Multi-transcriptomic analysis points to early organelle dysfunction in human astrocytes in Alzheimer's disease

The phenotypic transformation of astrocytes in Alzheimer's disease (AD) is still not well understood. Recent analyses based on single-nucleus RNA sequencing of postmortem Alzheimer's disease (AD) samples are limited by the low number of sequenced astrocytes, small cohort sizes, and low number of differentially expressed genes detected. To optimize the detection of astrocytic genes, we employed a novel strategy consisting of the localization of pre-determined astrocyte and neuronal gene clusters in publicly available whole-brain transcriptomes. Specifically, we used cortical transcriptomes from 766 individuals, including cognitively normal subjects (Controls), and people diagnosed with mild cognitive impairment (MCI) or dementia due to AD. Samples came from three independent cohorts organized by the Mount Sinai Hospital, the Mayo Clinic, and the Religious Order Study/Memory and Aging Project (ROSMAP). Astrocyte- and neuron-specific gene clusters were generated from human brain cell-type specific RNAseq data using hierarchical clustering and cell-type enrichment scoring. Genes from each cluster were manually annotated according to cell-type specific functional Categories. Gene Set Variation Analysis (GSVA) and Principal Component Analysis (PCA) were used to establish changes in these functional categories among clinical cohorts. We highlight three novel findings of the study. First, individuals with the same clinical diagnosis were molecularly heterogeneous. Particularly in the Mayo Clinic and ROSMAP cohorts, over 50% of Controls presented down-regulation of genes encoding synaptic proteins typical of AD, whereas 30% of patients diagnosed with dementia due to AD presented Control-like transcriptomic profiles. Second, down-regulation of neuronal genes related to synaptic proteins coincided, in astrocytes, with up-regulation of genes related to perisynaptic astrocytic processes (PAP) and down-regulation of genes encoding endolysosomal and mitochondrial proteins. Third, down-regulation of astrocytic mitochondrial genes inversely correlated with the disease stages defined by Braak and CERAD scoring. Finally, we interpreted these changes as maladaptive or adaptive from the point of view of astrocyte biology in a model of the phenotypical transformation of astrocytes in AD. The main prediction is that early malfunction of the astrocytic endolysosomal system, associated with progressive mitochondrial dysfunction, contribute to Alzheimer's disease. If this prediction is correct, therapies preventing organelle dysfunction in astrocytes may be beneficial in preclinical and clinical AD.