Search

Astrocytes have been shown to play a central role in Alzheimer’s Disease (AD). However, the genes and biological pathways underlying disease manifestation are unknown, and it is unclear whether regional molecular differences among astrocytes contribute to regional specificity of disease. Here, we began to address these challenges with integrated experimental and computational approaches. We constructed a human astrocyte-specific functional gene network using Bayesian integration of a large compendium of human functional genomics data, as well as regional astrocyte gene expression profiles we generated in the mouse. This network identifies likely region-specific astrocyte pathways that operate in healthy brains. We leveraged our findings to compile genome-wide astrocyte-associated disease-gene predictions, employing a novel network-guided differential expression analysis (NetDIFF). We also used this data to predict a list of astrocyte-expressed genes mediating region-specific human disease, using a network-guided shortest path method (NetPATH). Both the network and our results are publicly available using an interactive web interface athttp://astrocyte.princeton.edu. Our experimental and computational studies propose a strategy for disease gene and pathway prediction that may be applied to a host of human neurological disorders.

Alzheimer’s disease is a multifactorial disorder with a heterogeneous patient population. Comorbidities such as hypertension, hypercholesterolemia and diabetes are known contributors to the disease progression. Indeed, therapies targeting these disorders have been shown efficient in dementia prevention. However, their mechanistic contribution to Alzheimer’s pathology and neurodegeneration has not been fully clarified.In the current study, we used CSF samples from a memory clinic cohort of 90 patients without diagnosed hypertension, hypercholesterolemia, or diabetes nor other neurodegenerative disorder, to investigate 13 molecular markers representing key mechanisms underlying Alzheimer’s pathogenesis. Levels were compared between clinical groups of subjective cognitive decline, mild cognitive impairment, and Alzheimer’s disease. Associations between markers and groups of markers were analyzed by linear regression. Two-step cluster analysis was used to determine patient clusters. Two key markers were further analyzed by immunofluorescence staining in hippocampus from control and AD individuals without hypertension, hypercholesterolemia nor diabetes.CSF angiotensinogen, thioredoxin-1, and interleukin-15 were the biomarkers with the most prominent associations with Alzheimer’s pathology, synaptic and axonal damage. Synaptosomal-associated protein 25 kDa and neurofilament light chain were increased in mild cognitive impairment and Alzheimer cases. When we grouped biomarkers by biological function, we found that inflammatory and survival components were associated with Alzheimer’s pathology, synaptic dysfunction and axonal damage. Moreover, a vascular/metabolic component was associated with synaptic dysfunction. In data-driven analysis, two patient clusters were identified; Older participants with increased CSF markers of oxidative stress, vascular pathology and neuroinflammation were assigned to cluster 1, that was also smaller and characterized by increased synaptic and axonal damage, compared to individuals in cluster 2. Clinical groups were evenly distributed between the clusters.Analysis of post-mortem hippocampal tissue, showed that, compared to controls, angiotensinogen staining was higher in Alzheimer’s disease and was also found to co-localize with phosphorylated-tau.In a population free of common comorbidities, we could still find associations between Alzheimer’s disease biomarkers and markers of pathways associated with increased risk for Alzheimer’s disease (i.e., neuroinflammation, vascular function, oxidative stress and cholesterol homeostasis), suggesting that these pathways are contributing to Alzheimer’s disease mechanisms even in absence of clinically diagnosed comorbidities. The identification of distinct biomarker-driven endophenotypes of cognitive disorder patients, further highlights the biological heterogeneity of Alzheimer’s disease and the importance of developing tailored prevention and treatment strategies.

SUMMARYNeurons from layer II of the entorhinal cortex (ECII) are the first to accumulate tau protein aggregates and degenerate during prodromal Alzheimer’s disease. Here, we use a data-driven functional genomics approach to model ECII neuronsin silicoand identify the proto-oncogene DEK as a potential driver of tau pathology. By modulating DEK levels in EC neuronsin vitroandin vivo, we first validate the accuracy and cell-type specificity of our network predictions. We then show thatDeksilencing changes the inducibility of immediate early genes and alters neuron excitability, leading to dysregulation of neuronal plasticity genes. We further find that loss of function of DEK leads to tau accumulation in the soma of ECII neurons, reactivity of surrounding microglia, and eventually microglia-mediated neuron loss. This study validates a pathological gene discovery tool that opens new therapeutic avenues and sheds light on a novel pathway driving tau pathology in vulnerable neurons.

Oxidative stress and neuroinflammation play a central role in Alzheimer’s Disease (AD) pathogenesis. However, the mechanism by which these processes lead to neurodegeneration is still not fully understood. Thioredoxin-1 (Trx1) is an antioxidant protein that can be cleaved into a peptide known as Thioredoxin-80 (Trx80), which modulates monocyte function in the periphery and shows anti-amyloidogenic properties in the brain. In this study we aimed to further clarify the biological function of this peptide and its regulation in the brain. We show that neurons are the main producers of Trx80 in the brain. Trx80 levels increase in vivo both in normal aging and in young APPNL-G-F mouse model of amyloid pathology. Trx80 levels were increased in neurons in primary culture treated with either rotenone or 27-hydroxycholesterol, what suggests that Trx80 production is stimulated upon oxidative stress. RNA-sequencing followed by differential gene expression analysis revealed that Trx80 induces microglia activation into a phenotype compatible with interferon response microglia. Finally, we determine that the induction of this microglia phenotype by Trx80 is Trem2-dependent. This study identifies Trx80 as a novel neuron-derived signaling mechanism that modulates microglia function under stress conditions. Strategies to regulate Trx80 levels could be beneficial against AD pathology.