MFN2-associated astrocyte dysfunction contributes to neuronal hyperexcitability and depressive-like behaviors Authors: Feng-xia Ma, Min Wang, Shuo-yan Fang, Xiao-xin Wang, Shen-han Xu, Xin-xin Zhao, Yu-xiao Liu, Cong Wang, Chen Qiao, Qiu-ling Wang, Lei Cao Published: 2026 ABSTRACT Astrocyte pathology is increasingly recognized as a key factor in the pathogenesis of major depressive disorder (MDD), where disrupted glutamate clearance and impaired neuron-glia interactions contribute to synaptic dysfunction. However, the molecular mechanisms leading to astrocyte loss remain largely undefined. Mitochondrial dysfunction is a hallmark of MDD, yet most studies have focused on neuronal mitochondria, leaving astrocytic mitochondrial dynamics insufficiently investigated. Here, we identify mitofusin-2 (MFN2)–associated astrocyte dysfunction in depression. SMART-seq profiling of hippocampal astrocytes from chronic social defeat stress (CSDS) mice revealed Mfn2 up-regulation, accompanied by excessive mitochondrial fusion, impaired mitophagy, and astrocyte loss. Elevated MFN2 expression was detected in plasma, postmortem brain tissue, and iPSC-derived astrocytes from patients with MDD. Astrocyte-specific knockdown of Mfn2 restored mitochondrial morphology and mitophagy, thereby preserving astrocyte viability and alleviating depressive-like behaviors. Using mass spectrometry–based label-free proteomics, in vivo calcium imaging, and functional assays, we demonstrated that β-arrestin2 deficiency reduced STAT1 stability, releasing its transcriptional repression on Mfn2. Normalization of MFN2 expression through astrocytic Mfn2 knockdown or Stat1 overexpression reduced extracellular glutamate accumulation and prevented neuronal hyperexcitability in CSDS mice. Collectively, our findings define astrocytic MFN2-associated astrocyte dysfunction as a central pathological event in depressive behaviors, with the β-arrestin2–STAT1 pathway acting as an upstream regulatory mechanism. This work reveals an astrocytic pathway in depression and highlights a potential therapeutic target that directly addresses astrocyte pathology, beyond the neuron-centered focus of current antidepressant treatments. INTRODUCTION Mental disorders represent a leading cause of global disease burden, with depressive disorders contributing the greatest proportion (1, 2). Major depressive disorder (MDD) is among the most disabling psychiatric conditions. Multiple hypotheses have been proposed to explain its pathophysiology, including neurotransmitter/receptor imbalance, hypothalamic-pituitary-adrenal axis dysregulation, cytokine activation, impaired neuroplasticity, and systemic influences (3, 4). However, available antidepressants targeting these mechanisms remain limited by low efficacy, delayed onset, high relapse rates, and substantial side effects (5–7). These shortcomings highlight that current theories do not fully capture the molecular complexity of MDD, underscoring the need for mechanistic studies to identify promising therapeutic targets. Astrocytes comprise roughly one-third of all brain cells and are deeply integrated into neuronal circuits through structural, metabolic, and signaling interactions (8). Mounting evidence from both human and animal studies has revealed a reduction in astrocyte number or density in depression (9). Astrocyte dysfunction impairs neuron-glia communication by disrupting gap junctions, glutamate clearance, and energy metabolism (7). For instance, reduced expression of glutamate transporters has been observed in mouse models of depression (10), while knockdown of the excitatory amino acid transporters EAAT1/2 induces depressive-like phenotypes through neuronal hyperactivation (11). These studies highlight astrocytes as active participants in MDD pathophysiology across molecular, cellular, and circuit levels. Nevertheless, the molecular mechanisms driving astrocyte loss remain poorly understood. Elucidating these pathways may uncover previously unidentified mechanistic insights into MDD and open avenues for glia-targeted therapies beyond conventional neuron-centered approaches. Mitochondrial dysfunction, characterized by impaired bioenergetics, oxidative stress, and altered signaling, has emerged as a major contributor to MDD (12). Dysregulated mitochondrial dynamics, mitophagy, and apoptosis have been observed in peripheral blood mononuclear cells from patients with MDD (13). In the central nervous system, mitochondrial dysfunction has been extensively investigated in neurons, which are closely linked to disrupted synaptic function and plasticity (14). Chronic stress promotes mitochondrial fission in neurons of the medial prefrontal cortex (mPFC), resulting in mitochondrial dysfunction and impaired synaptic transmission (15). Mitophagy defects have also been reported in inhibitory neurons of the mPFC, contributing to depressive-like behaviors (16). Although mitochondrial abnormalities have been documented in astrocytes under depressive conditions (17), the mechanistic link between astrocytic mitochondrial dynamics and depression pathogenesis remains largely unexplored. In the present study, we performed SMART sequencing (SMART-seq) profiling of hippocampal astrocytes and identified MFN2-associated astrocyte dysfunction in chronic social defeat stress (CSDS) mice. Mechanistically, β-arrestin2–signal transducer and activator of transcription 1 (STAT1) signaling acts upstream to suppress Mfn2 expression, and disruption of this β-arrestin2–STAT1–MFN2 axis in astrocytes leads to glutamate accumulation, neuronal hyperexcitability, and depressive-like behaviors. Astrocyte-specific overexpression of Stat1 or knockdown of Mfn2 restored mitochondrial function and neuronal activity, alleviating depressive phenotypes. Together, our findings identify the β-arrestin2–STAT1–MFN2 axis as a critical regulator of astrocytic mitochondrial homeostasis and a promising therapeutic target for MDD. RESULTS Astrocyte loss correlates with depression-like behaviors in the CSDS mouse model To investigate astrocyte pathology in depression, we established a CSDS mouse model. Following CSDS exposure, a battery of behavioral assays, including the social interaction test (SIT), open-field test (OFT), tail suspension test (TST), forced swim test (FST), and sucrose preference test (SPT), was conducted (fig. S1A). Mice categorized as susceptible (SUS) to CSDS displayed hallmark depression-like behaviors, including reduced time spent in the interaction zone and a decreased social interaction (SI) ratio in the SIT, diminished time in the center zone during the OFT, increased immobility in both the TST and FST, and lower sucrose preference in the SPT. In contrast, resilient (RES) mice exhibited no apparent depressive phenotypes and were therefore excluded from further analyses (fig. S1, B to H). Postmortem analyses of patients with major depressive disorder (MDD) have consistently revealed astrocytic loss across multiple brain regions, including the hippocampus, mPFC, anterior cingulate cortex (ACC), and basolateral amygdala (BLA) (9). To determine whether similar astrocytic alterations occur in the CSDS model, we performed immunofluorescence staining for glial fibrillary acidic protein (GFAP) and glutamine synthetase (GS) across these MDD-associated brain regions. Consistent with human data, we observed a pronounced reduction in astrocyte number within the hippocampus, prelimbic cortex, ACC, and BLA of CSDS mice (Fig. 1A and fig. S2A). Notably, the hippocampus exhibited the most substantial loss, particularly within the CA1, CA3, and dentate gyrus (DG) subregions (Fig. 1A). Astrocyte numbers in these hippocampal subregions were inversely correlated with depression-like behaviors, supporting a link between astrocyte depletion and depressive phenotypes (Fig. 1, B to F, and fig. S2, B to K). To further elucidate the molecular underpinnings of astrocyte pathology, hippocampal astrocytes were isolated and subjected to SMART-seq transcriptomic profiling (Fig. 1G). This analysis identified 734 differentially expressed genes, including 624 up-regulated and 110 down-regulated (Fig. 1H and fig. S2L). Gene ontology enrichment revealed significant associations with mitochondrial dysfunction and autophagy-related pathways (Fig. 1I and fig. S2M). Notably, genes implicated in mitochondrial dynamics exhibited robust dysregulation in astrocytes from CSDS mice (Fig. 1J), suggesting that perturbations in mitochondrial homeostasis may contribute to astrocyte dysfunction in the context of depression. Imbalance