MS Severity Gene Variant Also Tied to Slower Thinking in Healthy Adults, UK Biobank Study Finds

A single genetic variant already known to worsen the course of multiple sclerosis appears to subtly dull cognitive performance even in people who never develop the disease, according to a new analysis of nearly 400,000 adults in the UK Biobank. The finding, published in Annals of Clinical and Translational Neurology, adds weight to an increasingly influential idea in neurology: that the same biological machinery governing how well the brain withstands injury may shape how fast neurological diseases progress, whether or not disease is present at all.

The variant in question, known as rs10191329, sits in the stretch of DNA between two genes called DYSF and ZNF638. It does not change any protein sequence, yet a large genome-wide association study of more than 12,000 people with multiple sclerosis previously linked carrying its A allele to greater cross-sectional physical disability. Subsequent work connected the same allele to brain atrophy, thinning of the retinal nerve fibre layers, and elevated levels of serum neurofilament light chain, a blood biomarker of neuroaxonal damage. Together, these findings pointed toward a mechanism centred on the central nervous system itself rather than on the immune attack that defines relapsing disease.

That pattern intrigued the study’s authors. If the A allele erodes the nervous system’s resilience, they reasoned, its influence might not stop at physical disability. Cognitive impairment is one of the most common and disabling features of multiple sclerosis, and the concept of cognitive reserve, the brain’s capacity to tolerate damage while preserving function, is well established across neurodegenerative conditions including Alzheimer’s and Parkinson’s disease. The team therefore hypothesised that rs10191329 might influence cognition not only in multiple sclerosis but in healthy people and in other neurological disorders, and turned to the UK Biobank to test the idea at population scale.

The UK Biobank recruited roughly half a million participants aged 40 to 69 between 2006 and 2010, collecting genetic data alongside questionnaires, physical assessments, imaging and biological samples. From this resource the researchers constructed five mutually exclusive cohorts of European ancestry: people with a recorded multiple sclerosis diagnosis, people with Parkinson’s disease, people with all-cause dementia, people with migraine as a negative disease control, and healthy controls. After excluding participants with missing genotype data, the final analytical sample comprised 399,031 individuals, including 373,530 controls, 2,026 people with multiple sclerosis, 2,337 with dementia, 1,466 with Parkinson’s disease and 19,672 with migraine.

The cognitive outcomes came from three tests administered at recruitment. Reaction time was measured in a computerised card-matching task, with higher scores indicating slower responses. Fluid intelligence was scored as the number of correct answers out of 13 on a reasoning test, with higher scores indicating better performance. Prospective memory, the ability to remember to carry out an intended action, was assessed as a binary outcome based on whether participants correctly recalled a task instruction on their first attempt. The researchers confirmed that the multiple sclerosis, Parkinson’s and dementia cohorts showed the expected pattern of poorer performance compared with controls and the migraine group, validating the measures.

Statistical models adjusted for age, sex, the first four genetic principal components and the Townsend deprivation index, a measure of socioeconomic status. Linear outcomes were rank-inverse normal transformed to satisfy model assumptions, and false discovery rate correction was applied across all tests simultaneously. Sensitivity analyses added educational leaving age as a covariate and repeated the models under dominant and recessive genetic frameworks. As a positive control, the team also examined the relationship between the variant and self-reported disability claims among the multiple sclerosis cohort.

The results were striking in their direction, if modest in their magnitude. Among healthy controls, each copy of the A allele was associated with slower reaction time, a beta of 0.01 standard deviations per allele; lower fluid intelligence, a beta of minus 0.02 standard deviations; and higher error rates on the prospective memory task, an odds ratio of 1.06. All three associations reached study-wide significance. In the multiple sclerosis group, carriers of two A alleles were 5.9 percent more likely to claim disability allowance than common-allele homozygotes, 52.4 percent versus 46.5 percent, although the regression estimate did not reach statistical significance. In the other disease cohorts, associations were directionally concordant but not significant, with slower reaction time in multiple sclerosis, Parkinson’s and dementia, lower fluid intelligence across all cohorts, and impaired prospective memory in dementia.

Power calculations help explain the pattern of significance. Using reaction time as an example, the study had 98 percent power to detect a per-allele effect larger than 0.15 standard deviations in the multiple sclerosis cohort of roughly 2,000 people, but only 63 percent power for a 0.1 standard deviation effect. In other words, the analysis was well equipped to detect moderate genetic effects within disease groups but underpowered for the small effects observed in the enormous control sample. The authors are careful to note that the cognitive tests were designed as pragmatic population measures, not as validated outcomes for multiple sclerosis, and that prevalent cases in the Biobank are biased toward milder disease, both of which would attenuate detectable associations.

One obvious alternative explanation is education. A previous Mendelian randomisation analysis suggested genetic overlap between the severity locus and educational attainment, and cognitive reserve is strongly shaped by education. Yet adjusting for educational leaving age, a proxy for educational attainment, did not meaningfully attenuate the associations, making it unlikely that schooling alone accounts for the signal. The variant’s effects, whatever their mechanism, appear to operate at least partly independently of educational background, consistent with a direct influence on brain health rather than a purely sociodemographic confound.

The broader implication is that a variant discovered as a modifier of multiple sclerosis severity may act through general properties of central nervous system resilience, influencing how the brain performs even in the absence of disease. The effect sizes are small, and the authors emphasise that large disease-specific cohorts are needed to determine whether the variant shapes cognition within multiple sclerosis and other disorders, and whether its influence extends to populations of non-European ancestry, which this study did not examine. Still, the convergence of evidence, spanning disability, brain atrophy, neurofilament light chain, retinal thinning and now cognition, sketches a coherent picture of a genetic locus that quietly sets the baseline health of the nervous system, and in doing so helps determine how hard neurological disease hits when it arrives.

Subject of Research: Association between the multiple sclerosis severity allele rs10191329A and cognitive function in the UK Biobank

Article Title: The Multiple Sclerosis Severity Allele rs10191329A and Cognitive Function: A UK Biobank Study

Article References: Zimianiti, I., Waters, S., Harroud, A., Stridh, P., Dobson, R., & Jacobs, B. M. (2026). The Multiple Sclerosis Severity Allele rs10191329 A and Cognitive Function: A UK Biobank Study. Annals of Clinical and Translational Neurology, 13(10), 2143-2147. https://doi.org/10.1002/acn3.70458

Image Credits: AI Generated

DOI: 10.1002/acn3.70458

Keywords: multiple sclerosis, rs10191329, UK Biobank, cognitive function, genetics, neurodegeneration, DYSF-ZNF638 locus, neurofilament light chain, cognitive reserve, fluid intelligence, reaction time, prospective memory