Serum Metabolomic Response to Low- and High-Dose Vitamin E Supplementation in Two Randomized Controlled Trials

The lipid-soluble, essential nutrient vitamin E is a critical cellular antioxidant that inhibits lipid peroxidation, platelet aggregation, and inflammation, and vitamin E supplementation has been tested for cancer and cardiovascular disease prevention for decades with mixed results (1–4). Prostate cancer has received substantial attention in this regard, with several large randomized controlled trials (RCT) examining the hypothesis. The first of these reported significant 32% and 40% reductions in prostate cancer incidence and mortality, respectively, in response to a modest daily dose of 50 IU alpha-tocopheryl acetate (ATA), albeit the findings were secondary hypotheses in the Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study (ATBC), which was focused primarily on lung cancer (4–6). On the basis of these and other basic and observational data, two additional RCTs subsequently tested vitamin E for prostate cancer prevention. The Physicians' Health Study–II (PHS-II) reported no effect of 400 IU every other day (2), while the Selenium and Vitamin E Cancer Prevention Trial (SELECT) found 17% increased prostate cancer incidence following 7–12 years of supplementation with 400 IU ATA daily (3). In comparison with the two supplementation dosages, the current recommended U.S. dietary allowance for vitamin E is 22.4 IU daily, and the majority of multivitamin supplements provide 30 IU (7). The biological basis and molecular responses for the qualitative difference between ATBC (50 IU daily) and SELECT (400 IU daily) has not been elucidated. In addition, a previous meta-analysis showed that high-dose vitamin E supplementation (≥400 IU daily) may increase all-cause mortality, whereas low doses appeared to have no effect on mortality (8).

The present investigation measured the serum biochemical changes in men receiving 400 IU ATA in a controlled clinical trial, the Vitamin E Atherosclerosis Prevention Study (VEAPS; ref. 9), and an additional set of men supplemented with 50 IU ATA in ATBC to gain biological insight into vitamin E dosage–related effects potentially relevant to the divergent prostate cancer findings in previous RCTs.

VEAPS was a randomized, double-blinded, placebo-controlled trial primarily designed to test whether vitamin E supplementation prevents progression of subclinical atherosclerosis in healthy individuals (9). Trial participants (n = 353 men and women ages 40 and older without clinical evidence of diabetes or cardiovascular disease) were randomly assigned to receive (i) placebo (n = 176), or (ii) alpha-tocopheryl acetate (n = 177; ATA, 400 IU) daily for two to three years and followed through study clinic visits every 6 months during which fasting blood samples were collected and stored at −80°C. This analysis included 154 men (81 placebo and 73 ATA) with available baseline and on-study fasting serum samples obtained at the 1-year clinic visit (participant flowchart in Supplementary Fig. S1). All study participants submitted written informed consent, the study protocol was approved by the University of Southern California Institutional Review Board, and the trial was registered on clinicaltrials.gov (NCT00114387; investigators of SELECT were invited to collaborate for a similar metabolomic profiling analysis of their trial population to compare to ATBC, but they declined to do so).

ATBC was a randomized, double-blinded, placebo-controlled primary prevention trial conducted to evaluate the effects of supplementation with vitamin E and beta-carotene on the incidence of lung and other cancers (10). The trial recruited 50–69 year old male smokers (n = 29,133) from 1985 to 1988 in southwestern Finland. The trial participants were randomly assigned to receive one of four supplements daily: (i) ATA (50 IU), (ii) beta-carotene (20 mg), (iii) both vitamins, or (iv) placebo for 5–8 years (a median of 6.1 years). Fasting serum was collected and stored at −70°C for all trial participants at baseline and for a random sample during the trial. For this analysis, 100 men with serum at baseline and after an average of 1.7 years of supplementation (range: 49–332 weeks) with ATA alone (n = 50) or placebo (n = 50) and not previously examined for metabolite profiles were selected (participant flowchart in Supplementary Fig. S2). All trial participants provided written informed consent, ATBC was approved by Institutional Review Boards in both the Finnish National Public Health Institute and the NCI. The trial was registered on clinicaltrials.gov (NCT00342992).

