Blocking A1 astrocyte conversion with semaglutide attenuates blood-brain barrier disruption in mice after middle cerebral artery occlusion

Qi Zhang Shanghai Jiao Tong University Chang Liu Shanghai Jiao Tong University Rubing Shi Shanghai Jiao Tong University Huimin Shan Shanghai Jiao Tong University Lidong Deng Shanghai Jiao Tong University Tingting Chen Shanghai Jiao Tong University Zhijun Zhang Shanghai Jiao Tong University Guo-Yuan Yang Shanghai Jiao Tong University Yongting Wang Shanghai Jiao Tong University Yaohui Tang (  yaohuitang@sjtu.edu.cn ) Shanghai Jiao Tong University https://orcid.org/0000-0002-9603-2650


Introduction
Ischemic stroke is one of the leading causes of morbidity and mortality worldwide. Many pathological processes are involved in stroke progression, including blood-brain barrier (BBB) dysfunction, in ammation [1], excitotoxicity, oxidative stress, neuronal loss, and glial activation. Glial cells, the largest cell population in the central nervous system (CNS) [2], have received extensive attention for their activation in response to CNS injury and subsequent conversion to different phenotypes, including neurotoxic and neuroprotective properties.
Astrocytes play critical roles in maintaining the essential function of the CNS. Astrocytes instruct the formation and elimination of synapses during development [3,4], provide trophic factors to support neuronal function [5], mediate the uptake and recycling of neurotransmitters [6], form the structure of the brain, and are also involved in the maintenance of the BBB integrity [7]. It is clear that astrocytes dysfunction is closely related to the pathogenesis of many disease, including ischemic stroke. Accumulating evidence demonstrates that astrocytes could have different phenotypes. It has been reported that reactive astrocytes can be classi ed into A1 and A2 phenotypes, which have neurotoxic and neuroprotective effects, respectively [8,9]. A1 astrocyte was induced by interleukin-1 alpha (IL-1α), tumor necrosis factor alpha (TNFα), and the classical complement component C1q, which were secreted by activated microglia. Evidence showed that A1 astrocytes lost many normal astrocyte functions such as promoting neuronal survival, inducing synapse formation and function and phagocytizing synapse, as well as releasing toxic factors to kill neurons and oligodendrocytes [10]. However, a comprehensive characterization of astrocyte phenotype conversion and the effects of A1 astrocytes on the BBB function after ischemic stroke are still unclear.
BBB is a highly complex and dynamic structure composed of tight junctions between endothelial cells lining blood vessels, astrocytic endfeet and a basement membrane, which plays a key role in regulating CNS homeostasis [11,12]. It is well established that BBB is disrupted after ischemic stroke, which contributes to the development of brain injury and subsequent neurological impairment [13,14].
Astrocytes directly encircled cerebral microvessels and modulated BBB functions via astrocyte-derived factors and endfeet [15,16]. Reactive astrocytes secreted not only vascular permeability factors to downregulate tight junction proteins, but also protective factors to protect endothelial cells from apoptosis, suggesting that astrocytes have dual effects on BBB after CNS injury [17]. Therefore, it is critical to elucidate the role of different phenotypes of astrocytes, especially A1 phenotype, on BBB integrity after ischemic stroke. Novel therapeutic strategies that attenuate BBB dysfunction by reversing A1 astrocytes are potential treatment options after ischemic stroke.
In this study, we aim to explore the effect of A1 astrocyte on the BBB integrity after ischemic stroke. We showed for the rst time that ischemic stroke induced resting astrocyte (A0 astrocyte) converted into A2 astrocyte at the early stage, with a gradual shift to A1 astrocyte. In vitro studies revealed that A1 astrocyte conditioned medium disrupted BBB integrity, which was potentially mediated by in ammatory factors and chemokines secreted by A1 astrocyte. We also demonstrated that e ciently blocking A1 astrocytes conversion through semaglutide, a glucagon-like peptide-1 receptor (GLP-1R) agonist reduced ischemia induced BBB disruption, and improved neurobehavioral recovery. Our study highlighted a critical role of A1 astrocytes in the BBB integrity and concluded that A1 astrocyte was a potential therapeutic target for treating ischemic stroke. (Shanghai, China) and housed in a standard facility with 12-hour light-dark cycle, free access to the food and water, ambient humidity of 20∼50%, and temperature of 21∼25°C. Mice were randomly assigned to 3 groups, sham group (n=12), transient middle cerebral artery occlusion (tMCAO) group (n=40) and tMCAO with semaglutide treated group (n=26). Experimental designs were summarized in Fig. 1a.

