case-matched sets. We assumed a log-additive genetic model, but also tested dominant and recessive models as the underlying genetic model for these SNPs is unknown. Further adjustment for body mass index (BMI; continuous), smoking status (never, former, current smokers, missing), physical activity (active, moderately active, moderately inactive, and inactive), alcohol intake (continuous), hormone therapy, and menopausal status did not substantially change the results, and thus these variables were not included in the final statistical model. Subgroup analyses were conducted by sex and tumor location (colon vs. rectum).
To examine the associations between genes (a combination of SNPs) and genetic pathways (a combination of genes) and CRC risk, we used the Adaptive Rank Truncated Product (ARTP) method [25] as implemented in the first step (no interaction) of the R package PIGE (http://cran. r-project.org/web/packages/PIGE/index.html). This method can combine associations of SNPs in each gene (or from the genes in a pathway) to provide a P-value at the gene or pathway level, respectively. Genetic markers in high LD (r 2 ≥ 0.8) were excluded using the AdaJoint R package (https://cran.r-project.org/web/packages/ARTP2). To investigate the multiplicative interaction between the genes and genetic pathways with 25(OH)D on CRC risk, we used the modified ARTP method as implemented in the R package PIGE. The P-values at the SNP and the gene levels were corrected for multiple testing for the number of SNPs and for the number of genes, respectively, using the false discovery rate (Benjamini-Hochberg or BH) method [26]. Furthermore, we used traditional methods to assess potential interactions between SNPs and 25(OH)D stratifying by categories of 25(OH)D concentrations and assuming a log-additive model for genetic markers. Also, we assessed the association of 25(OH)D (per 24.96 nmoL = 10 ng/mL) with CRC risk by genotype.
All statistical tests were two-sided with P-values < 0.05 considered statistically significant (SAS software, version 9.2; SAS Institute, Cary/NC; R, R Foundation for Statistical Computing, Vienna/Austria).

Baseline Characteristics of Cases and Controls
Selected baseline characteristics of the CRC cases and matched controls are shown in Table 1. The mean age at blood donation of cases and controls was 58 years. On average, CRC cases had 4 years between blood donation and the time of diagnosis. The dataset included 520 rectal cancer cases and 900 colon cancer cases.

SNPs in the Genes Related to Vitamin D Metabolism/Transcriptional Activity and 25(OH)D
Thirty-seven SNPs in the genes related to vitamin D metabolism, formation of the VDR complex, and VDR transcriptional activity were associated with season-adjusted 25(OH)D concentrations with unadjusted P ≤ 0.05 among controls (Supplementary Table S2). The top 10 SNPs are shown in Table 2. Of the 37, 17 SNPs were in the genes involved in vitamin D metabolism, and 20 SNPs in the genes involved in vitamin D transcriptional activity. None of these SNPs were statistically significantly associated with 25(OH)D after BH correction. The associations of all SNPs with 25(OH)D among controls only are shown in Supplementary Table S3A, and among cases and controls combined in Supplementary Table S3B.

SNPs in the Genes Related to Vitamin D Metabolism/Function and CRC Risk
We examined the associations between SNPs in the genes involved in vitamin D metabolism (genes = 9, SNPs = 274), mineral homeostasis and endocrine regulation of 1,25(OH) 2 D synthesis (genes = 5, SNPs = 58), vitamin D genomic effects including the VDR complex co-activators and co-regulators (genes = 30, SNPs = 538), and two SNPs in the intergenic regions previously associated with circulating 25(OH)D [27] and CRC risk (Supplementary Table S4). In Table 3, we show the top fifteen statistically significant SNPs associated with CRC risk defined by P unadjusted < 0.01. However, after BH correction, none of the associations remained statistically significant (all P BH > 0.2). The results did not differ by tumor location (Table 3 and Supplementary Table S4) or sex (Supplementary Table S5).

