MS Risk Loci Through Single Cells
A genome-wide association study returns positions on a chromosome, not cells. For most of the history of multiple sclerosis (MS) genetics the step from a risk locus to a mechanism went through bulk tissue, which averages whatever mixture of cell types happened to be in the sample and therefore tells you least about the rare population that may be doing the work. Single-cell sequencing changes what a locus can be asked to explain, and three recent papers take three different routes through that change. Fujimoto and colleagues assign MS risk to cell types and to positions inside a lesion. King and colleagues follow blood cells over time and find a pre-relapse programme sitting on MS risk loci. Jones and colleagues split single-cell expression by sex and find regulatory variants that only appear once you do.
Adding an Underrepresented Population to MS Genetics
Fujimoto and colleagues start with a Japanese GWAS of 688 cases against 205,199 controls. That case count is small by current standards, and the paper's value does not rest on it: what a distinct ancestry buys is different allele frequencies and different linkage structure, so associations that are invisible or unresolvable in European data can come through. They report associations in the major histocompatibility complex region and a population-specific variant at 11q24. The Japanese data were then folded into a cross-population meta-analysis of 29,374 cases and 1,843,563 controls over four ancestries, which yielded 22 novel loci. Loci found only when a fourth ancestry joins the analysis are a reminder that the European-dominated MS risk map has been incomplete rather than finished, and that incompleteness is structural rather than a matter of sample size alone.
Which Cells Carry the Risk, and Where in the Lesion
The genetic signal was then read against single-cell and single-nucleus RNA sequencing of peripheral blood mononuclear cells and of subcortical lesion tissue. Risk enrichment fell on the CD4 T helper lineage, on regulatory T cells, and on endothelial cells. The first two are expected company for an autoimmune disease; the endothelial signal is the one that moves the analysis out of the immune compartment, since endothelium is the barrier that circulating cells have to negotiate rather than a lymphocyte subset. The spatial transcriptomics is the part that is hard to get any other way: within subcortical lesions, MS genetic risk associations showed spatial and temporal heterogeneity, meaning the enrichment is not a property of the lesion as a whole but of particular regions and particular stages. A risk variant that matters at a lesion edge and not at its centre, or early and not late, is a variant whose effect bulk tissue would average away.
A Blood Signature Three Months Before Relapse
King and colleagues built a time-resolved atlas instead of a snapshot, sampling blood longitudinally from patients with relapsing-remitting MS and combining single-cell RNA sequencing with bulk transcriptomics, multiparameter flow cytometry and targeted viral RT-qPCR. The finding is a reproducible transcriptional signature in monocytes and B cells appearing up to three months before clinical onset of a relapse. Three months is a long lead time for an event that is usually treated as abrupt, and the signature was enriched for host genes responsive to Epstein-Barr virus lytic reactivation factors. Two independent assays were brought to bear on that reading rather than leaving it at the transcriptional level: RT-qPCR confirmed raised EBV LMP-1 transcripts in pre-relapse B cells, and flow cytometry showed expansion of CD11c-positive atypical B cells displaying gp350, a viral envelope protein. A transcriptional module can be read many ways; a viral transcript and a viral protein on a particular B cell subset narrow the options.
Inherited Risk and Viral Programmes on Shared Regulatory Elements
The genetic observation in that paper is the one most relevant to anyone working on MS risk loci. The pre-relapse transcriptional modules overlapped both MS GWAS risk loci and enhancers bound by EBNA-2, an EBV protein that reprogrammes host transcription. Two things that have been studied as separate risk factors for decades, inherited susceptibility and viral exposure, turn out to converge on the same regulatory elements, which suggests the genotype may set how strongly a cell responds to the virus rather than acting alongside it. The authors are careful about the boundary of the claim, stating that how this peripheral activation relates to central nervous system lesion formation remains to be established. That is the honest position: everything measured here is in blood, and the lesion is where relapse is defined.
Sex-Biased Regulatory Variants at MS Loci
Jones and colleagues come at the problem from the direction of a variable MS genetics cannot ignore, given the female bias in incidence. They performed single-cell RNA sequencing on 480 lymphoblastoid cell lines and found roughly 1,200 sex-biased genes, 71 per cent of which are sex-biased in at least one GTEx tissue, with a 21-gene core shared across all datasets examined. The mechanism is not what a chromosome count would predict: only 7.7 per cent of the sex bias was explained by sex-chromosome number, while 79 per cent of the genes were targets of sex-biased transcription factors, among them FOSL1, ZNF730, ZFX and ZNF726, which are themselves regulated by X-chromosome number. They then mapped 2,390 sex-biased expression quantitative trait loci, and these were enriched across more than 100 GWAS phenotypes including female-biased autoimmune diseases such as MS. The confirmation step is where the paper states its own weakness: agreement in an independent dataset was modest, and the support rests on permutation showing that sex-biased eQTLs defined with real sex labels were more likely to have concordant direction than those from permuted labels.
Three Resolutions, Three Different Gaps
Read together, the three papers add resolution along three separate axes, and each leaves a different thing unfinished. Fujimoto and colleagues gain ancestry and anatomy but read lesion tissue at whatever stage it was sampled, so the temporal heterogeneity they report is inferred across specimens rather than followed in one. King and colleagues gain time and a named viral mechanism but work entirely in peripheral blood, and say so. Jones and colleagues gain sex as an explicit variable but work in cell lines, with independent support that they describe as modest. What the three share is a change in what a risk locus is expected to deliver: not a gene name, but a cell type, a position, a moment, and now a sex in which the variant actually does something.
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:
Fujimoto, R., Ogawa, K., Namba, S., ... Okuno, T., Isobe, N., & Okada, Y. (2026). Multiancestry genome-wide association and multiomics analyses elucidate spatiocellular features of multiple sclerosis genetics. Nature Genetics, 58(9), 2192–2200. https://doi.org/10.1038/s41588-026-02741-5
King, D. A., Saxena, S., Caefer, D., Downer, K. C., Saucier, L. E., Goodman, E., Aracena, J., Zhirova, A., Alchanat, A., Nolan, S., Lokhande, H., Weiner, H. L., Gewurz, B. E., & Chitnis, T. (2026). EBV reactivation priming of the peripheral immune system in multiple sclerosis relapse. Nature Medicine. https://doi.org/10.1038/s41591-026-04665-3
Jones, A. G., Dalapati, T., Connelly, G. G., Wang, L., Schott, B. H., San Roman, A. K., & Ko, D. C. (2026). Transcriptional mechanisms of sex-biased gene expression and their connections to disease-associated variation. Human Molecular Genetics, 35(6). https://doi.org/10.1093/hmg/ddag019