The high-resolution accurate mass (HRAM) platform (global “HD4”) of ultrahigh-performance LC/MS-MS of Metabolon, Inc. was used to measure 1,217 baseline and follow-up serum metabolites in both VEAPS and ATBC. After excluding unknown compounds and those with values below the limit of detection in more than 90% of participants, 974 metabolites were included in the final analysis. Metabolites were batch normalized by dividing by the batch mean of all nonmissing values, missing values were imputed to one-half of the minimum detectable metabolite value, and metabolites were classified according to eight chemical classes: amino acids, carbohydrates, cofactors and vitamins, energy metabolites, lipids, nucleotides, peptides, or xenobiotics, where the categorization was based on the databases from Kyoto Encyclopedia of Genes and Genomics (KEGG) and Human Metabolome Database (HMDB). Samples from VEAPS and ATBC were measured at the same time in the same platform runs but in separate batches. The paired baseline and follow-up samples from each study were tested within the same batches. Coefficients of variations (CV) and intraclass correlation coefficients (ICC) were calculated for each metabolite using replicate quality control samples placed in each batch from pooled ATBC serum (32 total quality control samples in 16 batches). The median ICC and CV for all measured metabolites were 0.96 (interquartile range = 0.88 to 0.99) and 0.12 (interquartile range = 0.07 to 0.24) in VEAPS, and 0.96 (interquartile range = 0.87 to 0.99) and 0.11 (interquartile range = 0.06 to 0.24) in ATBC, respectively.

All batch-normalized metabolites were log-transformed and standardized (mean = 0 and variance = 1) in the VEAPS Study and ATBC Study, separately. In the main analysis, we investigated the association between change in log-metabolite concentrations and trial assignment (i.e., ATA or no ATA) using linear regression (using PROC GLM in SAS) in each study separately. Here, change is defined as the log-level at follow-up minus the log-level at baseline. In both studies, the threshold for statistical significance of the primary analysis was 5.1 × 10⁻⁵ according to Bonferroni correction for 974 tests to get a family-wise error rate of 0.05. In a sensitivity analysis, we further adjusted for potential confounding factors obtained at baseline, including age, height, weight, smoking status, smoking years, serum high-density and total lipoprotein cholesterol, and alcohol consumption. In the secondary analyses, we modeled the association between change in log-metabolite concentrations and change in log serum C₂₂ lactone sulfate (“X-12063”) by linear regression in VEAPS. For the metabolites significantly correlated with C₂₂ lactone sulfate at the Bonferroni correction threshold, we further examined their changes in response to ATA supplementation in linear regression in VEAPS and ATBC. A sensitivity analysis was restricted to ATBC men with on-study serum obtained within 1,000 days of randomization (n = 62). All analyses were conducted using SAS version 9.4 (SAS Institute), and all statistical tests are two-sided.

Characteristics of the study participants at trial entry in VEAPS and ATBC according to ATA supplement assignment are presented in Table 1. With the exception of significant increases in serum alpha-tocopherol concentrations (measured previously by reversed-phase HPLC with UV detection) during vitamin E supplementation in ATBC and VEAPS, there were no participant differences by intervention assignment in either trial.

