The Transcriptional Landscape of Pericytes in Acute Ischemic Stroke

The current treatment options for ischemic stroke aim to achieve reperfusion but are time critical. Novel therapeutic approaches that can be given beyond the limited time window of 3-4.5 h are still an unmet need to be addressed to improve stroke outcomes. The lack of oxygen and glucose in the area of ischemic injury initiates a pathological cascade leading to blood-brain barrier (BBB) breakdown, inflammation, and neuronal cell death, a process that may be intercepted to limit stroke progression. Pericytes located at the blood/brain interface are one of the first responders to hypoxia in stroke and therefore a potential target cell for early stroke interventions. Using single-cell RNA sequencing in a mouse model of permanent middle cerebral artery occlusion, we investigated the temporal differences in transcriptomic signatures in pericytes at 1, 12, and 24 h after stroke. Our results reveal a stroke-specific subcluster of pericytes that is present at 12 and 24 h and characterized by the upregulation of genes mainly related to cytokine signaling and immune response. This study identifies temporal transcriptional changes in the acute phase of ischemic stroke that reflect the early response of pericytes to the ischemic insult and its secondary consequences and may constitute potential future therapeutic targets.


Introduction
Stroke is the second leading cause of death worldwide and one of the main reasons behind adult disability [1].
In ischemic stroke, the interruption of nutrients and oxygen supply to the affected brain regions leads to a cascade of molecular and cellular events arising within seconds to minutes from the ischemic insult and progresses for days up to weeks. Stroke pathology develops in different phases that can be divided into a hyperacute, an acute and a chronic phase [2]. The rst few hours following the initial event correspond to the hyperacute phase, characterized by cell death, and breakdown of the blood-brain-barrier (BBB) [3,4], followed by the acute phase which is usually considered to last for a few days, characterized by microglial and astrocytic activation, increased BBB permeability, vascular leakage and in ammation [5].
The chronic phase starting days after the injury involves endogenous repair mechanisms including vascular remodelling, neural plasticity, and scar formation [6,7].
Currently, available treatments, like tissue plasminogen activator (tPA) [8,9] and thrombectomy can only be given during a narrow time window within the rst few hours after stroke onset, and are therefore available only to a small number of patients [8,10]. As such, the development of therapeutic interventions that could target stroke pathogenesis beyond the hyperacute phase and improve stroke outcome is a global health priority.
Pericytes are one of the rst cell types to respond to hypoxia [11,12]. Surrounding the endothelial cell layer at the microvasculature, they form a key component of the BBB [13] and maintain BBB integrity and the vascular tone [14]. The response of pericytes to hypoxia may result in detrimental effects e.g., early cell death contributing to restricted capillary blood-ow [15,16] and their detachment from the capillaries that in turn leads to BBB breakdown [17]. However, pericytes also mediate bene cial effects in ischemic stroke, such as contributing to angiogenesis and stabilization of newly formed vessels [18,19]. Pericytes may exert these different functions at different time points after stroke, however, the exact timeline and sequence of pericytes response associated with pathological events after ischemic stroke as well as the underlying molecular mechanisms governing this sequential response are currently not known [20].
Overall, their unique position at the blood/brain interface, their essential functions related to angiogenesis, and their ability to react to hypoxia potentially place pericytes at the top of the stroke cascade and makes them an ideal candidate for therapeutic interventions that could regulate several key pathological hallmarks [21,22].
One of the adaptive responses of pericytes to hypoxic environments is the expression of regulator of Gprotein signalling 5 (RGS5), a negative regulator of G-protein-coupled receptors that in the brain is exclusively expressed by vascular mural cells, and a valuable tool to speci cally identify pericytes in the brain [23][24][25]. Here we utilize RGS5 reporter mice [26] in order to characterize the speci c molecular pericyte response and potential differences between subpopulations of pericytes within the vascular niche in ischemic stroke at different time points of the acute ischemic cascade. To our knowledge, we present the rst study comparing speci c transcriptomic changes at a single-cell level between the ipsilateral and the non-affected contralateral hemispheres of the brain at 1, 12 and 24 hours after the ischemic stroke insult. Overall, this study suggests a role of a subcluster of pericytes in ischemic stroke that warrants future investigation as a potential target population to intersect the pathological progression in stroke evolution.