of mitochondrial dynamics contributes the loss of astrocytes in the hippocampus of CSDS mice. (A) Immunofluorescence staining of GFAP and GS, and quantification of GFAP+ and GS+ cells in the subregions of hippocampus including CA1, CA3, and DG in control/CSDS mice; n = 5 mice. Scale bars, 50 μm. ***P < 0.001 by unpaired Student’s t test. (B to F) Correlation analysis of astrocyte count in DG with depression-like behaviors; n = 10 mice. Pearson correlation coefficients (r). (G) Isolation of astrocytes from hippocampus tissue for SMART-seq. Created in BioRender. W. Cong (2026), https://BioRender.com/ohcvaai. (H) Volcano plot showing the differentially expressed genes (DEGs) detected by SMART-seq; n = 3 mice. (I) Gene Ontology enrichment analysis showing the involvement of mitochondrial dysfunction. BP, biological process; CC, cellular component; MF, molecular function; ATPase, adenosine triphosphatase. (J) Heatmap of DEGs associated with mitochondrial fusion/fission dynamics. MFN2-associated mitochondrial dysfunction impairs autophagy and promotes astrocyte pathology in depression Next, we explored the molecular basis of astrocyte pathology by examining the expression of genes regulating mitochondrial dynamics in the hippocampus. Reverse transcription quantitative polymerase chain reaction (RT-qPCR) analysis revealed increased expression of mitochondrial fusion-related genes, including Arl2, Mfn1, and Mfn2, as well as the fission-associated gene Rab24 (Fig. 2, A and B). Among these, Mfn2 exhibited the most robust up-regulation in CSDS mice. Western blotting analysis confirmed elevated MFN2 protein levels in the hippocampus following CSDS exposure (Fig. 2C). To assess the cell-type–specific expression of Mfn2, we performed RNAscope in situ hybridization combined with immunofluorescence staining. Consistent with transcriptomic data, Mfn2 expression was markedly increased in GFAP+ astrocytes from CSDS mice (Fig. 2D), whereas no significant changes were detected in NeuN+ neurons or IBA1+ microglia (Fig. 2, E to H). Western blotting analysis showed significant of MFN2 up-regulation and GFAP down-regulation in hippocampal astrocytes of CSDS mice (fig. S3, A and B). Ultrastructural analysis by transmission electron microscopy further revealed enlarged mitochondrial area without a corresponding change in density, indicating pronounced mitochondrial fusion in the hippocampus of CSDS mice (Fig. 2I). In parallel, autophagy impairment was evident, as reflected by elevated levels of P62 and LC3B (Fig. 2J). Given that mitochondrial hyperfusion is known to suppress mitophagy (18), these results suggest that MFN2-associated mitochondrial dysfunction contributes to astrocyte pathology by inhibiting mitophagy under depressive conditions. Notably, up-regulation of MFN2 was also observed in multiple sample types derived from patients with MDD, including postmortem brain tissue, plasma, and iPSC-derived astrocytes (Fig. 2K), supporting a conserved role of MFN2 dysregulation in human depression. Astrocytic MFN2 is up-regulated in CSDS mice and patients with MDD. (A and B) RT-qPCR analyses of mitochondrial fusion/fission-associated genes in the hippocampus of control/CSDS mice; n = 5 mice. (C) Representative blots and quantification of MFN2 in the hippocampus tissue; n = 6 mice. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. (D to H) Examination of Mfn2 expression in DG region within astrocytes (GFAP positive), neurons (NeuN positive), and microglia (IBA1 positive) by RNAscope assays; n = 6 mice. Scale bars, 25 μm (D) and 50 μm (E and F). (I) Observation and quantification of mitochondrial morphology by transmission electron microscopy; n = 23 mitochondria for mitochondrial area and n = 6 fields for mitochondrial density. Scale bars, 1 μm. (J) Representative blots and quantification of P62 and LC3B in the hippocampus tissue; n = 6 mice. (K) Analyses of GEO datasets from healthy individuals and patients with MDDs. The samples in GSE214921 [RNA sequencing (RNA-seq) from postmortem brain tissue], GSE217811 (RNA-seq from plasma exosomes), and GSE178071 (RNA-seq from iPSC-derived astrocytes) were derived from patients with MDD and matched healthy controls. The sample sizes in the GSE214921 and GSE217811 datasets correspond to the number of individual patients, whereas the GSE178071 dataset includes iPSC-derived astrocyte cell lines from six patients with MDD and three healthy controls. *P < 0.05, **P < 0.01, and ***P < 0.001 by unpaired Student’s t test; n.s., not significant. To directly test this hypothesis, we generated astrocyte-specific Mfn2 knockdown mice by stereotaxic injection of rAAV-GfABC1D-EGFP-shMfn2 into the hippocampus, followed by CSDS and behavioral assessment (Fig. 3A and fig. S4, A and B). Robust viral expression was detected in GFAP+ astrocytes within the hippocampal slices (Fig. 3B and fig. S4, C and D). Astrocytic Mfn2 knockdown alleviated depression-like behaviors across all behavioral paradigms, including SIT, OFT, TST, FST, and SPT (Fig. 3, C to J). Immunofluorescence analysis further revealed that Mfn2 knockdown prevented CSDS-induced astrocyte loss (Fig. 3, K and M). Consistently, both terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL) and cleaved-caspase3 staining revealed that Mfn2 knockdown significantly reduced astrocyte apoptosis in CSDS mice (fig. S4, E to H). Furthermore, astrocytic Mfn2 knockdown normalized mitochondrial morphology and autophagy markers in CSDS conditions (Fig. 3, L and N to P, and fig. S4, I to K). To further validate the role of astrocytic MFN2 in depression, we generated astrocyte-specific Mfn2 overexpression mice by hippocampal injection of rAAV-GfaABC1D-EGFP-Mfn2 (fig. S5, A to C). Consistent with the effects observed under CSDS conditions, Mfn2 overexpression in astrocytes induced depression-like behaviors (fig. S5, D to J). Additionally, we observed astrocyte apoptosis and autophagy dysfunction with Mfn2 overexpression (fig. S5, K to M). Collectively, these findings identify a causal role of MFN2-mediated mitochondrial fusion in astrocytes as a key driver of the depressive behaviors. Knockdown of astrocytic Mfn2 in the hippocampus improves depression-like behaviors through amelioration of astrocyte loss and mitophagy dysfunction in CSDS mice. (A) Experimental design for conditional knockdown of Mfn2 in astrocytes followed by CSDS model and behavioral tests. Created in BioRender. W. Cong (2026), https://BioRender.com/ohcvaai. (B) Observation of virus infection in astrocytes by immunofluorescence staining of GFAP and enhanced green fluorescent protein (EGFP). Scale bar, 20 μm. (C to J) Effects of astrocytic Mfn2 knockdown on CSDS-induced depression-like behaviors; n = 10 mice. (K and M) Immunofluorescence staining of GFAP and quantification of GFAP+ cells in DG region demonstrating effects of astrocytic Mfn2 knockdown on CSDS-induced astrocyte loss; n = 6 mice. Scale bar, 50 μm. (L, N, and O) Effects of astrocytic Mfn2 knockdown on mitochondrial morphology observed by transmission electron microscopy; n = 21 mitochondria in (N) and n = 7 fields in (O). Scale bar, 1 μm. (P) Representative blots showing effects of astrocytic Mfn2 knockdown on the protein levels of MFN2, P62, and LC3B in hippocampus tissue. *P < 0.05, **P < 0.01, and ***P < 0.001 by two-way analysis of variance (ANOVA) with Tukey’s multiple comparisons test; n.s., not significant. β-Arrestin2 deficiency enhances MFN2 signaling to drive astrocyte pathology and depressive behaviors Given the pronounced up-regulation of MFN2 in astrocytes from CSDS mice, we next explored the upstream signaling mechanisms contributing to this alteration. Gene set enrichment analysis of our SMART-seq data for astrocytic transcriptomes indicated enrichment of G protein–coupled receptor (GPCR) signaling pathways (Fig. 4A). Consistent with this finding, analysis of hippocampal lysates revealed marked MFN2 elevation accompanied by reduced β-arrestin2 expression in SUS mice, while RES mice showed no such changes. Other GPCR-associated proteins, including phosphorylated phosphatidylinositol 3-kinase, phosphorylated Akt (p-Akt), and β-arrestin1, remained unaltered between groups (Fig. 4B and fig. S6, A to C). To determine whether β-arrestin2 deficiency directly contributes to astrocyte pathology, we generated astrocyte-specific Arrb2 (encoding β-arrestin2) knockdown mice via hippocampal injection of rAAV-GfABC1D-EGFP-shArrb2, followed by CSDS and behavioral testing (Fig. 4C). Viral expression was restricted to GFAP+ astrocytes (Fig. 4D and fig. S6, D and