Transcriptomic profiling of sporadic Alzheimer's disease patients

Abstract

Alzheimer's disease (AD) manifested before age 65 is commonly referred to as early-onset AD (EOAD) (Reitz et al. Neurol Genet. 2020;6:e512). While the majority (> 90%) of EOAD cases are not caused by autosomal-dominant mutations in PSEN1, PSEN2, and APP, they do have a higher heritability (92-100%) than sporadic late-onset AD (LOAD, 70%) (Wingo et al. Arch Neurol. 2012;69:59-64, Fulton-Howard et al. Neurobiol Aging. 2021;99:101.e1-101.e9). Although the endpoint clinicopathological changes, i.e., Aβ plaques, tau tangles, and cognitive decline, are common across EOAD and LOAD, the disease progression is highly heterogeneous (Neff et al. Sci Adv Am Assoc Adv Sci. 2021;7:eabb5398). This heterogeneity, leading to temporally distinct age at onset (AAO) and stages of cognitive decline, may be caused by myriad combinations of distinct disease-associated molecular mechanisms. We and others have used transcriptome profiling in AD patient-derived neuron models of autosomal-dominant EOAD and sporadic LOAD to identify disease endotypes (Caldwell et al. Sci Adv Am Assoc Adv Sci. 2020;6:eaba5933, Mertens et al. Cell Stem Cell. 2021;28:1533-1548.e6, Caldwell et al. Alzheimers Demen. 2022). Further, analyses of large postmortem brain cohorts demonstrate that only one-third of AD patients show hallmark disease endotypes like increased inflammation and decreased synaptic signaling (Neff et al. Sci Adv Am Assoc Adv Sci. 2021;7:eabb5398). Areas of the brain less affected by AD pathology at early disease stages-such as the primary visual cortex-exhibit similar transcriptomic dysregulation as those regions traditionally affected and, therefore, may offer a view into the molecular mechanisms of AD without the associated inflammatory changes and gliosis induced by pathology (Haroutunian et al. Neurobiol Aging. 2009;30:561-73). To this end, we analyzed AD patient samples from the primary visual cortex (19 EOAD, 20 LOAD) using transcriptomic signatures to identify patient clusters and disease endotypes. Interestingly, although the clusters showed distinct combinations and severity of endotypes, each patient cluster contained both EOAD and LOAD cases, suggesting that AAO may not directly correlate with the identity and severity of AD endotypes.

Fig. 1

A Sex, age at onset (AAO), and age at death (AAD) of early-onset sporadic AD (onset age < 60 years), late-onset sporadic AD (onset age > 70 years), and nondemented control (NDC) patient Occipital Lobe samples. B Cluster dendrogram of all AD samples based on the expression of genes (8934) with 10 cpm across all samples. C DEqual plot of correlation between differential expression (relative to NDC) and reference patient brain RNA degradation in the 4 clusters. D Multi-Dimensional Scaling (MDS) plot of all patient samples for the top two dimensions. E Transcriptome assignment % (kallisto) across the sample groups. F AAO (left) and AAD (right) across the sample groups. G RNA-seq Volcano plots for the three AD clusters. Left, downregulated DEGs; right, upregulated DEGs. H Ranked enrichment analysis of gene expression signatures for the three AD clusters using the GOBP, Hallmark, Reactome, and StringDB databases by the tmod CERNO (left) and fgsea enrichment test (right); plotted data indicates adj. P < 0.05. I TFs with predicted significant activity change by ISMARA motif analysis curated into canonical ontological categories; [z-score] > 2 in at least one cluster shown. J GSVA heatmap, dendrogram, and gene size of the 22 co-expression modules identified by CEMiTool across the sample groups. K Camera enrichment analysis of the 22 co-expression modules in the three clusters relative to NDC; * = adj. P < 0.05. L StringDB PPI interaction networks for ontologically- and expression-related comodules across the three clusters; genes color-coded by limma t-value; upper right subpanel indicates number of DEGs within the comodule for each AD cluster. M Top enriched pathways and TFs for each comodule (hypergeometric test)