key: cord-0277085-ts5ujn81
authors: Zeitelhofer, M.; Stefanitsch, C.; Protzmann, J.; Torrente, D.; Adzemovic, M.Z.; Lewandowski, S.A.; Muhl, L.; Eriksson, U.; Nilsson, I.; Su, EJ.; Lawrence, D. A.; Fredriksson, L.
title: Reduced myofibroblast transdifferentiation and fibrotic scarring in ischemic stroke after imatinib treatment
date: 2021-10-28
journal: bioRxiv
DOI: 10.1101/2021.10.28.466225
sha: 23d5d60ce163653938661792761bc709e864ac47
doc_id: 277085
cord_uid: ts5ujn81

The tyrosine kinase inhibitor imatinib has been reported to improve outcome in patients following ischemic stroke but the exact mechanism remains elusive. Here, utilizing a photothrombotic murine model of middle cerebral artery occlusion (MCAO), we show that imatinib-mediated inhibition of stroke-induced blood-brain barrier (BBB) dysfunction coincided with decreased expression of genes associated with inflammation and fibrosis in the cerebrovasculature. We found that imatinib effectively dampened stroke-induced reactive gliosis and myofibroblast transdifferentiation, whilst having very limited effect on the rest of the glia scar and on peripheral leukocyte infiltration. Further, our data suggest that consolidation of the PDGFRα+ portion of the fibrotic scar in imatinib-treated mice contributes to the improvement in functional outcome compared to the vehicle controls, where the PDGFRα+ scar is expanding. Comparison with human stroke transcriptome databases revealed significant overlap with imatinib-regulated genes, suggesting imatinib may modulate reactive gliosis and fibrotic scarring also in human stroke.

Ischemic stroke represents a major public health challenge and is currently the 3rd leading cause of death and the leading cause of adult neurological disability in the Western world 1 .

Further, stroke has been reported in 2-6% of patients hospitalized with COVID-19 2 .

Treatment options are limited and currently the only approved interventions are intravenous thrombolysis (IVT) with recombinant tissue plasminogen activator (tPA) and mechanical thrombectomy (MT) 3 . However, despite combined IVT and MT, up to 40% of the patients will die or remain functionally dependent, and more importantly, due to various contraindications, only a fraction of all ischemic stroke patients will be treated with these therapies. Much effort has therefore been invested in researching additional treatment options, mainly focusing on direct neuroprotection. However, due to the limited success of neuroprotection approaches, studies examining the therapeutic potential of preserving or restoring the integrity of the blood-brain barrier (BBB) have been gaining interest 4 . This is supported by our published study showing for the first time that a treatment strategy aiming to reduce BBB damage using the tyrosine kinase inhibitor imatinib improves neurological outcome in humans suffering from ischemic stroke 5 . Despite these advances the dynamics and functional relationship of cerebrovascular dysfunction with ischemic stroke progression remain largely unclear and require further investigation.

The BBB forms a mechanical and functional barrier between the systemic circulation and the central nervous system (CNS) and it tightly controls trafficking of substances between the blood and the CNS 4 . The barrier properties are established by the brain endothelial cells. It is however widely recognized that other cells in the neurovascular unit (NVU), including perivascular astrocytes and vascular mural cells (vascular smooth muscle cells (vSMC) and pericytes) as well as neurons and microglia, all work together in a coordinated way to maintain BBB properties 4 . Damage to the BBB is an early pathological event in ischemic stroke, and the absence of a functional BBB will lead to profound disturbances in neuronal and glial signaling 6 . Several molecular pathways have been reported to play important roles in disease-induced BBB damage, including our previous studies in mice demonstrating the role of tPA-activated platelet-derived growth factor CC (PDGF-CC) signaling via its tyrosine kinase receptor PDGF receptor α (PDGFRα) on perivascular cells in the NVU [7] [8] [9] [10] [11] [12] [13] . As aforementioned, administration of imatinib, a small molecule tyrosine kinase inhibitor of Abl, c-Kit and PDGFR 14 , significantly improves outcome after both ischemic and hemorrhagic stroke in rodents 7, 15, 16 and in humans 5 . This beneficial effect has been ascribed to imatinib's potential to reduce stroke-induced BBB leakage, but how this exactly translates to reduced lesion volume and improved neurological outcome is yet unknown. However, activation of the endothelium and subsequent BBB damage in the acute phase (hours after onset) of cerebral ischemia is known to accelerate secondary degeneration of the surrounding, initially spared, neural tissue, and this continues during the subacute phase (hours -days after onset) 17 . In addition, cells in the NVU act as sensors for brain injury eliciting activation of the early reactive gliosis response, in which astrocytes, NG2 glia (also referred to as oligodendrocyte progenitor cells (OPCs) or polydendrocytes) and microglia become activated, leukocytes infiltrate and a whole cascade of post stroke neuropathology is initiated 18 .

The general view is that reactive gliosis is an important early injury response, orchestrating subsequent formation of a glial scar during the tissue remodeling phase after insult (daysweeks after onset) 19 . It is believed that demarcation of the lesion by activated glia cells sequesters the toxic environment at the lesion site, thereby protecting the relatively unaffected surrounding CNS tissue. Also, isolation of the lesion is thought to allow regeneration of the injured tissue and thus enable recovery of CNS function in the chronic phase after insult. However, the role of the scar is highly debated and sustained gliosis has been reported to be deleterious to functional recovery 20 . For example, reactive gliosis and glial scarring, especially astrogliosis, are regarded as barriers to CNS regeneration and have been shown to have inhibitory effects on CNS axon regrowth 21 . Thus, these discrepancies highlight the need for a deeper understanding of the CNS injury response and repair.

