Astrocytic glycogen aggregates induce ATAD3A oligomerization mediated mitochondrial fragmentation and impede stroke recovery

Astrocytic glycogen aggregates induce ATAD3A oligomerization mediated mitochondrial fragmentation and impede stroke recovery | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Astrocytic glycogen aggregates induce ATAD3A oligomerization mediated mitochondrial fragmentation and impede stroke recovery Hongrui Zhu, Mengmeng Yang, Zhong Li, Chunliu Li, Shuaijie Sun, and 6 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6650856/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Ischemic stroke induces pathological glycogen deposition in astrocytes, but its role in post-injury neural dysfunction remains undefined. We reveal that glycogen-laden astrocytes in the ischemic penumbra undergo ATAD3A-dependent mitochondrial fragmentation via a stress granule-mediated mechanism, exacerbating neuronal injury and hindering functional recovery. Mechanistic studies demonstrate that glycogen aggregates sequester cytoplasmic HDAC3, enabling its translocation to mitochondria. There, HDAC3 deacetylates outer mitochondrial membrane protein ATAD3A, promoting oligomerization-driven mitochondrial fission. Astrocyte-specific ATAD3A ablation prevents stroke-induced synaptic disorganization, neural circuit disruption, and cognitive deficits. Therapeutically, pharmacologic or genetic exhaustion of astrocytic glycogen and HDAC3 inhibition reverse glycogen accumulation, rescue mitochondrial architecture/function, and restore synaptic plasticity and circuit reorganization, thereby acting synergistically to enhance post-stroke recovery. Our work identifies glycogen stress granules as pathogenic signaling hubs linking astrocytic metabolic stress to mitochondrial failure through compartmentalized HDAC3-ATAD3A crosstalk, and proposes a dual-target paradigm addressing both substrate overload and protein acetylation dynamics for stroke neurorestoration. Biological sciences/Neuroscience Biological sciences/Neuroscience/Glial biology Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Introduction Stroke is a leading cause of severe disability worldwide, driving increased focus on recovery mechanisms due to limited therapeutic strategies. Current approaches to restore neural function include rehabilitation, physical stimulation, and pharmacological interventions 1 . Molecular, cellular, and tissue network remodeling and behavioral systems occur dynamically throughout the entire stroke recovery course. These process include growth-promoting gene upregulation, axonal sprouting, dendritic spine turnover, synaptogenesis, axonal projections, and neural plasticity, such as motor-sensor map plasticity and glial cell homeostasis 2 . Astrocytes, strategically positioned between capillaries and neurons, supply energy and maintain synaptic homeostasis 3 . Astrocytes and neurons can form tripartite synapse structures. This structure ensures rapid astrocyte–neuron nutrient transport and promotes the shuttle of mitochondria from connected astrocytes when neurons meet the increased glucose demand resulting from high neuronal activity or a shortage of substrate supply because of capillary occlusion, such as ischemia 4 . Astrocytes metabolize glycogen into lactate, which fuels distal neurons via gap junctions 3 . Recent studies show that insoluble glycogen accumulates in astrocytes in human, primate, and rodent brains after ischemic stroke 5 . These accumulated glycogens which are difficult to be utilized in time are not conducive to long-term functional recovery in stroke animal models 5 .Yet the pathological impact of astrocytic glycogen deposition remains unclear, hindering therapeutic innovation. Glycogens, as storage polysaccharides, are steadily and dynamically balanced between glycogenesis and glycogenolysis for endogenous energy reserves under slight energy deficiency. Glycogen maintain the normal physiological functions of cells, such as protein glycosylation 6 , lactate release 3 , ATP synthesis, and GABA production, and is closely related to movement, memory, and lifespan 7, 8 . Under pathological conditions, insoluble glycogen aggregates (called polyglucosan bodies (PGBs), glycogen stress, Lafora bodies in Lafora disease) form in the brain 9 and are associated with aging, cognitive decline and dementia 10 , suggesting a role in impaired synaptic plasticity. Inhibiting glycogen aggregates formation reduces astrogliosis and neuroinflammation 11 . These insoluble, dense cytoplasmic polysaccharides, in addition to encapsulating large-molecule polyglucosides, also rich in glycogen-binding proteins like Stbd1 and laforin 12 . These aggregates resist lysosomal degradation and act as membrane-free compartments in stress signaling 6, 13 . Here, we report that glycogen aggregates deposit mainly in a subset of astrocytes, exacerbating injury and impairing brain repair. We identified glycogen-binding proteins in the penumbra and found astrocytic glycogen closely associates with mitochondrial fission receptor ATAD3A. Glycogen stress granules anchor the outer mitochondrial membrane and recruit cytoplasmic HDAC3 to deacetylate the ATAD3A outer segment of the mitochondrial membrane. Deacetylation initiated ATAD3A oligomerization-dependent mitochondrial fragmentation. Knocking down astrocytic ATAD3A reduced neurotoxicity, synaptic defects, compromised neural circuit reorganization and cognitive decline after stroke. Depleting glycogen granules and blocking downstream signaling improved sensorimotor recovery by enhancing synaptic and circuit plasticity. Thus, our findings reveal how astrocytic glycogen stress delays stroke recovery and highlight a clinically translatable strategy for treatment. Results Glycogen aggregates mainly increase in astrocytes from the ischemic penumbra after stroke Glycogen staining in photothrombotic and MCAO mouse models showed glycogen lesions enriched in the ipsilateral penumbra but absent contralaterally (Fig. 1a, b and Extended Data Fig. 1a). In MCAO mouse models, glycogen deposition peaked early and persisted for ~2 weeks, declining by day 21 (Extended Data Fig. 1c, d). Co-staining for Stbd1 (glycogen granule marker) 13 and GFAP (astrocyte marker) confirmed glycogen-laden astrocytes were concentrated in the penumbra (Extended Data Fig.1 e). Some Stbd1⁺ cells were non-astrocytic; co-staining with Iba1 and NeuN revealed minor glycogen accumulation in neurons and microglia (Extended Data Fig. 1f, g). RNA analysis of glycogen-rich versus glycogen-sparse regions showed decreased glycogenolysis genes (such as G6pc, Pygb) and increased glycogenesis genes (such as Gbe1, Gys1) (Fig. 1c), consistent with previous studies 5, 13 . Then, single-cell RNA sequencing distinguished four astrocyte clusters from the public database of mouse ischemic stroke (Extended Data Fig. 1b). Clustering based on glycogenesis genes (Gys1, Gbe1) revealed that clusters 0, 1, and 2 were enriched for glycogen deposition signatures, unlike cluster 3 (Extended Data Fig. 1h). It has been suggested that reactive astrocytes diversify beyond neurotoxic and neuroprotective astrocytes 14 , which can be further subdivided into unique subpopulations. We next performed clustering mapping on distinct astrocytes (neurotoxic astrocyte cluster A1; for example, C3 and Serpina3n ) and neuroprotective (neuroprotective astrocyte cluster A2; for example, Emp1 ) reactive astrocytes and pan-reactive astrocytes ( Gfap , Vim ) 15-17 with glycogen-laden astrocytes subsets. Results indicated that cluster 0 represented glycogen-enriched genes but lacked pan-reactive characteristics, which is regarded as glycogen resting astrocytes. While astrocytes with neuroprotective and glycogen-enriched genes were regarded as glycogen-reactive and neuroprotective cluster. Astrocytes with neurotoxic and glycogen-enriched genes were regarded as glycogen-reactive and neurotoxic cluster. And resting astrocyte with neither glycogen accumulation nor activation (Extended Data Fig. 1i). Decreased glycogenolysis, increased glycogenesis, decreased glycogen branching points all contribute to low soluble glycogen granule deposition, containing many glycogenin such as Stbd1 (covalently bound to the glucose polymer) and other proteins (such as laforin (Epm2a, GBE) 12 . Immunofluorescence confirmed glycogen accumulated astrocytes contained higher GYS1 (Extended Data Fig. 1j, k; area 01) or GBE1 (Extended Data Fig. 1j, l; area 02) expression. Post-stroke, resting and glycogen-resting astrocytes decreased, while glycogen-reactive neurotoxic or neuroprotective subtypes increased (Extended Data Fig. 1m). Thus, insoluble glycogen granules increase mainly in penumbral astrocytes after stroke. Astrocytic glycogen deposition correlates with ATAD3A oligomerization and neuropathology after stroke Glycogen pulldown and mass spectrometry identified a glycogen granule proteome enriched for mitochondrial pathways: respiratory chain complexes, mitochondrial respirasomes, and mitochondrial membrane protein complexes (Fig. 1d, Extended Data Fig. 2a, b, Supplementary Table 1). Cross-referencing with a mitoproteome database highlighted oxidative phosphorylation (Extended Data Fig. 2c, Supplementary Table 1) 18 , dynamics (MFF, DNM1L, GDAP1), homeostasis, and quality control (AFG3L2, CLPB, ATAD3A) (Fig. 1e, Extended Data Fig. 2c). ATAD3, a mitochondrial AAA+ ATPase, has three paralogs (ATAD3A, ATAD3B, and ATAD3C) in humans and only ATAD3A in mice. As a fission receptor of mitochondria, ATAD3A regulates mitochondrial dynamics, morphology, contact sites and nucleoid trafficking 19 . ATAD3A mutation in human cause neurodegeneration. ATAD3A oligomers recruit cytoplasmic Drp1 and mark mitochondrial damage, directly initiate mitochondrial fragmentation in neurological diseases 20, 21 . Next, we directly incubated the tissue lysate with purified glycogen and grasped the glycogen pellet via centrifugation. Results indicated that ATAD3A co-precipitated with glycogen pellets (Fig. 1f). To confirm whether glycogen directly binds to ATAD3A or requires indirect glycogen binding, in vitro protein binding experiments were performed by sequentially assigning purified glycogen and purified ATAD3A-His protein (Fig. 1g). We directly incubated the tissue lysate with purified glycogen and detected endogenous ATAD3A enrichment (lane 2 vs lane 1) (Fig. 1g). We next reincubated His-ATAD3A with the proteinase K-treated glycogen protein complex and observed no enrichment of endogenous ATAD3A or His-ATAD3A compared with the nonproteinase K-treated group (lane 5 vs lane 6) (Fig. 1g). These findings suggest that glycogen anchors the ATAD3A protein, which is not dependent on charge affinity or direct binding but is indirectly bound through glycogen-binding proteins. Serial sections showed spatial correlation between glycogen aggregates and ATAD3A oligomers in the ischemic penumbra (Fig. 1h, i). We detected ATAD3A oligomers in the injury foci at different time points after ischemic stroke. Results indicated that in the physiological state of the brain, there is a presence of small amounts of ATAD3A oligomers, and during the acute phase of ischemic stroke, the level of ATAD3A oligomers sharply increases (Fig. 1j). Immunofluorescence staining further confirmed ATAD3A oligomers easily form in ischemic area, especially in astrocytes and rarely form in normal uninjured brain area (Fig. 1k, l). Further staining results revealed the colocalization of ATAD3A oligomers and glycogen aggregates in astrocytes were more obvious in glycogen accumulated astrocytes (Fig. 1m, n). Whlie, minor oligomerization occurred in neurons or microglia/macrophage (Extended Data Fig. 3a, b). To further verify whether ATAD3A oligomers are mainly derived from astrocytes in vivo , astrocyte-specific ablation markedly reduced ATAD3A oligomers after stroke (Fig. 1o, Extended Data Fig. 3c, d). G6PT⁻ / ⁻ mice, a glycogen storage disease model of glycogen storage disease type Ib, manifested glycogen accumulation in the brain (Extended Data Fig. 3e, f). We next used G6PT -/- mice to assess whether glycogen aggregates predisposition is adverse to ischemic stroke prognosis. To rule out G6PT knockout-induced defects in peripheral immunity, we constructed a parabiosis model in which glycogen storage disease mice and normal wild-type mice were compared (Extended Data Fig. 3g, h). In the brains of the recipient mice, the PTS model was established 2 weeks after surgery. The results revealed that the in situ predeposition of glycogen in the brain worse stroke injury (Extended Data Fig. 3i, j) and promoted ATAD3A oligomer formation (Extended Data Fig. 3k). Glycogen aggregates thus exhibit mitochondrial affinity and may disrupt homeostasis after stroke. Glycogen aggregation causes mitochondrial dysfunction via ATAD3A Next, we observed that mitochondria area in glycogen-rich areas was smaller than in low-glycogen regions (Fig. 2a, b). We next sought to investigate how ATAD3A can be recruited to glycogen droplets and identify the binding partners of ATAD3A in the glycogen compartment. Immunoprecipitation identified Epm2aip as an ATAD3A-binding glycogenin proteins, which directly interact with glycogen 13, 22 (Extended Data Fig. 4a). And Epm2aip directly binds ATAD3A from the outer to the inner segment of the mitochondrial membrane (Extended Data Fig. 4b, c). Exogenously overexpressing Epm2aip-Flag and ATAD3A-Myc in human astrocytes (SVG p12) did not result in the formation of ATAD3A oligomers (Extended Data Fig. 3d). The overexpression of GYS1 was used to mimic a cell model of glycogen stress granule accumulation in astrocytes (SVG p12 GYS1 OE cells) (Extended Data Fig. 4e, f) 11, 13 . Overexpressing GYS1 in astrocytes induced glycogen granules and ATAD3A oligomerization, reversed by Epm2aip knockdown (Fig. 2c). Knockdown of ATAD3A in SVG p12 GYS1 OE cells reduced mitochondrial fragmentation (Fig. 2d, e). Under physiological conditions, the ATAD3A monomer helps maintain the formation and morphology of mitochondrial cristae by composing mitochondrial contact sites and cristae organizing system (MICOS) complexes 23 . The transmission electron microscopy results revealed that the number of cristae in the mitochondria adjacent to glycogen stress granules decreased. While, ATAD3A knockdown restored cristae structure (Fig. 2f, g). ATAD3A oligomers have been previously reported to participate in mitochondrial nucleoid formation and perturb mtDNA maintenance and the machinery of protein synthesis 20 . Astrocytic glycogen aggregates deposition impaired ATP synthesis (Fig. 2h) and oxygen consumption, which can be rescued by ATAD3A knockdown (Fig. 2i, j). ATAD3A 200 aa contain the CC1~2 region, and ATAD3A 300 aa contain the CC1~2 and TMS regions, which were previously reported to anchor mitochondria 20, 24 . SVG p12 GYS1 OE cells were expressed with ATAD3A 200 aa or 300 aa GFP tag proteins. Only mitochondrial-anchored ATAD3A fragments (300 aa) oligomerized (Extended Data Fig. 4g). Together, astrocytic glycogen granules lead to ATAD3A oligomerization-triggered mitochondrial dysfunction. The glycogen chamber contains HDAC3, and ATAD3A oligomerization is dependent on deacetylation Above evidence shown that ATAD3A oligomerization requires the accumulation of astrocytic glycogen aggregates. However, under physiological conditions, ATAD3A does not dramatically oligomerize in non-glycogen aggregates deposited cells and un-ischemic brain tissue. Mitochondria also do not exhibit obvious dysfunction in non-glycogen stress granule-deposited cells. Therefore, we speculate that the glycogen chamber may carry some proteins to the mitochondria, promoting ATAD3A oligomerization. To explore what candidate partners are involved in it, the proximity-biotinylating enzyme TurboID was reconstituted with ATAD3A to detect biotinylated proteins adjacent to ATAD3A (Fig. 3a). TurboID proximity labeling identified 182 ATAD3A-neighboring proteins (Fig. 3b, Supplementary Table 1). Subcellular and pathway analyses highlighted mitochondrial protein localization (Extended Data Fig. 5a). KEGG pathway analysis revealed that protein polymerization and protein localization to mitochondria are associated with ATAD3A (Extended Data Fig. 5b). We intersected proteins between the glycogen proteome and the biotinylated proteome and found that many proteins, such as the nuclear protein histone deacetylase 3 (HDAC3), Dnm1l (Drp1), ATAD3A, and VDAC1, were enriched (Supplementary Table 1). The mitochondrial membrane outer segment of ATAD3A deacetylation has been reported to be an independent factor initiating ATAD3A self-oligomerization 21 , but the enzymes involved in ATAD3A deacetylation are currently unclear. We speculate that HDAC3 may be a critical candidate effector (Fig. 3c). We further confirmed that the glycogen protein Epm2aip can interact with HDAC3 (Extended Data Fig. 5c). Screening deacetylases in the Sirt family (Extended Data Fig. 5d) and HDAC superfamily (Extended Data Fig. 5e) revealed Sirt1, HDAC3 and HDAC6 as candidates. To determine whether ATAD3A can be acetylated by HDACs or Sirts, SVG p12 cells were transfected with Myc-tagged ATAD3A and treated with an inhibitor of sirtuin family deacetylases (nicotinamide; NAM) or the pan-HDAC inhibitor trichostatin A (TSA). Pan-HDAC inhibition, but not sirtuin inhibition, increased ATAD3A acetylation (Fig. 3d). GST pull-down and split-luciferase assays confirmed HDAC3–ATAD3A interaction (Fig. 3e, f , Extended Data Fig. 5f, g). Domain mapping of human HDAC3 and human ATAD3A revealed that the regions of homology are located only in the CC domain (mitochondrial membrane outer segment) of hATAD3A and the deacetylase domain of hHDAC3 (Fig. 3g). The human ATAD3A K135 site (K134 in mice) was previously identified as a critical acetylation site 20 , and its deacetylation triggers ATAD3A self-oligomerization. Cytoplasmic HDAC3 was previously reported to translocate to mitochondria to execute deacetylation as a ‘decision maker’ under conditions of cell stress 25 . Therefore, we hypothesized that HDAC3 can modify ATAD3A to participate in oligomerization. Cells were transfected with Myc-tagged ATAD3A with two deacetylase-dead HDAC3 mutants (H134/135A and Y298F) or pharmacological pan-HDAC inhibition (TSA). We found that acetylated ATAD3A levels subsequently decreased after HDAC3 WT overexpression. However, two deacetylase-dead HDAC3 mutants (mainly Y298F) or pan-HDAC inhibition restored and augmented acetylated ATAD3A levels, respectively (Fig. 3h). And HDAC3 knockout increased ATAD3A acetylation (Fig. 3i). When HDAC3 WT but HDAC3 Y298F were overexpressed in HDAC3 deficient cell, ATAD3A acetylation decreased correspondingly (Fig. 3i). However, the other candidate partner, HDAC6, did not yield positive results (Extended Data Fig. 5h). Lysine acetyltransferases (KATs) and deacetylases can both regulate protein acetylation. We next searched for KATs responsible for ATAD3A acetylation (Extended Data Fig. 5i). The results indicated that Kat8 is a potential effector. Kat8 is shuttled between the nucleus and mitochondria and is required to rescue transcriptional and respiratory defects via its catalytic activity 26 . Co-expression experiments confirmed that Kat8 increases ATAD3A acetylation, which is reversed by wild-type HDAC3 but not a catalytically dead mutant (HDAC3 Y298F ) or HDAC6 (Extended Data Fig. 5j, k). The lysine 134 residue in mice, which is conserved among species, resides in the linker between the CC1 and CC2 domains and is accessible for reversible acetylation 20 . To further investigate ATAD3A deacetylation mediated by HDAC3 in ATAD3A oligomerization, an acetylation-mimetic mutant (K134Q) and a deacetylation-mimetic mutant (K134E) were generated. The K134E mutant enhanced ATAD3A oligomerization, while K134Q had no effect, even with HDAC3 co-expression (Fig. 3j). In vitro assays with purified proteins confirmed that Kat8 directly acetylates ATAD3A (Fig. 3k), and HDAC3 deacetylates it but not catalytically dead mutant (HDAC3 Y298F ) or Stirt1(Fig. 3l, m). Subsequently, Proximity ligation assay (PLA) was performed to visualize the interaction between ATAD3A and HDAC3 proteins in astrocytes under physiological or pathological conditions. Under physiological conditions, we observed an interaction between HDAC3 and ATAD3A in astrocytes, with minimal levels detected in non-astrocytic cells. Following ischemic stroke, this interaction was markedly enhanced in astrocytes, which was quantified by an increase in HDAC3-ATAD3A interaction spots per GFAP area (Fig. 3n, o). Critically, the stroke-induced enhancement of the HDAC3-ATAD3A interaction was abolished by pre-knockdown of HDAC3 in astrocytes (Extended Data Fig. 5l, m). And under physiological conditions, there is an extremely weak presence of ATAD3A oligomers. When ischemic stroke occurs, the level of ATAD3A oligomerization sharply increases. While, ATAD3A oligomerization significantly decreases after astrocytic HDAC3 knockdown (Fig. 3p). In addition, astrocytic ATAD3A acetylation was significantly downregulated, which is rescued by astrocyte-specific HDAC3 knockdown (Fig. 3q). Thus, HDAC3 promotes ATAD3A oligomerization by direct deacetylation both in vivo and in vitro. HDAC3 regulates ATAD3A-mediated astrocytic mitochondrial dysfunction ATAD3A deacetylation promotes Drp1 recruitment and mitochondrial fragmentation 20 . To further validate the correlations among ATAD3A acetylation and Drp1 and HDAC3 protein levels in vivo after ischemic stroke, results indicated that ATAD3A acetylation decreased over time (Fig. 4a). Endogenous ATAD3A immunoprecipitation was performed to assess the relationship between ATAD3A acetylation and mitochondrial function. The phosphorylation of Drp1 at Ser 600 (the active site) or Ser 579 (the effector site) in mice, which is equivalent to Ser616 and Ser637 in humans, is required from the cytosol to the mitochondrial outer membrane for maximal mitochondrial fragmentation 27 . We found that the level of mitochondrial fragmentation increased, accompanied by decreased ATAD3A acetylation and increased HDAC3 binding after ischemic stroke, and that the increase in Drp1 activity for mitochondrial fragmentation manifested as increased serine 579 site phosphorylation (Fig. 4b). In addition, HDAC3-mediated deacetylation of ATAD3A increased the binding ability between ATAD3As (Fig. 4c). Treatment with an HDAC3-specific inhibitor decreased the binding ability between ATAD3As (Extended Data Fig. 6a~c). HDAC3 interacted with the ATAD3A N-terminus (1~200 aa) (Extended Data Fig. 6d); truncation mutants confirmed after transfecting the truncated mutants of ATAD3A in cells, the ATAD3A-GFP 300aa (containing CC1~2 and TMS domains) mutant strongly increased ATAD3A dimers under HDAC3-mediated deacetylation (Fig. 4d), whereas mitochondrial targeting transmembrane sequence (TMS) deficiency (ATAD3A-GFP 100aa or ATAD3A-GFP 200aa) abolished dimerization (Extended Data Fig. 6e). The acetylated ATAD3A CC domain is required for the recruitment of the Drp1 mitochondrial location 24 . The expression of the ATAD3A-GFP 200 aa mutant or full ATAD3A-GFP induced Drp1 translocation to interact with the ATAD3A outer mitochondrial membrane segment, which was enhanced by HDAC3 and blocked by RGFP966 treatment (Fig. 4e, f, Extended Data Fig. 6 f~h). We further investigated whether and how HDAC3 regulates ATAD3A-mediated astrocytic mitochondrial dysfunction. It was previously reported that mitochondria presented moderate fragmentation upon exogenous ATAD3A overexpression 20, 28 and that co-expression with HDAC3 augmented branched mitochondrial structures in SVG p12 cells. Compared with ATAD3A WT cells, SVG p12 cells expressing ATAD3A K134Q (acetylation gain-of-function mutation) coupled with HDAC3 presented relatively reduced fragmented mitochondria (Extended Data Fig. 6i, j). Furthermore, ATAD3A K134Q + HDAC3 overexpression attenuated mitochondrial respiratory defects compared with ATAD3A WT - + HDAC3-expressing SVG p12 cells (Extended Data Fig. 6k, l); ATAD3A K134Q + HDAC3 overexpression improved the maximal and spare respiratory capacity but had minor effects on mitochondrial function in ATAD3A-overexpressing SVG p12 cells. However, obvious mitochondrial fragmentation (Extended Data Fig. 7 i, j) and severe mitochondrial respiratory defects (Extended Data Fig. 6 k, l) were also observed in HDAC3-overexpressing cells, indicating that endogenous ATAD3A was also deacetylated. To eliminate the background effects of endogenous ATAD3A in SVG p12 cells, we next performed experiments in ATAD3 KD (ATAD3A knockdown) SVG p12 cells (SVG p12 ATAD3 KD ). Basal, maximal and spare respiratory capacity were significantly compromised after ATAD3A WT + HDAC3 overexpression in SVG p12 ATAD3 KD cells, whereas ATAD3A K134Q overexpression or HDAC3-specific inhibitor treatment throughout the entire stage effectively rescued these effects (Fig. 4g, h). SVG p12 ATAD3 - KD cells overexpressing ATAD3A WT + HDAC3 exhibited extensive mitochondrial fragmentation relative to ATAD3A K134Q overexpressing cells (Fig. 4i, j). Moreover, overexpression of HDAC3 or HDAC3 coupled with ATAD3A WT under RGFP966 did not cause obvious mitochondrial fragmentation (Fig. 4i, j). Thus, HDAC3-mediated deacetylation drives astrocytic mitochondrial dysfunction. Blocking astrocytic ATAD3A oligomerization alleviates damage and improves recovery after stroke The homeostasis of mitochondrial function in astrocytes is crucial for neuronal damage and repair after stroke injury 4, 29 . Aldh1l1 CreErt ; ATAD3A fl/Δ mice with astrocyte-specific heterozygous knockout (or called ATAD3A Ast Het ) showed normal development, memory and motor function (Fig. 5i, j, Extended Data Fig. 7a~d). After tMCAO, there was no difference in the areas with cerebral blood flow reduction between the two genotypes (Extended Data Fig. 7e, f). But ATAD3A Ast Het had smaller Infarct volume than controls (Fig. 5a~c). This indicated that astrocytic ATAD3A oligomerization signaling abrogation alleviates neuronal injury during the acute phase. The brain repair process after stroke during the subacute phase (especially at 1~2 weeks) 30 , which can last for 90 days after injury, is reported to be a critical period of rehabilitation. This process is strongly associated with astrocyte homeostasis 31 . Whole-brain MRI scanning was performed at 14 days revealed a reduced infarct edema volume and delayed deterioration in ATAD3A Ast Het mice (Fig. 5d). Fragmented mitochondria derived from astrocytes are toxic to oligodendrocytes for remyelination 32 . Quantification of Luxol fast blue (LFB) staining revealed that ATAD3A Ast Het mice presented more myelin coverage after stroke than ATAD3A fl/fl mice (Extended Data Fig. 7g, h). And directionally encoded color (DEC) maps of the brains showed improved circuit reconstruction (Fig. 5e) and white matter integrity in ATAD3A Ast Het mice (Fig. 5f). And thickness, discontinuity and defects of myelin sheaths in ipsilesional external capsule were less frequently observed in ATAD3A Ast Het mice than in control mice (Extended Data Fig. 7i~k). Behaviorally, ATAD3A Ast Het mice showed better sensorimotor recovery in adhesive removal test (Fig. 5g) and rotarod test (Fig. 5h) from day 14. The white matter (WM)-enriched corpus callosum and external capsule are associated with nerve fiber conduction, especially transcallosal connections to the denervated striatum after stroke 33 , which is responsible for sensorimotor function regulation. Mitochondrial dysfunction in astrocytes is not conducive to WM repair or nerve fiber conduction reconstruction 34 . Immunostaining for MBP (myelin basic protein) and SMI32 (a marker of demyelinated axons) in the external capsule and striatum revealed improved transcallosal connections in ATAD3A Ast Het mice (Extended Data Fig. 7i, m). Evoked compound action potentials (CAPs) display a biphasic wave with an early peak representing fast-conducting myelinated axons (N1) and a delayed peak representing slow-conducting unmyelinated axons (N2) (Extended Data Fig. 7n). Compared with those in ATAD3A Ast Het mice, impaired myelinated axons are more obvious in ATAD3A fl/fl mice after stroke and exacerbate the reduction in N1 amplitude. The amplitude of the N2 component did not markedly differ between the two stroke groups (Extended Data Fig. 7o). In addition, spatial recognition memory or spatial learning tests (novel object recognition (Fig. 5i), Morris water maze (Fig. 5j) also indicated better recovery in ATAD3A Ast Het mice. To eliminate differences in memory function recovery caused by lesion size variability, bilateral hippocampal photothrombosis was induced before astrocytic ATAD3A knockout (Fig. 6a). TTC staining confirmed confirmed comparable lesion areas between ATAD3A fl/fl and Aldh1l1 CreErt ; ATAD3A fl/Δ mice (Fig. 6b, c). Tamoxifen-induced astrocytic ATAD3A knockout was followed one week later by fear conditioning to assess hippocampal function. Post stroke contextual fear conditioning results indicated that ATAD3A fl/fl mice with hippocampal photothrombosis showed no increased freezing time upon re-exposure to the conditioning context, whereas ATAD3A Ast Het mice exhibited significantly elevated freezing time (Fig. 6d, Supplementary video 1). In addition, the calcium signal intensity of hippocampus in the conditioning context was significantly lower in ATAD3A fl/fl mice than in ATAD3A Ast Het mice (Fig. 6e~g). For sensorimotor assessment, photothrombosis was applied to the sensorimotor cortex prior to astrocytic ATAD3A knockout, followed by behavioral and electrophysiological tests (Fig. 6h). The ATAD3A fl/fl mice displayed higher asymmetry index in the cylinder test, compared with ATAD3A Ast Het mice (Fig. 6i). After AAV-CaMKII-ChR2-eGFP injection into the contralateral somatosensory area and prior astrocytic ATAD3A knockout, in vivo somatosensory recordings revealed stronger bursting activity in ATAD3A Ast Het mice post-stimulation (Fig. 6j, k, Supplementary video 2), along with a more pronounced cumulative percentage of evoked responses with rising stimulus power (Fig. 6l, m). Collectively, abrogating astrocytic ATAD3A oligomerization mitigates neuronal injury, improves white matter integrity, and promotes long-term functional recovery after stroke. Astrocytic ATAD3A oligomerization inhibition remodels the inflammatory microenvironment To address astrocyte heterogeneity directly after astrocytic ATAD3A oligomerization was abolished, scRNA-seq of astrocytes from ATAD3A fl/fl and ATAD3A Ast Het mice at day 7 post-stroke identified major brain cell types (Extended Data Fig. 8a). All expected cell types were identified in both sample sets, such as astrocytes ( Gfap , Aldh1l1 ), microglia ( Hexb , Tmem119 ), macrophages ( Mcr1 , Pf4 ), oligodendrocytes ( Olig1 , Olig2 and Mog ) and their precursors (OPCs), endothelial cells ( Hemk1 , Kcnj13 ), pericytes ( Rgs5 , Cspg4 ) and stromal cells (for example, T cells and Erythrocytes) 16, 17 (Fig. 7a, Extended Data Fig. 8b). And the percentage of astrocytes decreased notably with increasing percentages of oligodendrocytes, endothelial cells and neurons in ATAD3A Ast Het mice (Fig. 7b). Subsequently, using GYS1 and GBE1 as glycogen-enriched markers (Extended Data Fig. 1j), astrocytes clustered into five subsets: resting (c1, c2), glycogen-resting, glycogen-reactive and neuroprotective cluster (A2 cluster), and glycogen-reactive and neurotoxic cluster A1 according to scRNA-seq results (Fig. 7c, Extended Data Fig. 8c). Glycogen markers (GYS1, GBE1) were high in the glycogen resting astrocyte cluster, the glycogen reactive astrocyte A2 cluster and the glycogen reactive astrocyte A1 cluster (Fig. 7d). Mapping panreactive gene sets and A1-specific and A2-specific gene sets (Fig. 7e), HDAC3 was enriched in glycogen-reactive A1 astrocytes (Fig. 7h). All distinct astrocyte populations expressed Aqp4, the water channel located on vascular endfeet in all astrocytes in the brain 35 . We mapped distinct astrocytes from regions of olfactory-specific astrocytes ( Islr , Islr2 ), cerebellum astrocytes ( Gdf10 ), telencephalon-specific astrocytes ( Mfge8 , Lhx2 ), non-telencephalon astrocytes ( Agt (angiotensinogen)) and dorsal midbrain astrocytes ( Myoc ) (Extended Data Fig. 8d). We confirmed that all the distinct astrocyte populations were located mainly in telencephalon or non-telencephalon regions during ischemic cortical infarction (Fig. 7f). Immunofluorescence confirmed HDAC3 sequestration and ATAD3A oligomerization in glycogen-laden astrocytes (Fig. 7g). Pathway enrichment analysis revealed that the functional characteristics of resting astrocyte c1 cells included ‘generation of precursor metabolites and energy’, ‘small-molecule catabolic process’ and ‘small-molecule biosynthetic process’. Resting astrocyte c2 mainly represented ‘cytoplasmic ribosomal proteins’ and the ‘microglia pathogen phagocytosis pathway’. Glycogen-reactive A2 astrocytes mainly represented ‘regulation of synapse structure or activity’ and ‘neuron projection morphogenesis’. Glycogen-reactive A1 astrocytes mainly represented ‘SRP-dependent cotranslational protein targeting to the membrane’, ‘cellular responses to stimuli’, ‘aggrephagy’ and other pathways (Extended Data Fig. 8e, f, Supplementary Table 2). Compared with glycogen-reactive A2 astrocytes, glycogen-reactive A1 astrocytes displayed an imbalance in homeostasis and inflammatory stress after stroke (Fig. 7h, i). Using a validated ATAD3A and GYS1 antibodies to confirm their specificity (Extended Data Fig. 8l, m), we stained human brain samples from stroke patients (Supplementary Table 3). Human stroke samples confirmed more glycogen-laden astrocytes (GYS + GFAP + ) (Fig. 7j, k) and ATAD3A aggregates (ATAD3A + GFAP + ) (Fig. 7j, l) in ischemic regions. Transcriptional profiling suggested that abrogation of astrocytic ATAD3A oligomerization is associated with a shift in the ischemic microenvironment, including a reduction in the proportion of astrocytes with an A1-like transcriptional signature (Extended Data Fig. 8g). Unbiased clustering showed a reduction in the astrocyte subpopulation annotated as glycogen-reactive A1 in ATAD3A Ast Het mice (Fig. 7m). The A1 astrocytic response is directly attributed to fragmented mitochondria and activated microglia/macrophages 36, 37 , which propagate inflammatory reactions and tissue damage in the brain 38 . We extracted primary astrocytes from adult ATAD3A Ast Het and ATAD3A fl/fl mice. The results revealed that pDrp1 s579 and pDrp1 s600 are both increased in astrocytes after ischemic stroke and can be rescued after ATAD3A knockout (Fig. 6n). Moreover, macrophages/microglia all displayed an anti-inflammatory gene signature in ATAD3A Ast Het mice compared with ATAD3A fl/fl mice (Extended Data Fig. 8h). Using neural communication networks to analyze predefined cell groups from scRNA-seq expression data 39 , we found that connections within neuron c1, neuron c2 and neuron c3 and connections between resting astrocyte c2 (or c1) and distinct neurons were all enhanced in ATAD3A Ast Het mice (Extended Data Fig. 8i, j). The projection network of each individual interaction pair, gap junction-related gene links (such as Gjb2 , Gjb6 , and Gja1 ) and neurosynaptic-related gene pairs (such as Nrxn1 , Nrxn2 , Nrxn3 , and Nlgn3 ) was more obvious in ATAD3A Ast KO mice (Extended Data Fig. 8k). Transcriptional profiling suggested that abrogation of astrocytic ATAD3A oligomerization is associated with a shift in the ischemic microenvironment, including a reduction in the proportion of astrocytes with an A1-like transcriptional signature, increasement of anti-inflammatory gene signature and alterations in enhanced ligand-receptor pairs suggestive of modified cell-cell communication. Astrocytic glycogen depletion and HDAC3 inhibition act synergistically to enhance post-stroke recovery Considering that glycogen depletion agents and HDAC3 inhibitors have drugs developed before clinical trials, we next investigated whether astrocytic glycogen depletion and HDAC3 inhibition can act synergistically to enhance post-stroke recovery. Preclinical studies have shown that cotadutide, a dual glucagon-like peptide 1 (GLP-1R) and glucagon receptor (GCGR) agonist, can effectively act on the brain and peripheral organs 40 , reducing glycogen accumulation, inflammatory stress and fibrosis in some chronic diseases 41-43 . We next explored the changes in glycogen deposition in the brain of ischemic stroke under different drug treatment groups. RGFP966 alone treatment cannot reduce the deposition of glycogen in the brain, while adding cotadutide can effectively deplete the glycogen deposition (Extended Data Fig. 9a, b). Next, astrocyte-specific protein (Ast ER-BioID HA ) mice were used to label biotinylated proteins derived from astrocytes for global protein oxidation levels and ATAD3A oligomers detection (Extended Data Fig. 9c). Astrocytes lost the expression of EGFP and expressed ER-BioID2 HA after Aldh1l1 -CreERT ; ER-BioID HA mice were treated with tamoxifen (Extended Data Fig. 9d). These Ast ER-BioID HA mice were treated with RGFP966, cotadutide, or cotadutide combined with RGFP966 (CotR) after ischemic stroke. Biotinylated proteins were enriched and obtained for global protein oxidation levels and ATAD3A oligomers detection (Extended Data Fig. 9e). Treatment with either RGFP966 or cotadutide alone reduced global protein oxidation and ATAD3A oligomerization; these reductions were more pronounced with CotR combination therapy (Extended Data Fig. 9f~h). Therefore, cotadutide alone can effectively reduce the deposition of glycogen granules in the brain, it cannot completely reduce the oxidation level and ATAD3A oligomerization. When combination therapy, it can both reduce glycogen deposition and fully block ATAD3A oligomerization signals. Following sensory cortex ischemic stroke, mice received systemic CotR therapy, with or without pre-overexpression of oligomer-prone ATAD3A (AAV-ATAD3A K134E ) (Extended Data Fig. 10a, b). Pre-overexpression of ATAD3A K134E exacerbated pathological damage and mildly impaired motor recovery, increasing foot faults and asymmetry indices compared to controls (Extended Data Fig. 10c, d). It also consistently reduced synaptic spine density in the sensorimotor cortex, regardless of stroke (Extended Data Fig. 10e, f). Although daily CotR treatment markedly reduced the asymmetry index, this protective effect was counteracted by ATAD3A K134E pre-overexpression (Fig. 8a). Motor network excitability of the ipsilateral ischemic cortex during the repair phase is related to motor functional recovery during the repair phase of stroke 2, 44 . CotR also restored stroke-impaired cortical excitability, as measured by Ca²⁺ ultrasensitive intracellular Ca 2+ sensor for pyramidal neurons on day 14, an effect partially reversed by ATAD3A K134E pre-overexpression (Fig. 8b~c, Extended Data Fig. 10g), as shown by the locomotion heatmap results (Extended Data Fig. 10h). Whole-brain MRI scanning was subsequently performed after stroke. The results revealed a reduced infarct edema volume and augmented FA value after CotR therapy, which were counteracted by ATAD3A K134E pre-overexpression (Extended Data Fig. 10i, j). Synaptogenesis and synaptic transmission in the peri-infarct cortex are critical for circuit remapping and neural network plasticity 30, 45, 46 . On day 16, CotR rescued the stroke-induced reductions in dendrite length and spine density in the peri-infarct cortex, protective effects again attenuated by ATAD3A K134E overexpression (Extended Data Fig. 10k~m). Excitatory synaptic transmission was usually used to assess functional plasticity. Electrophysiological recordings on day 14 also confirmed that CotR reversed the stroke-induced decrease in mEPSC frequency, indicating enhanced neural functional plasticity in the peri-infarct region (Extended Data Fig. 10n~p). Axonal sprouting involves short-distance axonal sprouting from the uninjured peri-infarct region to the ischemic penumbra (Extended Data Fig. 10k) 46, 47 and long-distance axonal sprouting from uninjured corticospinal axons to denervated spinal cord and brain nuclei (such as the red nucleus and facial nucleus) (Extended Data Fig. 10q) 1, 48 , which are essential for post-stroke circuit remodeling. By BDA tracing, neural fibers from uninjured circuits crossed obviously the denervated region under CotR intervention. However, this effect was similarly blunted by ATAD3A K134E pre-overexpression (Extended Data Fig. 10r, s). Projection neurons in the cortex, which reside primarily in layers II/III, send axons to distant contralateral brain targets 49, 50 . The transcallosal connection is responsible for extending axons to mirror-image locations in the contralateral functional area, enabling information intercommunication and integration. The transcallosal connection contributes to the recovery process of sensorimotor function 51, 52 . After a focal stroke, the recovery of sensory responsiveness in the lesion periphery is mediated by interhemispheric synaptic inhibition, which is selectively activated by contralateral excitatory inputs 51 . We injected AAV-CaMK II-ChR2-eGFP in the contralateral mirror-image somatosensory area, and an in vivo multichannel electrophysiological electrode was implanted in the peri-infarct somatosensory cortex on the day of stroke (Fig. 8d). 