Association of Variants in Two Vitamin E Transport Genes with Circulating Vitamin E Concentrations and Prostate Cancer Risk

Significant reductions in prostate cancer incidence and mortality were observed in men randomized to receive 50 mg supplemental vitamin E (α-tocopherol) per day in the Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) Study. We hypothesized that variation in key vitamin E transport genes might directly affect prostate cancer risk or modify the effects of vitamin E supplementation. Associations between prostate cancer risk and 13 polymorphisms in two genes, TTPA and SEC14L2, were examined in 982 incident prostate cancer cases and 851 controls drawn from the ATBC Study. There was no association between the genetic variants and prostate cancer risk. Significant interactions were observed, however, between two variants in SEC14L2 (IVS11+931A>G and IVS11−896A>T) and the trial α-tocopherol supplement such that vitamin E supplementation reduced prostate cancer risk among men who were homozygous for either common allele [odds ratios (OR) and 95% confidence intervals (95% CI), 0.52 (0.30–0.90) and 0.64 (0.46–0.88), respectively] and nonsignificantly increased risk among those who carried one or two copies of either variant allele [ORs and 95% CIs, 1.27 (0.90–1.79) and 1.21 (0.96–1.52), respectively; both P for interaction < 0.05]. Genotype-phenotype analyses revealed significant but modest differences in baseline circulating concentrations of α-tocopherol and serum responses to the vitamin E supplementation for several polymorphisms. This study shows that genetic variation in TTPA and SEC14L2 is associated with serum α-tocopherol but does not have a direct effect on prostate cancer. Our results do, however, suggest that polymorphisms in SEC14L2 may modify the effect of vitamin supplementation regimens on prostate cancer risk. [Cancer Res 2009;69(4):1429–38]

Significant reductions in prostate cancer incidence and mortality were observed among male smokers randomized to receive 50 mg dl-α-tocopheryl acetate (vitamin E) daily for 5 to 8 years in the Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) Study (1). Several observational studies have also shown an inverse association between use of vitamin E supplements and prostate cancer risk, particularly in smokers (2–5). In response to these promising findings, the Selenium and Vitamin E Cancer Prevention Trial (SELECT) was initiated to test whether long-term supplementation with 400 IU all-rac-α-tocopheryl acetate, either alone or in combination with selenium, reduces the incidence of prostate cancer among 35,500 North American men (6).

Although all eight vitamin E isoforms, including four tocopherols and four tocotrienols, are absorbed in the intestine with equal efficiency, α-tocopherol is the most abundant form in blood and most tissues, including the prostate (7). This is due to the activity of the hepatic α-tocopherol transfer protein (α-TTP), which selectively incorporates this isoform into very low density lipoproteins (8). α-Tocopherol–enriched lipoproteins are then secreted into the circulation and delivered to peripheral tissues. α-TTP exhibits little affinity for the other vitamin E isomers, and consequently, they are metabolized to a large extent and excreted through bile and urine (9). Several rare insertion and deletion mutations in TTPA, the gene that encodes α-TTP, have been linked with ataxia with vitamin E deficiency (AVED), which is a severe neurologic disorder characterized by extremely low serum α-tocopherol concentrations (10).

Because vitamin E is highly lipophilic, additional proteins may be required to facilitate its intracellular transport and/or modulate its actions. The human α-tocopherol–associated protein (hTAP; identical to supernatant protein factor) is encoded by the SEC14L2 gene and binds to α-tocopherol with much higher affinity than any other vitamin E isoform; although expressed ubiquitously, it is most highly concentrated in liver, prostate, and brain tissue (11). Cell culture studies have shown that hTAP translocates from the cytosol to the nucleus on binding to α-tocopherol and subsequently activates gene expression (12). Other potential roles for this protein include mediating transport of α-tocopherol to organelles, such as Golgi or mitochondria, and assisting with the assembly of secretory granules (13). In addition, hTAP binds to several phospholipids and seems to play an important role in cholesterol biosynthesis (14, 15).