It has been proposed that CNS responses to injury resemble that of normal wound healing processes in other tissues and that the poor CNS recovery is reminiscent of chronic/unresolved wounds 17 . A key healing phase in normal wound healing is 'scar contraction', which is achieved by injury-induced transdifferentiation and proliferation of specialized subsets of mesenchymal cells, mainly localized in a perivascular position, into contractile myofibroblasts 22 . These myofibroblasts contribute to repair by generating contractile forces in the tissue and are identified by expression of a-smooth muscle actin (ASMA), a marker not normally expressed in tissue mesenchymal cells at steady state. It is not yet clear if a similar process occurs in CNS lesions; however, hypoxia has been shown to promote myofibroblast transition in other organs 23 . Although beneficial initially, persistent myofibroblast expansion triggers pathological tissue contraction and results in overproduction of extracellular matrix (ECM), a pathology known as fibrosis 22 . This contributes to distortion of parenchymal architecture, which compromises organ recovery and impairs function. Interestingly, PDGF-CC/PDGFRα signaling has been associated with the fibrotic response in other organs [24] [25] [26] [27] [28] and nintedanib, a tyrosine kinase inhibitor that blocks the activity of PDGF receptors as well as that of FGF and VEGF receptors, has been shown to inhibit myofibroblast transdifferentiation and proliferation in patients with idiopathic pulmonary fibrosis 23 .

In the present study, we utilized a photothrombotic murine model of experimentally induced ischemic stroke in order to elucidate the molecular and cellular mechanisms of imatinib treatment. We found that stroke-induced BBB leakage was inhibited by imatinib and coincided with preserved cellular organization in the NVU in the acute phase after ischemia onset. BBB transcriptome analyses of vascular fragments isolated from the ipsilateral hemisphere of imatinib-treated mice and their controls identified differential gene expression mainly associated with inflammation and fibrosis pathways. Immunohistological analyses confirmed that imatinib dampened the reactive gliosis response and myofibroblast transdifferentiation after stroke, whilst having very limited effect on peripheral leukocyte infiltration and on the rest of the glial scar. Assessment of sensory-motor integration after stroke revealed that functional benefit with imatinib treatment progressively improved over time. This suggests that imatinib might ameliorate stroke outcome by counteracting the spread of fibrosis without affecting the ability of the rest of the glial scar to protect the healthy parenchyma from the toxic environment of the lesion. This study also offers novel insight into the relationship between BBB breach and fibrotic/glia scarring as well as the role of myofibroblasts in CNS injury response and repair.

We have previously shown that tPA-mediated activation of PDGF-CC/ PDGFRa signaling in the NVU during ischemic stroke in mice induces opening of the BBB and augments brain injury 7 . Blocking this pathway with the tyrosine kinase inhibitor imatinib improves outcome following ischemic stroke in both mice 7 and human patients 5 as well as in other neuropathologies in mice [8] [9] [10] [11] [12] 15 . However, very little is known about the kinetics and the underlying mechanisms of this inhibition. To advance this knowledge we performed in vivo BBB leakage, BBB transcriptomics and immunofluorescent analyses at different time points following MCAO induction in mice treated with imatinib or PBS as vehicle control (Fig. 1a, study design).

To study the kinetics of imatinib attenuated BBB permeability, we first determined the time course of MCAO-induced BBB leakage during the first 24 hours after ischemic onset in our photothrombotic model (Fig. 1b) . Evans blue (EB) extravasation analyses revealed a significant, time-dependent increase in EB extravasation in the ipsilateral hemisphere of vehicle control mice after MCAO compared to sham operated controls. We found that already 1 hour post MCAO the BBB is losing its integrity. At 3 hours post MCAO we detected the highest leakage, which was followed by a time-dependent decrease in EB extravasation, displaying a similar level of EB extravasation at 24 hours as at 1 hour post MCAO. These findings correlate well with previous reports demonstrating an early transient rise in BBB permeability in various reperfusion models of MCAO 29, 30 . To determine the kinetics of using imatinib to block MCAO-induced BBB leakage, we compared EB extravasation 3 and 24 hours post MCAO in imatinib-treated mice to vehicle controls. The results showed a significant reduction in EB extravasation of approximately 34% 3 hours and 46% 24 hours post MCAO in imatinib-treated mice compared to vehicle controls (Fig. 1c) .

The latter correlates well to previously published data 7 . Representative images illustrate the transient nature of MCAO-induced BBB leakage, with extensive EB extravasation 3 hours post MCAO and less at 24 hours, which is strongly attenuated by imatinib treatment (Fig.   1d ).

Immunofluorescent stainings revealed that the extensive BBB breach observed 3 hours post MCAO coincided with changes in the NVU of PDGFRa + vessels in the ischemic region of vehicle control mice, whereas imatinib treatment preserved a more normal appearing organization similar to that seen in naïve unchallenged brains (compare Fig. 1e to Extended data Fig. 1) . In naïve brains, as well as in the non-ischemic contralateral hemisphere, PDGFRa was found in perivascular cells staining positive for aquaporin 4 (AQP4) and glialfibrillary acidic protein (GFAP) but not in AQP4 + cells distributed along capillaries (Extended data Fig. 1a -c) . PDGFRa was also detected in non-vascular NG2-glia cells (Extended data Fig. 1a, d) . Further, the PDGFRa + perivascular cells were located on the parenchymal side of the vascular smooth muscle cell (vSMC) layer in alpha smooth muscle actin (ASMA) positive vessels (Extended data Fig. 1c ). In our analyses of stroked brains 3 hours post MCAO we found that imatinib treatment preserved high expression of PDGFRa and GFAP around vessels within the ischemic area, whereas it was absent or scattered around vessels in the ischemic area of vehicle controls (Fig. 1e) . Meanwhile we noticed an increase of non-vascular GFAP signal in vehicle controls (asterisk, Fig. 1e ). Quantification of PDGFRa + and GFAP + vessels in the ischemic area confirmed these observations (Fig. 1f ). This is in line with our previous published data showing that increased BBB opening after MCAO is mediated via early activation of PDGFRa signaling in perivascular astrocytes 7, 8, 31 and suggests that strategies targeting BBB integrity might be most effective if provided in the early phase after ischemic stroke.