40-Hz optogenetic stimulation was performed 15 days after surgery and drug intervention. The behavior of the mice revealed that the transcortical corpus callosum connection was initially established when light stimulation was given to the opposite undamaged side (Supplementary video 3). CotR enhanced transcallosal connectivity, evidenced by greater optogenetically evoked bursting activity and response rates in the somatosensory cortex (Fig. 8e~h). So, pharmacologic exhaustion of astrocytic glycogen and HDAC3 inhibition enhances recovery after stroke. To eliminate the off-target effects and specificity issue of drugs, astrocyte-targeted viruses were used. Overexpression of glycogen-degrading enzyme PYGB effectively reduced glycogen deposits, while HDAC3 knockdown did not (Extended Data Fig. 11a~e). The stroke mice with astrocytic glycogen depletion or HDAC3 knockdown displayed significantly reduced foot faults in grid-walking test and asymmetry index in the cylinder test, compared with vehicle-treated mice. And their combination acted synergistically to produce the greatest functional recovery (Fig. 8i), consistent with greater optogenetically evoked bursting activity and response rates in the somatosensory cortex, compared with single treatment (Fig. 8j~l). Discussion Astrocytes support brain homeostasis by providing energy substrates and shuttle mitochondria to neurons 4, 5, 29 . Large insoluble glycogen aggregates are associated with neuronal dysfunction, aging, and cognitive decline 10, 53 . We found that glycogen-laden astrocytes in the ischemic penumbra exacerbate stroke injury and impair recovery. Astrocytic glycogen aggregates sequester cytoplasmic HDAC3, enabling its translocation to mitochondria. There, HDAC3 deacetylates outer mitochondrial membrane protein ATAD3A, promoting oligomerization-driven mitochondrial fragmentation. Exhaustion of astrocytic glycogen and HDAC3 inhibition reverse glycogen accumulation, rescue mitochondrial architecture/function, and restore synaptic plasticity and circuit reorganization, thereby acting synergistically to enhance post-stroke recovery (Fig. 8m). Glycogen granules form as electron-dense, organelle-like structures adjacent to the endoplasmic reticulum and mitochondria 12 . Under ischemic stress, dysregulated glycogen metabolism-marked by increased glycogenesis and reduced glycogenolysis-promotes the accumulation of insoluble, Stbd1-positive glycogen granules 12 , 54 , consistent with prior reports 5, 55 . While soluble glycogen supports neuronal energetics via lactate shuttling 3 , the role of insoluble aggregates has been obscure. Astrocytic mitochondria provide most of the energy for metabolism, and their homeostasis is critical for stroke recovery 4, 56 . Previous studies have reported that oxidative phosphorylation, mitochondrial metabolism, and mitochondrial ultrastructure are all impaired in cells in vitro and in vivo models of glycogen storage disease. The mitochondrial content also decreases, which is coupled with decreased mitochondrial biogenesis and increased mitochondrial fragmentation 57, 58 . These results were clearly observed in glycogen stress granule-enriched astrocytes in this study. However, the intricate relationship between mitochondrial dysfunction and glycogen storage has not been fully elucidated. A previous study revealed that Stbd1 can recruit glycogen to endoplasmic reticulum (ER)-mitochondria contact sites 59 , which may be correlated with mitochondrial morphology and mitochondrial dynamics, such as mitochondrial fusion (elongation) and fission (fragmentation) 60 . In addition, fragmented mitochondria are more obvious under nutrient oversupply conditions 60, 61 . Given that glycogen itself has the “sweet” side of ER‒mitochondria contact sites, we propose that glycogen overload may result in mitochondrial fragmentation. This hypothesis was also mentioned by Demetriadou et al . but lacks experimental evidence 60 . Our proteomic and functional analyses reveal that glycogen stress granules recruit mitochondrial proteins and localize to ER–mitochondria contact sites, directly implicating them in mitochondrial dysfunction. Notably, we identified the glycogen-anchored protein Epm2aip as a physical linker between glycogen granules and ATAD3A-a mitochondrial fission receptor situated at ER-mitochondria junctions. And Epm2a (Laforin) also participates in the formation of insoluble and neurotoxic glycogen aggregates and is enriched in glycogen condensates 9, 13 . ATAD3A is located at ER‒mitochondrion contact sites 19, 24 , and its deacetylation triggers self-oligomerization 20 . The CC domain of ATAD3A, an inhibitory segment that maintains the steady state, is key for mitochondrial fission and mtDNA stability. Deacetylation of the CC domain at residue K135 (K134 in mice) is essential for signaling-induced self-oligomerization 20 . We now establish HDAC3 as the deacetylase responsible for this modification. Under glycogen overload, cytoplasmic HDAC3 is recruited to glycogen granules and recruited to mitochondria, where it deacetylates ATAD3A, driving oligomerization mediated fragmentation. This pathway represents a molecular bridge between glycogen metabolism and mitochondrial dynamics, substantiating the hypothesis that glycogen overload disrupts mitochondrial integrity. ATAD3A oligomerization, a terminal pathological signaling after glycogen accumulation, is specifically genetically knocked down in astrocytes before stroke. Indeed, our results demonstrated that a reduction in ATAD3A oligomerization alleviates acute damage and improves recovery after stroke. Astrocytic mitochondrial dysfunction is not conducive to the survival of adjacent neuronal cells, especially during the course of stroke 56 . Electron microscopy revealed that morphologically damaged mitochondria were more apparent in non-ATAD3A -knockdown mice than in control mice. Previous studies have demonstrated that ATAD3A forms higher-order oligomers and acts as a molecular linker coupling Drp1-mediated mitochondrial fragmentation 20, 21 . The abolition of ATAD3A oligomerization helps delay the progression of neurodegenerative diseases such as Alzheimer’s disease and Huntington’s disease. Therefore, intervening in the ATAD3A oligomerization pathway may be a potential therapeutic target for stroke. To inhibit the ATAD3A oligomerization pathway after stroke, we aimed to accelerate the exhaustion of astrocytic glycogen by blocking the downstream signaling cascade with an HDAC3-specific inhibitor. However, although cotadutide involves in enhancing glycogenolysis and glycogen depletion, dynamic changes in astrocytic glycogen after stroke in vivo cannot be well monitored because of a lack of technical strategies 43 . The mitochondrial biogenesis regulator Pgc-1α is well known downstream of GCG, cotadutide can increase Pgc-1α expression and restore the basal and maximal respiratory rates of damaged mitochondria via the GCGR. HDAC3 has been found to regulate several oxidative stress-related processes and molecules through its deacetylase and nonenzymatic activities. HDAC3 inhibitors are widely used in many inflammatory diseases. Therefore, the ability of cotadutide or RGFP966 to improve mitochondrial function and resolve inflammation may be attributed to multiple pathways. This approach, leveraging clinically developed compounds, holds significant translational potential for stroke treatment. Several limitations warrant consideration. Although we identified key signaling components, other glycogen-mediated pathways involving metabolism, apoptosis, or epigenetics cannot be excluded. The dynamic interplay between glycogen and mitochondria in vivo remains challenging to track due to a lack of real-time glycogen probes. Furthermore, differences between murine ATAD3A and human ATAD3 paralogs necessitate caution in extrapolating findings. Finally, while cotadutide and RGFP966 showed synergistic efficacy, their pleiotropic effects suggest involvement of additional mechanisms beyond the pathway described here. MATERIALS & CORRESPONDENCE Further information and requests for resources, reagents and codes should be directed to and will be fulfilled by the lead contact, Dr. Hongrui Zhu ( [email protected] ). Methods Human samples The human brain tissue was provided by the National Health and Disease Human Brain Tissue Resource Center (Ethical number: S2024052). The detailed information of the human brain samples is presented in Supplementary Table 3. Animals All the animal experiments were approved by the Institutional Animal Care and Use Committee at the University of Science and Technology of China (Ethical number: No. 2022-N(A)-175). The mice were housed under a 12:12-h light‒dark cycle with ad libitum access to food and water. ATAD3A fl/fl (stock no: NM-CKO-2102515) and G6PT -/- mice were obtained from Shanghai Model Organisms Center, Inc. Aldh1l1 -CreERT (stock no: C001288) mice were purchased from Cyagen Biosciences, Inc. Aldh1l1 -CreERT mice were crossed with ATAD3A fl/fl mice to generate Aldh1l1 -CreERT mice. To obtain astrocytic ATAD3A heterozygous knockout mice (ATAD3A Ast Het ), male Aldh1l1 -CreERT ; ATAD3A fl/Δ mice aged six weeks were subjected to five daily intraperitoneal injections of 20 mg/kg tamoxifen (TAM, dissolved in corn oil) to generate ATAD3A Ast Het and control male littermates under corn oil treatment (ATAD3A Ast WT ). Astrocyte-specific secretion protein mice ( Aldh1l1 -CreERT ; ER-BioID HA ) were generated by crossing Aldh1l1 -CreERT mice with C57BL/6-Tg (CAG-EGFP, -birA*)1Fink/J mice (Jackson Laboratory, Stock No: 036203). Aldh1l1 -CreERT ; ER-BioID HA mice aged six weeks were subjected to five daily intraperitoneal injections of 20 mg/kg tamoxifen (TAM, dissolved in corn oil) to generate Ast ER-BioID HA mice. Astrocyte specific diphtheria toxin A (DTA) expression-based mouse strain was established by Aldh1l1-CreERT mice crossed with ROSA26iDTR (Jackson laboratory, Stock No: 007900). ROSA26iDTR; Aldh1l1-CreERT mice aged at six weeks were pretreated with administrated with 20 mg/kg tamoxifen for 5 consecutive days to induce Cre-inducible expression of DTR. 100 ng diphtheria toxin (DT, Aladdin, Cat# D684675-1mg) in 1 × PBS was injected intraperitoneally 3x daily for 10 d to obtain astrocyte specific ablation mouse. Cell lines SVG p12 cells (ATCC, CRL-8621) were cultured in Eagle's minimum essential medium (ATCC, 30–2003) supplemented with 10% fetal bovine serum and 100 U/ml penicillin/streptomycin. HeLa cells (ATCC, CRM-CCL-2) and HEK293 cells (Procell, CL-0005) were cultured in Dulbecco’s modified Eagle’s medium (Servicebio, G4511) with the same supplements. These cell lines were cultured in a humidified incubator with a 5% CO 2 atmosphere at 37°C. Stroke model Focal cortical stroke was induced via the photothrombotic (PTS) method. After anesthesia, mice were fixed in a stereotaxic apparatus (RWD Life Science), and a cranial window was created at the target coordinates relative to the bregma. The mice were then injected intraperitoneally with rose bengal at a dose of 100 mg/kg (Sigma, 330000). After 10 min, the light outlet was placed in direct contact with the skull removal site to transmit light from a cold light source (B5200, POFC-S6H-1000-F1) for 20 min. Following light exposure, the scalp was sutured, and the mice were placed on a warming pad to recover. Focal hippocampal stroke was induced in mice using the photothrombotic method. The experimental procedure was as follows: anesthetized mice were secured in a stereotaxic apparatus, the skull was exposed, and a fiber-optic cannula (400 μm core diameter, MFC_400/430-0.48_2.5 mm_MF2.5_FLT, Doric Lenses) was implanted at coordinates relative to bregma (Anteroposterior: 0 mm, Mediolateral: ± 2 mm, Dorsoventral: - 1.7 mm). Ten minutes after intraperitoneal injection of rose bengal (100 mg/kg), the hippocampal region was illuminated for 10 minutes at 6 mW using a photothrombosis induction laser (561 nm, LRS-0561-GFO-00100-03, LaserGlow Technologies) connected to the implanted fiber. This method relies on the interaction between the photosensitizer and laser light at a specific wavelength to induce focal thrombosis in the hippocampal area adjacent to the fiber tip. Focal ischemia was induced by occlusion of the middle cerebral artery (MCA). After being anesthetized, the mice were placed in a supine position under a microscope, and a longitudinal incision was made along the midline of the neck. The common carotid artery was dissected, with the proximal end ligated and the distal end occluded at the bifurcation of the common carotid artery. The internal carotid artery was clamped via vascular clamps. An incision was made between the proximal end and the clamp, and a silicone-coated filament (Beijing Cinontecech, 1622A4) was introduced. The clamp was released to facilitate filament insertion. After 1 h, the filament was removed, and the incision was sutured. The mice were then placed on a warming pad to facilitate recovery. Parabiosis surgery After being anesthetized, the skin and muscle of the donor mouse are removed to create a muscle flap. This flap is subsequently sutured onto the muscle of the recipient mouse, thereby facilitating the establishment of a shared circulatory system between the two organisms. On the 10th day, methylene blue was injected into the donor mouse, and urine from the recipient mouse was collected 2 h postinjection to analyze the physiological responses influenced by the donor. All animal surgeries were conducted in accordance with humane care policies and were approved. Laser speckle imaging Under anesthesia, laser speckle contrast imaging (RWD, RFSLI-ZW) was employed to monitor cortical blood flow changes before and after the induction of photothrombotic and middle cerebral artery occlusion (MCAO) infarction models in mice. Data analysis was performed via the CUDA Toolkit 10.2 software, following the experimental procedures outlined in the instrument manual. Magnetic resonance imaging and analysis For the magnetic resonance imaging (MRI) experiments, the mice were anesthetized and immobilized, and vital signs were monitored in real time. MRI scans were performed via a 9.4T uMR (United Imaging Life Science Instrument, Wuhan, China) within 14 days poststroke. DTI and T2-weighted images were obtained via the following parameters: (1) DTI: TR/TE = 5728/28 ms; acquisition matrix = 148×192; field of view (FOV) = 15×19 mm; slices = 30; slice thickness = 0.5 mm; number of excitations (NEX) = 4; b-values = 0, 1000, 2000 s/mm² (for the x, y, and z directions); and 5 averages. (2) T2W-MRI: TR/TE = 3000/49 ms; acquisition matrix = 331×368; FOV = 18×20 mm; slices = 20; slice thickness = 0.5 mm; rare factor = 13. Drug administration For the treatment with cotadutide (Aladdin, 1686108-82-6), mice were administered continuous subcutaneous (s.c.) injections of cotadutide at a dosage of 10 nmol/kg/day beginning three days poststroke and continuing until tissue collection. For the HDAC3 inhibitor RGFP966 (Selleck, S7229), the mice received continuous intraperitoneal (i.p.) injections of 10 mg/kg/day starting three days after stroke and lasting until tissue collection. The sham group was given continuous subcutaneous injections of saline. Western blotting The cells were lysed in SDS lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 0.1% SDS, 0.5% sodium deoxycholate, and 1% Triton X-100, pH 7.4) supplemented with a protease and phosphatase inhibitor cocktail (Abcam, ab201119) at 4°C for 10 min. The lysate was centrifuged at 12,000 × g for 15 min at 4°C, and the supernatant was collected and quantified via a BCA assay kit (Biosharp, BL521A). Equal amounts of proteins were then subjected to SDS‒PAGE and transferred to PVDF membranes. After blocking with 5% nonfat milk at room temperature (RT) for 1 h, the blots were probed by overnight incubation with the primary antibody at 4°C. The bound antibodies were detected with the secondary antibody and visualized via enhanced chemiluminescence (ECL) substrates (Biosharp, BL520A) on a Tanon 5200 imaging system. Detailed information about the antibodies is provided in Supplementary Table 4. Co immunoprecipitation The cells or tissue samples were lysed in IP buffer (50 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, 1 mM sodium vanadate, 1 M EGTA, 2.5 mM sodium pyrophosphate, and 1% Triton X-100, pH 7.4) at 4°C for 10 min. After centrifugation, a portion of the supernatant was retained as input, and the remainder was incubated with primary antibody at 4°C for 3 h. Then, protein G agarose beads (Yeasen, 36403ES03) were added and incubated at 4°C for 12 h. After washing with cold IP buffer, the mixture was centrifuged at 12,000 × g for 1 min, and the pellets were boiled in SDS sample buffer and analyzed via western blotting. Plasmids and transfection ATAD3A, Drp1, Epm2aip, GYS1, HDAC1, HDAC2, HDAC3, HDAC6, HDAC8, SIRT1, SIRT3, SIRT4, SIRT5, SIRT6, SIRT7, CLOCK, KAT1, KAT2A, KAT2B, KAT5, and KAT8 were subsequently cloned and inserted into the pCMV-Myc, pEGFP-N1, pCMV-His, and pCMV-Flag vectors. Truncated mutants of ATAD3A were constructed via overlap extension cloning techniques. The ATAD3A mutants K134E and K135Q, as well as the HDAC3 mutants Y298F and H133/H134A, were constructed via the Mut Express II Fast Mutagenesis Kit (Vazyme, C214-01). The tdTomato-tagged wild-type STBD1 constructs were constructed by subcloning the STBD1 fragments into pLV3-CMV-MCS-tdTomato (Miaolingbio, P50400). All plasmid constructs were confirmed through sequencing. Transfections of these plasmids were conducted via polyethyleneimine (Polysciences, 24765-1, 24765-100 (MW 40,000)) following the manufacturer's instructions. Detailed information about the plasmids is provided in Supplementary Table 4. Generation of HDAC3 - or HDAC6 - knockout cell lines and ATAD3A-knockdown cell lines To effectively deplete endogenous HDAC3, HDAC6, and ATAD3A in HEK293 cells, lentiviral vectors (GeneCopoeia, SG002) were used to deliver guide RNAs specifically targeting human HDAC3 (5′-GCGATGTGGGCAACTTCCAC-3′) and human HDAC6 (5′-GCCGGCCAAGATTCTTCTACT-3′). The single guide RNA (sgRNA) sequences were selected via Synthego sgRNA Designer (https://design.synthego.com) and subsequently incorporated into the lentiCRISPR vector through BsmBI-v2 (NEB, R0580) digestion. Knockdown of ATAD3A in SVG p12 cells was achieved by infecting the cells with the pLKO.1 lentivirus (Sigma, SHC001), which carries a guide RNA targeting human ATAD3A (5′-TGGACCATCTCATTGATGCGGT-3′). The lentivirus was produced by cotransfecting the lentiviral expression vector along with packaging and envelope plasmids, including psPAX2 and pMD2, into HEK293 cells via polyethyleneimine. The efficiency of knockout or knockdown was confirmed through Western blot analysis and Sanger sequencing. 2,3,5-Triphenyl tetrazolium chloride staining After the mice were anesthetized, they were perfused with PBS, and the brains were extracted. The brains were then rapidly frozen at -20°C for 20 min before sectioning. Coronal sections were continuously cut at 1 mm intervals and placed in 2,3,5-triphenyl tetrazolium chloride (TTC, Sigma, USA) at 37°C for 30 min. The infarct area was quantified via ImageJ software. Transmission electron microscopy For transmission electron microscopy analysis, samples were fixed in 2.5% glutaraldehyde at 4°C for 24 h, followed by fixation with 2% osmium tetroxide. They were then immersed in 2% uranyl acetate at RT for 2 h and dehydrated in a gradient of ethanol (30% to 100%). After being embedded in resin, ultrathin sections were stained with lead citrate and uranyl acetate. Images were acquired via a transmission electron microscope (Tecnai G2 F20; 200 kV; FEI). Immunostaining and imaging After the mice were anesthetized, they were perfused with PBS followed by 4% PFA. The brains were extracted and fixed in 4% PFA for 6‒8 h and then dehydrated in 20% and 30% sucrose solutions. Coronal sections (40 µm) were cut via a vibratome (Leica CM1950). The sections were air dried at 50°C for 1 h, followed by antigen retrieval in sodium citrate at 60°C. After permeabilization with 0.3% Triton X-100 and blocking with 3% BSA for 40 min, the sections were incubated overnight with primary antibodies at 4°C. Following three washes with PBS, secondary antibodies were applied at RT for 2 h. After another three washes with PBS, the sections were stained with DAPI for 10 min at RT. Images were acquired using a Leica STED confocal microscope. Detailed information about the antibodies is provided in Supplementary Table 4. Three-dimensional (3D) reconstruction The 40-μm- thick brain sections were co-stained with primary antibodies targeting the astrocytic marker GFAP and the ATAD3A protein, followed by fluorescence labeling with corresponding secondary antibodies. Image acquisition was performed using a near-infrared continuous-spectrum single-photon confocal system (Leica STED) microscope and imaging parameters (laser power, gain, and offset) were consistent across all experiments. Z stacking was performed with 1.0- μm steps in the Z direction, and 1024 × 1024-pixel resolution images were analyzed using IMARIS 9.6.2 software (Bitplane). Astrocytes were reconstructed in 3D using the IMARIS “Surface” function applied to GFAP signal, with consistent thresholds applied across all samples. Punctate signals for ATAD3A, GYS1, GBE1, and STBD1 were identified using the IMARIS “Spots” function. Colocalization was quantified by calculating the number of respective fluorescent puncta (ATAD3A; GYS1 and STBD1; GBE1 and STBD1) located within each GFAP-positive astrocytic surface, defined by a distance ≤ 0 μm using the “Split into Surface Objects” function. These counts served as the metrics for colocalization analysis between GFAP and the respective proteins. Immunohistochemistry The mice were anesthetized and perfused with 4% PFA. The tissues were fixed in 4% PFA for 8 h. The brains were embedded in paraffin. The brain tissue was sectioned to a thickness of 10 µm along the coronal plane via a paraffin microtome (Leica BIOCUT). The samples were sequentially processed through the following solutions: xylene, 100% ethanol, 90% ethanol, 80% ethanol, 75% ethanol, and 3% hydrogen peroxide. The samples were then blocked with 3% horse serum at RT for 10 min and incubated for 12 h at 4°C with anti-ATAD3A (Abnova, H00055210-D01). After being washed with TBST, the tissue sections were treated with biotinylated anti-rabbit IgG (Servicebio, G1213) at RT for 40 min. Immunohistochemical localization was performed via a DAB color development kit (Servicebio, G1212) according to the manufacturer's instructions. The images were then examined under a ZEISS Axio Imager M2 microscope. Detailed information about the antibodies is provided in Supplementary Table 4. Multiplex immunofluorescence staining Multiplex immunofluorescence staining of mouse brain frozen sections was conducted using the multiplex fluorescent immunohistochemistry kit (FMS-mIHC004, FCMACS) following the manufacturer's protocol. Briefly, sections were dried at 55°C for 30 min, followed by three washes with PBS. Subsequently, they were incubated in a decoloring and antigen repair buffer (Reagent A, 1:20 in deionized water). The sections were then heated to boiling in a microwave oven and maintained at approximately 95°C for 30 min. After cooling to room temperature, sections were permeabilized with 0.5% Triton X-100 for 20 min and then incubated with Reagent B to block endogenous peroxidase activity for 15 min. Subsequently, the staining was accomplished through four sequential labeling cycles. Each cycle consisted of sequential incubations with a specific primary antibody-dilutions were as follows: ATAD3A (1:500), HDAC3 (1:250), GFAP (1:4000), and STBD1 (1:1000) at 4°C overnight, followed by an HRP-conjugated polymer secondary antibody (Reagent C), the corresponding TSA fluorescent dye (D-488 for ATAD3A, D-647 for HDAC3, D-750 for GFAP, and D-594 for STBD1), and finally a microwave-based HRP inactivation step to enable subsequent cycling. Following the completion of all cycles, sections were counterstained with DAPI-containing mounting medium (Reagent G) and imaged using an Olympus FV4000 laser scanning confocal microscope. Contextual fear conditioning A standard two-day contextual fear conditioning paradigm was employed to assess the long-term recovery of hippocampal memory function one week after stroke. All experimental mice were transferred to the behavioral testing room at least 1 h prior to testing to acclimatize to the environment and minimize stress. On day 7 post-stroke (bilateral hippocampal stroke), mice were placed in a fear conditioning apparatus (Labmaze Conditional Fear Video Analysis System, ZS-KJ) constructed of plexiglass with a metal shock grid floor. The training session was conducted over a 7 min period according to the following protocol: The initial 1 min period allowed mice to freely explore the novel environment without any stimulation. Subsequently, the system automatically delivered three-foot shocks (0.7 mA, 2s duration) at programmed intervals. The software automatically recorded behavioral responses before and after each shock, along with baseline activity levels throughout the session. On day 8 post-stroke (bilateral hippocampal stroke), mice were returned to the identical apparatus used during the previous day’s conditioning session. During this 7 min test session, no foot shocks were administered. The analysis software primarily tracked and recorded freezing behavior, defined as complete immobility except for respiration movements. In situ proximity ligation assay (PLA) Following standard perfusion and fixation, mouse brains were cryoprotected in sucrose gradients and sectioned coronally at 40 μm thickness using a vibratome. For PLA, sections were washed three times after air-drying at 50°C. After permeabilization with 0.3% Triton X-100, sections were blocked with 3% bovine serum albumin (BSA) for 40 min at room temperature and incubated with primary antibodies against ATAD3A, HDAC3, and GFAP overnight at 4°C. PLA probes (PLUS and MINUS) were diluted 1:5 in antibody diluent and pre-incubated for 20 min at room temperature. After washing with Duolink Wash Buffer A (2×5 min), the probe mixture was applied and incubated for 1 h at 37°C. Sections were then incubated with ligation solution containing ligase (1:40 dilution) for 30 min at 37°C, followed by amplification with polymerase (1:80 dilution) for 100 min at 37°C in the dark. Final washes were performed with Duolink Wash Buffer B, and nuclei were counterstained with DAPI. Imaging was conducted using a high-resolution confocal microscope (Leica STED). Periodic acid -Schiff (PAS) staining of glycogen After the mice were anesthetized, they were perfused with PBS followed by 4% PFA. The brains were extracted and fixed in 4% PFA for 8 h. PAS staining (Biosharp, BL1120A) of the brain tissues was performed according to the manufacturer's instructions. For cultured SVG p12 cells, the cells were seeded onto poly-D-lysine-coated slides and fixed in 2% PFA for 15 min at RT. PAS staining of cultured SVG p12 cells was performed following the instructions of a PAS Stain Kit (Solarbio, G1360). The images of brain sections were then examined under a Nikon Eclipse E100 microscope (Japan), while the images of cultured cells were observed using a ZEISS Axio Imager M2 microscope. Millisect system To precisely isolate the glycogen-rich ischemic penumbra and contralateral normal brain tissue in a mouse model of cerebral ischemia-reperfusion injury, automated microdissection was performed using the AVENIO Millisect System (Roche Diagnostics, Indianapolis, IN). Consecutive 20 μm cryosections were prepared on PET membrane slides. A guidance slide was stained with PAS-hematoxylin to identify the ischemic penumbra, characterized by strong PAS positivity with early ischemic alterations, and the contralateral normal tissue. After target regions were marked, consecutive unstained sections were fixed with ice-cold acetone for 2–3 min (or briefly fixed with 70%–100% ethanol) and rinsed with RNase-free water to enhance tissue adhesion. Using the system software, the digital template generated from the guidance slide was aligned with the capture sections for automated dissection. Tissues from each region were separately collected into pre-chilled RNase-free microcentrifuge tubes, immediately supplemented with RNA stabilizer, and stored at −80°C for subsequent RNA analysis. Golgi staining For Golgi staining, mouse brains were collected 16 days poststroke and subsequently immersed in Golgi staining fixative for two weeks. A vibratome was used to cut the brain tissue into 100 µm coronal slices. All procedures adhered to the guidelines provided in the Golgi staining kit (FD NeuroTechnologies, PK401). Images were captured using a PanoBrain slide scanner (Meca Scientific) and analyzed with Panolyzer software. Dendritic lengths were measured via ImageJ. Sholl analysis and measurement of dendritic spine density Neurons in the ischemic penumbra region of the cortex were traced and analyzed. Using a 100 X oil objective on a ZEISS Axioskop2 plus Microscope (Carl Zeiss, Thornwood, NY) with AxioVision Rel.4.7 software, 10 neurons were randomly selected from each ischemic penumbra region. Sholl’s concentric circle method was employed to objectively examine the dendrites of the chosen neurons. Dendrites intersecting each circle were counted to determine the number of dendritic intersections at different radial distances from the neuronal soma to the dendritic tips, in addition to the total dendritic length and branch points. Each value was averaged per mouse, and the mean value for each mouse was taken as n=1. The values per group were averaged and expressed as mean ± SEM. For each mouse, spine density was quantified in 4 randomly selected high-power fields (HPF; 24 fields total per group) using the NeuronJ plugin (v1.4.3) in ImageJ software (Fiji distribution, NIH), followed by spine counting with the Cell Counter plugin. Luxol fast blue staining Myelin loss was evaluated via the Luxol Fast Blue Myelin Stain Kit (Solarbio, G3245). Frozen sections (10 µm) were preheated at 65°C for 30 min and then stained with Luxol fast blue solution at room temperature for 12 h. The sections were washed sequentially with 95% ethanol and tap water until colorless. They were then differentiated in the solution for 15 s, followed by immersion in 70% ethanol for 30 s, and this process was repeated 2–3 times. After dehydration in anhydrous ethanol, the samples were mounted with neutral gum. The analysis was performed via a Thunder Imager (Leica DM6B). Quantitative RT–PCR ( qRT‒PCR ) Total RNA was extracted via TRIzol reagent (GENESAND). The synthesis of cDNA was carried out with Prime Script™ RT Master Mix (TaKaRa). For the qRT‒PCR assays, TaKaRa SYBR qPCR Master Mix and the LightCycler 480II detection system were used. The mRNA expression levels were normalized to those of GAPDH. Detailed information about the parameters is provided in Supplementary Table 4. Measurement of the oxygen consumption rate After the samples were processed, they were transferred to a 96-well XF cell culture microplate at a density of 5 × 10 5 cells per well. The XF sensor cartridges were allowed to hydrate overnight at RT, and the following day, the hydration solution was replaced with Seahorse XF hydration solution. Before analysis, the media in the XF microplate was replaced with Seahorse XF DMEM base assay medium (1 mM glucose, 100 mM pyruvate, and 200 mM L-glutamine; pH 7.4) and incubated at 37°C in a non-CO 2 incubator for 1 h. The four injection ports were then filled with oligomycin A (2 μM), FCCP (2 μM), rotenone (0.5 μM), and antimycin A (0.5 μM). The oxygen consumption rate (OCR) was assessed via an XF96 Analyzer (Seahorse Bioscience), with the results normalized to the number of cells. ATP fluorescent protein biosensor In SVGP12 cells transfected with the ATP fluorescent probe plasmid pCMV-Mito-AT1.03 (Beyotime, D2606) for 48 hours, real-time monitoring of ATP levels and their dynamic changes in single living cells was performed via fluorescence resonance energy transfer (FRET). Real-time images were captured via a laser scanning confocal LSM980 two-photon imaging system. The excitation wavelength was set to 435 nm, with emission wavelengths of 475 nm (mseCFP) and 527 nm (cp173-Venus). The