Materials And Methods
Transient middle cerebral artery occlusion (tMCAO) The mouse model of tMCAO was established using a method described previously [18]. Brie y, mice were anesthetized with 1.5-2% iso urane in a mixture of oxygen/nitrous oxide (30%/70%). First, the common carotid artery (CCA), internal carotid artery (ICA), and external carotid artery (ECA) were carefully separated. Then, a 6-0 nylon suture (Dermalon, 1756-31, Covidien) coated with silica gel was inserted through an incision in the ECA, into the ICA, and advanced until the suture tip reached the bifurcation of the middle cerebral artery (MCA). The inserted suture length was about 0.95 0.05 cm. The success of occlusion was con rmed by the decrease of surface cerebral blood ow (CBF) in the MCA territory to 20% of baseline CBF using a laser Doppler owmetry (Moor Instruments, Devon, UK). Reperfusion was performed by withdrawing the suture 90 minutes after occlusion. The success of reperfusion was con rmed by more than 70% CBF recovered, compared to the baseline. Animals failed in occlusion or reperfusion were excluded from the study.

Neurobehavioral assessment
Neurobehavioral tests were carried out by an investigator blinded to the experimental design using the modi ed neurological severity score (mNSS), hanging wire test, and rotarod test. mNSS was performed at 1, 3, 7, 14, 21 and 28 days after tMCAO. Hanging wire test and rotarod test were carried out at 3, 7, 14 21 and 28 days after tMCAO. mNSS was a composite score of motor, re ex, and balance tests. The severity score was graded at a scale from 0 to 14, where 0 represents normal and 14 indicates the most severe injury [20].
Hanging wire test was used to assess muscle function and motor coordination [21]. In the hanging test, mice were hung on a horizontal wire. The wire was 1.6 mm in diameter, 50 cm in length, and elevated at 30 cm above the oor. Each mouse was given a score of 10 at the beginning of the test. The mice were then scored based on the number of times they reached the terminal (earn one point) and the number of falls (loss one point) in 180 seconds. If the mouse reached the terminal without falling in one trial, no score was given or taken. Therefore, the highest score an animal could receive is 10. The average score of three 180-second tests was used for analysis. A Kaplan-Meier-like curve was created using the scores, in which 10 indicates best muscle function and motor coordination, while a lower score indicated worse muscle function and motor coordination. Holding impulse (s*g), which was calculated by multiplying body mass (g) and hanging time (s), was used as an outcome measure of overall limb strength.
Rotarod test was used to evaluate motor coordination and balance [22]. Mice were trained for 3 consecutive days before tMCAO. On the rst and second day of training, the mice were placed on the nonrotating rod to adapt for 1 minute, then the rod was accelerated to 20 revolutions per minute (rpm) and maintained for 5 minutes. On the third day of training, baseline test after two training sessions was carried out. The rotating rod was accelerated to 40 rpm while the mice were monitored and the fell off rate within 5 minutes was recorded and analyzed.
Brain infarct volume and atrophy volume measurement Six mice from each group were euthanized with an excess of 10% chloral hydrate at 3 days and 14 days after tMCAO. Mouse brain samples were collected after intracardial perfusion with 0.1 ml/L PBS followed by 4% paraformaldehyde (PFA, Sinopharm Chemical Reagent, Shanghai, China). Brains were immediately placed in 4% PFA for 6 hours at 4°C and then transferred to 30% sucrose in PBS until the brain sank to the bottom. Processed mouse brains were placed in −42°C pre-chilled isopentane for 10 min and then stored at −80°C. The pre-OCT-embedded brains were cut into oating coronal sections at a thickness of 30 μm from the anterior commissure to the hippocampus and stored in a 24-well plate containing antifreeze (Meilunbio, Dalian, China). Cresyl violet staining was carried out by sampling 16 sections spaced 300 μm apart that collectively spans the entire injury region for each mouse. The selected brain slices were mounted on glass slides and air dried. The sections were then stained in 0.1% Cresyl violet solution (Meilunbio, Dalian, China) followed by de-staining in ethyl alcohol. The ratio of staining in the ipsilateral and contralateral hemispheres was calculated using ImageJ (National Institutes of Health, Bethesda, MD).
Immunostaining and quanti cation Brain slices or cell slides were xed with 4% PFA for 10 minutes and incubated in 0.3% Triton X-100 solution for 10 minutes followed by blocking with 10% BSA for 1 hour at room temperature. Astrocytes were incubated with antibodies against C3d (1:100, AF2655, R&D system, Minneapolis, MN), S100A10 (1:100, AF2377, R&D system), and glial fibrillary acidic protein (GFAP, 1 Brain sections stained for occludin, ZO-1, or claudin-5 were analyzed for tight junction gap length, which was presented as a percentage (%) of whole tight junction staining. GFAP staining results were analyzed by comparing the GFAP integral optical density (IOD). The number of C3d-positive, S100A10-positive, CD16/32-positive, GFAP-labelled, and IBA-1 labeled cells was counted from randomly selected elds in the peri-focal striatum of the ipsilateral hemisphere in three mice from each group. For each mouse, foursection were evaluated and three elds per section were sampled. The data were analyzed using ImageJ (National Institutes of Health, MD) and Prism Graphpad 8 (GraphPad Software, San Diego).

