Automated image segmentation uncovers the role of CD74^high human microglia in cognitive decline

The role of activated microglia in Alzheimer's disease (AD) is well established; the proportion of stage III activated microglia has been associated with AD and cognitive decline, but this morphologically defined subtype is relatively uncommon (1-2% of microglia) and its cellular function is unknown. Single-cell RNA-sequencing revealed CD74 as a marker gene that is enriched in immunologically active microglial subtypes associated with AD. Here, we evaluated the relationship between CD74 expression, AD-related traits, and microglial morphology using dorsolateral prefrontal cortex samples from two brain collections (ROSMAP: n=63, NYBB: n=91). An image segmentation pipeline using CellProfiler was developed to extract features from entire tissue sections. The pipeline automatically delineated gray and white matter regions and segmented 1,120,780 gray matter microglia. In a meta-analysis of the two datasets, we find an increase in frequency of microglia with high CD74 expression (CD74^high) in relation to AD dementia (p = 0.038), particularly in the phase of terminal, accelerated cognitive decline before death. These microglia have a more rounded, amoeboid shape (ROSMAP: p = 1.4×10^-6; NYBB: p = 2×10^-13) which is a characteristic morphology of activated stage III microglia. Results were consistent across both datasets, highlighting the robustness of our cellular segmentation approach. This study identifies a potential role for CD74^high microglia and the CD74 ligand MIF in cognitive decline, and it provides evidence for a partially overlapping but distinct role for CD74^high microglia and morphologically defined stage III microglia, whose functional properties have remained poorly understood. These CD74^high microglia appear to be enriched for genes involved in cytokine response for class I and II antigen presentation, as well as regulation of T cell proliferation. These findings begin to link microglial subtypes defined by single-cell transcriptomic data with those characterized by classical morphological criteria to resolve the roles of different microglial functions to distinct stages in the trajectory to AD.