Male Sprague-Dawley rats (230–250 g) were provided by the Animal Experiment Center of Xuzhou Medical University. The animals were housed in 12 h light/dark cycle conditions with controlled temperature and humidity and were supplied adequate food and water before the experiment. The animal experiment was conducted in accordance with the national and institutional guidelines on ethics and biosafety and has been conducted after the approval of the Local Animal Care Committees (License ID: 201907W079). All the animal experiments were approved and regulated by the Committee on the Use of Live Animals in Teaching and Research of Xuzhou Medical University.

In total, 30 rats were used to isolate serum exosomes for the whole study. About 7ml whole blood could be collected from one sham rat via the abdominal aorta into a vacutainer. The whole blood was allowed to clot at room temperature for 30 min. The serum (1–2 ml) was obtained by centrifugation at 2,500 g for 10 min. The serum was ultracentrifuged at 100,000 g for 60 min (Beckman Optima L-100XP, Beckman, United States) and the pellets were resuspended in phosphate-buffered saline (PBS) and then ultracentrifuged at 150,000 g for another 90 min. The pellets were resuspended in 1 ml PBS and passed through the 0.22 μm filter (Millipore, SLGPR33RB) to obtain sterile exosomes. The protein concentration of exosomes was determined with bicinchoninic acid (BCA) commercial kit (Vazyme, China) as manufacturer’s instruction. The protein concentration of 1 ml exosomes is 0.1–0.2 mg/μl. The exosomes were aliquoted and stored at –80°C until use.

The serum exosomes (labeled as Con-exo) were observed using transmission electron microscopy (TEM) to identify the morphology. The size and concentration distribution profile of serum exosomes were analyzed using a Multiple-Laser ZetaView^® f-NTA Nanoparticle Tracking Analyzers (Particle Metrix, Germany). The exosome surface markers were confirmed by Western blotting using antibodies against CD9 and TSG101.

Focal cerebral ischemia was induced by the intraluminal suture MCAO method (Longa et al., 1989; Gong et al., 2015). Briefly, after anesthesia with inhalation of 4% isoflurane, the left common carotid artery (CCA), internal carotid artery (ICA), and external carotid artery (ECA) were exposed through a midline incision of the neck. A 3–10 silica gel coated with nylon suture was used as an embolus and inserted to the origin of MCA via the ECA to block MCA for 2 h and then the suture was withdrawn for 24 h reperfusion. In the sham group, the suture was inserted 5 mm from the incision and no cerebral ischemia was induced. After the operation, animals were transferred into an intensive care chamber with the maintained temperature at 37°C.

The penetration ability of Con-exo crossing the BBB was evaluated using 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (Dil) labeled Con-exo. The label of Con-exo with Dil was performed, as previously described (Elashiry et al., 2021). Briefly, Con-exo was incubated with Dil (C1036, Beyotime, China) at 37°C in the dark for 15 min. After centrifugation at 100,000 g for 60 min, the pellet was resuspended and washed with PBS twice. Then, Dil-labeled Con-exo was intravenously injected to sham and MCAO rats. The rats were sacrificed after injection for 0.5, 6, and 24 h, respectively. The rats were perfused with 4% paraformaldehyde (PFA) by cardiac transfusion. Then, the brains were collected and fixed in 4% PFA. After dehydration with 30% sucrose solution, the brains were embedded with Tissue-Tek^® OCT Compound and prepared for cryosection. After cryosection, the slides were captured under fluorescence microscope (Olympus BX51, Japan). The fluorescence intensity of Dil was analyzed with Image J software (Maryland, United States).

For 2,3,5-triphenyltetrazolium chloride (TTC) staining experiments, 25 rats were totally used. Serum exosomes (400, 800, and 1,600 μg/kg) were injected intravenously after ischemia for 2 h, followed with reperfusion for 24 h. The optimal dosage of exosomes (800 μg/kg) was chosen from TTC staining of rat focal ischemia/reperfusion injury model for the subsequent experiments. Then, 21 rats were randomly divided into the three groups: sham (labeled as control, n = 5), MCAO 2 h/reperfusion 24 h injected with equal volume of PBS (labeled as MCAO, n = 8), and MCAO 2 h/reperfusion 24 h plus 800 μg/kg exosomes (labeled as Con-exo, n = 8). The MCAO rats were only subjected to Con-exo injection once during the experiment.

Infarct volume was determined by TTC (A610558, Shanghai Sangon, China) staining. After 2 h of ischemia and reperfusion for 24 h, the animals were sacrificed and the brains were quickly dissected. The brains were placed at –20°C for 30 min. Then, the brain was cut into 2 mm coronal sections and immersed in 2% TTC solution for 30 min at 37°C. The sections were preserved in 4% PFA solution overnight and then photographed. Infarct volume in the hemispheric lesion area was calculated by summation of unstained areas of all the slices and multiplied by the slice thickness (2 mm) and quantitatively analyzed with Image J software (Image J software, Maryland, United States). Relative infarct volume percentage was calculated with following formula: (uninfarct hemisphere area-infarct hemisphere uninfarct area)/uninfarct hemisphere area × 100.