On the basis of the 974 metabolites identified in both the VEAPS and ATBC trials, only one, C₂₂ lactone sulfate (X-12063), was significantly decreased by ATA supplementation after multiple comparisons correction in VEAPS (beta = −0.70, P = 8.1 × 10⁻⁶) but not in ATBC (beta = −0.17, P = 0.4; Table 2). In contrast, three metabolites were significantly altered in both VEAPS and ATBC: alpha-CEHC sulfate (β = 1.56 and 1.44, P = 10⁻³³ and 10⁻¹⁷, respectively), beta/gamma-tocopherol (β = −1.31 and −0.97, P = 10⁻²⁰ and 10⁻⁷, respectively), and alpha-tocopherol (β = 1.04 and 0.98, P = 10⁻¹² and 10⁻⁷, respectively; Table 2). The findings were not changed after further multivariable adjustment of the vitamin E supplement-metabolite change associations for the possible biological influences of age, weight, height, body mass index, smoking status, smoking years, serum high-density and low-density lipoprotein cholesterol, dietary vitamin E intake, and alcohol consumption (Supplementary Table S1). Our findings were not essentially changed when we restricted analyses to men who were supplemented with vitamin E for less than 1,000 days in ATBC (Supplementary Table S2). The correlations between changes of all measured metabolites and response to vitamin E supplementation in VEAPS and ATBC are presented in Supplementary Tables S3 and S4.

Among 147 metabolites significantly correlated with C₂₂ lactone sulfate in VEAPS after Bonferroni correction, 10 were positively correlated with P values <10⁻¹⁰, including six androgenic steroids, the carbohydrate 1,5-anhydroglucitol (1,5-AG), the amino acid 2,3-dihydroxy-5-methylthio-4-pentenoate, the cofactor/vitamin 2-O-methylascorbic acid, and the xenobiotic O-sulfo-L-tyrosine (Table 3).

In secondary analyses, we examined the response to vitamin E supplementation in VEAPS and ATBC of the androgenic steroids positively and significantly correlated with C₂₂ lactone sulfate in VEAPS (Table 4). Five of six androgen metabolites in Table 3 were decreased by the 400 IU ATA supplement in VEAPS (P < 0.05, although not achieving Bonferroni significance), but not in response to the 50 IU supplement in ATBC (weaker, nonsignificant reductions were suggested for androstenediol (3alpha, 17alpha) monosulfate; P values >0.05; Table 4). Four additional androgen metabolites correlated with C₂₂ lactone sulfate in VEAPS at Bonferroni significance (not presented in Table 3) were also decreased in VEAPS (P < 0.05) but not in ATBC (Table 4).

In this RCT-based serum metabolomic analysis testing the biochemical effects of supplementation with either 50 IU or 400 IU ATA daily, in addition to the anticipated significant increase in alpha-CEHC sulfate and alpha-tocopherol (and decrease in beta/gamma-tocopherol) in both trials, a novel C₂₂ lactone sulfate compound was significantly decreased only in the high-dose VEAPS trial. In addition, most of the androgenic steroid metabolites directly correlated with serum C₂₂ lactone sulfate were significantly reduced by ATA supplementation only in VEAPS. Our investigation points to a direct impact of high-dose vitamin E supplementation on this lactone-containing metabolite that correlated with androgen metabolites potentially relevant to elevated prostate cancer incidence in the SELECT controlled supplementation trial (3).

Although the molecular structure of C₂₂ lactone sulfate (Fig. 1) suggests possible involvement in the lanosterol synthase pathway, how it is produced and the precise biological pathways the molecule may interact with following high-dose ATA supplementation are unknown. Experimental evidence does support inhibitory effects of lactone-containing metabolites (e.g., sesquiterpene lactones) and lactone-based derivatives on prostate and other cancer cell line growth, mouse xenograft tumor formation, and NFĸB and STAT3 in the prostate carcinogenic src pathway (11–13). Lactone-based derivatives have also demonstrated androgen receptor antagonism in the presence of dihydrotestosterone or R1881 (13). Beyond such potential direct effects of the C₂₂ lactone on prostate cancer risk, the observed changes in several androgen metabolites, including DHEA-S, that were strongly positively correlated with the compound in VEAPS suggest an impact of high-dose vitamin E on sex steroid metabolism. For example, the decreased profile of these androgen metabolites may indicate higher concentrations of functional androgens upstream in the sex steroid pathway resulting from C₂₂ lactone sulfate alteration of specific enzymatic activity that inhibits androgen metabolism and clearance. Functional studies of this compound, with experiments that address how high-dose vitamin E might alter sex steroid metabolism, are needed to elucidate the precise biochemical actions and pathways involved.