Common variants in genes encoding α-TTP and TAP may affect the transport and delivery of α-tocopherol to the prostate and ultimately influence cancer risk at this site. No study to date has evaluated whether such genetic variation affects prostate cancer risk or risk at any other cancer site. We therefore examined whether polymorphisms in TTPA and SEC14L2 were associated with circulating concentrations of vitamin E and prostate cancer risk in a large case-control study nested within a vitamin E intervention trial. The trial component of the study allowed us to additionally investigate whether these genetic variants modified the effects of vitamin E supplementation on prostate cancer risk.

Study population. The ATBC Study was a randomized, double-blind, placebo-controlled trial that tested whether daily supplementation with β-carotene (20 mg) and/or vitamin E (50 mg dl-α-tocopheryl acetate) reduced the incidence of lung and other cancers. Details about study design, methods, participant characteristics, and compliance have been reported (16). Briefly, 29,133 participants meeting all eligibility criteria at entry (male resident of southwestern Finland aged 50–69 y who smoked five or more cigarettes per day) were successfully randomized between 1985 and 1988. Reasons for exclusion included a prior history of cancer (other than nonmelanoma skin cancer or carcinoma in situ), serious illness, or refusal to discontinue use of vitamin E, vitamin A, or β-carotene supplements in excess of predefined amounts. The trial ended on April 30, 1993 after 5 to 8 y of active intervention (median, 6.1 y), and ascertainment of morbidity and mortality end points continued thereafter. The institutional review boards of both the National Public Health Institute of Finland and the U.S. National Cancer Institute approved the study, and written informed consent was obtained from each participant before randomization.

Data collection. Before randomization, all subjects were asked to provide detailed demographic, smoking, and occupational information, to give a history of medical examinations and physician-confirmed diseases, and to complete a 276-item dietary questionnaire that ascertained both frequency of intake and portion size. An overnight fasting blood sample was collected from virtually all participants at baseline, protected from light exposure, divided into aliquots, and stored at −70°C until analyzed. Three years after randomization, a follow-up blood sample was collected from each participant that was still actively participating in the trial (n = 22,881). Collection of a whole blood sample for genotyping studies took place between 1992 and 1993 (n = 20,311). Baseline and on-study serum concentrations of α-tocopherol, β-carotene, and retinol were determined using high-performance liquid chromatography (17), whereas total and HDL-cholesterol concentrations were measured with an enzymatic assay (CHOD-PAP method; Boehringer Mannheim). The between-run coefficient of variation for serum α-tocopherol was 2.2%.

Case ascertainment. Cases comprised all men diagnosed with incident prostate cancer (ICD-9 code 185) by April 30, 2003 that had DNA available for genotyping. These cancers were identified through the Finnish Cancer Registry, which provides ∼100% case ascertainment nationwide (18). For cases diagnosed through August 2001, medical records were reviewed centrally by two independent clinical oncologists for diagnostic confirmation and staging. Information on prostate cancer cases diagnosed since September 2001 was derived exclusively from the Finnish Cancer Registry. Advanced cases (n = 248) were defined as those with stage III or IV of the tumor-node-metastasis staging system, as defined by the American Joint Committee on Cancer (19), and/or those with a Gleason grade of 8 or higher. Prostate-specific antigen screening has not been widely adopted in Finland; thus, few cases in the cohort were identified through this mechanism.

Eligible controls were those subjects that were alive at the time of case diagnosis, free of prostate cancer through the end of follow-up, and had DNA available for genotyping. Controls were individually match to cases on age (±5 y), trial intervention group, and date of baseline blood draw (±30 d).

Single nucleotide polymorphism selection and genotyping. Single nucleotide polymorphisms (SNP) in the two genes of interest were identified by searching dbSNP, a publicly available database of polymorphisms maintained by the National Center for Biotechnology Information in conjunction with the National Human Genome Research Institute (20). Priority was given to variants with potential functional significance, including those located in promoters, coding regions, and exon-intron junctions. We also selected SNPs at regular intervals across each gene to enhance coverage (average coverage of 1 SNP per 6,200 bp in TTPA and 1 SNP per 3,300 bp in SEC14L2).

