Regulatory landscape of Alzheimer’s disease variants in human microglia

Although defined by end-stage neuronal loss, neurodegenerative diseases are increasingly hypothesized to be initiated by early microglial dysfunction. Genetic risk for Alzheimer’s disease (AD) is particularly highly enriched for non-coding variants located in microglia-specific regulatory regions. Yet, the linkage-dense nature of these variants limits their functional interpretation. To elucidate how risk variants influence microglial gene regulation, we contrast fine-mapping of AD genome-wide association study (GWAS) loci by four complementary methods, and resolve a highly concordant set of putative causal variants. In total, 11,550 candidate regulatory sequences (CRSs), nominated from GWAS (n = 2,183), microglial expression quantitative trait loci (QTL) (n = 5,721) and microglial chromatin accessibility QTL (n = 1,519), are screened en masse by massively parallel reporter assay (MPRA) in human induced pluripotent stem cell (hiPSC)-derived microglia (iMGL) and immortalized HMC3 cell lines. Distinct sets of functional variants are identified between cellular models, underscoring the importance of using physiologically relevant microglia for variant interpretation. Functional regulatory variants (frVar) are enriched in promoters and splicing regions, with over half predicted to disrupt transcription factor (TF) binding. The affected TFs yield distinct regulatory modules, the first with an inflammatory program and the second with a DNA stress-response network. Mapping frVar to their predicted target genes reveals enrichment for biological processes related to amyloid biology, immune signaling, and neuron-microglia communication. Notably, we pinpoint functional variants within several understudied AD loci, including TSPOAP1 , providing direct evidence for causal regulatory effects. Together, these findings establish the first large-scale microglial-context MPRA, define a microglial regulome underlying AD genetic risk, and provide a functional framework to refine noncoding variant interpretation and therapeutic target discovery.