Abstract
Ischemic stroke causes a significant amount of cell death, including pericytes. Chronic perfused phase of ischemic stroke, the number of pericytes increases, and limited research has been explorated on the reasons for the recovery of pericytes. By genetic tracing, we identified ECs undergoing cell-fate changing, through TGFb-pSMAD3-EndoMT pathway transdifferentiation, into pericytes (E-pericytes). Some of the migrating E-pericytes induced by stroke are derived from detached ECs. The E-pericytes can integrate into vasculature again, suggesting ischemic brains can recycle ECs from unfunctional vessels to replenish lost pericytes as an innate self-maintenance program sustaining repair phenomena. We also found E-pericytes promotes BBB restoration, increases angiogenesis and restores cerebral blood flow(CBF), resulting more neurons survive and accelerating behavioral rehabilitation.
Main Text
Stroke kills millions each year and no effective treatment is available so far1. Brains possess an intrinsic, limited self-repair capacity2, 3, e.g. spontaneous angiogenesis after stroke. However, understanding the definitive spontaneous repairing processes is largely incomplete. ECs form vascular tubes, while pericytes encircle the endothelial tube. The observation that more pericytes die faster than ECs may explain neurodegeneration exacerbation and brain self-recovery hindering after stroke because pericytes are a type of pericytess known to be critical for the blood-brain barrier and cerebral blood flow4–6. However, little is known about cellular sources for pericyte regeneration in the adult ischemic brain while it has been well documented that the cephalic neural crest gives rise to pericytes in all blood vessels of the face and forebrain during embryonic development7, 8. Many capillaries collapse after stroke, however, the fate of endothelial cells from these unfunctional vessels to be recycled or to die is unclear. Recently, Yu et al. have demonstrated that pericytes can be differentiated locally from a subgroup of ECs in multiple tissues under homeostasis9. However, the case in brains with or without ischemia remains unknown. Identification of origin cells for regenerated pericytes and EC’s fate in the ischemic brain is of importance for enhancing endogenous capacity to repair a stroke-damaged brain.
First, we found that using flow cytometric analysis revealed that after 2 days of reperfusion, close to 50% of pericytes were lost, while endothelial cells accounted for less than 20% (Fig. 1a). With the prolonged reperfusion time, the numbers of pericytes and endothelial cells both show recovery, with the percentage of pericyte recovery exceeding 40% and the percentage of endothelial cell recovery approaching 10% at RP34D, comparion to RP2D(Fig. 1b-d). This indicates that in the chronic phase of ischemic reperfusion, there are ways for both endothelial cells and pericytes to increase the cell number. In ischemic stroke mice continuously treated with EdU by intraperitoneal injection every day for 34 days, we observed the presence of EdU+ pericytes(Fig. 1e-f). After statistical analysis(Fig. 1g), it was found that EdU+ pericytes accounted for less 20% of the total pericytes, suggesting that over 20% of pericytes may not be generated through self-proliferation, comparison with flow analysis results. Could these non-EdU+ pericytes originate from the transformation of endothelial cells?
To test if cerebral ECs differentiate into pericytes during adult angiogenesis as do those in the mammary gland, skin, and retina9, we employed a cell fate tracing system by crossing EC-specific DNA recombinase driver line Cdh5-CreERT210 with triple EGFP reporter (Ai47)11 or tdTomato reporter (Ai14) (The Jackson Laboratory) (Fig. 2a). ECs and their potential progeny cells are permanently labeled by EGFP or tdTomato fluorescent proteins once tamoxifen (TAM) is provided(Extended Data Fig. 2a). At the same time, When tamoxifen was not given, there was no EGFP signal(Extended Data Fig.2a-b). Upon long-term TAM exposure (Fig. 2b), although some of these EGFP+ ECs in skin tissue change their fates into EGFP+ pericytes9, brain EGFP+ ECs, however, seemed never (Fig. 2c-e). all EGFP+ cells remained expressing multiple classical endothelial markers, such as ERG, GluT1 and VE-cadherin, despite up to a two-year tracing (Extended Data Fig.2c-f). Consistent with previous reports9, 10, 12, these fluorescent proteins were restricted in ECs but not in any other types of brain cells short-term after TAM exposure (4 - 7 days, Extended Data Fig. 2g-m). At the same time, ECs and their potential progeny cells are never labeled by pericyte markers: PDGFRβ, NG2, CD13, Caldesmon and α-SMA (Fig.2f-h;Extended Data Fig.2n-q). Together, these results demonstrated unlike in the peripheral system and retina, cerebral ECs are not the cellular source for pericytes in adult brains under homeostasis.