The Modified Neurological Severity Score (mNSS) was assessed to evaluate the animal neurological deficit according to previous study (Chen et al., 2001). Neurological function was graded on the scales from 0 to 18 (normal score, 0; maximal deficit score, 18). The mNSS includes motor, sensory, reflex, and balance tests. The mNSS details were shown in Supplementary File. Tests were carried out by the researchers who were blind to the experimental design.

Rat spontaneous locomotor activity was measured using an open-field apparatus (50 × 50 × 50 cm, O’Hara and Corporation Ltd., Tokyo, Japan). Each rat was placed in the center of the open-field apparatus. The center zone was defined as a square, 10 cm away from the wall. The total distance moved, the distance moved in center, and the time spent in center by each animal were recorded for 5 min with a video imaging system (ANY-maze, Stoelting, United States), as previously described (Yoshizaki et al., 2016).

Golgi-Cox staining was performed to visualize dendritic spine structure in superficial and deep cortical layer neurons using the FD Rapid GolgiStain Kit (FD Neurotechnologies Incorporation, United States). In brief, rats were deeply anesthetized and intracardially perfused with saline. The whole brain was immersed in the A + B solution from the kit for 18 days in the dark at room temperature and then transfer the brain into the C solution from the kit for 72 h in the dark at room temperature. Coronal sections were cut at 100 μm and stained according to FD Rapid GolgiStain’s protocol. The images were captured by VS120 virtual slide microscope (Olympus, Tokyo, Japan).

Evans blue (EB) was used to analyze the BBB permeability, as previously described (Naderi et al., 2015). In brief, the rats were anesthetized and injected intravenously with 2% EB (Beijing Yinuokai Technology Corporation Ltd., 3 ml/kg) 1 h before sacrifice. Rats were perfused with saline 24 h after MCAO to remove circulating EB. Brains were harvested and cut into 2 mm slices and the digital photos of EB extravasation were captured. Then, the slice was divided into ischemic and non-ischemic, weighed and stored at –80°C until use. Slices were homogenized in cold PBS and homogenized tissue was soaked up and further homogenized with the same volume of 50% trichloroacetic acid (Sigma-Aldrich). Supernatants were collected after centrifuging at 15,000 rpm for 20 min. Optical density (OD) values of supernatants were measured at 620 nm with a microplate reader (Bio-Rad, Hercules, California, United States). The amount of extravagated EB dye was quantified as microgram per gram brain tissue.

For TUNEL and CD31 staining in striatum, the brain cryosections (slice thickness 20 μm) were prepared, as previously described (Yang et al., 2012). Briefly, rats were intracardially perfused with saline, followed by 4% PFA perfusion. Then, the brains were immersed in 30% sucrose solution. After dehydration, the brains were embedded in Tissue-TEK OCT Compound (Sakura, United States) and sectioned using freezing microtome (Leica, CM1950). TUNEL staining was performed in accordance with the instructions in the TUNEL Detection Kit (Cat#: PF00009, Proteintech, United States). Mouse antirat CD31 antibody (1:1,000, ab64543, Abcam) and goat antimouse Alexa Fluor 488-labeled secondary antibody (1:1,000; Cat# A11001, Thermo Fisher Scientific) were successively incubated. Then, 4′,6-diamidino-2-phenylindole (DAPI) staining was performed for 20 min and the nucleus was observed under a fluorescence microscope (Olympus BX51, Japan) and the TUNEL/CD31 double-positive cells were counted from ten different fields in each group using Image J software (Maryland, United States).

For cell apoptosis detection, TUNEL detection kit (Cat#: PF00006, Proteintech, United States) was used. The TUNEL⁺ cell percentage was calculated by TUNEL⁺ cell number divided by total cell number per field.

The bEnd.3 cell line, immortalized mouse brain endothelial cells, was grown in Dulbecco’s Modified Eagle’s Medium (DMEM) (with 4,500 mg/l D-glucose, 110 mg/l sodium pyruvate, 3,700 mg/l sodium bicarbonate, 584 mg/l L-glutamine, 80 units/ml of penicillin, and 80 μg/ml of streptomycin; KeyGEN BioTECH, Nanjing, China) supplemented with 10% fetal bovine serum (FBS) (Gibco, New York, United States). The bEnd.3 cells were maintained in a humidified incubator at 37°C with 5% CO₂ and 95% air. In order to simulate an ischemic stroke model in vitro, bEnd.3 cells were subjected to OGD/R (Jiang et al., 2000). For OGD/R, the medium was changed to glucose-free DMEM and cultured in an O₂/CO₂ (1.2%/5%) trigas incubator. After 9 h, the cells were replaced with the DMEM medium supplemented with 10% FBS and cultured in O₂/CO₂ (21/5%) for 24 h (reoxygenation). For cell experiments, the cells were divided into the three groups: control, OGD/R, and OGD/R plus 50 μg/ml Con-exo (labeled as Con-exo).