Animals
We utilized C57bl/6 wildtype (WT) mice (n = 10) and male RGS5 GFP/+ mice (n = 27) aged 8-12 weeks from a KO/knock-in reporter mouse strain that expresses green uorescent protein (GFP) under the promoter of RGS5 in a C57bl/6 background [26]. In this model, one allele of RGS5 is replaced by GFP, making it possible to track pericytes by GFP expression under the activated RGS5 promoter. RGS5 GFP/+ mice were used for single-cell RNA sequencing (n = 3); and real-time quantitative PCR (RT qPCR) veri cations (n = 24), of which n = 12 were selected for pMCAO and n = 12 for sham operations. Sham animals were used in order to further validate the use of the contralateral hemisphere as healthy control. Three WT mice for each RT-qPCR replicate were used as a negative control for GFP staining at the ow cytometer. N = 7 additional WT mice underwent pMCAO for protein veri cations by immunohistochemistry.
All animal experiments were approved by the Ethical Committee of Lund University (14205-199), and methods were carried out in accordance with the relevant guidelines and regulations. Animals were housed under standard conditions with a 12 h light/dark cycle and had access to food and water ad libitum. Every effort was made to keep animal numbers minimal according to the 3 R guidelines and principles of the Swedish Research Council (https://djurforsok.info/).

Permanent middle cerebral arterial occlusion (pMCAO)
In order to obtain a localized and reproducible stroke lesion, we used a pMCAO model. The distal part of the left MCA was occluded using electrocoagulation to induce focal ischemia as previously described [27]. Brie y, animals were anesthetized using 5% iso urane initially, then 1.5-2% iso urane was maintained during surgery. 100 µl Marcain (2 mg/kg) (AstraZeneca) was locally applied to the site of surgery and a ~ 1 cm long incision was made between the left lateral ear and eye. The temporal muscle was detached from the skull in its apical and dorsal parts and the parotid gland moved aside. After identifying the MCA in the rostral part of the temporal area, dorsal to the retro-orbital sinus, a small craniotomy was made at the anterior distal branch of the MCA, using a surgical drill. After exposing the MCA, the artery was occluded using electrocoagulation forceps (ICC50; Erbe), proximal and distal to the bifurcation branch point. In the case of anatomical variation where no bifurcation was present, the MCA was coagulated twice, before the wound was sutured. Sham operations were conducted in the same way, but without occluding the MCA.

Tissue processing
At the respective time point, mice were transcardially perfused with 0.9% saline for 5 min. Brains for immunohistochemistry were extracted and frozen at -80°C, then cut on a cryostat to coronal 20 µm-thick sections on glass slides. Sections were kept at -20°C until further analysis. Brains for uorescence activated cell sorting (FACS) and single-cell sequencing were kept on ice and processed as described below.
Tissue dissociation for 10x Genomics Bulk live non-neuronal cells were isolated according to Chang et al. [28]. After saline perfusion, the brains were minced with a razor blade and dissociated with enzymatic digestion using collagenase IV (400 U/ml, Worthington Biochemical, cat. LS004188), dispase I (1.2 U/ml, Worthington Biochemical, cat. LS02104) and DNAse I (32 U/ml, Thermo Fisher Scienti c, cat. 18047019) in 6 ml phosphate buffer saline (PBS) with 0.9mM CaCl 2 and 0.49 mM MgCl 2 per sample. The solution was incubated at 37°C for 1 hour with mechanical trituration every 10 min with 1000 µl followed by 100 µl pipette tips. After the tissue was disaggregated, 1 ml of fetal bovine serum (FBS) was added to the solution to stop the enzymatic reaction, ltered through a 40 µm lter, and centrifuged at 400 g for 5 min at 4°C. The cells were washed once with ice-cold PBS before being centrifuged again at 400 x g for 5 min at 4°C and resuspended in 10 ml of ice-cold 20% BSA in 1x PBS and centrifuged at 1000 x g for 25 min at 4°C for removal of myelin and neurons. The cells were then resuspended in FACS buffer (0.5% BSA in PBS) and kept on ice for subsequent staining.