absolute intensity value Fc (RFETcorrected) of the photonic emission in the RFET (cp173-Venus/mseCFP) channel was statistically analyzed. Cylinder task The cylinder Test (Spontaneous Forelimb Task) was conducted to evaluate sensorimotor function. Mice were tested at the following time points: pre-stroke (baseline), and post-stroke days. Each mouse was individually placed in a transparent glass cylinder (15 cm in height, 10 cm in diameter) and allowed to explore freely for 3 min. During this period, the mice spontaneously reared into a standing position, supporting their body weight with either a single forelimb or both forelimbs. Their spontaneous rearing activity was recorded on video for 5 min. The recorded videos were subsequently analyzed at 1/4 of the actual speed to assess the animal’s forelimb preference during exploratory behavior. Only rearing episodes in which both forelimbs were clearly visible were included in the analysis. The duration of support by the left forelimb, right forelimb, or both forelimbs simultaneously against the cylinder wall was quantified for each mouse. The percentage of time using each forelimb was calculated, and an asymmetry index was derived using the following formula: (% ipsilateral forelimb use) - (% contralateral forelimb use). Grid-walking task Each mouse was placed individually on an elevated wire grid (32 × 20 × 50 cm; mesh size 12 × 12 mm) and allowed to walk freely for 5 min while being video-recorded. Videos were analyzed at one-fourth normal speed by an experimenter blinded to treatment groups. A foot fault was defined as either a step through the grid opening without support or a resting position with the wrist below grid level. The percentage of foot faults was calculated as (number of foot faults / total steps) × 100. The ratio of foot faults to total steps was calculated to control for variations in locomotor activity across animals and test sessions. Morris water maze test To evaluate spatial recognition memory in the mice, the Morris water maze test was performed 21 to 27 days after photothrombotic infarction surgery. The maze had a diameter of 120 cm and a water depth of approximately 30 cm, featuring a submerged platform that measured 11 cm × 11 cm. During the initial training phase, the mice were placed in the water facing the wall of the pool from various starting points. The time taken to find the submerged platform was measured in seconds. If a mouse took longer than 60 s to find the platform, it was guided to it and allowed to stay there for 10 s. Each mouse underwent four training sessions daily for five consecutive days prior to surgery. Between days 21 and 26 post surgery, acquisition training was conducted with four sessions daily for five days. On day 27, the platform was removed, and the mice were placed in the water from the opposite quadrant of the platform. The time spent in the former platform quadrant was recorded. Adhesive removal test The adhesive removal test is used to evaluate sensory‒motor function asymmetry in mice following cerebral ischemia. A small piece of tape (2 mm wide and 3 mm long) was placed on the forepaw controlled by the injured cortex. The time taken by the mice to remove the tape was recorded. Rotarod test The rotarod apparatus was used to test the motor coordination of the mice. The mice were placed on a rotating rod, with the speed gradually increasing from 2 revolutions per min to 50 revolutions per min. Each trial lasted for 5 min, with a 30 min interval between trials. The time until the mice fell or lost their balance was recorded. Open field test The experiment was conducted in a white, open field cubic box (50 × 50 × 30 cm). After an acclimation period, the mice were recorded for 6 min, and their total distance traveled was measured using EthoVision XT 14 software (Noldus, Wageningen, The Netherlands) to assess overall motor function. The arena was cleaned with 75% ethanol followed by water after each test. Axon sprouting evaluation To evaluate corticospinal tract remodeling and motor function recovery poststroke, 2 μL of 10% Dextran, Biotin, 10,000 MW, Lysine Fixable (BDA-10,000, Invitrogen™, D1956) was injected at two sites in the contralateral cortex one month after the stroke, with the following stereotaxic coordinates: (1) anteroposterior: 0.6 mm, medial-lateral: 1.2 mm, dorsal-ventral: 0.82 mm; (2) anteroposterior: 0 mm, medial-lateral: 1.8 mm, dorsal-ventral: 0.82 mm. Two weeks later, the animals were perfused with PBS and subsequently with 4% PFA, after which both the brain and spinal cord were harvested. Transverse cervical spinal cord sections and coronal brain sections were sliced to a thickness of 30 μm, and BDA was labeled with Cy™3 streptavidin (Jackson ImmunoResearch, 016-160-084). Thunder Imager (Leica DM6B) was used to detect BDA in the spinal cord and red nucleus, and ImageJ was used to quantify the number of fibers crossing the denervated area from intact circuits. ATAD3A oligomer detection For the cell samples, cell lysates were prepared via RIPA buffer, followed by the addition of 1 mM BMH (Thermo Scientific™, 22330) for incubation at RT for 25 min. Subsequently, 0.1% β-mercaptoethanol (β-ME) was added, and the reaction was terminated by incubation at RT for 10 min. The proteins were then loaded in SDS loading buffer and analyzed via 8% SDS‒PAGE. For tissue samples, 1% paraformaldehyde was used for incubation at RT for 20 min, followed by three washes with 100 mM glycine to terminate the reaction. The tissue was subsequently homogenized in RIPA buffer without reducing agents. After BCA protein quantification, the samples were analyzed via 8% SDS‒PAGE, with or without β-ME. GST-pull down To investigate protein‒protein interactions, a GST pull-down assay was performed. GST- and His-tagged fusion proteins were expressed in E. coli Rosetta (DE3) cells by induction with 0.5mM isopropyl β-D-1-thiogalactopyranoside (IPTG) at 16°C overnight and purified using glutathione-Sepharose beads and Ni-NTA Agarose, respectively. The His-tagged protein was subsequently eluted from the Ni-NTA resin. The two purified proteins were incubated with an equal volume of pull-down binding buffer (50 mM Tris-HCl, pH 8.0, 250 mM NaCl, 1 mM EDTA, 1% NP-40, 10 mM MgCl₂, 0.2 mM PMSF, and 0.2 mM DTT) at 4°C overnight. The samples were then washed three times with TEN buffer (20 mM Tris-HCl, pH 7.4, 0.1 mM EDTA, and 100 mM NaCl). After centrifugation at 1,500 rpm for 1 min at 4°C, the precipitated components were analyzed by immunoblotting using anti-His or anti-GST antibodies. Following electrophoresis, the SDS-PAGE gel was incubated with Coomassie Blue staining solution for 1 min with low-heat assistance. The gel was then destained with destaining solution (20% methanol, 10% acetic acid) on a rocking platform until the background became clear and protein bands were clearly visible. In vitro acetylation assay Purified GST-KAT8 from E. coli Rosetta (DE3) cells was incubated with His-ATAD3A and 50μM acetyl-coenzyme A in HAT assay buffer (50 mM Tris-HCl pH 8.0, 10% glycerol, 0.1 mM EDTA, 1 mM DTT) at 30°C for 3 h. The acetylated His-ATAD3A (His-ATAD3A Ace ) protein was subsequently purified for downstream applications. The His-ATAD3A Ace was incubated with purified GST-HDAC3 or GST-HDAC6 from E. coli in deacetylase assay buffer (10 mM Tris-HCl pH 8.0, 10 mM NaCl) at 30°C for 30 min. Reaction mixtures were then analyzed by western blot. Purified GST-Sirt1 from E. coli Rosetta (DE3) cells was incubated with acetylated His-ATAD3A in deacetylase buffer (10 mM Tris-HCl pH 8.0, 150 mM NaCl, 10% glycerol) supplemented with 1 mM NAD⁺ at 25°C for 2 h. Samples were mixed with equal volumes of loading buffer and analyzed by SDS-PAGE. Split-luc In this study, two potentially interacting proteins were fused to the N-terminal and C-terminal fragments of firefly luciferase. A stable cell line overexpressing one of the fusion proteins (bait protein A) was transiently transfected with the other fusion protein (prey protein B). The interaction between the two proteins in 293T cells was assessed by measuring firefly luciferase activity via a microplate reader. Positive controls included a stable cell line overexpressing firefly luciferase, while negative controls included cell lines expressing either bait fusion protein A alone (stable transfection) or prey fusion protein B alone (transient transfection), as well as a cell line stably expressing the bait fusion protein (C-terminal fusion) followed by transient transfection with the bait fusion protein (N-terminal fusion). Overexpression plasmids were constructed on the pLJM1-MCS-linker-N-fluc and pLJM1-Cf-luc-Linker-MCS vectors: pLJM1-ATAD3A-linker-N-fluc; pLJM1-C-fluc-Linker-HDAC3. The lentivirus was used to infect 293T cells to establish a stable cell line overexpressing the bait protein HDAC3, designated 293T-HDAC3-C-fluc. The cells were then transiently transfected with ATAD3A-N-fluc. HDAC3-C-fluc and ATAD3A-N-fluc 293T cells were seeded in 96-well plates in suspension (nonadherent) at a density of 1×10 6 cells per well. Subsequent imaging was performed using an ID5 multifunctional microplate reader. Cellular mitochondria imaging For live-cell mitochondrial imaging, the cells were cultured in confocal glass dishes (Biosharp, BS-15-GJM). The medium was replaced with fresh serum-free medium prior to staining. Next, the cells were treated with 100 nM MitoTracker Red CMXRos (Thermo Fisher Scientific, M7512) at 37°C for 10 min. Following three washes with PBS, the mitochondrial morphology was imaged via a Leica STED microscope. The acquired images were analyzed via Imaris (Bitplane, Imaris 9.6.2). Mitochondrial morphology was assessed by quantifying the number of fragmented mitochondria relative to the total number of cells, with counts normalized to the total number of cells in each field of view for comparison and statistical analysis. Glycogen pull down The samples were collected in binding buffer containing protease inhibitors (10% glycerol, 1 mM EDTA, 20 mM Tris-HCl, 150 mM NaCl, pH 8.0). Following treatment with an ultrasonic disruptor, the samples were centrifuged at 30,000 rpm for 30 min at 4°C to obtain the supernatant. The supernatants were divided into two groups: one mixed with purified glycogen (MedchemExpress, 9005-79-2) and the other without glycogen and then incubated for 90 min at 4°C. After centrifugation at 100,000 rpm for 90 min at 4°C, the protein precipitate bound to glycogen was collected for mass spectrometry analysis. Both the protein precipitate and the supernatant were analyzed via SDS‒PAGE. TurboID protein purification In SVG p12 human fetal glial cells, after transfection with the ATAD3A-TurboID plasmid, the cells were cultured for 48 h. Biotin (200 µM) was then added to the culture medium, and the cells were incubated at 37°C for an additional 48 h. Next, the cells were washed three times with cold DPBS and digested with 0.5% trypsin-EDTA. Following centrifugation at 800 rpm for 5 min, the cell pellet was resuspended in urea lysis buffer (8 M urea, 10 mM Tris, 100 mM NaH 2 PO 4 , pH 8.5) and subjected to ultrasonic treatment for 3 min (30 W amplitude, 5 s pulse). The mixture was shaken at 4°C for 30 min and then centrifuged at 12,000 rpm for 10 min to collect the supernatant. Biotinylated beads (Thermo Fisher Scientific, 20353) were added (80 µL) and incubated overnight at 4°C. After centrifugation at 12,000 rpm for 2 min at 4°C, the beads containing biotinylated proteins were collected. The samples were then washed sequentially with RIPA buffer, 1 M KCl, and 0.1 M Na 2 CO 3 , followed by a quick rinse with RIPA buffer. Subsequently, mass spectrometry analysis, SDS‒PAGE analysis (4 to 12% polyacrylamide gel), and silver staining were performed. For purification of proteins synthesized specifically by astrocytes in vivo, Ast ER-BioID HA mice were sequentially injected with biotin with 100 μL of a 2-mg/mL biotin solution (Beyotime, Cat# ST2051) once a day intraperitoneally. After RIPA lysis of whole brain tissue, biotinylated proteins were purified according to the above method. Quantitative LC‒MS /MS analysis An appropriate amount of TEAB was added to the samples to adjust the pH to 8.0. Five microlitres of each suspension was subjected to SDS‒PAGE. The remaining samples were subjected to digestion. The protein mixture was reduced with 5 mM dithiothreitol for 30 min at 56°C and alkylated with 11 mM iodoacetamide for 15 min at room temperature in darkness. The samples were then diluted by adding 200 mM TEAB to a urea concentration less than 2 M. Next, trypsin (1:50 trypsin-to-protein mass ratio) was added for the first digestion overnight and 1:100 for a second 4 h digestion. Finally, the peptides were desalted on a Strata X SPE column and separated on an AvianQuick-Nano UPLC system (Thermo Fisher Scientific). The separated peptides were analyzed in an Orbitrap Exploris 480 instrument with a nanoelectrospray ion source. The precursors and fragments were analyzed with an Orbitrap detector. The full MS scan resolution was set to 60000 for a scan range of 350–1400 m/z. The MS/MS scan was fixed first mass at 120.0 m/z at a resolution of 15000. HCD fragmentation was performed at a normalized collision energy of 27%. The automatic gain control target was set at 1E6, with a maximum injection time of 22 ms. The DIA data were processed via the DIA-NN search engine. Tandem mass spectra were searched against Mus_musculus_10090_SP_20231220.fasta concatenated with the reverse decoy database. Trypsin/P was specified as a cleavage enzyme allowing up to 1 missing cleavage. Excision of N-term Met and carbamidomethyl on Cys were specified as fixed modifications. The FDR was adjusted to < 1%. Gene Ontology (GO) analysis, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway annotation, subcellular localization annotation, and protein‒protein interaction network analysis were subsequently performed. Isolation of astrocytes from adult murine brain Isolation of astrocytes from adult murine brain was performed according to the manufacturer's instructions (Miltenyi, Cat#130-107-677). First, enzyme mixtures were prepared in advance by mixing Enzyme Mix 1 (50 μL Enzyme P + 1900 μL Buffer Z) and Enzyme Mix 2 (20 μL Buffer Y + 10 μL Enzyme A) according to the specified ratios. Moreover, the gentleMACS Octo Dissociator was set to program 37C_ABDK_01, and the appropriate C-tubes were prepared. Tissue processing: Following anesthesia, the mice were perfused transcardially with ice-cold DPBS (Gibco, 14287080). The cerebral cortex was subsequently rapidly dissected and minced into small pieces, after which the tissue fragments were transferred into C Tubes containing the enzyme mixture. Immediately thereafter, tissue dissociation was performed via the gentleMACS Octo Dissociator (program: 37C_ABDK_01). Upon completion of the program, the cell suspension was centrifuged (300 ×g, 10 min, 4°C), and then the supernatant was carefully aspirated to collect the cell pellet. Next, the cell pellet was resuspended in ice-cold PBS, followed by the addition of Debris Removal Solution with gentle mixing. Thereafter, ice-cold PBS was carefully added on top of the suspension (avoiding mixing of phases), after which the sample was centrifuged (3000 ×g, 10 min, 4°C). After the upper two phases were removed, the remaining cells were resuspended in ice-cold PBS and centrifuged again (1000 ×g, 10 min, 4°C) to obtain a purified cell suspension. Magnetic Labeling and Separation. For magnetic labeling, the cells were first resuspended in buffer (PBS + 0.5% BSA) at a concentration of 80 μL per 10 7 cells and then incubated with 10 μL of FcR Blocking Reagent for 10 min at 4°C to prevent nonspecific binding. Subsequently, 10 μL Anti-ACSA-2 MicroBeads were added per 10 7 cells, after which the mixture was incubated for 15 min at 4°C. Following labeling, the cells were washed with 1~2 mL buffer (300 ×g, 10 min) and finally resuspended in 500 μL of buffer per 10 7 cells. Positive selection: Prior to separation, the LS columns were equilibrated with 3 mL of buffer. After sample loading, the flow-through containing ACSA-2-negative cells were collected, and the column was washed three times with 3 mL buffer each time. Upon removing the column from the magnetic field, ACSA-2-positive cells were eluted and collected with 5 mL buffer. For the analysis of ATAD3A acetylation and oligomerization, astrocytes from four mice per group were pooled to obtain sufficient protein. scRNA-seq library preparation, sequencing and analysis Freshly dissociated single-cell suspensions were loaded onto a 10X Genome GenCode single-cell instrument to generate GEMs (Gel Bead-In-EMlusion). ScRNA-seq libraries were subsequently prepared following the manufacturer’s protocol for the Chromium Next GEM Single Cell 3’ Regent Kit v3.1. Libraries were then sequenced via the NovaSeq 6000 platform in paired-end sequencing (PE150) mode. For scRNA-seq data processing, raw reads were aligned to the mouse reference genome GRCm39 via Cell Ranger software (v8.0.0) to generate cell × gene matrices. The FASTQ files were processed via the standard workflow of CellRanger to generate gene expression count matrices for each sample. The raw count matrices from all the samples were merged via Scanpy (V1.9.6), followed by quality control. Genes expressed in fewer than 50 cells were excluded, and cells were filtered on the basis of the following criteria: (1) the number of detected genes per cell ranged between 500 and 7000, (2) the mitochondrial RNA content was less than 20%, and (3) total counts per cell did not exceed 50,000. To eliminate potential doublets, Doublet Detection (http://doi.org/10.5281/zenodo.2678041) was applied with the boost_rate set to 0.5 and the voter_thresh set to 0.9. After filtering, 63,283 high-quality cells were retained for downstream analyses. The gene expression matrix of the retained cells was normalized by scaling the total UMI counts per cell to 10,000, followed by log transformation. The top 2,000 highly variable genes were subsequently selected for dimensionality reduction and clustering. Principal component analysis (PCA) was performed, and the first 40 principal components (PCs) were used to construct a UMAP. In the initial round of clustering, major cell types (including Astrocytes, Oligodendrocytes, Microglia, Macrophages, Ependymal Cells, ChP Epithelial Cells, Endothelium, Pericytes/SMCs, VLCMs, Neurons, T cells, and Erythrocytes) were identified via Louvain clustering at a resolution of 0.3 on the basis of canonical markers. A second round of clustering was performed to further resolve astrocyte subpopulations with similar clustering parameters. DEGs of each cell type were calculated via the function rank_genes_groups in Scanpy. The top 300 DEGs were selected for gene pathway enrichment analysis via the Metascape website. For ligand‒receptor analysis of the scRNA-seq data, ligand‒receptor interaction pairs between cell subtypes were inferred via NeuronChat (v1.0.0) 39 . We subsequently compared the cell type interaction counts and identified the changed ligand‒receptor pairs between the WT and KO groups. To recover the cellular dynamics of astrocyte subtypes, we performed RNA velocity analysis. Spliced and unspliced reads were counted via velocyto.py (v0.17.17). Virus injection For stereotaxic intracranial viral injection, mice were anesthetized and secured in a stereotaxic frame (RWD Life Science). Body temperature was maintained using a heating pad, and corneal dehydration was prevented by applying sterile ophthalmic ointment. Following a midline scalp incision, small craniotomies (0.5 mm in diameter) were drilled over the target regions using the following coordinates relative to bregma: for the sensorimotor cortex. Using a glass micropipette connected to a micro-syringe and an injection pump (UMP3, WPI, USA), 100 nL of viral solution was delivered at a rate of 5 nL/s. The following viruses were used to be delivered via stereotaxic injection into the sensorimotor cortex: rAAV-CMV-ATAD3A(K134E)-3XFlag-WPRE-hGH pA (abbreviated as AAV ATAD3A K134E ) and rAAV-CMV-3XFlag-WPRE-hGH pA (control). Experiments were conducted 3 weeks post-injection. For tail vein viral injection to infect astrocytes throughout the brain, a recombinant adeno-associated virus (rAAV) carrying the target gene was mixed with a glial cell-specific Cre recombinase virus (rAAV-GFaABC1D-CRE-4x6TPA) at a 2:1 ratio to prepare a viral mixture with a total volume of 150 μL. For knockdown experiments, the experimental group received a cocktail of rAAV-CMV-DIO-(EGFP-U6)-shRNA (Hdac3)-WPRE-hGH polyA and rAAV-GFaABC1D-CRE-4x6T-PA, while the control group received rAAV-CMV-DIO-(EGFP-U6)-shRNA(scramble)-WPRE-hGH polyA and rAAV-GFaABC1D-CRE-4x6T-PA. For overexpression experiments, the experimental group received a cocktail of rAAV-CMV-DIO-pygb-3XFlag-P2A-EGFP-WPREs and rAAV-GFaABC1D-CRE-4x6T-PA, whereas the control group received rAAV-CMV-DIO-EGFP-WPRE-hGH pA and rAAV-GFaABC1D-CRE-4x6T-PA. Please refer to the attached materials for the specific virus sequence. Whole-cell recordings Pyramidal neuron whole-cell currents were measured via an upright microscope (Olympus X51W) in conjunction with a patch-clamp amplifier (Axon Patch 700B). Recordings were low-pass filtered at 2 kHz and sampled at 10 kHz via a Digidata 1440A. Series resistance was monitored (5–25 MΩ), and only data with changes <20% were included. Data collection and analysis were performed with pClamp 10.3 and Clamfit 10.3. Biocytin (0.2%) was added to the internal solution for neuron labeling, and the slices were subsequently fixed for immunohistochemistry. For the tonic inhibitory current ( I tonic ) and average phase current, the electrode was filled with a solution containing 120 mM CsMeSO 4 (pH 7.25-7.30). To reduce extracellular GABA, 5 mM GABA was added during perfusion. I tonic displacement was recorded as baseline displacement following the administration of 100 µM BMI. The strong current density was calculated by dividing the current amplitude by the membrane capacitance. For action potential recordings, the electrode was filled with a solution containing 70 mM potassium. For miniature excitatory postsynaptic currents (mEPSCs), the electrode contained 132.5 mM Cs-gluconate, with tetrodotoxin and bicuculline added to block GABA receptor currents. The data were analyzed via the Mini Analysis Program 6.0, with up to 100 events selected for cumulative probability analysis. Fiber photometry The mice were fixed in a stereotaxic frame, and the scalp was incised along the midline. A cranial drill was used to expose the motor cortex at the following stereotaxic coordinates: anteroposterior: 0.47 mm, mediolateral: 1.5 mm, and dorsoventral: 0.82 mm (relative to bregma). An AAV-CaMKII-GCaMP6s virus (BrainVTA Co., Ltd. (Wuhan, China)) was injected at a volume of 100 nl at a rate of 5 nl/s. The scalp was sutured after the procedure. One month later, a photothrombotic stroke was induced in the sensory cortex at the following coordinates: anteroposterior: 0 mm, mediolateral: 2 mm (relative to bregma). Simultaneously, a ceramic optical fiber (RWD, R-FOC-BL200C-39NA) was implanted in the motor cortex and secured with bone cement (Super Bond C&B). Fourteen days poststroke, the mice were placed in a large beaker with a diameter of 10 cm, and recordings were conducted via an optical fiber photometry system (Inper). The light intensity at 470 nm was set within the range of 20-40 µW, whereas the intensity at 410 nm was set within the range of 10-20 µW. The data were further processed via signal acquisition software (InperStudio) and analysis software (InperDataProcess). The onset of movement from a standing position (designated as the marker) was used as the basis for trial division. Multiple trials were analyzed and statistically evaluated, resulting in the generation of heatmaps and Delta F/F result graphs. To record hippocampal neuronal activity, we injected AAV-CaMKII-GCaMP6s into the hippocampus (anteroposterior: 0 mm, mediolateral: ± 2 mm, dorsoventral: -1.7 mm relative to bregma) and induced a photothrombotic stroke at the same coordinates three weeks later. Hippocampal GCaMP6s signals were recorded on day 8 post-stroke. The onset of immobility, defined as the behavioral event for analysis, was recorded from the moment the animal was placed in the fear-conditioned context. Electrophysiological recordings Following previous studies, a preparatory surgery was performed at least two days before the in vivo multi-channel recording. General anesthesia was induced (3%) and maintained (1~1.5%) with isoflurane. The mouse was then fixed in a stereotaxic frame (RWD). After applying lidocaine gel, the skin above the target areas was removed. A craniotomy was performed in the contralateral somatosensory cortex relative to the viral injection site, and the surgical opening was sealed with a silicon elastomer upon completion. In the somatosensory cortex which had been injected with AAV virus two weeks earlier, an optical cannula (RWD, 200 μm diameter, 0.5 NA) was implanted and fixed with dental acrylic (C&B Super Bond). To facilitate head fixation during recording, a metal post was affixed to the skull using dental acrylic (Super Bond C&B). Additionally, a Teflon-coated silver wire was implanted in the contralateral hemisphere as the reference electrode. Following recovery from anesthesia, the mouse was returned to its home cage and monitored for the subsequent two days. Before in vivo recordings, mice were habituated to the head-fixation system. Each animal underwent three recording sessions per day, with each session lasting no longer than 30 minutes. Before recording, the silicon seal was removed, and a 64-channel silicon probe (NeuroNexus) was inserted into the somatosensory cortex. Neural signals were recorded at a sampling rate of 40 kHz and bandpass-filtered (0.3–3 kHz) using the OmniPlex Neural Recording Data Acquisition System (Plexon). Single-unit spikes were extracted through semi-automated spike sorting with Offline Sorter (Plexon) and subsequently analyzed using custom MATLAB scripts.To mark the recording site, DiI, a non-toxic lipophilic fluorescent dye, was applied to the electrode which was then reinserted into the recording site after all sessions were completed. Finally, mice were euthanized to confirm the recording location. Compound action potential measurements Composite action potentials (CAPs) are used to assess CC and EC. Mouse brain slices (300 μm) were cut starting from 1.06 mm relative to the bregma. The slices were placed in artificial cerebrospinal fluid (26 mmol/L NaHCO3, 2.5 mmol/L KCl, 10 mmol/L glucose, 1 mmol/L Na 2 H 2 PO 4 , 126 mmol/L NaCl, 2.5 mmol/L CaCl 2 , 1.3 mmol/L MgCl 2 ; pH 7.4, 95% O 2 ) at 34°C for 30 min and then allowed to rest at room temperature for 1 hour. During recording, a bipolar stimulating electrode was positioned laterally to the midline of the corpus callosum. A glass extracellular recording pipette was placed in the EC. The distance between the stimulating and recording electrodes was 0.75 mm. CAP signals are digitized via a digidata 1500B (Molecular Devices, San Jose, CA), amplified, and recorded via an Axoclamp 700B (Molecular Devices, San Jose, CA), with analysis performed via pClamp 10. An input‒output curve is generated at a stimulation intensity of 2 mA. The induced CAPs exhibit a biphasic waveform, with the early peak representing fast-conducting myelinated axons (N 1 ) and the delayed peak indicating slow-conducting unmyelinated axons (N 2 ). The amplitudes of N 1 and N 2 are calculated as the variance from the first and second peaks to the trough. Protein oxidation detection Protein oxidation was determined by protein carbonylation using the OxyBlot™ protein oxidation detection kit (Sigma Aldrich, Catalog No. S7150). Following RIPA lysis, protein samples were denatured by adding an equal volume of 12% SDS. Derivatization was performed by adding 20 μL of 1X DNPH solution per sample, while negative controls received 20 μL of 1X derivatization-control solution. The mixtures were incubated at room temperature for 15 min (not exceeding 30 min), after which 7.5 μL of neutralization solution was added. Samples lacking reducing agents were supplemented with 2-mercaptoethanol to a final concentration of 0.74 M. For western blot analysis, the derivatized and neutralized protein samples were loaded directly onto the gel. Separately, for the molecular weight standard, 2.5 µL of the DNP-labeled protein marker was mixed with 20 µL of 1X gel loading buffer prior to loading. Repeated freeze-thaw cycles of the protein standards were avoided. Proteins were transferred to nitrocellulose membranes which were then incubated in the primary antibody solution (anti-DNP, 1:150) for 2 h, followed by incubation in the secondary antibody solution (1:300) for 1 h at room temperature. The washing procedure was repeated eight times within 40 min. Immunoreactive bands were visualized by enhanced chemiluminescence. The oxidative index was calculated from digital blot images using a two-step normalization. First, the total lane intensity from the Anti-DNP blot was normalized to the corresponding Ponceau S total protein signal. This normalized value for each sample was then expressed relative to the normalized value of the Biotin (+) Ctrl group to establish the final Oxidative Index. Statistical analysis In this study, all the statistical analyses were conducted via GraphPad Prism 8.0 or R software. The data are presented as the means ± SEMs or SDs. The statistical methods followed these principles: comparisons between two groups were performed via a two-tailed unpaired Student’s t test, whereas multiple groups were analyzed via one-way ANOVA or two-way ANOVA analysis. p ≥ 0.05 was considered as no significance (ns); *p< 0.05; **p < 0.01; ***p < 0.005. The processing of specific data, such as electrophysiological and bioinformatics analysis, is detailed in the corresponding methods. Declarations DATA AVAILABILITY All unique/stable reagents generated in this study are available from the lead contact with a completed materials transfer agreement. scRNA sequencing data has been submitted and deposited in GEO database (GEO: GSE295882). The proteomics raw data are currently being uploaded to ProteomeXchange and make them public. Obtaining the accession ID or PRIDE Submission reference: https://www.ebi.ac.uk/pride/archive/projects/PXD064814 (glycogen binding proteomics); https://www.ebi.ac.uk/pride/archive/projects/PXD064694 (ATAD3A binding proteomics data). Microscopy data and any required information reported in this paper is available from the lead contact. And this study doesn’t involve any original code. ACKNOWLEDGMENTS We thank Dr. Tian Xue, Dr. Zhi Zhang and Dr. Qiang Liu for providing constructive suggestions. We would like to thanks to Dr. Ziqiu Zhang, Dr. Xiezong Hu and Dr. Qi Wu for their assistance with the bioinformatics analysis. We would like to extend special thanks to Chunying Yin for her assistance with electron microscopy sample preparation. We extend our sincere gratitude to the National Health and Disease Human Brain Tissue Resource for providing the human brain tissue samples critical to this study. This study was supported by: The National Natural Science Foundation of China (82272225); The National Natural Science Foundation of China (82572514);The National Natural Science Foundation of China (81860249); USTC Research Funds of the Double First-Class Initiative (YD9110002084); The National Natural Science Foundation of China (32471079); The Strategic Priority Research Program of the Chinese Academy of Sciences (Grant No. XDB0940101); Noncommunicable Chronic Diseases-National Science and Technology Major Project (Grant No. 2023ZD0507500); Research Funds of the Center for Advanced Interdisciplinary Science and Biomedicine of IHM (Grant No. QYPY20220005); Natural Science Foundation of Anhui Province (Grant No. 2208085MH241); USTC Research Funds of the Double First-Class Initiative (YD9100002053). AUTHOR CONTRIBUTIONS H.R. Zhu conceived and designed the study. M.M. Yang and H. Zhu wrote the manuscript. M.M. Yang and Z. Li conducted in vivo and in vitro experiments. H.R. Zhu, M.M. Yang, and Z. Li performed data analysis. C.L. Li, S.J. Sun, and X.R. Liu carried out stroke experiments and behavioral recordings. M.M. Yang, L. Wang, Q. Shao, and J.M. Zhang performed immunohistochemistry. S. Wang and K.Q. He revised the manuscript and provided fund support. All authors read, validated the underlying data, and approved the final manuscript. COMPETING INTERESTS The authors declare that they have no competing interests. References Wahl, A.S. et al. Neuronal repair. Asynchronous therapy restores motor control by rewiring of the rat corticospinal tract after stroke. Science 344 , 1250-1255 (2014). Joy, M.T. & Carmichael, S.T. Encouraging an excitable brain state: mechanisms of brain repair in stroke. Nat Rev Neurosci 22 , 38-53 (2021). Giaume, C., Koulakoff, A., Roux, L., Holcman, D. & Rouach, N. Astroglial networks: a step further in neuroglial and gliovascular interactions. Nat Rev Neurosci 11 , 87-99 (2010). Zhou, J. et al. Astrocytic LRP1 enables mitochondria transfer to neurons and mitigates brain ischemic stroke by suppressing ARF1 lactylation. Cell Metab 36 , 2054-2068.e2014 (2024). Cai, Y. et al. Glycogenolysis Is Crucial for Astrocytic Glycogen Accumulation and Brain Damage after Reperfusion in Ischemic Stroke. iScience 23 , 101136 (2020). Sun, R.C. et al. Brain glycogen serves as a critical glucosamine cache required for protein glycosylation. Cell Metab 33 , 1404-1417.e1409 (2021). Sinadinos, C. et al. Neuronal glycogen synthesis contributes to physiological aging. Aging Cell 13 , 935-945 (2014). Gusarov, I. et al. Glycogen controls Caenorhabditis elegans lifespan and resistance to oxidative stress. Nat Commun 8 , 15868 (2017). Nitschke, F., Ahonen, S.J., Nitschke, S., Mitra, S. & Minassian, B.A. Lafora disease - from pathogenesis to treatment strategies. Nat Rev Neurol 14 , 606-617 (2018). Johnson, E.C.B. et al. Large-scale proteomic analysis of Alzheimer's disease brain and cerebrospinal fluid reveals early changes in energy metabolism associated with microglia and astrocyte activation. Nat Med 26 , 769-780 (2020). Ahonen, S. et al. Gys1 antisense therapy rescues neuropathological bases of murine Lafora disease. Brain 144 , 2985-2993 (2021). Prats, C., Graham, T.E. & Shearer, J. The dynamic life of the glycogen granule. J Biol Chem 293 , 7089-7098 (2018). Liu, Q. et al. Glycogen accumulation and phase separation drives liver tumor initiation. Cell 184 , 5559-5576.e5519 (2021). Escartin, C. et al. Reactive astrocyte nomenclature, definitions, and future directions. Nat Neurosci 24 , 312-325 (2021). Guttenplan, K.A. et al. Neurotoxic Reactive Astrocytes Drive Neuronal Death after Retinal Injury. Cell Rep 31 , 107776 (2020). Cameron, E.G. et al. A molecular switch for neuroprotective astrocyte reactivity. Nature 626 , 574-582 (2024). Liddelow, S.A. et al. Neurotoxic reactive astrocytes are induced by activated microglia. Nature 541 , 481-487 (2017). Pagliarini, D.J. et al. A mitochondrial protein compendium elucidates complex I disease biology. Cell 134 , 112-123 (2008). Brar, K.K. et al. PERK-ATAD3A interaction provides a subcellular safe haven for protein synthesis during ER stress. Science 385 , eadp7114 (2024). Zhao, Y. et al. ATAD3A oligomerization causes neurodegeneration by coupling mitochondrial fragmentation and bioenergetics defects. Nat Commun 10 , 1371 (2019). Zhao, Y. et al. ATAD3A oligomerization promotes neuropathology and cognitive deficits in Alzheimer's disease models. Nat Commun 13 , 1121 (2022). Stapleton, D. et al. Analysis of hepatic glycogen-associated proteins. Proteomics 10 , 2320-2329 (2010). Pánek, T., Eliáš, M., Vancová, M., Lukeš, J. & Hashimi, H. Returning to the Fold for Lessons in Mitochondrial Crista Diversity and Evolution. Curr Biol 30 , R575-r588 (2020). Gilquin, B. et al. The AAA+ ATPase ATAD3A controls mitochondrial dynamics at the interface of the inner and outer membranes. Mol Cell Biol 30 , 1984-1996 (2010). Chi, Z. et al. Histone Deacetylase 3 Couples Mitochondria to Drive IL-1β-Dependent Inflammation by Configuring Fatty Acid Oxidation. Mol Cell 80 , 43-58.e47 (2020). Chatterjee, A. et al. MOF Acetyl Transferase Regulates Transcription and Respiration in Mitochondria. Cell 167 , 722-738.e723 (2016). Hu, D. et al. TMEM135 links peroxisomes to the regulation of brown fat mitochondrial fission and energy homeostasis. Nat Commun 14 , 6099 (2023). Ishihara, T., Ban-Ishihara, R., Ota, A. & Ishihara, N. Mitochondrial nucleoid trafficking regulated by the inner-membrane AAA-ATPase ATAD3A modulates respiratory complex formation. Proc Natl Acad Sci U S A 119 , e2210730119 (2022). Hayakawa, K. et al. Corrigendum: Transfer of mitochondria from