Novel microglial transcriptional signatures promote social and cognitive deficits following repeated social defeat

Psychosocial stress is linked to anxiety, social withdrawal, and cognitive deficits. Prior work shows that repeated social defeat (RSD) stress in mice sensitizes microglia, monocytes, and neurons, driving neuroinflammation and monocyte recruitment to brain regions involved in fear and threat appraisal. Microglia reactivity and endothelial activation promote IL-1β signaling and anxiety-like behavior, whereas some behavioral effects (e.g., social avoidance of a novel aggressor and enhanced fear memory) can occur independently of microglia. The hippocampus mediates anxiety-like behavior and memory formation and exhibits neuronal sensitization after RSD, including elevated pCREB and IL-1R1-dependent effects in glutamatergic neurons. The present study asks how microglia influence single-cell transcriptional programs across hippocampal cell types during RSD, and how these molecular changes relate to microglia-dependent versus independent behavioral outcomes.

Human studies of chronic stress report heightened inflammatory transcriptional profiles in monocytes associated with depressive symptoms. In rodents, RSD produces long-term sensitization of microglia and neurons, endothelial activation, and recruitment of IL-1β+ monocytes that promote anxiety-like behavior. Microglia depletion with CSF1R antagonists attenuates stress-induced monocyte recruitment and anxiety-like behavior. Endothelial IL-1R1 signaling and prostaglandin pathways mediate monocyte-to-brain communication after RSD. Neuronal sensitization in the hippocampus includes increased pCREB and is mediated by IL-1R1 on Vglut2+ neurons; stress also enhances glutamatergic and neurotrophic signaling affecting fear memory. Some RSD-induced behaviors are microglia-independent (e.g., social avoidance of an aggressor, enhanced fear memory), highlighting both microglia-dependent and independent pathways. Prior single-cell studies emphasized microglial activation in stress and disease; however, a comprehensive single-cell and single-nuclei dissection of microglia–astrocyte–endothelial–leukocyte–neuron crosstalk during RSD was lacking.

Male C57BL/6 mice (6–8 weeks) were fed vehicle chow or PLX5622 (CSF1R antagonist; 1200 mg/kg) for 14 days to deplete microglia, and remained on diets throughout experiments (~21 days). RSD was induced by introducing a novel male CD-1 aggressor into the cage of experimental mice (3 per cage) for 2 h per night on six consecutive nights; submissive behaviors were monitored and aggressors replaced if needed. Fourteen hours after the final defeat, hippocampi were collected. For scRNAseq, single-cell suspensions were prepared using the Miltenyi Adult Mouse Brain Dissociation Kit; myelin was removed, RBCs lysed, and viability assessed. Pooled hippocampi (3 per group) were processed on 10x Genomics Chromium; libraries were prepared (v3.1), sequenced on Illumina HiSeq, and aligned to mm10 with CellRanger. Low-quality cells were filtered (including >20% mitochondrial RNA). Clustering used UMAP in Seurat; differential expression used MAST (p-adj < 0.05; FC ≥ 0.25). Cell types were annotated by canonical markers (e.g., microglia P2ry12; astrocytes Slc1a3; endothelia Cldn5; leukocytes Plac8; neurons Meg3). Pathway analyses used Gene Ontology (GO) and Ingenuity Pathway Analysis (IPA; z-score ≥ ±1.5). For snRNAseq, nuclei from pooled hippocampi (3 per group) were isolated by dounce, myelin removal, fixed (Parse Biosciences), split-pool barcoded (Whole Transcription Kit), and sequenced (NovaSeq S4; ~40,000 reads/nucleus). Reads were aligned to mm39 using the Parse pipeline; nuclei with >20% mitochondrial RNA were removed. UMAP clustering and marker-based annotation identified neuronal subtypes (DG Prox1; CA1 Mpped1; CA2/3 Ptpn5/Mndal; inhibitory Gad2; etc.). Behavioral assays 14 h after RSD included social avoidance (CD1 aggressor), social interaction (novel juvenile), and Y-Maze working memory; distance, time in zones, and spontaneous alternations were recorded. Immunohistochemistry measured ΔFosB (neuronal activation) and IBA1 (microglia) in the dentate gyrus; percent area, counts, and fluorescence intensity were quantified. Statistics used two-way ANOVA with Tukey’s post hoc (p < 0.05; trends p ≤ 0.10); experimenters were blinded. Data integration across two independent scRNAseq experiments (n=6 mice per group) and pooled snRNAseq samples was performed.

