Proteomic Signals Across Demyelinating Diseases

Evidence

Two recent primary studies examined proteins in inflammatory demyelinating diseases, but they asked different questions and sampled different biological compartments. The first sought candidate diagnostic signals in plasma from people newly diagnosed with relapsing-remitting multiple sclerosis (RRMS). The second combined proteomics and immune-cell profiling to identify features associated with relapse history and phenotype in myelin oligodendrocyte glycoprotein antibody-associated disease (MOGAD), a disorder that can resemble multiple sclerosis clinically but is immunopathologically distinct.

In the RRMS study, researchers first compared five ways of preparing plasma for liquid chromatography-tandem mass spectrometry. They tested raw-plasma SP3, iST, ENRICH-iST, high-abundance-protein depletion followed by SP3 (DEPL-SP3), and extracellular-vesicle enrichment followed by SP3 (EV-SP3). The method comparison used plasma from five healthy controls. Raw-plasma SP3 quantified a mean of 391 proteins, DEPL-SP3 quantified 646, and EV-SP3 quantified 923. The deeper workflows did not simply produce interchangeable versions of the same proteome: in the five-method comparison, DEPL-SP3 uniquely quantified 152 proteins and EV-SP3 uniquely quantified 161.

The investigators then applied DEPL-SP3 and EV-SP3 to plasma from 15 women with RRMS at diagnosis, before treatment, and five sex- and age-matched healthy controls. DEPL-SP3 identified 54 regulated proteins, while EV-SP3 identified 35. Only four regulated proteins overlapped between those lists: A2M, IGKV3-15, APOA2, and HYOU1. The EV-enriched data emphasized coagulation, extracellular-matrix, and antigen-binding terms; the depleted-plasma data emphasized immune-response and complement/coagulation signals. Von Willebrand factor (VWF) was elevated in the RRMS group and proposed as a diagnostic candidate, not presented as a validated biomarker.

The MOGAD study addressed relapse-associated biology rather than initial diagnosis. It profiled samples from 67 people with MOGAD, 49 with multiple sclerosis, and 36 with Alzheimer disease. The investigators used NULISAseq CSF proteomics, Olink Explore 3072 CSF and serum proteomics, and high-dimensional mass cytometry, then integrated those measurements with longitudinal attack counts and clinical phenotypes in MOGAD. The resulting inflammatory and cardiometabolic protein profiles distinguished MOGAD from the comparison groups within the study.

Within MOGAD, relapsing disease was associated with fewer CD8-positive T cells carrying CCR7, CD31, and CTLA4 and with expansion of double-negative gamma-delta T-cell subsets. IL-13 correlated positively with relapse count and inversely with circulating regulatory T cells. IL-32 and CASP4 showed the opposite pattern, correlating negatively with relapse count and positively with regulatory-T-cell frequency. The abstract also reports that some immune-proteomic relationships differed between optic-neuritis and non-optic-neuritis presentations. These are exploratory associations, not evidence that the proteins cause relapse or can predict a future attack.

Analysis — Matching Methods to Questions

The cross-study pattern is that a “demyelinating-disease proteome” is not one fixed signal waiting to be read. This is analysis rather than an established conclusion. In the RRMS study, changing plasma preparation sharply changed which proteins were visible and which case-control differences emerged; only four regulated proteins overlapped across the two deepest workflows. In the MOGAD study, the relevant signal also depended on clinical framing: relapse history, regulatory immune-cell frequencies, and presentation phenotype were analyzed together across CSF, serum, and blood. The studies therefore converge on a design principle, not a replicated biomarker: proteomic measurements need to be matched to a defined question, biological compartment, and comparator population. They also complement rather than confirm one another. One asks whether plasma proteins separate a small RRMS group from healthy controls at diagnosis; the other asks which immune-proteomic features track relapse burden within MOGAD while comparing disease groups. Their signatures cannot be pooled or treated as a shared mechanism. A stronger next step would test prespecified protein panels prospectively, using standardized preparation, disease-mimic controls, blinded thresholds, and external cohorts.

Limitations

The RRMS cohort was very small: 15 patients and five controls, all women, from a single clinical setting. Its statistical testing identified discovery-stage differences, and the full text notes that distinguishing RRMS-specific markers from broader inflammatory or neurodegenerative signals remains difficult. The limited control group and the use of unadjusted significance for reported regulated proteins increase the risk of unstable candidates. VWF therefore requires validation in larger, independent cohorts that include relevant neurological and inflammatory comparators.

The MOGAD source was ingested at abstract depth, so this briefing does not infer unreported effect sizes, model specifications, correction procedures, or temporal prediction. Its associations link measured proteins and immune-cell subsets to accumulated relapse counts; they do not show that a baseline panel prospectively predicts relapse. Cross-sectional or retrospective associations can also reflect disease duration, prior attacks, treatment exposure, or phenotype differences. Finally, the two studies used different diseases, compartments, platforms, and endpoints. Their methodological convergence is informative, but it is not biological replication and does not establish a diagnostic or monitoring assay.