Serum metabolite changes observed in both VEAPS and ATBC included the expected increases in alpha-tocopherol and its primary terminal metabolite alpha-CEHC sulfate, and decreases in beta/gamma-tocopherol, consistent with earlier findings (14). Alpha-CEHC sulfate is produced following serial ω-oxidation and β-oxidation reactions of alpha-tocopherol and ATA catalyzed by cytochrome p450 enzymes (notably, CYP4F2; ref. 15). Although alpha-CEHC and its sulfated metabolite have not been thoroughly studied with respect to prostate cancer risk, experimental data demonstrate anti-proliferative, anti-inflammatory, and antioxidative activity for alpha-CEHC (16–18). Circulating beta- and gamma-tocopherol are decreased by alpha-tocopherol supplementation because of the affinity-based reduction in hepatic uptake and lipoprotein transfer by alpha-tocopherol transfer protein of the former tocopherols (19). Accumulating evidence demonstrates unique antioxidant and anti-inflammatory properties for gamma-tocopherol relevant to chronic disease prevention (20, 21). However, neither the large increase in alpha-CEHC, nor the substantial decrease in beta/gamma-tocopherol, in response to ATA supplementation can explain the divergent prostate cancer findings of ATBC and SELECT given the relatively similar magnitudes of their serum changes observed here and previously.

This is likely the first investigation of metabolomic responses to supplementation with a low- and high-dose of vitamin E in two randomized, placebo-controlled clinical trials. The untargeted metabolomic platform exhibited high laboratory validity and reproducibility and identified more than 970 known metabolites reflecting a broad array of biochemicals and biological pathways. Application of the stringent Bonferroni statistical threshold for the large number of compounds still permitted discovery of several individual metabolites directly related to the vitamin E dosages of the two trials. The relatively homogenous male smoker, European ancestry population of ATBC is a limitation. In addition, based on available follow-up serum samples in ATBC, men received the trial vitamin E supplementation for up to 6 years as compared with the one-year timepoint selected in VEAPS to maintain sample size; however, our findings remained essentially unchanged when only the first three follow-up years were included for ATBC. Also, despite the large number of measured compounds, there are possibly other unmeasured metabolites or biochemical pathways related to low- and high-dose vitamin E supplementation still to be identified.

In conclusion, high-dose (400 IU/day), but not low-dose (50 IU/day), vitamin E supplementation resulted in a significant reduction in serum C₂₂ lactone sulfate that was highly correlated with alterations of androgenic steroid metabolites. Our study provides evidence of distinct steroid hormone pathway responses based on vitamin E dosages that could have direct relevance to opposing prostate cancer incidence results from two large controlled trials, ATBC and SELECT. Reexamination of the observed responses to vitamin E supplementation in other populations, and further elucidation of the interrelationships among C₂₂ lactone sulfate, androgenic steroid hormones, and prostate cancer risk are needed.

No potential conflicts of interest were disclosed.

Conception and design: J. Huang, S.J. Weinstein, D. Albanes

Development of methodology: J. Huang, J.N. Sampson

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Huang, H.N. Hodis, W.J. Mack, A.M. Mondul, D. Albanes

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Huang, S.J. Weinstein, J.N. Sampson, D. Albanes

Writing, review, and/or revision of the manuscript: J. Huang, H.N. Hodis, S.J. Weinstein, W.J. Mack, J.N. Sampson, A.M. Mondul, D. Albanes

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Huang, W.J. Mack, D. Albanes

Study supervision: J. Huang, D. Albanes

The ATBC Study is supported by the Intramural Research Program of the NCI, NIH, U.S. Public Health Service, Department of Health and Human Services. The VEAPS Study was supported by the Extramural Research Program of NIH, National Institute on Aging, R01 AG13860.

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.