DNA was isolated from whole blood samples as previously described (21). All genotyping was conducted at the National Cancer Institute's Core Genotyping Facility using the Taqman or MGB Eclipse platform (sequence data and assay conditions can be found online⁸

). Selected SNPs were verified in a panel of 102 reference samples from four ethnically diverse groups by resequencing ∼300 bp of DNA on either side of the locus of interest. Genotyping assays were subsequently developed for known and newly discovered variants that exhibited a minor allele frequency of >5% in Caucasians and were considered to be validated when there was 100% concordance between sequence analysis and genotyping results on one or more of the aforementioned platforms.

For quality control purposes, duplicate specimens for 120 control samples were inserted into the sample set with their identity masked. Concordance between duplicates was >98% for all SNPs. The call rate exceeded 94% for all variants. One SNP in TTPA deviated significantly from Hardy-Weinberg equilibrium (HWE) in the controls (rs7818611, IVS2−2191A>G; P = 0.0005), indicating divergence from the expected distribution.

Statistical analysis. Differences in baseline characteristics between prostate cancer cases and controls were compared using the Wilcoxon rank sum test for continuous variables and the χ² test for categorical variables. Unconditional logistic regression was used to estimate odds ratios (OR) and 95% confidence intervals (95% CI) adjusted for age (continuous) and trial intervention assignment. Addition of other suspected risk factors for prostate cancer (body mass index, smoking habits, education, personal history of benign prostatic hyperplasia, personal history of diabetes mellitus, family history of prostate cancer, and dietary intakes of lycopene, selenium, red meat, phosphorous, calcium, and vitamin D) did not alter genotype-prostate cancer risk estimates by >10%, and these covariates were not included in the final models. The common-allele homozygous genotype served as the reference group in each analysis, and tests for trend were conducted by first assigning values of 0, 1, and 2 to the homozygous wild-type, heterozygote, and homozygous variant genotypes, respectively, and then modeling these scores as a continuous variable. In alternative analyses, we imposed a dominant model of inheritance by comparing risk in the combined group of heterozygotes and homozygous rare genotypes with risk among the homozygous common genotype.

We performed a global omnibus test for interaction for each gene with baseline serum α-tocopherol concentrations, vitamin E intake, and trial intervention assignment (α-tocopherol or no α-tocopherol) in relation to prostate cancer risk. This was accomplished by simultaneously including all of the polymorphisms in a given gene (coded by two dummy variables corresponding to the homozygous and heterozygous variant genotypes), the effect modifier of interest (continuous, where applicable), and all cross-product terms in a logistic regression model and then comparing it with a null model containing only the main effects of the genotypes and the putative modifier. These global tests automatically adjust for multiple testing based on the degrees of freedom of the corresponding likelihood ratio test. Interaction between individual variants and the aforementioned effect modifiers was also evaluated.

Pairwise measures of linkage disequilibrium (r²) were calculated between polymorphic loci in the same gene using Haploview (22). Haplotype frequencies, ORs, and 95% CIs were estimated using HaploStats,⁹

which reconstructs haplotypes and estimates ORs simultaneously based on a suitable expectation-maximization algorithm (23, 24).

Multiple linear regression was used to compute least squares means of baseline serum α-tocopherol concentrations according to individual genotypes and haplotypes among controls. Four observations were removed before analysis as they were identified as outliers (values beyond three times the interquartile range of serum α-tocopherol). We present both unadjusted and adjusted models, with the latter including significant predictors of baseline serum α-tocopherol concentrations in controls. We also investigated mean differences between baseline and postrandomization serum α-tocopherol concentrations across TTPA and SEC14L2 genotypes in controls that were randomized to the trial vitamin E intervention arm.