SNPs in the Vitamin D-Responsive Genes and CRC Risk
We also examined the associations between 434 SNPs in the genes responsive to vitamin D, including the genes in the TGFβ and IGF signaling pathways, inflammation, oxidative stress, cell cycle, and 19 SNPs located in the VDR binding sites as previously published [19] (Supplementary  Table S6). Twenty-five SNPs were significantly associated with CRC risk at P < 0.01. However, after BH correction, none of the associations (except for SMAD3 rs7180244; SMAD7 rs11874392, rs12953717 and rs4939827) remained statistically significant (Supplementary Table S7). Interestingly, three SNPs (rs3197999, rs3802842, rs762421) in previously identified VDR binding sites were associated with CRC risk. The results did not differ by tumor location (Supplementary Tables S6 and S7) or sex  (Supplementary Table S8).

Vitamin D Genes/Pathways and CRC Risk
At the pathway level, the VDR binding sites and TGFβ signaling pathway were statistically significantly associated with CRC risk (P < 0.04; Table 4). For colon cancer, in addition to the VDR binding sites (P = 0.008) and TGFβ signaling pathway (P = 0.0001), an association with cell cycle pathway was observed (P = 0.03). The TGFβ (P = 0.0001) and IGF (P = 0.007) signaling pathways, but not the VDR binding sites (P = 0.256), were statistically significantly associated with rectal cancer risk.

25. (OH)D-Gene and 25(OH)D Pathway Interactions and CRC Risk
At the pathway level, the VDR complex and its transcriptional co-regulators and co-activators demonstrated a potential interaction with 25(OH)D concentrations in the association with CRC risk (P = 0.04; Table 4). Within this pathway, the interaction P-values of <0.05 were observed for ARID1A, CARM1, CHAF1A, and SMARCA2, but none were statistically significant after BH correction. Similar associations were observed for colon cancer, but not rectal cancer (P for interaction for the VDR complex and its transcriptional co-regulators and co-activators were 0.105 and 0.727, respectively).
At the gene level, the interaction P-values of <0.05 were observed for CYP27B1 and GC (vitamin D metabolism) and IL10 (inflammation) for CRC and colon cancer. Also, the interaction between 25(OH)D and IGFBP2/IGFBP5 was statistically significant for colon cancer. For rectal cancer, the interaction P-values of <0.05 were observed for CYP24A1 (vitamin D metabolism) and BMP2 (TGF signaling). None of the gene-25(OH)D interactions were statistically significant after BH correction.
Next, we assessed the associations between 25(OH)D (per 24.96 nmol/L) with CRC risk, stratified by genotypes of SNPs in the genes that were identified in the step above as potentially modifying the association of 25(OH)D with CRC risk (CYP27B1, GC, ARID1A, CARM1, CHAF1A, SMARCA2, and IL10; Supplementary Table S9). Sixteen SNPs in these genes with P for interaction <0.05 are presented in Table 5. None were statistically significant after BH correction. Several SNPs had a very low number of minor allele homozygotes, with no effect estimates presented in the table.