To further investigate imatinib's effect after ischemia we performed transcriptome analysis on cerebrovascular fragments isolated 3 and 24 hours post MCAO from imatinib-treated mice or vehicle controls. Differential gene expression analysis was performed by microarray hybridization (GeneChip Gene 2.0 ST Array) or by qPCR. Using a P-value < 0.05 and a log2 (0.5) fold change as cutoff, we identified 121 and 85 differentially expressed transcripts in the cerebrovascular fragments from imatinib-treated mice at 3 hours (Supplementary table 1) and 24 hours (Supplementary table 2) 

Based on our BBB transcriptomics analysis we next investigated the effect of imatinib on reactive gliosis, the initial events of fibrosis and inflammation in response to insult. Among the glial cells taking part in the reactive gliosis response are astrocytes, NG2-glia cells and microglia/macrophages, which are recruited to the site of insult 19 . Since astrocytes are known to react to injury by hypertrophy and up-regulation of GFAP in the acute phase following insult 35 , we assessed astrogliosis by immunofluorescent staining for GFAP 3 hours post MCAO. These data demonstrated that there was significantly increased GFAP signal in the ischemic area (demarcated with a dashed line) compared to the non-ischemic surrounding tissue (Fig. 3a, b) . Our stainings revealed that imatinib inhibited this MCAO-induced activation of astrocytes, displaying approximately 60% less GFAP positive signal in the ischemic area compared to vehicle controls 3 hours post MCAO (quantified in Fig. 3c ). This pronounced effect seen on astrocytes is even more remarkable given that imatinib treatment preserves perivascular GFAP expression around vessels (see Fig. 1e , f and arrows, Fig. 3b) and is supported by our gene profiling analyses, showing that many of the imatinib regulated genes in the cerebrovascular fragments are normally expressed in perivascular astrocytes within the unchallenged NVU (Fig. 2d, Supplementary table 3 ). Further analysis of the reactive gliosis response showed that NG2-glia cell condensation, determined by staining for PDGFRa and neuron glia antigen-2/CSPG4 (NG2), appeared in the ischemic border of vehicle control brains 3 hours post MCAO (two headed arrows) but was inhibited by imatinib treatment (Fig. 3d -f ). Likewise, imatinib reduced microgliosis/macrophage recruitment after MCAO, as determined with staining for CD11b (microglia, arrows; activated microglia/macrophages, two headed arrows, Fig. 3g -h). Quantification revealed no detectable (ND) activated microglia/infiltrating macrophages 3 hours post MCAO whereas at 24 hours and 7 days post MCAO, pronounced microglia activation/macrophage infiltration was detected and this was significantly reduced by imatinib treatment (Fig. 3h ). Considering our earlier findings with the CX3CR1 GFP /CCR2 RFP reporter mice, showing very few infiltrating macrophages up to 24 hours post MCAO 31 , the effect of imatinib, at least up to 24 hours after MCAO, appears to be mainly mediated by inhibiting activation of intrinsic microglia. This is further supported by our gene expression analyses in Fig. 2 increased. It is of course possible that imatinib affects recruitment of macrophages in later phases after MCAO. This is supported by imatinib-induced upregulation of Nov expression, a factor reported to reduce monocyte adhesion. Apart from this potential effect on macrophage recruitment, our immunohistochemical analyses suggest that imatinib has only limited effect on peripheral immune cell infiltration after MCAO, including neutrophils and T cells (Extended data Fig. 2 ). Taken together this indicates that imatinib attenuates the reactive gliosis response after MCAO. This is supported by qPCR analyses showing that MCAO-increased expression of Il1a, Tnfa and Ccl2, known drivers of endothelial cell activation, microglia activation/macrophage infiltration as well as fibrosis, was dampened by imatinib treatment (Fig. 3i) .

Imatinib reduces PDGFRa fibrotic scar formation through inhibition of myofibroblast transdifferentiation without affecting the formation of the astrocyte-or NG2-glia scar 7

Reactive gliosis is an immediate and critical response to CNS injury, that is believed to orchestrate the subsequent formation of the glial scar. The glia scar is in turn believed to contain the damage and thereby improve outcome after injury, although, sustained gliosis has been reported to be deleterious to functional recovery 19 . Hence, the role of the scar is highly debated. To investigate whether the early effect of imatinib on reactive gliosis affects glial scar formation in the subacute/chronic tissue remodeling phase 7 days post MCAO, we performed immunofluorescent co-stainings of PDGFRa with markers of the glial scar; GFAP for the astroglia scar ( Fig. 4a) and NG2 for the NG2-glia scar (Fig. 4b) . These stainings revealed that imatinib treatment did not markedly affect the GFAP + or the NG2 + scar, whereas it significantly reduced formation of a PDGFRa + scar 7 days post MCAO compared to vehicle controls (Fig. 4c ). We noticed that the PDGFRa + scar was localized on the ischemic core side of the GFAP + scar and that there was very little, if any, co-localization of GFAP and PDGFRa expression in the scar (Fig. 4a , higher magnification shown in Extended data Fig. 3 ). The PDGFRa + scar appeared more disorganized in vehicle controls than in imatinib-treated animals and was embedded within the NG2 + scar (Fig. 4b) . However, contrary to the normal expression pattern of PDGFRa in the brain (see Extended data Fig. 1 ), the majority of the parenchymal PDGFRa expression within the scar was detected in NG2cells (Fig. 4d ). This ectopic expression of PDGFRa was found to co-localize with de novo parenchymal expression of ASMA ( Fig. 5a -d) , a commonly used marker for myofibroblast transdifferentiation/expansion following wound repair in peripheral tissue 38 . It should be noted that in the healthy murine brain ASMA expression is restricted to vSMC and no parenchymal expression of ASMA is observed (Extended data Fig. 4a ). Analysis of the ASMA stainings revealed that MCAO induced pronounced myofibroblast transdifferentiation/ expansion in the PDGFRa + portion of the scar of vehicle controls, which was much reduced by imatinib treatment (Fig. 5a, b) . It appeared as if the double positive ASMA + PDGFRa + cells were leaving the vessel wall (Extended data Fig. 4b , c), suggesting the myofibroblasts might originate from perivascular cells, although this will require in depth lineage tracing analysis. Since myofibroblasts in peripheral organs are known to secrete extracellular matrix (ECM) molecules that further promotes disease pathogenesis 38 , we next performed stainings for the ECM glycoprotein fibronectin 7 days post MCAO (Fig. 5e,f) .

