Scientists Finally Found Reliable Gut Microbiome Markers for Multiple Sclerosis

Multiple sclerosis research has a reproducibility problem when it comes to the gut. Study after study has implicated the microbiome in MS risk and progression, yet the specific bacteria named as culprits have shifted with each new paper. A new study published in Gut Microbes by Pauwels A et al. set out to fix that, applying rigorous confounder control and quantitative microbiome profiling to the largest such MS cohort to date, and the results upend several widely cited findings.

Bacteroides depletion and Blautia enrichment emerge as robust MS signals

The study enrolled 228 MS patients alongside 2,860 population controls drawn from the Flemish Gut Flora Project. Using both standard relative microbiome profiling and quantitative microbiome profiling (QMP), which counts absolute bacterial cell numbers rather than just proportions, the researchers identified 21 differentially abundant genera in MS. Three stood out as consistent across their own cohort and a pooled reanalysis of 10 previously published studies covering 1,065 MS patients and 874 controls: lowered Bacteroides, higher Blautia, and higher Methanobrevibacter. The finding is notable because it survived a cross-study validation that most individual papers never attempt.

Some well-known candidates did not hold up. Akkermansia, frequently cited as elevated in MS, showed a positive signal in relative profiling but disappeared after quantitative correction, explained by its negative correlation with fecal moisture rather than a true disease association. Roseburia, reported as decreased in several earlier studies, also lost significance once confounders were removed. The implication is uncomfortable: a substantial portion of the microbial imbalances previously linked to MS may reflect differences in stool consistency or methodology rather than genuine disease biology.

Lower bacterial cell counts and harder stools add complexity to MS profiles

Beyond individual taxa, the study found that total bacterial load was lower in relapsing-remitting MS patients compared to controls, while fecal moisture was lower in patients with primary progressive MS, meaning stiffer, drier stools. This matters because fecal moisture was the single largest contributor to gut microbiota variation in the entire dataset, accounting for 2.64% of explained variance, ahead of MS phenotype itself.

The clinical picture is further complicated by treatment. Oral disease-modifying therapies (DMTs) including dimethyl fumarate, teriflunomide, fingolimod, cladribine, and azathioprine were associated with higher fecal calprotectin, a marker of intestinal inflammation, in RRMS patients. Nine percent of RRMS patients had calprotectin levels above 200 micrograms per gram, a threshold suggestive of subclinical gut inflammation. For the roughly 2.8 million people worldwide living with MS, whose disease course and treatment responses are already difficult to predict, these overlapping signals make developing reliable biomarkers considerably harder.

Quantitative profiling and confounder adjustment as the methodological turning point

What separates this study from most predecessors is the analytical framework. Standard 16S rRNA sequencing produces relative abundances, essentially proportions that shift when any single taxon changes, creating compositional distortions. Quantitative microbiome profiling integrates absolute bacterial cell counts via flow cytometry, producing numbers that reflect actual microbial density. The team then applied generalized linear models and distance-based redundancy analysis to strip out variation attributable to fecal moisture, age, sex, BMI, overall health score, and calprotectin before looking at MS-specific signals.

Enterotype analysis added another layer. Four gut community types were identified: Ruminococcaceae, Bacteroides1, Bacteroides2, and Prevotella. The Bacteroides2 (Bact2) enterotype, characterized by low microbial load and depletion of Faecalibacterium, Coprococcus, Roseburia, and Gemmiger, was disproportionately prevalent in treated MS patients, particularly those on glatiramer acetate (53% Bact2) and teriflunomide (37.5%). This raises questions about whether certain DMTs shape gut ecology in ways that compound existing dysbiosis.

The Prevotella enterotype told a different story. Patients in this cluster had lower disease severity scores on three separate clinical measures (ARMSS, T25FW, and the MSIS physical subscore), lower BMI, better self-reported physical functioning, lower diastolic blood pressure, and higher vitamin D levels. Prevotella species are known producers of propionate, and propionic acid supplementation has shown beneficial effects in both an animal model of MS and in MS patients in previous work, suggesting this enterotype may reflect a genuinely protective gut configuration.

Serum GFAP ties neurodegeneration to microbial variation

One of the more unexpected findings involved serum glial fibrillary acidic protein (GFAP), a blood marker of astrocyte damage that is elevated in progressive MS phenotypes. GFAP accounted for 0.81% of explained microbiota variance within the MS group, comparable in magnitude to BMI. More specifically, GFAP levels inversely correlated with the abundance of Faecalibacterium and Roseburia in primary progressive MS patients, two genera associated with butyrate production and anti-inflammatory effects. Whether neurodegeneration is driving microbiome shifts, or gut dysbiosis is contributing to neurodegeneration, cannot be determined from this cross-sectional design.

The study also flagged lowered butyrate producers more broadly, including Butyricicoccus and Butyricimonas, as findings that warrant follow-up given the established role of butyrate in gut barrier integrity and immune regulation. These signals are consistent with a broader picture of gut bacteria changes associated with MS risk and progression identified in population-based research.

Open questions about causality and the path toward microbiome-based therapies

The study's cross-sectional design means causality remains out of reach. It is not possible to know from this data whether altered microbial communities precede MS onset, emerge as a consequence of disease, or arise primarily from the medications used to treat it. Dietary data were incompatible between the MS and control groups, leaving a major lifestyle confounder unresolved. Sample sizes for primary progressive MS (30 patients) and secondary progressive MS (19 patients) were too small to draw firm conclusions about those subtypes.

The authors recommend shotgun metagenomic sequencing for future work, which would provide the functional and metabolic resolution that 16S rRNA gene sequencing cannot. Identifying not just which bacteria are present but what they are doing metabolically would help clarify whether the butyrate and propionate signals translate into actionable targets. Researchers are already exploring novel microbiome-based therapies for MS including dietary interventions, prebiotics, and more targeted approaches, but those efforts depend on having stable, validated targets. By applying quantitative profiling and rigorous deconfounding to a large, well-characterized cohort, this study moves the field meaningfully closer to having them.

Evidence rating

EMERGING — 86/100

Caveats:

• Cross-sectional design: this study shows associations, not causation — it cannot determine whether gut microbiome changes drive MS or result from it.
• Funding was not disclosed by the authors, so independent verification of this limitation is not possible.
• 16S rRNA sequencing used here provides limited taxonomic and functional resolution; shotgun metagenomics would be needed to confirm metabolic implications.
• Subgroup analyses for progressive MS phenotypes (PPMS n=30, SPMS n=19) were underpowered and should be interpreted cautiously.

Large cross-sectional study (228 MS patients, 2860 controls) applying quantitative microbiome profiling and systematic confounder adjustment, including a comparative reanalysis of 10 prior published datasets. Findings are associative, not causal, and the study design cannot establish whether microbiome differences precede or follow MS onset. Results for progressive MS subtypes are limited by small subgroup sizes and have not yet been independently replicated using the same quantitative methodology.