Spatially-resolved integration of microglia morphological diversity and gene expression using Visium with protein co-detection

Microglia exhibit a dynamic range of morphologies that actively shift in response to local environmental cues. There has been a concerted effort as a field to move away from dualistic characterization of all microglia as resting or activated, which is often described in terms of their morphological differences alone, towards more integrated definitions of microglial states based on multiple data types. Recent advances in spatially-resolved transcriptomics offer new ways to understand microglial gene expression in situ. Here, we related microglia morphology to gene expression within a published, spatially-resolved proteogenomics (Visium-SPG) dataset of the human dorsolateral prefrontal cortex as proof of principle to combine the contributions of microglia morphology to regional transcriptomics differences in the brain. The published dataset uniquely combined spatial transcriptomics with immunofluorescent staining for four major brain cell types, including microglia, which enabled us to link microglial morphologies to gene expression in a spatially-resolved manner. Using a computational toolset, MicrogliaMorphology and MicrogliaMorphologyR, we classified individual microglia into morphological subtypes, assigned microglia to spots, and integrated these data with transcriptomic profiles for each Visium-SPG spot. We applied non-negative matrix factorization to account for contributions from multiple cell types, morphological features, and regional context in downstream differential expression modeling. Our approach offers a methodological framework for investigating the relationship between microglial morphology and gene expression in larger cohorts and within disease contexts where changes in microglia morphology are anticipated. Author summaryRecent advances in spatially-resolved transcriptomics and the development of methods to deconvolve cell type signals has offered new ways to understand microglial gene expression in various tissue contexts. In this work, we related microglia morphologies to gene expression within a published spatial transcriptomics dataset of the human dorsolateral prefrontal cortex to assess whether morphologically different microglia are transcriptomically different in the healthy human brain. This dataset uniquely combines spatial transcriptomics with immunofluorescent staining for four major brain cell types, including microglia, enabling the linkage of microglial morphologies to gene expression in a spatially-resolved manner. Using a computational toolset for microglia morphology analysis, we classified individual microglia into morphological subtypes and integrated these data with spatial transcriptomic profiles while accounting for contributions from multiple cell types, morphological features, and regional context. This allowed for the identification of microglia morphology-related genes that are differentially expressed between microglial subtypes and relevant to human brain disorders. Findings together provide proof of principle for exploring the relationship between microglial morphology and gene expression in larger cohorts and diverse experimental contexts.