E). Astrocytic Arrb2 knockdown significantly aggravated depression-like behaviors in CSDS mice across multiple assays, including SIT, OFT, TST, FST, and SPT (Fig. 4, E to H, and fig. S6, F to I). Notably, Arrb2 knockdown reduced astrocyte numbers even under basal conditions and further exacerbated CSDS-induced astrocytic loss (Fig. 4, I and K). Ultrastructural assessment by transmission electron microscopy demonstrated enlarged mitochondrial area in Arrb2-deficient mice at baseline, which was further amplified following CSDS exposure (Fig. 4, J, L, and M). Consistent with these morphological changes, Western blotting confirmed that Arrb2 knockdown enhanced MFN2 expression and aggravated autophagy impairment (Fig. 4N and fig. S6, J to M). These findings reveal a critical inhibitory role of β-arrestin2 in regulating MFN2-associated pathology in astrocytes. Knockdown of astrocytic Arrb2 exacerbates depression-like behaviors via boosting MFN2-associated mitochondrial dysfunction and astrocyte loss in CSDS mice. (A) Gene set enrichment analysis of SMART-seq data revealing the involvement of G protein–coupled receptor (GPCR) signaling in astrocyte pathology. (B) Western blotting analysis of MFN2 and GPCR-related proteins in hippocampus tissue collected from control (CON), susceptible (SUS), and resilient (RES) mice; n = 6 mice. p-PI3K, phosphorylated phosphatidylinositol 3-kinase. (C) Experimental design for conditional knockdown of Arrb2 in astrocytes followed by CSDS model and behavioral tests. Created in BioRender. W. Cong (2026), https://BioRender.com/ohcvaai. (D) Observation of virus infection in astrocytes by immunofluorescence staining of GFAP and EGFP. Scale bar, 20 μm. (E to H) Effects of astrocytic Arrb2 knockdown on CSDS-induced depression-like behaviors; n = 10 mice. (I and K) Immunofluorescence staining of GFAP and quantification of GFAP+ cells in DG region demonstrating effects of astrocytic Arrb2 knockdown on CSDS-induced astrocyte loss; n = 6 mice. Scale bar, 20 μm. (J, L, and M) Effects of astrocytic Arrb2 knockdown on mitochondrial morphology observed by transmission electron microscopy; n = 24 mitochondria in (L) and n = 7 fields in (M). Scale bar, 1 μm. (N) Representative blots showing effects of astrocytic Arrb2 knockdown on the protein levels of β-arrestin2, MFN2, P62, and LC3B in hippocampus tissue. *P < 0.05, **P < 0.01, and ***P < 0.001 by one-way ANOVA with Tukey’s multiple comparisons test in (B) and by two-way ANOVA with Tukey’s multiple comparisons test in others; n.s., not significant. To validate the β-arrestin2–MFN2 regulatory axis in vivo, we used Arrb2 global knockout (KO) mice and performed astrocyte-specific Mfn2 knockdown by hippocampal injection of rAAV-GfABC1D-EGFP-shMfn2 (Fig. 5, A to C). Consistent with our observations using astrocyte-specific Arrb2 knockdown, Arrb2 KO mice did not display significant depression-like behaviors under basal conditions but showed a pronounced exacerbation of CSDS-induced behavioral deficits (fig. S7, A to H). Viral expression was restricted to GFAP+ astrocytes (Fig. 5D and fig. S7, I to J). Mfn2 knockdown ameliorated CSDS-induced depression-like behaviors in Arrb2 KO mice (Fig. 5, E to H, and fig. S7, K to N). Moreover, astrocytic Mfn2 knockdown restored astrocyte number and normalized mitochondrial morphology and autophagy markers in the hippocampus (Fig. 5, I to O). Together, these results identify β-arrestin2 as an upstream regulator of MFN2-mediated mitochondrial and autophagic dysfunction in astrocytes. Astrocytic Mfn2 knockdown ameliorates CSDS-induced depression-like behaviors in Arrb2 KO mice. (A) Genotyping of wild-type (WT) and Arrb2 knockout (KO) mice. bp, base pairs. (B) Western blotting analysis of β-arrestin2 expression in hippocampus tissue of WT and Arrb2 KO mice. (C) Experimental design for conditional knockdown of astrocytic Mfn2 in Arrb2 KO mice followed by CSDS model and behavioral tests. Created in BioRender. W. Cong (2026), https://BioRender.com/ohcvaai. (D) Observation of virus infection in astrocytes by immunofluorescence staining of GFAP and EGFP. Scale bar, 20 μm. (E to H) Effects of astrocytic Mfn2 knockdown on CSDS-induced depression-like behaviors in Arrb2 KO mice; n = 10 mice. (I and J) Immunofluorescence staining of GFAP and quantification of GFAP+ cells in DG region demonstrating effects of astrocytic Mfn2 knockdown on CSDS-induced astrocyte loss in Arrb2 KO mice; n = 6 mice. Scale bar, 50 μm. (K to M) Effects of astrocytic Mfn2 knockdown on mitochondrial morphology in Arrb2 KO mice observed by transmission electron microscopy; n = 22 mitochondria in (L) and n = 6 fields in (M). Scale bar, 1 μm. (N and O) Representative blots and quantification showing effects of astrocytic Mfn2 knockdown on the protein levels of MFN2, P62, and LC3B in hippocampus tissue of Arrb2 KO mice; n = 6 mice. ***P < 0.001 by two-way ANOVA with Tukey’s multiple comparisons test; n.s., not significant. Loss of β-arrestin2 up-regulates MFN2 expression by attenuating STAT1-mediated transcriptional suppression of Mfn2 Primary astrocytes isolated from the hippocampus of WT and Arrb2 KO mice were used to dissect the molecular mechanism underlying β-arrestin2 loss-induced MFN2 up-regulation (Fig. 6A). Consistent with in vivo findings, Arrb2 KO astrocytes exhibited a marked increase in MFN2 and cleaved-caspase3 expression (Fig. 6, B to E). Ultrastructural and functional analyses further revealed pronounced mitochondrial fusion and reduced mitophagy in Arrb2 KO astrocytes (Fig. 6, F to J, and fig. S8, A to C), corroborating the regulatory role of β-arrestin2 in maintaining mitochondrial homeostasis and autophagic flux. To identify potential β-arrestin2–interacting proteins, we performed mass spectrometry–based label-free proteomics on astrocyte lysates immunoprecipitated with either anti–β-arrestin2 or immunoglobulin G (IgG) control antibodies. Comparative analysis identified 119 β-arrestin2–binding proteins after excluding nonspecific IgG-binding candidates (Fig. 6K). Integrative analysis with Mfn2 transcription factors predicted by PROMO revealed a single overlap, STAT1, as both a β-arrestin2–binding protein and a putative transcriptional regulator of Mfn2 (Fig. 6L). In Arrb2 KO astrocytes, both total STAT1 and phosphorylated STAT1 protein levels were significantly reduced, whereas Stat1 mRNA expression remained unchanged (Fig. 6, M to O, and fig. S8D), indicating that β-arrestin2 loss promotes STAT1 degradation rather than transcriptional suppression. Bioinformatic prediction identified potential STAT1-binding motifs within the Mfn2 promoter region (Fig. 6P and fig. S8E), which were experimentally validated using chromatin immunoprecipitation (ChIP)–PCR and ChIP-qPCR assays (Fig. 6, R and S). To test the functional relevance of this binding, luciferase reporter assays were performed using wild-type (WT; Mfn2-WT) and STAT1-binding-site-deletion (Mfn2-mutant) promoter constructs (Fig. 6Q). STAT1 overexpression significantly decreased luciferase activity in cells transfected with Mfn2-WT, but not Mfn2-mutant, constructs (Fig. 6T), confirming that STAT1 acts as a transcriptional repressor of Mfn2. Consistently, Stat1 knockdown in primary astrocytes had no effect on Arrb2 mRNA or protein levels but markedly elevated Mfn2 expression at both mRNA and protein levels (fig. S9, A to L). Collectively, these findings demonstrate that STAT1 suppresses Mfn2 transcription, whereas β-arrestin2 deficiency attenuates this repression by facilitating STAT1 degradation, thereby driving MFN2 up-regulation and mitochondrial dysfunction in astrocytes. Loss of β-arrestin2 up-regulates Mfn2 expression by reducing STAT1, a transcriptional repressor of Mfn2 in cultured astrocytes. (A) Illustration of hippocampus primary astrocyte culture from WT and Arrb2 KO mice. Created in BioRender. W. Cong (2026), https://BioRender.com/ohcvaai. (B and C) RT-qPCR analyses of Arrb2 and Mfn2 mRNA levels in WT and Arrb2 KO astrocytes; n = 6 independent cultures. (D and E) Western blotting analyses of β-arrestin2, MFN2, and cleaved-caspase3 (c-Caspase3) levels in WT and Arrb2 KO astrocytes; n = 6 independent cultures. (F to H) Observation and quantification of mitochondrial morphology in WT and Arrb2 KO astrocytes by transmission electron microscopy; n = 30 mitochondria in (G) and n = 7 fields in (H). Scale bar, 1 μm. (I and J) Immunofluorescence staining of LC3 with MitoTracker, and quantification of LC3-positive mitochondria in WT and Arrb2 KO astrocytes; n = 35 cells. Scale bar, 10 μm. (K) Venn diagram showing