MS Single-Cell Data by Perturbing It

Describing cell states in multiple sclerosis (MS) is no longer the hard part. Enough single-cell atlases of patient blood, cerebrospinal fluid and postmortem brain exist that the field has started doing something else with them: giving a patient a drug that removes one cell type, then watching which other states move. Four human studies from 2026 work that way, none through the autoimmune animal model, and they sit on a ladder that is worth climbing in order. One uses an approved therapy as a probe in a longitudinal cohort with independent confirmation. One does the most aggressive depletion available in two patients. One reanalyses public data to ask about a pathway nobody had checked. And one maps lesion stages in postmortem cortex and says plainly that part of it rests on a single lesion.

Using an Approved Drug as a Probe
Farooq and colleagues treated anti-CD20 therapy as a perturbation rather than as a treatment to be audited. They profiled patients longitudinally as they started ocrelizumab, high-dimensional and multimodal, and ran the data through a machine learning pipeline built to resolve shifts in continuous immune cell states rather than in discretely partitioned cell types. That design choice is the point: what moved were shared biological processes that cut across cluster boundaries, which a clustering-first analysis would have split across several labels. What they pulled out was a single distinct state, a chronically activated proinflammatory-cytotoxic T cell state with CNS-homing properties, and it was selectively depleted after B cells were removed. A T cell state that disappears when you take away B cells is a dependency, and the design is what made it visible.

Where Else That State Turns Up
A state found in one cohort under one drug could be an artefact of either, so the paper goes looking for it elsewhere. Across independent datasets the same state was enriched in the circulation of patients with clinically aggressive relapsing disease, in expanded clonotypes enriched in cerebrospinal fluid, including a subset with proven Epstein-Barr virus specificity, and in chronically inflamed lesion rims in end-stage MS. That is the same anatomical zone other groups keep landing on. The last experiment runs the logic backwards with a different drug: blocking lymphocyte movement across the blood-brain barrier with the anti-integrin natalizumab enriched this state in the circulation, which is what should happen if the cells are being held back from a destination they were heading for. Two therapies with different mechanisms, one concordant effect.

Two Patients, One Compartment Cleared
Gupta and colleagues pushed the same depletion logic further in a preprint reporting two participants with treatment-refractory progressive MS. The rationale is a known failure: CNS-resident B cells resist peripheral depletion by anti-CD20 antibodies, so they used CD19-targeted CAR-T cells, 0.33 x 10⁸ autologous cells of mivocabtagene autoleucel after preconditioning. CAR-T cells expanded in the cerebrospinal fluid by day 14 in both participants. What followed is the result worth noting: sustained disappearance of the disease-associated oligoclonal band in one and normalization of the IgG index in the other, and both held despite peripheral naive B cells coming back. Single-cell myeloid and proteomic signatures showed proinflammatory profiles declining, anti-Epstein-Barr virus seropositivity fell, and disability scores and MRI were stable at week 48. Two patients, reported as supporting continued investigation.

What a Reanalysis of Public Data Can Add
Fritzen went at an entirely different question with no new samples at all, reanalysing two publicly available human single-cell datasets covering cerebrospinal fluid and peripheral blood B cells from controls and from patients with clinically isolated syndrome or MS. The target was zinc handling, which regulates immune cell function and had not been examined in the B cell populations that matter in MS. A composite zinc transporter and metallothionein signature was raised in MS relative to controls in both compartments and in both CXCR3-positive and CXCR3-negative B cells, which makes it a feature of B cells broadly rather than of the intrathecally active subset alone. In cerebrospinal fluid CXCR3-positive cells the rise was driven by individual transporters including SLC39A11, SLC30A9 and SLC39A6, while in blood no single gene reached significance and only the composite did.

Lesion Stages, and a Stress Response in the Quietest Ones
Gasque-Rubio and colleagues ran spatial transcriptomics on postmortem cortical tissue, sorting grey matter lesions into early active, chronic active and chronic inactive by myelin integrity and macrophage and microglial activity. The stages came out transcriptionally distinct. The early active lesion carried raised immune gene expression with inflammatory and matrix remodelling pathways. Chronic active lesions carried innate immune signalling, including Toll-like receptor and complement pathways. Chronic inactive lesions carried something different in kind: raised heat shock proteins HSPA1A and HSPA1B and stress-response pathways, in the lesions with the least ongoing immune activity. Control samples kept their neuronal and myelin gene expression. The authors label the whole thing exploratory on sample size grounds, and state that the early active group is represented by a single lesion, which is the kind of disclosure that makes the rest readable.

Four Rungs on the Same Ladder
The four designs license different things, and each paper says so in its own words. Farooq and colleagues have a longitudinal perturbation plus the same state recovered in independent datasets and a second drug producing a concordant shift, which is about as much as human immunology gets without a knockout. Gupta and colleagues have the strongest intervention and two patients, and frame the work as supporting continued investigation rather than as evidence of efficacy. Fritzen has a pathway-level finding from public data, with the composite signature significant where individual genes were not, and calls for further investigation of zinc-regulatory mechanisms. Gasque-Rubio and colleagues have preliminary stage profiles and candidate biomarkers awaiting validation, with one lesion carrying a key group on its own. Read in that order, the heat shock signal in quiet lesions and the zinc signal in B cells are both hypotheses; the B cell-dependent T cell state is a finding.

Disclaimer: This blog post is based on the cited studies 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.

References:
Farooq, R., Cutler, B., Pisa, M., Balaram Kuttikkatte, S., Palace, J., DeLuca, G. C., Gold, R., Theis, F. J., Fugger, L., & Attfield, K. E. (2026). Multimodal discovery of a pathogenic B cell-dependent T cell state in multiple sclerosis. Brain. https://doi.org/10.1093/brain/awag303
Gupta, S., Seshadri, M. R., Dandekar, R., et al., Wilson, M. R., & Cree, B. A. C. (2026). Immune Profiling after treatment with an Anti-CD19 Chimeric Antigen Receptor T Cell Therapy in Treatment-Refractory Progressive Multiple Sclerosis. medRxiv preprint. https://doi.org/10.64898/2026.09.11.26361992
Fritzen, R. (2026). Zinc and metallothionein dysregulation in B cells in multiple sclerosis, with disease-specific CXCR3+/CXCR3- divergence in blood. Frontiers in Immunology, 17, 1922905. https://doi.org/10.3389/fimmu.2026.1922905
Gasque-Rubio, R., Cubas-Nuñez, L., Castillo-Villalba, J., et al., Pérez-Miralles, F., & Casanova, B. (2026). Stage-Specific Gene Expression Profiles in Multiple Sclerosis Cortical Lesions Identified via Spatial Transcriptomics. OMICS: A Journal of Integrative Biology. https://doi.org/10.1177/15578100261479076