Comprehensive Association Analysis of the Vitamin D Pathway Genes, VDR, CYP27B1, and CYP24A1, in Prostate Cancer

Genetic variation in vitamin D–related genes has not been investigated comprehensively and findings are equivocal. We studied the association between polymorphisms across the entire vitamin D receptor (VDR) gene and genes encoding for vitamin D activating enzyme 1-α-hydroxylase (CYP27B1) and deactivating enzyme 24-hyroxylase (CYP24A1) and prostate cancer risk among middle-aged men using a population-based case-control study design. DNA samples and survey data were obtained from incident cases (n = 630), 40 to 64 years old, identified through the Seattle-Puget Sound Surveillance, Epidemiology, and End Results cancer registry from 1993 to 1996 and from random controls (n = 565) of similar age without a history of prostate cancer. We selected and genotyped tag single-nucleotide polymorphisms to predict common variants across VDR (n = 22), CYP27B1 (n = 2), and CYP24A1 (n = 14). Haplotypes of VDR and CYP24A1 were not associated with prostate cancer risk. In the genotype analysis, homozygotes at two VDR loci (rs2107301 and rs2238135) were associated with a 2- to 2.5-fold higher risk of prostate cancer compared with the homozygote common allele [odds ratio, 2.47 (95% confidence interval, 1.52-4.00; P = 0.002) and 1.95 (95% confidence interval, 1.17-3.26; P = 0.007), respectively; P value corrected for multiple comparisons for VDR = 0.002]. We found no evidence that the two associated VDR single-nucleotide polymorphisms were modified by age at diagnosis, prostate cancer aggressiveness, first-degree family history of prostate cancer, or vitamin D intake. Genotypes of CYP27B1 and CYP24A1 were not associated with prostate cancer risk. Our findings suggest that polymorphisms in the VDR gene may be associated with prostate cancer risk and, therefore, that the vitamin D pathway might have an etiologic role in the development of prostate cancer. (Cancer Epidemiol Biomarkers Prev 2007;16(10):1990–9)

An increasing number of studies show a chemopreventive role of vitamin D in prostate cancer; however, the association remains equivocal (1-7). Experimental studies have shown prostate cancer growth inhibition by vitamin D as the result of an increase in apoptosis, inhibition of cell-cycle progression, interaction with the insulin-like growth factor growth factor axis, and a reduction of metastatic potential, among other potential mechanisms (8). If vitamin D is an important factor in the development of prostate cancer, then it might be expected that common functional sequence polymorphisms in the genes that influence vitamin D metabolism and function could predispose to the disease.

The cellular effects of 1,25-dihydroxy-vitamin D [1,25(OH)₂D], the hormonally active form of vitamin D, are primarily mediated through binding to the nuclear vitamin D receptor (VDR), which regulates the transcription of numerous genes including proto-oncogenes and tumor suppressor genes (9). A large number of studies have related common genetic variants in VDR to risk of prostate cancer (reviewed in refs. 10, 11). These studies have focused mainly on five polymorphisms: FokI (rs10735810), BsmI (rs1544410), ApaI (rs7975232), and TaqI (rs731236) restriction sites and the poly-A microsatellite, all of which except for the FokI (rs10735810) variant are in strong linkage disequilibrium (LD) within Caucasian populations. Several studies have shown significant associations with one or more of these five variants; however, overall results remain inconsistent. Recent comparative, high-resolution analysis of LD in the VDR gene showed substantial genetic variability and showed that the commonly studied VDR variants provide very incomplete coverage of the variation within this gene, which spans 63.4 kb (12).

The genes CYP27B1 and CYP24A1 are directly involved in vitamin D metabolism and hence are interesting genes of the vitamin D pathway. CYP27B1 encodes for 1-α-hydroxylase, which converts 25-hydroxy-vitamin D [25(OH)D] into the active vitamin D metabolite 1,25(OH)₂D. CYP24A1 encodes for 24-hydroxylase, the enzyme that catalyzes the inactivation of 1,25(OH)₂D. Only two studies investigated thus far the effect of the CYP27B1 (13) and CYP24A1⁵

polymorphisms on prostate cancer.