FACS
For 10x genomics, single cells suspensions were incubated with Fc-block for 15 min, followed by incubation with speci c antibodies or corresponding isotype-matched control antibodies at a concentration of 1 µg/µL at 4°C for 30 min in darkness in FACS buffer. To speci cally isolate pericytes, we used RGS5-reporter mice (see methods). We veri ed the presence of live single-cell pericytes (CD140b + , CD13 + , GFP + , DAPI − , PECAM1 − ) dissociated from endothelial cells among the processed cells before each experiment. We then sorted neuron depleted (see above) live DAPI − bulk brain cells on ARIA-II. After sorting, a fraction of the cells was re-stained as explained above and re-analyzed in the ow cytometer to check for the survival of pericytes and endothelial cells (PECAM1 + , DAPI − ) after sorting. Bulk-sorted cells were evaluated for > 80% viability with trypan blue staining. For each replicate of 10x analysis, 8500 cells were resuspended in 45 µl FACS buffer.
For qPCR-validation experiments of the scRNAseq data, single-cell suspensions were incubated with primary antibodies or corresponding isotype-matched control antibodies at 1 µg/µL at 4 ℃ for 30 min in darkness, after incubation with Fc-block as described above. Then, the cell suspensions were washed twice with PBS. For FACS sorting, quadrant gates were drawn to separate pericytes from the rest of the populations based on differences in speci c antigen expression: pericytes were selected based on GFP + , CD140b + , CD13 + , PECAM1 − , and all other cell populations were selected based on GFP − (indicated as bulk), in order to compare the relative gene expression. Cells were sorted on a FACS-ARIA II. To obtain a su cient number of pericytes for subsequent RNA isolation, n = 4 mice were pooled for each condition. For each replicate, one WT mouse was used as a negative control for GFP staining. for sequencing depth between experiments. The Seurat suite v.4 was used to analyze and normalize the scRNA-seq data. Cells with a number of features < 200 and a mitochondrial content above 10% were ltered out. The data was then log-normalized, and clusters were de ned according to the highest number of clusters from the nd-neighbors, nd-clusters function in Seurat. Uniform manifold approximation and projection (UMAP) was run with the same number of dimensions not to under-or over cluster the data. To de ne cluster identities, we compared the expression of cell type speci c transcripts reported in the literature [29][30][31][32][33][34][35][36]. Differentially regulated genes (DEGs) were identi ed with the ndallmarker-or ndconservedmarker-functions in Seurat. Fast gene set enrichment analysis (FGSEA) was performed on the DEGs to de ne Hallmark gene set enrichments within the data [37,38].

