Resting-state fMRI-based machine learning for predicting SSRI treatment response in major depressive disorder.

Alzheimer's disease (AD) is the most prevalent form of dementia, characterized by neuropathological aggregation of amyloid‐beta (Aβ) plaques and neurofibrillary tau tangles (Knopman et al. ). Work over the past two decades has highlighted the important role of neuroinflammation in AD pathophysiology (Brown et al. ; Butovsky et al. ; Heneka et al. ). Microglia, the resident immune cells in the brain parenchyma (Paolicelli et al. ), are implicated in AD, as further suggested by their enriched expression of AD risk genes identified via genome‐wide association (Karch and Goate ; Sudwarts and Thinakaran ; Nott et al. ). Microglia also respond strongly to both amyloid‐beta (Aβ) plaques (Heneka et al. ), and tau pathology (Falkon et al. ; Wang et al. ). The phagocytosis and neuroinflammation induced in microglia by these pathological features likely result in their contribution to the neurodegeneration seen in AD (Gao et al. ; Taddei et al. ). Advancements in (single cell) transcriptomic technologies have helped to better understand how microglia respond to and contribute to neurodegenerative diseases such as AD. Most of this work involved characterizing the spectrum of microglial transcriptional states. This includes work performed on mouse models of amyloidosis (Keren‐Shaul et al. ; Krasemann et al. ), but also work from postmortem human brain samples (Mathys et al. , ; Srinivasan et al. ; Sun et al. ), which has resulted in the identification of gene signatures enriched in microglia derived from diseased donors. Our group previously contributed to this by performing single‐nucleus RNA sequencing (snRNAseq) on frozen postmortem samples enriched for non‐neuronal and non‐oligodendrocyte lineage nuclei using fluorescence activated nuclear sorting (Gerrits et al. ). Two distinct microglial states were identified, whose relative abundance correlated with the extent of Aβ and tau pathology, respectively. These microglial states might be expected to occupy different niches (van Olst et al. ), and differentially interact with the local microenvironment they are in. This includes how they communicate with other cells in the CNS, which can be bioinformatically inferred based on high dimensional snRNAseq data (Almet et al. ; Jin et al. , ). However, because of the glial enrichment performed in the original study, we could not properly interrogate how microglia‐associated cell–cell interactions are affected in disease. To this end, we supplemented our original snRNAseq dataset (Gerrits et al. ) with newly generated neuron‐ and oligodendrocyte‐enriched snRNAseq data and used bioinformatic tools to predict cell–cell interactions. After snRNAseq data integration, CellChat (Jin et al. , ) was used to predict changes in cell–cell communication, and multiple signaling pathways were predicted to be altered in AD donors. In a separate set of control and AD tissue samples, RNAscope and immunofluorescence validations confirmed increased microglial expression of a component of one of these pathways, which involved .