Bidirectional Mendelian randomization study of causal associations between white-matter structural connectivity and multiple sclerosis.

Abstract

Background: Multiple sclerosis (MS) is associated with widespread network disruption, but whether specific white-matter structural connectivity (WMSC) phenotypes contribute causally to MS susceptibility remains unclear.

Methods: We performed bidirectional two-sample Mendelian randomization (MR) using genome-wide association study (GWAS) summary statistics for 206 tractography-derived WMSC phenotypes (UK Biobank; N = 26,333) and MS susceptibility (IMSGC; N = 115,803 of European ancestry). Primary inference used inverse-variance weighted (IVW) MR under a multiplicative random-effects model, complemented by MR-Egger, weighted median, weighted mode, and simple mode.

Results: In forward MR, 12 WMSC phenotypes remained associated with MS susceptibility after direction-concordance filtering and robustness assessment. These signals were not randomly distributed across the structural connectome: risk-increasing effects were concentrated in salience/control-related cortico-subcortical and sensorimotor couplings, particularly connections involving the amygdala, putamen, and contralateral somatomotor network, whereas inverse associations were more prominent in default-mode/limbic and cross-network connections, including limbic-accumbens, limbic-caudate, default-mode-hippocampus, and visual-default-mode links. Sensitivity analyses did not indicate directional pleiotropy, outlier-driven distortion, or single-variant dependence among the retained traits. No reverse causal effect of MS liability on the prioritized WMSC phenotypes was supported.

Conclusions: These bidirectional MR results support a circuit-selective model in which genetically influenced variation in specific cortico-subcortical WMSC phenotypes is associated with MS susceptibility, with risk-increasing effects concentrated in salience/control-basal ganglia-sensorimotor circuits and inverse associations enriched in default-mode/limbic and cross-network couplings. The absence of robust reverse effects is more consistent with predisposing connectivity architectures than with MS liability causally altering WMSC.