Real-Time quantitative PCR
To further con rm single-cell data, we isolated pericytes as previously described from the ipsi-and contralateral hemispheres either from stroke mice or sham-treated mice at 24 hours after the surgery. To ensure the collection of a su cient number of sorted pericytes to perform the subsequent RT qPCR, cells from 4 mice were pooled for each condition, and the experiment was repeated 3 times. Total mRNA was isolated with RLTplus lysis buffer (Quiagen) from pericytes or bulk cells previously isolated by FACS. After mRNA puri cation, cDNA was retrotranscribed using Thermo Scienti c Maxima First Strand cDNA Synthesis Kit for quantitative polymerase chain reaction (RT-qPCR). Samples were then prepared for qPCR using PowerUp™ SYBR® Green Master Mix (Thermo Fisher). Forward and reverse primers (TAG Copenhagen) of interest were added at a concentration of 0.05 µM/well. We then added 1 µl of cDNA from each sample and UltraPure water (Invitrogen) to reach the nal volume of 10 µl/well. Il11 master mix was supplemented with 0.5 M betaine (Sigma Aldrich) to enable PCR ampli cation of the GC-rich Il11 transcript. The analyses were run on a Bio-Rad CFX96 RT qPCR system. Thermal cycling started with incubation at 50°C for 2 min, followed by initial denaturation at 95°C for 2 min. Cycling was made with 15 seconds denaturation followed by the annealing and extension at 60°C for 1 minute, for a total of 44 cycles. To calculate the relative target gene expression, the 2 − ΔΔCt method was used [39]. Each qPCR round of the triplicate was performed in duplicates measuring the relative expression of the genes from stroke ipsilateral pericytes in comparison with stroke contralateral pericytes.
Samples with a normal β2m ampli cation curve but an ampli cation of the target gene above 36 qPCR cycles were considered undetected and imputed with the R package nondetects [40]. Platelet-derived growth factor receptor beta (PDGFRß) was used as pericyte marker. Sections were incubated with the corresponding secondary antibodies diluted in PBS for 1 hour at RT and with DAPI (1:500) for 5 min. Negative controls were treated similarly without primary antibodies. Sections were washed in PBS and rinsed in deionized water and mounted using PVA-DABCO. Image processing Immuno uorescence images were acquired on a Leica DMi8 confocal microscope using the 40x oil immersion objective. The tissues were visually scanned by the operator in order to nd IL11 signals and all the locations in which a signal was present were imaged, for a total of 70 images from 7 different animals. The images had xy dimensions of 387.69 x 387.69 µm and a depth of 6.5 µm and were acquired with a z-step size of 0.35 µm. Images were analyzed and assembled using ImageJ software v.1.53v (NIH, USA). 2D images were produced from the z-stack using Max Intensity function, followed by Split Channels function. The threshold for the IL11 channel was set using the Max Entropy automatic threshold, while for PDGFRβ channel, Moments threshold was set. A cut-off of the particles from the debris was set using Analyze Particles function (IL11 channel: 0.4-In nity; PDGFRβ channel: 1.5-In nity). Measurements of area density were conducted on the resulting images. IL11 density was normalized on PDGFRβ density.

Statistical analysis
To analyze qPCR data, One-way ANOVA with a 95% con dence level was used, followed by Dunnett´s multiple comparison correction test. To analyze confocal images, paired t-test with a 95% con dence level was used. Statistical analysis was performed using GraphPad Prism version 9.0.0 for Mac, GraphPad Software, San Diego, California USA, www.graphpad.com.
Major changes in cellular clustering between the ipsilateral and the contralateral hemispheres were detected 12 and 24 hours after stroke (Fig. 1d), however, subtle changes were also detected at 1 hour, including a reduction in the oligodendrocyte population density, passing from 7.3 to 1.6% of the total amount of cells, and the onset of a small cluster of neutrophils amounting to 0.85% of the total cells in the ipsilateral hemisphere compared to the contralateral one (Fig. 1d). At 12 hours, we observed an increased density of cells in the microglial cluster 5 (from 1 to 18.9% of the total) and a decrease in the density of cells in the microglial cluster 4 (from 18.8 to 8.3%) in the ipsilateral hemisphere compared to the contralateral. At 24 hours, the differences between the ipsilateral and contralateral hemispheres were more marked than at the 12-hour time point, with an increased density of microglial cluster 5 (0.9 to 17.6%), endothelial cells cluster 9 (0.4 to 11.1%) macrophages (0.2 to 8.5%) and a decrease in the microglial clusters 0, 3 and 4 and in the endothelial cell cluster 1 (13.2 to 5.4%) ( Fig. 1d; Supplementary).
Temporal heterogeneity of pericyte sub-populations in the acute phase of ischemic stroke To investigate the speci c pericyte transcriptomic response after ischemic stroke in depth, we extracted the pericytes and SMC clusters from the dataset and performed a new clustering analysis. Since pericytes are widely known as highly plastic cells and share common features with SMC, we decided to also include SMC in our re-clustering analysis to obtain a more comprehensive view of the temporal changes after stroke onset. The re-clustering analysis revealed 8-10 subclusters depending on the time-point and experimental condition (Fig. 2a).
When comparing the ipsilateral and contralateral hemispheres 1 hour after stroke, the frequency of cells in subcluster 3 increased, while we observed fewer cells in subcluster 0 (Fig. 2a). Similarly, at 12 and 24 hours after stroke, the cellular frequency in subcluster 3 increased while it decreased in subcluster 0 (Fig. 2a). Interestingly, the pericyte subcluster 5, characterized by the expression of Pdgfrβ and Rgs5 and the lack of expression of Acta2, was only present in the ipsilateral sample at both, 12 and 24 hours, but not in the contralateral hemisphere, suggesting a unique pericyte subpopulation in response to ischemic stroke (Fig. 2a).
Other subclusters that showed a different distribution across the timepoints were cluster 6, 8 and 10, all characterized by a low frequency of cells (Fig. 2a, c). Based on the co-expression of markers other than the pericyte-speci c ones, we classi ed cluster 6 as possible doublets of endothelial cells (partly Pecam1 + ), cluster 8 as possible doublets of microglia (partly Aif1 + ) and cluster 10 as possible doublets of oligodendrocytes (partly Mog + ) (Fig. 2b).