The ischemic cerebral cortex was dissected and brain tissue (or serum exosomes) was lysed in radio immunoprecipitation assay (RIPA) buffer with 1X protease and phosphatase inhibitor cocktail. The protein concentration was determined using Bradford protein assay (BCA Protein Assay Kit, Vazyme, China) and normalized for protein content. The lysates were loaded on sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gels and subsequently transferred onto a polyvinylidene difluoride (PVDF) membrane. The PVDF membrane was shaken slowly in the blocking solution (5% milk in Tris-buffered saline with 0.1% Tween 20) at room temperature for 1 h. Then, the membrane was incubated with the diluted primary antibody overnight at 4°C. The secondary antibody was incubated at room temperature for 1 h. The bands were visualized using an enhanced chemiluminescence reagent (KeyGen Biotechnology, China) and analyzed using the Image J software.

Primary antibodies include rabbit anti-CD9 (1:1,000, Proteintech, 20579-1-AP); rabbit anti-TSG101 (1:1,000, Proteintech, 28283-1-AP); rabbit anticleaved caspase-3 (1:1,000, Cell Signaling Technology, 9661); rabbit anti-Bcl-2 (1:1,000, Abcam, ab196495); rabbit anti-Bax (1:1,000, Abcam, ab32503); rabbit anti-ZO-1 (1:1,000, Abcam, ab190085); rabbit anti-MMP-9 (1:1,000, Abcam, ab76003); rabbit anticlaudin-5 (1:1,000, Abcam, ab172968); rabbit anti-LC3B (1:1,000, Sigma-Aldrich, L7543); rabbit anti-SQSTM1/p62 (1:1,000, HuaBio, R1309-8); mouse anti-Akt (1:1,000, Proteintech, 51077-1-AP); rabbit anti-p-Akt (1:1,000, Cell Signaling Technology, S473); and mouse anti-α-tubulin (1:10,000, Proteintech, 66031-1-lg). Secondary antibodies included horseradish peroxidase (HRP)-conjugated antirabbit (1:8,000, Proteintech, SA00001-2) or antimouse immunoglobulin G (IgG) secondary antibodies (1:8,000, Proteintech, SA00001-1).

Briefly, cells were harvested after treatment with OGD/R or OGD/R combined with Con-exo, fixed with ice-cold 2.5% glutaraldehyde in PBS (pH 7.3) for 2 h, postfixed in 1% osmium tetroxide, dehydrated in gradient ethanol series (50–100%) and acetone, and embedded in Epon. The sections were stained with 3% lead citrate-uranyl acetate and examined with a TEM (Hitachi, Japan).

All the experimental data are shown as means ± SE and analyzed by GraphPad Prism version 8 software (GraphPad Software Incorporation, La Jolla, California, United States). The data for the fluorescence intensity of Dil and spine density were counted by Image J software. The Dil-Con-exo penetration experiment was analyzed by two-way ANOVA. Except that, all the statistical analysis adopted one-way ANOVA; Tukey’s post hoc test was used for comparison between each group (α = 0.05); P < 0.05 indicated that the difference was statistical significant.

The serum exosomes (Con-exo) were separated from the blood of healthy rats by gradient centrifugation and ultracentrifugation (Figure 1A). TEM examination indicated Con-exo with typical round morphology (Figure 1B). Exosome-specific surface markers such as CD9 and TSG101 were examined by immunoblotting analysis in exosome lysate and the supernatant from last step of filtration was used as negative control (Figure 1C). The concentration and size distribution of Con-exo were quantified by nanoparticle tracing analysis (NTA). The NTA showed that the particle diameter of the Con-exo ranged from 80 to 200 nm. The distribution width of Con-exo was as following: the X10 was 68.7 ± 10.5 nm, the X50 was 109.6 ± 9.4 nm, and the X90 was 189.8 ± 14.2 nm. The mean diameter of Con-exo was 108.1 ± 15.7 nm (n = 3, Figure 1D).

