Adult male C57BL/6 mice weighting 20-25g (6-8 weeks old) were purchased from Beijing Vital River Laboratory Animal Technology. All mice were housed in an environment with controlled light (12 h light/dark), temperature (temperature 23 ± 1°C) and humidity (humidity 55-60%) and provided ad libitum access to food and water. All experimental protocols were approved by the Animal Ethics Review Committee of The First Affiliated Hospital of the University of Science and Technology of China. The study was implemented according to the National Institute of Health Guide for the Care and Use of Laboratory Animals (NIH Publications No. 80-23, revised 1996). All mice were acclimated 1 week to the environment before use and randomized into groups. The experiment scheme is shown in Supplementary Figure S1 . A total of 466 male mice were used in this study. The animal usage and the excluded reason are shown in Supplementary Table S1 . At least 4 mice were analyzed for each data point.

The middle cerebral artery occlusion (MCAO) model was induced by occluding the MCA using the intraluminal filament method (24). In brief, mice were anesthetized with 2% isoflurane in O₂ (RWD Life Science, China). The right common carotid artery (CCA), internal carotid artery (ICA), and external carotid artery (ECA) were isolated through a midline cervical incision. The ECA was ligated, and a nylon suture with silicon (Beijing Cinontech Co., Ltd., China) was inserted through the ECA stump and next advanced into the ICA to occlude the origin of the middle cerebral artery (MCA). Mice underwent reperfusion after 90 min of occlusion. The body temperature was maintained at 37 ± 0.5°C by a homoeothermic blanket. To confirm the occlusion of MCA, mice were monitored for cerebral blood flow (CBF) with Laser Speckle Doppler Flowmetry (PeriCam PSI Z; Perimed, Sweden) contrast Imaging before, during, and after surgery ( Supplementary Figure S2 ). A decline in regional CBF ≥ 75% of baseline was considered as a successful occlusion. Sham-operated mice were anesthetized and received the same surgical procedures except that the MCA was not occluded.

The mouse BV2 microglial cells and mouse HT22 hippocampal neurons were purchased from the Procell (Wuhan, China). Cells were grown in DMEM (Gibco, USA) medium supplemented with 10% FBS (Gibco, USA) and 1% penicillin-streptomycin (Gibco, USA) solution in a 37°C, 5% CO₂ incubator. OGD/R was implemented according to our previous methods (24, 25). Briefly, cells were transferred to glucose-free and serum-free DMEM and then incubated in an anaerobic chamber equipped with Anaero Pack-Anaero (MITSUBISHI GAS CHEMICAL CO., INC. Japan). After cells were subjected to OGD for 2 h in the chamber, the medium was replaced by DMEM and returned to normal incubator for reoxygenation. Control microglia cells were cultured with normal oxygen-conditioned incubator for the same periods. For microglia-neuron cocultures, Transwell^®plates (0.4-μm pore size, Corning, MA, USA) were used. HT22 cells were seeded in the lower chamber of the Transwell plates and cultured together with microglia, which were pretreated and underwent OGD/R.

STING inhibitor C-176 (6.1, 12.2, 24.4ug/g, MedChemExpress, USA) or vehicle (1% DMSO+corn oil) were administered intraperitoneally at 30min after MCAO surgery (16, 26). The mtDNA (5mg/kg), extracted from liver tissue by Mitochondrial DNA Isolation Kit (ab65321, Abcam, USA), was intracerebroventricular injected (right lateral ventricle, bregma: 0.23mm posterior, 1mm lateral, and 2.25 mm deep) immediately after MCAO modeling (27). To inhibit STING, BV2 cells were treated with 1 µM C-176 (28). Cells were collected 6 h after mtDNA (100 ng/ml) stimulation and used for subsequent analyses (29). The same volume of PBS was used as the control treatment.