Despite the large number of previously conducted studies, genetic variation in VDR has not been systematically analyzed with regard to prostate cancer, and very limited data are available on CYP27B1 and CYP24A1 polymorphisms. To address this, we conducted a population-based case-control study to comprehensively examine the relation between tag single-nucleotide polymorphisms (tag SNP) in three key genes in the vitamin D pathway, VDR, CYP27B1, and CYP24A1, and prostate cancer risk.

We used data from a population-based case-control study of risk factors for prostate cancer in middle-aged, male residents of King County, Washington, which has previously been described (14). Because the initial study was designed to examine the risk of prostate cancer in middle-aged men, all males ages 40 to 59 years and a random 75% sample of patients ages 60 to 64 years were invited to participate. Cases were diagnosed with histologically confirmed prostate cancer between January 1, 1993 and December 31, 1996, and identified through the Seattle-Puget Sound Surveillance, Epidemiology, and End Results cancer registry. Control subjects were men without a self-reported history of prostate cancer who resided in King County, Washington, identified by random digit dialing, frequency matched to cases by 5-year age groups, and recruited evenly throughout the ascertainment period of the cases. There were 917 eligible cases of prostate cancer. Men who completed the interview (n = 752) were eligible for genotyping; 630 (69%) provided blood samples yielding sufficient DNA for genotyping. Of the 941 eligible controls, 703 men completed the interview and were eligible for genotyping; 565 (60%) provided blood samples.

Detailed information about demographics, individual behaviors, and medical history was obtained from a structured in-person interview. Consenting subjects each provided a blood sample for genotyping and completed a self-administered 99-item food frequency questionnaire and a supplement use survey, which included vitamin D in multivitamins only. Exposure information was collected up to the reference date (date of diagnosis for cases and an assigned date for controls that approximated the distribution of the cases' diagnosis dates). For cases, the cancer registry provided information on stage and Gleason score of prostate cancer at diagnosis.

Using a previously described approach (15), we selected tag SNPs that have r² ≥ 0.8 with all common genetic variants (minor allele frequency ≥5% in Caucasian populations) in each of the three genes, VDR (12), CYP27B1,⁶

and CYP24A1.⁷ As we try to use the most comprehensive source of SNP discovery to select tag SNPs, we used resequencing data when available (VDR and CYP27B1). If such data were not available, we based SNP selection on HapMap data (CYP24A1). A total of 25 tag SNPs for VDR (chromosome 12q13, length 63.4 kb, 9 exons), 3 tag SNPs for CYP27B1 (chromosome 12q13, length 4.8 kb, 9 exons), and 20 tag SNPs for CYP24A1 (chromosome 20q13; length 20.5 kb, 12 exons) were chosen.

Genomic DNA was prepared using a standard alkaline lysis method followed by phenol-chloroform extraction and stored at −80°C. The Applied Biosystems (ABI) SNPlex Genotyping System was used to genotype SNPs in individual DNA samples. Proprietary GeneMapper software was used for calling alleles.⁸

The SNPlex assay could not be designed for 5 tag SNPs and 3 tag SNPs failed genotyping after the design stage; accordingly, we present results for 22 tag SNPs for VDR (Fig. 1), 2 tag SNPs for CYP27B1, and 14 tag SNPs for CYP24A1. Quality control included genotyping of 76 blind duplicate samples, which revealed ≥99% agreement on genotyping calls across all tag SNPs assayed. The call rate was ≥97% for all but two SNPs (VDR rs2238139, 96%; CYP24A1 rs6127118, 95%). Genotype frequencies for all loci were consistent with Hardy-Weinberg proportions (P > 0.05) among Caucasian controls, the predominant ethnic group in this study, with the only exceptions of VDR rs2238139 (P = 0.004) and CYP24A1 rs13038432 (P < 0.001).

Pairwise LD was estimated between tag SNPs based on D and r² values (16) using Haploview software (version 3.2; ref. 17). For CYP27B1 and CYP24A1, LD block structure was evaluated according to the criteria defined by Gabriel et al. (18) using Haploview. The LD block structure for VDR has previously been examined (12). For each haplotype block, haplotype frequencies and associated measures of effect were estimated using HaploStats (version 1.2.1)⁹

for R (version 2.1.0; ref. 19), which uses the expectation-maximization algorithm to estimate haplotype frequencies and an iterative two-step expectation-maximization model to estimate the association between individual haplotypes and risk, assuming an additive model (20, 21). A global score statistic, adjusted for age and ethnicity, was used to evaluate the overall difference in haplotype frequencies between cases and controls.

