Genetic Determinants of Multiple Sclerosis Severity: Polygenic Risk, CNS Resilience, and Prognostic Machine Learning
Multiple sclerosis is a complex neurological disease in which inflammatory demyelination, neuroaxonal injury, and repair failure interact over many years. Although more than 230 common genetic variants have been linked to susceptibility to multiple sclerosis, this article asks a different and clinically more urgent question: why do some patients accumulate disability rapidly while others remain comparatively stable? The authors emphasize that genetic variants associated with disease risk do not necessarily explain disease severity, and that predicting long-term outcome at diagnosis remains a major unmet need for treatment selection and trial stratification.
A Longitudinal Registry-Based Approach
The study used MSBase, a large international prospective multiple sclerosis registry, to assemble a deeply phenotyped cohort of European-ancestry individuals with relapse-onset multiple sclerosis. Participants were recruited from specialist centres in Australia, Spain, and the Czech Republic, and clinical data included EDSS scores, relapse history, treatment exposure, demographics, and disease phenotype. Rather than relying on a single cross-sectional disability score, the investigators calculated longitudinal ARMSS and MSSS measures, then classified patients into mild and severe outcome extremes; Figure 1 illustrates the screening process, and Figure 2 visualizes the divergent disability trajectories over age and symptom duration.
Genome-Wide Association: No Single Large Genetic Driver
The principal genome-wide association analyses did not identify any common single nucleotide variant that reached the conventional genome-wide significance threshold for longitudinal severity. The strongest signal was rs7289446, an intronic variant in SEZ6L on chromosome 22, followed closely by another SEZ6L-linked variant, rs1207401, but these remained suggestive rather than definitive in the primary analysis. This is scientifically important because it argues against a simple model in which multiple sclerosis severity is driven by one or a few common variants of large effect; Figure 3, including the Manhattan plot and chromosome 22 locus zoom, visually reinforces the absence of a dominant genome-wide signal.
Polygenic Severity and Machine Learning Prediction
Although no single variant explained severity, the study found evidence that many small-effect variants collectively contribute to clinical outcome. The authors used an xgboost machine learning model incorporating more than 62,000 variants, together with clinical and demographic variables available near disease onset, to classify mild versus severe disease. This model achieved an AUROC of approximately 0.84, with positive predictive value of 80% and negative predictive value of 88%, whereas a model based only on clinical and demographic variables performed near chance, with an AUROC of 0.54. Figure 4 is central here: it shows both the high-performing ROC curve for the genetic-clinical model and the weak performance of the clinical-only model.
Biological Signals Beyond the Immune System
A major conceptual contribution of the article is that severity-associated biology appears to point beyond classical peripheral immune mechanisms. Pathway and tissue-enrichment analyses implicated CNS-expressed genes, especially cerebellar genes, as well as pathways related to mitochondrial function, synaptic plasticity, oligodendroglial biology, cellular senescence, calcium signalling, Wnt signalling, and G-protein receptor signalling. This supports the authors’ title: not all roads lead to the immune system. In other words, susceptibility to developing multiple sclerosis may be strongly immunological, but severity may also depend on neuronal resilience, remyelination capacity, mitochondrial integrity, and the ability of neural networks to compensate for injury.
Sex-Stratified Findings and Disability Milestones
Secondary sex-stratified analyses identified two loci that met genome-wide significance: rs10967273 in females and rs698805, intronic to CAMKMT, in males. Survival analyses also linked several top variants to time to irreversible EDSS 3 and EDSS 6, including variants near or within SEZ6L, RCL1, MTSS1, RCAN3AS, TCF7L2, and SUCLA2. Figures 5 and 6 show Kaplan–Meier curves for disability milestone accumulation, illustrating that some variants were associated with faster or slower progression, including sex-specific effects for MTSS1, RCAN3AS, and TCF7L2. These findings are biologically plausible but should be interpreted cautiously because several variants may be tagging nearby causal mechanisms rather than being causal themselves.
Clinical Significance and Limitations
The study represents an important step toward precision prognostics in relapse-onset multiple sclerosis, suggesting that genetic information may eventually help clinicians estimate severity risk at diagnosis and tailor therapeutic intensity accordingly. However, the authors appropriately stress several limitations: the severe group included more patients diagnosed before widespread availability of disease-modifying therapy, treatment exposure could influence longitudinal severity scores, and the findings require independent validation in an equivalent cohort. Therefore, the article should not be read as delivering an immediately deployable clinical test, but as a rigorous proof of principle that multiple sclerosis severity is partly polygenic, partly modifiable, and biologically distinct from susceptibility.
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:
Jokubaitis, V. G., Campagna, M. P., Ibrahim, O., Stankovich, J., Kleinova, P., Matesanz, F., ... & Butzkueven, H. (2023). Not all roads lead to the immune system: the genetic basis of multiple sclerosis severity. Brain, 146(6), 2316-2331.
