Parallel single-nucleus chromatin accessibility and transcriptomic profiling of human late-onset Alzheimer's disease brains.

Abstract

In the post-GWAS era for late-onset Alzheimer's disease (LOAD), the precise disease-causing genes, the specific causal variants, and molecular mechanisms mediating their pathogenic effects remain unknown. Recent studies using single-nucleus (sn)RNA-seq on human LOAD tissue have achieved unprecedented resolution in identifying cell type-specific gene dysregulation, however the regulatory mechanisms and genetic variability underlying these LOAD-specific transcriptomic signatures remain to be identified. Nuclei were isolated from frozen post-mortem human temporal cortex tissue from LOAD patients and cognitively healthy controls (n = 24, all Caucasian and APOE 3/3). 10X Genomics technology was used to generate single-nucleus (sn)RNA-seq and snATAC-seq libraries in parallel from the same pool of nuclei isolated from each brain sample. Uniform manifold approximation and projection (UMAP) analysis showed a total of 29 clusters assigned to 8 cell types including astrocytes, excitatory neurons, inhibitory neurons, microglia, oligodendrocytes, and oligodendrocyte precursor cells (OPCs) in both snRNA-seq and snATAC-seq datasets. We identified differentially expressed genes (DEGs) and differentially accessible peaks (DAPs) for each cluster and found overlaps with known LOAD GWAS regions (SNP+/- 1Mb) as well as new LOAD loci that were not discovered previously. We next performed co-accessibility analysis and identified co-accessible peaks that were more open in LOAD and corresponded to dysregulation of nearby target genes. This analysis also produced networks of co-accessible LOAD peaks that were enriched in transcription factor (TF) motifs of TFs that are specifically overexpressed in LOAD samples. Integrative multi-omics is a powerful strategy to identify regulatory elements underlying gene dysregulation in LOAD. By integrating single-nucleus transcriptomic and open chromatin profiles from the same pool of nuclei, we are able to identify cell subtype-specific molecular mechanisms contributing to LOAD pathogenesis. This new genetic knowledge will progress the identification of new therapeutic targets.