Western blotting analysis
The ischemic tissue from the ipsilateral hemisphere of the striatum was dissected and lysed in the protein lysis buffer (RIPA, protease cocktail inhibitor and phosphatase inhibitor) on ice immediately after brain sample collection [24]. The protein concentration of each protein sample was determined by a BCA kit (Meilunbio, Dalian, China). An aliquot of 30 μg of total protein from each sample was loaded onto 12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) gel and subjected to electrophoresis. The protein samples on the SDS-PAGE gel were transferred to polyvinylidene uoride membrane (Merck KGaA, Darmstadt, Germany). The membrane was blocked with 5% non-fat milk for 1 hour at room temperature and incubated with the primary antibodies against GFAP (1:4000, AB5804, Millipore, MA), C3d (1:1000, AF2655, R&D system), and β-actin (1:1000, 66009, Invitrogen) overnight at 4°C. After being washed three times in TBST buffer (Meilunbio), the membranes were incubated with HRP-conjugated secondary antibody for 1 hour at room temperature. Immunoblots were developed by incubating with solutions from an enhanced chemiluminescence kit (ECL, Pierce). The results of ECL were analyzed using Image J software.

Real-Time PCR Analysis
The ischemic tissue extracted from the ipsilateral hemisphere of the striatum was used to isolate total RNA using TRIzol reagent (Invitrogen). The Extracted RNA was used for real-time PCR analysis to detect the expression levels of IL-1α, TNFα, and C1q. Single-strand cDNA was synthesized through a universal cDNA synthesis kit (Qiagen, Hilden, Germany) under the conditions of 42°C for 1 hour and then 95°C for 5 minutes. The expression of RNA was tested by a fast real-time PCR system (7900 HT, ABI, Foster City, CA) using an SYBR Green master mix (Qiagen) with the following cycling conditions: 95°C for 10 minutes followed by 40 cycles of 95°C for 10 seconds and 60°C for 1 minute. GAPDH was used as the control for tissues. The primers used for real-time PCR analysis are listed in Table 1.
Primary microglia and astrocyte cell cultures Primary microglia and astrocytes were isolated from postnatal ICR mice (JieSiJie, Shanghai, China). Brie y, the cortex was isolated from the brain of mice and then trypsinized for 10 minutes. After centrifugation, the cell pellet was resuspended in glial cell culture medium with 10% inactivated fetal bovine serum, ltered with a 70 μm lter (Millipore), and then inoculated on the culture dish prepared in advance. The mixed glial cells were kept in an incubator at 37°C and cultured for 10 days under the conditions of 95% humidity and 5% CO 2 . When the astrocytes were con uent, the primary microglia were isolated from the culture by brie y shaking. To obtain primary astrocytes, mixed glial cells were seeded into the petri dish and the medium was replaced with fresh culture medium every 3 days.  ). The genome-wide transcriptomic analysis was performed on 4 independent experiments in the A0 astrocytes and A1 astrocytes groups. Differential expression analysis was performed using the DESeq (2012) R package. p<0.05 and fold change>2 or fold change<0.5 were set as the threshold for signi cantly differential expression. Gene set enrichment analysis (GSEA; https://www.broadinstitute. org/gsea/index.jsp) was performed to nd differential phenotypes between A0 astrocytes and A1 astrocytes groups.