Ischemic stroke, modeled by middle cerebral artery occlusion (MCAO, Extended Data Fig.2r-t), induced cerebral ECs detaching from vessels, gradually losing EC markers, basal membrane protein, and undergoing drastic morphological changes such as losing hollow structure and growing multiple processes from cell soma (Fig. 2i; Extended Data Fig.2u-v). This finding suggested that ECs might undergo cell fate changing in response to ischemia attack. Thus, we referred to this subgroup of EGFP+ cells as ‘EGFP+ non-EC cells’. TAM was delivered to 1.5-month-old mice and then at least two weeks later we performed MCAO surgery (Fig. 2b), aiming to ensure no extra residual TAM. After 2-hour MCAO, sutures were withdrawn, permitting various periods of reperfusion, ranging from 8 all the way through 514 days (RP2D to RP514D) (Extended Data Fig.2w-z). These EGFP+ non-EC cells can be found anywhere possible in the ischemic region, showing no regional specificity (Extended Data Fig. 2a1). According to whether having secondary processes, we classified EGFP+ non-EC cells into two types: type I, short, and fewer primary processes, no secondary processes (Extended Data Fig. b1,d1-e1); type II, longer and more primary branches with secondary ones (Extended Data Fig. b1,d1-e1), while their cell bodies were comparable in shape and size (Extended Data Fig. f1).
We found that the prevalence of these two types of cells switched when reperfusion period prolonged. (Extended Data Fig. c1). This morphological switching suggests type I becoming type II gradually. The prevalence of EGPF+ non-EC cells started to rise at RP8D, peaked by RP52D, and finally declined back to a low level (Extended Data Fig.z). This curve suggested that these transformed avascular EGFP+ non-EC cells may die or be recruited back to vessels, becoming vascular cells again. We termed this possibility a recycling process. Furthermore, FACS analysis (RP34D) revealed an EGFP+ but CD31 negative population (EGFP+ non-EC cells, Fig. 2j). These EGFP+ transformed cells accounted for ∼1.2% of total EGFP+ cells which included ECs and non-EC cells (Fig. 2k).
Cdh5-CreERT2 may label a small portion of neutrophils and monocytes in the bone marrow and circulation12 (Extended Data Fig. 3 a-b), which are known to enter brain parenchyma shortly after stroke onset but dies out13. To discriminate whether parenchyma EGFP+ non-EC cells are derived from EGFP+ bone marrow cells in Cdh5CreERT2;Ai47 mice, we introduced a viral expression system restricted to ECs preventing EGFP expression in hematopoietic cells (Extended Data Fig. 3c-e). Specifically, we injected cranially the Cre-dependent AAV-CAG-DIO-EGFP virus to Cdh5-CreERT2 single-positive mice that do not carry Ai47 or Ai14 reporter (Extended Data Fig. 3c), resulting in high labeling specificity to ECs (Extended Data Fig. 3f-g). Once again, we found EGFP+ ECs transforming into EGFP+ non-EC cells after stroke (Extended Data Fig. 3h). Notably, some of these cells extended their long processes and enwrapped capillaries, reminiscent of pericyte characteristics (Extended Data Fig. 3h). Independent radiation experiment was implemented to exclude the same possibility, we replaced Cdh5CreERT2;Ai47 mice’s bone marrow with wild type ones (Extended Data Fig. 3i-j). Despite the depletion of EGFP+ bone marrow cells under this circumstance (Extended Data Fig. 3i-j), we still found comparable EGFP+ non-EC cells in ischemic regions (Extended Data Fig. 3k-n), demonstrating that EGFP+ immune cells were not the cell source of EGFP+ non-EC cells in parenchyma. These results together indicate that a small proportion (1.2%) of ECs detached from vessels in the parenchyma, lost EC features, and changed their cellular identity after subjecting them to ischemia insult.
To assess whether these EGFP+ non-EC cells are derived from vascular stem endothelial cells as reported previously9 which involves cell proliferation, we traced cell division events by injecting EdU for consecutive 34 days. No single EGFP+ non-ECs were found to be EdU positive after examining 15 brain slices from two mice (Fig. 2l-m). similarly, neither Ki67 signals were found (Extended Data Fig. 4a-b). Thus, we concluded these EGFP+ non-EC cells may not be progenies of stem cells, suggesting that a transdifferentiation process may be required. Fortunately, EGFP+ non-EC cells appear at the classical pericyte location (Fig. 2n-o).
