Rare Coding Variants and the Genetic Basis of Multiple Sclerosis
Multiple sclerosis (MS) is a complex immune-mediated disease in which inflammatory and neurodegenerative processes damage the central nervous system. Family and population studies demonstrate a substantial genetic contribution to susceptibility, yet the inheritance pattern does not conform to a single-gene model. Genome-wide association studies (GWAS) have identified more than 230 independent common-variant associations, including powerful effects within the major histocompatibility complex. Collectively, however, common autosomal variants account for only approximately 19.2% of the additive genetic component measured by genotyping arrays. This discrepancy constitutes part of the “missing heritability” problem: disease risk is clearly heritable, but a large proportion cannot be assigned to established common variants. The study by the International Multiple Sclerosis Genetics Consortium addresses the possibility that low-frequency and rare protein-coding variants contribute to this unexplained component. Its central conclusion is that coding variants below the frequency range efficiently interrogated by conventional GWAS explain nearly 5% of MS liability and reveal susceptibility genes that common-variant analyses had not detected.
An International Exome-Wide Association Study
The investigators assembled one of the largest rare-variant datasets then available for MS genetics, analysing 32,367 individuals with MS and 36,012 unaffected controls from Australia, ten European countries and multiple regions of the United States. They examined 120,991 coding variants distributed across autosomal exons, including 104,218 nonsynonymous and 2,276 nonsense variants with an increased probability of altering protein function. Genotyping was performed with the Illumina HumanExome BeadChip or a custom MS array containing the same exome-focused content. The analytical pipeline incorporated stringent exclusions for low call rates, abnormal heterozygosity, sex discrepancies, relatedness, population outliers and technical artefacts. Mixed linear models, genetic relatedness matrices and principal-component covariates were used to limit confounding by ancestry and cryptic population structure, after which association statistics were combined across geographical strata by inverse-variance-weighted meta-analysis. Although the exome array captured approximately 88% of low-frequency coding variants observed among non-Finnish Europeans in ExAC, it detected fewer than 5% of extremely rare alleles, an important boundary on the study’s scope.
Coding Associations Hidden from Conventional GWAS
The meta-analysis identified seven significant coding variants in six genes outside the extended major histocompatibility complex. Two signals, located in GALC and TYK2, occurred within regions previously implicated by common-variant GWAS and were correlated with established association signals. The remaining findings implicated PRF1, HDAC7, NLRP8 and PRKRA independently of known common-variant loci. These variants showed negligible linkage disequilibrium with nearby common markers and were therefore unlikely to have been detected through standard genotype imputation. The associated changes included PRF1 p.Ala91Val, HDAC7 p.Arg166His, NLRP8 p.Ile942Met and two linked PRKRA substitutions, p.Asp33Gly and p.Pro11Leu. Most effect sizes were modest, as expected for a complex polygenic disease, although the very rare NLRP8 allele had a comparatively larger protective association. Six of the seven minor alleles were associated with reduced disease risk; nevertheless, balanced resampling experiments indicated that this pattern did not reflect a systematic tendency for low-frequency alleles to be protective. The result illustrates how functional coding analyses can identify disease-relevant genes even when individual variants exert relatively small population-level effects.
Quantifying the Contribution of Rare Variation
Individual association tests reveal only variants whose effects and frequencies are sufficient to exceed stringent genome-wide significance thresholds. To estimate the aggregate contribution of variants that remained individually undetectable, the researchers applied genome-wide complex trait analysis using restricted maximum-likelihood modelling. Variants were partitioned according to minor allele frequency, permitting separate estimates for common, intermediate-frequency and rare coding variation. Low-frequency variants collectively explained 11.34% of the observed case–control variance, corresponding to an average of 4.1% on the disease-liability scale. Variants with frequencies below 1% alone explained 9.0% of observed-scale variance, or approximately 3.2% on the liability scale. The distinction between these scales is important: observed-scale estimates describe variance within the sampled case–control dataset, whereas liability-scale estimates adjust for the assumed prevalence of MS in the population. The figure on page 5 demonstrates that the signal is not attributable solely to the small number of genome-wide significant variants. Instead, it suggests a broader polygenic burden composed of many rare nonsynonymous alleles whose individual effects cannot yet be resolved, even with tens of thousands of participants.