potential β-arrestin2–binding proteins detected in astrocyte lysates immunoprecipitated with either anti–β-arrestin2 or IgG control antibodies by liquid chromatography–tandem mass spectrometry. (L) Venn diagram showing STAT1 as one of Mfn2 transcription factors that binding to β-arrestin2. (M to O) Representative blots and quantification of STAT1 and phosphorylated STAT1 (p-STAT1) in WT and Arrb2 KO astrocytes; n = 6 independent cultures. (P) Illustration of predicted STAT1 binding motif within Mfn2 promoter region. (Q) Design of Mfn2-WT and Mfn2-mutant plasmids for luciferase reporter assay. (R and S) Validation of STAT1 binding to Mfn2 promoter by ChIP-PCR in R and ChIP-qPCR in (S); n = 3 independent cultures. (T) Measurement of luciferase activity in human embryonic kidney–293 T cells transfected with WT/Mutant Mfn2 plasmids in normal or Stat1 overexpression condition; n = 6 independent cultures. ***P < 0.001 by two-way ANOVA with Tukey’s multiple comparisons test in (R) and (T) and by unpaired Student’s t test in others; n.s., not significant. Overexpression of Stat1 ameliorates β-arrestin2 deficiency–induced astrocyte pathology by reducing MFN2 expression To determine whether STAT1 directly regulates astrocytic function downstream of β-arrestin2, we overexpressed Stat1 in both WT and Arrb2 KO astrocytes. Consistent with results from Stat1 knockdown experiments, Stat1 overexpression did not affect Arrb2 mRNA or protein levels but markedly reduced Mfn2 mRNA and protein expression in both WT and Arrb2 KO astrocytes (Fig. 7, A to G). Transmission electron microscopy revealed that Stat1 overexpression diminished the enlarged mitochondrial area observed in Arrb2 KO astrocytes (Fig. 7, H to J). Immunofluorescence staining with MitoTracker and LC3 further showed a decreased proportion of LC3-positive mitochondria in Arrb2 KO astrocytes, which was restored by Stat1 overexpression (Fig. 7, K and L). These findings indicate that Stat1 overexpression effectively alleviates MFN2-associated mitochondrial dysfunction in Arrb2-deficient astrocytes. To further evaluate mitochondrial function, we measured mitochondrial membrane potential using the JC-1 probe. Flow cytometry analysis showed an increased proportion of depolarized mitochondria in Arrb2 KO astrocytes, as indicated by elevated JC-1 green fluorescence, whereas Stat1 overexpression markedly reduced mitochondrial depolarization (Fig. 7, M and N). We observed that Arrb2 KO led to reduced adenosine 5′-triphosphate (ATP) production and a lower nicotinamide adenine dinucleotide (oxidized form) (NAD+)/reduced form of NAD+ (NADH) ratio, both of which were ameliorated by Stat1 overexpression (fig. S10, A and B). In addition, Stat1 overexpression in these astrocytes reduced mitochondrial reactive oxygen species (ROS) levels (fig. S10, C and D). Proximity ligation assay (PLA) showed a significant increase in MFN2-IP3R interaction in Mfn2-overexpressing astrocytes, suggesting that ER-mitochondria contacts were enhanced (fig. S10E). Furthermore, we used the endoplasmic reticulum (ER) calcium indicator G-CEPIA1er to measure ER calcium release upon ATP stimulation. Our findings revealed that Mfn2 overexpression significantly enhanced ER calcium release in response to ATP stimulation (fig. S10F). These results suggest that Mfn2 overexpression promotes ER-to-mitochondria calcium transfer via MFN2-IP3R interactions, which likely contributes to mitochondrial dysfunction in astrocytes. Together, we identify a previously unrecognized β-arrestin2–STAT1–MFN2 regulatory axis that governs mitochondrial function in astrocytes. Overexpression of Stat1 protects against MFN2-associated mitochondrial dysfunction in cultured astrocytes. (A to C) RT-qPCR analyses of Stat1, Arrb2, and Mfn2 mRNA levels in WT and Arrb2 KO astrocytes with Stat1 overexpression; n = 6 independent cultures. (D to G) Representative blots and quantification of STAT1, β-arrestin2, and MFN2 protein levels in WT and Arrb2 KO astrocytes with Stat1 overexpression; n = 6 independent cultures. (H to J) Observation and quantification of mitochondrial morphology in WT and Arrb2 KO astrocytes with Stat1 overexpression by transmission electron microscopy; n = 20 mitochondria in (H) and n = 7 fields in (I). Scale bar, 1 μm. (K and L) Immunofluorescence staining of LC3 with MitoTracker, and quantification of LC3-positive mitochondria in WT and Arrb2 KO astrocytes with Stat1 overexpression; n = 20 cells. Scale bar, 20 μm. (M and N) JC-1 staining for mitochondrial membrane potential and quantification of depolarized mitochondria by flow cytometry; n = 7 independent cultures. *P < 0.05, **P < 0.01, and ***P < 0.001 by two-way ANOVA with Tukey’s multiple comparisons test; n.s., not significant. Astrocytic Mfn2 knockdown or Stat1 overexpression reduces neuronal hyperexcitability in Arrb2 KO CSDS mice We next explored how astrocytic dysfunction contributes to depression-like behaviors. Immunofluorescence staining of c-Fos in hippocampal sections revealed a significant increase in c-Fos–positive neurons in SUS, but not RES, mice (Fig. 8A). Correspondingly, SUS mice displayed elevated glutamate levels and reduced expression of the astrocytic glutamate transporters EAAT1 and EAAT2 (Fig. 8, B and C, and fig. S11A). These findings suggest that astrocyte impairment may diminish glutamate clearance, thereby elevating extracellular glutamate and driving neuronal hyperexcitability in the hippocampus. To test this hypothesis, we directly measured astrocyte Ca2+ signaling in vivo using rAAV-GfaABC1D-GCaMP6m (fig. S11B). Our results showed that astrocytic calcium transients were significantly reduced in CSDS mice during the SIT (fig. S11C). This reduction in calcium signaling aligns with the observed astrocytic dysfunction following chronic stress. However, knockdown of astrocytic Mfn2 preserved calcium transients in CSDS mice (fig. S11, D and E). Furthermore, astrocytes with Mfn2 overexpression exhibited reduced calcium transients, similar to what was observed in CSDS mice (fig. S11, F and G). These findings suggest that MFN2 up-regulation in astrocytes contributes to impaired calcium signaling. Furthermore, we performed astrocytic Mfn2 knockdown in Arrb2 KO mice and coinjected rAAV-CaMKIIα-GCaMP6m into the hippocampus to monitor calcium dynamics in glutamatergic neurons (Fig. 8D and fig. S12A). In vivo calcium imaging demonstrated that astrocytic Mfn2 knockdown markedly reduced Ca2+ transients in glutamatergic neurons of Arrb2 KO CSDS mice during SIT, TST, and FST (Fig. 8, E and F, and fig. S12, B to E). Consistently, Mfn2 knockdown decreased hippocampal glutamate levels while restoring EAAT1 and EAAT2 expression (Fig. 8, G to I). These results confirm that alleviating MFN2-mediated astrocyte dysfunction preserves neuronal excitability homeostasis by reducing glutamate accumulation and enhancing glutamate uptake. Given that STAT1 transcriptionally represses Mfn2 expression, we next overexpressed Stat1 in astrocytes of Arrb2 KO mice and monitored neuronal calcium activity (Fig. 8J). Similar to Mfn2 knockdown, astrocytic Stat1 overexpression significantly attenuated Ca2+ activity in hippocampal glutamatergic neurons during SIT, TST, and FST (Fig. 8, K and L, and fig. S12, F to I), accompanied by decreased glutamate levels and increased EAAT1 and EAAT2 expression (Fig. 8, M to O). We also found that knockdown of astrocytic Mfn2 significantly reduced the levels of interleukin-1β (IL-1β), IL-6, and tumor necrosis factor–α (TNF-α) in Arrb2 KO mice following CSDS (fig. S12J). This reduction in inflammatory markers was also observed in astrocytic Stat1 overexpression mice (fig. S12K). Together, these findings reveal that loss of the β-arrestin2–STAT1 inhibitory control on MFN2 expression triggers astrocyte dysfunction, which includes impaired glutamate uptake, increased inflammation, and mitochondrial dysfunction, and disrupts astrocyte-neuron communication. This, in turn, leads to the neurophysiological alterations and depression-like behaviors. Astrocytic Mfn2 knockdown or Stat1 overexpression reduces neuronal hyperexcitability in hippocampus of Arrb2 KO CSDS mice. (A) Immunofluorescence staining of c-Fos and quantification of c-Fos–positive puncta in hippocampal slices; n = 6 mice. (B) Detection of glutamate levels in hippocampal tissue; n = 10 mice. (C) Western blotting analyses of EAAT1 and EAAT2 protein levels in hippocampal tissue. The cartoon graphical elements were created in BioRender. W. Cong (2026), https://BioRender.com/ohcvaai. (D) Injection of AAVs targeting astrocytic