- scRNAseq of hippocampus captured 17,682 cells across Con-Veh, Stress-Veh, Con-PLX, Stress-PLX; 19 clusters identified with enrichment of microglia (59%), astrocytes (11%), and endothelia (15). - Microglia (P2ry12+) subclustering (n=6,589) revealed nine clusters (MC1–MC9). Stress increased stress-associated clusters MC2 (Inflammatory; Il1b, Ccl3, Ccl4, Nfkbia), MC3 (Cell stress; EIF2), MC4 (Phagocytosis; IEGs Jun, Fos), MC5 (Antigen presentation; S100 family), and decreased homeostatic MC8 (P2ry12, Ttr, Maf). CellChat indicated enhanced CCL and galectin signaling from stress-associated microglia. - Across microglia, stress upregulated Th2, autophagy, EIF2, P2YR, clathrin-mediated endocytosis, IL-6 signaling; upstream regulators included EIF4E, ZBTB16, MLXIPL, MYC, HNF4A, IL1B, IL6. Homeostatic genes (P2ry12, Sall1, Mef2c, Tgfb1, Tmem119) were reduced; cytokine/chemokine/complement genes (Ccl2, Ccl4, Il1a, Il1b, C1qa, Tnf) increased. SAMs were transcriptionally distinct from DAM1/2. - Leukocytes (Plac8+ or Cd3g+; n≈466–488) formed seven clusters with stress-driven increases in neutrophils (LC1: 14%→23%) and MO/monocytes (LC7: 0%→10%), decreases in BAMs (LC5: 41%→21%) and monocytes (LC6: 18%→8%); Il1b expression increased in neutrophils and monocyte/macrophage clusters. PLX groups had ~60% fewer leukocytes than vehicle. - Astrocytes (Slc1a3+; n=1,281) yielded four clusters: stress increased AC3 (44%→70%; neuronal support genes Spry2, Gria2, Slitrk2, Malat1) and AC4 (11%→25%; gliogenesis genes Gfap, Gpm6b, Sparc, Dner, S100a6), while decreasing AC1 (36%→3%) and AC2 (9%→3%). Microglia depletion prevented the stress-induced increase in AC3/AC4. Of 100 stress-induced DEGs, 78% (78) were reversed by PLX. IPA: oxytocin signaling, RAR activation, IL-8 signaling increased with stress and were microglia-dependent; oxidative phosphorylation changes were microglia-independent. - Endothelia (Cdh5+; n=1,659) formed six clusters. Stress increased EC2 (1%→14%) and EC3 (4%→12) and decreased EC4 (38%→12%). EC3 (stress-enriched) upregulated Ackr1, Lcn2, Lrg1, Vcam1, Ptgs2, Icam1, Il1r1; EC2 increased metabolic/vesicular genes. Of 244 stress-influenced DEGs, 71% (174) were microglia-dependent. IPA showed increased inflammatory signaling (IL-12, cytokine storm) and leukocyte extravasation with stress that were microglia-dependent; mitochondria dysfunction and IL-17A signaling were microglia-independent. - Behavioral outcomes: Splenomegaly increased after RSD (F(1,25)=26.13, p<0.0001), independent of PLX. ΔFosB in hippocampus showed main effects of stress (F(1,11)=5.702, p=0.0360) and intervention (F(1,11)=6.529, p=0.0267). IBA1 percent area increased with stress (F(1,11)=10.8, p=0.0073); IBA1+ cell counts increased (F(1,21)=4.144, p=0.05); PLX reduced IBA1+ cells (F(1,21)=108.6, p<0.001). Social avoidance of aggressor: time in interaction zone decreased by stress (F(1,32)=16.6, p<0.05) and was microglia-independent; corner time increased by stress (F(1,32)=4.512, p<0.05). Social interaction with juvenile: interaction tended to decrease with stress (F(1,24)=3.749, p=0.06) and was microglia-dependent (interaction F(1,24)=4.289, p=0.0493); corner time interaction F(1,24)=5.557, p=0.0269. Y-Maze: spontaneous alternation showed interaction (F(1,25)=6.028, p<0.05), with Stress-Veh lowest (p=0.06 trend); microglia depletion prevented the deficit. - snRNAseq (27,879 nuclei; neurons 86%; Snap25+ n=23,955) revealed robust neuronal responses to stress. The fraction of stress-induced DEGs reversed by PLX: DG 22%, CA2/3 35%, CA1 43%, inhibitory 51%. Microglia-dependent neuronal pathways included CREB, glutamatergic receptor, calcium, reelin (excitatory), and oxytocin and chemokine signaling (inhibitory); DNA methylation transcription repression and spliceosome cycle were largely microglia-independent.