These analyses reveled that the PDGFRa + cells were surrounded by fibronectin-positive ECM and that less fibronectin-positive ECM was detected in the ASMA + PDGFRa + scar of imatinib-treated mice compared to controls, thus further supporting a role of imatinib in inhibiting myofibroblast transdifferentiation/expansion after MCAO. Interestingly it appeared as if the fibrotic scar within the lesion core was unaltered by imatinib treatment and that imatinib thus specifically targeted the expansion of the ASMA + PDGFRa + scar (Extended data Fig. 5 ). Because detrimental scarring is attributed to sustained myofibroblast expansion in pathologic wound healing in peripheral tissue, it is therefore possible that inhibition of this expansion is central to the beneficial effect seen when using imatinib in ischemic stroke.

To test functional outcome following imatinib treatment we assessed lateralized sensorymotor integration 3 and 7 days after MCAO (Fig. 6a , study design) using the corridor task modified for mice 39, 40 . This test is based on the fact that mice with unilateral brain lesions display contralateral neglect and thus, preferentially explore/retrieve objects/food placed on the side ipsilateral to the lesion. Functional benefit of a treatment will thus result in reduced ipsilateral bias. We found that 3 days after MCAO all vehicle-treated mice preferentially explored sugar pellets from the side ipsilateral to the lesion (Fig. 6b ). This ipsilateral exploration bias in vehicle controls largely remained 7 days after MCAO. However, imatinib treatment significantly reduced ipsilateral exploration bias compared to vehicle controls, both at 3 and 7 days post MCAO, although the effect was more pronounced at 7 days (Fig. 6b ). In fact, at 7 days there was no significant difference between imatinib-treated mice and sham operated controls. As expected, imatinib significantly reduced lesion volume compared to vehicle controls 7 days post MCAO (Fig. 6c) . Our analysis indicated a 37% reduction in lesion volume, which is in line with previously published data where therapeutic imatinib reduced the lesion by 34% 3 days post MCAO 7 . We found that lesion volume significantly correlated with exploration bias 7 days post MCAO, with all imatinib-treated animal clustering in the lower, left quartile whereas vehicle controls mainly clustered in the upper, right (Fig. 6d ). This could suggest that functional outcome might merely reflect lesion size.

However, imatinib-treated mice displayed a greater reduction in ipsilateral exploration bias between day 3 and day 7 than vehicle controls (52% vs. 10% decrease, respectively) (individual values and group means shown in Fig. 6e ), despite similarities in lesion volume reduction after imatinib treatment at the two different time points. Further, correlation of the 3-day bias data with the 7-day data showed a clear separation of the lines and significantly different correlations in vehicle vs. imatinib-treated mice (Fig. 6f) . These data are consistent with the hypothesis that the consolidation of the fibrotic scar in imatinib-treated mice contributes to the improvement in functional outcome compared to the vehicle controls where the fibrotic scar is expanding. This suggests that the inhibition of myofibroblast expansion in the subacute/chronic phase of stroke recovery may enhance regenerative capacity and thereby contribute to improved functional recovery.

In conclusion, our findings show that imatinib dampens MCAO-induced cerebrovascular activation and the reactive gliosis response in the acute phase after ischemia. This is associated with reduced myofibroblast transdifferentiation and expansion in the subacute/chronic phase, which potentially improves regenerative capacity and thereby contributes to functional recovery in stroke management (Schematic illustration Fig. 7 ).

The molecular and cellular mechanisms of scar formation in the CNS are still poorly understood. Here, we found that imatinib dampened the early reactive gliosis response in the acute phase of ischemia and reduced myofibroblast transdifferentiation/expansion and fibrotic scarring in the subacute/chronic phase. We have previously postulated that the mechanism for the effect of imatinib in stroke, downstream of BBB preservation, potentially involves inhibition of immune cell infiltration 5, 31 . However, our analyses here point at a very limited effect of imatinib on peripheral leukocyte recruitment post MCAO, suggesting that it is unlikely that the attenuation of ischemia-induced systemic inflammatory processes account for a significant part of the beneficial outcome associated with imatinib treatment 5, 7 . This is in line with studies failing to show beneficial effects with immunomodulatory treatments in clinical settings of ischemic stroke 41 . Instead our data reveal that in addition to protecting the BBB, imatinib also modulates formation of the PDGFRa + myofibroblast portion of the fibrotic scar following MCAO. We hypothesize this might be central to the beneficial effect seen on neurological and functional outcome, especially given that sustained myofibroblast expansion is well known to be detrimental in healing processes in many organs, although less is known about this process in the CNS 38 . This hypothesis is further supported by our findings that consolidation of the PDGFRa + fibrotic scar in imatinib-treated mice seemingly works to discriminate behavioral outcome.

