Cell-type-specific Mendelian randomization analysis of brain single-nucleus eQTLs and neuromyelitis optica spectrum disorder risk.

BACKGROUND

Neuromyelitis optica spectrum disorder (NMOSD) is a rare central nervous system autoimmune disease characterized by aquaporin-4 antibody-mediated astrocyte injury. Although genetic risk variants in the major histocompatibility complex (MHC) region have been consistently implicated in NMOSD susceptibility, the cellular context in which these genetic risk factors operate in the brain remains uncharacterized. We aimed to identify cell-type-specific genetic associations between brain gene expression and NMOSD risk through a systematic Mendelian randomization screen.

METHODS

Cis-eQTL summary statistics from control-only single-nucleus RNA sequencing data of 183 individuals across eight major brain cell types were used as exposures, and the largest European-ancestry NMOSD GWAS (215 cases, 1244 controls) as the outcome. Two-sample Mendelian randomization was performed using the Wald ratio method with Bonferroni correction (P < 2.87 × 10⁻⁵). Steiger directionality testing was performed. Colocalization analysis (coloc.abf) was performed for the Bonferroni-significant loci. Exploratory subtype-stratified analyses were conducted for Bonferroni-significant exposures using NMOSD-IgG+ (132 cases) and NMOSD-IgG- (83 cases) GWAS data.

RESULTS

Of 1465 cell-type-gene exposures screened, three reached Bonferroni significance, all in the MHC region: HLA-DRB1 in excitatory neurons (OR = 2.50, 95% CI 1.84-3.41), HLA-B in oligodendrocytes (OR = 0.23, 95% CI 0.12-0.44), and HLA-B in inhibitory neurons (OR = 0.32, 95% CI 0.20-0.51). Steiger testing supported the assumed direction of effect for all three. All three associations were substantially stronger in NMOSD -IgG+ and attenuated or absent in NMOSD -IgG-. Colocalization supported a shared causal signal for HLA-DRB1 in excitatory neurons (PP.H4 = 0.90) but not for either HLA-B association (PP.H4 <0.5), consistent with the latter tagging the shared risk haplotype.

CONCLUSIONS

Cell-type-specific Mendelian randomization identified three MHC-region genetic associations with NMOSD that exhibited clear AQP4-IgG+ subtype specificity, providing a cell-type-resolved perspective on the genetic architecture of seropositive NMOSD. Because of the extensive linkage disequilibrium across the MHC, these signals most plausibly tag the shared HLA-DRB1×03:01/HLA-B*08:01 risk haplotype rather than reflecting independent cell-type-specific causal effects, and should be regarded as hypothesis-generating.