Mfn2 knockdown and glutamatergic neuronal Ca2+ signal into the hippocampus of Arrb2 KO mice. Created in BioRender. W. Cong (2026), https://BioRender.com/ohcvaai. (E and F) Detection of glutamatergic neuronal Ca2+ signal in the hippocampus of Arrb2 KO CSDS mice with astrocytic Mfn2 knockdown during SIT; n = 19 events. (G) Detection of glutamate levels in the hippocampus of Arrb2 KO CSDS mice with astrocytic Mfn2 knockdown; n = 8 mice. (H and I) Effects of astrocytic Mfn2 knockdown on the expression of EAAT1 and EAAT2 in hippocampal tissue of Arrb2 KO CSDS mice; n = 6 mice. (J) Injection of AAVs targeting astrocytic Stat1 overexpression and glutamatergic neuronal Ca2+ signal into the hippocampus of Arrb2 KO mice. Created in BioRender. W. Cong (2026), https://BioRender.com/ohcvaai. (K and L) Detection of glutamatergic neuronal Ca2+ signal in the hippocampus of Arrb2 KO CSDS mice with astrocytic Stat1 overexpression during SIT; n = 19 events. (M) Detection of glutamate levels in the hippocampus of Arrb2 KO CSDS mice with astrocytic Stat1 overexpression; n = 6 mice. (N and O) Effects of astrocytic Stat1 overexpression on the expression of EAAT1 and EAAT2 in hippocampal tissue of Arrb2 KO CSDS mice; n = 6 mice. ***P < 0.001 by one-way ANOVA with Tukey’s multiple comparisons test in (A) and (B) and by unpaired Student’s t test in others; n.s., not significant. DISCUSSION Astrocyte pathology has long been implicated in MDD, yet the underlying molecular mechanisms remain poorly understood. We demonstrate that β-arrestin2 deficiency in astrocytes destabilizes STAT1, releasing its transcriptional repression of Mfn2 and thereby promoting MFN2-associated mitochondrial dysfunction. These mitochondrial abnormalities lead to astrocyte dysfunction, glutamate accumulation, inflammation, neuronal hyperexcitability, and, ultimately, depression-like behaviors (fig. S13). Conversely, restoring Stat1 expression or suppressing Mfn2 in astrocytes reinstates mitochondrial integrity, normalizes glutamate homeostasis and inflammatory cytokine levels, and ameliorates depressive phenotypes. Collectively, our data define a β-arrestin2–STAT1–MFN2 signaling axis that governs astrocyte homeostasis and emotional behavior. Historically, MDD was regarded as a neuron-centric disorder (19). However, postmortem analyses have revealed substantial glial loss, without evident neuronal reduction, in the prefrontal cortex of patients with MDD (20). Subsequent studies further demonstrated a marked decrease in astrocyte density across multiple regions, including the hippocampus, mPFC, ACC, and BLA (9, 21), shifting the pathophysiological focus from neurons to glia. Consistent with these human findings, our work identified a pronounced astrocyte loss in the hippocampus of CSDS mice. Although astrocytic loss has been observed across several MDD-associated brain regions, the underlying mechanisms remain poorly characterized. A recent study implicated microglia-derived IL-6 in astrocytic apoptosis (22). Our results extend this understanding by identifying mitochondrial dysregulation as a direct pathogenic driver within astrocytes. Through SMART-seq profiling of hippocampal astrocytes, we uncovered MFN2-associated astrocyte dysfunction as key pathological features in depression. MFN2 up-regulation was also detected in postmortem brains, plasma, and iPSC-derived astrocytes from patients with MDD (23, 24), supporting the clinical relevance of our findings. Together, these data delineate a conserved mitochondrial pathology within astrocytes that contributes to depressive disorders. Previous reports demonstrate reduced mitochondrial bioenergetics in adipose-specific Mfn2 KO mice (25), whereas our study further shows that MFN2 up-regulation in astrocytes leads to impaired mitochondrial function and astrocyte apoptosis. Therefore, we propose that MFN2 dysregulation in either direction, whether through up-regulation or down-regulation, can disrupt mitochondrial bioenergetics, contributing to mitochondrial dysfunction and cellular pathology. Mitochondrial dysfunction has emerged as a core feature of psychiatric illnesses, including MDD and anxiety (14, 26), yet the cell-type– and region-specific mechanisms remain elusive. Our study reveals that MFN2 expression is selectively increased in hippocampal astrocytes, but not neurons or microglia, of a depression model. Notably, astrocyte-specific knockdown of Mfn2 mitigated depression-like behaviors by preventing mitochondrial impairment. Chronic stress has been shown to impair autophagy (27), and excessive mitochondrial fusion suppresses mitophagy (18). Consistent with this, we found that MFN2-associated mitochondrial dysfunction disrupts autophagy and promotes astrocytic loss in the hippocampus. This observation contrasts with findings in the nucleus accumbens, where reduced neuronal MFN2 correlates with enhanced depressive phenotypes (28), highlighting a distinct and cell-type–specific role of MFN2 in astrocytes during MDD pathogenesis. We also recognize that the hippocampus plays a key role in learning and memory. We propose that the MFN2-induced astrocyte loss and the resulting disruption of hippocampal function may affect these cognitive processes, potentially leading to long-term consequences, including memory impairments or even dementia. GPCR signaling pathways are central to MDD pathophysiology and represent promising therapeutic targets (29). Whole transcriptome sequencing of iPSC-derived astrocytes from patients with MDD revealed dysregulation of GPCR signaling (24), consistent with our SMART-seq data showing disrupted GPCR signaling in hippocampal astrocytes of CSDS mice. Moreover, we explored downstream molecules of GPCR signaling and found that β-arrestin2–biased GPCR signaling was diminished in CSDS-susceptible mice. While global β-arrestin2 KO enhances vulnerability to chronic stress-induced astrocyte loss (30), our astrocyte-specific β-arrestin2 knockdown model demonstrates that this effect is cell autonomous. Mechanistically, β-arrestin2 loss leads to STAT1 degradation, derepressing Mfn2 transcription and resulting in mitochondrial impairments. Unlike prior findings implicating β-arrestin2–STAT3 signaling in neuroinflammation (30), our data reveal a previously unrecognized β-arrestin2–STAT1–MFN2 axis that directly regulates mitochondrial homeostasis in astrocytes. Notably, we found that Arrb2-deficient mice did not exhibit significant depression-like behaviors under basal conditions, despite showing partial molecular and mitochondrial alterations. We propose that the emergence of depression-like behaviors correlates with the magnitude of MFN2-mediated astrocytic dysfunction and cell loss, rather than the mere presence of pathway dysregulation. Furthermore, Arrb2 deficiency alone at baseline induces only modest MFN2 up-regulation, whereas CSDS exposure produces a substantially greater effect, indicating that stress-associated signaling pathways further converge on MFN2 regulation. Our findings that astrocytic Mfn2 knockdown or Stat1 overexpression attenuated inflammatory responses in Arrb2 KO mice subjected to CSDS, supporting functional cross-talk between MFN2 dysregulation and stress-induced inflammatory signaling. These findings suggest that β-arrestin2/STAT1 signaling constitutes one critical modulatory pathway within a broader regulatory network controlling MFN2 expression under chronic stress. Astrocytes are multifaceted glial cells that maintain structural, metabolic, and synaptic homeostasis in the brain. Their bidirectional communication with neurons is fundamental to shaping synaptic activity and circuit function (31). Disruption of this astrocyte-neuron cross-talk is increasingly recognized as a driver of MDD-related synaptic and network dysfunction (32, 33). Chronic stress has been shown to enhance astrocytic glutamate release, overstimulating hippocampal neurons and inducing depressive behaviors (34). In agreement, we observed elevated glutamate levels in the hippocampus of CSDS-SUS mice, accompanied by reduced expression of astrocytic glutamate transporters EAAT1 and EAAT2. In vivo calcium imaging further revealed neuronal hyperexcitability resulting from impaired astrocytic glutamate clearance. Further research is needed to elucidate the mechanisms by which MFN2 suppresses astrocytic Ca2+ signaling. We hypothesize that this suppression results from MFN2-induced astrocyte apoptosis, a dysfunctional state that indirectly impairs Ca2+ channel activity. These findings emphasize that astrocytic dysfunction, particularly