Integrative multiomics reveals common endotypes across PSEN1, PSEN2, and APP mutations in familial Alzheimer's disease

Abstract

Background: PSEN1, PSEN2, and APP mutations cause Alzheimer's disease (AD) with an early age at onset (AAO) and progressive cognitive decline. PSEN1 mutations are more common and generally have an earlier AAO; however, certain PSEN1 mutations cause a later AAO, similar to those observed in PSEN2 and APP.

Methods: We examined whether common disease endotypes exist across these mutations with a later AAO (~ 55 years) using hiPSC-derived neurons from familial Alzheimer's disease (FAD) patients harboring mutations in PSEN1^A79V, PSEN2^N141I, and APP^V717I and mechanistically characterized by integrating RNA-seq and ATAC-seq.

Results: We identified common disease endotypes, such as dedifferentiation, dysregulation of synaptic signaling, repression of mitochondrial function and metabolism, and inflammation. We ascertained the master transcriptional regulators associated with these endotypes, including REST, ASCL1, and ZIC family members (activation), and NRF1 (repression).

Conclusions: FAD mutations share common regulatory changes within endotypes with varying severity, resulting in reversion to a less-differentiated state. The regulatory mechanisms described offer potential targets for therapeutic interventions.

Fig. 1

Transcriptomic profiling of FAD hiPSC-derived neurons. A Patient-derived Non-Demented Control (NDC), APP^V717I, PSEN1^A79V, and PSEN2^N141I hiPSCs were differentiated into neurons and purified by CD44⁻/CD184.⁻ selection. B RNA-Seq volcano plots of differentially expressed genes (DEGs) across the three FAD mutations relative to non-demented control (NDC) as determined by limma with an FDR p-value (p) < 0.05. C Quasi-proportional Venn diagram overlap of DEGs across the three FAD mutant hiPSC-derived neurons. D Gene Ontology: Biological Process (GOBP) and Hallmark database geneset enrichment using the fgsea multilevel enrichment test (left) or tmod CERNO enrichment test (right); dot plots indicate significant (-log₁₀ FDR p-value < 0.05) pathways in each mutation relative to NDC. E–F Common Transcription Factors (TFs) across the FAD mutations with predicted significant activity change by (E) ISMARA motif analysis (based on z-score, TF-gene Pearson correlation, and average gene target expression change) or (F) DoRothEA TF-gene target analysis (Normalized Enrichment Score)

Fig. 2

Co-expression module detection in FAD hiPSC-derived neurons. A Co-expression modules identified by CEMiTool module detection; left, fgsea enrichment of each module across the FAD mutations; right, gene size for each co-expressed module. B-C Hypergeometric enrichment of CEMiTool module and first neighbor genes using B ENCODE-ChEA Consensus and ReMap TF-gene target databases (ENCODE, E; ChEA, C; ReMap, R) or (C) GOBP and Hallmark ontology databases (Hallmark, H). D Combined PPI and TF-gene target networks of the CEMiTool co-expression module 3 for PSEN1^A79V hiPSC-derived neurons

Fig. 3

Regions of differential chromatin accessibility are enriched for transcriptional regulators and pathways mirroring gene expression signatures. A TSS and PEREGRINE enhancer heatmap coverage plots of Tn5-accessible chromatin in NDC, APP^V717I, PSEN1^A79V, and PSEN2^N141I hiPSC-derived neurons as determined by ATAC-seq. B Differential accessibility plots (log₂FC) of ATAC-seq peaks for each FAD mutation relative to NDC (significant peaks: red, up; blue, down). C Annotation (promoter, PEREGRINE enhancer, or distal/intergenic) and directionality of significant differential ATAC-seq peaks for each FAD condition. D HINT TF footprinting analysis in all accessible ATAC-seq regions using the CIS-BP motif database to identify TFs with a change in footprinting activity. E Top differentially activated and repressed TFs across the FAD mutations based on HINT-ATAC footprinting analysis. F Tn5 insertion density in each FAD mutation relative to NDC around ASCL1 (top) or NRF1 (bottom) motifs as determined by HINT

Fig. 4

Transcription factor motif enrichment of chromatin accessibility reveals endotype-associated regulator differential activity. A TF motif enrichment of accessible ATAC-seq peaks (all peaks, promoter-associated peaks, and enhancer-associated peaks) using GimmeMotifs maelstrom with the SwissRegulon motif database. B chipenrich enrichment analysis of differentially accessible promoter-associated regions with increased accessibility (top) or decreased accessibility (bottom) using the GOBP and Hallmark ontology databases (Hallmark, H). C-D chipenrich enrichment analysis of differentially accessible enhancer-associated regions with differential accessibility using the (C) GOBP and Hallmark ontology databases or (D) ENCODE-ChEA Consensus and ReMap TF-gene target databases (ENCODE, E; ChEA, C; ReMap, R); FDR p-value < 0.05

Fig. 5

Chromatin accessibility change drives differential gene expression and dedifferentiation in FAD mutant hiPSC-derived neurons. A Differential ATAC-seq peaks with corresponding differential gene expression change in APP^V717I, PSEN1^A79V, and PSEN2.^N141I hiPSC-derived neurons relative to NDC; right, annotation (promoter, PEREGRINE enhancer, or distal/intergenic) and direction of differential ATAC-seq peak change. B CONfident efFECT size (confect) of differential chromatin accessibility (ATAC-seq) and gene expression (RNA-seq) for significant genes (by ATAC-seq) using topconfects. C quasi-proportional Venn diagram overlap of genes with significant differential accessibility and gene expression change between the FAD mutations. D Union of all genes with significant differential accessibility and gene expression in three FAD mutations; left, peak-centered ATAC-seq coverage in all four conditions; right, differential confect score for ATAC-seq and RNA-seq for each gene (relative to NDC), with corresponding z-score correlation; far right, genes with high correlation and increased (red) or decreased (blue) expression and accessibility change. E TFs with differential activity based on chromatin accessibility change (ATAC-seq) around TF motifs and target gene expression change (RNA-seq) across the three FAD mutations relative to NDC using DiffTF with the CIS-BP motif database. F-G Schematic for intepareto ranking of genes characterized by ATAC-seq and RNA-seq to identify functional programs with the highest correlation of chromatin accessibility change and gene expression change; z-scores of log₂FC change of ATAC-seq peak accessibility change (promoter- or enhancer-located) and log₂FC of RNA-seq gene expression for each gene across all FAD mutations relative to NDC, followed by pareto optimization ranking for each gene and subsequent CERNO ranked geneset enrichment test. H-I intepareto-CERNO ranked enrichment using the (H) ENCODE-ChEA Consensus and ReMap TF-gene target databases (ENCODE, E; ChEA, C; ReMap, R) and (I) GOBP and Hallmark databases (Hallmark, H)

Fig. 6

Endotype dysregulation driven by chromatin accessibility change or key regulator activity leads to precursor lineage state in FAD neurons. A-D ATAC-seq coverage plots (left) and RNA-seq expression (right) showing differential ATAC-seq peaks common across FAD mutant hiPSC-derived neurons occurring in promoter and enhancer regions for factors related to (A) inflammation (CXCL12), B-C neuronal development (ZIC2, NEUROG2), and (D) neuronal lineage (DLX2). E The hallmark disease mechanism in FAD mutations is dedifferentiation to a precursor-like state. Left, differentiation of pluripotent cell to a terminal neuron, with mechanistic TFs differentially regulated in FAD neurons (red, increased activity; blue, decreased activity). Right, qualitative comparison of severity of dedifferentiation across mutations in PSEN1, PSEN2, and APP