CNS Resilience in Multiple Sclerosis: Insights from Single-Cell Data
Single-cell sequencing changed the question the multiple sclerosis (MS) field asks about resilience. Instead of asking whether a gene protects tissue, it became possible to ask which cell is holding out and what that particular cell is expressing while its neighbours fail. Four papers from 2025 and 2026 work at that resolution, none of them through the autoimmune animal model, and across all four the answer comes back indexed rather than general. One gene is protective in one immune lineage and risk-raising in another. One transcription factor programme shifts in microglia, and shifts most at one zone of the lesion rather than across the white matter. One ion channel decides which axons keep their myelin, sorted by fibre calibre. And the fourth paper asks the question underneath the other three, which is whether the transcript you counted was assigned to the right cell at all.
One Gene, Two Directions, Depending on the Lineage
Zhou and colleagues built a single-cell genetic framework rather than a tissue one. They took cis-eQTL data from 14 immune cell types, paired it with genome-wide association data for MS and for an Epstein-Barr virus-related clinical phenotype, and ran Mendelian randomization inside each cell type with colocalization alongside. That produced 43 candidate eGenes with convergent genetic support, 6 for the viral phenotype and 37 for MS. For the viral phenotype, SERPINB1 in natural killer cells came out risk-raising and HLA-G protective. For MS, APOM and MSH5 were protective. The result that matters for resilience is AHI1, which showed cell-type-dependent and bidirectional associations across immune lineages. The same gene, instrumented the same way, points in opposite directions depending on which cell its expression was measured in.
What That Framework Settles and What It Does Not
The paper is careful about the weight those 43 genes can carry, and reporting that carefully is part of reading it. Candidate eGenes were evaluated in independent cohorts, and colocalization supported the findings. For instruments built from multiple variants they ran heterogeneity tests, MR-Egger, weighted median, mode-based estimators and MR-PRESSO. Among the FDR-significant multi-variant associations, MR-Egger intercept tests gave no sign of directional pleiotropy, although a small subset did show heterogeneity or MR-PRESSO signals, which the authors state rather than bury. Phenome-wide analysis found no significant adverse phenotypic associations among the genes that could be evaluated at their prespecified threshold. Querying DrugBank then nominated sodium nitroprusside, fasudil, artenimol and choline as compounds hitting prioritised genes, and the authors frame these explicitly as hypothesis-generating rather than as candidates ready for trial.
A Transcription Factor Programme That Moves in Microglia
Ma and colleagues went at human MS white matter with single-nucleus sequencing and asked which regulatory programmes had shifted, using gene regulatory network analysis alongside subpopulation grouping, functional enrichment and pseudotime. One module, M3, carried transcription factor dysregulation in MS white matter relative to controls, its genes expressed predominantly in microglia and enriched in inflammation-related signalling. The candidate regulator connected to those changes was IKZF1, the IKAROS family zinc finger. The microglia themselves were not two populations but a continuum, under dynamic regulation from homeostatic towards pro-inflammatory phenotypes, with the latter end carrying disease-associated microglia characteristics. The authors are explicit that this heterogeneity extends beyond the traditional polarization models, which is a statement about the vocabulary the field has been using as much as about the cells.
Where in the Lesion That Programme Sits
The spatial half of the same paper is what makes the microglial result addressable. Spatial transcriptomics with deconvolution showed strong heterogeneity within MS white matter and distinct functions by niche, rather than a uniformly inflamed tissue. The lesion rim, a transition zone between lesion and surrounding white matter, correlated strongly with microglia and was heavily enriched for disease-associated microglia characteristics, along with altered IKZF1 regulon activity, which is the same signal the single-nucleus arm produced from dissociated tissue. The authors read the rim as an inflammatory hotspot, and offer sustained microglial activation there as a reason remyelination is hindered at that particular zone. Resilience in this reading is not a property of a cell type but of a cell type at a tissue position, and the rim is where the position is worst.
A Channel That Decides Which Axons Keep Their Myelin
Dyckow-Schubart and colleagues used single-nucleus sequencing to put Piezo2, a mechanosensitive ion channel better known from touch and pain in the peripheral nervous system, in late differentiated oligodendrocytes of the optic nerve, with minor expression in retinal ganglion cells. Deleting it in oligodendrocytes produced age-dependent motor impairment and a selective defect: myelin compaction was disrupted in small-calibre optic nerve axons specifically, the fibre population already known to be particularly vulnerable in demyelinating disease. Differential expression analysis put the channel's job in mature oligodendrocytes at myelin compaction and white matter integrity. The human arm is what ties it to MS, with oligodendrocyte-encoded PIEZO2 expression reduced in optic nerve lesion areas from MS patients, which the authors read as a convergent mechanism behind small-calibre fibre vulnerability.
Which Cell Did That Transcript Belong To
Kotah and colleagues, working from an MS centre, examined the step every claim above rests on. Running in situ sequencing on formalin-fixed paraffin-embedded postmortem human brain, they compared ways of drawing cell boundaries and assigning transcripts to them, using nuclear borders against expanded nuclear borders. The broad cell-type proportions held up across the methods, which is reassuring. The specific cell types did not: microglia, neurons and neurovascular cells were each affected by how transcripts were assigned, in tissue they describe as morphologically complex for exactly this reason. Their fix was to bring in fluorescent imaging of 18S RNA and IBA1 protein to steer transcripts toward RNA-rich cells and toward microglia respectively, with a second pass for transcripts falling outside the imaging masks, which raised both the number of microglia detected and the specificity of the transcripts assigned to them. The cell types most sensitive to that choice are the ones the other three papers are making claims about.
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:
Zhou, Y., Huang, X., She, X., Hao, Q., Mi, Z., & Yang, Y. (2026). Mapping the immune-genetic architecture of Epstein-Barr virus-related phenotypes and multiple sclerosis through a single-cell genetic framework for target prioritization and pharmacologic hypothesis generation. Multiple Sclerosis and Related Disorders, 114, 107410. https://doi.org/10.1016/j.msard.2026.107410
Ma, K., Zhang, Q., Du, H., Peng, H., Kang, J., Cai, Y., Tan, L., Chen, X., Bian, C., & Qu, X. (2026). Microglial dysregulation and spatiotemporal dynamics of inflammation in multiple sclerosis white matter: an integrative transcriptomic analysis. Journal of Translational Medicine, 24(1). https://doi.org/10.1186/s12967-026-08360-2
Dyckow-Schubart, J., Rabitsch, A. M., Geywitz, C., et al., Möbius, W., & Schirmer, L. (2026). Oligodendrocyte Piezo2 is a regulator of age-dependent myelin integrity and dysregulated in multiple sclerosis. Communications Biology, 9(1). https://doi.org/10.1038/s42003-026-10530-3
Kotah, J. M., Rust, T., van Weering, H. R. J., Bosma, J., Woudstra, A. L., Kooistra, S. M., & Eggen, B. J. L. (2025). Beyond the nuclear border: single-cell analysis of in situ sequenced human brain tissue using cellular features. Communications Biology, 8(1), 1089. https://doi.org/10.1038/s42003-025-08518-6