CNS injury repair has long been viewed as being markedly different from repair processes in other organs, mainly because of the poor regenerative capacity and immune-privileged nature of the CNS 17 . The insufficient regenerative capacity has largely been ascribed to the formation of a glial scar, in particular the astrocytic scar has been regarded as a barrier to CNS axon regrowth 20 . Recent studies, however, have shown that the presence of an astrocytic scar is not a principal cause for the failure of CNS axons to regrow. Instead, scar-forming astrocytes have been found to permit and support robust amounts of CNS axon regeneration in an experimental murine model of spinal cord injury 21 . These controversies have raised the question whether there might be different subtypes of reactive astrocytes eliciting different protective/harmful functions following injury and whether these potentially different astrocyte pools are activated in a time-and disease-specific manner. Toward this end, recent work has demonstrated that there are at least two different types of reactive astrocytes, neurotoxic A1 astrocytes and neuroprotective A2 astrocytes, which are induced in disease-specific manners 42, 43 . Our data support the existence of different subtypes of disease activated astrocytes. For example, we report a strong effect of imatinib on preserving perivascular astrocyte coverage in the NVU in the acute phase after MCAO, likely via direct inhibition of MCAO-induced PDGFRa signaling in these cells. Thus, supporting a protective role of these perivascular cells in maintaining vessel health and integrity. However, imatinib treatment also inhibited MCAO-induced activation of, what appears to be harmful, non-perivascular astrocytes in the ischemic core in the acute phase after MCAO induction. As these parenchymal astrocytes do not express PDGFRa, nor any of the other receptor tyrosine kinases known to be targeted by imatinib, including PDGFRb, Abl and c-Kit 14 , it seems likely this is mediated through an indirect mechanism. We speculate these parenchymal astrocytes might be activated by cytokines, including IL-1α, TNFa and C1q, which we found to be strongly upregulated in the vascular fragments after ischemia and normalized by imatinib, since these factors are known to stimulate astrocyte activation [44] [45] [46] and potently induce activation of neurotoxic A1 astrocytes 43 . Interestingly though, the astrogliosis dampening effect of imatinib within hours following MCAO did not translate into a differential thickness of the GFAP + astrocyte scar a week after injury. This might suggest that the astroglia scar originates from a subset of reactive astrocytes, not targeted by imatinib.

The formation of a scar following CNS injury, despite the potential inhibitory effect on axon regeneration, is thought to be crucial in order to seal off the lesion site from unaffected brain regions, thereby allowing temporal and spatial control of tissue remodeling of the injured tissue 17 . However, equally important as the formation of the scar is the termination of the scarring process. In most tissues this final key event of the normal healing process includes the appearance of injury-induced myofibroblasts contributing to 'scar contraction' and resolution of the injury by generation of strong contractile forces 17, 22 , although very little data is available on this process in CNS healing. In addition to their contractile role, myofibroblasts also secrete ECM molecules that influence mechanical signaling and cell adhesion. Persistence and/or dysregulation of 'scar contraction' and excessive ECM deposition leads to distortion of the parenchymal architecture, promoting disease pathogenesis and in worst cases tissue failure, a process referred to as fibrosis. Here we report that genes important for the fibrotic response, including Ccl2, Il1a, Tnfa, Ccl5, Ccl7, Fgr, and The ASMA + myofibroblasts were found to co-express PDGFRa. It is possible that these cells arise from perivascular cells as has been suggested in wound healing processes in peripheral organs 22 and that they upregulate ASMA as they transdifferentiate into contractile myofibroblasts leaving the vessel wall, although this will require further analysis e.g. by genetic lineage tracing. Imatinib treatment reduced this myofibroblast response, as seen in the immunohistochemical analyses, but also indicated in the gene expression analyses where genes highly implicated in activation and proliferation of myofibroblasts 48 , including Pdgfra, Il1a, Il1r2, Ccl2, Ccl5, and Hpse, were downregulated. Similar to the heterogeneity with injury-induced astrocytes, it has been suggested that the ASMA + myofibroblasts only accounts for a subset of injury-induced mesenchymal cells 22 . Different subsets of activated mesenchymal cells have been proposed to vary spatiotemporally after injury as well as to elicit different protective/harmful functions in the repair processes. Since imatinib treatment resulted in a myofibroblast scar that appeared to be more well-structured than in the untreated controls and inhibited, but did not completely block, its formation, this might indicate differential effects on distinct subsets of activated mesenchymal cells. Understanding the cellular origin and potential multilineage differentiation of the injury-induced mesenchymal cells is therefore a central issue for future research.