By employing a list of cell type markers, we identified that 90% of EGPF+ non-EC cells were pericytes (namely, E-pericytes) as they were positive for typical pericyte markers of CD13(90%), PDGFRβ(75%), and NG2 (Fig. 2p-s;Extended Data Fig. 4c), as well as that they were negative for all of the markers of immature and mature neurons (doublecortin and NeuN), astrocytes (GFAP), CNS macrophages (Iba1, CD68, and F4/80), immuce cells (CD45), and smooth muscle cells (α-SMA) (Extended Data Fig. 4d-f). The cell identity of the rest 10% of EGFP+ non-EC cells remained unknown (Fig. 3g) though as these cells were not labeled by any cell markers that we used. Furthermore, more than a quarter of EGFP+ pericytes were recruited back to vessels (Fig. 2p-q,v), suggesting ECs were fully recycled not only by expressing pericyte markers but also by locating the right anatomical position. A transformed tdTomato+ non-ECs (here, Ai14 reporter was used instead of Ai47) that were also tracked by AAV-DIO-EGFP virus exhibited one process attaching a capillary while the other did not (Extended Data Fig. 4g), implying capturing a recruiting moment. It has been repeatedly reported that pericytes detach from vessels after stroke14. We found actually a third of these avascular, migrating pericytes were derived from ECs (Fig. 2v). Further analysis revealed that 23.3% of total vascular pericytes were EGFP+ (Fig. 2t), suggesting ECs are a fair contributor cell source for pericyte regeneration. The virus-tracked ECs also can generate pericytes, coinciding with results from genetic tracing (Extended Data Fig. 4h-i). Meanwhile, we also employed other methods to specifically label endothelial cells to validate the E-pericytes phenomenon. Tie2Dre;Msfd2aXER mice were used for specific tracing of endothelial cells in the brain15. Firstly, we verified their specificity using endothelial cell markers(Extended Data Fig. 4j-k). After MCAO/RP34D, we still observed the presence of CD13+E-periytes (Extended Data Fig. 4n). We also use AAV-Bi30-Cre virus which can be traced ECs in mouse brain16. AAV-Bi30-Cre virus was injected via the eye socket in Ai47 mice and we still observed the presence of CD13+E-periytes (Extended Data Fig. 4i-m,o).These results together demonstrate that after ischemic stroke, ECs transdifferentiated into E-pericytes, likely capable of integrating back into the cerebral vasculature.
To further verify the identified EGFP+&CD31- cells as pericyte, we analyzed the transcriptomic characteristics of EGFP+&CD31- cells. First, we obtained EGFP+&CD31- cells (MCAO:2H/RP34D) using flow cytometric sorting. The flow cytometric gating strategy was CD45-&Viability+EGFP+&CD31- cells(Fig. 3a). The sorted cells from each group were as follows: non-stroke endothelial cells (Con-ECs), ipsilateral stroke endothelial cells (Ipsi-ECs), and EGFP+&CD31- cells derived from the ipsilateral side (Ipsi-ETCs). After library construction, sequencing was performed. PCA analysis revealed that Con-ECs and Ipsi-ECs had similar principal component relationships because they were both endothelial cells, although they were derived from different environments of the endothelial cells (non-ischemia and ischemia brain tissue, respectively)(Fig. 3b). On the other hand, Ipsi-ETCs showed greater dissimilarity to the principal component relationships of Con-ECs and Ipsi-ECs(Fig. 3b). This result remained consistent different batches, suggesting significant differences in the transcriptomes between Ipsi-ETCs and the endothelial cells represented by Con-ECs and Ipsi-ECs, even though Ipsi-ETCs were derived from EGPF+ normal endothelial cells. Heatmap analysis of all genes indicated that the transcriptomes of Con-ECs and Ipsi-ECs were more similar and clustered closely, despite some differences caused by the ischemic environment(Fig. 3c). In contrast, the transcriptome differences between Ipsi-ETCs and Con-ECs/Ipsi-ECs were more pronounced, and Ipsi-ETCs from different batches tended to cluster together(Fig. 3c). These findings indicated significant differences between the transcriptomes of CD45-&Viability+EGFP+&CD31- cells and endothelial cells. Subsequently, we analyzed the expression of endothelial cell marker proteins among different groups. We found that the expression of endothelial cell markers genes in Ipsi-ETCs significantly decreased compared to Con-ECs and Ipsi-ECs, and this decrease was observed across most endothelial cell marker genes(Fig. 3d). Con-ECs and Ipsi-ECs still expressed numerous endothelial cell marker genes, although differences existed between them(Fig. 3d). Additionally, we analyzed the expression of pericyte-related transcriptomes genes among different groups. We observed a significant increase in the expression of pericyte marker genes in Ipsi-ETCs, particularly in specific transcription factors(Fig. 3e), membrane proteins(Fig. 3f), and ligand proteins associated with pericytes(Fig. 3g). Conversely, the expression of pericyte-related marker genes was lower in Con-ECs and Ipsi-ECs, suggesting that the transcriptome of Ipsi-ETCs exhibited characteristics of pericyte cell transcriptomes genes. Furthermore, volcano plot analysis of ETCs/Ipsi-ECs identified differentially expressed genes with statistical significance(Fig. 3h). GO enrichment analysis revealed that the Ipsi-ETCs group was more enriched in functionally related genes associated with fate conversion, such as high expression of genes involved in cell differentiation, cell movement, cell migration, and endothelial cell migration, as well as low expression of genes related to endothelial cell junction genes(Fig. 3i). These findings were consistent with the observations in ischemic brain slices, where 2/3 of Ipsi-ETCs detached from the blood vessels. Additionally, KEGG pathway enrichment analysis revealed that genes related to metabolic pathways were upregulated, while genes related to endothelial cell tight junctions were downregulated(Fig. 3j). GSEA analysis showed that genes associated with endothelial cell features, such as endothelial cell differentiation, cell junctions, and cell apoptosis, were downregulated(Fig. 3k-m), while genes associated with glucose metabolism pathways and ATP production through aerobic respiration were upregulated(Fig. 3n-p). It is known that endothelial cells have fewer mitochondria and primarily produce ATP through glycolysis, whereas Ipsi-ETCs showed a significant increase in genes related to glucose metabolism, as well as genes involved in ATP production. This suggests that important genes in the metabolic pathway of glucose metabolism are altered in Ipsi-ETCs. These results collectively indicate that Ipsi-ETCs exhibit significant differences in the genomic profile compared to endothelial cells and show transcriptional characteristics of pericyte cells, providing further evidence from the transcriptome analysis that EGFP+ non-EC cells are E-pericytes.