astrocytes to neurons after stroke. Nature 539 , 123 (2016). Murphy, T.H. & Corbett, D. Plasticity during stroke recovery: from synapse to behaviour. Nat Rev Neurosci 10 , 861-872 (2009). Joy, M.T. et al. CCR5 Is a Therapeutic Target for Recovery after Stroke and Traumatic Brain Injury. Cell 176 , 1143-1157.e1113 (2019). Rawji, K.S., Gonzalez Martinez, G.A., Sharma, A. & Franklin, R.J.M. The Role of Astrocytes in Remyelination. Trends Neurosci 43 , 596-607 (2020). Carmichael, S.T. & Chesselet, M.F. Synchronous neuronal activity is a signal for axonal sprouting after cortical lesions in the adult. J Neurosci 22 , 6062-6070 (2002). Nguyen, H., Zerimech, S. & Baltan, S. Astrocyte Mitochondria in White-Matter Injury. Neurochem Res 46 , 2696-2714 (2021). Zeisel, A. et al. Molecular Architecture of the Mouse Nervous System. Cell 174 , 999-1014.e1022 (2018). Joshi, A.U. et al. Fragmented mitochondria released from microglia trigger A1 astrocytic response and propagate inflammatory neurodegeneration. Nat Neurosci 22 , 1635-1648 (2019). Rao, J. et al. A novel role for astrocytic fragmented mitochondria in regulating morphine addiction. Brain Behav Immun 113 , 328-339 (2023). Yun, S.P. et al. Block of A1 astrocyte conversion by microglia is neuroprotective in models of Parkinson's disease. Nat Med 24 , 931-938 (2018). Zhao, W., Johnston, K.G., Ren, H., Xu, X. & Nie, Q. Inferring neuron-neuron communications from single-cell transcriptomics through NeuronChat. Nat Commun 14 , 1128 (2023). Spezani, R., Marinho, T.S., Reis, T.S., Aguila, M.B. & Mandarim-de-Lacerda, C.A. Cotadutide (GLP-1/Glucagon dual receptor agonist) modulates hypothalamic orexigenic and anorexigenic neuropeptides in obese mice. Peptides 173 , 171138 (2024). Parker, V.E.R. et al. Cotadutide promotes glycogenolysis in people with overweight or obesity diagnosed with type 2 diabetes. Nat Metab 5 , 2086-2093 (2023). Ambery, P. et al. MEDI0382, a GLP-1 and glucagon receptor dual agonist, in obese or overweight patients with type 2 diabetes: a randomised, controlled, double-blind, ascending dose and phase 2a study. Lancet 391 , 2607-2618 (2018). Boland, M.L. et al. Resolution of NASH and hepatic fibrosis by the GLP-1R/GcgR dual-agonist Cotadutide via modulating mitochondrial function and lipogenesis. Nat Metab 2 , 413-431 (2020). Lin, Y.H. et al. Ketone bodies promote stroke recovery via GAT-1-dependent cortical network remodeling. Cell Rep 42 , 112294 (2023). Di Pino, G. et al. Modulation of brain plasticity in stroke: a novel model for neurorehabilitation. Nat Rev Neurol 10 , 597-608 (2014). Mostany, R. et al. Local hemodynamics dictate long-term dendritic plasticity in peri-infarct cortex. J Neurosci 30 , 14116-14126 (2010). Li, S. et al. GDF10 is a signal for axonal sprouting and functional recovery after stroke. Nat Neurosci 18 , 1737-1745 (2015). Reitmeir, R. et al. Post-acute delivery of erythropoietin induces stroke recovery by promoting perilesional tissue remodelling and contralesional pyramidal tract plasticity. Brain 134 , 84-99 (2011). Molyneaux, B.J., Arlotta, P., Menezes, J.R. & Macklis, J.D. Neuronal subtype specification in the cerebral cortex. Nat Rev Neurosci 8 , 427-437 (2007). Greig, L.C., Woodworth, M.B., Galazo, M.J., Padmanabhan, H. & Macklis, J.D. Molecular logic of neocortical projection neuron specification, development and diversity. Nat Rev Neurosci 14 , 755-769 (2013). Kokinovic, B. & Medini, P. Loss of GABA(B) -mediated interhemispheric synaptic inhibition in stroke periphery. J Physiol 596 , 1949-1964 (2018). Tang, X. et al. Pyramidal and parvalbumin neurons modulate the process of electroacupuncture stimulation for stroke rehabilitation. iScience 27 , 109695 (2024). Duran, J., Saez, I., Gruart, A., Guinovart, J.J. & Delgado-García, J.M. Impairment in long-term memory formation and learning-dependent synaptic plasticity in mice lacking glycogen synthase in the brain. J Cereb Blood Flow Metab 33 , 550-556 (2013). Demetriadou, A. et al. Mouse Stbd1 is N-myristoylated and affects ER-mitochondria association and mitochondrial morphology. J Cell Sci 130 , 903-915 (2017). Hossain, M.I., Roulston, C.L. & Stapleton, D.I. Molecular basis of impaired glycogen metabolism during ischemic stroke and hypoxia. PLoS One 9 , e97570 (2014). Zhang, L.Y. et al. The Role of Astrocytic Mitochondria in the Pathogenesis of Brain Ischemia. Mol Neurobiol 61 , 2270-2282 (2024). Farah, B.L. et al. Hepatic mitochondrial dysfunction is a feature of Glycogen Storage Disease Type Ia (GSDIa). Sci Rep 7 , 44408 (2017). Mishra, K. & Kakhlon, O. Mitochondrial Dysfunction in Glycogen Storage Disorders (GSDs). Biomolecules 14 (2024). Jiang, S. et al. Starch binding domain-containing protein 1/genethonin 1 is a novel participant in glycogen metabolism. J Biol Chem 285 , 34960-34971 (2010). Demetriadou, A., Drousiotou, A. & Petrou, P.P. The “sweet” side of ER-mitochondria contact sites. Communicative & Integrative Biology 10 , e1329787 (2017). Gomes, L.C., Di Benedetto, G. & Scorrano, L. During autophagy mitochondria elongate, are spared from degradation and sustain cell viability. Nat Cell Biol 13 , 589-598 (2011). Additional Declarations There is no duality of interest Supplementary Files RS.pdf Reporting Summary SourceData.xlsx Source Data Authorliststatement.pdf Author list statement supData.zip supData ExtendedFigures.pdf Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6650856","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":584333821,"identity":"33bcd29e-7a6d-4e07-80e3-3dcd42d113cb","order_by":0,"name":"Hongrui Zhu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABEklEQVRIiWNgGAWjYDACCQY2CIOZ+QADg4EFTJyZGC1sCUAtEjDVxGhh4DEAcwlqkZ/dY/bg447afIPjPF83/CiQYDA4f/7gB4YK68QG9rMHsGlhnHPG3HDmmeOWGw7zbrvZA3SYwY1kZgmGM+mJDTx5Cdi0MEvkmEnzth0zMABqucED1sLMxsDYdjixQQLsVAzAhtDC8+zmH5CW84eBWv7h1sID0VID0sJ2G2zLgWSglgbcWiQk0sokZ7YdMJA8zGZ2W8ZAgkfyRrKxRMKxdOM2nhysWuRnJG+T+NhWZ8B3/vCzm2/+2MjxnT/48MOHGmvZfvYzWLVAwWGES8EkKKjYsCuFgTr80qNgFIyCUTCyAQC2a1dSe2UisQAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0001-6823-6827","institution":"University of Science and Technology of China","correspondingAuthor":true,"prefix":"","firstName":"Hongrui","middleName":"","lastName":"Zhu","suffix":""},{"id":584333822,"identity":"0312289e-0a68-470b-97bd-8ccde02e37f8","order_by":1,"name":"Mengmeng Yang","email":"","orcid":"","institution":"University of Science and Technology of China","correspondingAuthor":false,"prefix":"","firstName":"Mengmeng","middleName":"","lastName":"Yang","suffix":""},{"id":584333823,"identity":"eb71798a-19c4-46bd-a4a7-bc94f5d8d4c8","order_by":2,"name":"Zhong Li","email":"","orcid":"","institution":"University of Science and Technology of China","correspondingAuthor":false,"prefix":"","firstName":"Zhong","middleName":"","lastName":"Li","suffix":""},{"id":584333824,"identity":"a935ec8a-0b55-4f1b-b7d9-11c25fa7ec3b","order_by":3,"name":"Chunliu Li","email":"","orcid":"","institution":"Department of Anesthesiology and Pain Medicine, The First Affiliated Hospital of USTC；Key Laboratory of Precision and Intelligent Chemistry, Department of Chemistry, USTC","correspondingAuthor":false,"prefix":"","firstName":"Chunliu","middleName":"","lastName":"Li","suffix":""},{"id":584333825,"identity":"d76d1690-67e8-49e1-8a65-8e6753541c4b","order_by":4,"name":"Shuaijie Sun","email":"","orcid":"","institution":"University of Science and Technology of China","correspondingAuthor":false,"prefix":"","firstName":"Shuaijie","middleName":"","lastName":"Sun","suffix":""},{"id":584333826,"identity":"ca78098c-ce83-4a6e-978c-6c54817cea72","order_by":5,"name":"Xiangru Liu","email":"","orcid":"","institution":"University of Science and Technology of China","correspondingAuthor":false,"prefix":"","firstName":"Xiangru","middleName":"","lastName":"Liu","suffix":""},{"id":584333827,"identity":"459a84a7-1d23-4ec6-8bcc-6708cafe4f58","order_by":6,"name":"Li Wang","email":"","orcid":"","institution":"University of Science and Technology of China","correspondingAuthor":false,"prefix":"","firstName":"Li","middleName":"","lastName":"Wang","suffix":""},{"id":584333828,"identity":"2322596b-d272-40e2-9e36-af1d101672db","order_by":7,"name":"Jiaming Zhang","email":"","orcid":"","institution":"University of Science and Technology of China","correspondingAuthor":false,"prefix":"","firstName":"Jiaming","middleName":"","lastName":"Zhang","suffix":""},{"id":584333829,"identity":"382dc3e5-b146-4b0c-8091-096512ef15cc","order_by":8,"name":"Qing Shao","email":"","orcid":"","institution":"The First Affiliated Hospital of USTC","correspondingAuthor":false,"prefix":"","firstName":"Qing","middleName":"","lastName":"Shao","suffix":""},{"id":584333830,"identity":"569e94ea-345e-4759-a44b-c817751a745c","order_by":9,"name":"Keqiang He","email":"","orcid":"","institution":"The First Affiliated Hospital of USTC","correspondingAuthor":false,"prefix":"","firstName":"Keqiang","middleName":"","lastName":"He","suffix":""},{"id":584333831,"identity":"b7eaf1b4-6729-4efd-b528-0cdd6b26ed7a","order_by":10,"name":"Sheng Wang","email":"","orcid":"","institution":"University of Science and Technology of China","correspondingAuthor":false,"prefix":"","firstName":"Sheng","middleName":"","lastName":"Wang","suffix":""}],"badges":[],"createdAt":"2025-05-13 03:15:33","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6650856/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6650856/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":101733833,"identity":"a6b49680-1f66-471d-8776-4310b062a836","added_by":"auto","created_at":"2026-02-03 06:50:49","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":2392837,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGlycogen stress granule mainly increases in astrocytes from the ischemic penumbra and is closely related to ATAD3A oligomerization after stroke. a.\u003c/strong\u003e Periodic acid-Schiff (PAS)\u0026nbsp;staining of the contralateral hemisphere (01, 02), ipsilateral penumbra (03), and ischemic core (04) at 36 h after reperfusion in the MCAO model. Black arrows indicate glycogen aggregates. Scale bars, 1 mm (overview) and 20 μm (insets). \u003cstrong\u003eb.\u003c/strong\u003e Quantification of glycogen-positive spots per high-power field (HPF) (n = 5).\u0026nbsp; \u003cstrong\u003ec. \u003c/strong\u003eThe RNA expression of genes related to glycogenesis and glycogenolysis was measured in dissected regions (n = 3). \u003cstrong\u003ed.\u003c/strong\u003e Schematic of the glycogen pull-down mass spectrometry approach for identifying the glycogen-associated proteins after ischemic stroke. \u003cstrong\u003ee. \u003c/strong\u003eThe overlap between glycogen-enriched proteins and the mitochondrial functional gene set (Mitoproteome database). \u003cstrong\u003ef.\u003c/strong\u003e Immunoblot analysis of ATAD3A in the supernatant (S) and the protease K-pretreated (100 μg/mL) glycogen pellet (P) isolated from the ischemic penumbra in MCAO mice. \u003cstrong\u003eg.\u003c/strong\u003e Schematic workflow of the glycogen-protein co-incubation assay. Tissue lysate was incubated with glycogen, followed by proteinase K treatment (100 μg/mL), ATAD3A-His co-incubation, glycogen pull-down, and bioassay detection (Lanes 1-6).\u003cstrong\u003e h.\u003c/strong\u003e PAS staining and ATAD3A immunohistochemistry (IHC) on serial sections of the ischemic penumbra. Scale bar, 100 μm. Boxed regions (01, 02 in PAS; 03, 04 in IHC) highlight detailed morphology. Scale bar, 20 µm. \u003cstrong\u003ei.\u003c/strong\u003e The scatterplot shows the quantification of the correlation between the glycogen spot signals and the ATAD3A spot signals. \u003cstrong\u003ej.\u003c/strong\u003e Immunoblot analysis shows the oligomerization states of ATAD3A in different groups at the indicated time points. \u003cstrong\u003ek. \u003c/strong\u003eImmunofluorescence images of ATAD3A (green), GFAP (gray, astrocytes), and DAPI (blue) in Sham and MCAO mice. 3D rendering (01, 02) of ATAD3A\u003csup\u003e+\u003c/sup\u003e puncta in GFAP\u003csup\u003e+\u003c/sup\u003e astrocytes. Scale bar, 40 µm. \u003cstrong\u003el.\u003c/strong\u003e Quantification of the ratio of ATAD3A\u003csup\u003e+\u003c/sup\u003e puncta to GFAP\u003csup\u003e+\u003c/sup\u003e area per cell (n = 6). \u003cstrong\u003em.\u003c/strong\u003e Representative immunofluorescence of the ischemic penumbra showing GFAP (gray), ATAD3A (green), STBD1 (red, glycogen aggregates), and DAPI (blue). Magnified views (01, 02) of ATAD3A in glycogen-accumulating and non-accumulating astrocytes. Scale bar, 50 µm. \u003cstrong\u003en.\u003c/strong\u003e Quantification of the ratio of ATAD3A\u003csup\u003e+\u003c/sup\u003e puncta to GFAP\u003csup\u003e+\u003c/sup\u003e area (n = 10). \u003cstrong\u003eo. \u003c/strong\u003eImmunoblot analysis of ATAD3A oligomerization states in brain tissue from control (Ast\u003csup\u003ewt\u003c/sup\u003e) and astrocyte-specific KO (Ast\u003csup\u003edepletion\u003c/sup\u003e) mice with or without ischemic stroke. All data are mean ± SEM. Two-tailed t test (\u003cstrong\u003eb\u003c/strong\u003e, \u003cstrong\u003el\u003c/strong\u003e and \u003cstrong\u003en\u003c/strong\u003e), **\u003cem\u003ep\u003c/em\u003e \u0026lt; 0.01; ***\u003cem\u003ep\u003c/em\u003e \u0026lt; 0.005.\u003c/p\u003e","description":"","filename":"Fig1.png","url":"https://assets-eu.researchsquare.com/files/rs-6650856/v1/f922e1856a296a366fc5d2ca.png"},{"id":101733836,"identity":"f1050797-eadf-4907-a79f-cf73124a445b","added_by":"auto","created_at":"2026-02-03 06:50:49","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":1164022,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eATAD3A signaling abolition mitigates glycogen aggregation trigged mitochondrial dysfunction. a. \u003c/strong\u003eRepresentative transmission electron microscopy (TEM) images of brain glycogen aggregates and mitochondria in glycogen-accumulating tissue. Arrows indicate glycogen aggregates; blue boxes denote few glycogen aggregate-enriched areas, while red boxes denote glycogen aggregate-enriched areas. Scale bar, 1 µm. \u003cstrong\u003eb.\u003c/strong\u003e Quantification of mitochondrial area (µm²) between glycogen aggregate-enriched areas and few glycogen aggregate-enriched areas in glycogen-accumulating tissue (n = 28 mitochondria per HPFs from 5 mice). \u003cstrong\u003ec.\u003c/strong\u003e Representative western blot analysis of ATAD3A oligomerization in SVG p12\u003csup\u003eGYS1 OE\u003c/sup\u003e cells with or without Epm2aip knockdown. \u003cstrong\u003ed.\u003c/strong\u003e Leica super-resolution images showing the mitochondrial morphology in the indicated SVG p12\u003csup\u003eGYS1 OE\u003c/sup\u003e cells with or without Epm2aip knockdown.Scale bars, 20 μm. \u003cstrong\u003ee.\u003c/strong\u003e Quantification of fragmented mitochondria cell/ total cells (Ref. Ctrl) in different groups from (\u003cstrong\u003ed\u003c/strong\u003e) (n = 5 independent experiments). \u003cstrong\u003ef.\u003c/strong\u003e Representative TEM images of SVG p12\u003csup\u003eGYS1 OE\u003c/sup\u003e cells in indicated groups. Red arrowheads denote glycogen stress granules. The white boxes denote the regions magnified below. Scale bars, 2 µm (overview) and 0.5 µm (zoom). \u003cstrong\u003eg.\u003c/strong\u003e Quantification of the number of cristae / mitochondrial area (µm²) in different groups from (\u003cstrong\u003ef\u003c/strong\u003e) (n = 50). \u003cstrong\u003eh.\u003c/strong\u003e SVG p12\u003csup\u003eGYS1 OE\u003c/sup\u003e cells transfected with an ATP biosensor were used to perform Venus/CFP fluorescence ratio detection upon addition of 4 mM ADP. \u003cstrong\u003ei. \u003c/strong\u003eOxygen consumption rate (OCR) in SVG p12 cells and SVG p12\u003csup\u003eGYS1 OE\u003c/sup\u003e cells, with or without ATAD3A knockdown, was\u003cstrong\u003e \u003c/strong\u003edetermined using a Seahorse analyzer following treatment with oligomycin (2 µM), FCCP (2 µM), and rotenone (0.5 µM). \u003cstrong\u003ej.\u003c/strong\u003e The histogram shows the maximal OCR (pmol/min) across different groups (n = 4 independent experiments). All data are mean ± SEM. Two-tailed t test (\u003cstrong\u003eb\u003c/strong\u003e); one-way ANOVA with Bonferroni post hoc test (\u003cstrong\u003ee\u003c/strong\u003e, \u003cstrong\u003eg\u003c/strong\u003e and \u003cstrong\u003ej\u003c/strong\u003e); *\u003cem\u003ep\u003c/em\u003e\u0026lt; 0.05; **\u003cem\u003ep\u003c/em\u003e \u0026lt; 0.01; ***\u003cem\u003ep\u003c/em\u003e \u003cem\u003e\u0026lt; \u003c/em\u003e0.005.\u003c/p\u003e","description":"","filename":"Fig3.png","url":"https://assets-eu.researchsquare.com/files/rs-6650856/v1/f3ebde84f1fe7d44e1f66ada.png"},{"id":101733834,"identity":"91ea676e-4233-4c4a-894e-8e37eebc0e0e","added_by":"auto","created_at":"2026-02-03 06:50:49","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":1436724,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eATAD3A oligomerization is dependent on the deacetylation of HDAC3. a.\u003c/strong\u003e Schematic of proximity-dependent biotinylation using an ATAD3A-TurboID fusion construct. \u003cstrong\u003eb.\u003c/strong\u003e Biotinylation pull-down assay of the ATAD3A-TurboID fusion protein in SVG p12 cells under different conditions. \u003cstrong\u003ec.\u003c/strong\u003e Proteomic analysis of ATAD3A-TurboID-enriched proteins. \u003cstrong\u003ed.\u003c/strong\u003e SVG p12 cells were transfected with ATAD3A-Myc and treated with or without NAM (5 mM, 12 h) or TSA (1 mM, 12 h). \u003cstrong\u003ee.\u003c/strong\u003e Schematic of the split-luciferase complementation assay for detecting ATAD3A-HDAC3 interaction. \u003cstrong\u003ef.\u003c/strong\u003e Quantification of relative luminescence units (RLU) from split-luciferase assays analyzing ATAD3A-Nluc and HDAC3-Cluc interaction in HEK293 cells under indicated treatments. \u003cstrong\u003eg.\u003c/strong\u003e Domain mapping of HDAC3 and ATAD3A identifies interaction interfaces. \u003cstrong\u003eh.\u003c/strong\u003e HEK293 cells transfected with ATAD3A-Myc and different forms of HDAC3-Flag (WT, H133/134A, Y298F), with or without TSA (1μM, 12h). \u003cstrong\u003ei.\u003c/strong\u003e Co‑immunoprecipitation (Co‑IP) assays in HEK293 cells transfected with ATAD3A‑Myc and HDAC3‑Flag (WT or Y298F). \u003cstrong\u003ej.\u003c/strong\u003e Immunoblot analysis of ATAD3A oligomers in HEK293 cells co-transfected with HDAC3\u003csup\u003eWT\u003c/sup\u003e-Flag and Myc-tagged ATAD3A (WT or the K134 mutant forms). \u003cstrong\u003ek. \u003c/strong\u003eIn vitro acetylation assay of His‑ATAD3A by GST‑Kat8 in the presence of acetyl‑CoA (50 µM, 3 h). Coomassie brilliant blue (CBB) staining is shown (right). \u003cstrong\u003el.\u003c/strong\u003e In vitro deacetylation assay of pre‑acetylated His‑ATAD3A (His-ATAD3A\u003csup\u003eace\u003c/sup\u003e) by GST‑HDAC3, GST‑HDAC3\u003csup\u003eY298F\u003c/sup\u003e, or GST‑HDAC6. CBB staining shown at right. \u003cstrong\u003em. \u003c/strong\u003eIn vitro deacetylation assay of pre‑acetylated His‑ATAD3A (His-ATAD3Aace) by GST‑SIRT1 in the presence or absence of NAD+ (1 mM, 2 h). CBB staining shown at right. \u003cstrong\u003en. \u003c/strong\u003eProximity ligation assay (PLA) of GFAP\u003csup\u003e+\u003c/sup\u003e astrocytes in brain sections from Sham, MCAO, and astrocyte-specific HDAC3 knockdown (Ast\u003csup\u003eshHDAC3\u003c/sup\u003e) mice, co-stained for GFAP (green), ATAD3A-HDAC3 interaction (PLA, purple), and DAPI (blue). Scale bar, 20 µm. \u003cstrong\u003eo. \u003c/strong\u003eQuantification of HDAC3-ATAD3A interaction spots per GFAP\u003csup\u003e+\u003c/sup\u003e area in the indicated groups (n = 6). \u003cstrong\u003ep. \u003c/strong\u003eImmunoblot analysis of ATAD3A oligomerization in primary astrocytes (pooled from n=4 mice per group) from Sham and MCAO mice with or without HDAC3 knockdown. \u003cstrong\u003eq. \u003c/strong\u003eCo‑IP of primary astrocytes from Sham and MCAO mice with or without astrocytic HDAC3 knockdown (pooled from n=4 mice per group). All data are mean ± SEM. One-way ANOVA with Bonferroni post hoc test (\u003cstrong\u003eo\u003c/strong\u003e); **\u003cem\u003ep\u003c/em\u003e \u0026lt; 0.01.\u003c/p\u003e","description":"","filename":"Fig5.png","url":"https://assets-eu.researchsquare.com/files/rs-6650856/v1/c98bf893655873a0dc6cfb1e.png"},{"id":101753816,"identity":"dba8a5f2-ac49-481f-a848-567d916b5ca3","added_by":"auto","created_at":"2026-02-03 10:40:56","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":577316,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eHDAC3 regulates ATAD3A-mediated astrocytic mitochondrial dysfunction. a.\u003c/strong\u003e Immunoprecipitation of acetylated proteins from cortical lysates isolated from ischemic penumbra at different time points. \u003cstrong\u003eb.\u003c/strong\u003e Endogenous ATAD3A Co-IP assay assessing ATAD3A acetylation and mitochondrial function proteins. \u003cstrong\u003ec.\u003c/strong\u003e HEK293 cells were transfected with the HDAC3-Flag and/or ATAD3A-GFP and/or ATAD3A-Myc for Co-IP detection. \u003cstrong\u003ed. \u003c/strong\u003eHeLa cells were transfected with the indicated Myc-tagged ATAD3A mutants and/or HDAC3-Flag for ATAD3A oligomers detection. \u003cstrong\u003ee.\u003c/strong\u003e HeLa cells were transfected with the HDAC3-Flag and/or ATAD3A-Myc and/or GFP-tagged ATAD3A 200 aa mutant for 48 h. Proteins were then extracted for Co-IP. \u003cstrong\u003ef. \u003c/strong\u003eHeLa cells were transfected with ATAD3A-GFP and/or HDAC3-Flag and/or Drp1-HA for 48 h. Proteins were extracted for Co-IP. \u003cstrong\u003eg. \u003c/strong\u003eOxygen consumption rate (OCR) in SVG p12\u003csup\u003e\u003cem\u003eATAD3 KD\u003c/em\u003e\u003c/sup\u003e cells was determined using a Seahorse analyzer with treatments of oligomycin (2 µM), FCCP (2 µM), and rotenone (0.5 µM) in order. \u003cstrong\u003eh.\u003c/strong\u003e The histogram shows the maximal and basal OCR (pmol/min) from different groups (n = 4). \u003cstrong\u003ei.\u003c/strong\u003e SVG p12\u003csup\u003e\u003cem\u003eATAD3 KD\u003c/em\u003e\u003c/sup\u003e cells were transfected with ATAD3A-Myc and/or HDAC3-Flag and/or ATAD3A\u003csup\u003eK134Q\u003c/sup\u003e-Myc for 48 h and with or without RGFP966 treatment (10 µM). Leica super-resolution images showing the morphology of mitochondria in different groups. Scale bars, 20 µm. \u003cstrong\u003ej.\u003c/strong\u003e Quantification of relative fragmented mitochondria per cells (Ref. Ctrl) in different groups from (\u003cstrong\u003ei\u003c/strong\u003e) (n = 10). All data are mean ± SEM. One-way ANOVA with Bonferroni post hoc test (\u003cstrong\u003eh\u003c/strong\u003e and \u003cstrong\u003ej\u003c/strong\u003e); *\u003cem\u003ep\u003c/em\u003e\u0026lt; 0.05; **\u003cem\u003ep\u003c/em\u003e \u0026lt; 0.01; ***\u003cem\u003ep\u003c/em\u003e \u0026lt; 0.005.\u003c/p\u003e","description":"","filename":"Fig7.png","url":"https://assets-eu.researchsquare.com/files/rs-6650856/v1/59e98e5c191d0008c70e166b.png"},{"id":101733842,"identity":"f39d6ae8-86f3-48ae-bb9b-dcb023750951","added_by":"auto","created_at":"2026-02-03 06:50:50","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":710316,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAstrocytic ATAD3A oligomerization signaling abrogation alleviates damage and improves neural recovery after stroke. a.\u003c/strong\u003e Experimental design. \u003cstrong\u003eb.\u003c/strong\u003e Representative images of TTC staining of ATAD3A\u003csup\u003efl/fl\u003c/sup\u003e and ATAD3A\u003csup\u003eAst Het \u003c/sup\u003emice after tMCAO. \u003cstrong\u003ec.\u003c/strong\u003e Quantification of cerebral infarct volume (%) shown in (b) (n = 5 mice per group). \u003cstrong\u003ed \u003c/strong\u003eand\u003cstrong\u003e e.\u003c/strong\u003e Representative DTI axial views of fractional anisotropy (FA) and directionally encoded color (DEC) maps acquired 21 days after tMCAO in ATAD3A\u003csup\u003efl/fl\u003c/sup\u003e and ATAD3A\u003csup\u003eAst Het \u003c/sup\u003emice. \u003cstrong\u003ef. \u003c/strong\u003eQuantification of fractional anisotropy (FA ipsilateral/FA contralateral) in the external capsule and internal capsule (n = 5 mice per group). \u003cstrong\u003eg.\u003c/strong\u003e Adhesive removal test. The time taken to remove the tape from the affected forepaw was recorded (n = 8 mice per group). \u003cstrong\u003eh. \u003c/strong\u003eRotarod test. The latency to fall from the rotating rod was recorded (n = 8 mice per group). \u003cstrong\u003ei. \u003c/strong\u003eNovel object recognition test 14 days after tMCAO in ATAD3A\u003csup\u003efl/fl\u003c/sup\u003e and ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e mice. The image shows representative exploration tracks for novel (white square) and familiar (black circle) objects. Histogram: the preference for novel object and the recognition index (n = 8 mice per group). \u003cstrong\u003ej.\u003c/strong\u003e Morris water maze test 21~28 days after tMCAO in ATAD3A\u003csup\u003efl/fl\u003c/sup\u003e and ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e mice. The image shows representative swim paths. Histogram: the escape latency and the time in quadrant (% of total time in all quadrants) (n = 7~8 mice per group). All data are mean ± SEM. Two-tailed t test (\u003cstrong\u003ef\u003c/strong\u003e); one-way ANOVA with Bonferroni post hoc test (\u003cstrong\u003ec\u003c/strong\u003e, \u003cstrong\u003ei\u003c/strong\u003e and \u003cstrong\u003ej\u003c/strong\u003e); two-way ANOVA analysis (\u003cstrong\u003eg\u003c/strong\u003e, \u003cstrong\u003eh\u003c/strong\u003e and \u003cstrong\u003ej\u003c/strong\u003e); ns, no significance; *\u003cem\u003ep\u003c/em\u003e\u0026lt; 0.05; **\u003cem\u003ep\u003c/em\u003e \u0026lt; 0.01; ***\u003cem\u003ep\u003c/em\u003e \u003cem\u003e\u0026lt; \u003c/em\u003e0.005.\u003c/p\u003e","description":"","filename":"Fig9.png","url":"https://assets-eu.researchsquare.com/files/rs-6650856/v1/a95c4ae077b670fa39250b05.png"},{"id":101733839,"identity":"d9e3dde4-63e4-4a32-ae4e-02e5bc2bcdd6","added_by":"auto","created_at":"2026-02-03 06:50:49","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":238337,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAstrocytic ATAD3A deletion contributes to the recovery of memory and motor functions after stroke. a\u003c/strong\u003e. Experimental design. Bilateral hippocampal photothrombosis followed by astrocytic ATAD3A knockout induced with tamoxifen in Aldh1l1\u003csup\u003eCreERT\u003c/sup\u003e; ATAD3A\u003csup\u003efl/Δ\u003c/sup\u003e mice\u003cstrong\u003e. b.\u003c/strong\u003e Representative TTC staining of hippocampal sections from Sham, ATAD3A\u003csup\u003efl/fl\u003c/sup\u003e and Aldh1l1\u003csup\u003eCreERT\u003c/sup\u003e; ATAD3A\u003csup\u003efl/Δ\u003c/sup\u003e mice after photothrombosis. \u003cstrong\u003ec\u003c/strong\u003e. Quantification of lesion area (mm\u003csup\u003e2\u003c/sup\u003e) shown in (\u003cstrong\u003eb\u003c/strong\u003e) (n = 5 mice per group). \u003cstrong\u003ed. \u003c/strong\u003ePost-stroke contextual fear conditioning one week after sham/stroke. Day 7: 3 evenly spaced foot shocks (FS) were delivered in the fear chamber over a 7 min session. Day 8: mice return to fear chamber without FS (7 min). Freezing time (%) was quantified on Day 7 (pre-conditioning baseline) and Day 8 (24 h after foot-shock conditioning) in hippocampal-infarcted ATAD3A\u003csup\u003efl/fl\u003c/sup\u003e and ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e mice, with Sham as control (n=8 mice of sham group, n=7 mice of ATAD3A\u003csup\u003efl/fl\u003c/sup\u003e group, n=6 mice of ATAD3A\u003csup\u003eAst Het \u003c/sup\u003egroup). \u003cstrong\u003ee. \u003c/strong\u003eIn vivo fiber photometry recording of hippocampal calcium activity in the fear conditioning context. \u003cstrong\u003ef. \u003c/strong\u003eRepresentative Ca²⁺ traces (ΔF/F) aligned to the onset of context re-exposure from Sham, ATAD3A\u003csup\u003efl/fl\u003c/sup\u003e and ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e after hippocampal infarction. \u003cstrong\u003eg\u003c/strong\u003e. Quantification of ∆F\u003csub\u003emax\u003c/sub\u003e /F (%) (n = 5 mice per group). \u003cstrong\u003eh\u003c/strong\u003e. Experimental design. \u003cstrong\u003ei\u003c/strong\u003e. Forelimb symmetry in the cylinder task (n = 5 mice per group). \u003cstrong\u003ej\u003c/strong\u003e. Electrophysiological recording traces. Time scale, 250 ms/division. \u003cstrong\u003ek\u003c/strong\u003e. Burst activity detection in ipsilateral somatosensory cortex following peripheral stimulation. \u003cstrong\u003el\u003c/strong\u003e. Quantification of probability %. \u003cstrong\u003em\u003c/strong\u003e. Quantification of evoked Response (Z-Score) (n = 3 mice PTS+ ATAD3A\u003csup\u003efl/fl\u003c/sup\u003e group, unit n=101; n = 3 mice PTS+ ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e group, unit n=329). All data are mean ± SEM. Two-tailed t test (\u003cstrong\u003ed\u003c/strong\u003e); one-way ANOVA with Bonferroni post hoc test (\u003cstrong\u003ec\u003c/strong\u003e, \u003cstrong\u003eg\u003c/strong\u003e and \u003cstrong\u003em\u003c/strong\u003e); two-way ANOVA analysis (\u003cstrong\u003ei\u003c/strong\u003e); ns, no significance; *\u003cem\u003ep\u003c/em\u003e\u0026lt; 0.05; **\u003cem\u003ep \u003c/em\u003e\u0026lt; 0.01; ***\u003cem\u003ep\u003c/em\u003e \u0026lt; 0.005.\u003c/p\u003e","description":"","filename":"Fig11.png","url":"https://assets-eu.researchsquare.com/files/rs-6650856/v1/fbb9c9bfee99ef28f7f8cd57.png"},{"id":101733840,"identity":"60e70b0d-be16-4458-aae4-deacf124cd76","added_by":"auto","created_at":"2026-02-03 06:50:50","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":1137518,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eSingle-cell transcriptomics reveals ATAD3A ablation remodels astrocyte heterogeneity and inflammatory responses. a. \u003c/strong\u003eUMAP\u003cstrong\u003e \u003c/strong\u003eplot of\u003cstrong\u003e \u003c/strong\u003eall cells\u003cstrong\u003e \u003c/strong\u003ecolored by cell types. \u003cstrong\u003eb.\u003c/strong\u003e Stacked barplots comparing cell type proportions between ATAD3A\u003csup\u003eWT\u003c/sup\u003e and ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e groups. \u003cstrong\u003ec.\u003c/strong\u003e UMAP\u003cstrong\u003e \u003c/strong\u003eplot of\u003cstrong\u003e \u003c/strong\u003eastrocytes\u003cstrong\u003e \u003c/strong\u003ecolored by astrocyte subtypes. \u003cstrong\u003ed.\u003c/strong\u003e Dot plot showing the expression level of glycogen metabolism related genes in astrocyte subtypes. \u003cstrong\u003ee. \u003c/strong\u003eDot plot depicting the expression level of astrocyte reactive genes in all kinds of astrocyte subtypes. \u003cstrong\u003ef.\u003c/strong\u003e Violin plot showing the expression level of astrocyte brain region specific genes in various astrocyte subtypes. \u003cstrong\u003eg. \u003c/strong\u003eMulti-color immunofluorescence staining of the ischemic penumbra for GFAP (gray, astrocytes), STBD1 (red, glycogen aggregates), HDAC3 (purple), ATAD3A (green), and DAPI (blue). Scale bar, 30 μm. Insets show magnified views of representative non-glycogen and glycogen-accumulated astrocytes. Scale bar, 5 μm. \u003cstrong\u003eh.\u003c/strong\u003e Heatmaps displaying the expression of astrocyte subtype differentially expressed genes (DEGs). \u003cstrong\u003ei.\u003c/strong\u003e Barplots displaying pathway enrichment results of astrocyte subtype DEGs. \u003cstrong\u003ej.\u003c/strong\u003e Human brain staining pictures indicate the expression of astrocytes (GFAP, white), ATAD3A (green) and GYS1 (red) in ischemic and non-ischemic areas. \u003cstrong\u003ek.\u003c/strong\u003e Statistical analysis of Ratio GYS1 spots/GFAP area (per cell) in ischemic and non-ischemic areas from (\u003cstrong\u003ej\u003c/strong\u003e) (n = 10). \u003cstrong\u003el.\u003c/strong\u003e Statistical analysis of Ratio ATAD3A spots/GFAP area (per cell) in ischemic and non-ischemic areas from (\u003cstrong\u003ej\u003c/strong\u003e) (n = 10). \u003cstrong\u003em. \u003c/strong\u003eStatistical analysis of the cellular proportion of glycogen reactive astrocyte A1 between ATAD3A\u003csup\u003eWT\u003c/sup\u003e and ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e groups. \u003cstrong\u003en.\u003c/strong\u003e Western blotting analysis depicts the protein levels in isolated adult astrocytes of pDrp1\u003csup\u003es579\u003c/sup\u003e, pDrp1\u003csup\u003es600\u003c/sup\u003e, Drp1, HDAC3, GFAP and actin after astrocytic ATAD3A knockout or not. Two-tailed t test (\u003cstrong\u003ek\u003c/strong\u003e, \u003cstrong\u003el\u003c/strong\u003e and \u003cstrong\u003em\u003c/strong\u003e); ***\u003cem\u003ep\u003c/em\u003e \u003cem\u003e\u0026lt; \u003c/em\u003e0.005.\u003c/p\u003e","description":"","filename":"Fig13.png","url":"https://assets-eu.researchsquare.com/files/rs-6650856/v1/f84a55af4d1be6ab1872248f.png"},{"id":101753865,"identity":"168d7808-336e-4f6f-bdfe-fc5ac9563d6d","added_by":"auto","created_at":"2026-02-03 10:41:04","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":263096,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAstrocytic glycogen depletion and HDAC3 inhibition act synergistically to enhance post-stroke recovery. a.\u003c/strong\u003e Forelimb use was assessed by cylinder test at indicated time points before and after photothrombotic stroke (PTS). Daily administration of RGFP966 (10 mg/kg, i.p.), Cotadutide (10 nmol/kg, s.c.), or both (CotR) started at 3 days post-PTS. The time spent on left paw relative to right paw (%) was recorded (n = 5 mice per group). \u003cstrong\u003eb.\u003c/strong\u003e The time courses of normalized GCaMP6s fluorescence in peri-infarct pyramidal neurons were examined before and during spontaneous rearing behavior. \u003cstrong\u003ec.\u003c/strong\u003e Quantification of ∆F\u003csub\u003emax\u003c/sub\u003e /F (%) (fluorescence changes after spontaneous rear) (n = 8 mice per group). \u003cstrong\u003ed.\u003c/strong\u003e Schematic of in vivo multichannel electrophysiology (Left). The expression of AAV-CaMK II-ChR2-eGFP (green) in cortex (M: motor cortex; S: sensory cortex) (Right). Scale bar, 1 mm. \u003cstrong\u003ee.\u003c/strong\u003e Electrophysiological recording traces. Time scale, 250 ms/division. \u003cstrong\u003ef.\u003c/strong\u003e Burst activity detection in ipsilateral somatosensory cortex following peripheral stimulation. \u003cstrong\u003eg. \u003c/strong\u003eQuantification of cumulative % of Evoked Responses. \u003cstrong\u003eh. \u003c/strong\u003eQuantification of cumulative evoked response (Z-Score) (n = 3 mice Stroke + Vehicle group, unit n=238; n = 3 mice Stroke + Cotadutide + RGFP966 group, unit n=240). \u003cstrong\u003ei. \u003c/strong\u003eLeft: foot faults of the left forelimb in the grid-walking task. Right: forelimb symmetry in the cylinder task (n = 5 mice per group). \u003cstrong\u003ej\u003c/strong\u003e. Experimental design. \u003cstrong\u003ek\u003c/strong\u003e. Burst activity detection in ipsilateral somatosensory cortex following peripheral stimulation (Top). Quantification of Probability (%) (bottom). \u003cstrong\u003el\u003c/strong\u003e. Quantification of cumulative evoked response (Firing Rate) (Top); Quantification of cumulative evoked response (Z-Score) (bottom) (n = 2 mice Sham group, unit n=142; n = 4 mice AAV-Vector+PTS group, unit n=268; n = 3 mice AAV-shHDAC3+PTS group, unit n=323; n = 2 mice AAV-Pygb+PTS group, unit n=82; n = 2 mice AAV-Pygb + AAV-shHDAC3+PTS group, unit n=141). \u003cstrong\u003em.\u003c/strong\u003e Mechanism schematic description. This study highlights that (1) Excessive astrocytic glycogen granules contribute to ischemic brain injury; (2) accumulated glycogen aggregates initiates ATAD3A oligomerization signaling; (3) Glycogen chamber exists ATAD3A, and its oligomerization requires HDAC3 catalysis; (4) ATAD3A oligomerization signaling abrogation improves long-term outcomes after stroke. All data are mean ± SEM. One-way ANOVA with Bonferroni post hoc test (\u003cstrong\u003ec,\u003c/strong\u003e \u003cstrong\u003eh \u003c/strong\u003eand \u003cstrong\u003el\u003c/strong\u003e); two-way ANOVA analysis (\u003cstrong\u003ea \u003c/strong\u003eand\u003cstrong\u003e i\u003c/strong\u003e); *\u003cem\u003ep\u003c/em\u003e\u0026lt; 0.05; **\u003cem\u003ep\u003c/em\u003e \u0026lt; 0.01; ***\u003cem\u003ep\u003c/em\u003e \u0026lt; 0.005.\u003c/p\u003e","description":"","filename":"Fig15.png","url":"https://assets-eu.researchsquare.com/files/rs-6650856/v1/658ab2437a62a9d46f3b7e12.png"},{"id":102295826,"identity":"31b9bb76-4323-4e26-9eb0-2bb2bc2ae676","added_by":"auto","created_at":"2026-02-10 10:15:19","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":10443980,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6650856/v1/a1ee5264-7059-4764-8cd8-66b89d1c3606.pdf"},{"id":101753529,"identity":"2e930683-1963-4ca2-9758-726cf5721d8b","added_by":"auto","created_at":"2026-02-03 10:40:12","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":127080,"visible":true,"origin":"","legend":"Reporting Summary","description":"","filename":"RS.