Targeting the astrocyte-microglia EFEMP1-GALNT10 axis: a spatially programmable therapeutic strategy for hippocampal vulnerability in Alzheimer’s disease

The hippocampus is selectively vulnerable to Alzheimer’s disease (AD), but the spatially resolved, cell-type-specific drivers of this susceptibility remain unknown, hindering the development of targeted therapies. We employed an integrated causal-spatial approach, combining single-nucleus RNA sequencing (snRNA-seq) from 53 post-mortem AD brains, multi-omics triangulation (Mendelian randomization and Bayesian colocalization), and cross-species spatial mapping in human cohorts and 5xFAD mice to identify drivers of hippocampal vulnerability. We identified a novel astrocyte-microglia co-pathology axis centered on EFEMP1 and GALNT10. Causal inference prioritized this axis, supported by an EFEMP1 plasma protein quantitative trait locus (pQTL) with an exceptional effect size (OR = 7.96) and dual epigenetic-transcriptional regulation of microglial GALNT10. The axis demonstrated hippocampus-specific co-expression in humans (R = 0.81, p < 0.001), pathological amplification in 5xFAD mice (tM1 = 0.77 ± 0.04 vs. WT 0.69 ± 0.04, p < 0.001), and focal enrichment near Aβ plaques (EFEMP1: r = -0.79; GALNT10: r = -0.67, p < 0.001). Mechanistically, it forms a core interactome with EGFR and TIMP3, coupling extracellular matrix (ECM) dysregulation with neuroinflammation. Our study defines the EFEMP1-GALNT10 axis as a spatially coordinated driver of hippocampal vulnerability in AD. The integrated causal-spatial pipeline provides a generalizable framework for translating genetic associations into spatially resolved therapeutic targets.