The potential progenitors for myofibroblasts are a matter of some debate and have been proposed to include epithelial-/endothelial-mesenchymal transition; circulating bone marrow-derived fibrocytes; tissue-resident fibroblasts or other mesenchymal cells related to blood vessels, such as pericytes, adventitial cells, and mesenchymal stem cells 22 . The currently prevailing dogma however seems to lean toward the latter, where specific subsets of tissue-resident mesenchymal cells, mainly localized in a perivascular position, serve as the major source for ECM-producing cells after injury. Some of these perivascular progenitors have in peripheral organs been shown to express PDGFRα. For example, in the liver it has been shown that the primary progenitors for myofibroblasts are perivascular hepatic stellate cells and specific deletion of PDGFRα from these cells in vivo resulted in reduced myofibroblast transdifferentiation and fibrosis in a model of hepatotoxic liver injury 28 . In the CNS, lineage tracing analyses suggest that perivascular cells in the NVU are important for fibrotic scar tissue formation 49, 50 and that reduction of this scarring response promotes recovery after spinal cord injury in mice 51 . These perivascular cells were shown to express PDGFRα and the authors referred to them as type A pericytes 49 , although this has later been challenged by single-cell analyses suggesting the perivascular PDGFRα + cells are not pericytes and should instead be referred to as fibroblast-like cells 33 . We, on the other hand, have been referring to the perivascular PDGFRα + cells in the CNS as perivascular astrocytes based on immunofluorescent studies indicating co-expression of GFAP and AQP4 with PDGFRα along medium to large size vessels throughout the CNS (Extended data Fig. 1 In conclusion, our data suggest that imatinib improves outcome of ischemic stroke by inhibiting disease-induced cerebrovascular changes and thereby expression of profibrotic genes from the activated BBB in the acute phase after ischemia onset. Consequently, this leads to reduced myofibroblast transdifferentiation/expansion in the subacute/chronic phase. This order of events is supported by our findings that imatinib appeared to be more beneficial if administered to patients within 5 hours from stroke onset, coinciding with the timing of the highest peak of cerebrovascular permeability, whereas the beneficial effect tailed off if imatinib treatment was initiated at a later time point 5 . We propose the effect is mediated through inhibition of ischemia-induced activation of PDGFRα signalling in perivascular cells in the NVU, which is in line with our previous findings showing that PDGFRα phosphorylation is induced in the NVU within hours following MCAO induction 31 . It should be noted though that imatinib is not exclusively inhibiting PDGFRα but also PDGFRb, Abl and c-kit, of which PDGFRb is of particular interest given that this receptor is also expressed in perivascular PDGFRα + cells of the NVU 33,49 and has been implicated in myofibroblast activation in injury responses in peripheral organs 54 . Contrary to PDGFRα however, PDGFRb is also expressed in vascular mural cells, both in pericytes distributed along the capillaries and vSMC 33 . Also, during ischemic stroke, expression of PDGFRb is highly upregulated throughout the ischemic core region, and not selectively only at the rim of the astroglia scar 50, 55 . More importantly, reduced ischemia-induced expression of PDGFRb in the lesion core has been found to correlate with an enlargement of infarct volume after MCAO 56 and stromal PDGFRb + pericytes have been reported not to contribute to ECM production in the fibrotic lesion after stroke 57 . The latter is however in stark contrast to the lineage tracing studies of type A pericytes 50 , where type A pericytes were shown to give rise to nearly the entire fibrotic scar in the lesion core after cortical MCAO, a region of the fibrotic scar that appeared unaffected by imatinib treatment in our analyses. Nevertheless, this highlights a heterogeneity of the fibrotic scar, where it seems that reducing the PDGFRα + myofibroblast portion is beneficial, whereas tampering with the fibrotic core portion of the scar might be detrimental, although this requires further investigation. In addition, it appears the different parts of the fibrotic scar stem from discrete progenitors, thus opening the possibility of differential targeting. Further studies, e.g. utilizing more specific ways to explicitly target the PDGFRα pathway genetically or pharmacologically, including the use of a monoclonal PDGF-C antibody 58,59 , as well as longitudinal in vivo imaging approaches to lineage trace the response of these enigmatic perivascular PDGFRα + cells to ischemia, are therefore warranted.

Taken together, this offers new opportunities to study and understand the multidimensional roles and complex cellular interactions of fibrotic scar formation and function in ischemic stroke but potentially also in other neurological pathologies.

All experiments in this study were approved and performed in accordance with the guidelines from the Swedish National Board for Laboratory Animals and the European Community Council Directive (86/609/EEC) and were approved by the North Stockholm Animal Ethics Committee and the Institutional Animal Care and Use Committee of Unit for Laboratory Animal Medicine at the University of Michigan, respectively.

Age-and gender matched C57BL/6 wild type mice were anesthetized with isoflurane and securely placed under a dissecting microscope. The left middle cerebral artery (MCA) was exposed and a laser Doppler flow probe (Type N (18 gauge), Transonic Systems) was placed on the surface of the cerebral cortex located 1.5 mm dorsal median from the bifurcation of the MCA as described before 7 . The probe was connected to a flowmeter (Transonic model BLF22) and relative tissue perfusion units (TPU) data was recorded with a continuous data acquisition program (Windaq, DATAQ Instruments). The photoactivatable dye Rose Bengal (Fisher Scientific) was diluted to 10 mg/ml in phosphate buffered saline (PBS) and then injected intravenously with the final dose of 40 mg/kg. A 3.5-mW green light laser (540 nm, Melles Griot) was directed at the MCA from a distance of 6 cm at the onset of the injection, and the TPU of the cerebral cortex was recorded. Total MCA occlusion (MCAO) was achieved when the TPU dropped to less than 30% of pre-occlusion levels and to achieve a stable clot the laser was left on for 20 minutes. As controls, C57BL/6 mice from the same cohort were shaved, skin was cut, the muscle retracted and either laser or Rose Bengal was used (sham operated mice). To alleviate postoperative pain, carprofen (Rimadyl Vet.®, Pfizer) was administered by subcutaneous injection at the start of the surgical procedure.

To block PDGFRα activation after stroke, mice were treated with the tyrosine kinase inhibitor imatinib (Novartis, Switzerland alternatively Mylan AB, Sweden) three times 

For analysis of cerebrovascular permeability after MCAO, stroked mice were injected with 100µl of 4% Evans blue dye (Sigma-Aldrich) intravenously 1 hour prior to sacrifice. The animals were then transcardially perfused with PBS for 5 minutes under isoflurane anesthesia and the brains removed and photographed using a Canon PowerShot SX200IS camera.

