Same Dose, Different Response: Vitamin D and the MS Metabolome

Bhargava and colleagues put two questions in one paper, and the second is the one worth the read. First, does the plasma metabolome of multiple sclerosis (MS) patients differ from healthy people. Second, when both groups take the same vitamin D dose, does their metabolism respond the same way. The cross-sectional cohort was 27 patients with relapsing-remitting MS and 27 controls matched to them on age and sex, 96% of both groups of European descent, recruited at Johns Hopkins, with early and mild disease: median duration two years, median disability score of 1.0, and only a third on treatment. The supplementation cohort was 24 patients and 27 controls, all given 5,000 IU of cholecalciferol daily for 90 days (NCT01667796). Profiling ran on a commercial platform combining liquid and gas chromatography with mass spectrometry, yielding 492 and 529 compounds through quality control. One design choice deserves credit before the results: with 27 people per group against roughly 500 metabolites, the authors analysed at the level of metabolite modules and pathway sets rather than testing hundreds of molecules one at a time, which is the right response to that ratio.

What Separates Patients From Controls
Forty-three metabolites differed after adjusting for age and sex, with an average standardised difference of 0.62 ± 0.20. The discriminant model separated the two groups with a predictive ability of 0.56, and permutation testing on 1,000 relabelled datasets gave that a p of 0.001. Its goodness of fit was 0.95, but permutation put that at p = 0.047, which is the weaker half of the result and worth carrying alongside the stronger half. The metabolites driving the separation cluster on redox chemistry. 5-Oxoproline led the importance ranking at 2.22, followed by phenylpyruvate, N-acetylaspartate, prolylglycine, and ornithine, with S-adenosylhomocysteine, γ-glutamyl valine, and cysteine glutathione disulfide close behind. Network analysis sorted the 492 metabolites into 15 modules, of which two differed between groups, both landing at p = 0.05: a green module of γ-glutamyl amino acids and oxidative stress metabolites, and a brown module of benzoate and xanthine metabolites.

One Module Is Redox Chemistry, the Other Is Diet
The green module is the biologically coherent one. γ-Glutamyl amino acids are intermediates of the γ-glutamyl cycle, which recycles glutathione, so a shift across the whole group of them points at disturbed redox handling rather than at any single molecule. Individually, γ-glutamyl leucine (p = 0.002), γ-glutamyl valine (p = 0.003), leucine and valine (both p = 0.002), and 2-aminoadipate (p = 0.0009) differed, although γ-glutamyl isoleucine did not (p = 0.16) despite sitting centrally in the module. The brown module is a different kind of finding. Benzoate, produced by bacterial metabolism of phenylalanine and polyphenols, was reduced in patients, and caffeine (p = 0.02) and paraxanthine (p = 0.03) also differed. These are dietary and microbial in origin, and the study recorded neither diet, time of blood draw, nor time of last meal, which the authors list among their limitations. A caffeine difference between two groups of people is as readily an observation about coffee as about disease.

The Supplementation Result Is the Real Finding
Everyone in the second cohort took the same 5,000 IU daily for 90 days, so the comparison is not vitamin D against nothing but how two groups respond to the same intervention. Two modules changed differently between patients and controls: the green module again, carrying γ-glutamyl amino acids and oxidative stress metabolites, with an interaction p of 0.006, and a red module of lysolipids and fatty acids at p = 0.03. The direction is the point. In healthy controls the green module fell after supplementation. In MS patients it stayed flat. Vitamin D lowered markers of oxidative stress in controls and did not do so in patients. Within that module, γ-glutamyl histidine (p = 1.86E-05), γ-glutamyl glycine (p = 0.0001), γ-glutamyl valine (p = 0.0007), and γ-glutamyl isoleucine (p = 0.002) carried most of the effect. Changes in the two modules were uncorrelated (r = 0.03, p = 0.92), so the lipid and redox responses are independent rather than one driving the other.

It Is Not Simply That Patients Reached Lower Levels
The obvious explanation is dose. Patients did finish lower than controls on identical supplementation, at 47.8 ng/ml against 54.9, from near-identical baselines of 21.5 and 22.3, which matches this group's earlier report that MS patients raise 25-hydroxyvitamin D less for a given oral dose. That explanation does not survive the analysis, because the models included the change in 25-hydroxyvitamin D as a covariate alongside age and body mass index. The blunted metabolic response remains after adjusting for how far the level actually rose. The authors read this as resistance to the metabolic effects of vitamin D in MS, and list candidate explanations they are careful to mark as unconfirmed: altered vitamin D metabolism or signalling, vitamin D receptor genotype, interactions between MS risk variants and ongoing inflammation, or an effect of inflammation on vitamin D metabolism itself.

What the Design Supports and What It Does Not
Three constraints deserve stating precisely. The cross-sectional comparison cannot tell whether the metabolic differences contribute to MS or follow from it, which the authors acknowledge. The supplementation study has no untreated arm, since everyone received vitamin D, so the statement that oxidative stress markers fell in controls is a before-and-after observation across 90 days with nothing to exclude a seasonal or other time-related effect. The comparative finding is protected from that problem, because a pure time effect would move both groups and produce no difference between them, but the single-group claim is the weaker of the two and should not be quoted as if it were the stronger. Third, reliability: blinded repeat samples gave a median intraclass correlation of 0.77 with a lower quartile of 0.53, estimated from six pairs. A quarter of these metabolites were measured with modest reproducibility, and six pairs is a thin basis for knowing which quarter.

Who These Numbers Apply To
Both cohorts are narrow, and narrow in different ways, which matters for anyone tempted to generalise. The cross-sectional patients had a median disease duration of two years and a median disability score of 1.0, with a third on treatment; the supplementation cohort was 88% treated with a median score of 1.5. Neither included progressive patients, so nothing here speaks to the phase of disease where treatment options are thinnest. The supplementation study enrolled only women of European descent, aged 18 to 60, with body mass index between 18 and 30 and baseline vitamin D at or below 30 ng/ml. Most patients across the study were untreated or taking glatiramer acetate, which the authors flag as a real problem for applying these results to a population now largely on disease-modifying therapy. The follow-up they name is genotype. Trials in other populations have found the effect of supplementation on oxidative stress and inflammation to depend on vitamin D receptor genotype, and testing whether that explains the blunted response here would turn a described phenomenon into a mechanism, and would begin to answer which patients, if any, should be given vitamin D at all.

Disclaimer: This blog post is based on the cited research article and is intended for informational purposes only. It is not intended to provide medical advice. Please consult with a healthcare professional for any health concerns.

Reference:
Bhargava, P., Fitzgerald, K. C., Calabresi, P. A., & Mowry, E. M. (2017). Metabolic alterations in multiple sclerosis and the impact of vitamin D supplementation. JCI Insight, 2(19), e95302. https://doi.org/10.1172/jci.insight.95302