RSD induces a distinct set of stress-associated microglia (SAMs) in the hippocampus characterized by cytokine/chemokine expression (e.g., Il1b, Ccl3/4), ER stress (EIF2), phagocytosis, and antigen presentation, with enhanced paracrine CCL and galectin signaling. These SAMs modulate astrocyte and endothelial transcriptional states and are associated with increased Il1b-expressing leukocytes (monocytes/macrophages and neutrophils), aligning with prior evidence that microglia orchestrate monocyte recruitment and IL-1R1/prostaglandin signaling at the neurovascular interface. The majority of stress-induced DEGs and pathways in astrocytes and endothelia were microglia-dependent, indicating that microglial activation is a primary driver of non-neuronal remodeling after stress. In neurons, stress elicited both microglia-dependent and independent changes; microglia specifically supported increases in CREB, glutamatergic, calcium, reelin, oxytocin, and chemokine signaling, whereas chromatin/splicing programs were less dependent on microglia. Behavioral data map onto these molecular patterns: microglia depletion reversed social withdrawal (juvenile interaction) and working memory deficits (Y-Maze), paralleling microglia-dependent excitatory neuronal signaling and CA1/DG pathway alterations, while social avoidance of an aggressor and splenomegaly remained microglia-independent. Together, the findings delineate microglial mechanisms linking stress to hippocampal circuit and glial remodeling that underlie specific social and cognitive outcomes.

Single-cell and single-nuclei transcriptomics reveal novel stress-associated microglia in the hippocampus that drive widespread, microglia-dependent transcriptional changes in astrocytes and endothelia and selectively shape neuronal signaling pathways after repeated social defeat. Microglia-dependent modulation of neuronal CREB, glutamatergic, calcium, and oxytocin signaling aligns with reversal of stress-induced social withdrawal and working memory deficits upon microglia depletion, whereas social avoidance behavior is microglia-independent. These data define cell-type-specific microglia–brain communication pathways that couple stress to social and cognitive dysfunction. Future work should assess long-term memory domains, delineate inhibitory neuron subtypes, examine females, and integrate high-resolution microglial morphology with single-cell states to further resolve SAM functions.

Only male mice were used, limiting generalizability across sexes. scRNAseq provided low neuronal resolution, necessitating snRNAseq for neuronal profiling. Enzymatic dissociation may induce immediate early gene signatures, potentially confounding some microglial clusters. Single-cell counts do not reflect absolute cell numbers, complicating comparisons of leukocyte abundance. PLX5622 may affect long-lived tissue macrophages, posing off-target considerations. Inhibitory neuron subtypes were not fully resolved. Long-term and reference memory tasks (e.g., Morris water maze, novel object localization) were not assessed.