the loss of glutamate homeostasis, plays a causal role in the breakdown of neuron-glia communication in depression. Although mitochondrial hyperfusion remains a plausible interpretation based on the established function of MFN2 and our ultrastructural observations, our current data do not exclude the possibility that altered ER-mitochondria communication, calcium signaling, or other MFN2-dependent cellular processes also contribute to astrocyte dysfunction and depressive-like behaviors. Our study has several limitations. Although mitochondrial hyperfusion was observed by transmission electron microscopy in hippocampal tissue from CSDS mice, definitive astrocyte-specific evidence will require future studies using approaches such as immunogold electron microscopy or live imaging in genetically labeled astrocytes. We therefore present mitochondrial hyperfusion as one possible interpretation of the current findings rather than a definitive mechanistic conclusion. In conclusion, our study identifies a β-arrestin2–STAT1–MFN2 signaling axis that links astrocyte dysfunction to depressive pathology. Restoration of STAT1 or inhibition of MFN2 in astrocytes protects against astrocyte loss and depressive behaviors, underscoring astrocytic mitochondria as therapeutic targets in MDD. We propose that targeting astrocytic MFN2 through the β-arrestin2/STAT1 signaling pathway represents a promising therapeutic strategy for depression as it addresses the underlying astrocytic dysfunction that drives the neuronal hyperexcitability and depression-like behaviors observed. More broadly, these findings highlight the pivotal contribution of astrocytic metabolic integrity to mood regulation and provide a mechanistic framework for developing glia-focused antidepressant strategies. MATERIALS AND METHODS Animals Adult male C57BL/6 (8 to 10 weeks old, 22 to 25 g) and male CD-1 (5-month-old, 40 g) mice were obtained from the Laboratory Animal Center of Nanjing Medical University. Arrb2 global KO (Arrb2 KO) mice were generated by Nanjing Biomedical Research Institute of Nanjing University. All mice were housed under a 12-hour light/dark cycle with ad libitum access to food and water. Sample sizes were based on prior variance/pilot data in CSDS models. Mice were randomized, and investigators were blinded to allocation for behavioral testing, stereology/cell counting, imaging acquisition/quantification, and Western blot. No data were excluded. All animal experiments were conducted in accordance with the ARRIVE guidelines and were approved by Institutional Animal Care and Use Committee of Nanjing Medical University (no. 2008067). CSDS mouse model The CSDS paradigm was performed as previously described (35). Briefly, experimental mice (C57BL/6) were exposed daily to an aggressive CD-1 mouse for 10 consecutive days. During each session, the C57BL/6 mouse was subjected to 10 min of physical interaction followed by 24 hours of sensory contact through a perforated divider. Control mice were housed in similar cages with no exposure to aggressors. Behavioral assays were performed 24 hours after the last defeat session. Behavioral tests Social interaction test SI behavior was assessed using a two-phase paradigm. In the initial “no-target” session, an empty wire-mesh cage was positioned within the interaction zone of the testing arena to establish baseline exploratory activity. During the subsequent “target” session, a novel CD-1 mouse was introduced into the cage. Each session lasted 2.5 min, with a 5-min acclimation period before testing. Mouse trajectories were recorded and analyzed using EthoVision software (Noldus Technology). The SI ratio was calculated as: SI ratio = (time spent in interaction zone with target)/(time spent in interaction zone without target). Mice exhibiting SI ratios of ≥1 were classified as RES, whereas those <1 were considered SUS. All apparatus were cleaned with 75% ethanol between trials to remove olfactory residues. Open-field test Locomotor activity and exploratory behavior were evaluated in a 50 cm–by–50 cm–by–40 cm opaque arena under dim lighting. After a 3-hour acclimation period in the testing room, mice were individually placed in the center of the arena for a 10-min session. Movement was tracked with a ceiling-mounted camera and analyzed using TopScan software (CleverSys Inc.). The total distance traveled, mean velocity, and time spent in the central zone (35 cm by 35 cm) were quantified as primary parameters. The arena was sanitized with 75% ethanol after each trial. Tail suspension test Each mouse was suspended by the tail using adhesive tape affixed ∼1 cm from the tip and attached to a fixed horizontal bar. The 6-min testing session was video recorded, and immobility, defined as the absence of active movement except those required for breathing, was quantified using TailSuspScan software (CleverSys Inc., USA). The total duration of immobility was used as an index of behavioral despair. Forced swim test Mice were individually placed in transparent cylindrical containers (height, 25 cm; diameter, 15 cm) filled with 14 cm of water maintained at 20° ± 2°C. Each session lasted 6 min, and immobility time, defined as minimal movement sufficient to keep the head above water, was automatically recorded using TailSuspScan software. Water was replaced between sessions to maintain consistent temperature and cleanliness. Sucrose preference test To evaluate anhedonia, mice were deprived of water for 12 hours before testing and then given simultaneous access to two preweighed bottles for 10 hours: one containing 1% sucrose solution and the other containing tap water. Bottle positions were exchanged after 5 hours to avoid side bias. Fluid consumption was determined by weighing the bottles before and after the session. Sucrose preference was calculated as sucrose preference (%) = sucrose intake/(sucrose + water intake) × 100%. Immunofluorescence Mice were transcardially perfused with phosphate-buffered saline (PBS) followed by 4% paraformaldehyde (PFA). Brains were harvested and sectioned at 30-μm thickness using a cryostat (Leica CM1950). Fixed samples, including brain sections and primary astrocytes, were permeabilized and blocked in PBS containing 0.3% Triton X-100 (PBST) and 5% bovine serum albumin for 1 hour at room temperature. Samples were then incubated overnight at 4°C with validated primary antibodies, followed by three washes in PBST (5 min each). Appropriate species-matched fluorophore-conjugated secondary antibodies were applied for 1 hour at room temperature. Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI), and images were captured using a Leica SP8 confocal microscope. Quantitative image analysis was performed with ImageJ software (version 1.53q). Astrocyte numbers were quantified by counting GFAP+ or GS+ astrocytes using ImageJ software. Antibodies used for the immunofluorescence assays are provided in table S1. SMART-seq and bioinformatics analysis Hippocampal astrocytes were purified by fluorescence-activated cell sorting using the Anti–ACSA-2 MicroBead Kit (Miltenyi Biotec, Germany) according to the manufacturer’s guidelines. Total RNA was extracted from sorted cells and subjected to SMART-seq analysis, performed by LC Bio (Zhejiang, China). Sequencing libraries were prepared using the NEBNext Ultra RNA Library Prep Kit for Illumina, with sample-specific index barcodes incorporated during adapter ligation. Cluster generation was carried out on an Illumina cBot system with the TruSeq PE Cluster Kit v4-cBot-HS, followed by paired-end sequencing on an Illumina HiSeq 2500 platform. Raw reads were processed to remove adaptor sequences and low-quality fragments, yielding high-confidence clean reads for downstream analysis. Clean reads were aligned to the reference mouse genome using the TopHat2 alignment tool, allowing for a single mismatch per read. Only uniquely mapped reads were retained for further quantification and annotation. Gene expression levels were calculated and normalized according to standard RNA-sequencing analysis pipelines. Bioinformatics analysis of the SMART-seq data was performed using OmicStudio tools (www.omicstudio.cn/tool). Real-time RT-qPCR Total RNA was extracted from hippocampal tissue and cultured astrocytes using TRIzol reagent (Invitrogen, USA) according to the supplier’s protocol. RNA integrity and concentration were assessed using a NanoDrop ND-1000 spectrophotometer. cDNA was synthesized from purified RNA with the 5× HiScript II qRT SuperMix (Vazyme, China) following the standard reaction conditions. qPCR