Statistical analyses were performed using Statistical Analysis Systems software, version 8.02 (SAS, Inc.).

There were no significant differences in age, smoking habits, body mass index, dietary intakes of red meat, α-tocopherol, and lycopene, or self-reported history of benign prostatic hyperplasia and diabetes between prostate cancer cases and controls (Table 1). Family history of prostate cancer was, however, twice as common among cases as controls. Circulating concentrations of serum α-tocopherol at baseline were similar in cases and controls. Among controls, the mean prerandomization serum α-tocopherol concentration was 12.0 mg/L; those randomized to the vitamin E intervention arm had 3-year on-study serum α-tocopherol concentrations of 18.2 mg/L, whereas those who did not receive the trial vitamin E supplement had 3-year on-study concentrations similar to baseline levels (mean, 12.7 mg/L). The study population was entirely Caucasian.

Carriage of both copies of the variant G allele of IVS2−2191A>G in TTPA was significantly inversely associated with the risk of prostate cancer overall (Table 2), although this finding must be interpreted cautiously as the observed genotype frequencies among controls deviated significantly from HWE. When the ORs and 95% CIs for this polymorphism were recalculated using expected rather than observed counts in controls, the point estimate among homozygous variant versus homozygous common genotypes was attenuated and the CI included unity [OR, 0.33 (0.10–1.06)]. None of the other variants in TTPA or SEC14L2 showed associations with prostate cancer risk, including advanced disease (Table 2). Accordingly, imputed haplotypes were also unrelated to disease risk (Table 2).

Although global omnibus tests did not show any significant effect modification of the relation between overall variation in each gene and prostate cancer risk by trial vitamin E supplementation, baseline serum α-tocopherol concentrations, or baseline dietary intake of α-tocopherol (all P values > 0.05), there were significant interactions when individual polymorphisms were examined (Table 3). Two SNPs in SEC14L2 (IVS11+931A>G and IVS11−896A>T; pairwise r² = 0.23) modified the previously reported effects of the trial vitamin E supplement on prostate cancer risk (1). In each instance, vitamin E supplementation reduced total prostate cancer risk among men who were homozygous for the common allele, whereas risk was somewhat elevated among those who carried one or two copies of the variant allele. For IVS11−896A>T, this interaction was also evident for advanced prostate cancer.

Two intronic SNPs in SEC14L2 (IVS11+931A>G and IVS11+1032C>T; pairwise r² = 0.02) modified the association between baseline dietary intakes of α-tocopherol and advanced prostate cancer. For IVS11+931A>G, higher vitamin E consumption seemed to be most protective against advanced prostate cancer among men who carried both common alleles, whereas suggestive increases in risk were evident in high vitamin E consumers who also carried one or two copies of the variant allele. The reverse was apparent for IVS11+1032C>T.

To assess whether the alleles studied were related to circulating vitamin E levels, we calculated raw and adjusted mean serum α-tocopherol concentrations (determined at baseline) for each genotype in controls (Table 4). Although unadjusted models showed no significant differences in blood concentrations across variants, in models that adjusted for predictors of serum vitamin E status, including serum cholesterol, education, prior vitamin E supplement use, and dietary α-tocopherol, we observed significant, albeit modest, variations in serum concentrations for one locus in TTPA (−980T>A) and six loci in SEC14L2 (EX1−23G>A R11K, IVS1+111G>A, IVS2+182A>G, IVS11+1032C>T, EX12−201T>C, and EX12−71G>A). The TTPA (−980T>A) variant is located in the promoter region, and carriers of both copies of the variant T allele had ∼3% lower baseline α-tocopherol concentrations than men with both copies of the wild-type A allele (P = 0.03). A common haplotype containing this variant allele (T-G-C-A) also significantly predicted lower baseline α-tocopherol concentrations (−0.35 mg/L; P = 0.006) relative to the haplotype bearing the common allele at each locus (A-G-C-A). Three of the alleles in SEC14L2 are located within exons, and for each, serum concentrations were significantly higher among men who were homozygous for the variant allele versus those homozygous for the common allele. The remaining three SNPs in this gene are located in noncoding intronic regions, with two (IVS1+111G>A and IVS11+1032C>T) showing 5% to 10% increases in serum α-tocopherol concentrations among men with the homozygous variant genotype. A haplotype containing five of six of these variants (A-A-G-A-T-G-T-C-A), including all three exonic mutations, significantly predicted increased serum α-tocopherol concentrations (0.26 mg/L; P = 0.04) compared with the referent haplotype (G-G-G-A-C-del-A-T-G).