STING small interfering RNA (si-STING, sequence: 5’-CGAAAUAACUGCCGCCUCATT-3’) was used in our experiments according to a previous study (17). The 2′OMe+5′Chol+5′Cy5 modified STING siRNA (si-STING) and scrambled siRNA (si-NC) were synthesized by RiboBio (Guangzhou, China) for STING knockdown. After the mice were anesthetized with isoflurane, they were fixed with a stereotaxic apparatus (RWD Life Science, China). A 10µl stainless-steel microsyringe (Shanghai Gaoge Industry & Trade Co., Ltd., China) with siRNAs against STING (si-STING or si-NC, 1.5nmol in 3μl) was inserted into the right lateral ventricle (bregma: 0.23mm posterior, 1mm lateral, and 2.25 mm deep) at a flow rate of 300 nl/min. The injector was administered over 10 min and withdrawn slowly after the infusion, and MCAO surgical procedures were performed 48h after the infusion. After completion of the injection, the burr hole was sealed with bone wax and the incision was closed with sutures. For in vitro experiments, 1.25µl siRNA (20µM) and 3µl riboFECT™ CP Reagent (RiboBio, China) were mixed in 30µl riboFECT™ CP Buffer to generate a transfection mixture with a final concentration of 50nM, according to the manufacturer’s instruction. Then the 50nM transfection mixture of scrambled siRNA (si-NC) or interfering RNA (si-STING) was added into DMEM and cultured with BV2 microglia for 48 h. Immunofluorescent staining and western blot analysis were performed to examine the transfection efficiency.

Modified Neurological Severity Scores (mNSS) and Corner-turning test were performed to assess sensorimotor deficits and motor coordination at day 1, 3, 7, 14, and 21 after cerebral ischemia, as previously described (30). Behavioral function assessments were performed by two researchers who were blinded to the experiments.

Spatial learning and memory was investigated with the Morris Water Maze (MWM) test on days 22-28 after reperfusion (24, 31). Briefly, blind tests were performed on day 22 to exclude blind mice, in which mice were allowed to reach the platform with the flag when the platform was over the water. In the next 5 days, animals were trained to find the platform below the water in four trials. Five days later, the platform was removed and each mouse was explored to search the platform for 90s. The time spent in the target quadrant, the number of platform crossings, the escape latency to find the platform and the swim path length were recorded by the Smart video software (Panlab Harvard Apparatus, USA). The observer and recorder were blinded to animal grouping.

At 24 h after reperfusion, mice were anesthetized, and brains were quickly removed. The brains were frozen rapidly at -80°C with PBS for 6-8min, and coronally sliced in a brain mold and incubated in 2% 2,3,5-triphenyl tetrazolium chloride (TTC, Sigma Aldrich Inc., USA) solution for 10 min at 37°C (25). Then, the brain sections were photographed, and the infarct volume of each slice was determined using Image J software (NIH, USA) by an investigator blinded to the experimental design. To exclude the effects of brain edema, each slice infarct area was calculated as the volume of the contralateral hemisphere minus the non-infarcted volume of the ipsilateral hemisphere and summed to determine the whole infarct volume.

The brain water content was measured with the wet-dry method to evaluate the severity of brain edema at 24 h post-modeling (32). Brains were collected and quickly divided into the left hemisphere and the right hemisphere. Each hemisphere was weighed to get the wet weight and then weighed again at 105°C overnight to get the dry weight. The percentage of water content was calculated according to the following formula: [(wet weight-dry weight)/wet weight] ×100%.