was then performed on a StepOnePlus Real-Time PCR System (Applied Biosystems, USA) using SYBR Green Master Mix (Vazyme, China). Specificity of amplification was verified by melt-curve analysis. Relative mRNA abundance was determined using the 2–ΔΔCt method, with Gapdh serving as the internal reference gene. All primers were synthesized by GenScript (China), and their sequences are listed in table S2. Western blotting assay Protein extracts from hippocampal tissue and cultured astrocytes were obtained using radioimmunoprecipitation assay (RIPA) buffer (Bio-Rad) supplemented with a protease and phosphatase inhibitor cocktail (Roche). Lysates were clarified by centrifugation at 13,000g for 20 min at 4°C, and protein concentration was determined using the Micro BCA Protein Assay Kit (Beyotime, China). Equal amounts of total protein were separated on 8 to 12% SDS–polyacrylamide gels and transferred onto polyvinylidene difluoride membranes. After blocking in 5% nonfat milk, membranes were incubated overnight at 4°C with the indicated primary antibodies, followed by horseradish peroxidase–conjugated secondary antibodies for 1 hour at room temperature. Blots were developed using Pierce ECL reagents (Thermo Fisher Scientific) and visualized with an ImageQuant LAS 4000 system (GE Healthcare). Details of all antibodies used are provided in table S1. RNAscope assay Fresh-frozen brain sections were mounted onto glass slides, baked at 60°C, and postfixed with 4% PFA. Samples were sequentially dehydrated through graded ethanol and processed using the RNAscope Pretreatment Kit (ACD) following the manufacturer’s instructions. After treatment with hydrogen peroxide and target retrieval reagents, sections were permeabilized with Protease Plus and hybridized with an Mfn2-specific probe at 40°C for 2 hours. Signal amplification was achieved using the RNAscope Multiplex Fluorescent Detection Kit. Following the hybridization steps, sections were blocked and incubated overnight at 4°C with primary antibodies against GFAP, NeuN, and IBA1 to label astrocytes, neurons, and microglia, respectively. Fluorophore-conjugated secondary antibodies were applied for visualization, and slides were mounted with antifade medium. Images were acquired using a confocal microscope under identical settings for all groups. Transmission electron microscopy For ultrastructural analysis of mitochondria, freshly dissected hippocampal tissue and primary cultured astrocytes were immediately rinsed three times in 0.1 M phosphate buffer (PB) and postfixed in 1% osmium tetroxide for 1 hour at room temperature. Samples were again washed three times in PB and sequentially dehydrated through a graded ethanol series (30, 50, 70, 80, 90, and 100%; 8 min per step), followed by two acetone washes. Dehydrated samples were infiltrated with EPON812 resin mixed with acetone (1:1 for 2 hour at 37°C and then 2:1 overnight at 37°C), embedded in fresh EPON812, and polymerized at 60°C for 48 hours. Ultrathin sections (70 nm) were prepared using an ultramicrotome and stained with 2% uranyl acetate for 10 min, followed by lead citrate for 10 min. The sections were examined using a Hitachi HT7800 transmission electron microscope, and mitochondrial morphology was analyzed under identical acquisition settings across experimental groups. Viral constructs and stereotactic injection Recombinant adeno-associated viruses (rAAVs) were purchased from BrainVTA (Wuhan, China). The following vectors were used for astrocyte-specific gene manipulation: rAAV-GfaABC1D-EGFP-shMfn2, rAAV-GfaABC1D-mCherry-shMfn2, rAAV-GfaABC1D-EGFP-shArrb2, rAAV-GfaABC1D-mCherry-Stat1, rAAV-GfaABC1D-EGFP-Mfn2, and rAAV-GfaABC1D-mCherry-Mfn2. rAAV-GfaABC1D-GCaMP6m was used to monitor calcium dynamics in astrocytes. For neuronal calcium imaging, rAAV-CaMKIIα-GCaMP6m was used to monitor calcium dynamics in glutamatergic neurons. Mice were anesthetized with isoflurane and secured in a stereotaxic frame. Viral suspensions were bilaterally injected into the DG of the hippocampus using the following coordinates relative to bregma: anterior-posterior, −2.0 mm; mediolateral, ±1.3 mm; and dorsoventral, −2.3 mm. A total of 300 nl per site was delivered at a rate of 60 nl/min using a Hamilton microsyringe. The needle was left in place for 10 min after injection to minimize reflux along the needle track. Mice were allowed to recover and maintained for 3 weeks to ensure sufficient viral expression before undergoing CSDS paradigm and subsequent behavioral assays. Detailed viral constructs and target sequences are provided in table S3. TUNEL staining Apoptotic cells were detected using a commercial TUNEL assay kit (Elabscience, E-CK-A322). Frozen tissue sections were rinsed in PBS, fixed in 4% PFA for 30 min at room temperature, permeabilized with 200 μl of Proteinase K solution (37°C, 20 min), and washed three times in PBS. Sections were then equilibrated with 200 μl of terminal deoxynucleotidyl transferase (TdT) equilibration buffer in a humidified chamber (37°C, 30 min), followed by incubation in 200 μl of freshly prepared labeling reaction mixture (TdT equilibration buffer, 35 μl; nucleotide labeling solution, 10 μl; and TdT enzyme, 5 μl) under light-protected conditions in a humidified chamber (37°C, 60 min). After three PBS washes, nonspecific binding sites were blocked with 0.3% Triton X-100 in PBS (PBST) containing 10% donkey serum (room temperature, 2 hours, dark). Sections were subsequently immunolabeled with the primary antibody GFAP (4°C, overnight). Following three PBS washes, sections were incubated with corresponding secondary antibody conjugated to Alexa Fluor 488, and nuclei were counterstained with DAPI. Fluorescence images were acquired using a confocal laser scanning microscope under standardized acquisition parameters. TUNEL-positive astrocytes were quantified using ImageJ software. Enzyme-linked immunosorbent assays Hippocampal tissue was collected to measure inflammatory cytokines in mice. The hippocampus was homogenized in lysis buffer and subjected to centrifugation at 12,000g for 15 min to acquire supernatant. Enzyme-linked immunosorbent assay (ELISA) was performed following the manufacturer’s instructions to quantify the levels of IL-1β (Beyotime Biotechnology, PY2725), TNF-α (Beyotime Biotechnology, PT512), and IL-6 (Beyotime Biotechnology, PY2782). Quantification was normalized to tissue weight. Hippocampal tissues were dissected and homogenized in ice-cold RIPA lysis buffer supplemented with protease inhibitors. Homogenates were centrifuged at 12,000g for 15 min at 4°C, and supernatants were collected. Concentrations of proinflammatory cytokines IL-1β (Beyotime Biotechnology, PY2725), TNF-α (Beyotime Biotechnology, PT512), and IL-6 (Beyotime Biotechnology, PY2782) were quantified in supernatants using commercial ELISA kits according to the manufacturer’s protocols. Cytokine levels were normalized to tissue wet weight (picograms per milligram of tissue). Primary astrocyte culture and transfection Primary astrocytes were isolated from the hippocampus of neonatal WT and Arrb2 KO mice. Hippocampal tissue was carefully dissected, and meninges were removed before enzymatic digestion with 0.25% trypsin at 37°C. Digestion was terminated by adding Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin. The resulting cell suspension was plated in culture flasks and maintained in fresh medium, which was replaced every 3 days. Cells were used at day in vitro 14 for further experiments. For gene manipulation, astrocytes were seeded in six-well plates and transfected at 70 to 80% confluency. Small interfering RNAs (siRNAs) targeting Mfn2 or Stat1 were delivered using siRNA-Mate transfection reagent (Jima, China), whereas Stat1-expressing and Mfn2-expressing plasmids were introduced using Lipofectamine 3000 (Invitrogen, USA). All downstream assays were performed 48 hours posttransfection. All cell culture experiments were performed using at least three biologically independent astrocyte preparations derived from separate litters of mice, with each preparation tested in technical triplicates. The sequences of siRNAs used are listed in table S3. Mass spectrometry–based label-free proteomics Astrocyte lysates were subjected to immunoprecipitation using either anti–β-arrestin2 or IgG control antibodies. Proteins captured in the immunoprecipitates were digested with trypsin according to the Filter-Aided Sample Preparation protocol. The resulting peptides were desalted using C18 Empore SPE cartridges, concentrated via vacuum centrifugation, and