Statistical analysis
The parametric data were analyzed using Prism Graphpad 8. Comparisons between two groups were carried out using Student's t-test. For comparison among multiple groups, statistical signi cance between each group were examined by one-way ANOVA followed by Bonferroni correction for multiple analyses. All data were expressed as mean±standard error of the mean (SEM), and two-tail p<0.05 was considered statistically signi cant [25].

Ischemic stroke induced A1 astrocytes formation
To determine if ischemia initiates astrocytes conversion, we characterized the phenotypic changes of astrocytes at different time points after tMCAO. We found that the number of C3d + /GFAP + cells at the ischemic perifocal region was gradually increased from day 1 to day 14, while the number of S100A10 + /GFAP + cells was gradually decreased (Fig. 1b). The percentage of C3d + /GFAP + cells reached 60% of total cells, in comparison, S100A10 + /GFAP + cells decreased to 40% of total cells at 14 days of tMCAO, indicating that ischemia could promote resting astrocytes transform to the A1 or A2 phenotype astrocytes dynamically. Besides, we found that at the third day after tMCAO, the A1 astrocytes endfeet wrapped surrounding the microvessels in the peri-focal region (Fig. 1c), suggesting a crosstalk between A1 astrocytes and endothelial cells during tMCAO.
Semaglutide reduced M1 microglia polarization and A1 astrocytes formation after ischemic stroke GLP1R agonists have been proven as potential anti-in ammatory agents [26]. Thus, we hypothesized that GLP1R agonists could block A1 astrocytes formation and microglia polarization towards the M1 phenotype microglia. To test our hypothesis, we injected semaglutide, a GLP1R agonist, into ischemic mice 2 hours after tMCAO. We found that semaglutide treatment signi cantly reduced the number of Iba-1 + cells compared to the control at day 3 after tMCAO (p<0.01, Fig. 2a). Besides, semaglutide treatment reduced the number of CD16/32 + /Iba-1 + cells compared to the control (p<0.05, Fig. 2a), suggesting that semaglutide attenuated microglia M1 polarization in the acute phase after ischemic brain injury.
It has been reported that IL-1α, TNFα and C1q together trigger A1 conversion of reactive astrocytes, and subsequently cause neuronal death. Therefore, we examined the expression of IL-1α, TNFα and C1q after tMCAO. Real time-PCR analysis of the perifocal region of the ipsilateral hemisphere revealed that tMCAO induced IL-1α, TNFα and C1q hyperexpression (Fig. 2b), and the IL-1α, TNFα and C1q expression were reduced in the semaglutide treated mice compared to the control (p<0.05, Fig. 2b). To validate if semaglutide could reduce A1 astrocytes formation, western blot and immuno uorescence staining were performed to detect C3d, S100A10 and GFAP protein levels. We found that the A1 astrocytic marker C3d and GFAP expression increased apparently after ischemic brain injury (p<0.001, Fig. 3a, b), and semaglutide treatment decreased their expression at day 3 after tMCAO (p<0.01, Fig. 3a, b), suggesting semaglutide could reduce A1 astrocytes after tMCAO. In contrast, the expression of A2 astrocytes marker S100A10 increased signi cantly in the semaglutide treatment group at day 3 after injury (p<0.05, Fig. 3a, b).