The above experiments have demonstrated that the ischemic microenvironment after stroke promotes the transdifferentiation of endothelial cells into E-pericytes. Next, we attempted to explore the mechanism by which endothelial cells transform into E-pericyte. In 2016, Yu QC et al. reported the identification of endothelial cells expressing Protein C Receptor (Procr) in breast tissue, which possess characteristics of vascular endothelial stem cells (VESCs). Procr+ VESCs have strong proliferative abilities in vitro and also participate in vascular remodeling in live animals. Lineage tracing studies have found that this group of cells can transform into pericytes during breast development, and the transcriptome of Procr+ VESCs exhibits characteristics of Endothelial-to-Mesenchymal Transition (EndoMT) and Epithelial-to-Mesenchymal Transition (EMTs), suggesting that the transformation of endothelial cells into pericytes in breast development may occur through EndoMT9. We found that after ischemic stroke, the expression of EndoMT marker proteins in ischemic regions of endothelial cells was significantly increased. Specifically, at MCAO:2H/RP2D, the expression of phosphorylated SMAD3 (an impor-tant kinase in the TGF β pathway-activated by phosphorylation) as an EndoMT marker protein increased in the ischemic region(Fig. 4a-c). Furthermore, MCAO:2H/RP8D, the expression of p-SMAD3 in ECs also increased in the ischemic region(Fig. 4a-c). However, MCAO:2H/RP34D, endothelial cells almost did not express p-SMAD3 as an EndoMT marker protein, and E-pericytes cells also almost did not express p-SMAD3(Fig. 4a-c). At 8 days post ischemia, we also observed two adjacent EFGP+ cells on the same blood vessel, one cell exhibiting CD31+&p-SMAD3-, still maintaining endothelial cell characteristics, while the other cell exhibiting CD31-& p-SMAD3+, indicating that after p-SMAD3 is expressed in some endothelial cells, the characteristic molecule CD31 is reduced or not expressed(Fig. 4d), indicating successfully catching cells undergoing transition on time. At 2 days post ischemia, the expression of KLF4 (an important transcription factor in the EndoMT pathway) as an EndoMT marker protein in the ischemic region of endothelial cells increased(Extended Data Fig.5a-c). At 8 days post ischemia, the expression of KLF4 in the ischemic region of endothelial cells decreased, and at 34 days post ischemia, the expression of KLF4 as an EndoMT marker protein in endothelial cells further decreased, and E-pericytes cells almost did not express KLF4(Extended Data Fig.5a-c). At 2 days post ischemia, the expression of α-SMA (an important marker protein in EndoMT) in the ischemic region of endothelial cells increased(Extended Data Fig.5d-f). At 8 days post ischemia, the expression of α-SMA as an EndoMT marker protein in the ischemic region of endothelial cells decreased, and at 34 days post ischemia, the expression of α-SMA in endothelial cells also decreased, and E-pericytes cells almost did not express α-SMA (Extended Data Fig.5d-f). Furthermore, cytoplasmic flow cytometry demonstrated that at 2 days post ischemia, the expression of α-SMA in endothelial cells increased, reaching 1.6% of the proportion of endothelial cells in the ischemic region, which is close to the proportion of E-pericytes cells (1.2%) in the ischemic region(Extended Data Fig.5g-h). Next, we investigate the potential molecular signaling pathway that might mediate this E-pericytes. Since TGFβ signaling is well documented for cell fate changing including the endothelial mesenchymal transition (EndMT) process, a newly recognized type of transdifferentiation17,18. Therefore, we injected LY364947 and SB431542, TGFβRI inhibitors when MCAO was performed, followed by additional daily injections for consecutive 2 days of reperfusion. The expression of p-SMA3, KLF4 and α-SMA protein in the ischemic region of endothelial cells decreased in TGFβRI inhibitors group(Extended Data Fig.5i-j), and cytoplasmic flow cytometry demonstrated that at 2 days post ischemia, the expression of α-SMA in endothelial cells decrease(Extended Data Fig.5k-i). For consecutive 34 days of reperfusion, We found significantly fewer EGFP+ non-EC cells appeared overall(Extended Data Fig.5m-n). FACS analysis further confirmed this conclusion quantitively (Extended Data Fig.5o-p). The specific knockout of the Tgfbr2 gene in endothelial cells also resulted in a decrease of EndoMT markers (p-SMAD3, KLF4, and α -SMA) at the RP2D(Fig. 4e-h), along with a decrease in the number of E-pericytes(Fig. 4i-l). These results strongly suggested that TGFβ-pSMAD3-EndoMT signaling involved in E-pericytes is centrally important. Together, under ischemic conditions, ECs exploited TGF β signaling to transform ECs into pericytes during reperfusion to replenish the shrunk pericyte pool occurring in the acute ischemia period.
