Mapping the Hidden Pathological Diversity of Multiple Sclerosis
Multiple sclerosis is increasingly understood not as a single uniform disease, but as a heterogeneous disorder shaped by interacting inflammatory, neurodegenerative, genetic, and repair-related processes. The article “Identification of neuropathology-based subgroups in multiple sclerosis using a data-driven approach” addresses this complexity by asking whether post-mortem brain pathology can reveal biologically meaningful subgroups of MS. The authors emphasize that traditional clinical categories such as relapsing-remitting, primary progressive, and secondary progressive MS do not fully capture the underlying pathological diversity of the disease, particularly because relapse activity and progression can overlap in complex ways. This study therefore shifts attention from clinical labels to tissue-level disease mechanisms, using a large Netherlands Brain Bank autopsy cohort as the foundation for MS stratification.
The investigators analyzed neuropathological data from 228 MS brain donors, focusing primarily on white matter lesion characteristics. Their input variables included lesion load, reactive site load, proportions of active, mixed active/inactive, inactive, and remyelinated lesions, microglial morphology within lesions, and the presence of microglial nodules, perivascular cuffs, and cortical lesions. To make these heterogeneous data suitable for unbiased discovery, the authors applied centered log-ratio transformation to proportional lesion data, imputed missing values, and then used factor analysis of mixed data to reduce redundancy and balance continuous and categorical variables before clustering. Hierarchical clustering followed by K-means consolidation produced four neuropathology-based MS subgroups.
The resulting clusters revealed distinct patterns of white matter pathology. Cluster 1 was characterized by relatively low lesion and reactive site load, together with higher proportions of inactive and remyelinated lesions, suggesting a comparatively less aggressive pathological profile. Cluster 2 showed increased active lesions with ramified and rounded microglia and relatively more remyelination than clusters 3 and 4. Cluster 3 was distinguished by a higher number of mixed lesions with ramified and rounded microglia, while cluster 4 showed the greatest proportion of lesions containing foamy microglia and the lowest proportion of inactive lesions. These patterns suggest that microglial state, lesion activity, and remyelination capacity are central axes of neuropathological heterogeneity in MS.
The clinical validation of these subgroups was particularly informative. The authors found no significant differences among clusters in sex, age at MS onset, or conventional clinical phenotype, reinforcing the conclusion that traditional clinical descriptors are insufficient proxies for tissue-level disease biology. However, cluster 4 stood out as clinically severe: donors in this group died at a younger age, had a shorter disease duration, and reached EDSS-6 more rapidly than donors in the other clusters. Interestingly, the frequency and type of clinical signs and symptoms were broadly similar across groups; what differed was timing. Cluster 4 donors tended to experience motor, sensory, and other symptoms earlier, indicating that disease tempo rather than symptom category distinguished the most severe subgroup.
The study also connected white matter-defined subgroups to gray matter pathology and immune-cell distribution. Cluster 2 displayed a distinct cortical lesion pattern, with fewer leukocortical lesions and more intracortical lesions, suggesting that white and gray matter pathology may be linked through subgroup-specific mechanisms. In contrast, cluster 4 showed a higher proportion of donors with B cells in the brain parenchyma and perivascular space of the brainstem. This observation is biologically important because B-cell-associated inflammation has been linked to more severe MS pathology, and cluster 4 also showed features consistent with stronger immune infiltration, including more pronounced perivascular cuffing and lesions rich in foamy microglia.
The genetic component of the study provided additional support, although not definitive proof, for biological differences among the subgroups. The authors examined MS-associated variants and calculated polygenic risk scores using available genotype data. While overall MS polygenic risk scores were higher in MS donors than in controls, they did not significantly differ among the four clusters. Nevertheless, the polygenic risk score correlated significantly with the first two dimensions from the factor analysis, suggesting that the principal neuropathological axes may partly reflect genetic susceptibility architecture. The HLA-DRB1*15:01 tagging SNP rs3135388 was notably associated with clusters 2 and 4, consistent with the prominent role of the major histocompatibility complex in MS risk and immune-mediated pathology.
Overall, this article provides a compelling example of how computational neuropathology can refine the biological interpretation of MS beyond conventional clinical phenotypes. Its major strength lies in the integration of post-mortem lesion characterization, microglial morphology, clinical trajectory data, cortical pathology, B-cell distribution, and genetics. However, the authors appropriately caution that the clusters may reflect a combination of true biological subtypes, disease duration, severity gradients, treatment-era effects, and autopsy-procedure differences. Because the study is cross-sectional and based on post-mortem tissue, translation into living-patient stratification will require biomarkers that connect these neuropathological patterns to imaging, genetic, immunological, or molecular signatures. The article is also a medRxiv preprint and has not been certified by peer review, so its conclusions should be interpreted as research findings rather than clinical guidance.
Disclaimer: This blog post is based on the provided research article 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:
de Boer, A., van den Bosch, A. M., Mekkes, N. J., Fransen, N., Hoekstra, E., Smolders, J., ... & Holtman, I. R. (2023). Identification of neuropathology-based subgroups in multiple sclerosis using a data-driven approach. medRxiv, 2023-05.