Most of the variants examined did not affect serum responses to the trial vitamin E intervention, measured as the mean difference between baseline and on-study serum α-tocopherol concentrations, in either unadjusted or fully adjusted models (Table 5). A notably lower serum response to supplementation was observed, however, among men who carried one or both copies of the T allele of TTPA −980T>A compared with those who carried both copies of the more common allele. A common haplotype containing this variant allele (T-G-C-A) also significantly predicted lower serum response to the trial intervention (−1.22 mg/L; P = 0.0003) versus the referent haplotype.

This is the first study to examine whether common variants in two key genes involved in vitamin E transport are associated with prostate cancer risk. Our study is nested within the ATBC Trial, which enabled us to additionally investigate whether the selected variants modified the effects of the trial vitamin E supplementation on prostate cancer risk. Another unique strength is the availability of serum α-tocopherol concentrations, obtained at baseline and 3 years after randomization, for each participant. Although none of the selected polymorphisms was associated with prostate cancer risk, two variants in SEC14L2 seemed to modify the effects of the vitamin E intervention on risk. In addition, genotype-serum phenotype analysis revealed modest yet statistically significant differences in α-tocopherol concentrations between genotypes for several of the variant alleles.

We chose to investigate TTPA because it encodes a protein that is directly responsible for the preferential retention and eventual transport of α-tocopherol, the most biologically active form of vitamin E (8). Rare insertion and deletion mutations in TTPA are associated with severely reduced plasma (<1% of normal) and tissue concentrations of α-tocopherol, leading to AVED, a neurologic disorder that is easily treated with vitamin E supplementation (10). Importantly, point mutations in this gene have also been found in patients diagnosed with AVED, with substitutions that yield at least partially functional proteins conferring a milder disease phenotype than more severe truncating mutations (25). To our knowledge, only one study to date has investigated the temporal expression pattern of α-TTP during carcinogenesis; this study showed a marked decrease in α-TTP mRNA expression during the early stages of hepatocarcinogenesis in rats, with low levels evident throughout disease progression (26).

Although vitamin E regulatory mechanisms at the tissue level are largely unknown, hTAP (encoded by SEC14L2) has emerged as a promising candidate because it preferentially binds α-tocopherol compared with other isoforms and seems to be an important carrier protein, mediator of vitamin E function, or both (12). hTAP is highly expressed in prostate tissue, with levels in this organ second only to those found in liver (11); this suggests that hTAP may serve a crucial role in facilitating vitamin E–related activities in the prostate. Of note is a recent study that showed that hTAP mRNA and protein expression levels were significantly down-regulated in human prostate cancer tissues compared with normal prostatic epithelium (27). Mutations in the gene encoding hTAP, SEC14L2, have not been linked to any disease phenotype thus far.

We found modest differences in serum α-tocopherol concentrations, as well as serum responses to vitamin E supplementation, for several of the variants examined, yet only the TTPA IVS2−2191A>G polymorphism, which deviated significantly from HWE, showed any association with prostate cancer risk. Mendelian randomization dictates that associations between disease and functional genetic variants (i.e., those that predict phenotypic differences in a modifiable exposure) provide an unbiased estimate of risk associated with the environmental exposure of interest (28, 29). Although our findings do not provide strong genetic support for a causal relationship between vitamin E and prostate cancer, coverage of all variation in TTPA and SEC14L2 was not comprehensive; it is possible that other common SNPs in these genes are associated with circulating α-tocopherol concentrations and could have a greater biological effect on prostate cancer than the known variants that we were able to examine. As noted below, more extensive characterization of functional (or other) variation in these two genetic loci is needed.

