Observational studies associate Epstein–Barr virus (EBV) with multiple sclerosis (MS), but causality across autoimmune neuroinflammatory diseases (ANDs) remains uncertain. This study aimed to assess the causal effects of 5 EBV antibodies on ANDs using Mendelian randomization (MR). Bidirectional, two-sample MR was performed using generalized summary-data-based MR (GSMR). Genetic instruments for EBV antibodies (anti-EBNA-1, VCA p18, ZEBRA, EA-D, IgG) were sourced from UK Biobank (UKB). Outcome data for ANDs (multiple sclerosis MS, neuromyelitis optica NMO, myasthenia gravis MG, Guillain–Barré syndrome GBS, and chronic inflammatory demyelinating polyneuropathy CIDP) were from FinnGen (discovery), UKB, and International Multiple Sclerosis Genetics Consortium (replication). Applying an evidence-tiered interpretation, we found a high-confidence, protective causal effect of increased ZEBRA antibody levels on MS risk (odds ratio OR = 0.705, P = 2.967 × 10 −6 ), which was robust to multiple testing nominally supported in an independent cohort (IMSGC). In contrast, the association between EBNA-1 antibodies and increased MS risk (OR = 1.284, P = 2.450 × 10 −4 ) was graded as suggestive as it showed nominal support only in one of 2 validation cohorts (UKB) and was substantially attenuated after excluding HLA-region single nucleotide polymorphisms, indicating shared genetic architecture rather than direct causation. Steiger directionality tests confirmed the primary causal direction was from antibodies to MS. Results for NMO, MG, GBS, and CIDP were inconclusive due to limited statistical power, and thus no definitive conclusions can be drawn regarding EBV antibodies and these diseases. This study provides high-confidence genetic evidence for a protective role of ZEBRA antibodies in MS, which was robust to multiple testing and nominally supported in an independent cohort International MS Genetics Consortium (IMSGC), and suggestive evidence for an HLA-mediated link between EBV-encoded nuclear antigen-1 (EBNA-1) and MS which showed nominal support only in one of 2 validation cohorts. The role of EBV in other ANDs warrants investigation in larger, well-powered cohorts.
Sun et al. (Fri,) studied this question.