c = contralateral; i = ipsilateral; h = hour
DEGs and GSEA analyses on the most responsive pericytes subclusters To further explore the speci c response of selected subclusters we investigated the differentially expressed genes (DEGs) between the selected subclusters and the remaining pericyte clusters at each speci c timepoint. Motif 1 (Adamts1) and others (Fig. 3a).
To investigate the top pathways that could best explain the variations among the most interesting pericytes subclusters based on DEGs analysis, we performed an unbiased search across the MSigDB "Hallmark" gene sets.
First, we evaluated the most frequently differentially expressed gene sets between the ipsilateral and contralateral pericytes clusters 1 hour after stroke. Hallmark analysis showed an upregulation of TNFa signalling, P53 pathway, Hypoxia and Apoptosis (Fig. 3b).
Secondly, to gain insights into the functions associated with the pericyte subcluster 5, we performed the GSEA on the DEGs of cluster 5 vs the rest of the mural cells subclusters. 24 hours after stroke, cells from subcluster 5 upregulated pathways related to downstream Myc targets, Tumor necrosis factor (Tnfα) via nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB), interleukin-2 (Il2)/signal transducer and activator of transcription 5 (Stat5), mammalian target of rapamycin complex 1 (Mtorc1), hypoxia, apoptosis, unfolded protein response, glycolysis and pathways related to in ammation. Downregulated pathways included epithelial mesenchymal transition, estrogen response, UV response, apical junction, adipogenesis and myogenesis. GSEA analysis on the markers of the stroke-speci c cluster 5 at 12 hours compared to the remaining mural cell subclusters from the same time point revealed a similar pathways pro le to 12 hours ( Fig. 3d and f).

RT qPCR validation of the top 10 DEGs associated with the stroke-speci c subcluster of pericytes
Since the transcriptomic pro le of the pericytes subcluster 5 showed the largest changes at 24 hours after ischemic stroke, we decided to validate the increased gene expression of the DEGs characterizing the stroke-speci c pericyte subcluster using RT-qPCR. Thus, we isolated pericytes by cell sorting using the same RGS5 GFP/+ reporter mice and compared the expression of the genes selected from the transcriptomic analysis to the remaining cell types 24 hours after stroke.
Overall, qPCR analysis con rmed our sc-RNA seq data (Fig. 4b). Il6, Il11, and Adamts4 mRNA expression by ipsilateral pericytes was respectively ~ 180, ~110 and ~ 120 times increased compared the other groups, where the mRNA expression was low or non-detectable. Rdh10 and Stc1 mRNA levels showed a 150 fold change increase in the ipsilateral pericytes and only a slight upregulation of ~ 50 and ~ 20 in the contralateral pericytes. Ccl2 mRNA levels were ~ 120 times upregulated in ipsilateral pericytes, although the remaining cells from the ipsilateral hemisphere also showed a ~ 20 fold change increase in Ccl2 expression. Ednrb mRNA levels were mostly increased in ipsilateral pericytes with a ~ 120 fold change increase, despite one sample of contralateral pericytes (from the sham-treated animals) that also expressed Ednrb mRNA. qPCR results on Fstl1 levels only partly con rmed the transcriptomic data, being upregulated in all groups of pericytes in stroke animals, both in the ipsilateral and the contralateral hemisphere, but not in the groups of bulk cells. Mt2 mRNA was expressed by ipsilateral pericytes, but with a wide variation in the expression in the other groups, therefore not con rming the computational prediction on the differentially expressed genes from the transcriptomic analysis. Finally. Dbn1 expression was too low to be accurately detected with qPCR analysis (data not shown).