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6650856/v1/ac8dcd0b18d509ec5eff14fc.pdf"},{"id":101733843,"identity":"926bd4a1-1802-4215-acd1-6574caf6056a","added_by":"auto","created_at":"2026-02-03 06:50:50","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":26147422,"visible":true,"origin":"","legend":"Source Data","description":"","filename":"SourceData.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6650856/v1/6e374a0663ca3e4f1c3c2e4e.xlsx"},{"id":101733841,"identity":"9ba6b3df-5ec5-4bd9-b5d8-7cc9fb3a40c8","added_by":"auto","created_at":"2026-02-03 06:50:50","extension":"pdf","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":7361148,"visible":true,"origin":"","legend":"Author list statement","description":"","filename":"Authorliststatement.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6650856/v1/b4ceae310e7c0e5d382475b0.pdf"},{"id":101733846,"identity":"7ef2990d-7410-4b5a-8d3a-c933f2ab4ec0","added_by":"auto","created_at":"2026-02-03 06:50:56","extension":"zip","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":503733144,"visible":true,"origin":"","legend":"supData","description":"","filename":"supData.zip","url":"https://assets-eu.researchsquare.com/files/rs-6650856/v1/f548ec39057f6aeec90198b0.zip"},{"id":101733844,"identity":"f5cb4413-b152-451a-a77f-9fc53ec01ecd","added_by":"auto","created_at":"2026-02-03 06:50:51","extension":"pdf","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":77786794,"visible":true,"origin":"","legend":"","description":"","filename":"ExtendedFigures.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6650856/v1/3572e54d6dce49ce34cba335.pdf"}],"financialInterests":"There is no duality of interest","formattedTitle":"Astrocytic glycogen aggregates induce ATAD3A oligomerization mediated mitochondrial fragmentation and impede stroke recovery","fulltext":[{"header":"Introduction","content":"\u003cp\u003eStroke is a leading cause of severe disability worldwide, driving increased focus on recovery mechanisms due to limited therapeutic strategies. Current approaches to restore neural function include rehabilitation, physical stimulation, and pharmacological interventions\u003csup\u003e1\u003c/sup\u003e. Molecular, cellular, and tissue network remodeling and behavioral systems occur dynamically throughout the entire stroke recovery course. These\u0026nbsp;process include growth-promoting\u0026nbsp;gene\u0026nbsp;upregulation, axonal sprouting, dendritic spine turnover, synaptogenesis, axonal projections,\u0026nbsp;and\u0026nbsp;neural plasticity,\u0026nbsp;such as motor-sensor map plasticity\u0026nbsp;and\u0026nbsp;glial cell homeostasis\u003csup\u003e2\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eAstrocytes, strategically positioned between capillaries and neurons, supply energy and maintain synaptic homeostasis\u003csup\u003e3\u003c/sup\u003e.\u0026nbsp;Astrocytes and neurons can form tripartite synapse structures. This structure ensures rapid astrocyte\u0026ndash;neuron nutrient transport and promotes the shuttle of mitochondria from connected astrocytes when neurons meet the increased glucose demand resulting from high neuronal activity or a shortage of substrate supply because of capillary occlusion,\u003cstrong\u003e\u0026nbsp;such as ischemia\u003c/strong\u003e\u003csup\u003e4\u003c/sup\u003e. Astrocytes metabolize glycogen into lactate, which fuels distal neurons via gap junctions\u003csup\u003e3\u003c/sup\u003e. Recent studies show that insoluble glycogen accumulates in astrocytes in human, primate, and rodent brains after ischemic stroke\u003csup\u003e5\u003c/sup\u003e. These accumulated glycogens which are difficult to be utilized in time are not conducive to long-term functional recovery in stroke animal models\u003csup\u003e5\u003c/sup\u003e.Yet the pathological impact of astrocytic glycogen deposition remains unclear, hindering therapeutic innovation.\u003c/p\u003e\n\u003cp\u003eGlycogens, as storage polysaccharides, are steadily and dynamically balanced between glycogenesis and glycogenolysis for endogenous energy reserves under slight energy deficiency. Glycogen maintain the normal physiological functions of cells, such as\u003cstrong\u003e\u0026nbsp;protein glycosylation\u003c/strong\u003e\u003csup\u003e6\u003c/sup\u003e, lactate release\u003csup\u003e3\u003c/sup\u003e, ATP synthesis, and GABA production, and is closely related to movement, memory, and lifespan\u003csup\u003e7, 8\u003c/sup\u003e. Under pathological conditions, insoluble glycogen aggregates (called polyglucosan bodies (PGBs), glycogen stress, Lafora bodies in Lafora disease) form in the brain\u003csup\u003e9\u003c/sup\u003e and are associated with aging, cognitive decline and dementia\u003csup\u003e10\u003c/sup\u003e, suggesting a role in impaired synaptic plasticity. Inhibiting glycogen aggregates formation reduces astrogliosis and neuroinflammation\u003csup\u003e11\u003c/sup\u003e. These insoluble, dense cytoplasmic polysaccharides, in addition to encapsulating large-molecule polyglucosides, also rich in glycogen-binding proteins like Stbd1 and laforin\u003csup\u003e12\u003c/sup\u003e. These aggregates resist lysosomal degradation and act as membrane-free compartments in stress signaling\u003csup\u003e6, 13\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eHere, we report that glycogen aggregates deposit mainly in a subset of astrocytes, exacerbating injury and impairing brain repair. We identified glycogen-binding proteins in the penumbra and found astrocytic glycogen closely associates with mitochondrial fission receptor ATAD3A. Glycogen stress granules anchor the outer mitochondrial membrane and recruit cytoplasmic HDAC3 to deacetylate the ATAD3A outer segment of the mitochondrial membrane. Deacetylation initiated ATAD3A oligomerization-dependent mitochondrial fragmentation. Knocking down astrocytic ATAD3A reduced neurotoxicity, synaptic defects, compromised neural circuit reorganization and cognitive decline after stroke. Depleting glycogen granules and blocking downstream signaling improved sensorimotor recovery by enhancing synaptic and circuit plasticity. Thus, our findings reveal how astrocytic glycogen stress delays stroke recovery and highlight a clinically translatable strategy for treatment.\u0026nbsp;\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eGlycogen aggregates mainly increase in astrocytes from the ischemic penumbra after stroke\u003c/p\u003e\n\u003cp\u003eGlycogen staining in photothrombotic and MCAO mouse models showed glycogen lesions enriched in the ipsilateral penumbra but absent contralaterally (Fig. 1a, b and Extended Data Fig. 1a). In MCAO mouse models, glycogen deposition peaked early and persisted for ~2 weeks, declining by day 21 (Extended Data Fig. 1c, d). Co-staining for Stbd1 (glycogen granule marker)\u003csup\u003e13\u003c/sup\u003e and GFAP (astrocyte marker) confirmed glycogen-laden astrocytes were concentrated in the penumbra (Extended Data Fig.1 e). Some Stbd1⁺ cells were non-astrocytic; co-staining with Iba1 and NeuN revealed minor glycogen accumulation in neurons and microglia (Extended Data Fig. 1f, g). RNA analysis of glycogen-rich versus glycogen-sparse regions showed decreased glycogenolysis genes (such as G6pc, Pygb) and increased glycogenesis genes (such as Gbe1, Gys1) (Fig. 1c), consistent with previous studies\u003csup\u003e5, 13\u003c/sup\u003e. Then, single-cell RNA sequencing distinguished four astrocyte clusters from the public database of mouse ischemic stroke (Extended Data Fig. 1b). Clustering based on glycogenesis genes (Gys1, Gbe1) revealed that clusters 0, 1, and 2 were enriched for glycogen deposition signatures, unlike cluster 3 (Extended Data Fig. 1h). It has been suggested that reactive astrocytes diversify beyond\u0026nbsp;neurotoxic and neuroprotective astrocytes\u003csup\u003e14\u003c/sup\u003e, which can be further subdivided into unique subpopulations. We next performed clustering mapping on distinct astrocytes (neurotoxic astrocyte cluster A1; for example, \u003cem\u003eC3\u003c/em\u003e\u003csup\u003e\u0026nbsp;\u003c/sup\u003eand \u003cem\u003eSerpina3n\u003c/em\u003e) and neuroprotective (neuroprotective astrocyte cluster A2; for example, \u003cem\u003eEmp1\u003c/em\u003e) reactive astrocytes and pan-reactive astrocytes (\u003cem\u003eGfap\u003c/em\u003e, \u003cem\u003eVim\u003c/em\u003e)\u003csup\u003e15-17\u003c/sup\u003e with glycogen-laden astrocytes subsets. Results indicated that cluster 0 represented glycogen-enriched genes but lacked pan-reactive characteristics, which is regarded as glycogen resting astrocytes. While astrocytes with neuroprotective and glycogen-enriched genes were regarded as glycogen-reactive and neuroprotective cluster. Astrocytes with neurotoxic and glycogen-enriched genes were regarded as glycogen-reactive and neurotoxic cluster. And resting astrocyte with neither glycogen accumulation nor activation (Extended Data Fig. 1i). Decreased glycogenolysis, increased glycogenesis, decreased glycogen branching points all contribute to low soluble glycogen granule deposition, containing many glycogenin such as Stbd1 (covalently bound to the glucose polymer) and other proteins (such as laforin (Epm2a, GBE)\u003csup\u003e12\u003c/sup\u003e. Immunofluorescence confirmed glycogen accumulated astrocytes contained higher GYS1 (Extended Data Fig. 1j, k; area 01) or GBE1 (Extended Data Fig. 1j, l; area 02) expression. Post-stroke, resting and glycogen-resting astrocytes decreased, while glycogen-reactive neurotoxic or neuroprotective subtypes increased (Extended Data Fig. 1m). Thus, insoluble glycogen granules increase mainly in penumbral astrocytes after stroke.\u003c/p\u003e\n\u003cp\u003eAstrocytic glycogen deposition correlates with ATAD3A oligomerization and neuropathology after stroke\u003c/p\u003e\n\u003cp\u003eGlycogen pulldown and mass spectrometry identified a glycogen granule proteome enriched for mitochondrial pathways: respiratory chain complexes, mitochondrial respirasomes, and mitochondrial membrane protein complexes (Fig. 1d, Extended Data Fig. 2a, b, Supplementary Table 1). Cross-referencing with a mitoproteome database highlighted oxidative phosphorylation (Extended Data Fig. 2c, Supplementary Table\u0026nbsp;1)\u003csup\u003e18\u003c/sup\u003e, dynamics (MFF, DNM1L, GDAP1), homeostasis, and quality control (AFG3L2, CLPB, ATAD3A) (Fig. 1e, Extended Data Fig. 2c).\u003c/p\u003e\n\u003cp\u003eATAD3, a mitochondrial AAA+ ATPase, has three paralogs (ATAD3A, ATAD3B, and ATAD3C) in humans and only ATAD3A in mice. As a fission receptor of mitochondria, ATAD3A regulates mitochondrial dynamics,\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003emorphology, contact sites and nucleoid trafficking\u003csup\u003e19\u003c/sup\u003e. ATAD3A mutation in human cause neurodegeneration. ATAD3A oligomers recruit cytoplasmic Drp1 and mark mitochondrial damage, directly initiate mitochondrial fragmentation in neurological diseases\u003csup\u003e20, 21\u003c/sup\u003e. Next, we directly incubated the tissue lysate with purified glycogen and grasped the glycogen pellet via centrifugation. Results indicated that ATAD3A co-precipitated with glycogen pellets (Fig. 1f). To confirm whether glycogen directly binds to ATAD3A or requires indirect glycogen binding, \u003cem\u003ein vitro\u003c/em\u003e protein binding experiments were performed by sequentially assigning purified glycogen and purified ATAD3A-His protein (Fig. 1g). We directly incubated the tissue lysate with purified glycogen and detected endogenous ATAD3A enrichment (lane 2 vs lane 1) (Fig. 1g). We next reincubated His-ATAD3A with the proteinase K-treated glycogen protein complex and observed no enrichment of endogenous ATAD3A or His-ATAD3A compared with the nonproteinase K-treated group (lane 5 vs lane 6) (Fig. 1g). These findings suggest that glycogen anchors the ATAD3A protein, which is not dependent on charge affinity or direct binding but is indirectly bound through glycogen-binding proteins. Serial sections showed spatial correlation between glycogen aggregates and ATAD3A oligomers in the ischemic penumbra (Fig. 1h, i). We detected ATAD3A oligomers in the injury foci at different time points after ischemic stroke. Results indicated that\u0026nbsp;in the physiological state of the brain, there is a presence of small amounts of ATAD3A oligomers, and during the acute phase of ischemic stroke, the level of ATAD3A oligomers sharply increases (Fig. 1j). Immunofluorescence staining further confirmed ATAD3A oligomers easily form in ischemic area, especially in astrocytes and rarely form in normal uninjured brain area (Fig. 1k, l). Further staining results revealed the colocalization of ATAD3A oligomers and glycogen aggregates in astrocytes were more obvious in glycogen accumulated astrocytes (Fig. 1m, n). Whlie, minor oligomerization occurred in neurons or microglia/macrophage (Extended Data Fig. 3a, b). To further verify whether ATAD3A oligomers are mainly derived from astrocytes \u003cem\u003ein vivo\u003c/em\u003e, astrocyte-specific ablation markedly reduced ATAD3A oligomers after stroke (Fig. 1o, Extended Data Fig. 3c, d).\u003c/p\u003e\n\u003cp\u003eG6PT⁻\u003csup\u003e/\u003c/sup\u003e⁻ mice,\u0026nbsp;a glycogen storage disease model of glycogen storage disease type Ib, manifested glycogen accumulation in the brain (Extended Data Fig. 3e, f). We next used G6PT\u003csup\u003e-/-\u003c/sup\u003e mice to assess whether glycogen aggregates predisposition is adverse to ischemic stroke prognosis. To rule out G6PT knockout-induced defects in peripheral immunity, we constructed a parabiosis model in which glycogen storage disease mice and normal wild-type mice were compared (Extended Data Fig. 3g, h). In the brains of the recipient mice, the PTS model was established 2 weeks after surgery. The results revealed that the \u003cem\u003ein situ\u003c/em\u003e predeposition of glycogen in the brain worse stroke injury (Extended Data Fig. 3i, j) and promoted ATAD3A oligomer formation (Extended Data Fig. 3k). Glycogen aggregates thus exhibit mitochondrial affinity and may disrupt homeostasis after stroke.\u003c/p\u003e\n\u003cp\u003eGlycogen aggregation causes mitochondrial dysfunction via ATAD3A\u003c/p\u003e\n\u003cp\u003eNext, we observed that mitochondria area in glycogen-rich areas was smaller than in low-glycogen regions (Fig. 2a, b). We next sought to investigate how ATAD3A can be recruited to glycogen droplets and identify the binding partners of ATAD3A in the glycogen compartment. Immunoprecipitation identified Epm2aip as an ATAD3A-binding glycogenin proteins, which directly interact with glycogen\u003csup\u003e13, 22\u003c/sup\u003e (Extended Data Fig. 4a). And Epm2aip directly binds ATAD3A from the outer to the inner segment of the mitochondrial membrane (Extended Data Fig. 4b, c). Exogenously overexpressing Epm2aip-Flag and ATAD3A-Myc in human astrocytes (SVG p12) did not result in the formation of ATAD3A oligomers (Extended Data Fig. 3d). The overexpression of GYS1 was used to mimic a cell model of glycogen stress granule accumulation in astrocytes (SVG p12\u003csup\u003eGYS1 OE\u003c/sup\u003e cells) (Extended Data Fig. 4e, f)\u003csup\u003e11, 13\u003c/sup\u003e. Overexpressing GYS1 in astrocytes induced glycogen granules and ATAD3A oligomerization, reversed by Epm2aip knockdown (Fig. 2c). Knockdown of ATAD3A in SVG p12\u003csup\u003eGYS1 OE\u003c/sup\u003e cells\u0026nbsp;reduced mitochondrial fragmentation\u0026nbsp;(Fig. 2d, e). Under physiological conditions,\u0026nbsp;the\u0026nbsp;ATAD3A monomer helps maintain the formation and morphology of mitochondrial cristae by composing mitochondrial contact\u0026nbsp;sites and cristae organizing system (MICOS) complexes\u003csup\u003e23\u003c/sup\u003e. The\u0026nbsp;transmission electron\u0026nbsp;microscopy results revealed that the number of cristae in the\u0026nbsp;mitochondria adjacent to glycogen stress granules decreased. While,\u0026nbsp;ATAD3A knockdown restored cristae structure\u0026nbsp;(Fig. 2f, g). ATAD3A oligomers have been\u0026nbsp;previously\u0026nbsp;reported to participate in mitochondrial\u0026nbsp;nucleoid formation and perturb mtDNA maintenance and the machinery of protein synthesis\u003csup\u003e20\u003c/sup\u003e. Astrocytic glycogen aggregates deposition impaired ATP synthesis (Fig. 2h) and oxygen consumption, which can be rescued by ATAD3A knockdown (Fig. 2i, j). ATAD3A 200 aa contain the CC1~2 region, and ATAD3A 300 aa contain the CC1~2 and TMS regions, which were previously reported to anchor mitochondria\u003csup\u003e20, 24\u003c/sup\u003e. SVG p12\u003csup\u003eGYS1 OE\u003c/sup\u003e cells were expressed with ATAD3A 200 aa or 300 aa GFP tag proteins. Only mitochondrial-anchored ATAD3A fragments (300 aa) oligomerized (Extended Data Fig. 4g). Together, astrocytic glycogen granules lead to ATAD3A oligomerization-triggered mitochondrial dysfunction.\u003c/p\u003e\n\u003cp\u003eThe glycogen chamber contains HDAC3, and ATAD3A oligomerization is dependent on deacetylation\u003c/p\u003e\n\u003cp\u003eAbove evidence shown that ATAD3A oligomerization requires the accumulation of astrocytic glycogen aggregates. However, under physiological conditions, ATAD3A does not dramatically oligomerize in non-glycogen\u0026nbsp;aggregates deposited cells and un-ischemic brain tissue. Mitochondria also do not exhibit obvious dysfunction in non-glycogen stress granule-deposited cells.\u0026nbsp;Therefore, we speculate that the glycogen chamber may carry some proteins to the mitochondria, promoting ATAD3A oligomerization. To explore what candidate partners are involved in it, the proximity-biotinylating enzyme TurboID was reconstituted with ATAD3A to detect biotinylated proteins adjacent to ATAD3A (Fig. 3a). TurboID proximity labeling identified 182 ATAD3A-neighboring proteins (Fig. 3b, Supplementary Table\u0026nbsp;1). Subcellular and pathway analyses highlighted mitochondrial protein localization (Extended Data Fig. 5a). KEGG pathway analysis revealed that protein polymerization and protein localization to mitochondria are associated with ATAD3A (Extended Data Fig. 5b). We intersected proteins between the glycogen proteome and the biotinylated proteome and found that many proteins, such as the nuclear protein histone deacetylase 3 (HDAC3), Dnm1l (Drp1), ATAD3A, and VDAC1, were enriched (Supplementary\u0026nbsp;Table\u0026nbsp;1). The mitochondrial membrane outer segment of ATAD3A deacetylation has been reported to be an independent factor initiating ATAD3A self-oligomerization\u003csup\u003e21\u003c/sup\u003e, but the enzymes involved in ATAD3A deacetylation are currently unclear. We speculate that HDAC3 may be a critical candidate effector (Fig. 3c). We further confirmed that the glycogen protein Epm2aip can interact with HDAC3 (Extended Data Fig. 5c). Screening deacetylases in the Sirt family (Extended Data Fig. 5d) and HDAC superfamily (Extended Data Fig. 5e) revealed\u0026nbsp;Sirt1, HDAC3 and HDAC6 as candidates. To determine whether ATAD3A can be acetylated by HDACs or Sirts, SVG p12 cells were transfected with Myc-tagged ATAD3A and treated with an inhibitor of sirtuin family deacetylases (nicotinamide; NAM) or the pan-HDAC inhibitor trichostatin A (TSA). Pan-HDAC inhibition, but not sirtuin inhibition, increased ATAD3A acetylation (Fig. 3d). GST pull-down and split-luciferase assays confirmed HDAC3\u0026ndash;ATAD3A interaction (Fig. 3e, f , Extended Data Fig. 5f, g). Domain mapping of human HDAC3 and human ATAD3A revealed that the regions of homology are located only in the CC domain (mitochondrial membrane outer segment) of hATAD3A and the deacetylase domain of hHDAC3 (Fig. 3g).\u003c/p\u003e\n\u003cp\u003eThe human ATAD3A K135 site (K134 in mice) was previously identified as a critical acetylation site\u003csup\u003e20\u003c/sup\u003e, and its deacetylation triggers ATAD3A self-oligomerization. Cytoplasmic HDAC3 was previously reported to translocate to mitochondria to execute deacetylation as a \u0026lsquo;decision maker\u0026rsquo; under conditions of cell stress\u003csup\u003e25\u003c/sup\u003e. Therefore, we hypothesized that HDAC3 can modify ATAD3A to participate in oligomerization. Cells were transfected with Myc-tagged ATAD3A with two deacetylase-dead HDAC3 mutants (H134/135A and Y298F) or pharmacological pan-HDAC inhibition (TSA). We found that acetylated ATAD3A levels subsequently decreased after HDAC3\u003cem\u003e\u003csup\u003eWT\u003c/sup\u003e\u003c/em\u003e overexpression. However, two deacetylase-dead HDAC3 mutants (mainly Y298F) or pan-HDAC inhibition restored and augmented acetylated ATAD3A levels, respectively (Fig. 3h). And HDAC3 knockout increased ATAD3A acetylation (Fig. 3i). When HDAC3\u003cem\u003e\u003csup\u003eWT\u003c/sup\u003e\u003c/em\u003e but HDAC3\u003cem\u003e\u003csup\u003eY298F\u003c/sup\u003e\u003c/em\u003e\u003cem\u003e\u003csup\u003e\u0026nbsp;\u003c/sup\u003e\u003c/em\u003ewere overexpressed in\u0026nbsp;HDAC3 deficient cell, ATAD3A acetylation decreased correspondingly (Fig. 3i). However, the other candidate partner, HDAC6, did not yield positive results (Extended Data Fig. 5h). Lysine acetyltransferases (KATs) and deacetylases can both regulate protein acetylation. We next searched for KATs responsible for ATAD3A acetylation (Extended Data Fig. 5i). The results indicated that Kat8 is a potential effector. Kat8 is shuttled between the nucleus and mitochondria and is required to rescue transcriptional and respiratory defects via its catalytic activity\u003csup\u003e26\u003c/sup\u003e.\u0026nbsp;Co-expression experiments confirmed that Kat8 increases ATAD3A acetylation, which is reversed by wild-type HDAC3 but not a catalytically dead mutant (HDAC3\u003cem\u003e\u003csup\u003eY298F\u003c/sup\u003e\u003c/em\u003e) or HDAC6 (Extended Data Fig. 5j, k).\u0026nbsp;The lysine 134 residue in mice, which is conserved among species, resides in the linker between the CC1 and CC2 domains and is accessible for reversible acetylation\u003csup\u003e20\u003c/sup\u003e. To further investigate ATAD3A deacetylation mediated by HDAC3 in\u0026nbsp;ATAD3A oligomerization, an acetylation-mimetic mutant (K134Q) and a deacetylation-mimetic mutant (K134E) were generated.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eThe K134E mutant enhanced ATAD3A oligomerization, while K134Q had no effect, even with HDAC3 co-expression (Fig. 3j).\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eIn vitro assays with purified proteins confirmed that Kat8 directly acetylates ATAD3A (Fig. 3k), and HDAC3 deacetylates it but not catalytically dead mutant (HDAC3\u003cem\u003e\u003csup\u003eY298F\u003c/sup\u003e\u003c/em\u003e) or Stirt1(Fig. 3l, m). Subsequently, Proximity ligation assay (PLA) was performed to visualize the interaction between ATAD3A and HDAC3 proteins in astrocytes under physiological or pathological conditions. Under physiological conditions, we observed an interaction between HDAC3 and ATAD3A in astrocytes, with minimal levels detected in non-astrocytic cells. Following ischemic stroke, this interaction was markedly enhanced in astrocytes, which was quantified by an increase in HDAC3-ATAD3A interaction spots per GFAP area (Fig. 3n, o). Critically, the stroke-induced enhancement of the HDAC3-ATAD3A interaction was abolished by pre-knockdown of HDAC3 in astrocytes (Extended Data Fig. 5l, m). And under physiological conditions, there is an extremely weak presence of ATAD3A oligomers. When ischemic stroke occurs, the level of ATAD3A oligomerization sharply increases. While, ATAD3A oligomerization significantly decreases after astrocytic HDAC3 knockdown (Fig. 3p). In addition, astrocytic ATAD3A acetylation was significantly downregulated, which is rescued by astrocyte-specific HDAC3 knockdown (Fig. 3q).\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eThus, HDAC3 promotes ATAD3A oligomerization by direct deacetylation both in vivo and in vitro.\u003c/p\u003e\n\u003cp\u003eHDAC3 regulates ATAD3A-mediated astrocytic mitochondrial dysfunction\u003c/p\u003e\n\u003cp\u003eATAD3A deacetylation promotes Drp1 recruitment and mitochondrial fragmentation\u003csup\u003e20\u003c/sup\u003e. To further validate the correlations\u0026nbsp;among ATAD3A acetylation\u0026nbsp;and\u0026nbsp;Drp1 and HDAC3 protein levels\u0026nbsp;\u003cem\u003ein vivo\u003c/em\u003e after ischemic stroke,\u0026nbsp;results indicated that ATAD3A acetylation decreased over time (Fig. 4a). Endogenous ATAD3A immunoprecipitation was performed to assess the relationship between ATAD3A acetylation and mitochondrial function. The phosphorylation of Drp1 at Ser 600 (the active site) or Ser 579 (the effector site) in mice, which is equivalent to Ser616 and Ser637 in humans, is required from the cytosol to the mitochondrial outer membrane for maximal mitochondrial fragmentation\u003csup\u003e27\u003c/sup\u003e. We found that the level of mitochondrial fragmentation increased, accompanied by decreased ATAD3A acetylation and increased HDAC3 binding after ischemic stroke, and that the increase in Drp1 activity for mitochondrial fragmentation manifested as increased serine 579 site phosphorylation (Fig. 4b).\u003c/p\u003e\n\u003cp\u003eIn addition, HDAC3-mediated deacetylation of ATAD3A increased the binding ability between ATAD3As (Fig. 4c). Treatment with an HDAC3-specific inhibitor decreased the binding ability between ATAD3As (Extended Data Fig. 6a~c). HDAC3 interacted with the ATAD3A N-terminus (1~200 aa) (Extended Data Fig. 6d); truncation mutants confirmed after transfecting the truncated mutants of ATAD3A in cells, the ATAD3A-GFP 300aa (containing CC1~2 and TMS domains) mutant strongly increased ATAD3A dimers under HDAC3-mediated deacetylation (Fig. 4d), whereas mitochondrial targeting transmembrane sequence (TMS) deficiency (ATAD3A-GFP 100aa or ATAD3A-GFP 200aa) abolished dimerization (Extended Data Fig. 6e). The acetylated\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eATAD3A CC domain is required for the recruitment of the Drp1 mitochondrial location\u003csup\u003e24\u003c/sup\u003e.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eThe expression of the ATAD3A-GFP 200 aa mutant or full ATAD3A-GFP induced Drp1 translocation to interact with the ATAD3A outer mitochondrial membrane segment, which was enhanced by HDAC3 and blocked by RGFP966 treatment (Fig. 4e, f, Extended Data Fig. 6 f~h).\u003c/p\u003e\n\u003cp\u003eWe further investigated whether and how HDAC3 regulates ATAD3A-mediated astrocytic mitochondrial dysfunction. It was previously reported that mitochondria presented moderate fragmentation upon exogenous ATAD3A overexpression\u003csup\u003e20, 28\u003c/sup\u003e and that co-expression with HDAC3 augmented branched mitochondrial structures in SVG p12 cells. Compared with ATAD3A\u003csup\u003eWT\u003c/sup\u003e cells, SVG p12 cells expressing ATAD3A\u003cem\u003e\u003csup\u003eK134Q\u003c/sup\u003e\u003c/em\u003e\u003cem\u003e\u0026nbsp;\u003c/em\u003e(acetylation gain-of-function mutation) coupled with HDAC3 presented relatively reduced fragmented mitochondria (Extended Data Fig. 6i, j). Furthermore, ATAD3A\u003cem\u003e\u003csup\u003eK134Q\u003c/sup\u003e\u003c/em\u003e + HDAC3 overexpression attenuated mitochondrial respiratory defects compared with ATAD3A\u003cem\u003e\u003csup\u003eWT\u003c/sup\u003e\u003c/em\u003e\u003cem\u003e\u003csup\u003e-\u003c/sup\u003e\u003c/em\u003e + HDAC3-expressing SVG p12 cells (Extended Data Fig. 6k, l); ATAD3A\u003cem\u003e\u003csup\u003eK134Q\u003c/sup\u003e\u003c/em\u003e + HDAC3 overexpression improved the maximal and spare respiratory capacity but had minor effects on mitochondrial function in ATAD3A-overexpressing SVG p12 cells. However, obvious mitochondrial fragmentation (Extended Data Fig. 7 i, j) and severe mitochondrial respiratory defects (Extended Data Fig. 6 k, l) were also observed in HDAC3-overexpressing cells, indicating that endogenous ATAD3A was also deacetylated. To eliminate the background effects of endogenous ATAD3A in SVG p12 cells, we next performed experiments in ATAD3 KD (ATAD3A knockdown) SVG p12 cells (SVG p12\u003cem\u003e\u003csup\u003eATAD3 KD\u003c/sup\u003e\u003c/em\u003e). Basal, maximal and spare respiratory capacity were significantly compromised after ATAD3A\u003cem\u003e\u003csup\u003eWT\u003c/sup\u003e\u003c/em\u003e + HDAC3 overexpression in SVG p12\u003cem\u003e\u003csup\u003eATAD3 KD\u003c/sup\u003e\u003c/em\u003e cells,\u0026nbsp;whereas\u0026nbsp;ATAD3A\u003cem\u003e\u003csup\u003eK134Q\u003c/sup\u003e\u003c/em\u003e overexpression or HDAC3-specific inhibitor treatment throughout the entire stage\u0026nbsp;effectively\u0026nbsp;rescued\u0026nbsp;these effects\u0026nbsp;(Fig. 4g, h).\u0026nbsp;SVG p12\u003csup\u003eATAD3\u003c/sup\u003e\u003csup\u003e-\u003c/sup\u003e\u003csup\u003eKD\u003c/sup\u003e cells\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eoverexpressing ATAD3A\u003cem\u003e\u003csup\u003eWT\u003c/sup\u003e\u003c/em\u003e + HDAC3 exhibited extensive mitochondrial fragmentation relative to ATAD3A\u003cem\u003e\u003csup\u003eK134Q\u0026nbsp;\u003c/sup\u003e\u003c/em\u003eoverexpressing\u0026nbsp;cells (Fig. 4i, j).\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eMoreover, overexpression of HDAC3 or HDAC3 coupled with ATAD3A \u003csup\u003eWT\u003c/sup\u003e under RGFP966\u0026nbsp;did not cause\u0026nbsp;obvious mitochondrial fragmentation (Fig. 4i, j).\u0026nbsp;Thus, HDAC3-mediated deacetylation drives astrocytic mitochondrial dysfunction.\u003c/p\u003e\n\u003cp\u003eBlocking astrocytic ATAD3A oligomerization alleviates damage and improves recovery after stroke\u003c/p\u003e\n\u003cp\u003eThe homeostasis of mitochondrial function in astrocytes is crucial for neuronal damage and repair after stroke injury\u003csup\u003e4, 29\u003c/sup\u003e. Aldh1l1\u003csup\u003eCreErt\u003c/sup\u003e; ATAD3A\u003csup\u003efl/\u0026Delta;\u003c/sup\u003e mice with astrocyte-specific heterozygous knockout (or called ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e) showed normal development, memory and motor function (Fig. 5i, j,\u0026nbsp;Extended Data Fig. 7a~d). After tMCAO, there was no difference in the areas with cerebral blood flow reduction between the two genotypes (Extended Data Fig. 7e, f). But ATAD3A \u003csup\u003eAst Het\u003c/sup\u003e had smaller Infarct volume than controls (Fig. 5a~c). This indicated that astrocytic ATAD3A oligomerization signaling abrogation alleviates neuronal injury during the acute phase.\u003c/p\u003e\n\u003cp\u003eThe brain\u0026nbsp;repair\u0026nbsp;process after stroke during the subacute phase (especially at 1~2 weeks)\u003csup\u003e30\u003c/sup\u003e, which can last for 90 days after injury, is reported to be a critical period of rehabilitation. This process is strongly associated with astrocyte homeostasis\u003csup\u003e31\u003c/sup\u003e. Whole-brain MRI scanning was performed at 14 days revealed a reduced infarct edema volume and delayed deterioration in ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e mice (Fig. 5d). Fragmented mitochondria derived from astrocytes are toxic to oligodendrocytes for remyelination\u003csup\u003e32\u003c/sup\u003e. Quantification of Luxol fast blue (LFB) staining revealed that\u0026nbsp;ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e mice\u0026nbsp;presented\u0026nbsp;more myelin coverage after stroke\u0026nbsp;than\u0026nbsp;ATAD3A \u003csup\u003efl/fl\u003c/sup\u003e mice (Extended Data Fig. 7g, h). And directionally encoded color (DEC) maps\u0026nbsp;of the brains\u0026nbsp;showed improved circuit reconstruction (Fig. 5e) and white matter integrity\u0026nbsp;in ATAD3A\u003csup\u003eAst Het\u0026nbsp;\u003c/sup\u003emice (Fig. 5f). And thickness, discontinuity and defects of myelin sheaths in ipsilesional external capsule were less frequently observed in ATAD3A\u003csup\u003eAst Het\u0026nbsp;\u003c/sup\u003emice than in control mice (Extended Data Fig. 7i~k). Behaviorally, ATAD3A\u003csup\u003eAst Het\u0026nbsp;\u003c/sup\u003emice showed better sensorimotor recovery in adhesive removal test (Fig. 5g) and rotarod test (Fig. 5h) from day 14. The white matter (WM)-enriched corpus callosum and external capsule are associated with nerve fiber conduction, especially transcallosal connections to the denervated striatum after stroke\u003csup\u003e33\u003c/sup\u003e, which is responsible for sensorimotor function regulation. Mitochondrial dysfunction in astrocytes is not conducive to WM repair or nerve fiber conduction reconstruction\u003csup\u003e34\u003c/sup\u003e. Immunostaining for MBP (myelin basic protein) and SMI32 (a marker of demyelinated axons) in the external capsule and striatum revealed improved transcallosal connections in ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e mice (Extended Data Fig. 7i, m).\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eEvoked compound action potentials (CAPs) display a biphasic wave with an early peak representing fast-conducting myelinated axons (N1) and a delayed peak representing slow-conducting unmyelinated axons (N2)\u003csup\u003e\u0026nbsp;\u003c/sup\u003e(Extended Data Fig. 7n). Compared with those in ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e mice, impaired\u0026nbsp;myelinated axons are more obvious in ATAD3A \u003csup\u003efl/fl\u003c/sup\u003e mice after stroke\u0026nbsp;and exacerbate\u0026nbsp;the reduction\u0026nbsp;in\u0026nbsp;N1 amplitude. The amplitude of the N2 component\u0026nbsp;did not markedly differ between the two stroke groups\u0026nbsp;(Extended Data Fig. 7o). In addition, spatial recognition memory or spatial learning tests (novel object recognition (Fig. 5i), Morris water maze (Fig. 5j) also indicated better recovery in\u0026nbsp;ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e mice.\u003c/p\u003e\n\u003cp\u003eTo eliminate differences in memory function recovery caused by\u0026nbsp;lesion size variability, bilateral hippocampal photothrombosis was induced before astrocytic ATAD3A knockout (Fig. 6a). TTC staining confirmed confirmed comparable lesion areas between ATAD3A\u003csup\u003efl/fl\u003c/sup\u003e and Aldh1l1\u003csup\u003eCreErt\u003c/sup\u003e; ATAD3A\u003csup\u003efl/\u0026Delta;\u003c/sup\u003e mice (Fig. 6b, c). Tamoxifen-induced astrocytic ATAD3A knockout was followed one week later by fear conditioning to assess hippocampal function. Post stroke contextual fear conditioning results indicated that ATAD3A\u003csup\u003efl/fl\u003c/sup\u003e mice with hippocampal photothrombosis showed no increased freezing time upon re-exposure to the conditioning context, whereas ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e mice exhibited significantly elevated freezing time (Fig. 6d, Supplementary video 1). In addition, the calcium signal intensity of hippocampus\u0026nbsp;in the conditioning context was significantly lower in ATAD3A\u003csup\u003efl/fl\u003c/sup\u003e mice\u0026nbsp;than in ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e mice (Fig. 6e~g). For sensorimotor assessment, photothrombosis was applied to the sensorimotor cortex prior to astrocytic ATAD3A knockout, followed by behavioral and electrophysiological tests (Fig. 6h). The ATAD3A\u003csup\u003efl/fl\u003c/sup\u003e mice displayed higher asymmetry index in the cylinder test, compared with ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e mice (Fig. 6i). After AAV-CaMKII-ChR2-eGFP injection into the contralateral somatosensory area and prior astrocytic ATAD3A knockout, in vivo somatosensory recordings revealed stronger bursting activity in ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e mice post-stimulation (Fig. 6j, k, Supplementary video 2), along with a more pronounced cumulative percentage of evoked responses with rising stimulus power (Fig. 6l, m). Collectively, abrogating astrocytic ATAD3A oligomerization mitigates neuronal injury, improves white matter integrity, and promotes long-term functional recovery after stroke.