Graph-based subclustering was performed for several cell types separately, using data from the 143 samples. There were four subclusters of microglia, which roughly correspond to different states of microglial activation (Fig. 2C, Extended Data Fig. 2, Supplementary Tables 3-4). Subcluster 1 (“Resting Microglia”) uniquely expressed genes specific to quiescent microglia, such as P2RY12 and CX3CR1, and was enriched for pathways relating to microglia migration. Subclusters 2 and 3 were both enriched for immune response-related genes, with subcluster 2 (“Inflammatory Microglia”) highly expressing genes involved in inflammation, such as TLR2. Subcluster 3 (“CD83+ Microglia”) was enriched for genes governing microglia activation, such as CD83.39 Subcluster 4 (“Phagocytosing Microglia”) was marked by high expression of genes involved in endocytosis and phagocytosis. There was a significant increase in the mean proportion of “Inflammatory Microglia” (subcluster 2) in individuals with AUD: 31%, as opposed to 23% in those without AUD (adjusted p value (padj) = 0.027). Also, 70% of individuals that had at least half of their microglia cells in the inflammatory state (subcluster 2) had AUD, while only 52% of individuals who had below half of their microglia in the inflammatory state had AUD (Fig. 2D).