Discussion
In this large European prospective case-control study nested within the EPIC cohort, we investigated whether genetic variation in the genes and pathways related to vitamin D metabolism and vitamin D genomic effects is associated with CRC risk, and whether these associations are modified by 25(OH)D concentrations.
We identified several genes related to vitamin D metabolism, the VDR complex, and VDR transcriptional activity associated with 25(OH)D concentrations, with an unadjusted P < 0.01 before BH correction among controls. We confirmed two genes related to vitamin D metabolism, CYP27B1 and GC, and one in the VDR, which were identified in previous GWAS studies [28][29][30][31][32]. We also identified other genes in our study including 1) two genes that encode the transcription-related factors HDAC9 and NCOA7, involved in vitamin D transcriptional activity and VDR complex formation, and 2) two genes that encode the vitamin D-related transporters LRP2 and CUBN [33,34]. LRP2, commonly known as megalin, is responsible for the endocytosis of the 25(OH)D vitamin D binding protein complex [35]. CUBN is an important co-receptor in the megalin-mediated endocytic pathway and patients without functioning CUBN were found to have abnormal 25(OH)D metabolism [33].
The genes associated with circulating 25(OH)D concentrations were also associated with CRC risk at unadjusted P < 0.01 before BH correction. HDAC9 is located in a region on chromosome 7p21 [36] in which chromosomal gains were observed in primary CRC [37]. Furthermore, HDAC9 has been observed via chromatin immunoprecipitation (ChIP) assay in human osteosarcoma tissues to suppress p53 transcription and, thereby, promote cell proliferation [38]. An association of CUBN with CRC was previously reported in a meta-analysis of six GWAS studies [39], while no studies have investigated a possible association of LRP2 with CRC. LRP2 is expressed in multiple epithelial cell lines, including colon [35,40] and is often co-expressed with CUBN [34]. Additionally, there are no previous GWAS regarding CRC and NCOA7, although Higginbotham et al., found statistically significant associations of NCOA7 gene variants with reduced breast cancer risk across three different study cohorts [41]. The NCOA7 SNPs identified in our study, however, differed from those identified by Higginbotham suggesting a possible novel CRC susceptibility locus.
Three VDR binding site SNPs were associated with CRC risk in our study population, but the associations were not statistically significant after BH correction. We a priori selected these VDR binding sites for genotyping based on the results of a previous study that used ChIP followed by DNA sequencing to identify 2776 VDR binding sites in lymphoblastoid cell lines treated with calcitriol, an active form of vitamin D [19]. The study found a statistically significant 4-fold increase in the enrichment of VDR binding sites located in genes associated with CRC, and a 3.5-fold increase in enrichment located in genes associated with Crohn's disease [19]. Our findings suggest that genetic variation in these VDR binding sites, upregulated in response to treatment with vitamin D and relevant to colorectal carcinogenesis and inflammatory bowel diseases, may be associated with CRC risk.
TGFβ has an important role in the regulation of cell proliferation, differentiation, migration and apoptosis [42], and may be modulated by vitamin D [43]. SMAD7 and SMAD3, in the TGFβ signaling pathway, were statistically significantly associated with CRC risk after BH correction for multiple testing. SMAD7 SNPs were previously identified to be associated with CRC risk in several different populations [44][45][46] as well as in a meta-analysis of 2906 cases and 3416 controls from four previous GWAS studies [47]. SMAD7 is transcriptionally induced by cytokines from the TGFβ family and regulates the TGFβ signaling pathway via a negative feedback loop [42]; therefore, the overexpression of SMAD7 inhibits the pathway and its associated anti-neoplastic effects [42]. The active form of vitamin D was shown to inhibit SMAD7 in experimental models [48]. The role of SMAD3 in the development of CRC is less understood and somatic tumor mutations in this gene have been observed in only 4.3% of CRC cases [49]. SMAD3 has been identified to interact with VDR and mediate a cross-talk between TGFβ and vitamin D signaling pathways [50]. An animal model found that SMAD3 may also play an important role in the TGFβ response to inflammation and bacteria-induced colon carcinogenesis [51]. Inflammation is further associated with CRC risk in our study as indicated by the interaction between circulating 25(OH)D concentrations, which has anti-inflammatory properties [52,53], and genetic variation in the IL10 gene encoding anti-inflammatory cytokine interleukin (IL)-10 involved in immune response to pathogens [54]. The induction of IL-10 is mediated by 1,25(OH) 2 D and is repressed with SMAD3 inhibition [48].
Chromosome 8q24 polymorphisms in the cell cycle pathway were previously identified to be strongly associated with CRC risk [47,55,56]. However, they were not statistically significant after BH correction. Although the 8q24 region is described as a gene desert, it is closely located to the region encoding c-MYC oncogene [57]. C-MYC controls processes related to cell growth regulation, metabolism and proliferation and is not only activated by numerous oncogenic pathways but also stimulates metabolic changes which can lead to malignant transformation [58]. Multiple studies have identified long-range physical interaction of the 8q24 region with c-MYC via enhancer elements and chromatin loops [57,59,60]. In an experimental study, 1,25(OH) 2 D and the VDR were shown to affect the c-MYC/MXD1 pathway leading to inhibition of c-MYC protein expression [61]. Using 8q24 SNPs as a proxy, our results confirm an association between c-MYC and CRC risk, but do not indicate a potential modification by 25(OH)D despite a previously reported possible interaction for fatal prostate cancer risk [62].
The IGF signaling pathway plays a key role in cell growth [63]. In our study, IGF-related genetic variation was associated with rectal cancer risk at the pathway level as well as for several individual genes before BH correction. Contrary to our results, IGF genetic variants [64,65] in addition to high circulating IGF peptides [66] have been previously associated more strongly with increased colon versus rectal cancer risk.
The strengths of our study include its prospective design and high follow-up rate. The hypothesis-driven selection of pathways, genes, and SNPs, and relatively large samples size within a large cohort study allowed an extensive investigation of vitamin D-related and -responsive genetic variation and the effect modification by established biomarker of vitamin D status with CRC risk. We used the detailed data from EPIC to address potential confounding by body size and other factors; with our careful analyses suggesting no or little confounding. However, we cannot altogether discount the possibility of residual confounding nor changes in lifestyle habits between enrolment into the cohort and the eventual cancer diagnosis. Although our study was large, most interaction and stratified analyses had limited power, especially by sex and tumor location. Our power analyses (Supplementary Table S10) showed that we have sufficient power (80%) to detect the effect associations in the range of 1.17 to 1.27 for relatively common SNPs with MAF between 40 and 10%, respectively, using our full data set (n = 1419 matched case-control sets). In addition, most of our results were not statistically significant after BH correction for multiple testing. As to the selection of genes and pathways, we were limited by published literature on vitamin D at the time of genotyping, so we may have not included all vitamin D-responsive genes. Additional experimental studies are needed to understand the biological mechanisms of the identified associations.

