Topological Analysis of Single-Cell Hierarchy Reveals Inflammatory Glial Landscape of Macular Degeneration

Abstract

Age-related macular degeneration (AMD) is a degenerative disease of the retina with unknown cellular drivers.  To identify, characterize, and compare rare pathogenic clusters of cells, we developed CATCH, a coarse graining framework that learns the cellular hierarchy. CATCH includes a suite of tools that identify salient levels of the hierarchy, automatically characterize clusters, identify pathogenic populations, and rapidly compute differentially expressed genes. We applied CATCH to the first single-cell atlas of AMD and identified two populations of activated glia enriched in the early phase of disease, one microglial and one astrocyte. Applying CATCH to other degenerative diseases revealed the same activated states in early disease, indicating a common glial signature during early neurodegeneration. In late-stage AMD, CATCH identified a microglia-to-astrocyte signaling axis mediated by IL-1β which drives VEGFA expression and pathologic angiogenesis. We validated this mechanism using in vitro and in vivo assays, identifying a new therapeutic target for AMD.

Keywords: retina, neurodegeneration, age-related macular degeneration, manifold learning, data-diffusion, clustering, single-cell transcriptomics, machine learning, glia, earth mover's distance