Thereafter the brains were separated into hemispheres and each hemisphere was then homogenized in N,N-dimethylformamide (Sigma-Aldrich) in Precellys lysing tubes and centrifuged twice at 16,100 x g for 20 minutes. The supernatants were collected, EB extravasation determined by absorbance measurement and quantified separately in the ipsiand contralateral hemispheres as described in 60 . Background EB level in the non-ischemic contralateral hemisphere was subtracted from the ischemic hemisphere. EB levels in each hemisphere were calculated using the following formula:

(A620nm -((A500nm + A740nm) / 2)) / mg wet weight

Tissue preparation for sectioning and immunostaining was conducted using standard protocols. Mice were anesthetized with isoflurane and transcardially perfused with PBS followed by fixation with 4% paraformaldehyde (PFA) in PBS. The brains were dissected and postfixed in 4% PFA 1 hour at room temperature (RT) and then kept in 30% sucrose 

Real-time quantitative PCR was performed using KAPA SYBR FAST qPCR Kit Master Mix Subsequent analysis of the gene expression data was carried out in the freely available statistical computing language R (http://www.r-project.org) using packages available from the Bioconductor project (www.bioconductor.org). In order to search for the differentially expressed genes between the different groups an empirical Bayes moderated t-test was then applied, using the 'limma' package. To address the problem with multiple testing, the Pvalues were adjusted using the method of Benjamini and Hochberg.

To compare gene expression in vascular fragments from imatinib-treated mice versus vehicle controls 3 hours and 24 hours post MCAO, molecules from the dataset that met the < or > log2 (0.5) fold change and P-value < 0.05 cutoff were uploaded to the ingenuity pathways analysis platform (Ingenuity Systems, CA, USA, www.ingenuity.com). The molecules in this dataset were grouped in biological functions and/or diseases or were associated with a canonical pathway in Ingenuity's knowledge base. Right-tailed Fisher's exact test was used to calculate a P-value determining the probability that each biological function and/or disease assigned to that data set is due to chance alone. The significance of the association between the data set and the canonical pathway was measured in 2 ways: 1) A ratio of the number of molecules from the data set that map to the pathway divided by the total number of molecules that map to the canonical pathway is displayed. 2) Fisher's exact test was used to calculate a P-value determining the probability that the association between the genes in the dataset and canonical pathway is by chance alone. Raw data are deposited on the NCBI Gene Expression Omnibus database (accession no. GSE137534). To compare our dataset to the harmonizome database 32 we used the harmonizome dataset named "fibrosis, CTD Gene-Disease Associations". In the harmonizome database every gene included in a certain functional annotation has a standardized value indicating the relative strength of the association.

Standardized values range from 1 to 2,88. We compared our dataset with all genes from the fibrosis dataset showing a higher standardized value than 1,5. The association with inflammation and metabolism was done manually and according to the current scientific knowledge.

To assess functional recovery we measured lateralized sensory-motor integration using a corridor task adapted from 39, 40 . Briefly, as the testing corridor a plexiglass box (L=60 cm x W=4 cm x H=15 cm) was used, where ten pairs of adjacent Eppendorf caps, each containing 4-5 sugar pellets (20 mg per pellet; TestDiet), had been placed at 5-cm intervals. A corridor with the same dimensions but without adjacent Eppendorf caps was used as the habituation corridor. 24 hours before MCAO, imatinib or vehicle-treated mice were habituated to the corridor by scattering sugar pellets along the corridor floor and allowing them to freely explore for 10 minutes. Sham operated mice were used as controls. Lateralized sensorymotor integration was tested 3 and 7 days after MCAO. On the testing day, mice were placed in the habituation corridor for 5 minutes in the absence of sugar pellets, then mice were transferred to one end of the testing corridor containing sugar pellets and video recorded for 5 minutes. All video recordings were analyzed by an investigator blinded to the experiment. A second investigator analyzed randomly selected videos independently to confirm scoring by the first investigator (approx. 16% of all videos were analyzed by two investigators). The number of ipsilateral and contralateral explorations relative to the stroked hemisphere were counted until the mouse made a total of 40 explorations or the video ended. An exploration was defined as a nose-poke into an Eppendorf cap, whether the sugar pellet was poked or eaten, and a new exploration was only counted by exploring a new cap. Data is expressed as ipsilateral bias, calculated as:

Ipsilateral bias (%) = ((ipsilateral (left) explorations -contralateral (right) explorations) / total explorations (left+right))*100

Following the last day of functional testing stroke volume analysis was performed as described previously 7, 31 . Briefly, brains were removed 7 days after MCAO and 2 mm thick coronal sections of the whole brain were stained with 4% 2,3,5-triphenyltetrazolium chloride (TTC) in PBS for 20 minutes at 37°C and then fixed in 4% paraformaldehyde solution for 10 minutes. TTC stained sections were captured with an Olympus digital C-3030 color camera attached to an SZ-60 Olympus microscope. The sections were analyzed with NIH Image J software using the following formula: V%stroke = ∑(lesion area)/∑(total area of ipsilateral hemisphere)*100, where V%stroke is stroke volume calculated as percent of the ipsilateral hemisphere.

Data analysis was performed using GraphPad Prism 9 statistical software (GraphPad Software, La Jolla, CA, USA). For statistical analysis in any experiment with only two groups a two-tailed t-test was used. For experiments with more than two groups, statistical evaluation was performed using one-way ANOVA with Fisher´s LSD with statistical significance defined as *P ≤ 0.05, **P ≤ 0.01 and ***P 

Yes. All applicable international, national, and/or institutional guidelines for the care and use of animals were followed. All procedures performed in studies involving animals were in accordance with the ethical standards of the institution at which the studies were conducted.

Drs. U. Eriksson, D.A. Lawrence, E.J. Su, and L. Fredriksson hold a patent on modulating blood-neural barrier using PDGFR-alpha antagonist. The other authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All animal experiments were reviewed and approved by the local animal ethics committees.

