Microglia in autism spectrum disorder: heterogeneity, immunometabolism, and synapse-related pathways

Microglia are increasingly implicated in autism spectrum disorder (ASD), but their role remains difficult to define because the available evidence is heterogeneous in cohort composition, developmental stage, sampled brain region, and experimental modality. This Review summarizes current evidence on three related aspects of ASD-relevant microglial biology: microglial heterogeneity, immunometabolic regulation, and synapse-related pathways. Human postmortem studies, bulk transcriptomics, single-cell and spatial atlases, methylomic deconvolution, and in vivo neuroimmune imaging collectively support the presence of immune- and glia-associated alterations in at least a subset of ASD brains, but these findings do not support a single ASD-wide microglial phenotype. Instead, current evidence is more consistent with region-, stage-, sex-, and context-dependent microglial variation that should be interpreted together with neuronal, astrocytic, vascular, and broader tissue-level changes. We further review how lipid handling, mitochondrial function, phagocytic-lysosomal load, and bioactive lipid signaling may influence microglial competence in ASD-relevant settings, while noting that much of the detailed mechanistic immunometabolism literature still derives from aging and neurodegeneration. At the microglia-synapse interface, complement deposition, phosphatidylserine exposure, anti-engulfment checkpoints, and astrocyte-microglia crosstalk provide more informative mechanistic entry points than broad activation terminology. Across studies, the major challenge is not whether microglia are involved in ASD, but how to distinguish primary pathogenic effects from secondary adaptation, and how to relate molecular signatures to excessive, insufficient, or mistargeted synaptic remodeling. Overall, the literature supports a more precise interpretation of ASD-related microglial biology based on developmental timing, cellular context, and mechanism-linked readouts rather than non-specific inflammatory labels alone.