Why Multiple Sclerosis Starts—and Why It Progresses: What the New Genetics Is Revealing
Multiple sclerosis (MS) is often described in textbooks as “heterogeneous,” but that word lands differently when you are the clinician trying to counsel a newly diagnosed patient—or the patient trying to plan a life around an unpredictable disease. In their 2025 review, Sahi and colleagues frame MS heterogeneity as the product of shifting biology over time: early focal inflammatory damage can give way to chronic compartmentalized inflammation, neuroaxonal loss, and imperfect repair. Against that moving backdrop, genetics becomes more than a list of risk loci; it becomes a structured way to ask two separate questions: “Why did MS start?” and “Why does disability accumulate faster in some people than others?” The review’s central argument is that these questions may have overlapping inputs but partially distinct genetic architectures—an insight that reshapes how we interpret genetic studies and how close we are to clinically useful prediction.
From family aggregation to a 233-variant susceptibility map
MS is not “genetically inherited” in a simple Mendelian sense, yet classic family studies clearly show enrichment: monozygotic twin concordance is substantially higher than dizygotic concordance, and first-degree relatives carry markedly higher lifetime risk than the general population. With genome-wide association studies (GWAS) scaled to very large cohorts, the field moved from single-locus signals to a broad susceptibility map: the International Multiple Sclerosis Genetics Consortium (IMSGC) identified 233 MS susceptibility variants in tens of thousands of cases and controls, estimating MS risk heritability at roughly 19%. Importantly, the review emphasizes what heritability is not: it does not mean inevitability, and it is not “genetic determination.” Instead, MS susceptibility is a complex trait—polygenic, modest-effect, and tightly interwoven with environmental and stochastic factors.
The HLA-DRB1*1501 story: strong on risk, nuanced on course
If MS genetics has a “main character,” it is HLA-DRB11501—the strongest established risk allele, consistently linked to earlier age at onset. The temptation has been to extend that importance forward into prognosis, but the evidence is mixed unless you look carefully at what outcome is being measured and when. Sahi et al. highlight long-term clinically isolated syndrome (CIS) cohorts that help disentangle inflammatory activity from later disability: in a 15-year CIS follow-up, HLA-DRB11501 associated with more MRI inflammatory burden (e.g., lesion volume and gadolinium-enhancing lesions) and faster EDSS worsening; in a separate 30-year CIS cohort, it associated with higher lesion accrual and relapse rates yet did not track cleanly with EDSS or conversion to secondary progressive MS (SPMS). The synthesis is subtle but important: HLA-DRB1*1501 looks more like a driver of inflammatory activity than a direct dial for long-term severity, and its apparent effect on “severity” may change as dominant pathology changes over decades.
Why “risk genes” rarely explain “severity” (and why that’s not a failure)
A recurring lesson in the review is that it is logically—and biologically—unsafe to assume susceptibility loci will also predict severity. Polygenic risk scores built from MS susceptibility variants can meaningfully stratify relative risk (the review notes higher scores may increase susceptibility several-fold), but they have performed poorly when asked to predict relapses, disease course, or disability outcomes. That mismatch becomes more comprehensible when you consider tissue and pathway enrichment: susceptibility loci disproportionately implicate peripheral immune mechanisms involved in disease initiation, whereas emerging severity signals appear more enriched in CNS compartments and pathways plausibly tied to neurodegeneration and repair. In other words, the genetics of “crossing the threshold into MS” may be immunologically biased, while the genetics of “how the CNS withstands and repairs injury over time” may be governed by different biology. This is not disappointing—it is clarifying—because it argues for different therapeutic hypotheses in progressive disease than in early inflammatory disease.
The first genome-wide severity locus: rs10191329 and what biology it points to
The review treats the identification of rs10191329 (near DYSF–ZNF368) as a pivotal moment: one of the first variants to reach genome-wide significance for MS severity using age-related MS severity metrics (ARMSS) in large collaborative datasets. In the IMSGC severity GWAS, rs10191329 associated with higher ARMSS and—crucially—longitudinal disability trajectories: faster EDSS progression, increased risk of confirmed disability worsening, and a shorter median time to EDSS 6.0 (needing a walking aid). The variant’s plausibility increases when aligned with pathology and biomarkers: rs10191329 risk-allele homozygosity associated with higher lesion counts in brainstem and cortex in postmortem material; other cohorts linked the risk allele to faster brain volume loss and higher serum neurofilament levels, consistent with increased neuroaxonal injury. The review also reminds readers that these are noncoding variants: the mechanistic story will likely run through regulation (e.g., gene expression/splicing) rather than altered protein sequence, making functional follow-up essential.
Replication is hard in MS severity: phenotype noise, power, and treatment-era effects
One of the most useful “humanizing” contributions of this review is its candid diagnosis of why severity genetics has been slow: replication is intrinsically difficult when outcomes are noisy, time-dependent, and treatment-modified. The standard disability anchor (EDSS) is heavily weighted toward ambulation and fluctuates with relapse/recovery; derived scales (ARMSS/MSSS) improve sensitivity but inherit limitations. Even when studies nominally use the same metric, longitudinal choices matter—median longitudinal severity (L-ARMSS/L-MSSS) may damp relapse-related volatility but can miss “end-point” disability, as illustrated by the review’s discussion of median-versus-last severity interpretation. Add statistical realities (small effect sizes, low minor allele frequency limiting homozygotes, winner’s curse) and cohort realities (different covariates, enrichment strategies, and disease-modifying therapy exposure), and it becomes easier to see how a true signal can look inconsistent across datasets. The take-home is not that findings are unreliable, but that severity GWAS needs rigor in phenotyping, scale, and harmonization that matches the complexity of progression biology.
What comes next: better phenotypes, diverse ancestries, rare variants, and usable prediction
Sahi et al. outline a pragmatic roadmap for turning severity genetics into clinical utility. First, “severity” needs deeper phenotyping: progression independent of relapse activity (PIRA), advanced MRI markers (including those capturing gray matter pathology and chronic inflammation), cognition, and patient-reported or wearable-derived outcomes may better map onto underlying mechanisms than EDSS alone. Second, MS genetics must expand beyond European-ancestry cohorts—both to reduce bias and to leverage cross-ancestry linkage disequilibrium differences for fine-mapping and causal inference. Third, missing heritability may sit in low-frequency or rare variants better captured by whole-exome or whole-genome sequencing. Finally, prediction is likely to emerge from models that combine genetics with early clinical/MRI features—potentially using nonlinear machine learning to capture non-additive effects—while mechanistic translation will require pathway- and gene-set approaches that move beyond single-variant associations. The review’s bottom line is clear: the field is finally beginning to separate the genetics of starting MS from the genetics of withstanding MS, and that separation may be exactly what progressive MS therapeutics has been waiting for.
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:
Sahi, N., Ciccarelli, O., Houlden, H., & Chard, D. T. (2025). Unlocking Multiple Sclerosis Genetics: From Susceptibility to Severity. Neurology, 105(8), e214141.