Raw data are deposited on the NCBI Gene Expression Omnibus database (accession no. GSE137534). Representative maximum intensity projections of 10 -11 µm confocal z-stacks are shown.

Scale bars, 1000 µm (a); 25 µm (c, e) and 10 µm (d). 

Heart disease and stroke statistics

Neurological associations of COVID-19

Targets for vascular protection after acute ischemic stroke

Blood-brain barrier dysfunction in ischemic stroke: targeting tight junctions and transporters for vascular protection

Randomized assessment of imatinib in patients with acute ischaemic stroke treated with intravenous thrombolysis

Pathophysiology of the neurovascular unit: disease cause or consequence

Activation of PDGF-CC by tissue plasminogen activator impairs blood-brain barrier integrity during ischemic stroke

Identification of a neurovascular signaling pathway regulating seizures in mice

Presymptomatic activation of the PDGF-CC pathway accelerates onset of ALS neurodegeneration

Imatinib treatment reduces brain injury in a murine model of traumatic brain injury

Imatinib enhances functional outcome after spinal cord injury

Imatinib ameliorates neuroinflammation in a rat model of multiple sclerosis by enhancing blood-brain barrier integrity and by modulating the peripheral immune response

Blocking PDGF-CC signaling ameliorates multiple sclerosis-like neuroinflammation by inhibiting disruption of the blood-brain barrier

Imatinib: a breakthrough of targeted therapy in cancer

PDGFR-alpha inhibition preserves blood-brain barrier after intracerebral hemorrhage

Longitudinal assessment of imatinib's effect on the blood-brain barrier after ischemia/reperfusion injury with permeability MRI

CNS sterile injury: just another wound healing?

The stem cell potential of glia: lessons from reactive gliosis

Reactive gliosis and the multicellular response to CNS damage and disease

The dual role of astrocyte activation and reactive gliosis

Astrocyte scar formation aids central nervous system axon regeneration

The perivascular origin of pathological fibroblasts

Nintedanib for the treatment of idiopathic pulmonary fibrosis

Platelet-derived growth factor C induces liver fibrosis, steatosis, and hepatocellular carcinoma

PDGF-C is a proinflammatory cytokine that mediates renal interstitial fibrosis

Elevated expression of PDGF-C in coxsackievirus B3-induced chronic myocarditis

Modulation of PDGF-C and PDGF-D expression during bleomycin-induced lung fibrosis

Hepatic Stellate Cell-Specific Platelet-Derived Growth Factor Receptor-alpha Loss Reduces Fibrosis and Promotes Repair after Hepatocellular Injury

Stroke-induced brain parenchymal injury drives blood-brain barrier early leakage kinetics: a combined in vivo/in vitro study

Rapid endothelial cytoskeletal reorganization enables early blood-brain barrier disruption and long-term ischaemic reperfusion brain injury

Microglial-mediated PDGF-CC activation increases cerebrovascular permeability during ischemic stroke

The harmonizome: a collection of processed datasets gathered to serve and mine knowledge about genes and proteins

A molecular atlas of cell types and zonation in the brain vasculature

Expression data from human brain samples, geo

Astrocyte barriers to neurotoxic inflammation

Ciliary neurotrophic factor (CNTF) plus soluble CNTF receptor alpha increases cyclooxygenase-2 expression, PGE2 release and interferon-gamma-induced CD40 in murine microglia

Decay accelerating factor (CD55) protects neuronal cells from chemical hypoxia-induced injury

Cytokine mediated tissue fibrosis

The Corridor Task: a simple test of lateralised response selection sensitive to unilateral dopamine deafferentation and graft-derived dopamine replacement in the striatum

Characterisation of behavioural and neurodegenerative changes induced by intranigral 6-hydroxydopamine lesions in a mouse model of Parkinson's disease

Anti-inflammatory treatments for stroke: from bench to bedside

Genomic Analysis of Reactive Astrogliosis

Neurotoxic reactive astrocytes are induced by activated microglia

Astrocytes: biology and pathology

Origin and progeny of reactive gliosis: A source of multipotent cells in the injured brain

Neuroinflammatory astrocyte subtypes in the mouse brain

Fibrotic scarring following lesions to the central nervous system

Chemokines in tissue fibrosis

A pericyte origin of spinal cord scar tissue

Pericyte-derived fibrotic scarring is conserved across diverse central nervous system lesions

Reducing Pericyte-Derived Scarring Promotes Recovery after Spinal Cord Injury

Platelet-derived growth factor C deficiency in C57BL/6 mice leads to abnormal cerebral vascularization, loss of neuroependymal integrity, and ventricular abnormalities

CNS fibroblasts form a fibrotic scar in response to immune cell infiltration

Novel insights into pericyte-myofibroblast transition and therapeutic targets in renal fibrosis

Early loss of pericytes and perivascular stromal cell-induced scar formation after stroke

Involvement of platelet-derived growth factor receptor beta in fibrosis through extracellular matrix protein production after ischemic stroke

Parenchymal pericytes are not the major contributor of extracellular matrix in the fibrotic scar after stroke in male mice

Development of monoclonal anti-PDGF-CC antibodies as tools for investigating human tissue expression and for blocking PDGF-CC induced PDGFRalpha signalling in vivo

Preclinical toxicological assessment of a novel monoclonal antibody targeting human platelet-derived growth factor CC (PDGF-CC) in PDGF-CChum mice

Tissue-type plasminogen activator induces opening of the blood-brain barrier via the LDL receptor-related protein

Microarray analysis of blood microvessels from PDGF-B and PDGF-Rbeta mutant mice identifies novel markers for brain pericytes

We would like to thank the Array and Analysis Facility, Science for Life Laboratory at