The mice were deeply anesthetized and intracardially perfused with cold PBS and 4% paraformaldehyde (PFA). The brains were removed and fixed in 4% PFA at 4°C for 24 h, and then transferred into 30% sucrose solution until dehydration. Brains were embedded and frozen in the Tissue-Tek O.C.T. compound (Sakura Finetek, USA). The brains were cut into 15 μm or 20 μm frozen coronal sections with a Leica CM1950 cryostat for immunofluorescence staining. Sections were fixed with 4% PFA for 10 min, and then incubated with 10% donkey serum, 1% bovine serum albumin (BSA) and 0.1% Triton X-100 at room temperature for 1 h. Then the slices were incubated overnight at 4°C with indicated primary antibodies ( Supplementary Table S2 ). After three washes with PBS, the slices were incubated for 2 h at room temperature with secondary antibodies ( Supplementary Table S2 ). Coverslips were immunostained as brain slices did. Immunofluorescence images were acquired with an Olympus FV3000 fluorescence microscope (Japan). Quantification analysis of positive signals was performed with Image J software (NIH, USA) by an investigator blinded to the experimental design.

For NeuN and TUNEL co-staining, the frozen sections were first stained with NeuN antibody overnight at 4°C, followed by terminal deoxynucleotidyl transferase-mediated dUTP nick end-labeling (TUNEL) staining with One Step TUNEL Assay Kit (Beyotime, China), according to the manufacturer’s protocol. Three brain sections were examined per mouse, with each section containing three microscopic fields from the ischemic boundary zone. The means numbers of target cells were measured in sections per brain, by observers blinded to the experiments. Data were presented with the number of double-positive for TUNEL and NeuN neurons in the fields as cells/mm². Meanwhile, HT22 cells death was determined with LIVE/DEAD™ Viability/Cytotoxicity Kit (Thermo Fisher Scientific, USA) after microglia-neuron co-cultures. The percentage of live cells (green) compared with the dead cells (red) was used to assess cell death. Images were captured using a confocal laser scanning microscope (Olympus FV3000, Japan).

Fluoro-Jade C (FJC, Millipore, USA) was performed at day 1 and day 28 after MCAO induction for identifying degenerated neurons (32, 33). Briefly, brain frozen sections were sequentially immersed in 1% NaOH/80% ethanol solution, 70% ethanol, and 0.06% potassium permanganate solution. Then, the sections were incubated with 0.0001% solution of FJC. Three brain slices were examined per mouse, and the number of FJC-positive neurons was quantified in three different fields of view for each section. Data were presented as TUNEL staining did.

Total RNA from cells or tissue was extracted with TRIzol reagent (Invitrogen, USA) according to the manufacturer’s instruction and was reverse-transcribed into cDNA using HiScript III RT SuperMix for qPCR Kit (Vazyme, China). Then quantitative real-time PCR was performed using quantitative PCR (Light Cycler 96 System, Roche, China) with corresponding primers ( Supplementary Table S3 ) in the presence of a fluorescent dye (AceQ qPCR SYBR Green Master Mix). The levels of mRNA were normalized in relevance to GAPDH.

Total lysates of peri-infarct tissue or BV2 microglia were prepared using RIPA lysis buffer (Beyotime, China). Protein concentration was quantified by NanoPhotometer-N50 (IMPLEN GMBH, Germany). Equal amounts of protein were loaded and separated by 8-12% SDS-PAGE gel, then electrophoresed and transferred onto PVDF membranes (Millipore, USA). PVDF membranes were blocked with 5% BSA and incubated with primary antibodies overnight at 4°C and then incubated with appropriate HRP-conjugated secondary antibodies at room temperature for 1 h. The used antibodies were listed in Supplementary Table S2 . Membranes were then incubated with ECL reagents (Willget biotech, China) before visualization using Azure 500 (Azure Biosystems, USA). β-Tubulin was served as the internal control. The immunoblots were analyzed by Image J software (NIH, USA).

All statistical tests were performed using SPSS 22.0 software (IBM, Armonk, NY, USA). Differences between groups were evaluated by the one-way ANOVA followed by Tukey’s post hoc test. The escape latency and swimming path length in MWM tests were analyzed using two-way ANOVA followed by Tukey’s post hoc test. All data were presented as mean ± SD. All tests were considered statistically significant at P < 0.05.

