Computational dissection of cellular and network vulnerability in Alzheimer's and related dementias

点击翻译按钮获取中文摘要

Project Summary Alzheimer's disease (AD) is a heterogeneous, multifactorial disease that selectively affects certain regions of the brain, e.g. locus coeruleus, entorhinal and hippocampus. Factors underlying this selective vulnerability (SV) remain unclear: Why is progression so stereotyped? Why is pathology seen in specific structures at early stages? What about certain cells makes them susceptible to AD? Current hypotheses have focused on specific features, e.g. cytoarchitecture, cell morphology, neurotransmitter system or molecular composition. The concept of cellular vulnerability (“SV-C”) has gained currency due to advances in single cell sequencing and spatial transcriptomics. Another vulnerability relates to network-based trans-neuronal “prion-like” transmission of pathology, due to which certain circuits, fiber pathways and regions (network hubs) may become selectively vulnerable (“SV-N”). This proposal will quantitatively test and validate hypotheses regarding SV-C and SV-N: 1) Protein aggregation, clearance and transmission on the network underly the spatiotemporal progression of pathology; hence SV of certain regions (e.g. EC and Hipp) may simply be a result of their location within the network topology. Alternatively, 2) SV is dictated by distribution and composition of certain neural cell types (e.g. large pyramidal neurons) that are selectively targeted by AD pathology. Beyond these are competing hypotheses is the possibility that both factors combine: 3) Pathology is governed by network transmission, but whose local and spread parameters are mediated by certain cell types (e.g. microglia). Unfortunately, AD research has so far been unable to fully test between these hypotheses or to identify which of these vulnerabilities are germane. Much of available bench, animal or human data are descriptive and do not accommodate quantitative models. We propose to develop network models for SV-C and SV-N and formal statistical models to test them. We capitalize and build on two enabling technologies that have recently come out of our laboratory: Matrix Inversion and Subset Selection (MISS) algorithm for creating whole-brain cell type maps; and Network Diffusion Model (NDM) which mathematically recapitulates transmission of tau along fiber projections. With further development of these enabling technologies, we will explore SV-C and SV-N in mouse tauopathy data and human tau- and amyloid-PET scans. We will also develop and test models where cells or genes mediate network vulnerability indirectly. If successful this project could lead to conclusive evidence for or against each of the identified SV hypotheses. We will explore in future work the morphological, molecular or electrophysiological properties of short-listed cells, genes, neural pathways and network epicenters. Our approach could become a computational test bed for future hypothesis generation and testing, without requiring expensive and laborious experiments. Proposed platform technologies (MISS and NexIS) may have even broader applicability in neuroscience.