IL11 is exclusively expressed in PDGFR-expressing cells after stroke
Interestingly, both bioinformatic analysis and qPCR indicated that Il11 expression was among the genes with the highest fold change and speci c to pericytes from subcluster 5 at 12 and 24 hours. IL11 is a cytokine known to be primarily produced by stromal broblasts within the gastrointestinal tract [41], heart [42], liver [43], and lungs [44], but it has also been reported that epithelial and immune cells can be a source of IL11 under pathological conditions [45,46]. Hence, we decided to further evaluate IL11 protein spatial expression in pMCAO brains using immunohistochemistry. Confocal imaging con rmed the signi cant increased expression of IL11 in the ipsilateral hemispheres (P = 0.028; CI, 0.01985-0.05979) and highlighted that IL11 presence was only restricted to the cells expressing PDGFRβ, and not expressed in other cell types (Fig. 5a, c).

Discussion
Extending previous studies by our group investigating functions of pericytes in ischemic stroke [47][48][49][50], we here perform a further characterization and an in-depth analysis of the molecular response of RGS5expressing pericytes to ischemic stroke using the same stroke model, pMCAO. To explore the events that determine BBB breakdown and vascular disruption, we examined the transcriptomic signature of pericytes at three different time points after stroke and we identi ed several transcripts uniquely expressed in pericytes that change over time and may constitute potential targets for the treatment of early vascular dysfunctions associated with ischemic stroke. To our knowledge, this is the rst study to focus speci cally on the temporal dynamics of molecular mechanisms engaged by pericytes during the acute phase of the pMCAO model using single-cell transcriptomic analysis.
Our data show that pericytes in the ischemic stroke area upregulate Il6 compared to the other cell types at both 12 and 24 hours after stroke. In addition, our data also revealed that Il11 expression is only upregulated in pericytes in the stroke-speci c subcluster 5 but absent in the contralateral pericytes or in other cell types. IL6 and IL11 belong to the same cytokine family [51]. IL6 has been described to have a bidirectional role in ischemic stroke, being both harmful and protective [52], while IL11 seem to have a protective role after ischemic stroke [53], despite having controversial roles in other processes such as brosis and in ammation [27]. Our ndings are consistent with previous studies, describing pericytes production of IL6 as early as 2 hours after hypoxic treatment [21] and Il11 expression 24 hours after ischemic stroke [20,54].
We validated the selective expression of IL11 protein in pericytes after stroke using immunohistochemistry, showing IL11 exclusively colocalizing with pericytes, and not with neurons or other cell types [53].
IL6 and IL11 act mainly on the JAK-STAT pathway [55]. Our group has previously shown that in hypoxia pericytes activate STAT3 pathways even before the widely studied hypoxia-inducible factor 1α (HIF1α) ones, leading to overexpression of genes that are involved in early hypoxic responses, such as c-MYC [21]. In addition, Gene Ontology term analysis suggested that the major STAT3 bound regulated genes responses controlled metabolic and angiogenic processes [21]. Overall, our data suggest that STAT3 in pericytes plays a key role in the hypoxic responses in the hyperacute phase after stroke, and the upstream cytokines IL6 and IL11 may be produced by pericytes to further activate the STAT3 pathway both in an autocrine and a paracrine way, which in turn could have an important role in the early responses to hypoxia.
In our dataset, pericytes residing in the ipsilateral stroke region showed highly upregulated levels of Ccl2 compared to the other cells in the brain. CCL2 is a chemokine with chemotactic activity for monocytes and basophils and it is produced by a wide variety of cells [56]. Previous studies have shown that Ccl2 levels are highly increased in the brain after stroke and that silencing the Ccl2 gene is protective in stroke models [56], promoting repair [57]. Furthermore, CCL2 de ciency has been shown to decrease macrophage in ltration to the infarct core after stroke [58]. In a transient MCAO mouse model, CCL2 inhibition resulted in a reduction of brain edema, leukocyte in ltration, and in ammation [59]. Considering the role of CCL2 in chemotaxis and BBB remodeling, our results suggest that pericytes may contribute to immune cell recruitment in the acute phase after stroke.
