Multi-Ancestry Immunogenetics of Multiple Sclerosis: MHC Signals, HLA Architecture, and the Limits of PRS Portability in South Asian and Afr

Multiple Sclerosis (MS) is a complex autoimmune disorder with a substantial heritable component, and large genome-wide association studies (GWAS) in predominantly European-ancestry cohorts have mapped hundreds of susceptibility loci, with the strongest signals concentrated in the Major Histocompatibility Complex (MHC). However, the historical imbalance toward European ancestry limits both mechanistic inference (e.g., fine-mapping causal variants) and clinical translation (e.g., equitable polygenic risk scoring), because allele frequencies and linkage disequilibrium (LD) patterns differ across populations. The study by Jacobs and colleagues addresses this gap by conducting an ancestrally diverse UK-based genetic analysis of MS susceptibility, focusing specifically on individuals of South Asian (SAS) and African (AFR) genetic ancestry.

Cohort Assembly and Phenotyping Strategy
The investigators established the ADAMS project (A Genetic Association study in Diverse Ancestries of Multiple Sclerosis), recruiting MS participants through UK clinical sites, an online portal, and the UK MS Register, and collecting baseline phenotypes via a standardized questionnaire. The phenotype instrument captured demographics, MS history (including subtype, age at diagnosis, and treatment exposures), and established MS risk factors (e.g., smoking history and glandular fever). Disability and health status were quantified using widely used instruments such as the Expanded Disability Status Scale (EDSS), MSIS-29, EQ5D, and the age-adjusted gARMSS severity metric, enabling clinical characterization alongside genetic analyses.

Genotyping, Imputation, and Genetic Ancestry Inference
DNA was obtained primarily from saliva samples and genotyped on the Illumina Global Screening Array. After stringent sample- and variant-level quality control, the authors imputed genotypes (initially to HRC, then re-imputed within ancestry using 1000 Genomes) and combined ADAMS cases with ancestrally similar controls (and a small number of additional cases) from UK Biobank. Genetic ancestry was inferred using PCA-derived features and a random forest classifier trained on reference panels (HGDP and 1000 Genomes), after which participants were partitioned into inferred SAS and AFR ancestry groups for within-ancestry association testing—an analytical choice intended to reduce confounding from population structure while preserving ancestry-specific LD patterns relevant to fine mapping.

Within-Ancestry GWAS Signals Emphasize the MHC
In ancestry-stratified case–control GWAS using mixed logistic models (REGENIE) with adjustment for sex and principal components (but not age in the primary model due to collinearity with ascertainment), the dominant association evidence in both ancestries localized to the MHC on chromosome 6—consistent with decades of MS immunogenetic findings. The curated datasets comprised SAS ancestry (175 MS cases, 6,744 controls) and AFR ancestry (113 MS cases, 5,177 controls). The lead SAS association mapped to the class II region near HLA-DRB1 (chr6:32600515:G:A; OR ≈ 1.84; P ≈ 4.6×10⁻⁶), whereas the lead AFR association mapped nearer class I HLA-A (chr6:29919337:A:G; OR ≈ 2.24; P ≈ 4.3×10⁻⁵). Importantly, the study reports no genome-wide significant non-MHC loci (P < 5×10⁻⁸), aligning with power limitations given the modest case counts despite careful control of test statistic inflation.

Cross-Ancestry Concordance of Established European Risk Alleles
To quantify shared genetic architecture, the authors compared effect directions for 164 independent European-ancestry MS susceptibility signals (from IMSGC discovery statistics) against the SAS and AFR results. They observed over-representation of European-derived susceptibility alleles among cases in both groups, with stronger statistical evidence in SAS (e.g., concordant direction for 104/154 variants; Spearman ρ ≈ 0.31) than AFR (80/152; weaker correlation). Although the MHC can dominate concordance metrics due to its effect size and allelic complexity, the overall pattern persisted even after excluding extended MHC variants, supporting a model in which much of MS risk biology is shared, while marginal SNP effects differ because of LD structure, allele frequency differences, and estimation noise in smaller cohorts.

HLA Alleles and Population-Attributable Burden Differ by Ancestry
Given the MHC signals, the study performed classical HLA allele imputation at six loci (HLA-A, -B, -C, -DPB1, -DRB1, -DQB1) using multi-ancestry models (HIBAG) and validated effect patterns using an independent approach (SNP2HLA). In the SAS cohort, several alleles showed suggestive association at false discovery rate < 10%, including risk-increasing effects for HLA-DPB1*10:01 (OR ≈ 3.1) and HLA-DRB1*15:01 (OR ≈ 1.7), alongside protective effects such as HLA-DRB1*13:01 and HLA-DQB1*06:03 (both OR ≈ 0.4). In the AFR cohort, HLA-A*66:01 emerged as the strongest reported association (OR ≈ 3.5), while HLA-DRB1*15:01 showed a directionally consistent risk effect (OR ≈ 2.3) but with less robust multiple-testing support. A key interpretive point is that even when an allele has a similar effect size across ancestries, its population-level impact depends on frequency: the authors estimate a much lower population attributable fraction for DRB1*15:01 in SAS (~9–10%) and AFR (~4–5%) compared with European-ancestry MS (~44% under their assumptions), underscoring why ancestry-aware immunogenetic models are necessary for both epidemiologic interpretation and risk prediction.

Polygenic Risk Score Portability, Limitations, and the Path Forward
Using European-ancestry discovery GWAS weights, the authors built polygenic risk scores (PRS) and evaluated discrimination in SAS and AFR cohorts. The PRS performed better than chance in both ancestries but with attenuated explanatory power: the best-performing scores explained ~1.6% of MS liability in SAS (empirical P ≈ 1.0×10⁻⁴) and ~0.5% in AFR (empirical P ≈ 0.08), compared with higher values typically observed in European-ancestry settings. The paper is explicit about limitations—most notably sample size, potential case–control batch effects from combining ADAMS with UK Biobank, and reliance on imputed HLA rather than gold-standard HLA typing—while also emphasizing methodological mitigations (stringent QC, ancestry inference, mixed models, cross-imputation validation). Scientifically, the study reinforces that MS susceptibility architecture is broadly shared across ancestries but not identical in marginal SNP effects, and it motivates substantially larger multi-ancestry cohorts to (i) improve PRS equity, (ii) fine-map causal immunoregulatory variants, and (iii) identify ancestry-enriched risk alleles that may reveal new therapeutic targets.

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:
Jacobs, B. M., Schalk, L., Tregaskis-Daniels, E., Scalfari, A., Nandoskar, A., Dunne, A., ... & Dobson, R. (2025). Genetic determinants of Multiple Sclerosis susceptibility in diverse ancestral backgrounds. medRxiv, 2025-01.