An MS Severity Variant Measured as Brain Atrophy

What determines how severely multiple sclerosis (MS) runs, and what the long-term outcome will be, is largely unknown. A recent genome-wide association study in more than 12,000 people with mostly longer disease duration, mean age 51.7 years and mean duration 18.2 years, found the minor A allele of rs10191329 in the DYSF–ZNF638 locus, at a frequency of 17%, associated with an unfavorable long-term disability outcome. Heritability enrichment analysis for MS severity in that study pointed to central nervous system tissues, in contrast to the immune-related nature of MS susceptibility. Gasperi and colleagues took that result and tested one directed hypothesis against it: that the minor allele is associated with accelerated brain atrophy. They ran it in 748 patients at the Technical University of Munich and then again in an independent cohort of 360 at the Karolinska Institute in Stockholm.

An Inclusion Rule Built Around the Measurement
The entry criteria are narrower than a typical genetic study, because the outcome demanded it. Patients had to be 18 to 70 years old at baseline with relapsing MS or a clinically isolated syndrome fulfilling MRI dissemination criteria, to have data on disease-modifying therapy at both baseline and follow-up, to have genotyping data, and to have a pair of T1-weighted and fluid-attenuated inversion recovery sequences acquired under the identical protocol on the same scanner at least 12 months apart, with a successful visual quality check of both raw and processed images. The Munich cohort had a median age of 35, was 65.4% women, and split 42 clinically isolated syndrome to 706 relapsing remitting with no secondary progressive patients, at a median EDSS of 1 and a median interscan interval of 3.0 years. Stockholm ran slightly older and more female at 37 and 76.4%, with 336 relapsing remitting and 24 secondary progressive patients, median EDSS 2 and interval 2.9 years. Minor allele frequency was 16.1% and 15.6%. Overlap with the severity cohort of the earlier GWAS was 32% in Munich and 87% in Stockholm.

What Was Measured, and the Statistics Declared in Advance
Brain atrophy, the primary outcome, was measured as percentage brain volume change per year using SIENA. Secondary outcomes were white matter lesion volume at baseline and new lesion volume, along with the volumes of cerebral cortex, thalamus, putamen and white matter, with baseline values and percentage change per year from SAMSEG in FreeSurfer; apart from lesion and intracranial volume, volumes were corrected for head size by mean scaling. Genotyping ran on Illumina arrays with quality control in PLINK, phasing with SHAPEIT2 or EAGLE, and imputation against the 1000 Genomes Phase 3 panel or the Haplotype Reference Consortium panel, with imputation quality for rs10191329 high across all three genotyping streams at 0.934, 0.969 and 0.973. Analysis used imputed dosages rather than hard genotype calls, in multivariate linear regression adjusted for age, sex, eight ancestry components, the MRI scanner and the genotyping array. Because the hypothesis had a direction, a one-sided p below 0.05 counted as significant, and the authors confirmed their p values by nonparametric permutation testing across 10,000 permutations.

The Primary Result, and Then the Same Result Again
In Munich, rs10191329*A was associated with higher rates of brain atrophy at an estimate of −0.109 (SE 0.036), one-sided p = 1.26×10⁻³. In Stockholm the estimate came out at −0.115 (SE 0.044), p = 4.86×10⁻³. Joined, the two give −0.111 (SE 0.028) at p = 6.54×10⁻⁵, meaning each copy of the A allele decreased yearly percentage brain volume change by 0.11, with a confidence interval of 0.06 to 0.17. Set against the mean yearly change of 0.40 in this population, that comes to 27.5%, with a confidence interval running from 15.0% to 42.5%. The forest plot puts both cohorts at −0.11 and the meta-analysis at −0.11 across 1,108 individuals.

Exploratory Metrics, and a Test of Whether Lesions Explain It
Joint exploratory analyses across ten further MRI metrics and EDSS did not point to specific brain regions being primarily affected. Three associations reached nominal significance: higher baseline white matter lesion volume (1.258, SE 0.441, p = 4.32×10⁻³), and higher yearly percentage volume change of the thalamus (−0.134, SE 0.049, p = 6.73×10⁻³) and the putamen (−0.074, SE 0.033, p = 2.62×10⁻²). Baseline cortex, thalamus, putamen and white matter volumes showed nothing, as did yearly change in lesion volume, cortex and white matter. Neither baseline EDSS nor yearly absolute EDSS change was associated with the variant, in subsets covering 731 patients (98%) in Munich and 281 (78%) in Stockholm. The authors state that these secondary analyses were exploratory and that they did not adjust for multiple testing. One further check addressed whether the atrophy finding simply followed the lesion finding: repeating the primary analysis with baseline white matter lesion volume as a covariate left rs10191329 significantly associated with yearly brain volume change at a joint p of 0.002.

A Comparison Group They Did Not Have to Recruit
For a reference outside MS, the authors downloaded summary statistics from ENIGMA consortium genome-wide association studies of longitudinal brain changes in more than 15,000 participants. There, rs10191329 showed only a nominally significant association with the yearly change rate of the lateral ventricles (p = 0.032), and nothing across the other 14 metrics, including yearly change rates of total brain volume (p = 0.690), thalamus (p = 0.953) and putamen (p = 0.613). Reading that absence alongside their own result, the authors write that carriers of the minor allele are more prone to MS-related brain tissue damage already at the stage of white matter lesion formation.

What the Authors Claim, and What They Ask For Next
They chose yearly percentage brain volume change as the most reliable in vivo marker of brain atrophy in MS, and they call the biological effect, an estimated 28% increase per additional allele, considerable. Having shown it in two cohorts drawn from MRI data across six scanners, they read that as pointing toward a high degree of generalizability, while stating in the same breath that the main result and particularly its effect size necessitate further confirmation and distinction. The practical suggestion they offer follows from the size of the effect: given the impact of rs10191329 on brain atrophy, it seems reasonable to consider stratifying by this genotype in clinical trials that involve brain atrophy measurements. On mechanism they are deliberately narrow, writing that at this point they have only demonstrated an association of a noncoding genetic variant with brain atrophy in MS, and that working out the molecular and cellular mechanisms behind it may reveal mechanisms of disease progression and open new avenues for treatment. Their closing claim is methodological rather than biological. These results illustrate that genetic variants can be related to the course of MS by using brain MRI as a proxy, and variants exerting effects in the same order of magnitude as rs10191329 should be detectable in large-scale multicenter MRI studies, which they describe as an ambitious but feasible step toward understanding the pathomechanisms that lead to brain atrophy and disability progression.

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:
Gasperi, C., Wiltgen, T., McGinnis, J., Cerri, S., Moridi, T., Ouellette, R., Pukaj, A., Voon, C., Bafligil, C., Lauerer, M., Andlauer, T. F. M., Held, F., Aly, L., Shchetynsky, K., Stridh, P., Harroud, A., Wiestler, B., Kirschke, J. S., Zimmer, C., Baras, A., Piehl, F., Berthele, A., Granberg, T., Kockum, I., Hemmer, B., & Mühlau, M. (2023). A genetic risk variant for multiple sclerosis severity is associated with brain atrophy. Annals of Neurology, 94(6), 1080–1085. https://doi.org/10.1002/ana.26807