ADAMTS4 was among the highest and most speci c genes upregulated in the ipsilateral pericytes. EDNRB is one of the receptors that mediate the effects of endothelin-1 (ET-1) [64]. Interestingly, a previous study showed Ednrb upregulation in pericytes at 24 hours after ischemic stroke using a temporal occlusion of the MCA [20], suggesting that the upregulation of Ednrb is not dependent on the occlusion model of the MCA, but rather due to the ischemic insult itself. Pericytes that upregulate EDNRB might be more responsive to ET-1, which could regulate the vasodilation of the blood vessels around the ischemic core promoting reperfusion mechanisms, both in the permanent and the temporal occlusion models.
RDH10 is one of the enzymes responsible for the synthesis of the endothelial retinoic acid (RA) [65].
Following ischemic brain injury, exogenous RA administration has been shown to promote STC1 is a secreted hormone with antioxidant effects [69] and it has been observed to reduce brain dysfunction after cerebral ischemia/reperfusion by decreasing BBB permeability and oxidative stress parameters [70]. Our study is the rst work, to our knowledge, suggesting that STC1 is produced by pericytes in response to an ischemic injury.
Importantly, our results are supported in ndings from a previous work, investigating cellular changes with sc-RNA seq 24 hours after the ischemic insult using a transient MCAO model [20]. In one of the pericytes clusters, the authors observed the upregulation of transcripts that we also identi ed in this study, such as Il11, Ednrb, Adamts4, Ccl2, Il6. The authors describe that the pericytes belonging to this cluster highly expressed gene sets involved in immune functions after ischemic stroke, including Oncostatin-M induced BBB breakdown, HIF-1, and cytokine-mediated signaling pathway. Therefore, despite the difference in stroke models, certain subclusters of pericytes respond comparably 24 hours after stroke.
Interestingly, in both studies, most of the transcriptional changes is associated with one speci c pericytes subcluster, identi ed in our analysis as subcluster 5, that is only found in the ipsilateral stroke hemisphere and only detectable from 12 hours after the stroke, pointing at pericytes as plastic cells where subsets of pericytes may have different functions in response to ischemic stroke.
Previously published transcriptomic studies on ischemic stroke were either performed using bulksequencing methods only and therefore not suitable to detect changes in gene expression from different cell types and cellular subtypes [71][72][73], or restricted to one time point after the ischemic insult [20,54,73,74]. Our work extends those studies and provides speci c new insights on the transcriptomic signature of pericytes in ischemic stroke, adding important knowledge on the emerging key role pericytes have at the blood/brain interface in particular as regulators of BBB integrity and in ammation after stroke, consistent with a role of pericytes as neuroin ammatory mediators at the BBB [75]. Further studies are warranted to con rm the functional implications of potential target genes in pericytes in the future and to pave the way for therapeutics modulating the pericyte response in stroke.

Declarations
Competing interests  Validation of IL11 increased expression in the stroke ipsilateral hemisphere by immunohistochemistry (a) IL11 is only detected in the ipsilateral hemisphere of mice subjected to stroke in the proximity of PDGFR + cells (b) Location of the selected images. The infarct core is outlined(c) Area fraction occupied by IL11 and PDGFR signals. Data are presented as mean ±SEM. p = 0.028, paired t-test 95% con dence level (con dence intervals 0,01985 to 0,05979). Scale bar = 20mm

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