We first investigated the expression change of STING in MCAO mice. Western blot results showed that the protein levels of cGAS, p-STING, and STING were significantly increased with time following MCAO ( Figure 1A ). BV2 microglia OGD/R model was used to mimic the in vitro conditions of I/R injury. Immunoblots showed that the expression levels of cGAS, p-STING, and STING in BV2 cells were significantly increased after OGD/R, and continuously increased as the reperfusion time prolongs ( Figure 1B ). Double immunostaining of STING with cell markers (Iba1, GFAP, or NeuN) demonstrated that STING was mainly expressed in microglia cells, rather than neurons or astrocytes and was increased at 1d after MCAO ( Figure 1C and Supplementary Figure S3 ). Furthermore, STING was co-stained with Iba1 in BV2 cells. STING immunofluorescence signal was enhanced at 24h after reoxygenation ( Figure 1D ). Thus, our data indicated that STING was activated after I/R injury.

To investigate the function of STING in the pathophysiological process of ischemic stroke, we intraperitoneally injected mice with C-176 to block STING. TTC staining showed that C-176 administration notably decreased brain infarction and brain edema in a dose-dependent manner 1d post-MCAO ( Figures 2A–C , for brain infarction: P = 0.0344, P < 0.001 and P < 0.001, for brain edema: P = 0.0044, P < 0.001 and P < 0.001). The dosage of 12.2 and 24.4 ug/g were more effective than the dosage of 6.1 ug/g on brain infarction and cerebral edema 1d post-MCAO ( Figures 2A–C ). However, there were no significant differences between 12.2 ug/g and 24.4 ug/g dose on stroke outcomes ( Figures 2A–C ). Therefore, 12.2 ug/g dose was chosen for subsequent experiments.

We next investigated whether blockade of STING could rescue I/R-induced neuronal injury. TUNEL/NeuN staining was performed to assess neuronal apoptosis at 1d after reperfusion and found a significant decrease (~25%) in the number of apoptotic neurons in the MCAO+C-176 mice compared to the MCAO+vehicle mice ( Figures 2D, E , P = 0.0023). We evaluated neuronal degeneration by FJC staining and found a decrease of ~49% in the number of neuronal degeneration in MCAO+C-176 mice at 1d post-modeling compared with that in MCAO mice ( Figures 2F, G , P < 0.001).

The mNSS score was prominently improved in C-176-administered MCAO mice compared to the vehicle-treated MCAO mice on day 7, 14, and 21 post-stroke ( Figure 3A , P = 0.0035, P < 0.001 and P = 0.0047). Similarly, the C-176-treated group exhibited a significant improvement in sensorimotor function with Corner test compared to the vehicle-treated group at 7 and 14 days after MCAO ( Figure 3B , P = 0.0412, P = 0.0156). Cognitive function was assessed by the MWM test on days 22-28 after reperfusion. As shown in Figures 3C, D , the escape latency and the path length for the mice to find the hidden platform in the MCAO group were inferior to that of Sham-operated mice, suggesting that I/R injury induced severe learning and memory impairments. C-176-treated MCAO mice displayed a better performance with decreased escape latency and shorter path length on days 3 to 5 (for escape latency: P = 0.008, P < 0.001 and P = 0.0012, for path length: P = 0.0138, P < 0.001 and P < 0.001). Moreover, in the probe phase, treatment with C-176 conspicuously increased the crossovers of the platform and the time spent in the target quadrant compared to the MCAO+vehicle group ( Figures 3E–G , P = 0.0096 and P = 0.0073).

We then used FJC staining to test whether STING could affect neuronal degeneration for a long time after ischemic stroke. Robust FJC staining in CA1, CA2, and CA3 regions of the hippocampus were found at 28 d after MCAO. The density of FJC-positive cells was increased after MCAO, while C-176 treatment significantly decreased the number of FJC-positive neurons in CA1, CA2 and CA3 regions of the hippocampus ( Figures 3H, I , P = 0.0018, P < 0.001 and P = 0.0468). Taken together, blockade of STING may have substantially advantageous effects on long-term behavioral outcomes following ischemic stroke.