The fluorescence intensity (FI) of Dil in striatum was defined as the ability of Con-exo penetrating BBB. The intensity was quantified in control and MCAO rats at 0.5, 6, and 24 h, respectively. The FI increased over time in MCAO rats. However, compared with the control group, the FI in the MCAO group was strikingly higher at 24 h (compared with MCAO, P < 0.05; Figures 2A,B). Then, the rats were subjected to MCAO or sham surgery, with immediate intravenous administration of 400, 800, or 1,600 μg/kg of Con-exo. Rats that received 800 μg/kg Con-exo were found to have significantly reduced infarct volumes compared with PBS-treated animals for 24 h after stroke by TTC staining (compared with MCAO, P < 0.0001; Figures 2C,D). Rats that received 800 μg/kg Con-exo treatment at the onset of stroke had significantly improved neurobehavioral scores at 24 h after stroke (P < 0.0001) compared with their PBS-treated counterparts (Figure 2E). Con-exo-treated rats had significantly longer total moving distance, less staying time, and shorter moving distance in the center in open field test (compared with MCAO, ***P < 0.001, **P < 0.01, P<0.001; Figures 2F–H). The dendritic spines are the postsynaptic terminals that mediate synaptic transmission and plasticity and effects on the sensorimotor and cognitive deficits of ischemic stroke. We used Golgi-Cox staining to explore the effects of Con-exo on ischemia-induced structural alterations in neuronal dendrites and dendritic spines. The neurons in infarct area lost their dendrites compared with those in peri-infarct area (Figure 2I). However, Con-exo remarkably protected the neuronal spine density against ischemic injury (compared with MCAO, P < 0.01; Figures 2J,K).

Next, we assessed the permeability of the BBB by measuring EB leakage to explore the neuroprotective mechanism of serum exosome in ischemic brain injury. EB leakage assay revealed that Con-exo obviously reduced the BBB permeability compared with MCAO rats injected with PBS (P < 0.0001; Figures 3A,B). To investigate the mechanisms of serum exosomes acting on the BBB modulation, we measured the expression of tight junction proteins, claudin-5 and ZO-1, in ischemic brains. Con-exo remarkably reversed claudin-5 and ZO-1 expression compared with the MCAO group (P < 0.05; Figures 3C,E,F). As MMP-9 was the primary proteolytic enzyme, which was reported to degrade claudin-5 and ZO-1 after ischemic stroke, Con-exo significantly inhibited the level of MMP-9 compared with the MCAO group (P < 0.01; Figures 3C,D). These data indicated that Con-exo reduced BBB leakage by preserving tight junction proteins in ischemic brains, at least partially owing to the inhibition of MMP-9.

As the primary cell type of the BBB, cerebrovascular endothelial cell survival is responsible for the BBB integrity. In order to explore the effect of Con-exo on endothelial cell apoptosis after ischemia injury, first, we used TUNEL staining to observe the apoptosis in the ipsilateral cerebrum and simultaneously, endothelial cell marker CD31 was stained for tracing endothelial cell apoptosis. In the MCAO group, the co-localization of TUNEL-positive and CD31-positive cells were obviously increased, compared to the control group. Con-exo treatment significantly reduced the number of TUNEL⁺/CD31⁺ cells compared with the MCAO group (P < 0.0001, Figures 4A,B). Compared to the control group, the OGD/R group exhibited significantly increased apoptotic cells. Con-exo treatment reduced the apoptotic cells in bEnd.3 cells (P < 0.01, Figures 4C,D). In addition, we analyzed the expression of the apoptosis-related proteins Bcl-2, Bax, and cleaved caspase-3 in the ipsilateral cortex. We found that the ratio of Bcl-2/Bax decreased and cleaved caspase-3 increased in the MCAO group; Con-exo treatment reversed the ratio of Bcl-2/Bax (P < 0.01, Figures 4E,G) and the increase of cleaved caspase-3 (P < 0.01, Figures 4E,H). We also found that Con-exo could significantly increase the phosphorylation of Akt (Ser473) (P < 0.01, Figures 4E,I). Consistently, we obtained the similar results in bEnd.3 cells administrated with Con-exo under OGD/R context (P < 0.001, Figures 4F,J). These results demonstrated that Con-exo could reduce endothelial cell apoptosis induced by ischemic injury.

Recent evidence indicates that autophagy is involved in claudin-5, occludin, and ZO-1 degradation after ischemic stroke and we further examined whether serum exosomes preserved tight junction proteins through autophagy. First, we examined the effect of Con-exo on the expression of autophagic proteins in ischemic brains. Con-exo treatment reduced the ratio of LC3B-II/LC3B-I compared to the MCAO group (P < 0.01, Figures 5A,C). During the process of autophagy, SQSTM1/p62 links LC3B-II with ubiquitin moieties on ubiquitinated proteins, which are degraded together. Hence, SQSTM1/p62 expression was reduced in the MCAO group; consistently, SQSTM1/p62 expression was reversed by Con-exo compared with the MCAO group (P < 0.01, Figures 5A,D). These findings were further confirmed in bEnd.3 cells (P < 0.01, P < 0.05, Figures 5B,E,F). TEM revealed that the autolysosomes were remarkably induced after OGD/R, while Con-exo obviously reduced autolysosomes production (Figure 5G).