Unconditional logistic regression models were used to estimate odds ratios (OR) and 95% confidence intervals (95% CI), adjusted for age and ethnicity, for the association between individual VDR, CYP27B1, and CYP24A1 tag SNPs and prostate cancer risk by using SAS software version 9.1 (22). Factors that changed the risk estimates by >10% were considered potential confounders: first-degree family history of prostate cancer; prostate-specific antigen (PSA) testing history (number of PSA tests done in the 5 years before reference date); history of benign prostatic hyperplasia; education; body mass index (kg/m²); physical activity; smoking history; alcohol consumption; and dietary and supplemental intake of vitamin D and calcium did not alter the observed associations. Genotypes were evaluated using indicator variables and the fit of an additive genetic model (analogous to the P value for linear trend) was assessed by using a single continuous variable that coded for the number of variant alleles present. We assessed the robustness of the findings by calculating the false-discovery rate (23), defined as the expected ratio of erroneous rejections of the null hypothesis to the total number of rejected hypotheses, which yields a P value corrected for multiple comparisons.

Individual tag SNP results were stratified by age (median, 58 years), first-degree family history of prostate cancer (yes/no), and ethnicity (Caucasian/African American) because younger men, men with a family history of disease, as well as African American men could be at an increased risk based on genetic background. Because associations may differ according to clinical characteristics of the tumor, we stratified by stage of disease, Gleason score, and a composite variable reflecting prostate cancer aggressiveness {defined as less aggressive [local stage, Gleason score 2-6 or 7 (3 + 4), and diagnostic serum PSA <20 ng/mL] or more aggressive [regional/distant stage, Gleason score 7 (4 + 3) or 8-10, and serum PSA ≥20 ng/mL]}. Lastly, associations were examined by total vitamin D (total daily intake from diet and supplements; median, 400 IU/d) because circulating levels of vitamin D may modify the association between VDR polymorphisms and prostate cancer risk.

Most prostate cancer cases were <60 years of age at diagnosis and cases were slightly older than controls (average, 57.5 and 56.7 years, respectively; Table 1). The study population was predominately Caucasian and the percentage of African Americans was higher among cases. Compared with controls, cases were more likely to have a first-degree family history of prostate cancer, a history of benign prostatic hyperplasia, and had a PSA test in the last 5 years. Education, body mass index, physical activity, smoking history, alcohol consumption, and total vitamin D and calcium intake were comparable for cases and controls. The majority of cases was diagnosed with localized stage disease (73%), had moderate grade [Gleason score 5-7 (3 + 4)] tumors (76%), and had less aggressive disease (65%).

None of the VDR or CYP24A1 haplotypes were significantly associated with prostate cancer risk when compared with men who carried the most common haplotype within each block (Table 2; the two CYP27B1 SNPs did not fall in one haplotype block). Global tests for any block also indicated a lack of an association.

Examining individual polymorphisms, risk of total prostate cancer was significantly associated in a dose-dependent manner with two tag SNPs in the VDR gene (Table 3). Men who were homozygote for the rare allele for VDR tag SNP rs2107301 had a 2.5-fold higher risk of prostate cancer compared with those who were homozygote for the common allele (95% CI, 1.52-4.00; P = 0.002). Furthermore, men who were homozygote for the rare allele for the VDR tag SNP rs2238135 had a 2-fold higher risk of prostate cancer compared with those who were homozygote for the common allele (95% CI, 1.17-3.26; P = 0.007). These two tag SNPs were located in two different LD blocks in the VDR gene and, accordingly, the LD between these two SNPs was low (D = 0.004). The estimate from the test for trend was borderline statistically significant after adjustment by the false-discovery rate method (P = 0.002). Simultaneous adjustment for the VDR tag SNPs rs2107301 or rs2238135 in a model did not affect association of the other SNP, indicating independent effects for each of these two SNPs (data not shown). We examined whether the combined effects of VDR tag SNPs rs2107301 and rs2238135 were associated with a substantial elevated risk of prostate cancer, but found no evidence of an interaction (P = 0.30). We also observed a suggestive association between VDR TaqI and prostate cancer risk; men carrying one or two variant alleles had a 23% lower risk. However, this result did not remain significant after adjustment for multiple comparisons by the false-discovery rate method. None of the tag SNPs in the CYP27B1 or CYP24A1 gene was significantly associated with prostate cancer risk.