Pericyte-deficient mouse mutants shows that pericyte deficiency increases the permeability of the BBB to water and a range of low-molecular-mass and high-molecular-mass tracers19,20. Therefore, we speculate that E-pericytes participate in the recovery of the blood-brain barrier (BBB) during the chronic recovery period after ischemic stroke. E-pericytes can return to the blood vessels and perform functions such as maintaining BBB integrity, reducing the entry of harmful substances into the brain, promoting angiogenesis, increasing capillary density, and improving blood supply to facilitate the recovery of neurons and other cells, ultimately promoting self-repair.
Studies in mouse models showed that, in the absence of key TGF β receptors, angiogenesis stalls in the yolk sac at an early stage with fatal consequences21. The effect of endothelial cell-specific knockout of Tgfbr2 on blood vessel development and endothelial function in adult mice is unknown. Firstly, we found that specific knockout of the Tgfbr2 gene in endothelial cells does not affect the function of BBB (Extended Data Fig.6a). At the same time, it also did not affect the expression of the BBB-related protein ZO-1(Extended Data Fig.6b-c). Furthermore, under electron microscopy, the basement membrane remains intact(Extended Data Fig.6d-e), and the tight connections between endothelial cells are normal(Extended Data Fig.6f). There is also no obvisous occurrence of transcytosis phenomena caused by BBB disruption in the endothelial cells(Extended Data Fig.6g). This indicates that knocking out Tgfbr2 in adult mice does not affect the BBB function of vascular endothelial cells.
At different reperfusion time points, reducing E-pericytes by specifically knocking out Tgfbr2 in endothelial cells can aggravate the chronic BBB leakage. In the group with Tgfbr2 knocked out specifically in endothelial cells, the BBB function was significantly impaired when the reperfusion time exceeded 7 days(Fig.5a-b). When reperfused timely for 118 days, it still showed low Evan blue leakage(Fig.5a-b). This suggests that reducing E-pericytes through specific knockout of Tgfbr2 in endothelial cells can affect the recovery of chronic BBB during reperfusion.
We observed a decrease in NeuN+ neurons after ischemic stroke, with a more significant decrease in the reducing E-pericytes mouse group(Fig.5c-d). Additionally, we quantified the number of CD13+ pericytes in the ischemic area and found a significant decrease in pericyte density(Fig.5e-f), which is consistent with previous studies. In the reducing E-pericytes group, the reduction in pericyte density was more significant (Fig.5e-f). The ratio of endothelial cells to pericytes on the blood vessels was also significantly lower than the normal group(Fig.5g). We measured the length of CD31+ endothelial cells in the ischemic area and found a decrease in length(Fig.5h-i), and flow cytometry analysis showed a more pronounced decrease in CD31+ endothelial cells in the reducing E-pericytes mouse group, indicating decreased angiogenesis(Fig.5j-k). Furthermore, laser speckle contrast analysis revealed poorer blood flow recovery in the reducing E-pericytes mouse group(Fig.5i-m), as well as more pronounced brain atrophy (Fig.5n-o). This suggests that in the E-pericytes mouse group, reduced E-periytes lead to impaired blood flow recovery during reperfusion, resulting in more cell death and more significant brain atrophy. To test animal locomotion function, Tgfbr2fl/fl, ECTgfbr2fl/+, and ECTgfbr2fl/fl mice groups were subjected to 2 hours of ischemia followed by 34 days of reperfusion, and behavioral tests were conducted at different time points (0D, 1 day, 3 days, 7 days, 14 days, 21 days, and 34 days of reperfusion). Animal locomotion function tests related to motor function, including pole test, beam walking test, turning test, and adhesive removal test, were performed. Compared to the Tgfbr2+/+ group, mice in ECTgfbr2fl/fl group showed shorter pole test duration and had statistically significant differences starting from 21 days of reperfusion(Fig.5p); mice in ECTgfbr2fl/fl group spent less time on the beam and had statistically significant differences starting from 21 days of reperfusion(Fig.5q); mice in ECTgfbr2fl/fl group exhibited more pronounced differences in the turning test compared to the normal values, with statistically significant differences starting from 14 days of reperfusion(Fig.5r); mice in ECTgfbr2fl/fl group were less likely to remove the adhesive tape from their forepaws in the adhesive removal test, with statistically significant differences starting from 14 days of reperfusion(Fig.5s). These results indicate that the motor recovery ability of ECTgfbr2fl/fl group is poorer, which may be associated with reducng E-periytes.