Having demonstrated that STING modulates functional outcomes after stroke, we sought to explore whether STING could affect post-ischemic inflammation through microglial polarization. We found that the protein levels of cGAS, p-STING, STING, p-p65, and p-IRF3 were increased substantially in MCAO mice and OGD/R microglia; whilst the STING-related proteins expression levels were remarkably decreased with C-176 administration ( Figures 4A, B ). Moreover, immunoblots showed that the levels of the M1-related marker iNOS decreased and the M2-related marker Arginase-1 increased with C-176 treatment 3d after MCAO and 24h after OGD/R ( Figure 4C , both P < 0.001, Figure 4D , P = 0.0133 and P = 0.0018). Real-time PCR analysis indicated that the levels of M1 phenotype factors (TNF-α, iNOS, IL-1β, IL-6) and the M2 phenotype factors (IL-10 and Arg-1) increased at 3d after reperfusion ( Figure 4E , P < 0.001, P < 0.001, P < 0.001, P < 0.001, P < 0.001 and P = 0.006). However, C-176 treatment notably mitigated the expression of M1 phenotype related genes and markedly enhanced the expression of M2 phenotype related genes ( Figure 4E , all P < 0.001). Consistent with observation in vivo, C-176 incubation markedly upregulated the expression level of M2 phenotype factors and notably suppressed the levels of M1 phenotype factors in BV2 microglia after 24h reoxygenation ( Figure 4F , all P < 0.001).

We then investigated the effects of STING on microglial phenotype transition. CD16/32 and CD206 staining were used to examine the polarization of microglia in the peri-infarct cortex. As depicted in Figures 4G, H , the number of CD16/32 and Iba1 double-positive M1-like cells was significantly increased on day 3 after MCAO when compared with the Sham group, which was markedly decreased in the MCAO+C-176 group (P = 0.0019). On the other hand, the number of CD206 and Iba1 double-positive M2-like cells was further increased by C-176 administration at 3 days after MCAO ( Figures 4G, H , P < 0.001). Coincident with the observation on day 3 after MCAO, C-176 treatment also notably decreased the CD16/32-positive microglia and increased the CD206-positive microglia on day 7 after MCAO ( Supplementary Figure S4 , P = 0.0210 and P = 0.0007). We further performed immunostaining for the BV2 cells to evaluate the effects of C-176 on the M1/M2 polarization state at 24h after OGD/R. Blockade of STING reduced the number of CD16/32⁺ microglia cells, and increased the number of CD206⁺ microglia cells ( Figures 4I, J , P = 0.0074 and P < 0.001). OGD/R induced a substantial increment of neuronal death as determined by quantification of live/dead staining, while incubation with C-176 maximally preserved neuronal viability ( Supplementary Figure S5 , P = 0.0026). These results demonstrated that C-176 alleviated cerebral I/R injury by promoting the microglial polarization to M2 state.

To further verify the role of STING post-MCAO, siRNA was administrated i.c.v to silence STING. Double immunostaining images at the siRNA injection site and the surrounding area showed that Cy5-conjugated si-STING was co-stained with microglia ( Figure 5A ). And we found that the levels of p-STING and STING were suppressed by si-STING ( Figures 5B, C , both P < 0.001). Similarly, si-STING transfection could effectively knockdown p-STING and STING expression in BV2 microglia 24h after OGD/R ( Supplementary Figure S6 , P = 0.0099 and P = 0.0087). These results suggested that si-STING could transfect into microglia and knockdown of STING expression.

