One protein fragment in spinal fluid faded step by step as multiple sclerosis progressed. In the disease that most often gets mistaken for MS, that fragment was not the story at all. Almost every fragment was lower.
The finding comes from a study published online October 5 in Proteomics ↗ by a group at the University of Cagliari, with Cristina Contini as first author and Eleonora Cocco and Barbara Manconi as joint senior authors. They read the small leftover peptides floating in cerebrospinal fluid, the clear liquid drawn by spinal tap that bathes the brain and spinal cord, from untreated patients across the demyelinating-disease spectrum and from people with other, non-inflammatory neurological problems as a comparison.
Reading the fragments whole
Most protein analysis works by chopping every protein into standard pieces with an enzyme first, then reading the pieces. That is efficient, but it throws away the part of the story this study cared about: how the body itself cut the protein, and what chemical tags it left behind. The Cagliari team used a top-down approach instead, reading the naturally occurring peptides exactly as the cerebrospinal fluid delivered them, cuts and tags intact.
That matters because the brain and its immune cells are constantly trimming proteins into shorter peptides, and the pattern of those cuts is a readout of what the tissue is doing. A standard digestion workflow erases it. Reading the fragments whole keeps it.
They catalogued 381 of these endogenous peptides, many never reported before, along with their oxidative modifications, the small chemical marks that get stamped onto a peptide when it is exposed to inflammatory stress.
Two diseases, two opposite signatures
The clearest single marker was a newly described tail-end fragment of a protein called secretogranin-5, part of the machinery neurons use to package and release signaling molecules. That fragment dropped in a stepwise way across the disease course, highest in clinically isolated syndrome (the earliest hint of MS, often a single first attack), lower in established relapsing-remitting multiple sclerosis. When the authors fed their data to a machine-learning classifier, that one fragment was the single most useful feature for sorting patients.
Neuromyelitis optica spectrum disorder, or NMOSD, looked nothing like that. NMOSD is a separate autoimmune attack on the spinal cord and optic nerves that mimics MS on scans and symptoms but is driven in most patients by an antibody against aquaporin-4 (AQP4), a water channel on the support cells of the brain, and, critically, is made worse by some of the drugs used to treat MS. Telling the two apart early is a real clinical problem. In this data they separated by opposite behavior: where MS showed a specific fragment fading against a background of oxidative marks, NMOSD showed broad peptide depletion and fewer oxidized species across the board. The disease-associated signal ran through fragments of granins, osteopontin, proSAAS, and fibrinogen, the oxidized versions most of all.
Why this is worth watching
Spinal fluid is already the fluid clinicians test when MS is on the table. The standard test looks for oligoclonal bands, a sign of antibody production inside the central nervous system, and this study stratified its MS patients by that same band status. What it adds is a different molecular layer underneath: not whether antibodies are being made, but how the tissue is cutting and chemically marking its own proteins, and how that pattern diverges between two diseases that otherwise resist separation.
The usual cautions apply, and the authors raise them first. This is an exploratory, cross-sectional snapshot, not a longitudinal track of individual patients, and the peptide candidates need validation in independent cohorts before anyone builds a test on them. A fragment that sorts patients in one dataset has to repeat in another to mean anything.
What makes it a peptide story rather than a protein story is the resolution. The signal did not live in whole proteins going up or down. It lived in which fragments the tissue produced and how oxidized they were, the kind of endogenous-peptide detail that only survives if you read the molecules whole. On peptidemodel these diseases sit under the neuroprotective target ↗ and, for the autoimmune side, the immune target ↗. The fragments themselves are not catalog peptides. They are the body's own, and for now they are a signature to be confirmed, not a drug.