Regulatory T Cells, Cytotoxicity and Interferon-γ Biology
The biological interpretation of the newly implicated genes reinforces the central role of immune dysregulation in MS. PRF1 encodes perforin, a pore-forming protein required for granzyme-mediated cytotoxicity in cytotoxic T lymphocytes, natural killer cells and subsets of regulatory T cells. The associated p.Ala91Val substitution has previously been linked to reduced target-cell killing and increased interferon-γ secretion by natural killer cells. Such impaired cytotoxic efficiency may prolong interactions between immune cells and their targets, intensifying T-cell receptor signalling and altering cytokine production. This mechanism is especially relevant because regulatory T cells from patients with MS can acquire an abnormal T-helper-like phenotype characterized by interferon-γ secretion. HDAC7, meanwhile, encodes a class II histone deacetylase that supports the transcriptional repression mediated by FOXP3, the principal regulator of regulatory T-cell identity. HDAC7 also contributes to thymocyte survival and T-cell development. Together, the PRF1 and HDAC7 associations connect inherited MS susceptibility to cytotoxic regulation, regulatory T-cell stability, thymic selection and the maintenance of peripheral immune tolerance.
Innate Immunity and Antiviral Stress Responses
The findings also broaden the prevailing view of MS genetics beyond mature adaptive lymphocytes. PRKRA encodes an interferon-inducible activator involved in the cellular response to double-stranded RNA, a molecular pattern commonly associated with viral infection. Through interaction with protein kinase R, PRKRA influences translational arrest, interferon production, apoptosis and nuclear factor-κB signalling. Because nuclear factor-κB regulates inflammatory activation and has repeatedly been implicated in autoimmunity, altered PRKRA function offers a plausible route through which antiviral stress responses could modify MS susceptibility. NLRP8 encodes an intracellular pattern-recognition receptor associated with innate immune signalling, although its precise function in relevant neural and immune populations remains comparatively poorly characterized. The NLRP8 p.Ile942Met variant was exceptionally rare and was reported in individuals of European ancestry in the reference dataset used by the authors. The simultaneous implication of PRKRA, NLRP8 and the thymic regulator HDAC7 therefore expands the pathogenic framework of MS: inherited risk may arise not only from aberrant peripheral T- and B-cell activity, but also from innate immune sensing, antiviral signalling and early immune-cell development.
Significance, Limitations and Future Directions
This study does not fully resolve the missing heritability of MS, but it demonstrates that rare coding variation represents a measurable and biologically informative component of disease susceptibility. Several limitations should guide interpretation. The exome array was designed around previously catalogued variants and therefore had limited sensitivity for extremely rare, population-specific or newly arising mutations. Its coverage was assessed primarily against variants found in individuals of European ancestry, making extrapolation to underrepresented populations uncertain. Statistical power also remained insufficient to identify most rare alleles individually, and some heritability models failed to converge in smaller cohorts. Most importantly, statistical association does not establish the molecular mechanism operating in a particular cell type. Functional studies will be required to determine how the implicated substitutions affect protein activity, immune-cell differentiation and interactions between peripheral immunity, microglia and the central nervous system. Nevertheless, coding variants are generally more experimentally tractable than noncoding associations because they directly identify a candidate gene and amino-acid change. The work consequently provides a focused set of hypotheses for mechanistic investigation while demonstrating that future progress will require larger, ancestry-diverse sequencing studies integrated with cell-specific functional genomics.
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:
Mitrovič, M., Patsopoulos, N. A., Beecham, A. H., Dankowski, T., Goris, A., Dubois, B., ... & Cotsapas, C. (2018). Low-frequency and rare-coding variation contributes to multiple sclerosis risk. Cell, 175(6), 1679-1687.