Inhibiting A1 astrocyte transformation reduced brain infarct volume and improved neurobehavioral outcomes after tMCAO in mice
To demonstrate the effect of blocking A1 astrocytes on infarction and neurobehavioral recovery of stroke mice, we examined infarct volume and neurobehavioral tests including mNSS score, rotarod test and hanging wire test in tMCAO mice. We found that infarct volume was greatly decreased in the semaglutide treated mice compared to the control at day 3 of tMCAO (p<0.05, Fig. 4a). Similarly, atrophy volume was reduced in the semaglutide treated mice compared to the control at 28 days of tMCAO (p<0.01, Fig. 4b).
We found that the mNSS score was lower in the semaglutide treated mice compared to the control at 1, 3, 7, 14, 21 and 28 days following tMCAO (p<0.05, Fig. 4c). Furthermore, the performance of motor functions in rotarod test and hanging wire test were better in the semaglutide treated mice compared to the control mice (p<0.05, Fig. 4d, f). The changes of body weight in the semaglutide treated mice were signi cantly recovered at 21 and 28 days of tMCAO (p<0.05, Fig. 4g), suggesting that the general condition was also improved in semaglutide treated mice after tMCAO.

Inhibiting A1 astrocyte transformation attenuated BBB disruption after tMCAO in mice
To evaluate the effect of A1 astrocytes on the BBB after tMCAO, we examined the IgG leakage in the mouse brain at 3 days after tMCAO. The results showed that IgG leakage reduced in the ipsilateral hemisphere in semaglutide treated mice compared to the controls (p<0.01, Fig. 5a). Immunostaining of tight junction proteins, including occludin, ZO-1 and claudin-5, in combination with endothelial cell marker CD31 was used to assess endothelial tight junction gap formation. We found that the disruption of such arrangement and microvessel gap formation was observed in the perifocal region of tMCAO mouse brain. However, the distance of gap formation revealed was less in the semaglutide treated mice compared to the control (p<0.05, Fig. 5b), suggesting a reduction of tight junction gap formation in the perifocal region in the acute phase after tMCAO.

A1 astrocytes derived medium reduced tight junction protein expression of endothelial cells in vitro
It was previously reported that IL-1α, TNFα and C1q are required for converting resting A0 astrocytes to A1 astrocytes [9]. As our immuno uorescence staining images and real-time PCR results showed that IL-1α, TNFα and C1q treatment increased the expression of A1 astrocytes marker C3d in astrocytes, as well as A1 astrocytes related genes including H2-T23, Serping1, H2D1 and Ligp1 (p<0.01, Fig. 6a, b), suggesting that we successfully induced A1 astrocytes in vitro.
To investigate the effect of A1 astrocytes on tight junction proteins of endothelial cells, bEnd.3 cells were treated with medium from A0 astrocytes (A0-CM), or medium from A1 astrocytes (A1-CM) or medium from astrocytes that were treated with LPS stimulated microglia (LPS-MCM-AS-CM). We found that A1-CM treatment and LPS-MCM-AS-CM treatment reduced the ZO-1 and claudin-5 expression in endothelial cells compared to the control (A0-CM treatment) (Fig. 6c, d).
Gene Ontology (GO) pathway enrichment analyses of the upregulated genes showed that compared to A0 astrocytes, multiple altered biological processes including T cell differentiation, immune response, in ammatory response and chemokine-mediated signaling pathway were increased in A1 astrocytes (Fig.  7b). KEGG analyses further demonstrated altered canonical signaling pathways that related to in ammation, chemokine signaling and immune response in A1 astrocytes were enriched (Fig. 7c). Many in ammatory factors and chemokines were found upregulated in A1 astrocytes (Suppl. table 1), indicating A1 astrocytes may induce in ammatory and immune responses to disrupt BBB integrity after stroke.

