Two Loci for How MS Progresses, Not Whether It Starts

Genetic work in multiple sclerosis (MS) has mapped risk well, with more than 200 confirmed genome-wide significant variants accounting for roughly 38.2% of SNP heritability once the major histocompatibility region is included, and a large recent study has begun to map severity. Progression has been harder to reach, partly because the usual clinical measure resists genetic analysis: the Expanded Disability Status Scale is numerical but nonlinear, runs from 0 to 10, leans heavily on mobility, and captures insidious injury only partly, while early relapses do not predict long-term outcomes in natural history studies. Loomis and colleagues took a different measurement. Working through a precompetitive partnership between Biogen and Roche/Genentech, they ran genome-wide association studies of two longitudinal quantitative brain MRI traits across six randomized controlled trials: change in brain volume in 3,401 participants, and change in T2 lesion volume in 3,513. The first tracks progressive tissue loss, the second accumulating disease burden.

Why Trial Data Is the Right Material for This Question
The six trials are ADVANCE, ASCEND and DECIDE from Biogen and OPERA I, OPERA II and ORATORIO from Roche/Genentech, spanning relapsing remitting, secondary progressive and primary progressive disease. Participants were 51–74% female, with mean age around 36 to 38 in the relapsing remitting trials and 45 to 47 in the progressive ones, and mean study duration of about 1.8 years. Brain volume change was measured with the automated SIENA method, and T2 lesion volume with a fully automated segmentation method followed by visual review. The authors re-baselined brain volume at week 24 rather than at study entry, to avoid counting the rapid volume change that follows treatment initiation as real atrophy, and used week 48 for the untreated arm of ADVANCE, where participants then switched to active treatment. Changes were annualized and rank-transformed, and models adjusted for randomized treatment arm, age, sex and genetic ancestry. For pooling, every trial sent its MRI to the same centralized reading center.

Two Peaks, and How Steady They Are Across Six Trials
The meta-analysis covered 10,382,375 SNPs for brain volume and 10,608,740 for T2 lesion volume, with genomic inflation factors of 1.008 and 1.019 showing no ancestry-driven bias. For brain volume, the strongest signal was rs77321193 C/A, an intronic variant in PTPRD on chromosome 9, with a beta of −0.18 (SE 0.04) at P = 5.33×10⁻⁷; the minor C allele, at a frequency of 0.18, tracked with greater volume reduction over time. What makes this worth attention is its steadiness: heterogeneity was essentially zero (I² = 0, P = 0.96), several SNPs across the region showed similar effect sizes, and all six trials pointed the same direction. For T2 lesion volume the peak was rs11398377, a GC/G variant in a regulatory region 2 kb upstream of NEDD4L on chromosome 18, with beta −0.18 (SE 0.03) at P = 9.52×10⁻⁸, where the C deletion at a frequency of 0.16 tracked with greater lesion accumulation. Heterogeneity here was higher (I² = 51, P = 0.07), though the direction again held across trials.

Colocalization Gives One Gene a Direction to Act On
To test whether the association signals and gene expression signals share a causal variant, the authors ran colocalization against 48 GTEx tissues and the Database of Immune Cell eQTLs, taking all SNPs within a 250 kb window of each query SNP. The T2 lesion peak colocalized with NEDD4L expression in whole blood at a posterior probability of 0.78. That result carries a usable direction, because the minor allele associated with greater lesion accumulation is also the one associated with increased expression, which is what makes downregulating the gene a therapeutic idea rather than an open question of sign. No colocalization appeared for PTPRD, and the authors offer three candidate reasons: the study may be underpowered for colocalization, or the locus may act through mechanisms that are time-dependent, cell-lineage-specific, or independent of PTPRD expression. Gene-based tests in MAGMA, at a Bonferroni threshold of 2.6×10⁻⁶ across 17,837 genes, reached significance for neither trait, with NEDD4L among the top hits for T2 lesion change at P = 6.94×10⁻⁵.

Both Genes Sit in Brain Cells Rather Than Immune Cells
Single-nucleus expression in brain tissue from 12 people with MS showed PTPRD expressed in neurons, oligodendrocytes and oligodendrocyte precursor cells, and NEDD4L expressed mainly in neurons, including interneurons and excitatory neurons. The biology on each side fits the trait it came from. PTPRD encodes a receptor protein tyrosine phosphatase involved in cellular signalling, growth and differentiation, and its presence in oligodendrocytes matters for a brain volume trait because those are the cells that myelinate neurons. NEDD4L encodes an E3 ubiquitin-protein ligase that regulates epithelial sodium channels and voltage-gated sodium channels, both of which have been tied to demyelination and MS pathophysiology, and it also has roles in axon guidance, neurite growth and synaptic transmission. The authors connect this to ongoing work on sodium channel blockers for preventing axonal damage, and note that dalfampridine, a channel blocker already in use, improves walking in people with MS.

The Separation From Risk Genetics Is Itself the Result
The authors then asked directly whether progression genetics is the same genetics as susceptibility, and answered it four ways. Of 197 non-MHC autosomal MS risk SNPs, 161 and 162 met inclusion criteria for the two traits, and none showed association with either imaging outcome at P < 1×10⁻⁶. Neither of the two peaks was associated with MS risk, and no putative peak colocalized with risk. A polygenic risk score built from the non-MHC variants and IMSGC weights showed no association with brain volume change (P = 0.54) or T2 lesion change (P = 0.89). And the index variant of the IMSGC severity locus, rs10191329 at DYSF–ZNF638, was not associated with brain volume change (beta −0.004, P = 0.91) nor with T2 lesion change (beta 0.07, P = 0.04). The authors read this carefully rather than triumphantly, attributing part of it to the greater power of the IMSGC analysis and raising the possibility that imaging traits and clinical severity scores capture different aspects of the disease course. A previous finding in roughly 500 patients, linking HLA-DRB1*15:01 to reduced brain parenchymal volume and higher T2 lesion load, was likewise not reproduced here.

What the Authors Say Bounds This, and What Comes Next
The limitations are set out plainly and most of them describe a designable next study. Neither peak reached genome-wide significance at P < 5×10⁻⁸, or the stricter P < 2.5×10⁻⁸ the authors apply for testing two traits, and while the analysis was powered for moderate effects of common SNPs, a larger sample would change that. Additional well-characterized cohorts are needed for independent confirmation. Four of the six trials enrolled relapsing remitting participants, and the mechanisms driving progression may differ between early and late disease. Trial participants may not represent the wider MS population, given strict entry criteria. MRI follow-up ran under two years, a short window against the span of the disease, and the authors chose to stay inside the controlled treatment period rather than fold in open-label extension data, because variable follow-up lengths and differing handling of treatment arms would have muddied the pseudoatrophy adjustment. They also allow that re-baselining at 24 weeks may not have caught all of it. What the study establishes is the method as much as the loci: randomized trials come with assigned treatment arms, systematically collected imaging and a single reading center, and that combination made this analysis possible. Two putative progression loci, PTPRD and NEDD4L, now wait on other cohorts to confirm them.

Disclaimer: This blog post is based on the cited study 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.

Reference:
Loomis, S. J., Sadhu, N., Fisher, E., Gafson, A. R., Huang, Y., Yang, C., Hughes, E. E., Marshall, E., Herman, A., John, S., Runz, H., Jia, X., Bhangale, T., & Bronson, P. G. (2023). Genome-wide study of longitudinal brain imaging measures of multiple sclerosis progression across six clinical trials. Scientific Reports, 13, 14313. https://doi.org/10.1038/s41598-023-41099-0