Two SNPs in SEC14L2 modified the effect of trial vitamin E supplementation on prostate cancer risk, although neither significantly affected serum responses to the intervention. The association between baseline dietary α-tocopherol intake and prostate cancer risk, previously reported to be null in this study population (30), was also modified by two intronic SNPs in SEC14L2. Of particular note was that men who carried both copies of the variant allele of SEC14L2 IVS11+1032, in whom protective associations between dietary α-tocopherol intake and advanced prostate cancer risk were observed, also had significantly higher mean baseline serum α-tocopherol concentrations compared with subjects who were homozygous for the common allele. This lends credence to the observed interaction but requires confirmation in other studies. Because all global tests for interaction, which effectively account for multiple comparisons, were null, these single SNP interactions could be due to chance and should be interpreted with caution.

In addition to the strong a priori biological plausibility of our hypothesis, strengths of our investigation include the relatively large sample size and the nested prospective case-control design, which limited the influence of disease on serum nutrient concentrations. We were also able to rule out confounding of genotype-prostate cancer associations by testing a large number of potential risk factors for prostate cancer.

This study has several limitations. First, we evaluated publicly available SNPs rather than resequencing the functional domains of each gene; the latter approach would provide a more comprehensive assessment of variation. Second, although we focused on two promising candidate genes involved in vitamin E transport, other genes encoding proteins that bind α-tocopherol with relatively high affinity and potentially play an important role in determining its bioavailability should be examined. These include, for example, genes involved in the intestinal uptake, plasma transport, cellular uptake and efflux, and metabolism of lipids and cholesterol, as α-tocopherol is extremely lipophilic and bound to lipoproteins in plasma (31, 32). Additional candidate genes include those that influence the status of other micronutrients that act in concert with vitamin E in the prostate; for example, in vitro studies have shown that vitamin E acts synergistically with both selenium and lycopene to inhibit prostate cancer cell growth (33, 34). Genome-wide association studies will also contribute to the exploration of these hypotheses. Third, Finland is an isolated founder population, which has implications when extrapolating our results to other populations (35). Genotype frequencies among the controls in our population were generally similar to those reported among Caucasians in two publicly available databases: the SNP500Cancer¹⁰

database and the International HapMap Project¹¹ database. In these databases, however, there are notable differences in genotype distributions across racial subgroups for several of the SNPs that we examined; our results may therefore not be generalizable to more admixed populations. Fourth, our study population consisted entirely of current smokers, which could limit the generalizability of the gene-serum interactions and genotype-phenotype findings because smokers seem to have higher vitamin E requirements than never smokers due to enhanced antioxidant depletion rates (36). Verification of the findings of the current study in nonsmokers is desirable. Finally, hTAP seems to have multiple biological functions [including stimulation of cholesterol biosynthesis (14, 15)] and our results concerning SEC14L2 may not be entirely attributable to its purported role as a vitamin E transport protein and mediator of α-tocopherol–dependent gene expression.

In summary, this study provides evidence that common polymorphisms in two vitamin E transport genes, TTPA and SEC14L2, may modify the association of supplemental and dietary vitamin E intake with prostate cancer and also modestly affect circulating concentrations of vitamin E. Future efforts should concentrate on fully characterizing the extent of variation across these and related genes and testing our findings with respect to prostate cancer risk in other populations, including nonsmokers and racially diverse groups. If our observations are confirmed, they could have important implications about the design, analysis, and interpretation of ongoing and future chemoprevention trials, including SELECT (6).

No potential conflicts of interest were disclosed.

Grant support: Intramural Research Program of the National Cancer Institute, NIH.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.