Discussion
In present study, we demonstrated that the number of A1 astrocytes increased within 14 days after tMCAO while A2 astrocytes decreased during this time. Semaglutide treatment reduced M1 microglia polarization and A1 astrocytes transformation, and subsequently reduced brain infarct volume, improved neurobehavioral outcomes, reduced neuroin ammation, and attenuated BBB disruption in tMCAO mice. Further RNA-seq study found in ammation-, immune-and chemotaxis-related signaling pathways involved in A1 astrocytes function. Our results suggested that A1 astrocyte was a speci c cell group, which played an important role in BBB integrity. Inhibiting A1 astrocyte conversion provides a novel approach for ischemic stroke therapy.
It is well known that astrocytes play essential roles in the formation and maintenance of the BBB [27][28][29]. Accumulating evidence demonstrated that the heterogeneity of astrocytes existed in CNS, both morphologically and functionally [30][31][32]. In particular, the recently proposed disease-associated astrocytes attracted widespread attention [33]. IL-1α, TNFα, and C1q could activate astrocytes to transform into a neurotoxic A1 phenotype [9]. However, its effect on BBB integrity was largely unexplored.
Using a tMCAO model, we demonstrated that resting A0 astrocyte transformed into A1 phenotype after tMCAO, and the number of A1 astrocytes increased over time, while inhibition of A1 astrocytes transformation by semaglutide attenuated BBB disruption. The results of in vitro experiments showed that conditional medium from A1 astrocytes reduced tight junction protein expression in endothelial cells, suggesting the destructive effect of A1 astrocytes on the BBB integrity.
IL-1α, TNFα and C1q are three in ammatory factors that are speci cally expressed in microglia, and they are required for the A1 astrocyte transformation [10]. Our study showed that IL-1α, TNFα, and C1q were highly upregulated in the brain at 3 days after tMCAO, and an increase of A1 astrocytes transformation was detected in the perifocal area after ischemic stroke, while such phenomenon was absent in the normal brain, indicating that ischemic stroke activated microglia could release these cytokines, and promote resting A0 astrocyte towards A1 astrocytes transformation. Our in vitro study further showed that adding IL-1α, TNFα and C1q into the medium of resting astrocytes could promote them to A1 astrocyte transformation.
Astrocytes play important roles in maintaining BBB integrity by safeguarding endothelial tight junctions [34][35][36]. We found that 3 days after ischemic stroke, a large number of A1 astrocytes endfeet wrapped surround the wall of blood vessels, leading us to suspect the potential crosstalk between A1 astrocytes and endothelial cells, which may affect BBB integrity. Our data strongly supported the hypothesis as conditional medium derived from A1 astrocytes or LPS-MCM treated astrocytes reduced tight junction proteins in endothelial cells, suggesting toxic factors secreted from A1 astrocytes degrade tight junction proteins.