Genetic knockdown of STING decreased the infarct volume and brain edema, and alleviated neurological deficits post-MCAO ( Figures 5D–G , P < 0.001, P = 0.0011 and P = 0.0207). Moreover, administration with si-STING significantly suppressed the mRNA level of M1 phenotype markers (TNF-α, iNOS, IL-1β, and IL-6) 1d post-MCAO when compared to the MCAO+si-NC group ( Figure 5H , P < 0.001, P = 0.0291, P < 0.001 and P < 0.001). We also found that silence STING upregulated the expression of M2 phenotype mRNA markers (IL-10 and Arg-1), while injection si-NC had no such effects ( Figure 5I , both P < 0.001).

Recently, it has been reported that mtDNA could release into the neuronal cytoplasm and participate in the pathogenesis of amyotrophic lateral sclerosis (ALS) (34). As a potent DAMPs, mtDNA is closely involved in the immune process after stroke (35). To this end, we triple-labeled stained against dsDNA, HSP60 (a mitochondrial marker), and Iba1 to assess whether mtDNA could leak into the microglial cytoplasm after I/R injury. Immunofluorescence and 3D-reconstructed images revealed that quantities of mtDNA were released into microglial cytoplasm following MCAO, which were alleviated by treatment with C-176 ( Figures 6A, B , P = 0.0016). A similar trend was found in the cell experiments. Incubation with C-176 remarkably reduced the level of mtDNA, which was released into microglial cytoplasm induced by OGD/R ictus ( Figures 6D, E , P < 0.001). I/R injury also induced the levels of mtDNA mRNA markers (UUR, COXI, ND1, and COX3) to noticeably increase, and these increments were all reversed by C-176 injection ( Figure 6C , P = 0.0030, P < 0.001, P < 0.001 and P < 0.001).

mtDNA has been found to be capable of triggering STING signaling (21). We next sought to explore the role of microglial mtDNA in the STING-mediated I/R injury. Mice were received an intracerebroventricular injection of 5 mg/kg mtDNA. Compared with the vehicle-treated MCAO mice, mtDNA injection could markedly enhance the protein levels of cGAS, p-STING, STING, p-p65 and p-IRF3 1d after MCAO, which were alleviated by C-176 administration ( Figures 7A, B ). A similar tendency was observed in BV2 cells that C-176 incubation notably reversed the increments of STING-related proteins induced by mtDNA treatment ( Supplement Figure S7 ). In addition, aggravated brain infarction and brain edema were observed as a result of mtDNA administration in MCAO mice, which was reversed by C-176 administration ( Figures 7C–E , both P < 0.001). mtDNA injection remarkably worsened neurobehavioral performance in MCAO mice compared to the MCAO+vehicle group throughout the testing periods ( Figure 7F , P = 0.0301, P = 0.0317 and P = 0.0067). The mtDNA+C-176 mice displayed significantly better neurological performance than the mice in the mtDNA group on days 7 to 21 post-MCAO ( Figure 7F , P < 0.001, P < 0.001 and P = 0.0036). As illustrated in Figures 7G, H , MCAO mice injected with mtDNA induced the increment of TUNEL-positive cells at 1d after reperfusion, which was alleviated by treatment of C-176 (P < 0.001).

We further evaluated the effects of STING activation, which is induced by mtDNA injection, on microglial polarization in ischemic stroke. Immunofluorescent staining results showed that mtDNA treatment increased the numbers of Iba1⁺CD16/32⁺ pro-inflammatory microglia and reduced Iba1⁺CD206⁺ anti-inflammatory microglia at 3d after MCAO compared with vehicle-treated mice ( Figures 7I–L , P < 0.001 and P = 0.0040). However, co-administration of mtDNA and C-176 could suppress the number of the M1-specific marker CD16/32 and increase the number of the M2-specific marker CD206 in microglia after MCAO attack ( Figures 7I–L , P < 0.001 and P = 0.0040). These results suggested that mtDNA could trigger microglial phenotype shift towards the M1 modality through activation of STING, resulting in neurological function deterioration following ischemic stroke.