\u003c/p\u003e\n\u003cp\u003eAstrocytic ATAD3A oligomerization inhibition remodels the inflammatory microenvironment\u003c/p\u003e\n\u003cp\u003eTo address astrocyte heterogeneity directly after astrocytic ATAD3A oligomerization was abolished, scRNA-seq of astrocytes from ATAD3A\u003csup\u003efl/fl\u003c/sup\u003e and ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e mice at day 7 post-stroke identified major brain cell types (Extended Data Fig. 8a). All expected cell types were identified in both sample sets, such as astrocytes (\u003cem\u003eGfap\u003c/em\u003e,\u003cem\u003e\u0026nbsp;Aldh1l1\u003c/em\u003e), microglia (\u003cem\u003eHexb\u003c/em\u003e,\u003cem\u003e\u0026nbsp;Tmem119\u003c/em\u003e), macrophages (\u003cem\u003eMcr1\u003c/em\u003e,\u003cem\u003e\u0026nbsp;Pf4\u003c/em\u003e), oligodendrocytes (\u003cem\u003eOlig1\u003c/em\u003e,\u003cem\u003e\u0026nbsp;Olig2\u003c/em\u003e and\u003cem\u003e\u0026nbsp;Mog\u003c/em\u003e) and their precursors (OPCs), endothelial cells (\u003cem\u003eHemk1\u003c/em\u003e, \u003cem\u003eKcnj13\u003c/em\u003e), pericytes (\u003cem\u003eRgs5\u003c/em\u003e,\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003cem\u003eCspg4\u003c/em\u003e) and stromal cells (for example, T cells and Erythrocytes)\u003csup\u003e16, 17\u003c/sup\u003e (Fig. 7a, Extended Data Fig. 8b). And the percentage of astrocytes decreased notably with increasing percentages of oligodendrocytes, endothelial cells and neurons in ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e mice (Fig. 7b). Subsequently, using GYS1 and GBE1 as glycogen-enriched markers (Extended Data Fig. 1j), astrocytes clustered into five subsets: resting (c1, c2), glycogen-resting, glycogen-reactive and neuroprotective cluster (A2 cluster), and glycogen-reactive and neurotoxic cluster A1 according to scRNA-seq results (Fig. 7c, Extended Data Fig. 8c).\u0026nbsp;Glycogen markers (GYS1, GBE1) were high in the glycogen resting astrocyte cluster, the glycogen reactive astrocyte A2 cluster and the glycogen reactive astrocyte A1 cluster (Fig. 7d). Mapping panreactive gene sets\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eand A1-specific and A2-specific gene sets (Fig. 7e), HDAC3 was enriched in glycogen-reactive A1 astrocytes (Fig. 7h). All distinct astrocyte populations expressed Aqp4, the water channel located on vascular endfeet in all astrocytes in the brain\u003csup\u003e35\u003c/sup\u003e. We mapped distinct astrocytes from regions of olfactory-specific astrocytes (\u003cem\u003eIslr\u003c/em\u003e, \u003cem\u003eIslr2\u003c/em\u003e), cerebellum astrocytes (\u003cem\u003eGdf10\u003c/em\u003e), telencephalon-specific astrocytes (\u003cem\u003eMfge8\u003c/em\u003e, \u003cem\u003eLhx2\u003c/em\u003e), non-telencephalon astrocytes (\u003cem\u003eAgt\u003c/em\u003e (angiotensinogen)) and dorsal midbrain astrocytes (\u003cem\u003eMyoc\u003c/em\u003e) (Extended Data Fig. 8d). We confirmed that all the distinct astrocyte populations were located mainly in telencephalon or non-telencephalon regions during ischemic cortical infarction (Fig. 7f). Immunofluorescence confirmed HDAC3 sequestration and ATAD3A oligomerization in glycogen-laden astrocytes (Fig. 7g).\u003c/p\u003e\n\u003cp\u003ePathway enrichment analysis revealed that the functional characteristics of resting astrocyte c1 cells included \u0026lsquo;generation of precursor metabolites and energy\u0026rsquo;, \u0026lsquo;small-molecule catabolic process\u0026rsquo; and \u0026lsquo;small-molecule biosynthetic process\u0026rsquo;. Resting astrocyte c2 mainly represented \u0026lsquo;cytoplasmic ribosomal proteins\u0026rsquo; and the \u0026lsquo;microglia pathogen phagocytosis pathway\u0026rsquo;. Glycogen-reactive A2 astrocytes mainly represented \u0026lsquo;regulation of synapse structure or activity\u0026rsquo; and \u0026lsquo;neuron projection morphogenesis\u0026rsquo;. Glycogen-reactive A1 astrocytes mainly represented \u0026lsquo;SRP-dependent cotranslational protein targeting to the membrane\u0026rsquo;, \u0026lsquo;cellular responses to stimuli\u0026rsquo;, \u0026lsquo;aggrephagy\u0026rsquo; and other pathways (Extended Data Fig. 8e, f, Supplementary Table 2). Compared with glycogen-reactive A2 astrocytes, glycogen-reactive A1 astrocytes displayed an imbalance in homeostasis and inflammatory stress after stroke (Fig. 7h, i). Using a validated ATAD3A and GYS1 antibodies to confirm their specificity (Extended Data Fig. 8l, m), we stained human brain samples from stroke patients (Supplementary Table 3). Human stroke samples confirmed more glycogen-laden astrocytes (GYS\u003csup\u003e+\u003c/sup\u003e GFAP\u003csup\u003e+\u003c/sup\u003e) (Fig. 7j, k) and ATAD3A aggregates (ATAD3A\u003csup\u003e+\u003c/sup\u003e GFAP\u003csup\u003e+\u003c/sup\u003e) (Fig. 7j, l) in ischemic regions.\u003c/p\u003e\n\u003cp\u003eTranscriptional profiling suggested that abrogation of astrocytic ATAD3A oligomerization is associated with a shift in the ischemic microenvironment, including a reduction in the proportion of astrocytes with an A1-like transcriptional signature (Extended Data Fig. 8g). Unbiased clustering showed a reduction in the astrocyte subpopulation annotated as glycogen-reactive A1 in ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e mice (Fig. 7m). The A1 astrocytic response is directly attributed to fragmented mitochondria and activated microglia/macrophages\u003csup\u003e36, 37\u003c/sup\u003e, which propagate inflammatory reactions and tissue damage in the brain\u003csup\u003e38\u003c/sup\u003e. We extracted primary astrocytes from adult ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e and ATAD3A \u003csup\u003efl/fl\u003c/sup\u003e mice. The results revealed that pDrp1 \u003csup\u003es579\u003c/sup\u003e and pDrp1 \u003csup\u003es600\u003c/sup\u003e are both increased in astrocytes after ischemic stroke and can be rescued after ATAD3A knockout (Fig. 6n). Moreover, macrophages/microglia all displayed an anti-inflammatory gene signature in ATAD3A\u003csup\u003eAst Het\u003c/sup\u003e mice compared with ATAD3A\u003csup\u003efl/fl\u0026nbsp;\u003c/sup\u003emice (Extended Data Fig. 8h). Using neural communication networks to analyze predefined cell groups from scRNA-seq expression data\u003csup\u003e39\u003c/sup\u003e, we found that connections within neuron c1, neuron c2 and neuron c3 and connections between resting astrocyte c2 (or c1) and distinct neurons were all enhanced in ATAD3A\u003csup\u003eAst\u003c/sup\u003e\u003csup\u003e\u0026nbsp;\u003c/sup\u003e\u003csup\u003eHet\u003c/sup\u003e mice (Extended Data Fig. 8i, j). The projection network of each individual interaction pair, gap junction-related gene links (such as \u003cem\u003eGjb2\u003c/em\u003e, \u003cem\u003eGjb6\u003c/em\u003e,\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eand\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003cem\u003eGja1\u003c/em\u003e) and neurosynaptic-related gene pairs (such as \u003cem\u003eNrxn1\u003c/em\u003e, \u003cem\u003eNrxn2\u003c/em\u003e, \u003cem\u003eNrxn3\u003c/em\u003e, and \u003cem\u003eNlgn3\u003c/em\u003e) was more obvious in ATAD3A\u003csup\u003eAst KO\u003c/sup\u003e mice (Extended Data Fig. 8k). Transcriptional profiling suggested that abrogation of astrocytic ATAD3A oligomerization is associated with a shift in the ischemic microenvironment, including a reduction in the proportion of astrocytes with an A1-like transcriptional signature, increasement of anti-inflammatory gene signature and alterations in enhanced ligand-receptor pairs suggestive of modified cell-cell communication.\u003c/p\u003e\n\u003cp\u003eAstrocytic glycogen depletion and HDAC3 inhibition act synergistically to enhance post-stroke recovery\u003c/p\u003e\n\u003cp\u003eConsidering that glycogen depletion agents and HDAC3 inhibitors have drugs developed before clinical trials, we next investigated whether astrocytic glycogen depletion and HDAC3 inhibition can act synergistically to enhance post-stroke recovery. Preclinical studies have shown that cotadutide, a dual glucagon-like peptide 1 (GLP-1R) and glucagon receptor (GCGR) agonist, can effectively act on the brain and peripheral organs\u003csup\u003e40\u003c/sup\u003e, reducing glycogen accumulation, inflammatory stress and fibrosis in some chronic diseases\u003csup\u003e41-43\u003c/sup\u003e. We next explored the changes in glycogen deposition in the brain of ischemic stroke under different drug treatment groups. RGFP966 alone treatment cannot reduce the deposition of glycogen in the brain, while adding cotadutide can effectively deplete the glycogen deposition (Extended Data Fig. 9a, b). Next, astrocyte-specific protein (Ast ER-BioID \u003csup\u003eHA\u003c/sup\u003e)\u0026nbsp;mice\u0026nbsp;were used\u0026nbsp;to label biotinylated proteins derived from astrocytes for global protein oxidation levels and ATAD3A oligomers detection (Extended Data Fig. 9c). Astrocytes lost the expression of EGFP and expressed ER-BioID2 HA after Aldh1l1\u003csup\u003e-CreERT\u003c/sup\u003e;\u0026nbsp;ER-BioID \u003csup\u003eHA\u003c/sup\u003e mice were treated with tamoxifen (Extended Data Fig. 9d). These\u0026nbsp;Ast ER-BioID \u003csup\u003eHA\u003c/sup\u003e mice were treated with\u0026nbsp;RGFP966, cotadutide,\u0026nbsp;or\u0026nbsp;cotadutide combined with RGFP966 (CotR)\u0026nbsp;after ischemic stroke. Biotinylated\u0026nbsp;proteins were enriched and obtained\u0026nbsp;for\u0026nbsp;global protein oxidation levels and ATAD3A oligomers detection\u0026nbsp;(Extended Data Fig. 9e). Treatment with either RGFP966 or cotadutide alone reduced global protein oxidation and ATAD3A oligomerization; these reductions were more pronounced with CotR combination therapy (Extended Data Fig. 9f~h). Therefore, cotadutide alone can effectively reduce the deposition of glycogen granules in the brain, it cannot completely reduce the oxidation level and ATAD3A oligomerization. When combination therapy, it can both reduce glycogen deposition and fully block ATAD3A oligomerization signals.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFollowing sensory cortex ischemic stroke, mice received systemic CotR therapy, with or without pre-overexpression of oligomer-prone ATAD3A (AAV-ATAD3A\u003csup\u003eK134E\u003c/sup\u003e) (Extended Data Fig. 10a, b). Pre-overexpression of ATAD3A\u003csup\u003eK134E\u003c/sup\u003e exacerbated pathological damage and mildly impaired motor recovery, increasing foot faults and asymmetry indices compared to controls (Extended Data Fig. 10c, d). It also consistently reduced synaptic spine density in the sensorimotor cortex, regardless of stroke (Extended Data Fig. 10e, f). Although daily CotR treatment markedly reduced the asymmetry index, this protective effect was counteracted by ATAD3A\u003csup\u003eK134E\u003c/sup\u003e pre-overexpression (Fig. 8a). Motor network excitability of\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003ethe\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eipsilateral ischemic cortex during the repair phase is related to motor functional recovery during the repair phase of stroke\u003csup\u003e2, 44\u003c/sup\u003e. CotR also restored stroke-impaired cortical excitability, as measured by Ca\u0026sup2;⁺ ultrasensitive intracellular Ca\u003csup\u003e2+\u003c/sup\u003e sensor for pyramidal neurons on day 14, an effect partially reversed by ATAD3A\u003csup\u003eK134E\u003c/sup\u003e pre-overexpression (Fig. 8b~c, Extended Data Fig. 10g), as shown by the locomotion heatmap results (Extended Data Fig. 10h).\u0026nbsp;Whole-brain MRI scanning was subsequently performed after stroke. The results revealed a reduced infarct edema volume and augmented FA value after CotR therapy, which were counteracted by ATAD3A\u003csup\u003eK134E\u0026nbsp;\u003c/sup\u003epre-overexpression (Extended Data Fig. 10i, j).\u003c/p\u003e\n\u003cp\u003eSynaptogenesis and synaptic transmission in the peri-infarct cortex are critical for circuit remapping and neural network plasticity\u003csup\u003e30, 45, 46\u003c/sup\u003e. On day 16, CotR rescued the stroke-induced reductions in dendrite length and spine density in the peri-infarct cortex, protective effects again attenuated by ATAD3A\u003csup\u003eK134E\u003c/sup\u003e overexpression (Extended Data Fig. 10k~m). Excitatory synaptic transmission was usually used to assess functional plasticity. Electrophysiological recordings on day 14 also confirmed that CotR reversed the stroke-induced decrease in mEPSC frequency, indicating enhanced neural functional plasticity in the peri-infarct region (Extended Data Fig. 10n~p).\u003c/p\u003e\n\u003cp\u003eAxonal sprouting involves short-distance axonal sprouting from the uninjured peri-infarct region to the ischemic penumbra (Extended Data Fig. 10k)\u003csup\u003e46, 47\u003c/sup\u003e and long-distance axonal sprouting from uninjured corticospinal axons to denervated spinal cord and brain nuclei (such as the red nucleus and facial nucleus) (Extended Data Fig. 10q)\u003csup\u003e1, 48\u003c/sup\u003e, which are essential for post-stroke circuit remodeling. By BDA tracing, neural fibers from uninjured circuits crossed obviously the denervated region under CotR intervention. However, this effect was similarly blunted by ATAD3A \u003csup\u003eK134E\u003c/sup\u003e pre-overexpression (Extended Data Fig. 10r, s).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eProjection neurons in the cortex, which reside primarily in layers II/III, send axons to distant\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003econtralateral brain targets\u003csup\u003e49, 50\u003c/sup\u003e. The transcallosal connection is responsible for extending axons to mirror-image locations in the contralateral functional area, enabling information intercommunication and integration. The transcallosal connection contributes to the recovery process of sensorimotor function\u003csup\u003e51, 52\u003c/sup\u003e. After a focal stroke, the recovery of sensory responsiveness in the lesion periphery is mediated by interhemispheric synaptic inhibition, which is selectively activated by contralateral excitatory inputs\u003csup\u003e51\u003c/sup\u003e. We injected AAV-CaMK II-ChR2-eGFP in the contralateral mirror-image somatosensory area, and an\u0026nbsp;\u003cem\u003ein vivo\u003c/em\u003e multichannel electrophysiological electrode was implanted in the peri-infarct somatosensory cortex on the day of stroke (Fig. 8d). 40-Hz optogenetic stimulation was performed 15 days after surgery and drug intervention. The behavior of the mice revealed that the transcortical corpus callosum connection was initially established when light stimulation was given to the opposite undamaged side (Supplementary video 3). CotR enhanced transcallosal connectivity, evidenced by greater optogenetically evoked bursting activity and response rates in the somatosensory cortex (Fig. 8e~h). So, pharmacologic exhaustion of astrocytic glycogen and HDAC3 inhibition enhances recovery after stroke. To eliminate the off-target effects and specificity issue of drugs, astrocyte-targeted viruses were used. Overexpression of glycogen-degrading enzyme PYGB effectively reduced glycogen deposits, while HDAC3 knockdown did not (Extended Data Fig. 11a~e). The stroke mice with astrocytic glycogen depletion or HDAC3 knockdown displayed significantly reduced foot faults in grid-walking test and asymmetry index in the cylinder test, compared with vehicle-treated mice. And their combination acted synergistically to produce the greatest functional recovery (Fig. 8i), consistent with greater optogenetically evoked bursting activity and response rates in the somatosensory cortex, compared with single treatment (Fig. 8j~l).\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eAstrocytes support brain homeostasis by providing energy substrates and shuttle mitochondria to neurons\u003csup\u003e4, 5, 29\u003c/sup\u003e. Large insoluble glycogen aggregates are associated with neuronal dysfunction, aging, and cognitive decline\u003csup\u003e10, 53\u003c/sup\u003e. We found that glycogen-laden astrocytes in the ischemic penumbra exacerbate stroke injury and impair recovery. Astrocytic glycogen aggregates sequester cytoplasmic HDAC3, enabling its translocation to mitochondria. There, HDAC3 deacetylates outer mitochondrial membrane protein ATAD3A, promoting oligomerization-driven mitochondrial fragmentation. \u0026nbsp;Exhaustion of astrocytic glycogen and HDAC3 inhibition reverse glycogen accumulation, rescue mitochondrial architecture/function, and restore synaptic plasticity and circuit reorganization, thereby acting synergistically to enhance post-stroke recovery (Fig. 8m).\u003c/p\u003e\n\u003cp\u003eGlycogen granules form as electron-dense, organelle-like structures adjacent to the endoplasmic reticulum and mitochondria\u003csup\u003e12\u003c/sup\u003e. Under ischemic stress, dysregulated glycogen metabolism-marked by increased glycogenesis and reduced glycogenolysis-promotes the accumulation of insoluble, Stbd1-positive glycogen granules\u003csup\u003e12\u003c/sup\u003e\u003csup\u003e,\u003c/sup\u003e\u003csup\u003e54\u003c/sup\u003e, consistent with prior reports\u003csup\u003e5, 55\u003c/sup\u003e. While soluble glycogen supports neuronal energetics via lactate shuttling\u003csup\u003e3\u003c/sup\u003e, the role of insoluble aggregates has been obscure. Astrocytic mitochondria provide most of the energy for metabolism, and their homeostasis is critical for stroke recovery\u003csup\u003e4, 56\u003c/sup\u003e. Previous studies have reported that oxidative phosphorylation, mitochondrial metabolism, and mitochondrial ultrastructure are all impaired in cells \u003cem\u003ein vitro\u003c/em\u003e and \u003cem\u003ein vivo\u003c/em\u003e models of glycogen storage disease. The mitochondrial content also decreases, which is coupled with decreased mitochondrial biogenesis and increased mitochondrial fragmentation\u003csup\u003e57, 58\u003c/sup\u003e. These results were clearly observed in glycogen stress granule-enriched astrocytes in this study. However, the intricate relationship between mitochondrial dysfunction and glycogen storage has not been fully elucidated. A previous study revealed that Stbd1 can recruit glycogen to endoplasmic reticulum (ER)-mitochondria contact sites\u003csup\u003e59\u003c/sup\u003e, which may\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003ebe correlated with mitochondrial morphology and mitochondrial dynamics, such as mitochondrial fusion (elongation) and fission (fragmentation)\u003csup\u003e60\u003c/sup\u003e. In addition, fragmented mitochondria are more obvious under nutrient oversupply conditions\u003csup\u003e60, 61\u003c/sup\u003e. Given that glycogen itself has the \u0026ldquo;sweet\u0026rdquo; side of ER‒mitochondria contact sites, we propose that glycogen overload may result in mitochondrial fragmentation. This hypothesis was also mentioned by Demetriadou \u003cem\u003eet al\u003c/em\u003e. but lacks experimental evidence\u003csup\u003e60\u003c/sup\u003e. Our proteomic and functional analyses reveal that glycogen stress granules recruit mitochondrial proteins and localize to ER\u0026ndash;mitochondria contact sites, directly implicating them in mitochondrial dysfunction.\u003c/p\u003e\n\u003cp\u003eNotably, we identified the glycogen-anchored protein Epm2aip as a physical linker between glycogen granules and ATAD3A-a mitochondrial fission receptor situated at ER-mitochondria junctions. And Epm2a (Laforin) also participates in the formation of insoluble and neurotoxic glycogen aggregates and is enriched in glycogen condensates\u003csup\u003e9, 13\u003c/sup\u003e. ATAD3A is located at ER‒mitochondrion contact sites\u003csup\u003e19, 24\u003c/sup\u003e, and its deacetylation triggers self-oligomerization\u003csup\u003e20\u003c/sup\u003e. The CC domain of ATAD3A, an inhibitory segment that maintains the steady state, is key for mitochondrial fission and mtDNA stability. Deacetylation of the CC domain at residue K135 (K134 in mice) is essential for signaling-induced self-oligomerization\u003csup\u003e20\u003c/sup\u003e. We now establish HDAC3 as the deacetylase responsible for this modification. Under glycogen overload, cytoplasmic HDAC3 is recruited to glycogen granules and recruited to mitochondria, where it deacetylates ATAD3A, driving oligomerization mediated fragmentation. This pathway represents a molecular bridge between glycogen metabolism and mitochondrial dynamics, substantiating the hypothesis that glycogen overload disrupts mitochondrial integrity.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eATAD3A oligomerization, a terminal pathological signaling after glycogen accumulation, is specifically genetically knocked down in astrocytes before stroke. Indeed, our results demonstrated that a reduction in ATAD3A oligomerization alleviates acute damage and improves recovery after stroke. Astrocytic mitochondrial dysfunction is not conducive to the survival of adjacent neuronal cells, especially during the course of stroke\u003csup\u003e56\u003c/sup\u003e. Electron microscopy revealed that\u003cstrong\u003e\u0026nbsp;morphologically damaged mitochondria were more apparent in non-ATAD3A\u003c/strong\u003e-knockdown mice than in control mice. Previous studies have demonstrated that ATAD3A forms higher-order oligomers and acts as a molecular linker coupling Drp1-mediated mitochondrial fragmentation\u003csup\u003e20, 21\u003c/sup\u003e. The abolition of ATAD3A oligomerization helps delay the progression of neurodegenerative diseases such as Alzheimer\u0026rsquo;s disease and Huntington\u0026rsquo;s disease. Therefore, intervening in the ATAD3A oligomerization pathway may be a potential therapeutic target for stroke.\u003c/p\u003e\n\u003cp\u003eTo inhibit the ATAD3A oligomerization pathway after stroke, we aimed to accelerate the exhaustion of astrocytic glycogen by blocking the downstream signaling cascade with an HDAC3-specific inhibitor. However, although cotadutide involves in enhancing glycogenolysis and glycogen depletion, dynamic changes in astrocytic glycogen after stroke \u003cem\u003ein vivo\u003c/em\u003e cannot be well monitored because of a lack of technical strategies\u003csup\u003e43\u003c/sup\u003e. The mitochondrial biogenesis regulator Pgc-1\u0026alpha; is well known downstream of GCG, cotadutide can increase Pgc-1\u0026alpha; expression and restore the basal and maximal respiratory rates of damaged mitochondria via the GCGR. HDAC3 has been found to regulate several oxidative stress-related processes and molecules through its deacetylase and nonenzymatic activities. HDAC3 inhibitors are widely used in many inflammatory diseases. Therefore, the ability of cotadutide or RGFP966 to improve mitochondrial function and resolve inflammation may be attributed to multiple pathways.\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003eThis approach, leveraging clinically developed compounds, holds significant translational potential for stroke treatment.\u003c/p\u003e\n\u003cp\u003eSeveral limitations warrant consideration. Although we identified key signaling components, other glycogen-mediated pathways involving metabolism, apoptosis, or epigenetics cannot be excluded. The dynamic interplay between glycogen and mitochondria in vivo remains challenging to track due to a lack of real-time glycogen probes. Furthermore, differences between murine ATAD3A and human ATAD3 paralogs necessitate caution in extrapolating findings. Finally, while cotadutide and RGFP966 showed synergistic efficacy, their pleiotropic effects suggest involvement of additional mechanisms beyond the pathway described here.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMATERIALS \u0026amp; CORRESPONDENCE\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFurther information and requests for resources, reagents and codes should be directed to and will be fulfilled by the lead contact, Dr. Hongrui Zhu ([email protected]).\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eHuman samples\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe human brain tissue was provided by the National Health and Disease Human Brain Tissue Resource Center (Ethical number: S2024052). The detailed information of the human brain samples is presented in Supplementary Table\u0026nbsp;3.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAnimals\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll the animal experiments were approved by the Institutional Animal Care and Use\u0026nbsp;Committee at the University of Science and Technology of China (Ethical number: No. 2022-N(A)-175).\u0026nbsp;The mice were housed under a 12:12-h light‒dark cycle with ad libitum access to food and water. \u003cem\u003eATAD3A\u003c/em\u003e \u003cem\u003e\u003csup\u003efl/fl\u003c/sup\u003e\u003c/em\u003e (stock no: NM-CKO-2102515) and G6PT\u003csup\u003e-/-\u003c/sup\u003e mice were obtained from Shanghai Model Organisms Center, Inc. \u003cem\u003eAldh1l1\u003csup\u003e-CreERT\u003c/sup\u003e\u003c/em\u003e (stock no: C001288) mice were purchased from Cyagen Biosciences, Inc. \u003cem\u003eAldh1l1\u003csup\u003e-CreERT\u003c/sup\u003e\u0026nbsp;\u003c/em\u003emice were crossed with \u003cem\u003eATAD3A\u003csup\u003efl/fl\u003c/sup\u003e\u003c/em\u003e mice to generate \u003cem\u003eAldh1l1\u003c/em\u003e\u003cem\u003e\u003csup\u003e-CreERT\u003c/sup\u003e\u003c/em\u003e mice. To obtain astrocytic \u003cem\u003eATAD3A\u0026nbsp;\u003c/em\u003eheterozygous knockout mice (ATAD3A \u003cem\u003e\u003csup\u003eAst Het\u003c/sup\u003e\u003c/em\u003e), male \u003cem\u003eAldh1l1\u003csup\u003e-CreERT\u003c/sup\u003e\u003c/em\u003e; \u003cem\u003eATAD3A\u003csup\u003efl/\u0026Delta;\u0026nbsp;\u003c/sup\u003e\u003c/em\u003emice aged six weeks were subjected to five daily intraperitoneal\u0026nbsp;injections of 20 mg/kg tamoxifen (TAM, dissolved in corn oil) to\u0026nbsp;generate\u0026nbsp;ATAD3A\u003csup\u003eAst\u0026nbsp;\u003c/sup\u003e\u003csup\u003eHet\u003c/sup\u003e and\u0026nbsp;control male littermates under corn oil treatment (ATAD3A \u003csup\u003eAst WT\u003c/sup\u003e). Astrocyte-specific secretion protein mice (\u003cem\u003eAldh1l1\u003csup\u003e-CreERT\u003c/sup\u003e\u003c/em\u003e; ER-BioID \u003csup\u003eHA\u003c/sup\u003e) were generated by\u0026nbsp;crossing\u0026nbsp;\u003cem\u003eAldh1l1\u003csup\u003e-CreERT\u003c/sup\u003e\u0026nbsp;\u003c/em\u003emice with C57BL/6-Tg (CAG-EGFP, -birA*)1Fink/J mice (Jackson\u0026nbsp;Laboratory,\u0026nbsp;Stock No: 036203). \u003cem\u003eAldh1l1\u003csup\u003e-CreERT\u003c/sup\u003e\u003c/em\u003e; ER-BioID \u003csup\u003eHA\u003c/sup\u003e mice aged six weeks were subjected to five daily intraperitoneal\u0026nbsp;injections of\u0026nbsp;20 mg/kg tamoxifen (TAM, dissolved in corn oil) to\u0026nbsp;generate\u0026nbsp;Ast ER-BioID \u003csup\u003eHA\u003c/sup\u003e mice.\u0026nbsp;Astrocyte specific diphtheria toxin A (DTA) expression-based mouse strain was established by Aldh1l1-CreERT mice crossed with ROSA26iDTR (Jackson laboratory, Stock No: 007900). ROSA26iDTR; Aldh1l1-CreERT mice aged at six weeks were pretreated with administrated with 20 mg/kg tamoxifen for 5 consecutive days to induce Cre-inducible expression of DTR. 100 ng diphtheria toxin (DT, Aladdin,\u0026nbsp;Cat# D684675-1mg) in 1 \u0026times; PBS was injected intraperitoneally 3x daily for 10 d to obtain astrocyte specific ablation mouse.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCell lines\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSVG p12 cells (ATCC, CRL-8621) were cultured in Eagle\u0026apos;s minimum essential medium (ATCC, 30\u0026ndash;2003) supplemented with 10% fetal bovine serum\u0026nbsp;and 100 U/ml penicillin/streptomycin. HeLa cells (ATCC, CRM-CCL-2) and HEK293 cells (Procell, CL-0005) were cultured in Dulbecco\u0026rsquo;s modified Eagle\u0026rsquo;s medium (Servicebio, G4511) with the same supplements. These cell lines were cultured in a humidified incubator with\u0026nbsp;a\u0026nbsp;5% CO\u003csub\u003e2\u003c/sub\u003e atmosphere at 37\u0026deg;C.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStroke model\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFocal cortical stroke was induced via the photothrombotic (PTS) method. After anesthesia, mice were fixed in a stereotaxic apparatus (RWD Life Science), and a cranial window was created at the target coordinates relative to the bregma. The mice were then injected intraperitoneally with rose bengal at a dose of 100 mg/kg (Sigma, 330000). After 10 min, the light outlet was placed in direct contact with the skull removal site to transmit light from a cold light source (B5200, POFC-S6H-1000-F1) for 20 min. Following light exposure, the scalp was sutured, and the mice were placed on a warming pad to recover. Focal hippocampal stroke was induced in mice using the photothrombotic method. The experimental procedure was as follows: anesthetized mice were secured in a stereotaxic apparatus, the skull was exposed, and a fiber-optic cannula (400 \u0026mu;m core diameter, MFC_400/430-0.48_2.5 mm_MF2.5_FLT, Doric Lenses) was implanted at coordinates relative to bregma (Anteroposterior: 0 mm, Mediolateral: \u0026plusmn; 2 mm, Dorsoventral: - 1.7 mm). Ten minutes after intraperitoneal injection of rose bengal (100 mg/kg), the hippocampal region was illuminated for 10 minutes at 6 mW using a photothrombosis induction laser (561 nm, LRS-0561-GFO-00100-03, LaserGlow Technologies) connected to the implanted fiber. This method relies on the interaction between the photosensitizer and laser light at a specific wavelength to induce focal thrombosis in the hippocampal area adjacent to the fiber tip. Focal ischemia\u0026nbsp;was induced by occlusion of the middle cerebral artery (MCA). After being anesthetized, the mice were\u0026nbsp;placed in a supine position under a microscope, and a longitudinal incision was made along the midline of the neck. The common carotid artery was dissected, with the proximal end ligated and the distal end occluded at the bifurcation of the common carotid artery. The internal carotid artery was clamped via vascular clamps. An incision was made between the proximal end and the clamp, and a silicone-coated filament (Beijing Cinontecech, 1622A4) was introduced. The clamp was released to facilitate filament insertion. After 1 h, the filament was removed, and the incision was sutured. The mice were then placed on a warming pad to facilitate recovery.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eParabiosis\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003esurgery\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAfter being anesthetized, the skin and muscle of the donor mouse are removed to create a muscle flap. This flap is subsequently sutured onto the muscle of the recipient mouse, thereby facilitating the establishment of a shared circulatory system between the two organisms. On the 10th day, methylene blue was injected into the donor mouse, and urine from the recipient mouse was collected 2 h postinjection to analyze the physiological responses influenced by the donor. All animal surgeries were conducted in accordance with humane care policies and were approved.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLaser speckle imaging\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eUnder anesthesia, laser speckle contrast imaging (RWD, RFSLI-ZW) was employed to monitor cortical blood flow changes before and after the induction of photothrombotic and middle cerebral artery occlusion (MCAO) infarction models in mice. Data analysis was performed via the CUDA Toolkit 10.2 software, following the experimental procedures outlined in the instrument manual.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMagnetic resonance imaging and\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eanalysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor the magnetic resonance imaging (MRI) experiments, the mice were anesthetized and immobilized, and vital signs were monitored in real time. MRI scans were performed via a 9.4T uMR (United Imaging Life Science Instrument, Wuhan, China) within 14 days poststroke. DTI and T2-weighted images were obtained via the following parameters: (1) DTI: TR/TE = 5728/28 ms; acquisition matrix = 148\u0026times;192; field of view (FOV) = 15\u0026times;19 mm; slices = 30; slice thickness = 0.5 mm; number of excitations (NEX) = 4; b-values = 0, 1000, 2000 s/mm\u0026sup2; (for the x, y, and z directions); and 5 averages. (2) T2W-MRI: TR/TE = 3000/49 ms; acquisition matrix = 331\u0026times;368; FOV = 18\u0026times;20 mm; slices = 20; slice thickness = 0.5 mm; rare factor = 13.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDrug administration\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor the treatment with cotadutide (Aladdin, 1686108-82-6), mice were administered continuous subcutaneous (s.c.) injections of cotadutide at a dosage of 10 nmol/kg/day beginning three days poststroke and continuing until tissue collection. For the HDAC3 inhibitor RGFP966 (Selleck, S7229), the mice received continuous intraperitoneal (i.p.) injections of 10 mg/kg/day starting three days after stroke and lasting until tissue collection. The sham group was given continuous subcutaneous injections of saline.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWestern blotting\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe cells were lysed in SDS lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 0.1% SDS, 0.5% sodium deoxycholate, and 1% Triton X-100, pH 7.4) supplemented with a protease and phosphatase inhibitor cocktail (Abcam, ab201119) at 4\u0026deg;C for 10 min. The lysate was centrifuged at 12,000 \u0026times; g for 15 min at 4\u0026deg;C, and the supernatant was collected and quantified via a BCA assay kit (Biosharp, BL521A). Equal amounts of proteins were then subjected to SDS‒PAGE and transferred to PVDF membranes. After blocking with 5% nonfat milk at room temperature (RT) for 1 h, the blots were probed by overnight incubation with the primary antibody at 4\u0026deg;C. The bound antibodies were detected with the secondary antibody and visualized via enhanced chemiluminescence (ECL) substrates (Biosharp, BL520A) on a Tanon 5200 imaging system. Detailed information about the antibodies is provided in Supplementary Table 4.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCo\u003c/strong\u003e\u003cstrong\u003eimmunoprecipitation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe cells or tissue samples were lysed in IP buffer (50 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, 1 mM sodium vanadate, 1 M EGTA, 2.5 mM sodium pyrophosphate, and 1% Triton X-100, pH 7.4) at 4\u0026deg;C for 10 min. After centrifugation, a portion of the supernatant was retained as input, and the remainder was incubated with primary antibody at 4\u0026deg;C for 3 h. Then, protein G agarose beads (Yeasen, 36403ES03) were added and incubated at 4\u0026deg;C for 12 h. After washing with cold IP buffer, the mixture was centrifuged at 12,000 \u0026times; g for 1 min, and the pellets were boiled in SDS sample buffer and analyzed via western blotting.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePlasmids and transfection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eATAD3A, Drp1, Epm2aip, GYS1, HDAC1, HDAC2, HDAC3, HDAC6, HDAC8, SIRT1, SIRT3, SIRT4, SIRT5, SIRT6, SIRT7, CLOCK, KAT1, KAT2A, KAT2B, KAT5, and KAT8 were subsequently cloned and inserted into the pCMV-Myc, pEGFP-N1, pCMV-His, and pCMV-Flag vectors. Truncated mutants of ATAD3A were constructed via overlap extension cloning techniques. The ATAD3A mutants K134E and K135Q, as well as the HDAC3 mutants Y298F and H133/H134A, were constructed via the Mut Express II Fast Mutagenesis Kit (Vazyme, C214-01). The tdTomato-tagged wild-type STBD1 constructs were constructed by subcloning the STBD1 fragments into pLV3-CMV-MCS-tdTomato (Miaolingbio, P50400). All plasmid constructs were confirmed through sequencing. Transfections of these plasmids were conducted via polyethyleneimine (Polysciences, 24765-1, 24765-100 (MW 40,000)) following the manufacturer\u0026apos;s instructions. Detailed information about the plasmids is provided in Supplementary Table 4.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGeneration of HDAC3\u003c/strong\u003e\u003cstrong\u003e-\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;or HDAC6\u003c/strong\u003e\u003cstrong\u003e-\u003c/strong\u003e\u003cstrong\u003eknockout cell lines and ATAD3A-knockdown cell lines\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo effectively deplete endogenous HDAC3, HDAC6, and ATAD3A in HEK293 cells, lentiviral vectors (GeneCopoeia, SG002) were used to deliver guide RNAs specifically targeting human HDAC3 (5\u0026prime;-GCGATGTGGGCAACTTCCAC-3\u0026prime;) and human HDAC6 (5\u0026prime;-GCCGGCCAAGATTCTTCTACT-3\u0026prime;). The single guide RNA (sgRNA) sequences were selected via Synthego sgRNA Designer (https://design.synthego.com) and subsequently incorporated into the lentiCRISPR vector through BsmBI-v2 (NEB, R0580) digestion. Knockdown of ATAD3A in SVG p12 cells was achieved by infecting the cells with the pLKO.1 lentivirus (Sigma, SHC001), which carries a guide RNA targeting human ATAD3A (5\u0026prime;-TGGACCATCTCATTGATGCGGT-3\u0026prime;). The lentivirus was produced by cotransfecting the lentiviral expression vector along with packaging and envelope plasmids, including psPAX2 and pMD2, into HEK293 cells via polyethyleneimine. The efficiency of knockout or knockdown was confirmed through Western blot analysis and Sanger sequencing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2,3,5-Triphenyl tetrazolium chloride\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;staining\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAfter the mice were anesthetized, they were perfused with PBS, and the brains were extracted. The brains were then rapidly frozen at -20\u0026deg;C for 20 min before sectioning. Coronal sections were continuously cut at 1 mm intervals and placed in 2,3,5-triphenyl tetrazolium chloride\u0026nbsp;(TTC, Sigma, USA) at 37\u0026deg;C for 30 min. The infarct area was quantified via\u0026nbsp;ImageJ software.