Conclusions
This large and comprehensive study has confirmed genetic variations in several previously identified vitamin D-related pathways associated with CRC risk in European populations, and has suggested potential new pathways related to vitamin D genomic effects and colorectal carcinogenesis.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6643/11/8/1954/s1, Table S1: Characteristics of SNPs used in the study, Table S2: Vitamin D metabolism and transcriptional activity-related SNPs associated (unadjusted P-value < 0.05) with season-adjusted 25(OH)D concentrations among controls, the EPIC study, 1992-2003, Table S3A: Associations between SNPs in the genes involved in the vitamin D metabolism, mineral homeostasis/endocrine regulation of 1,25(OH)2D synthesis, and vitamin D transcriptional activity with season-adjusted 25(OH)D concentrations among controls, Table S3B: Associations between SNPs in the genes involved in the vitamin D metabolism, mineral homeostasis/endocrine regulation of 1,25(OH)2D synthesis, and vitamin D transcriptional activity with season-adjusted 25(OH)D concentrations among cases and controls combined, Table S4: Associations of SNPs with CRC risk overall and by tumor location (colon vs. rectum) using unconditional logistic regression with adjustment for age at recruitment, study center and sex, the EPIC study, 1992-2003, Table S5 Associations of SNPs with CRC risk overall and by tumor location (colon vs. rectum) among men and women using unconditional logistic regression with adjustment for age at recruitment, study center and sex, the EPIC study, 1992-2003, Table S6: Associations of SNPs in vitamin D-responsive genes with CRC risk overall and by tumor location (colon vs. rectum) using unconditional logistic regression with adjustment for age at recruitment, study center and sex, the EPIC study, 1992-2003, Table S7: Statistically significant associations (unadjusted p < 0.01) of SNPs in vitamin D-responsive genes with CRC risk overall and by tumor location (colon vs. rectum) using unconditional logistic regression with adjustment for age at recruitment, study center and sex, the EPIC study, 1992-2003, Table S8: Associations of SNPs in vitamin D-responsive genes with CRC risk overall and by tumor location (colon vs. rectum) among men and women using unconditional logistic regression with adjustment for age at recruitment, study center and sex, the EPIC study, 1992-2003,