Endotype Characterization Reveals Mechanistic Differences Across Brain Regions in Sporadic Alzheimer's Disease

Abstract

Background: While Alzheimer's disease (AD) pathology is associated with altered brain structure, it is not clear whether gene expression changes mirror the onset and evolution of pathology in distinct brain regions. Deciphering the mechanisms which cause the differential manifestation of the disease across different regions has the potential to help early diagnosis.

Objective: We aimed to identify common and unique endotypes and their regulation in tangle-free neurons in sporadic AD (SAD) across six brain regions: entorhinal cortex (EC), hippocampus (HC), medial temporal gyrus (MTG), posterior cingulate (PC), superior frontal gyrus (SFG), and visual cortex (VCX).

Methods: To decipher the states of tangle-free neurons across different brain regions in human subjects afflicted with AD, we performed analysis of the neural transcriptome. We explored changes in differential gene expression, functional and transcription factor target enrichment, and co-expression gene module detection analysis to discern disease-state transcriptomic variances and characterize endotypes. Additionally, we compared our results to tangled AD neuron microarray-based study and the Allen Brain Atlas.

Results: We identified impaired neuron function in EC, MTG, PC, and VCX resulting from REST activation and reversal of mature neurons to a precursor-like state in EC, MTG, and SFG linked to SOX2 activation. Additionally, decreased neuron function and increased dedifferentiation were linked to the activation of SUZ12. Energetic deficit connected to NRF1 inactivation was found in HC, PC, and VCX.

Conclusions: Our findings suggest that SAD manifestation varies in scale and severity in different brain regions. We identify endotypes, such as energetic shortfalls, impaired neuronal function, and dedifferentiation.

Keywords: Alzheimer’s disease; NRF1; REST; SOX2; SUZ12; dedifferentiation; endotype; energetics; sporadic Alzheimer’s disease; transcriptome.

Fig. 1

Differential Gene Expression in SAD laser-dissected neurons. (A) Top: neuron samples from NDC (Braak I-II) and SAD patients (Braak III-VI) were collected from the entorhinal cortex (EC), hippocampus (HC), medial temporal gyrus (MTG), posterior cingulate (PC), superior frontal gyrus (SFG), and visual cortex (VCX) by Liang et al.; a dashed line around the brain region indicates interior anatomical location. Bottom: table of the number of control and afflicted samples alongside age and gender breakdown. (B) Volcano plots of differential gene expression across all 6 brain regions; the directional profile of DEGs is indicated in each sub-panel; the numbers indicate downregulated (blue) and upregulated (red) DEGs, respectively; the horizontal red line indicates an adjusted p-value < 0.05 cutoff for significance. (C) Quasi-proportional Venn diagram of shared and unique DEGs between each brain region.

Fig. 2

Functional enrichment, TF-target enrichment, and motif-based TF-activity prediction. Categories (gray): (1) Dedifferentiation, (2) Energetics, (3) Neuron Function, (4) Inflammation, (5) Stress Response. A-C) Geneset enrichment analysis by fgsea in all six brain regions using (A) Hallmark, (B) GOBP, and (C) ENCODE-ChEA Consensus (ECC) TF-gene target collection. GSEA score is the -log₁₀ of the adjusted p-value multiplied by the sign of the NES (net enrichment score). D) ISMARA TF activity prediction using the Swiss regulon motif database across all six brain regions. ISMARA score is a function of the z-score, the direction of mean TF target expression change, and the direction of Pearson correlation between TF gene expression and target gene expression. The red line delineates the ISMARA score of ±2, outside of which terms are considered significant.

Fig. 3

CEMiTool Co-expressed gene module identification. Categories (gray): (1) Dedifferentiation, (2) Energetics, (3) Neuron Function, (4) Inflammation, (5) Stress Response. A) Number of genes in co-expressed modules across all six brain regions. Modules with over 100 genes were selected for further functional and TF Target enrichment. B) Enrichment of each module across all brain regions using the fgsea multilevel test on limma t-value ranked gene lists. Enrichment utilized modules as the collection. C, D) The tmod hypergeometric enrichment for each module. C) Functional enrichment with Hallmark (H) or GOBP (G). D) TF-target enrichment with ENCODE (E) or ChEA (C) libraries.

Fig. 4

TF target differential gene expression for Module-specific TFs. A-D) Gene expression heatmaps for the targets of key TFs identified for modules M1, M3, and M4; (A) M1 REST targets in EC ∪ MTG ∪ PC ∪ VCX; (B) M3 SOX2 targets in EC ∪ MTG ∪ SFG; and (C) M4 NRF1 targets that are in HC ∪ PC ∪ VCX.

Fig. 5

Comparison of key regulators in HC from AD, Demented, and Control samples. For GSE4757: (A) multidimension scaling (MDS) comparing tangle-free and tangle samples; (B) ontological enrichment using GOBP library; and (C) TF-target enrichment using the ECC library. For Allen Brain Atlas samples: (D) hippocampus (HC) z-score expression values from Allen Brain Atlas’ ADT data for NRF1, REST, and SOX2 are plotted for two conditions (1) Non-demented control (NDC) (Braak I-II) and (2) probable AD samples (AD) (Braak III-VI). Traumatic brain injury samples were excluded for all conditions assessed.