DOI: 10.1002/alz.074328 ISSN: 1552-5260

Discovery of disease‐associated, tau, and inflammatory microglial subpopulations and molecular drivers in Alzheimer’s disease using network‐based deep learning integration of human brain single‐cell RNA‐sequencing data

Feixiong Cheng, Jielin Xu, Andrew A. Pieper, Jeffrey L. Cummings, James B Leverenz
  • Psychiatry and Mental health
  • Cellular and Molecular Neuroscience
  • Geriatrics and Gerontology
  • Neurology (clinical)
  • Developmental Neuroscience
  • Health Policy
  • Epidemiology
Show PDF Cite

Abstract

Background

Microglia have been implicated in the pathophysiology of Alzheimer’s disease (AD) for over 100 years, and substantial progress has been made in characterizing microglial heterogeneities and biology, including disease‐associated microglia (or DAM, amyloid‐pathology related, neuro‐protective), tau microglia (tau‐pathology related), and inflammatory microglia (neuro‐toxic). However, the crucial next step knowing how to apply these findings to develop new treatments for AD, through uncovering the functional roles and molecular drivers of each microglial subtype, has not yet been accomplished.

Method

In this study, we conducted single‐cell RNA‐sequencing data integration analyses of ∼0.8 million cells/nuclei by leveraging frozen brain samples from AD subjects across different brain regions from The Alzheimer’s Cell Atlas (https://taca.lerner.ccf.org/) using a deep learning analytic approach. We annotated three microglial subtypes (including DAM, tau, and inflammatory microglia) using well‐established marker genes and then used a human protein‐protein interactome network‐based approach to identify potential molecular drivers of different microglial subtypes across the spectrum of AD. We also applied transition gene network analysis to characterize microglial subtype transition from non‐disease associated to disease‐associated stages at transcriptomic levels.

Result

Via trajectory analysis we pinpointed the existence of three microglial subtypes across AD progression. We also identified that tau microglia were significantly associated with synaptic processes. Compared to DAM, upregulated genes within inflammatory microglia were more significantly enriched within key immune pathways (e.g., NF‐kappa B and toll‐like receptor signaling pathways). In addition, transition gene networks of inflammatory microglia and DAM were found to contain potential AD pathobiology regulators (e.g., SYK, LYN, IRF8, and CSF1R) and genetic risk genes (including INPP5D, PICALM, and MEF2A). We further conducted network‐based drug repurposing prediction by simultaneously activating DAM while inhibiting inflammatory microglia, which indicated that several predicted repurposable drugs (i.e., fluticasone and mometasone) are significantly associated with reduced risk of AD in large patient databases.

Conclusion

This study identifies potential underlying mechanisms of microglial subtypes involved in human AD brains, leading to potential drug targets for future development of microglia‐targeted therapies for AD.

More from our Archive