New study published in Clinical Epigenetics using Using MSBase Registry data. Congratulations Vilija Jokubaitis and committee members!

Abstract

Background: The variation in multiple sclerosis (MS) disease severity is incompletely explained by genetics, suggesting

genetic and environmental interactions are involved. Moreover, the lack of prognostic biomarkers makes it

difficult for clinicians to optimise care. DNA methylation is one epigenetic mechanism by which geneenvironment

interactions can be assessed. Here, we aimed to identify DNA methylation patterns associated with mild and severe

relapse-onset MS (RMS) and to test the utility of methylation as a predictive biomarker.

Methods: We conducted an epigenome-wide association study between 235 females with mild (n = 119) or severe

(n = 116) with RMS. Methylation was measured with the Illumina methylationEPIC array and analysed using logistic

regression. To generate hypotheses about the functional consequence of differential methylation, we conducted

gene set enrichment analysis using ToppGene. We compared the accuracy of three machine learning models in classifying

disease severity: (1) clinical data available at baseline (age at onset and first symptoms) built using elastic net

(EN) regression, (2) methylation data using EN regression and (3) a weighted methylation risk score of differentially

methylated positions (DMPs) from the main analysis using logistic regression. We used a conservative 70:30 test:train

split for classification modelling. A false discovery rate threshold of 0.05 was used to assess statistical significance.

Results: Females with mild or severe RMS had 1472 DMPs in whole blood (839 hypermethylated, 633 hypomethylated

in the severe group). Differential methylation was enriched in genes related to neuronal cellular compartments

and processes, and B-cell receptor signalling. Whole-blood methylation levels at 1708 correlated CpG sites classified

disease severity more accurately (machine learning model 2, AUC = 0.91) than clinical data (model 1, AUC = 0.74) or

the wMRS (model 3, AUC = 0.77). Of the 1708 selected CpGs, 100 overlapped with DMPs from the main analysis at the

gene level. These overlapping genes were enriched in neuron projection and dendrite extension, lending support to

our finding that neuronal processes, rather than immune processes, are implicated in disease severity.

Conclusion: RMS disease severity is associated with whole-blood methylation at genes related to neuronal structure

and function. Moreover, correlated whole-blood methylation patterns can assign disease severity in females with RMS

more accurately than clinical data available at diagnosis.

Back