Single-Cell Transcriptome-Wide Mendelian Randomization and Colocalization Analysis Reveals Immune Cell-Specific Mechanisms in Alzheimer's Disease

Objective: To integrate single-cell expression quantitative trait locus (sc-eQTL), Alzheimer's disease (AD) genome-wide association study (GWAS), and Bayesian colocalization analyses, in order to systematically identify candidate causal genes and potential immune mechanisms that specifically regulate AD risk in peripheral immune cells. Methods: Using the cis-eQTLs of 14 peripheral blood immune cell types from the OneK1K single-cell eQTL dataset as exposure data and the AD GWAS (GCST90565439) as outcome data, a two-sample Mendelian randomization (MR) analysis was performed. Weak instrumental variables were removed using F-statistics, and approximately independent genetic instruments were obtained by linkage disequilibrium pruning. For significant gene-cell type associations, compareCluster functional enrichment, Bayesian colocalization, and phenome-wide association study (PheWAS) were further performed, and the 1M-scBloodNL dataset was used for replication validation. Results: The MR analysis identified 187 gene-cell type associations significantly associated with AD risk, involving 161 candidate causal genes. Among them, 141 genes showed significant associations in only a single immune cell type, suggesting that AD-related immune genetic regulation has marked cell type specificity. Functional enrichment analysis showed that the significant genes in different immune cell types were enriched in biological processes including MHC class II antigen presentation, phagocytosis, membrane invagination, and positive regulation of T cell activation. Colocalization analysis indicated a strong shared genetic signal between TNF expression in CD4+SOX4+ T cells and AD risk (PP.H4 = 0.833). In the 1M-scBloodNL replication analysis, several candidate genes showed consistent direction and nominal significance. The PheWAS results suggested that the significant phenotypic associations of the TNF instrumental variable were mainly concentrated in immune- and hematology-related traits, with no genome-wide significant neurological phenotypic associations observed. Conclusion: This study suggests that single-cell eQTL-MR combined with colocalization analysis can be used to dissect the cell type-specific genetic regulatory mechanisms of peripheral immunity in AD. The genetically predicted expression of TNF in CD4+SOX4+ T cells may represent an important immune-related signal for AD risk, but further validation in independent populations and functional experiments is still required.