Thirty Years, No Signature: What the MS Metabolomics Literature Adds Up To
Zahoor and colleagues set out to survey metabolomics in multiple sclerosis (MS) across every biological matrix people have tried: cerebrospinal fluid, blood, urine, tears, brain tissue, and the animal model. Their tables list roughly 35 patient studies and 13 preclinical ones, each with its analytical platform, sample size, and headline metabolites. Read as a list of findings, it is a useful reference. Read as a whole, it demonstrates something the individual papers cannot show on their own, and the review states it plainly in its conclusion: there has been some overlap between the metabolite profiles reported so far, but a reproducible MS-specific metabolomic signature has not been identified. Putting three decades of studies side by side is what makes that visible, and it is more valuable than any single row in the tables.
The Same Molecule, the Same Fluid, Opposite Directions
The disagreements are not subtle and they are not between different tissues. Lactate in cerebrospinal fluid was reported as raised in patients with active lesions, correlating with the number of plaques in the brain, and separately as reduced alongside glutamine in MS patients against controls. The review does not smooth this over, stating that findings from these studies need validating in large patient and control cohorts before any conclusions are drawn. Choline and myo-inositol run the same way. An 800 MHz proton NMR study of 15 patients found both raised in cerebrospinal fluid; a later study of 19 patients found both reduced in the same fluid. Acetate rose in a 1990s cerebrospinal fluid study and fell in a 2019 one. Even the question of whether relapsing and progressive patients differ metabolically got opposite answers, with one 19-patient study finding no difference between the two forms and later work building whole classifiers on that distinction.
The Review Names the Reasons, and the Tables Show One More
The causes listed are specific: disease heterogeneity, a lack of stringent inclusion criteria, batch effects from diverse sample types, variable sample size, heterogeneity in study populations, and differences in sample handling, preparation, and data analysis. The tables add a further one that the text mentions only in passing. Many of the listed studies carry fewer than 30 patients per group, several fewer than 15, and only a handful pass 100 patients, the largest being an HPLC study of 518 patients against 167 controls. The review also observes that most early work used a single analytical platform and suggests this is part of why only limited alterations turned up, since no one method detects the whole metabolome. Its recommendation follows from that: combine untargeted discovery, which casts wide but leaves compounds unidentified for lack of standards, with targeted validation, which quantifies precisely but only about 20 metabolites at a time.
The Animal Model Shares One Metabolite With Patients
The most quotable number in the review concerns the experimental autoimmune encephalomyelitis model. Singh and colleagues profiled urine and plasma from mice with chronic disease, then compared the metabolites they found altered against published human MS metabolome data. Across the two, the only metabolite commonly disrupted was creatinine. The review draws the conclusion directly, calling this a limitation of using the model to study the metabolome in the context of MS. That single comparison deserves weight, because a large share of the field's mechanistic and therapeutic work runs through this model, and the review devotes a second table to 13 studies built on it. The review is careful elsewhere to say that none of the animal models reproduces the exact pathophysiology of human disease, and that findings from them need confirming in patients.
Where the Approach Has Actually Produced a Drug Candidate
The therapeutics section is the strongest argument the review makes for its own subject. Poisson and colleagues measured 324 metabolites in plasma from mice with chronic disease, found 100 altered, mostly lipids, and linked the most perturbed pathways to polyunsaturated fatty acid metabolism, specifically the alpha-linolenic and linoleic acid routes. They then took resolvin D1, a metabolite downstream of that pathway, gave it orally at 100 ng per mouse daily, and saw disease progression reduced, with autoreactive T cells inhibited and myeloid-derived suppressor cells, regulatory T cells, and anti-inflammatory microglia induced. That is the full arc the field promises: a signature narrowed to a pathway, and a molecule from that pathway tested as treatment. Bile acids followed the same route. Reduced bile acid metabolites were found in patients across discovery, validation, and paediatric cohorts, more pronounced in progressive disease; the receptors FXR and GPBAR1 turned out to be expressed on glial and immune cells in white matter lesions; tauroursodeoxycholic acid reduced disease in mice through GPBAR1; and a clinical trial is now registered.
Reading Treatment Response, and One Caution About Vitamin D
A second use is monitoring rather than discovery. Dimethyl fumarate shifted the plasma lipid profile of 18 patients over six months, raising phospholipids, lysophospholipids, and plasmalogens while lowering circulating free fatty acids, with those changes correlating with lymphocyte counts and particularly with CD8+ T cells. An interferon-β study separated responders from non-responders by metabolic profile using absence of disease activity as the endpoint, though it rested on 16 responders and 5 non-responders, and the review says outright that the sample was too small and the study needs repeating. The vitamin D work carries a caution the review states rather than buries: supplementation does measurably alter the metabolome, but clinical trials of vitamin D in MS have been unsuccessful, which the review says brings the therapeutic potential of the supplement into question. It also lists what those supplementation studies left out, including progressive patients, treatment data, meal timing, and genotype.
What Would Have to Change, and What This Review Cannot Settle
The recommendations are concrete enough to act on: better analytical platforms, extraction and recovery methods, consistency in study design including sample size and control of confounders such as diet and drugs, running multiple platforms rather than one, and integrating metabolomics with other omics data and with gut microbiome profiling. Two constraints of the review itself are worth stating. It is narrative rather than systematic, with no search strategy, inclusion criteria, or quality assessment described, so the tables reflect the authors' selection rather than a defined sample of the literature. And it reports findings as directions of change without effect sizes or confidence intervals, so a reader cannot tell from the tables whether two disagreeing studies genuinely conflict or merely differ in the sign of a small and imprecise estimate. That distinction matters for every contradiction listed above. The step the review implies without quite specifying is the one the field now needs: before another discovery cohort, take the metabolites that recur most often across these tables and measure them on one standardised platform, in a single large multi-site cohort, with diet, meal timing, and treatment recorded.
Disclaimer: This blog post is based on the cited review 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:
Zahoor, I., Rui, B., Khan, J., Datta, I., & Giri, S. (2021). An emerging potential of metabolomics in multiple sclerosis: a comprehensive overview. Cellular and Molecular Life Sciences, 78(7), 3181–3203. https://doi.org/10.1007/s00018-020-03733-2