Examination of the individual tag SNPs for VDR, CYP27B1, or CYP24A1 by age, first-degree family history of prostate cancer, prostate cancer aggressiveness, or total vitamin D intake did not reveal any appreciable heterogeneity of the risk estimates (data not shown). Restricting analyses to Caucasians did not substantially change the estimates. For example, comparing the homozygous rare allele to the homozygous common allele, the risk of prostate cancer among Caucasians at VDR loci rs2107301 (cases = 545) and rs2238135 (cases = 565) was 2.45 (95% CI, 1.51-3.98; P = 0.003) and 2.01 (95% CI, 1.18-3.42; P = 0.01), respectively.

Our systematic analysis of genetic variation in three key genes of the vitamin D pathway showed that risk of prostate cancer was significantly associated in a dose-dependent manner with two VDR tag SNPs (rs2107301 and rs2238135). We found no evidence that the two associated VDR SNPs have an effect on the age at diagnosis or prostate cancer aggressiveness. There was also no evidence that their association was modified by the first-degree family history of prostate cancer or dietary vitamin D intake. Tag SNPs in CYP27B1 and CYP24A1 and haplotypes in VDR and CYP24A1 were not associated with prostate cancer risk.

The nuclear receptor VDR is a member of the steroid hormone receptor superfamily of transcriptional regulators. On binding to 1,25(OH)₂D, VDR regulates the transcription of numerous target genes by direct binding to the promoter/enhancer sequence elements of these genes. The VDR gene encompasses two promoter regions, eight protein-coding exons (2-9), and six untranslated exons (1a-1f; Fig. 1). The presence of VDR has been shown to be essential for growth inhibitory effects of 1,25(OH)₂D (24) and VDR is expressed in both normal and malignant prostatic epithelial cells (25, 26). Five common VDR polymorphisms have intensively been studied. As summarized in two recent meta-analyses (10, 11) including 26 studies, overall, these studies do not support an effect of any of the five polymorphisms on risk of prostate cancer, which is consistent with our findings for BsmI (rs1544410), TaqI (rs731236), and FokI (rs10735810). However, our study indicated an association of two variant alleles located in introns 2 and 3, which remained borderline significant after adjusting for multiple comparisons. Both tag SNPs rs2107301 and rs2238135 are located in different LD blocks, and combined analysis suggested independent effects for each SNP. These two SNPs are not located in evolutionary conserved regions or known splicing sites; therefore, it is possible that these SNPs are in LD with functional polymorphisms. Although both significant VDR SNPs were located within a separate haplotype block, we observed no significant association with prostate cancer risk for any haplotype. However, this is not surprising because the haplotype analysis was conducted under the additive model and we observed a significant genotype association under the dominant model. Furthermore, possibly due to the large number of tag SNPs per haplotype block (n = 10), a large fraction of haplotypes were rare (<5%) and lumped together (40% and 66% in blocks 1 and 2 of the VDR gene, respectively). In contrast to our study, no association between rs2107301 and prostate cancer risk was observed in a hospital-based case-control study of prostate cancer [n = 430; OR for TT versus CC, 1.34 (95% CI, 0.65-2.73); P = 0.43; ref. 27]. As it has been shown that vitamin D inhibits growth of benign prostatic hyperplasia in cell culture studies (28), it is possible that genetic variants of VDR affect risk of benign prostatic hyperplasia. Therefore, the use of benign prostatic hyperplasia controls may have attenuated any associations between VDR polymorphisms and prostate cancer risk. Furthermore, recent summary data from a genome-wide scan, the Cancer Genetic Markers of Susceptibility, including 1,188 prostate cancer cases, have become publicly available.⁵

This genome-wide study observed no association with prostate cancer for any of the SNPs genotyped within the VDR gene, including rs2107301, which was significantly associated in our study.⁵ The discrepancy between studies may be explained, in part, by the use of early-onset compared with late-onset cases. Our study included men ages 40 to 64 years; Cancer Genetic Markers of Susceptibility included men ages 55 to 74 years. Early-onset cases may be enriched for cases that are related to genetic risk factors; however, our results require further corroboration.

