Pseudotime analysis of 2,106 brains across nine regions reveals conserved immune, neuronal, and myelin regulatory programs in Alzheimer’s disease

Abstract

Pseudotime trajectories can reconstruct latent disease progression from cross-sectional transcriptomic data. However, whether Alzheimer’s disease (AD) progression follows a conserved molecular architecture across brain regions remains unclear. We applied pseudotime analysis to harmonized bulk RNA-seq data from 2,106 postmortem brain samples (1,364 AD, 742 controls) across nine brain regions from three AMP-AD cohorts (ROSMAP, Mayo, MSBB). Pseudotime was significantly associated with AD diagnosis in all nine regions and with Braak stage in seven of nine. We identified 21 genes with concordant pseudotime associations across all regions, increasing to 234 when the cerebellum was excluded. Pathway analysis revealed 1,268 significant associations, with synaptic deregulation as the most conserved process, and immune/ECM programs showing greater regional specificity. The cerebellum followed a distinct pattern, with enrichment for protein refolding and chaperone pathways rather than neurodegeneration. Co-expression network analysis identified six conserved metamodules, including immune/glial (MM1) and excitatory neuronal (MM2) programs spanning all nine regions, and an oligodendrocyte/myelin program (MM3) in seven cortical regions. Key driver analysis identified 76 unique genes across 35 modules, with HCK and LAPTM5 as the most broadly replicated immune regulators in seven regions. Oligodendrocyte-associated key drivers (MYRF, CNP, MOBP) increased along pseudotime in cortical regions, supporting active myelin remodeling during AD progression. These findings reveal a conserved transcriptional architecture underlying AD progression, organized around coordinated immune activation, synaptic loss, and myelin remodeling, with the cerebellum following a distinct trajectory.