To gain insight into what are the A1 astrocytes derived toxic factors that disrupt BBB integrity, transcriptomic analyses of A0 and A1 astrocytes were performed. Our data suggested that multiple biological processes related to BBB function were increased in A1 astrocytes. Growing evidence has shown that in ammatory and immune response are two key driving forces that critically affect BBB function after stroke. In ammation is involved in tight junction decomposition which ultimately causes BBB breakdown. We demonstrated that TNF, IL-17 and NF-kappa B signaling pathways were enriched in A1 astrocytes, and many in ammatory factors such as TNF and IL-6 were increased. It should be noted that matrix metalloproteinases (MMPs) including MMP9, MMP14 and MMP3 were highly upregulated in A1 astrocytes. MMPs are zinc-dependent endopeptidases produced by astrocytes, microglia, endothelial cells that degrade extracellular matrix proteins during ischemic stroke [37]. MMP9 and MMP3 are reported to have detrimental effects on BBB integrity. Blocking MMP9 and MMP3 signi cantly attenuate BBB disruption and improve neurobehavioral recovery [38]. A1 astrocytes may induce BBB disruption by MMP9 and MMP3, as we found MMP9 and MMP3 were upregulated in A1 astrocytes.
In addition to in ammatory response, chemokine signaling pathways and chemokines including CCL12, CCL5, CXCL1, CXCL2 and CXCL10 were also enriched in A1 astrocytes. These chemokines are well known for mediating in ammatory and immune responses by guiding immune cells migration into the lesion area of the brain. For example, CCL5 is a chemoattractant of T cells to the site of in ammation, which mediates the cerebral in ammation and causes BBB disruption [39]. CXCL1 produced by astrocytes is a critical ligand required for neutrophil trans-endothelial migration and exacerbates brain damage [40]. Activating MMPs degraded endothelial cell tight junction proteins, secreted in ammatory cytokines and adhesion molecules, and induced endothelial cell death [37]. All these processes are closely related to the BBB disruption.
Besides A1 astrocyte, A2 astrocyte was also detected in the brain after ischemic stroke. Previous studies showed that activated astrocytes promoted neuronal function recovery and repair [41][42][43]. A2 astrocyte could up-regulate neurotrophic factors and vascular growth factors, which presumably promoted neuronal survival, aided axon regeneration, and promoted BBB repair [44]. Therefore, manipulation of astrocyte phenotype was regarded as a potential therapeutic strategy. Understanding the heterogeneity of astrocytes helps to comprehend how the function of astrocytes shapes the function and dysfunction of the brain.
GLP-1 is a 30-amino acid peptide hormone for stimulating insulin secretion. Currently, GLP-1 and GLP-1 receptor agonists including exendin-4, liraglutide, semaglutide, etc., were developed and approved for treating type 2 diabetes [45,46]. In addition, recent studies demonstrated their potential for neurological disorders treatment [47]. In a recent study, semaglutide was shown to have neuroprotective effects in a rat MCAO model [19]. Besides, a study demonstrated that NLY01, a GLP1R agonist, protected against the death of dopaminergic neurons and improved behavioral recovery in Parkinson's disease, which was mediated by the block of A1 astrocyte transformation [9]. In the present study, we found that similar to NLY01, semaglutide could also reduce A1 astrocyte transformation. This may be caused by attenuation of microglial in ammation as semaglutide treatment reduced expression of IL-1α, TNFα, and C1q and M1 microglial polarization. Together, these results highlighted the important role of A1 astrocytes on the BBB disruption after ischemic stroke.