Overall, using mouse genetic tracing tools combined with virus infection, pharmacology and genetically modified mice, we demonstrated adult ischemic brains spontaneously repair microvasculatures by regenerating new pericytes that are transformed from ECs after detaching from vessels. Distinct from the previous findings in peripheral organs in which a small group of endothelial stem cells Procr+ ECs differentiate into pericytes during adult angiogenesis under homeostasis9, ECs do not become pericytes in normal brains but transdifferentiate into pericytes without cell division. The death rate is dozens of times higher in pericytes than in ECs within 24 hours after a transient focal ischemia4, suggesting abundant capillaries become damaged by being naked. What is the fate of these naked capillaries made up by ECs? The present study suggests that these naked vessels may release ECs, some of which may not die but rather be recycled and repurposed by the ischemic brain. By observing type I simpler form switching to type II more complex EGFP+ E-pericytes (Extended Data Fig.2c1) and the medium stage of the transformed EGFP+ E-pericytes (Extended Data Fig. 4g), we extrapolate them likely being recruited and serve as building blocks to repair the existing damaged vasculatures. This could be an efficient system to reuse brain cells that have lost their original functions. Further studies are required to live imaging the whole dynamic process during reperfusion periods.
Identifying new cellular sources for pericyte replenishment is important not only for stroke but also for many other either acute or chronic brain injuries. This study may suggest a novel therapy for generating EC-derived pericyte without involving stemness which has long been a concern for tumorigenesis.
Methods
Experimental design
Transgenic Cdh5CreERT2 transgenic mice10 cross with reporter mice Ai47 (EGFPf/f) reporter or Ai14 (tdTomatof/f) reporter mice. Thus, ECs are labeled in green or red initially. TAM was delivered when mice were 1.5 months old (6 weeks), i.e. genetic tracing started. After waiting for 2 weeks, MCAO was performed. After a 2-hour occlusion, reperfusion began. Contralateral sides of the ischemic brain were used as objects for homeostasis conditions, except for the time point of tracing 4 days. Thus, it was traced for 22 days in the PR8D mice (N = 3 mice), 1.6 months in RP34D (N = 27 mice), 2.8 months in RP71D (N = 3 mice), 14 months in RP408D (N = X mice), 8.2 months in RP232D (N = 3 mice), 14 months in RP408D (N = 2 mice),17.6 months in RP51D mice (N = 2 mice). The tracing time point of 1.6 months (RP34D) was the most frequently examined. FACS and immunohistochemistry were the major two techniques used for cell identity confirmation.
Animal care and tissue dissection
All animal experiments were carried out in accordance with protocols approved by the Institutional Animal Care and Use Committee (IACUC) at the School of Life Sciences, Westlake University. Wild-type C57BL6/J mice were purchased from the Laboratory Animal Resources Center of Westlake University. Cdh5CreERT2 mice were a gift from Le-Ming Zheng (Peking University). Gt(ROSA)26Sortm47(CAG-EGFP*)Hze (Ai47) was share by Zilong Qiu (ION). We crossed Cdh5Cre-ERT2 mice to Ai47 mice to generate Cdh5CreERT2:Ai47 mice. Standard chow and water were provided to mice ad libitum. Four mice were housed in each cage. All animals were housed in a standard animal room with a 12/12-hour light/dark cycle at 25 °C. Both male and female mice were used in this study. Mice were anaesthetized by intraperitoneal injection pentobarbital sodium (100 mg/kg), and brain dissection was performed immediately for FACS experiments or cardiac PFA perfusion was performed for immunohistochemistry experiment. After fixation, brains were dissected and kept in 4% PFA for additional 4 hours and then transferred to PBS until further processing.
