Dissecting cell‐subtype specific regulatory variants and their target genes underpinning the pathogenesis of late onset Alzheimer’s disease

Abstract

AbstractBackgroundThe genetic architecture of late‐onset Alzheimer’s disease (LOAD) is yet to be fully explored. While GWAS discovered numerous LOAD‐associated loci, the causal variants and their target genes remain largely unknown. We aim to advance our understanding of the genetic architecture of LOAD by systematic interrogation of elements and variants underlying gene dysregulation in LOAD in a cell‐subtype resolution.MethodsWe conducted a single‐nucleus (sn)multi‐omics study on the 10X Genomics platform using 24 frozen brain tissues, 12 normal and 12 LOAD. The parallel snRNA‐seq and snATAC‐seq collected from the same nuclei samples and at the same time were used in multimodal analysis to profile in gene expression and chromatin accessibility.ResultssnRNA‐seq data from a total of 202,223 nuclei were grouped into 33 clusters and snATAC‐seq from 79,120 nuclei were separated into 25 clusters. We identified multiple cell subtype specific, LOAD‐associated differential expressed genes (DEGs), differential accessible peaks (DAPs) and cis co‐accessibility networks (CCANs) in LOAD. Integrative analysis focused on 110 LOAD CCANs out of ∼15,000 with at least one peak overlap promoter/intron 1 of DEG that contain >1 DAPs, identified 331 functional candidate cis regulatory elements (cCRE) and 118 linked genes across all cell subtypes. Out of which 12 CCANs encompassing a total of 54 cCREs and their 13 target genes overlap LOAD‐GWAS regions. Finally, we catalogue putative functional SNPs within cell subtype‐specific cCREs that interrupt with transcription factor (TF) motifs. For example, in subtypes of excitatory neurons we identified SNPs in cCREs linked to RPS15, FKBP5 that strongly affect the TF motifs of GSC2 and TBX3; and found in microglia multiple SNPs within APOE region with weaker effects on four TFs.ConclusionTo our knowledge this study represents the most comprehensive integrative single‐cell genomics study in LOAD. We provide new insights into the interactions between the genome, epigenome, and transcriptome in LOAD brains in an unprecedented cell‐subtype specific resolution. The outcomes of this work prioritize CREs, genetic variants and their linked genes for experimental validation in disease model systems using genome editing technologies. In summary, our findings enhance the translation of LOAD‐genetic risk into mechanistic understanding of causation.