resuspended in 40 μl of 0.1% (v/v) formic acid. Peptides were analyzed by liquid chromatography–tandem mass spectrometry (MS/MS) using a timsTOF Pro mass spectrometer (Bruker Daltonics) coupled to a NanoElute system (Bruker Daltonics). Peptides were loaded onto a reverse-phase trap column (Thermo Scientific Acclaim PepMap100) and separated on a C18 analytical column (Thermo Scientific Easy Column) using a linear gradient of buffer B (84% acetonitrile and 0.1% formic acid) at 300 nl/min. MS spectra were acquired in positive ion mode across mass/charge ratio 100 to 1700 and ion mobility 1/k0 of 0.6 to 1.6. Each sample underwent 10 cycles of PASEF (parallel accumulation serial fragmentation) MS/MS, with a target intensity of 1500, threshold of 2500, and active exclusion with 0.4 min release time. Raw MS data were processed using MaxQuant 1.5.3.17 for protein identification and quantification. For motif analysis, amino acid sequences surrounding modified sites (±6 residues, 13 amino acids total) were extracted and analyzed using the MEME Suite (https://meme-suite.org/meme/tools/meme) with parameters set to width of 13, occurrences of ≥20, and species-specific background. Prediction of transcription factors for Mfn2 The Mfn2 gene was analyzed using the PROMO database to predict potential transcription factor binding sites. Analysis parameters were set to a maximum matrix dissimilarity rate of 10, with the species specified as mouse. This prediction yielded 284 candidate transcription factors. To refine the list, a Venn diagram was generated to identify transcription factors that overlapped with β-arrestin2–binding proteins obtained from proteomic analysis. ChIP assay ChIP assays were conducted using the Millipore ChIP Assay Kit (17-295) according to the manufacturer’s guidelines. The Mfn2 promoter region was retrieved from NCBI (www.ncbi.nlm.nih.gov/), and potential STAT1 binding motifs were predicted using the PROMO database. A motif spanning 387 to 396 base pairs with the lowest dissimilarity score was selected for further analysis. Astrocytes were cross-linked with 1% formaldehyde and protease inhibitors at 37°C for 20 min. Cross-linking was quenched with 125 mM glycine and incubated for an additional 5 min at room temperature. Chromatin was then sonicated, and DNA-protein complexes were immunoprecipitated using anti-STAT1 or control IgG antibodies. Enriched DNA fragments were quantified by RT-qPCR using the following primers for the Mfn2 promoter: forward, GCCTCTGGCTTGAGTAAGAT; and reverse, AGAGACACACCCAGCACTTG. Luciferase reporter assay Human embryonic kidney–293 T cells were plated in 24-well plates and cultured to ∼70% confluence before transfection. Cells were cotransfected with plasmids encoding Stat1, WT Mfn2, mutant Mfn2 (Mutant), or a negative control, according to the experimental design. PRL-TK Renilla luciferase plasmid was included to monitor transfection efficiency. After 48 hours, luciferase activity was measured using the Dual-Luciferase Reporter Assay System (Promega), with Renilla luminescence quantified on a GloMax 20/20 Luminometer (Promega). Flow cytometry assay Primary astrocytes derived from WT and Arrb2 KO mice were transfected with Stat1 plasmids or a negative control vector. Mitochondrial membrane potential was assessed using the JC-1 Mitochondrial Membrane Potential Assay Kit (Thermo Fisher Scientific, catalog no. T3168), following the manufacturer’s protocol. Mitochondrial ROS levels were examined by MitoSOX Red Kit (MedChemExpress, HY-D1055), following the manufacturer’s protocol. After staining, astrocytes were immediately analyzed on a Guava easyCyte 8 flow cytometer (Merck Millipore, Billerica, MA, USA). The ratio of red (aggregated JC-1) to green (monomeric JC-1) fluorescence was used as an indicator of mitochondrial membrane potential integrity. Proximity ligation assay Protein-protein interactions between MFN2 and IP3R in astrocytes were analyzed using the Duolink In Situ Proximity Ligation Assay kit (Sigma-Aldrich, DUO92101) according to the manufacturer’s specifications. Cultured astrocytes were fixed with 4% PFA (15 min, room temperature), permeabilized with 0.3% Triton X-100 in PBS (10 min, room temperature), and blocked with Duolink Blocking Solution (37°C, 1 hour). Primary antibodies were applied overnight at 4°C. Following PBS washes, species-specific PLA probes were incubated (37°C, 1 hour). Subsequent ligation (30 min, 37°C) and amplification (100 min, 37°C, light protected) steps were performed. Nuclei were counterstained with DAPI. PLA signals were captured using a confocal microscope with consistent settings. PLA dots were quantified using ImageJ software. Measurement of calcium levels in astrocytes Primary astrocytes were transfected with the ER-targeted genetically encoded calcium indicator G-CEPIA1er (Addgene, plasmid no. 105012) using Lipofectamine 3000 according to the manufacturer’s protocol. Forty-eight hours posttransfection, cells were stimulated with 100 μM ATP for 5 min to induce calcium release. Live-cell imaging was performed using a confocal microscope equipped with a 37°C/5% CO2 environmental chamber. G-CEPIA1er fluorescence was captured, and mean fluorescence intensity was quantified using ImageJ software. Measurement of ATP levels and NAD+/NADH ratio Primary astrocytes derived from WT and Arrb2 KO mice were transfected with Stat1 plasmids or a negative control vector. Forty-eight hours posttransfection, cells were lysed in ice-cold extraction buffer. ATP concentrations were determined using an ATP assay kit (Beyotime Biotechnology, S0027) based on luciferase bioluminescence, whereas NAD+/NADH ratios were quantified colorimetrically using an NAD+/NADH assay kit (Beyotime Biotechnology, S0175S). Both assays were performed according to the manufacturer’s protocols. Glutamate measurement Hippocampal tissues were harvested from WT and Arrb2 KO mice following CSDS exposure. Tissue glutamate concentrations were quantified using a colorimetric Glutamate Assay Kit (Solarbio Life Science, catalog no. BC1585) according to the manufacturer’s instructions. Absorbance was measured spectrophotometrically, and glutamate levels were normalized to the corresponding tissue weight for comparison across experimental groups. In vivo fiber photometry recordings Calcium dynamics were monitored using a fiber photometry system (Thinker Tech Co., Nanjing, China) in freely moving mice. GCaMP6m fluorescence was excited by a 470-nm LED, which was passed through a band-pass filter, collimated, and reflected by dichroic mirrors before being focused through a 20× objective. To minimize photobleaching, the optical power at the fiber tip was maintained between 40 and 60 μW. A 405-nm LED served as an isosbestic reference to control for motion artifacts and nonspecific fluorescence. Fluorescent signals from the DG of the hippocampus were continuously recorded and synchronized with behavioral paradigms, including the SIT, TST, and FST. For the SIT, we recorded Ca2+ signals before and after the mice entered the interaction zone. Similarly, for the TST and FST, Ca2+ signals were collected before and after the movement events. Astrocytic or neuronal activity was elicited when mice entered the interaction zone in the SIT or when they moved during the TST and FST. The dotted line in the figures indicates the time points representing Ca2+ signals measured before and after these behavioral events. The transient nature of the Ca2+ signal reflects the brief increase in astrocytic or neuronal activity triggered by these events. Because astrocytic or neuronal activity was elevated only during the specific behaviors, the Ca2+ signals exhibited a transient pattern. Data acquisition and initial processing were carried out using Thinker Tech software, whereas post hoc alignment of calcium traces and behavioral events was performed in MATLAB using custom scripts provided by the manufacturer. Statistical analysis All quantitative data are expressed as means ± SEM. Statistical analyses were conducted using GraphPad Prism version 8.0. Unpaired two-tailed Student’s t tests were used for comparisons between two groups, whereas one-way or two-way analysis of variance (ANOVA) followed by Tukey’s post hoc test was applied for multiple-group analyses. A threshold of P < 0.05 was considered statistically significant. Each in vitro experiment was performed using at least three independent biological replicates and two technical replicates, with specific sample sizes and statistical details provided in the corresponding figure legends.