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTransmission electron\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003emicroscopy\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor transmission electron microscopy analysis, samples were fixed in 2.5% glutaraldehyde at 4\u0026deg;C for 24 h, followed by fixation with 2% osmium tetroxide. They were then immersed in 2% uranyl acetate at RT for 2 h and dehydrated in a gradient of ethanol (30% to 100%). After being embedded in resin, ultrathin sections were stained with lead citrate and uranyl acetate. Images were acquired via a transmission electron microscope (Tecnai G2 F20; 200 kV; FEI).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eImmunostaining and imaging\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAfter the mice were anesthetized, they were perfused with PBS followed by 4% PFA. The brains were extracted and fixed in 4% PFA for 6‒8 h and then dehydrated in 20% and 30% sucrose solutions. Coronal sections (40 \u0026micro;m) were cut via a vibratome (Leica CM1950). The sections were air dried at 50\u0026deg;C for 1 h, followed by antigen retrieval in sodium citrate at 60\u0026deg;C. After permeabilization with 0.3% Triton X-100 and blocking with 3% BSA for 40 min, the sections were incubated overnight with primary antibodies at 4\u0026deg;C. Following three washes with PBS, secondary antibodies were applied at RT for 2 h. After another three washes with PBS, the sections were stained with DAPI for 10 min at RT. Images were acquired using a Leica STED confocal microscope.\u0026nbsp;Detailed information about the antibodies is provided in Supplementary\u0026nbsp;Table 4.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eThree-dimensional (3D) reconstruction\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe 40-\u0026mu;m- thick brain sections were co-stained with primary antibodies targeting the astrocytic marker GFAP and the ATAD3A protein, followed by fluorescence labeling with corresponding secondary antibodies. Image acquisition was performed using a near-infrared continuous-spectrum single-photon confocal system (Leica STED) microscope\u0026nbsp;and imaging parameters (laser power, gain, and offset) were consistent across all experiments. Z stacking was performed with 1.0- \u0026mu;m steps in the Z direction, and 1024 \u0026times; 1024-pixel resolution images were analyzed using IMARIS 9.6.2 software (Bitplane). Astrocytes were reconstructed in 3D using the IMARIS \u0026ldquo;Surface\u0026rdquo; function applied to GFAP signal, with consistent thresholds applied across all samples. Punctate signals for ATAD3A, GYS1, GBE1, and STBD1 were identified using the IMARIS \u0026ldquo;Spots\u0026rdquo; function. Colocalization was quantified by calculating the number of respective fluorescent puncta (ATAD3A; GYS1 and STBD1; GBE1 and STBD1) located within each GFAP-positive astrocytic surface, defined by a distance \u0026le; 0 \u0026mu;m using the \u0026ldquo;Split into Surface Objects\u0026rdquo; function. These counts served as the metrics for colocalization analysis between GFAP and the respective proteins.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eImmunohistochemistry\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe mice were anesthetized and perfused with 4% PFA. The tissues were fixed in 4% PFA for 8 h. The brains were embedded in paraffin. The brain tissue was sectioned to a thickness of 10 \u0026micro;m along the coronal plane via a paraffin microtome (Leica BIOCUT). The samples were sequentially processed through the following solutions: xylene, 100% ethanol, 90% ethanol, 80% ethanol, 75% ethanol, and 3% hydrogen peroxide. The samples were then blocked with 3% horse serum at RT for 10 min and incubated for 12 h at 4\u0026deg;C with anti-ATAD3A (Abnova, H00055210-D01). After being washed with TBST, the tissue sections were treated with biotinylated anti-rabbit IgG (Servicebio, G1213) at RT for 40 min. Immunohistochemical localization was performed via a DAB color development kit (Servicebio, G1212) according to the manufacturer\u0026apos;s instructions. The images were then examined under a ZEISS Axio Imager M2 microscope.\u0026nbsp;Detailed information about the antibodies is provided in Supplementary\u0026nbsp;Table 4.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMultiplex immunofluorescence staining\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMultiplex immunofluorescence staining of mouse brain frozen sections was conducted using the multiplex fluorescent immunohistochemistry kit (FMS-mIHC004, FCMACS) following the manufacturer\u0026apos;s protocol. Briefly, sections were dried at 55\u0026deg;C for 30 min, followed by three washes with PBS. Subsequently, they were incubated in a decoloring and antigen repair buffer (Reagent A, 1:20 in deionized water). The sections were then heated to boiling in a microwave oven and maintained at approximately 95\u0026deg;C for 30 min. After cooling to room temperature, sections were permeabilized with 0.5% Triton X-100 for 20 min and then incubated with Reagent B to block endogenous peroxidase activity for 15 min. Subsequently, the staining was accomplished through four sequential labeling cycles. Each cycle consisted of sequential incubations with a specific primary antibody-dilutions were as follows: ATAD3A (1:500), HDAC3 (1:250), GFAP (1:4000), and STBD1 (1:1000) at 4\u0026deg;C overnight, followed by an HRP-conjugated polymer secondary antibody (Reagent C), the corresponding TSA fluorescent dye (D-488 for ATAD3A, D-647 for HDAC3, D-750 for GFAP, and D-594 for STBD1), and finally a microwave-based HRP inactivation step to enable subsequent cycling. Following the completion of all cycles, sections were counterstained with DAPI-containing mounting medium (Reagent G) and imaged using an Olympus FV4000 laser scanning confocal microscope.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eContextual fear conditioning\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA standard two-day contextual fear conditioning paradigm was employed to assess the long-term recovery of hippocampal memory function one week after stroke. All experimental mice were transferred to the behavioral testing room at least 1 h prior to testing to acclimatize to the environment and minimize stress. On day 7 post-stroke (bilateral hippocampal stroke), mice were placed in a fear conditioning apparatus (Labmaze Conditional Fear Video Analysis System, ZS-KJ) constructed of plexiglass with a metal shock grid floor. The training session was conducted over a 7 min period according to the following protocol: The initial 1 min period allowed mice to freely explore the novel environment without any stimulation. Subsequently, the system automatically delivered three-foot shocks (0.7 mA, 2s duration) at programmed intervals. The software automatically recorded behavioral responses before and after each shock, along with baseline activity levels throughout the session. On day 8 post-stroke (bilateral hippocampal stroke), mice were returned to the identical apparatus used during the previous day\u0026rsquo;s conditioning session. During this 7 min test session, no foot shocks were administered. The analysis software primarily tracked and recorded freezing behavior, defined as complete immobility except for respiration movements.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIn situ proximity ligation assay (PLA)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFollowing standard perfusion and fixation, mouse brains were cryoprotected in sucrose gradients and sectioned coronally at 40 \u0026mu;m thickness using a vibratome. For PLA, sections were washed three times after air-drying at 50\u0026deg;C. After permeabilization with 0.3% Triton X-100, sections were blocked with 3% bovine serum albumin (BSA) for 40 min at room temperature and incubated with primary antibodies against ATAD3A, HDAC3, and GFAP overnight at 4\u0026deg;C. PLA probes (PLUS and MINUS) were diluted 1:5 in antibody diluent and pre-incubated for 20 min at room temperature. After washing with Duolink Wash Buffer A (2\u0026times;5 min), the probe mixture was applied and incubated for 1 h at 37\u0026deg;C. Sections were then incubated with ligation solution containing ligase (1:40 dilution) for 30 min at 37\u0026deg;C, followed by amplification with polymerase (1:80 dilution) for 100 min at 37\u0026deg;C in the dark. Final washes were performed with Duolink Wash Buffer B, and nuclei were counterstained with DAPI. Imaging was conducted using a high-resolution confocal microscope (Leica STED).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePeriodic\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eacid\u003c/strong\u003e\u003cstrong\u003e-Schiff (PAS)\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003estaining of glycogen\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAfter the mice were anesthetized, they were perfused with PBS followed by 4% PFA. The brains were extracted and fixed in 4% PFA for 8 h. PAS staining (Biosharp, BL1120A)\u0026nbsp;of the brain tissues was performed according to the manufacturer\u0026apos;s instructions. For cultured SVG p12 cells, the cells were seeded onto poly-D-lysine-coated slides and fixed in 2% PFA for 15 min at RT. PAS staining of cultured SVG p12 cells was performed following the instructions of a PAS Stain Kit (Solarbio, G1360). The images of brain sections were then examined under a Nikon Eclipse E100 microscope (Japan), while the images of cultured cells were observed using a ZEISS Axio Imager M2 microscope.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMillisect system\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo precisely isolate the glycogen-rich ischemic penumbra and contralateral normal brain tissue in a mouse model of cerebral ischemia-reperfusion injury, automated microdissection was performed using the AVENIO Millisect System (Roche Diagnostics, Indianapolis, IN). Consecutive 20 \u0026mu;m cryosections were prepared on PET membrane slides. A guidance slide was stained with PAS-hematoxylin to identify the ischemic penumbra, characterized by strong PAS positivity with early ischemic alterations, and the contralateral normal tissue. After target regions were marked, consecutive unstained sections were fixed with ice-cold acetone for 2\u0026ndash;3 min (or briefly fixed with 70%\u0026ndash;100% ethanol) and rinsed with RNase-free water to enhance tissue adhesion. Using the system software, the digital template generated from the guidance slide was aligned with the capture sections for automated dissection. Tissues from each region were separately collected into pre-chilled RNase-free microcentrifuge tubes, immediately supplemented with RNA stabilizer, and stored at \u0026minus;80\u0026deg;C for subsequent RNA analysis.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGolgi staining\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor Golgi staining, mouse brains were collected 16 days poststroke and subsequently immersed in Golgi staining fixative for two weeks. A vibratome was used to cut the brain tissue into 100 \u0026micro;m coronal slices. All procedures adhered to the guidelines provided in the Golgi staining kit (FD NeuroTechnologies, PK401). Images were captured using a PanoBrain slide scanner (Meca Scientific) and analyzed with Panolyzer software. Dendritic lengths were measured via ImageJ.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSholl analysis and measurement of dendritic spine density\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNeurons in the ischemic penumbra region of the cortex were traced and analyzed. Using a 100 X oil objective on a ZEISS Axioskop2 plus Microscope (Carl Zeiss, Thornwood, NY) with AxioVision Rel.4.7 software, 10 neurons were randomly selected from each ischemic penumbra region. Sholl\u0026rsquo;s concentric circle method was employed to objectively examine the dendrites of the chosen neurons. Dendrites intersecting each circle were counted to determine the number of dendritic intersections at different radial distances from the neuronal soma to the dendritic tips, in addition to the total dendritic length and branch points. Each value was averaged per mouse, and the mean value for each mouse was taken as n=1. The values per group were averaged and expressed as mean \u0026plusmn; SEM. For each mouse, spine density was quantified in 4 randomly selected high-power fields (HPF; 24 fields total per group) using the NeuronJ plugin (v1.4.3) in ImageJ software (Fiji distribution, NIH), followed by spine counting with the Cell Counter plugin.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLuxol\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003efast blue\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;staining\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMyelin loss was evaluated via the Luxol Fast Blue Myelin Stain Kit (Solarbio, G3245). Frozen sections (10 \u0026micro;m) were preheated at 65\u0026deg;C for 30 min and then stained with Luxol fast blue solution at room temperature for 12 h. The sections were washed sequentially with 95% ethanol and tap water until colorless. They were then differentiated in the solution for 15 s, followed by immersion in 70% ethanol for 30 s, and this process was repeated 2\u0026ndash;3 times. After dehydration in anhydrous ethanol, the samples were mounted with neutral gum. The analysis was performed via a Thunder Imager (Leica DM6B).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eQuantitative RT\u0026ndash;PCR (\u003c/strong\u003e\u003cstrong\u003eqRT‒PCR\u003c/strong\u003e\u003cstrong\u003e)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTotal RNA was extracted via TRIzol reagent (GENESAND). The synthesis of cDNA was carried out with Prime Script\u0026trade; RT Master Mix (TaKaRa). For\u0026nbsp;the qRT‒PCR assays, TaKaRa SYBR qPCR Master Mix and the LightCycler 480II detection system were used. The mRNA expression levels were normalized to those of GAPDH.\u0026nbsp;Detailed information about the\u0026nbsp;parameters\u0026nbsp;is provided in Supplementary\u0026nbsp;Table 4.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMeasurement of\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ethe\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eoxygen consumption rate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAfter the samples were processed, they were transferred to a 96-well XF cell culture microplate at a density of 5 \u0026times; 10\u003csup\u003e5\u003c/sup\u003e cells per well. The XF sensor cartridges were allowed to hydrate overnight at RT, and the following day, the hydration solution was replaced with Seahorse XF hydration solution. Before analysis, the media in the XF microplate was replaced with Seahorse XF DMEM base assay medium (1 mM glucose, 100 mM pyruvate, and 200 mM L-glutamine; pH 7.4) and incubated at 37\u0026deg;C in a non-CO\u003csub\u003e2\u003c/sub\u003e incubator for 1 h. The four injection ports were then filled with oligomycin A (2 \u0026mu;M), FCCP (2 \u0026mu;M), rotenone (0.5 \u0026mu;M), and antimycin A (0.5 \u0026mu;M). The oxygen consumption rate (OCR) was assessed via an XF96 Analyzer (Seahorse Bioscience), with the results normalized to the number of cells.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eATP fluorescent protein biosensor\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn SVGP12 cells transfected with the ATP fluorescent probe plasmid pCMV-Mito-AT1.03 (Beyotime, D2606) for 48 hours, real-time monitoring of ATP levels and their dynamic changes in single living cells was performed via fluorescence resonance energy transfer (FRET). Real-time images were captured via a laser scanning confocal LSM980 two-photon imaging system. The excitation wavelength was set to 435 nm, with emission wavelengths of 475 nm (mseCFP) and 527 nm (cp173-Venus). The absolute intensity value Fc (RFETcorrected) of the photonic emission in the RFET (cp173-Venus/mseCFP) channel was statistically analyzed.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCylinder task\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe cylinder Test (Spontaneous Forelimb Task) was conducted to evaluate sensorimotor function. Mice were tested at the following time points: pre-stroke (baseline), and post-stroke days. Each mouse was individually placed in a transparent glass cylinder (15 cm in height, 10 cm in diameter) and allowed to explore freely for 3 min. During this period, the mice spontaneously reared into a standing position, supporting their body weight with either a single forelimb or both forelimbs. Their spontaneous rearing activity was recorded on video for 5 min. The recorded videos were subsequently analyzed at 1/4 of the actual speed to assess the animal\u0026rsquo;s forelimb preference during exploratory behavior. Only rearing episodes in which both forelimbs were clearly visible were included in the analysis. The duration of support by the left forelimb, right forelimb, or both forelimbs simultaneously against the cylinder wall was quantified for each mouse. The percentage of time using each forelimb was calculated, and an asymmetry index was derived using the following formula: (% ipsilateral forelimb use) - (% contralateral forelimb use).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGrid-walking task\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eEach mouse was placed individually on an elevated wire grid (32 \u0026times; 20 \u0026times; 50 cm; mesh size 12 \u0026times; 12 mm) and allowed to walk freely for 5 min while being video-recorded. Videos were analyzed at one-fourth normal speed by an experimenter blinded to treatment groups. A foot fault was defined as either a step through the grid opening without support or a resting position with the wrist below grid level. The percentage of foot faults was calculated as (number of foot faults / total steps) \u0026times; 100. The ratio of foot faults to total steps was calculated to control for variations in locomotor activity across animals and test sessions.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMorris\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003ewater maze\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;test\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo evaluate spatial recognition memory in the mice, the Morris water maze test was performed 21 to 27 days after photothrombotic infarction surgery. The maze had a diameter of 120 cm and a water depth of approximately 30 cm, featuring a submerged platform that measured 11 cm \u0026times; 11 cm. During the initial training phase, the mice were placed in the water facing the wall of the pool from various starting points. The time taken to find the submerged platform was measured in seconds. If a mouse took longer than 60 s to find the platform, it was guided to it and allowed to stay there for 10 s. Each mouse underwent four training sessions daily for five consecutive days prior to surgery. Between days 21 and 26 post surgery, acquisition training was conducted with four sessions daily for five days. On day 27, the platform was removed, and the mice were placed in the water from the opposite quadrant of the platform. The time spent in the former platform quadrant was recorded.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAdhesive removal test\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe adhesive removal test is used to evaluate sensory‒motor function asymmetry in mice following cerebral ischemia. A small piece of tape (2 mm wide and 3 mm long) was\u0026nbsp;placed on the forepaw controlled by the injured cortex. The time taken by the mice to remove the tape\u0026nbsp;was\u0026nbsp;recorded.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRotarod test\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe rotarod apparatus was used to test the motor coordination of the mice. The mice were placed on a rotating rod, with the speed gradually increasing from 2 revolutions per min to 50 revolutions per min. Each trial lasted for 5 min, with a 30 min interval between trials. The time until the mice fell or lost their balance was recorded.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eOpen field test\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe experiment was conducted in a white, open field cubic box (50 \u0026times; 50 \u0026times; 30 cm). After an acclimation period, the mice were recorded for 6 min, and their total distance traveled was measured using EthoVision XT 14 software (Noldus, Wageningen, The Netherlands) to assess overall motor function. The arena was cleaned with 75% ethanol followed by water after each test.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAxon sprouting evaluation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo evaluate corticospinal tract remodeling and motor function recovery poststroke, 2 \u0026mu;L of 10% Dextran, Biotin, 10,000 MW, Lysine Fixable (BDA-10,000, Invitrogen\u0026trade;, D1956) was injected at two sites in the contralateral cortex one month after the stroke, with the following stereotaxic coordinates: (1) anteroposterior: 0.6 mm, medial-lateral: 1.2 mm, dorsal-ventral: 0.82 mm; (2) anteroposterior: 0 mm, medial-lateral: 1.8 mm, dorsal-ventral: 0.82 mm. Two weeks later, the animals were perfused with PBS and subsequently with 4% PFA, after which both the brain and spinal cord were harvested.\u003c/p\u003e\n\u003cp\u003eTransverse cervical spinal cord sections and coronal brain sections were sliced to a thickness of 30 \u0026mu;m, and BDA was labeled with Cy\u0026trade;3 streptavidin (Jackson ImmunoResearch, 016-160-084). Thunder Imager (Leica DM6B) was used to detect BDA in the spinal cord and red nucleus, and ImageJ was\u0026nbsp;used to quantify the number of fibers crossing the denervated area from intact circuits.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eATAD3A\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eoligomer\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;detection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor the cell samples, cell lysates were prepared via RIPA buffer, followed by the addition of 1 mM BMH (Thermo Scientific\u0026trade;, 22330) for incubation at RT for 25 min. Subsequently, 0.1% \u0026beta;-mercaptoethanol (\u0026beta;-ME) was added, and the reaction was terminated by incubation at RT for 10 min. The proteins were then loaded in SDS loading buffer and analyzed via 8% SDS‒PAGE. For tissue samples, 1% paraformaldehyde was used for incubation at RT for 20 min, followed by three washes with 100 mM glycine to terminate the reaction. The tissue was subsequently homogenized in RIPA buffer without reducing agents. After BCA protein quantification, the samples were analyzed via 8% SDS‒PAGE, with or without \u0026beta;-ME.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGST-pull down\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo investigate protein‒protein interactions, a GST pull-down assay was performed. GST- and His-tagged fusion proteins were expressed in \u003cem\u003eE. coli\u0026nbsp;\u003c/em\u003eRosetta (DE3) cells by induction with 0.5mM isopropyl \u0026beta;-D-1-thiogalactopyranoside (IPTG) at 16\u0026deg;C overnight and purified using glutathione-Sepharose beads and Ni-NTA Agarose, respectively. The His-tagged protein was subsequently eluted from the Ni-NTA resin. The two purified proteins were incubated with an equal volume of pull-down binding buffer (50 mM Tris-HCl, pH 8.0, 250 mM NaCl, 1 mM EDTA, 1% NP-40, 10 mM MgCl₂, 0.2 mM PMSF, and 0.2 mM DTT) at 4\u0026deg;C overnight. The samples were then washed three times with TEN buffer (20 mM Tris-HCl, pH 7.4, 0.1 mM EDTA, and 100 mM NaCl). After centrifugation at 1,500 rpm for 1 min at 4\u0026deg;C, the precipitated components were analyzed by immunoblotting using anti-His or anti-GST antibodies.\u0026nbsp;Following electrophoresis, the SDS-PAGE gel was incubated with Coomassie Blue staining solution for 1 min with low-heat assistance. The gel was then destained with destaining solution (20% methanol, 10% acetic acid) on a rocking platform until the background became clear and protein bands were clearly visible.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIn vitro acetylation assay\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePurified GST-KAT8 from \u003cem\u003eE. coli\u0026nbsp;\u003c/em\u003eRosetta (DE3) cells was incubated with His-ATAD3A and 50\u0026mu;M acetyl-coenzyme A in HAT assay buffer (50 mM Tris-HCl pH 8.0, 10% glycerol, 0.1 mM EDTA, 1 mM DTT) at 30\u0026deg;C for 3 h. The acetylated His-ATAD3A (His-ATAD3A\u003csup\u003eAce\u003c/sup\u003e) protein was subsequently purified for downstream applications. The His-ATAD3A\u003csup\u003eAce\u003c/sup\u003e was incubated with purified GST-HDAC3 or GST-HDAC6 from E. coli in deacetylase assay buffer (10 mM Tris-HCl pH 8.0, 10 mM NaCl) at 30\u0026deg;C for 30 min. Reaction mixtures were then analyzed by western blot. Purified GST-Sirt1 from \u003cem\u003eE. coli\u003c/em\u003e Rosetta (DE3) cells was incubated with acetylated His-ATAD3A in deacetylase buffer (10 mM Tris-HCl pH 8.0, 150 mM NaCl, 10% glycerol) supplemented with 1 mM NAD⁺\u0026nbsp;at 25\u0026deg;C for 2 h. Samples were mixed with equal volumes of loading buffer and analyzed by SDS-PAGE.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSplit-luc\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn this study, two potentially interacting proteins were fused to the N-terminal and C-terminal fragments of firefly luciferase. A stable cell line overexpressing one of the fusion proteins (bait protein A) was transiently transfected with the other fusion protein (prey protein B). The interaction between the two proteins in 293T cells was assessed by measuring firefly luciferase activity via a microplate reader. Positive controls included a stable cell line overexpressing firefly luciferase, while negative controls included cell lines expressing either bait fusion protein A alone (stable transfection) or prey fusion protein B alone (transient transfection), as well as a cell line stably expressing the bait fusion protein (C-terminal fusion) followed by transient transfection with the bait fusion protein (N-terminal fusion). Overexpression plasmids were constructed on the pLJM1-MCS-linker-N-fluc and pLJM1-Cf-luc-Linker-MCS vectors: pLJM1-ATAD3A-linker-N-fluc; pLJM1-C-fluc-Linker-HDAC3. The lentivirus was used to infect 293T cells to establish a stable cell line overexpressing the bait protein HDAC3, designated 293T-HDAC3-C-fluc. The cells were then transiently transfected with ATAD3A-N-fluc. HDAC3-C-fluc and ATAD3A-N-fluc 293T cells were seeded in 96-well plates in suspension (nonadherent) at a density of 1\u0026times;10\u003csup\u003e6\u003c/sup\u003e cells per well. Subsequent imaging was performed using an ID5 multifunctional microplate reader.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCellular mitochondria imaging\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor live-cell mitochondrial imaging, the cells were cultured in confocal glass dishes (Biosharp, BS-15-GJM). The medium was replaced with fresh serum-free medium prior to staining. Next, the cells were treated with 100 nM MitoTracker Red CMXRos (Thermo Fisher Scientific, M7512) at 37\u0026deg;C for 10 min. Following three washes with PBS,\u0026nbsp;the mitochondrial morphology was imaged via a Leica STED microscope. The acquired images were analyzed via Imaris (Bitplane, Imaris 9.6.2). Mitochondrial morphology was assessed by quantifying the number of fragmented mitochondria relative to the total number of cells, with counts normalized to the total number of cells in each field of view for comparison and statistical analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGlycogen pull down\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe samples were collected in binding buffer containing protease inhibitors (10% glycerol, 1 mM EDTA, 20 mM Tris-HCl, 150 mM NaCl, pH 8.0). Following treatment with an ultrasonic disruptor, the samples were centrifuged at 30,000 rpm for 30 min at 4\u0026deg;C to obtain the supernatant. The supernatants were divided into two groups: one mixed with purified glycogen (MedchemExpress, 9005-79-2) and the other without glycogen and then incubated for 90 min at 4\u0026deg;C. After centrifugation at 100,000 rpm for 90 min at 4\u0026deg;C, the protein precipitate bound to glycogen was collected for mass spectrometry analysis. Both the protein precipitate and the supernatant were analyzed via SDS‒PAGE.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTurboID protein purification\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn SVG p12 human fetal glial cells, after transfection with the ATAD3A-TurboID plasmid, the cells were cultured for 48 h. Biotin (200 \u0026micro;M) was then added to the culture medium, and the cells were incubated at 37\u0026deg;C for an additional 48 h. Next, the cells were washed three times with cold DPBS and digested with 0.5% trypsin-EDTA. Following centrifugation at 800 rpm for 5 min, the cell pellet was resuspended in urea lysis buffer (8 M urea, 10 mM Tris, 100 mM NaH\u003csub\u003e2\u003c/sub\u003ePO\u003csub\u003e4\u003c/sub\u003e, pH 8.5) and subjected to ultrasonic treatment for 3 min (30 W amplitude, 5 s pulse). The mixture was shaken at 4\u0026deg;C for 30 min and then centrifuged at 12,000 rpm for 10 min to collect the supernatant. Biotinylated beads (Thermo Fisher Scientific, 20353) were added (80 \u0026micro;L) and incubated overnight at 4\u0026deg;C. After centrifugation at 12,000 rpm for 2 min at 4\u0026deg;C, the beads containing biotinylated proteins were collected. The samples were then washed sequentially with RIPA buffer, 1 M KCl, and 0.1 M Na\u003csub\u003e2\u003c/sub\u003eCO\u003csub\u003e3\u003c/sub\u003e, followed by a quick rinse with RIPA buffer. Subsequently, mass spectrometry analysis, SDS‒PAGE analysis\u0026nbsp;(4 to 12% polyacrylamide gel), and silver staining were performed. For purification of proteins synthesized specifically by astrocytes in vivo, Ast ER-BioID HA mice were sequentially injected with biotin with 100 \u0026mu;L of a 2-mg/mL biotin solution (Beyotime, Cat# ST2051) once a day intraperitoneally. After RIPA lysis of whole brain tissue, biotinylated proteins were purified according to the above method.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eQuantitative\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003eLC‒MS\u003c/strong\u003e\u003cstrong\u003e/MS analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAn appropriate amount of TEAB was added to the samples to adjust the pH to 8.0. Five microlitres of each suspension was\u0026nbsp;subjected to SDS‒PAGE. The\u0026nbsp;remaining samples were\u0026nbsp;subjected to\u0026nbsp;digestion. The protein\u0026nbsp;mixture\u0026nbsp;was reduced with 5 mM dithiothreitol for 30 min at 56\u0026deg;C and alkylated with 11 mM iodoacetamide for 15 min at room temperature in darkness. The samples were then diluted by adding 200 mM TEAB to\u0026nbsp;a\u0026nbsp;urea concentration less than 2 M. Next, trypsin (1:50 trypsin-to-protein mass ratio) was added for the first digestion overnight and 1:100 for a second 4 h\u0026nbsp;digestion. Finally, the peptides were desalted\u0026nbsp;on\u0026nbsp;a\u0026nbsp;Strata X SPE column and separated on\u0026nbsp;an AvianQuick-Nano\u0026nbsp;UPLC system (Thermo Fisher\u0026nbsp;Scientific).\u0026nbsp;The separated peptides were analyzed in an\u0026nbsp;Orbitrap Exploris 480\u0026nbsp;instrument\u0026nbsp;with a\u0026nbsp;nanoelectrospray\u0026nbsp;ion source.\u0026nbsp;The precursors\u0026nbsp;and fragments were analyzed\u0026nbsp;with an\u0026nbsp;Orbitrap detector. The full MS scan resolution was set to 60000 for a scan range of\u0026nbsp;350\u0026ndash;1400\u0026nbsp;m/z. The MS/MS scan was fixed first mass\u0026nbsp;at\u0026nbsp;120.0 m/z at a resolution of 15000. HCD fragmentation was performed at a normalized collision energy of 27%.\u0026nbsp;The automatic\u0026nbsp;gain control target was set at 1E6, with a maximum injection time of 22 ms.\u0026nbsp;The DIA data were processed via the\u0026nbsp;DIA-NN search engine. Tandem mass spectra were searched against Mus_musculus_10090_SP_20231220.fasta concatenated with\u0026nbsp;the\u0026nbsp;reverse decoy database. Trypsin/P was specified as\u0026nbsp;a\u0026nbsp;cleavage enzyme allowing up to 1 missing cleavage. Excision\u0026nbsp;of\u0026nbsp;N-term Met and carbamidomethyl on Cys were specified as fixed\u0026nbsp;modifications. The\u0026nbsp;FDR was adjusted to \u0026lt; 1%. Gene\u0026nbsp;Ontology (GO)\u0026nbsp;analysis,\u0026nbsp;Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway\u0026nbsp;annotation, subcellular localization annotation, and\u0026nbsp;protein‒protein\u0026nbsp;interaction network\u0026nbsp;analysis\u0026nbsp;were\u0026nbsp;subsequently performed.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIsolation of astrocytes from adult murine brain\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIsolation of astrocytes from adult murine brain was performed according to the manufacturer\u0026apos;s instructions (Miltenyi, Cat#130-107-677). First, enzyme mixtures were prepared in advance by mixing Enzyme Mix 1 (50 \u0026mu;L Enzyme P + 1900 \u0026mu;L Buffer Z) and Enzyme Mix 2 (20 \u0026mu;L Buffer Y + 10 \u0026mu;L Enzyme A) according to the specified ratios. Moreover, the gentleMACS Octo Dissociator was set to program 37C_ABDK_01, and the appropriate C-tubes were prepared. Tissue processing: Following anesthesia, the mice were perfused transcardially with ice-cold DPBS (Gibco, 14287080). The cerebral cortex was subsequently rapidly dissected and minced into small pieces, after which the tissue fragments were transferred into C Tubes containing the enzyme mixture. Immediately thereafter, tissue dissociation was performed via the gentleMACS Octo Dissociator (program: 37C_ABDK_01). Upon completion of the program, the cell suspension was centrifuged (300 \u0026times;g, 10 min, 4\u0026deg;C), and then the supernatant was carefully aspirated to collect the cell pellet. Next, the cell pellet was resuspended in ice-cold PBS, followed by the addition of Debris Removal Solution with gentle mixing. Thereafter, ice-cold PBS was carefully added on top of the suspension (avoiding mixing of phases), after which the sample was centrifuged (3000 \u0026times;g, 10 min, 4\u0026deg;C). After the upper two phases were removed, the remaining cells were resuspended in ice-cold PBS and centrifuged again (1000 \u0026times;g, 10 min, 4\u0026deg;C) to obtain a purified cell suspension. Magnetic Labeling and Separation. For magnetic labeling, the cells were first resuspended in buffer (PBS + 0.5% BSA) at a concentration of 80 \u0026mu;L per 10 \u003csup\u003e7\u003c/sup\u003e cells and then incubated with 10 \u0026mu;L of FcR Blocking Reagent for 10 min at 4\u0026deg;C to prevent nonspecific binding. Subsequently, 10 \u0026mu;L Anti-ACSA-2 MicroBeads were added per 10\u003csup\u003e\u0026nbsp;7\u003c/sup\u003e cells, after which the mixture was incubated for 15 min at 4\u0026deg;C. Following labeling, the cells were washed with 1~2 mL buffer (300 \u0026times;g, 10 min) and finally resuspended in 500 \u0026mu;L of buffer per 10 \u003csup\u003e7\u003c/sup\u003e cells. Positive selection: Prior to separation, the LS columns were equilibrated with 3 mL of buffer. After sample loading, the flow-through containing ACSA-2-negative cells were collected, and the column was washed three times with 3 mL buffer each time. Upon removing the column from the magnetic field, ACSA-2-positive cells were eluted and collected with 5 mL buffer. For the analysis of ATAD3A acetylation and oligomerization, astrocytes from four mice per group were pooled to obtain sufficient protein.