MCAO
Adult mice were anesthetized with pentobarbital sodium (100 mg/kg) and body temperature was maintained during surgery with a heating pad. A midline neck incision was made, the right common carotid artery was carefully separated from the vagus nerve, and the artery was ligated using a 5.0-string. A second knot was made on the left external carotid artery. The right internal carotid artery (ICA) was isolated, and a knot was left loose with a 5.0-string. This knot was not tightened until the intraluminal insertion was done. A small hole was cut on the common carotid artery before it was bifurcated to the external carotid artery and ICA. A silicon-coated monofilament (tip diameter = 230 μm, Doccol Corporation, 602256PK5Re) was gently advanced into the ICA until it stopped at the origin of the MCA in the circle of Willis. The third knot on the ICA was closed to fix the filament in position. During MCA occlusion (2 h), mice were kept in a warm cage in which the temperature was maintained at 35 °C.
Laser speckle contrast imaging (LSCI)
LSCI provides a measure of blood flow velocity by quantifying the extent of blurring of dynamic speckles caused by the motion of red blood cells through the vessels. Briefly, mice were placed under an RFLSI III device (RWD Life Sciences) before and after the suture was successfully inserted through CCA. The skull over both hemispheres was exposed by making an incision along the midline of the scalp. When a 785 nm laser is used to illuminate the brain, it produces a random interference effect that represented blood flow in the form of a speckle pattern. Scattering light was detected by a charge-coupled device (CCD) camera, and the images were acquired by custom software from RWD Life Sciences Company. For each acquisition, a total of 160 images, each of which measured 2048 ×2048 pixels, were collected at 16 Hz.
Tamoxifen
Tamoxifen, 10 mg/ml dissolved in 200 μl corn oil, was administrated intragastrically for 4 consecutive days. Prior to further experiments, fluorescent protein expressions in mouse ear vasculatures were first examined. Tamoxifen was administrated at least 2 weeks before MCAO surgery.
Immunohistochemistry
Cryosections of fixed mouse brains (50 μm) were handled free-floating and washed for 5 minutes in PBS prior to further procedures. Tissue sections were permeablized and blocked in the PBS containing 0.5% Triton and 5% BSA at room temperature for 1 hour. Then, brain sections were incubated with a different primary antibodies, including CD31(1:400, BD, Cat# 557355), VE-Cadherin(1:200, BiCellScientific, Cat# 00105), ERG(1:300, Sigma, Cat# ab110639), GLUT1 (1:300, Sigma, Cat# 07-1401), NeuN (1:300, Merck, Cat# ABN90P), Doublefortin (1:300, Santa clus, Cat# sc-271390), Iba1 (1:300, HUABIO, Cat# ET1705-78), GFAP (1:300, Proteintech, Cat# CL488-60190), PDGFRβ (1:300, ThermoFisher, Cat# 14-1402-82), NG2 (1:300, Merck, Cat# AB5320), CD68 (1:300, BIO-RAD, Cat# MCA1957T), CD45-APC (1:200, Biogend, Cat# 103112), Ki67 (1:300, thermoscientific, Cat# RM-9106-S0), F4/80 (1:300, Cell Signaling, Cat# 30325), CD13 (1:300, R&D system, Cat# AF2335), α-SMA-CY3 (1:300, Sigma, Cat# C6198), which were all diluted in the blocking solution (1% percelin, 5% BSA, and 0.1% Triton in PBS) at 4 °C overnight. After rinsing in PBS 3 times for 5 minutes each, brain slices were incubated with secondary antibodies accordingly, including donkey anti-mouse IgG (A21202, Thermo Fisher Scientific, 1:1,000) and donkey anti-rabbit IgG (A10040, Thermo Fisher Scientific, 1:1,000). They were diluted in the blocking solution. After a 2-hour incubation, brain slices were washed with PBS three times for 15 minutes each. Finally, Hoechst counterstaining was performed for each specimen. After mounting in an anti-fade mounting medium, images were acquired by using a Zeiss LSM800 laser-scanning confocal microscope with a 20X or 63X objective, with or without Airyscan mode.
Flow cytometry
Primary cell preparation was conducted as described below. Mouse brains were dissected into cold HBSS with glucose and BSA (HBSS++, Vendor info) were cut into small pieces by sterile scissors, followed by a centrifuge at 1600 rpm/min for 5 mins to get rid of supernatant. The minced tissues after the centrifuge were incubated with collagenase IV (2.5 mg/ml, ROCHE, Cat# 11088858001) at 37 °C with gentle rotation for 30 min in 2 ml of HBSS++. Next, samples were passed through a 70-μm filter and washed with HBSS++. To obtain higher cell yield, grind the tissue on the 70-μm filter using a syringe piston and guarantee all tissue pass the filter. The following antibodies in 1:200 dilutions were used: CD31-PE-594 (1:200, Biolegend, Cat# 102520), CD45-APC (1:200, Biogend, Cat# 103112), Ly6C-APC (1:200, Biolegend, Cat# 128016), CD11b-PE (1:200, Biolegend, Cat#101212), Viability dye (1:1000, ThermoFisher, Cat#65-0863-14). Antibody incubation was performed on ice for 30min in HBSS with glucose and BSA. All analyses were performed with CytoFLEX LX-5L1(Beckman). FlowJo x software was used for data analyses.