Conclusions
Our study demonstrated that stroke induced A1 astrocyte disrupted blood-brain barrier integrity, and block of A1 astrocytes conversion by semaglutide attenuated brain injury, suggesting the detrimental role of A1 astrocyte in stroke. We can envision that A1 astrocytes to be a novel therapeutic target for the ischemic stroke therapy. Jiao Tong University, Shanghai, China. All animal procedures were performed to minimize pain or discomfort in accordance with current protocols.

Consent for publication
Not applicable.
Availability of data and materials: The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests:
No competing interest. Author's contributions QZ and CL participated in the research design, all experimental procedures, animal surgery, RNA seq, data analysis and drafting of the rst manuscript. RS, HS and LD contributed to animal surgery, behavioral tests and data collection. TC assisted with data collection. ZZ and GYY were involved in the discussion of the research design, the results and edited the manuscript. YW and YT supervised all aspects including research design, data analysis and manuscript preparation.  Figure 1 Phenotypic changes of astrocytes after tMCAO in mice a. Experimental scheme. Mice were trained for 3 days before tMCAO. Neurobehavioral tests and body weight were examined at 1, 3, 7, 14 and 28 days following tMCAO. Animals were sacri ced at 1, 3, 7, 14 and 28 days following tMCAO. b. Photomicrographs showed C3d+/GFAP+ cells (top, C3d in green color; GFPA in red color) and S100A10+/GFAP+ cells (bottom, S100A10 in green color; GFAP in red color) in the perifocal area of ipsilateral hemisphere at 1, 3, 7 and 14 days after tMCAO. Bar graph showed the percentage of C3d+/GFAP+ cells and S100A10+/GFAP+ in the ipsilateral hemisphere of brain at 1, 3, 7 and 14 days after tMCAO. Scale bar=50 μm. Data are mean±SEM, n=4-5 per group. c. Photomicrographs showed C3d+/GFAP+ cells closely wrapped around lectin+ microvessels (arrowheads) in the perifocal area of the ipsilateral hemisphere in the mouse brain at 3 day after tMCAO. Scale bar=50 μm.  Semaglutide attenuated A1 reactive astrocyte transformation after tMCAO a. Representative photomicrographs showed that GFAP+ cells, C3d+/GFAP+ cells (C3d in green color; GFAP in red color) and S100A10+/GFAP+ cells (S100A10 in green color; GFAP in red color) in perifocal area of ipsilateral hemisphere in the semaglutide treated tMCAO mice and control mice at 3 days following tMCAO. Scale bar=50 μm. Bar graph showed that semi-quanti cation of GFAP intensity, the percentage of C3d+/GFAP+ cells, and S100A10+/GFAP+ cells. Data are presented as mean±SEM, n=3 per group. *p<0.05, **p<0.01. b.
Western blotting analysis data showed relative GFAP and C3d levels in the perifocal area of ipsilateral hemisphere in the semaglutide treated tMCAO mice and control mice. Data are mean±SEM, n=3 per group. *p<0.05.

Figure 4
Inhibiting A1 astrocytes attenuated infarct volume and neurobehavioral de cit in the semaglutide treated tMCAO mice a-b. Cresyl violet-stained coronal sections of the brain in control mice, tMCAO mice, and semaglutide treated tMCAO mice following 3 (a, infarct) and 28 days (b, atrophy) of tMCAO. The brain infarct area and brain atrophy were circled by the dashed line. Bar graph showed the semi-quantitative analysis of the infarct volume and atrophy volume. Data are mean±SEM, n=6 per group. *p<0.05, **p<0.01. c-f. Neurobehavioral outcomes were assessed by three neurobehavioral tests including the modi ed neurological severity score (mNSS, c), rotarod test (d), and hanging wire test (e-f). Line graph showed body weight (g), Data are mean±SEM, n=9-12 per group, **p<0.01, *p<0.05.

Figure 5
Inhibiting A1 astrocytes reduced BBB disruption and gap formation of tight junction in tMCAO mice a.
Photographs showed IgG staining in the coronal section of the brain following 3 days of tMCAO in the semaglutide treated tMCAO mice, tMCAO mice, and control mice. Scale bar=50 μm. Bar graph showed the mean IOD of IgG intensity in these 3 groups. Data are mean±SEM, n=6 per group. **p<0.01. b.

Figure 6
A1 astrocyte derived medium reduced tight junction protein expression in vitro a. Immuno uorescence images showed resting astrocytes were converted to C3d+/GFAP+ cells (C3d in red color; GFAP in green value showed top 10 upregulated biological process. c. KEGG enrichment top 20 analysis suggested that TNF signaling pathway and cytokine-cytokine receptor interaction were related with the phenotypic changes of astrocytes.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.