CYP27B1, which encodes for the vitamin D activating enzyme 1-α-hydroxylase, and CYP24A1, which encodes for the vitamin D–deactivating enzyme 24-hydroxylase, are both expressed in prostate epithelial cells (29). Consistent with this finding, a similar antiproliferative effect was described for 25(OH)D as compared with 1,25(OH)₂D in human prostate cancer cells in primary cultures that express 1-α-hydroxylase (30-32). In prostate cancer cells, 1-α-hydroxylase activity is down-regulated whereas high levels of 24-hydroxylase are found in prostate cancer cells (31, 33), suggesting lower concentration of the metabolically active form of vitamin D in malignant cells.

Findings for CYP27B1 and CYP24A1 variants have not been widely reported. A single hospital-based case-control study with 245 prostate cancer cases observed no association between CYP27B1 SNPs (−1260 C>A, +2838 T>C, +3545 A>C) and prostate cancer risk (13). In addition, the genome-wide scan for prostate cancer found no significant association with prostate cancer risk for CYP27B1 (including rs4646537) or CYP24A1 (including rs927650, rs912505, rs6068816, rs3787557, and rs4809960).⁵

Overall, these studies, including ours, do not provide evidence for a major effect of the CYP27B1 and CYP24A1 polymorphisms on prostate cancer risk. Given the sample size of these studies, we cannot rule out any weak associations because the studies were only powered to observe moderate effect sizes. Furthermore, it is possible that genetic variants may mediate prostate cancer risk via a mechanism involving availability of 1,25(OH)₂D (e.g., certain variants may only be relevant in men with vitamin D deficiency, a common condition with aging; refs. 34, 35). We tested for interaction with intake of vitamin D; however, we only had information on dietary vitamin D and vitamin D from multivitamins and had no information on individual vitamin D supplements and, importantly, sun exposure (UV-B radiation), the major source of vitamin D (36). The importance of considering vitamin D exposure is supported by findings for significant interactions of VDR polymorphisms and serum vitamin D concentration, an integrative measure of endogenous vitamin D production via UV-B radiation and vitamin D intake, and interaction of VDR and sun exposure (37-39).

Prostate cancer occurs more frequently among African Americans, and this characteristic is consistent with a role for vitamin D deficiency in the etiology of this disease (40). In addition, different genotype distributions for Caucasians and African Americans have been observed (12); thus, genetic polymorphisms in vitamin D–related genes may partly explain variations in prostate cancer risk among ethnic groups. In our study, associations were similar when analyses were restricted to Caucasians and when all men were considered. Given the small number of African American men (n = 49) in our study, we were not able to specifically investigate the effect of genetic variants among this group.

The strengths of our study include the comprehensive analysis of common genetic variants in VDR, CYP27B1, and CYP24A1; its population-based design; the relatively large sample size; the use of early-onset cases (which may be the more important phenotype related to genetic risk factors); and the ability to stratify the data by potential effect modifiers and clinical features of disease.

The potential for selection bias, a possible limitation of our study design, may be limited because it is unlikely that the genotype of the selected vitamin D–related genes is related to study participation except for the less likely event that VDR tag SNPs are linked to unknown SNPs that are associated with behavior or risk factors affecting participation rates. Furthermore, genotyping results were only available for a subset of cases and controls; however, we observed no significant differences in disease characteristics or pertinent prostate cancer risk factors, including age and family history of disease, between genotyped men and all eligible men. Despite our effort to comprehensively investigate the genetic variation in these three genes, the SNPlex assay could not be designed for five tag SNPs and three tag SNPs failed genotyping after the design stage; these tag SNPs could potentially be associated with prostate cancer risk. Lastly, our analysis was mainly among Caucasian men, and results may not be generalizable to other ethnic groups.

With more complete coverage of genetic variation in vitamin D pathway genes compared with previous studies, our findings suggest that polymorphisms in the VDR gene may be associated with prostate cancer risk and, therefore, that the vitamin D pathway might have an etiologic role in the development of prostate cancer. However, given the discrepancies with other studies on VDR polymorphisms and prostate cancer, these results require further corroboration in large statistically powerful sample collections.

We thank Ilir Agalliu and Claudia Salinas for their help with data management and valuable comments, and all the men for their time and cooperation in participating in this study.