EdU injection
EdU (200 mg/kg, Alfa Aesar Chemical) was injected intraperitoneally for consecutive 34 days since when MCAO was performed. EdU was initially dissolved in DMSO at the stock concentration of XX mg/ml and was diluted with saline prior to injection.
Virus injection
The following adenovirus vector AAV-CAG-DIO-EGFP was purchased for ShuMi, Wuhan, China (PT-0168). Half of the one-micron virus was delivered to postnatal day 2 mouse pups by intracerebroventricular injection to both sides. After 1.5 months, tamoxifen was administrated. One month later, these mice were subjected to MCAO to induce ischemic stroke.
Radiation and bone marrow transplantation
At two weeks post tamoxifen injection, adult Cdh5CreERT2:Ai47 mice received radiation (total: 11Gy, 2 times, 5.5Gy per time, interval 2 hours) and bone marrow cell transplantation from wild-type mice with the same genetic background. RBCs were removed from the bone marrow cells by using ACK Lysis Buffer (4 Million cells/mouse). These mice were subjected to MCAO 2 months after bone marrow reconstruction.
Mouse cranial imaging by two-photon microscopy
Adult mice were anesthetized with pentobarbital sodium, and analgesia was provided by subcutaneous injection of 0.2% meloxicam. The scalp was removed, and the skull was exposed and cleaned. A dental drill with a 0.6-mm-diameter bit was used to engrave and thin the bone around the circular craniotomy area at a size of 3 mm. The piece of skull was carefully peeled off with fine-pointed forceps, and a cranial window was generated. A coverslip with a 3-mm diameter was placed on the cranial window, and its perimeter was completely sealed with a 1% agarose gel. The metal headplate was glued onto skull with dental acrylic, through which the mouse brain was fixed on a headplate holder. The headplate holder together with the mouse was placed under a two-photon microscope (FVMPE-RS, Olympus). The cerebral vasculature Cdh5CreERT2:Ai47 mice was visualized and imaged once a day before and after stroke. All surgeries were performed with sterilized instruments and an environment.
Statistics
The quantified data in all figures were analyzed with GraphPad Prism 7.0 (La Jolla, CA, USA) and presented as the mean ± SEM with individual data points shown. Unpaired two-tailed Student’s t-test was used for assessing the statistical significance between two groups. Statistical significance was determined by calculation of p-value (*p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001, ns: not significant). The repetition of our data is independent biological replicates and the number of replicates for each experiment is noted in the corresponding figure legend.
Data availability
The source data underlying the graphs and charts shown in the figures and tables are provided in Supplementary Data 1. All data generated or analyzed during this study are included in this published article (and its supplementary information file).
Author contributions
J.-M.J. conceived the research. T.L. and J.-M.J. designed experiments. T.L. performed all the experiments and data quantification, except for the virus cranial injection. Y.W. performed FACS together with T.L.. Z.Z. conducted some stroke models. Q.G. and B.Z. conducted mouse genotyping.
Competing interests
All authors declare no competing interests.
Acknowlegements
We thank Xuzhao Li, Zhu Zhu, Dongdong Zhang, Jinze Li, Jiayu Ruan, Lili Zhou for their constructive discussions. We thank Danping Lu for her responsive and timely purchasing support, Jiongfang Xie for sharing AAV virus. We thank the animal facility for its technical assistance with rodent housing, Biomedical Research center platform for technique support. This work is supported by the National Natural Science Foundation of China (31800864 and 31970969 to J.-M.J, 82101475 to Z.Z.), Westlake startup funds, Westlake Education Foundation and MRIC funds (103536022011) to J.-M.J.; Westlake Education Foundation to B.D.
Footnotes
1.We increased the number of pericytes changes at different reperfusion times after strok. 2.Lineage tracing: We increased other endothelial cell-specific mice to identify ECs-Transition-E-pericytes. 3.The Bulk RNA-seq identified the identity of E-pericyte. 4.Tgfbr2 gene deletion in endothelial cell increase the leakage of BBB in chronic stage of stroke. 5.Tgfbr2 gene deletion in endothelial cell in normal mice does not effect on BBB. 6.E-pericyte formation after stroke via TGFβ-pSMAD3-EndoMT pathway.