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003escRNA-seq\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003elibrary\u003c/strong\u003e\u003cstrong\u003e\u0026nbsp;preparation, sequencing and analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFreshly dissociated single-cell suspensions were loaded onto a 10X Genome GenCode single-cell instrument to generate GEMs (Gel Bead-In-EMlusion). ScRNA-seq libraries were subsequently prepared following the manufacturer\u0026rsquo;s protocol for the Chromium Next GEM Single Cell 3\u0026rsquo; Regent Kit v3.1. Libraries were then sequenced via the NovaSeq 6000 platform in paired-end sequencing (PE150) mode. For scRNA-seq data processing, raw reads were aligned to the mouse reference genome GRCm39 via Cell Ranger software (v8.0.0) to generate cell \u0026times; gene matrices. The FASTQ files were processed via the standard workflow of CellRanger to generate gene expression count matrices for each sample. The raw count matrices from all the samples were merged via Scanpy (V1.9.6), followed by quality control. Genes expressed in fewer than 50 cells were excluded, and cells were filtered on the basis of the following criteria: (1) the number of detected genes per cell ranged between 500 and 7000, (2) the mitochondrial RNA content was less than 20%, and (3) total counts per cell did not exceed 50,000. To eliminate potential doublets, Doublet Detection (http://doi.org/10.5281/zenodo.2678041) was applied with the boost_rate set to 0.5 and the voter_thresh set to 0.9. After filtering, 63,283 high-quality cells were retained for downstream analyses. The gene expression matrix of the retained cells was normalized by scaling the total UMI counts per cell to 10,000, followed by log transformation. The top 2,000 highly variable genes were subsequently selected for dimensionality reduction and clustering. Principal component analysis (PCA) was performed, and the first 40 principal components (PCs) were used to construct a UMAP. In the initial round of clustering, major cell types (including Astrocytes, Oligodendrocytes, Microglia, Macrophages, Ependymal Cells, ChP Epithelial Cells, Endothelium, Pericytes/SMCs, VLCMs, Neurons, T cells, and Erythrocytes) were identified via Louvain clustering at a resolution of 0.3 on the basis of canonical markers. A second round of clustering was performed to further resolve astrocyte subpopulations with similar clustering parameters. DEGs of each cell type were calculated via the function rank_genes_groups in Scanpy. The top 300 DEGs were selected for gene pathway enrichment analysis via the Metascape website. For ligand‒receptor analysis of the scRNA-seq data, ligand‒receptor interaction pairs between cell subtypes were inferred via NeuronChat (v1.0.0)\u003csup\u003e39\u003c/sup\u003e.\u0026nbsp;We subsequently compared the cell type\u0026nbsp;interaction counts and\u0026nbsp;identified\u0026nbsp;the changed\u0026nbsp;ligand‒receptor\u0026nbsp;pairs between\u0026nbsp;the\u0026nbsp;WT and KO\u0026nbsp;groups.\u0026nbsp;To recover the cellular dynamics of astrocyte subtypes, we performed RNA velocity analysis. Spliced and\u0026nbsp;unspliced\u0026nbsp;reads were counted\u0026nbsp;via\u0026nbsp;velocyto.py (v0.17.17).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eVirus injection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFor stereotaxic intracranial viral injection, mice were anesthetized and secured in a stereotaxic frame (RWD Life Science). Body temperature was maintained using a heating pad, and corneal dehydration was prevented by applying sterile ophthalmic ointment. Following a midline scalp incision, small craniotomies (0.5 mm in diameter) were drilled over the target regions using the following coordinates relative to bregma: for the sensorimotor cortex. Using a glass micropipette connected to a micro-syringe and an injection pump (UMP3, WPI, USA), 100 nL of viral solution was delivered at a rate of 5 nL/s. The following viruses were used to be delivered via stereotaxic injection into the sensorimotor cortex: rAAV-CMV-ATAD3A(K134E)-3XFlag-WPRE-hGH pA (abbreviated as AAV ATAD3A \u003cem\u003e\u003csup\u003eK134E\u003c/sup\u003e\u003c/em\u003e) and rAAV-CMV-3XFlag-WPRE-hGH pA (control). Experiments were conducted 3 weeks post-injection. For tail vein viral injection to infect astrocytes throughout the brain, a recombinant adeno-associated virus (rAAV) carrying the target gene was mixed with a glial cell-specific Cre recombinase virus (rAAV-GFaABC1D-CRE-4x6TPA) at a 2:1 ratio to prepare a viral mixture with a total volume of 150 \u0026mu;L. For knockdown experiments, the experimental group received a cocktail of rAAV-CMV-DIO-(EGFP-U6)-shRNA (Hdac3)-WPRE-hGH polyA and rAAV-GFaABC1D-CRE-4x6T-PA, while the control group received rAAV-CMV-DIO-(EGFP-U6)-shRNA(scramble)-WPRE-hGH polyA and rAAV-GFaABC1D-CRE-4x6T-PA. For overexpression experiments, the experimental group received a cocktail of rAAV-CMV-DIO-pygb-3XFlag-P2A-EGFP-WPREs and rAAV-GFaABC1D-CRE-4x6T-PA, whereas the control group received rAAV-CMV-DIO-EGFP-WPRE-hGH pA and rAAV-GFaABC1D-CRE-4x6T-PA. Please refer to the attached materials for the specific virus sequence.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eWhole-cell recordings\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePyramidal neuron whole-cell currents were measured via an upright microscope (Olympus X51W) in conjunction with a patch-clamp amplifier (Axon Patch 700B). Recordings were low-pass filtered at 2 kHz and sampled at 10 kHz via a Digidata 1440A. Series resistance was monitored (5\u0026ndash;25 M\u0026Omega;), and only data with changes \u0026lt;20% were included. Data collection and analysis were performed with pClamp 10.3 and Clamfit 10.3. Biocytin (0.2%) was added to the internal solution for neuron labeling, and the slices were subsequently fixed for immunohistochemistry. For the tonic inhibitory current (\u003cem\u003eI\u003c/em\u003e\u003csub\u003etonic\u003c/sub\u003e) and average phase current, the electrode was filled with a solution containing 120 mM CsMeSO\u003csub\u003e4\u003c/sub\u003e (pH 7.25-7.30). To reduce extracellular GABA, 5 mM GABA was added during perfusion. \u003cem\u003eI\u003c/em\u003e\u003csub\u003etonic\u003c/sub\u003e displacement was recorded as baseline displacement following the administration of 100 \u0026micro;M BMI. The strong current density was calculated by dividing the current amplitude by the membrane capacitance. For action potential recordings, the electrode was filled with a solution containing 70 mM potassium. For miniature excitatory postsynaptic currents (mEPSCs), the electrode contained 132.5 mM Cs-gluconate, with tetrodotoxin and bicuculline added to block GABA receptor currents. The data were analyzed via the Mini Analysis Program 6.0, with up to 100 events selected for cumulative probability analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFiber photometry\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe mice were fixed in a stereotaxic frame, and the scalp was incised along the midline. A cranial drill was used to expose the motor cortex at the following stereotaxic coordinates: anteroposterior: 0.47 mm, mediolateral: 1.5 mm, and dorsoventral: 0.82 mm (relative to bregma). An AAV-CaMKII-GCaMP6s virus (BrainVTA Co., Ltd. (Wuhan, China)) was injected at a volume of 100\u0026nbsp;nl\u0026nbsp;at a rate of 5 nl/s. The scalp was sutured after the procedure. One month later, a photothrombotic stroke was induced in the sensory cortex at the following coordinates: anteroposterior: 0 mm, mediolateral: 2 mm (relative to bregma). Simultaneously, a ceramic optical fiber (RWD, R-FOC-BL200C-39NA) was implanted in the motor cortex and secured with bone cement (Super Bond C\u0026amp;B). Fourteen days\u0026nbsp;poststroke, the mice were placed in a large beaker with a diameter of 10 cm, and recordings were conducted\u0026nbsp;via\u0026nbsp;an optical fiber photometry system (Inper). The light intensity\u0026nbsp;at\u0026nbsp;470 nm was set within\u0026nbsp;the\u0026nbsp;range of 20-40 \u0026micro;W,\u0026nbsp;whereas\u0026nbsp;the intensity\u0026nbsp;at\u0026nbsp;410 nm was set within\u0026nbsp;the\u0026nbsp;range of 10-20 \u0026micro;W.\u0026nbsp;The data\u0026nbsp;were further processed\u0026nbsp;via\u0026nbsp;signal acquisition software (InperStudio) and analysis software (InperDataProcess). The onset of movement from a standing position (designated as the\u0026nbsp;marker) was used as the basis for trial division. Multiple trials were analyzed and statistically evaluated, resulting in the generation of heatmaps and Delta F/F result graphs.\u0026nbsp;To record hippocampal neuronal activity, we injected AAV-CaMKII-GCaMP6s into the hippocampus (anteroposterior: 0 mm, mediolateral: \u0026plusmn; 2 mm, dorsoventral: -1.7 mm relative to bregma) and induced a photothrombotic stroke at the same coordinates three weeks later. Hippocampal GCaMP6s signals were recorded on day 8 post-stroke. The onset of immobility, defined as the behavioral event for analysis, was recorded from the moment the animal was placed in the fear-conditioned context.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eElectrophysiological recordings\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFollowing previous studies, a preparatory surgery was performed at least two days before the in vivo multi-channel recording. General anesthesia was induced (3%) and maintained (1~1.5%) with isoflurane. The mouse was then fixed in a stereotaxic frame (RWD). After applying lidocaine gel, the skin above the target areas was removed. A craniotomy was performed in the contralateral somatosensory cortex relative to the viral injection site, and the surgical opening was sealed with a silicon elastomer upon completion. In the somatosensory cortex which had been injected with AAV virus two weeks earlier, an optical cannula (RWD, 200 \u0026mu;m diameter, 0.5 NA) was implanted and fixed with dental acrylic (C\u0026amp;B Super Bond). To facilitate head fixation during recording, a metal post was affixed to the skull using dental acrylic (Super Bond C\u0026amp;B). Additionally, a Teflon-coated silver wire was implanted in the contralateral hemisphere as the reference electrode. Following recovery from anesthesia, the mouse was returned to its home cage and monitored for the subsequent two days. Before in vivo recordings, mice were habituated to the head-fixation system. Each animal underwent three recording sessions per day, with each session lasting no longer than 30 minutes. Before recording, the silicon seal was removed, and a 64-channel silicon probe (NeuroNexus) was inserted into the somatosensory cortex. Neural signals were recorded at a sampling rate of 40 kHz and bandpass-filtered (0.3\u0026ndash;3 kHz) using the OmniPlex Neural Recording Data Acquisition System (Plexon). Single-unit spikes were extracted through semi-automated spike sorting with Offline Sorter (Plexon) and subsequently analyzed using custom MATLAB scripts.To mark the recording site, DiI, a non-toxic lipophilic fluorescent dye, was applied to the electrode which was then reinserted into the recording site after all sessions were completed. Finally, mice were euthanized to confirm the recording location.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompound action potential measurements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eComposite action potentials (CAPs) are used to assess CC and EC. Mouse brain slices (300 \u0026mu;m) were cut starting from 1.06 mm relative to the bregma. The slices were placed in artificial cerebrospinal fluid (26 mmol/L NaHCO3, 2.5 mmol/L KCl, 10 mmol/L glucose, 1 mmol/L Na\u003csub\u003e2\u003c/sub\u003eH\u003csub\u003e2\u003c/sub\u003ePO\u003csub\u003e4\u003c/sub\u003e, 126 mmol/L NaCl, 2.5 mmol/L CaCl\u003csub\u003e2\u003c/sub\u003e, 1.3 mmol/L MgCl\u003csub\u003e2\u003c/sub\u003e; pH 7.4, 95% O\u003csub\u003e2\u003c/sub\u003e) at 34\u0026deg;C for 30 min and then allowed to rest at room temperature for 1 hour. During recording, a bipolar stimulating electrode was positioned laterally to the midline of the corpus callosum. A glass extracellular recording pipette was placed in the EC. The distance between the stimulating and recording electrodes was 0.75 mm. CAP signals are digitized via a digidata 1500B (Molecular Devices, San Jose, CA), amplified, and recorded via an Axoclamp 700B (Molecular Devices, San Jose, CA), with analysis performed via pClamp 10. An input‒output curve is generated at a stimulation intensity of 2 mA. The induced CAPs exhibit a biphasic waveform, with the early peak representing fast-conducting myelinated axons (N\u003csub\u003e1\u003c/sub\u003e) and the delayed peak indicating slow-conducting unmyelinated axons (N\u003csub\u003e2\u003c/sub\u003e). The amplitudes of N\u003csub\u003e1\u003c/sub\u003e and N\u003csub\u003e2\u003c/sub\u003e are calculated as the variance from the first and second peaks to the trough.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eProtein oxidation detection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eProtein oxidation was determined by protein carbonylation using the OxyBlot\u0026trade; protein oxidation detection kit (Sigma Aldrich, Catalog No. S7150). Following RIPA lysis, protein samples were denatured by adding an equal volume of 12% SDS. Derivatization was performed by adding 20 \u0026mu;L of 1X DNPH solution per sample, while negative controls received 20 \u0026mu;L of 1X derivatization-control solution. The mixtures were incubated at room temperature for 15 min (not exceeding 30 min), after which 7.5 \u0026mu;L of neutralization solution was added. Samples lacking reducing agents were supplemented with 2-mercaptoethanol to a final concentration of 0.74 M. For western blot analysis, the derivatized and neutralized protein samples were loaded directly onto the gel. Separately, for the molecular weight standard, 2.5 \u0026micro;L of the DNP-labeled protein marker was mixed with 20 \u0026micro;L of 1X gel loading buffer prior to loading. Repeated freeze-thaw cycles of the protein standards were avoided. Proteins were transferred to nitrocellulose membranes which were then incubated in the primary antibody solution (anti-DNP, 1:150) for 2 h, followed by incubation in the secondary antibody solution (1:300) for 1 h at room temperature. The washing procedure was repeated eight times within 40 min.\u0026nbsp;Immunoreactive bands were visualized by enhanced chemiluminescence. The oxidative index was calculated from digital blot images using a two-step normalization. First, the total lane intensity from the Anti-DNP blot was normalized to the corresponding Ponceau S total protein signal. This normalized value for each sample was then expressed relative to the normalized value of the Biotin (+) Ctrl group to establish the final Oxidative Index.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStatistical analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn this study, all the statistical analyses were conducted via GraphPad Prism 8.0 or R software. The data are presented as the means \u0026plusmn; SEMs or SDs. The statistical methods followed these principles: comparisons between two groups were performed via a two-tailed unpaired Student\u0026rsquo;s t test, whereas multiple groups were analyzed via one-way ANOVA or two-way ANOVA analysis. p \u0026ge; 0.05 was considered as no significance (ns); *p\u0026lt; 0.05; **p \u0026lt; 0.01; ***p \u0026lt; 0.005. The processing of specific data, such as electrophysiological and bioinformatics analysis, is detailed in the corresponding methods.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eDATA AVAILABILITY\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll unique/stable reagents generated in this study are available from the lead contact with a completed materials transfer agreement. scRNA sequencing data has been submitted and deposited in GEO database (GEO: GSE295882). The proteomics raw data are currently being uploaded to ProteomeXchange and make them public. Obtaining the accession ID or PRIDE Submission reference: https://www.ebi.ac.uk/pride/archive/projects/PXD064814 (glycogen binding proteomics); https://www.ebi.ac.uk/pride/archive/projects/PXD064694 (ATAD3A binding proteomics data). Microscopy data and any required information reported in this paper is available from the lead contact. And this study doesn\u0026rsquo;t involve any original code.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eACKNOWLEDGMENTS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe thank Dr. Tian Xue, Dr. Zhi Zhang and Dr. Qiang Liu for providing constructive suggestions. We would like to thanks to Dr. Ziqiu Zhang, Dr. Xiezong Hu and Dr. Qi Wu for their assistance with the bioinformatics analysis. We would like to extend special thanks to\u0026nbsp;Chunying Yin for her assistance with electron microscopy sample preparation. We extend our sincere gratitude to the National Health and Disease Human Brain Tissue Resource for providing the human brain tissue samples critical to this study. This study was supported by: The National Natural Science Foundation of China (82272225);\u0026nbsp;The National Natural Science Foundation of China (82572514);The National Natural Science Foundation of China (81860249); USTC Research Funds of the Double First-Class Initiative (YD9110002084); The National Natural Science Foundation of China (32471079); The Strategic Priority Research Program of the Chinese Academy of Sciences (Grant No. XDB0940101); Noncommunicable Chronic Diseases-National Science and Technology Major Project (Grant No. 2023ZD0507500); Research Funds of the Center for Advanced Interdisciplinary Science and Biomedicine of IHM (Grant No. QYPY20220005); Natural Science Foundation of Anhui Province (Grant No. 2208085MH241); USTC Research Funds of the Double First-Class Initiative (YD9100002053).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAUTHOR CONTRIBUTIONS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eH.R. Zhu conceived and designed the study. M.M. Yang and H. Zhu wrote the manuscript. M.M. Yang and Z. Li conducted in vivo and in vitro experiments. H.R. Zhu, M.M. Yang, and Z. Li performed data analysis. C.L. Li, S.J. Sun, and X.R. Liu carried out stroke experiments and behavioral recordings. M.M. Yang, L. Wang, Q. Shao, and J.M. Zhang performed immunohistochemistry. S. Wang and K.Q. He revised the manuscript and provided fund support. All authors read, validated the underlying data, and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCOMPETING INTERESTS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eWahl, A.S.\u003cem\u003e et al.\u003c/em\u003e Neuronal repair. Asynchronous therapy restores motor control by rewiring of the rat corticospinal tract after stroke. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e344\u003c/strong\u003e, 1250-1255 (2014).\u003c/li\u003e\n\u003cli\u003eJoy, M.T. \u0026amp; Carmichael, S.T. Encouraging an excitable brain state: mechanisms of brain repair in stroke. \u003cem\u003eNat Rev Neurosci\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 38-53 (2021).\u003c/li\u003e\n\u003cli\u003eGiaume, C., Koulakoff, A., Roux, L., Holcman, D. \u0026amp; Rouach, N. Astroglial networks: a step further in neuroglial and gliovascular interactions. \u003cem\u003eNat Rev Neurosci\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 87-99 (2010).\u003c/li\u003e\n\u003cli\u003eZhou, J.\u003cem\u003e et al.\u003c/em\u003e Astrocytic LRP1 enables mitochondria transfer to neurons and mitigates brain ischemic stroke by suppressing ARF1 lactylation. \u003cem\u003eCell Metab\u003c/em\u003e \u003cstrong\u003e36\u003c/strong\u003e, 2054-2068.e2014 (2024).\u003c/li\u003e\n\u003cli\u003eCai, Y.\u003cem\u003e et al.\u003c/em\u003e Glycogenolysis Is Crucial for Astrocytic Glycogen Accumulation and Brain Damage after Reperfusion in Ischemic Stroke. \u003cem\u003eiScience\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 101136 (2020).\u003c/li\u003e\n\u003cli\u003eSun, R.C.\u003cem\u003e et al.\u003c/em\u003e Brain glycogen serves as a critical glucosamine cache required for protein glycosylation. \u003cem\u003eCell Metab\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 1404-1417.e1409 (2021).\u003c/li\u003e\n\u003cli\u003eSinadinos, C.\u003cem\u003e et al.\u003c/em\u003e Neuronal glycogen synthesis contributes to physiological aging. \u003cem\u003eAging Cell\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 935-945 (2014).\u003c/li\u003e\n\u003cli\u003eGusarov, I.\u003cem\u003e et al.\u003c/em\u003e Glycogen controls Caenorhabditis elegans lifespan and resistance to oxidative stress. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 15868 (2017).\u003c/li\u003e\n\u003cli\u003eNitschke, F., Ahonen, S.J., Nitschke, S., Mitra, S. \u0026amp; Minassian, B.A. Lafora disease - from pathogenesis to treatment strategies. \u003cem\u003eNat Rev Neurol\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 606-617 (2018).\u003c/li\u003e\n\u003cli\u003eJohnson, E.C.B.\u003cem\u003e et al.\u003c/em\u003e Large-scale proteomic analysis of Alzheimer\u0026apos;s disease brain and cerebrospinal fluid reveals early changes in energy metabolism associated with microglia and astrocyte activation. \u003cem\u003eNat Med\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 769-780 (2020).\u003c/li\u003e\n\u003cli\u003eAhonen, S.\u003cem\u003e et al.\u003c/em\u003e Gys1 antisense therapy rescues neuropathological bases of murine Lafora disease. \u003cem\u003eBrain\u003c/em\u003e \u003cstrong\u003e144\u003c/strong\u003e, 2985-2993 (2021).\u003c/li\u003e\n\u003cli\u003ePrats, C., Graham, T.E. \u0026amp; Shearer, J. The dynamic life of the glycogen granule. \u003cem\u003eJ Biol Chem\u003c/em\u003e \u003cstrong\u003e293\u003c/strong\u003e, 7089-7098 (2018).\u003c/li\u003e\n\u003cli\u003eLiu, Q.\u003cem\u003e et al.\u003c/em\u003e Glycogen accumulation and phase separation drives liver tumor initiation. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e184\u003c/strong\u003e, 5559-5576.e5519 (2021).\u003c/li\u003e\n\u003cli\u003eEscartin, C.\u003cem\u003e et al.\u003c/em\u003e Reactive astrocyte nomenclature, definitions, and future directions. \u003cem\u003eNat Neurosci\u003c/em\u003e \u003cstrong\u003e24\u003c/strong\u003e, 312-325 (2021).\u003c/li\u003e\n\u003cli\u003eGuttenplan, K.A.\u003cem\u003e et al.\u003c/em\u003e Neurotoxic Reactive Astrocytes Drive Neuronal Death after Retinal Injury. \u003cem\u003eCell Rep\u003c/em\u003e \u003cstrong\u003e31\u003c/strong\u003e, 107776 (2020).\u003c/li\u003e\n\u003cli\u003eCameron, E.G.\u003cem\u003e et al.\u003c/em\u003e A molecular switch for neuroprotective astrocyte reactivity. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e626\u003c/strong\u003e, 574-582 (2024).\u003c/li\u003e\n\u003cli\u003eLiddelow, S.A.\u003cem\u003e et al.\u003c/em\u003e Neurotoxic reactive astrocytes are induced by activated microglia. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e541\u003c/strong\u003e, 481-487 (2017).\u003c/li\u003e\n\u003cli\u003ePagliarini, D.J.\u003cem\u003e et al.\u003c/em\u003e A mitochondrial protein compendium elucidates complex I disease biology. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e134\u003c/strong\u003e, 112-123 (2008).\u003c/li\u003e\n\u003cli\u003eBrar, K.K.\u003cem\u003e et al.\u003c/em\u003e PERK-ATAD3A interaction provides a subcellular safe haven for protein synthesis during ER stress. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e385\u003c/strong\u003e, eadp7114 (2024).\u003c/li\u003e\n\u003cli\u003eZhao, Y.\u003cem\u003e et al.\u003c/em\u003e ATAD3A oligomerization causes neurodegeneration by coupling mitochondrial fragmentation and bioenergetics defects. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 1371 (2019).\u003c/li\u003e\n\u003cli\u003eZhao, Y.\u003cem\u003e et al.\u003c/em\u003e ATAD3A oligomerization promotes neuropathology and cognitive deficits in Alzheimer\u0026apos;s disease models. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 1121 (2022).\u003c/li\u003e\n\u003cli\u003eStapleton, D.\u003cem\u003e et al.\u003c/em\u003e Analysis of hepatic glycogen-associated proteins. \u003cem\u003eProteomics\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 2320-2329 (2010).\u003c/li\u003e\n\u003cli\u003eP\u0026aacute;nek, T., Eli\u0026aacute;\u0026scaron;, M., Vancov\u0026aacute;, M., Luke\u0026scaron;, J. \u0026amp; Hashimi, H. Returning to the Fold for Lessons in Mitochondrial Crista Diversity and Evolution. \u003cem\u003eCurr Biol\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, R575-r588 (2020).\u003c/li\u003e\n\u003cli\u003eGilquin, B.\u003cem\u003e et al.\u003c/em\u003e The AAA+ ATPase ATAD3A controls mitochondrial dynamics at the interface of the inner and outer membranes. \u003cem\u003eMol Cell Biol\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 1984-1996 (2010).\u003c/li\u003e\n\u003cli\u003eChi, Z.\u003cem\u003e et al.\u003c/em\u003e Histone Deacetylase 3 Couples Mitochondria to Drive IL-1\u0026beta;-Dependent Inflammation by Configuring Fatty Acid Oxidation. \u003cem\u003eMol Cell\u003c/em\u003e \u003cstrong\u003e80\u003c/strong\u003e, 43-58.e47 (2020).\u003c/li\u003e\n\u003cli\u003eChatterjee, A.\u003cem\u003e et al.\u003c/em\u003e MOF Acetyl Transferase Regulates Transcription and Respiration in Mitochondria. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e167\u003c/strong\u003e, 722-738.e723 (2016).\u003c/li\u003e\n\u003cli\u003eHu, D.\u003cem\u003e et al.\u003c/em\u003e TMEM135 links peroxisomes to the regulation of brown fat mitochondrial fission and energy homeostasis. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 6099 (2023).\u003c/li\u003e\n\u003cli\u003eIshihara, T., Ban-Ishihara, R., Ota, A. \u0026amp; Ishihara, N. Mitochondrial nucleoid trafficking regulated by the inner-membrane AAA-ATPase ATAD3A modulates respiratory complex formation. \u003cem\u003eProc Natl Acad Sci U S A\u003c/em\u003e \u003cstrong\u003e119\u003c/strong\u003e, e2210730119 (2022).\u003c/li\u003e\n\u003cli\u003eHayakawa, K.\u003cem\u003e et al.\u003c/em\u003e Corrigendum: Transfer of mitochondria from astrocytes to neurons after stroke. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e539\u003c/strong\u003e, 123 (2016).\u003c/li\u003e\n\u003cli\u003eMurphy, T.H. \u0026amp; Corbett, D. Plasticity during stroke recovery: from synapse to behaviour. \u003cem\u003eNat Rev Neurosci\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 861-872 (2009).\u003c/li\u003e\n\u003cli\u003eJoy, M.T.\u003cem\u003e et al.\u003c/em\u003e CCR5 Is a Therapeutic Target for Recovery after Stroke and Traumatic Brain Injury. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e176\u003c/strong\u003e, 1143-1157.e1113 (2019).\u003c/li\u003e\n\u003cli\u003eRawji, K.S., Gonzalez Martinez, G.A., Sharma, A. \u0026amp; Franklin, R.J.M. The Role of Astrocytes in Remyelination. \u003cem\u003eTrends Neurosci\u003c/em\u003e \u003cstrong\u003e43\u003c/strong\u003e, 596-607 (2020).\u003c/li\u003e\n\u003cli\u003eCarmichael, S.T. \u0026amp; Chesselet, M.F. Synchronous neuronal activity is a signal for axonal sprouting after cortical lesions in the adult. \u003cem\u003eJ Neurosci\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 6062-6070 (2002).\u003c/li\u003e\n\u003cli\u003eNguyen, H., Zerimech, S. \u0026amp; Baltan, S. Astrocyte Mitochondria in White-Matter Injury. \u003cem\u003eNeurochem Res\u003c/em\u003e \u003cstrong\u003e46\u003c/strong\u003e, 2696-2714 (2021).\u003c/li\u003e\n\u003cli\u003eZeisel, A.\u003cem\u003e et al.\u003c/em\u003e Molecular Architecture of the Mouse Nervous System. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e174\u003c/strong\u003e, 999-1014.e1022 (2018).\u003c/li\u003e\n\u003cli\u003eJoshi, A.U.\u003cem\u003e et al.\u003c/em\u003e Fragmented mitochondria released from microglia trigger A1 astrocytic response and propagate inflammatory neurodegeneration. \u003cem\u003eNat Neurosci\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 1635-1648 (2019).\u003c/li\u003e\n\u003cli\u003eRao, J.\u003cem\u003e et al.\u003c/em\u003e A novel role for astrocytic fragmented mitochondria in regulating morphine addiction. \u003cem\u003eBrain Behav Immun\u003c/em\u003e \u003cstrong\u003e113\u003c/strong\u003e, 328-339 (2023).\u003c/li\u003e\n\u003cli\u003eYun, S.P.\u003cem\u003e et al.\u003c/em\u003e Block of A1 astrocyte conversion by microglia is neuroprotective in models of Parkinson\u0026apos;s disease. \u003cem\u003eNat Med\u003c/em\u003e \u003cstrong\u003e24\u003c/strong\u003e, 931-938 (2018).\u003c/li\u003e\n\u003cli\u003eZhao, W., Johnston, K.G., Ren, H., Xu, X. \u0026amp; Nie, Q. Inferring neuron-neuron communications from single-cell transcriptomics through NeuronChat. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 1128 (2023).\u003c/li\u003e\n\u003cli\u003eSpezani, R., Marinho, T.S., Reis, T.S., Aguila, M.B. \u0026amp; Mandarim-de-Lacerda, C.A. Cotadutide (GLP-1/Glucagon dual receptor agonist) modulates hypothalamic orexigenic and anorexigenic neuropeptides in obese mice. \u003cem\u003ePeptides\u003c/em\u003e \u003cstrong\u003e173\u003c/strong\u003e, 171138 (2024).\u003c/li\u003e\n\u003cli\u003eParker, V.E.R.\u003cem\u003e et al.\u003c/em\u003e Cotadutide promotes glycogenolysis in people with overweight or obesity diagnosed with type 2 diabetes. \u003cem\u003eNat Metab\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 2086-2093 (2023).\u003c/li\u003e\n\u003cli\u003eAmbery, P.\u003cem\u003e et al.\u003c/em\u003e MEDI0382, a GLP-1 and glucagon receptor dual agonist, in obese or overweight patients with type 2 diabetes: a randomised, controlled, double-blind, ascending dose and phase 2a study. \u003cem\u003eLancet\u003c/em\u003e \u003cstrong\u003e391\u003c/strong\u003e, 2607-2618 (2018).\u003c/li\u003e\n\u003cli\u003eBoland, M.L.\u003cem\u003e et al.\u003c/em\u003e Resolution of NASH and hepatic fibrosis by the GLP-1R/GcgR dual-agonist Cotadutide via modulating mitochondrial function and lipogenesis. \u003cem\u003eNat Metab\u003c/em\u003e \u003cstrong\u003e2\u003c/strong\u003e, 413-431 (2020).\u003c/li\u003e\n\u003cli\u003eLin, Y.H.\u003cem\u003e et al.\u003c/em\u003e Ketone bodies promote stroke recovery via GAT-1-dependent cortical network remodeling. \u003cem\u003eCell Rep\u003c/em\u003e \u003cstrong\u003e42\u003c/strong\u003e, 112294 (2023).\u003c/li\u003e\n\u003cli\u003eDi Pino, G.\u003cem\u003e et al.\u003c/em\u003e Modulation of brain plasticity in stroke: a novel model for neurorehabilitation. \u003cem\u003eNat Rev Neurol\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 597-608 (2014).\u003c/li\u003e\n\u003cli\u003eMostany, R.\u003cem\u003e et al.\u003c/em\u003e Local hemodynamics dictate long-term dendritic plasticity in peri-infarct cortex. \u003cem\u003eJ Neurosci\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 14116-14126 (2010).\u003c/li\u003e\n\u003cli\u003eLi, S.\u003cem\u003e et al.\u003c/em\u003e GDF10 is a signal for axonal sprouting and functional recovery after stroke. \u003cem\u003eNat Neurosci\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 1737-1745 (2015).\u003c/li\u003e\n\u003cli\u003eReitmeir, R.\u003cem\u003e et al.\u003c/em\u003e Post-acute delivery of erythropoietin induces stroke recovery by promoting perilesional tissue remodelling and contralesional pyramidal tract plasticity. \u003cem\u003eBrain\u003c/em\u003e \u003cstrong\u003e134\u003c/strong\u003e, 84-99 (2011).\u003c/li\u003e\n\u003cli\u003eMolyneaux, B.J., Arlotta, P., Menezes, J.R. \u0026amp; Macklis, J.D. Neuronal subtype specification in the cerebral cortex. \u003cem\u003eNat Rev Neurosci\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 427-437 (2007).\u003c/li\u003e\n\u003cli\u003eGreig, L.C., Woodworth, M.B., Galazo, M.J., Padmanabhan, H. \u0026amp; Macklis, J.D. Molecular logic of neocortical projection neuron specification, development and diversity. \u003cem\u003eNat Rev Neurosci\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 755-769 (2013).\u003c/li\u003e\n\u003cli\u003eKokinovic, B. \u0026amp; Medini, P. Loss of GABA(B) -mediated interhemispheric synaptic inhibition in stroke periphery. \u003cem\u003eJ Physiol\u003c/em\u003e \u003cstrong\u003e596\u003c/strong\u003e, 1949-1964 (2018).\u003c/li\u003e\n\u003cli\u003eTang, X.\u003cem\u003e et al.\u003c/em\u003e Pyramidal and parvalbumin neurons modulate the process of electroacupuncture stimulation for stroke rehabilitation. \u003cem\u003eiScience\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 109695 (2024).\u003c/li\u003e\n\u003cli\u003eDuran, J., Saez, I., Gruart, A., Guinovart, J.J. \u0026amp; Delgado-Garc\u0026iacute;a, J.M. Impairment in long-term memory formation and learning-dependent synaptic plasticity in mice lacking glycogen synthase in the brain. \u003cem\u003eJ Cereb Blood Flow Metab\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 550-556 (2013).\u003c/li\u003e\n\u003cli\u003eDemetriadou, A.\u003cem\u003e et al.\u003c/em\u003e Mouse Stbd1 is N-myristoylated and affects ER-mitochondria association and mitochondrial morphology. \u003cem\u003eJ Cell Sci\u003c/em\u003e \u003cstrong\u003e130\u003c/strong\u003e, 903-915 (2017).\u003c/li\u003e\n\u003cli\u003eHossain, M.I., Roulston, C.L. \u0026amp; Stapleton, D.I. Molecular basis of impaired glycogen metabolism during ischemic stroke and hypoxia. \u003cem\u003ePLoS One\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, e97570 (2014).\u003c/li\u003e\n\u003cli\u003eZhang, L.Y.\u003cem\u003e et al.\u003c/em\u003e The Role of Astrocytic Mitochondria in the Pathogenesis of Brain Ischemia. \u003cem\u003eMol Neurobiol\u003c/em\u003e \u003cstrong\u003e61\u003c/strong\u003e, 2270-2282 (2024).\u003c/li\u003e\n\u003cli\u003eFarah, B.L.\u003cem\u003e et al.\u003c/em\u003e Hepatic mitochondrial dysfunction is a feature of Glycogen Storage Disease Type Ia (GSDIa). \u003cem\u003eSci Rep\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 44408 (2017).\u003c/li\u003e\n\u003cli\u003eMishra, K. \u0026amp; Kakhlon, O. Mitochondrial Dysfunction in Glycogen Storage Disorders (GSDs). \u003cem\u003eBiomolecules\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e (2024).\u003c/li\u003e\n\u003cli\u003eJiang, S.\u003cem\u003e et al.\u003c/em\u003e Starch binding domain-containing protein 1/genethonin 1 is a novel participant in glycogen metabolism. \u003cem\u003eJ Biol Chem\u003c/em\u003e \u003cstrong\u003e285\u003c/strong\u003e, 34960-34971 (2010).\u003c/li\u003e\n\u003cli\u003eDemetriadou, A., Drousiotou, A. \u0026amp; Petrou, P.P. The \u0026ldquo;sweet\u0026rdquo; side of ER-mitochondria contact sites. \u003cem\u003eCommunicative \u0026amp; Integrative Biology\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, e1329787 (2017).\u003c/li\u003e\n\u003cli\u003eGomes, L.C., Di Benedetto, G. \u0026amp; Scorrano, L. During autophagy mitochondria elongate, are spared from degradation and sustain cell viability. \u003cem\u003eNat Cell Biol\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 589-598 (2011).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-6650856/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6650856/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Ischemic stroke induces pathological glycogen deposition in astrocytes, but its role in post-injury neural dysfunction remains undefined. We reveal that glycogen-laden astrocytes in the ischemic penumbra undergo ATAD3A-dependent mitochondrial fragmentation via a stress granule-mediated mechanism, exacerbating neuronal injury and hindering functional recovery. Mechanistic studies demonstrate that glycogen aggregates sequester cytoplasmic HDAC3, enabling its translocation to mitochondria. There, HDAC3 deacetylates outer mitochondrial membrane protein ATAD3A, promoting oligomerization-driven mitochondrial fission. Astrocyte-specific ATAD3A ablation prevents stroke-induced synaptic disorganization, neural circuit disruption, and cognitive deficits. Therapeutically, pharmacologic or genetic exhaustion of astrocytic glycogen and HDAC3 inhibition reverse glycogen accumulation, rescue mitochondrial architecture/function, and restore synaptic plasticity and circuit reorganization, thereby acting synergistically to enhance post-stroke recovery. Our work identifies glycogen stress granules as pathogenic signaling hubs linking astrocytic metabolic stress to mitochondrial failure through compartmentalized HDAC3-ATAD3A crosstalk, and proposes a dual-target paradigm addressing both substrate overload and protein acetylation dynamics for stroke neurorestoration.","manuscriptTitle":"Astrocytic glycogen aggregates induce ATAD3A oligomerization mediated mitochondrial fragmentation and impede stroke recovery","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-03 06:50:44","doi":"10.21203/rs.3.rs-6650856/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"a53562ad-72d5-431a-9b9c-7fdbe858d944","owner":[],"postedDate":"February 3rd, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":62161415,"name":"Biological sciences/Neuroscience"},{"id":62161416,"name":"Biological sciences/Neuroscience/Glial biology"}],"tags":[],"updatedAt":"2026-02-06T13:01:01+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-03 06:50:44","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6650856","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6650856","identity":"rs-6650856","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}