Ultrastructural studies of the neurovascular unit reveal enhanced endothelial transcytosis in hyperglycemia‐enhanced hemorrhagic transformation after stroke

Abstract Aims Pre‐existing hyperglycemia (HG) aggravates the breakdown of blood–brain barrier (BBB) and increases the risk of hemorrhagic transformation (HT) after acute ischemic stroke in both animal models and patients. To date, HG‐induced ultrastructural changes of brain microvascular endothelial cells (BMECs) and the mechanisms underlying HG‐enhanced HT after ischemic stroke are poorly understood. Methods We used a mouse model of mild brain ischemia/reperfusion to investigate HG‐induced ultrastructural changes of BMECs that contribute to the impairment of BBB integrity after stroke. Adult male mice received systemic glucose administration 15 min before middle cerebral artery occlusion (MCAO) for 20 min. Ultrastructural characteristics of BMECs were evaluated using two‐dimensional and three‐dimensional electron microscopy and quantitatively analyzed. Results Mice with acute HG had exacerbated BBB disruption and larger brain infarcts compared to mice with normoglycemia (NG) after MCAO and 4 h of reperfusion, as assessed by brain extravasation of the Evans blue dye and microtubule‐associated protein 2 immunostaining. Electron microscopy further revealed that HG mice had more endothelial vesicles in the striatal neurovascular unit than NG mice, which may account for their deterioration of BBB impairment. In contrast with enhanced endothelial transcytosis, paracellular tight junction ultrastructure was not disrupted after this mild ischemia/reperfusion insult or altered upon HG. Consistent with the observed increase of endothelial vesicles, transcytosis‐related proteins caveolin‐1, clathrin, and hypoxia‐inducible factor (HIF)‐1α were upregulated by HG after MCAO and reperfusion. Conclusion Our study provides solid structural evidence to understand the role of endothelial transcytosis in HG‐elicited BBB hyperpermeability. Enhanced transcytosis occurs prior to the physical breakdown of BMECs and is a promising therapeutic target to preserve BBB integrity.


| INTRODUC TI ON
Hyperglycemia (HG) is a highly prevalent comorbid condition in acute ischemic stroke patients and can be a chronic condition persisting for days. 1 HG is associated with larger infarct size, more severe brain edema, poorer clinical outcome, and a higher risk of mortality and hemorrhagic transformation (HT). The mechanisms underlying increased occurrence of post-stroke HT under HG condition are not fully understood, but exacerbation of blood-brain barrier (BBB) damage may play a crucial role. [2][3][4][5][6][7] High blood glucose leads to endothelial cell dysfunction and increased intracellular reactive oxygen species (ROS), which are involved in a plethora of pathophysiological changes such as inhibition of nitric oxide (NO) synthesis, vascular inflammation, insulin resistance, neovascularization, leukocyte adhesion, and protein and polymer glycosylation. [8][9][10][11][12] Intriguingly, whether the efficacy of drugs targeting endothelial cell dysfunction is a direct result of improving endothelial functions is disputable under different disease conditions and at different vascular sites. 8,[13][14][15][16][17] A thorough understanding of the regulatory mechanisms of endothelial functions under HG condition is warranted to develop effective therapeutic approaches to ameliorate BBB injury.
Brain microvascular endothelial cells (BMECs) are major contributors to the permeability barrier in the neurovascular unit (NVU).
BMECs regulate substance exchange at the BBB interface and restrict BBB permeability by forming paracellular barriers that permit limited diffusion of blood-derived solutes and allowing only sparse endocytic vesicles for transcytosis. 18,19 Previous studies have shown that enhanced transcellular transport may be the initial response of CNS endothelial cells during stroke. 20 Uptake and transport of Alexa594-labeled albumin into the brain parenchyma by vascular endothelial cells occur as early as 6 h after stroke, accompanied by an increase in the number of endocytic vesicles. 20 That study indicates that the increase in BBB permeability at the early stage after stroke is not due to the degradation of tight junctions, but instead is related to increased endothelial transcytosis. In another subsequent study, ultrastructural analysis also favored the number of endothelial vesicles as the best temporal and spatial indicator of BBB destruction after cerebral ischemia. 21 A limitation of that study, however, is that it did not investigate the type of the transcytotic vesicles or the underlying mechanisms.
Endocytosis is a cellular process in which extracellular materials are transported into the cell through the deformation and internalization of cell membrane. Depending on the type of materials being taken up by the cell, endocytosis can be classified into phagocytosis and pinocytosis. On the basis of the size of the molecules and receptor involved in the process, pinocytosis is further divided into macropinocytosis (0.2-10 µm) and micropinocytosis (<0.2 µm), 22 and the micropinocytosis is conveniently classified by the nature of the coat proteins associated with the endocytic process into clathrin-mediated endocytosis (~150 nm), caveolin-mediated endocytosis (~100 nm), and clathrin-and caveolin-independent pinocytosis. [23][24][25] Among them, caveolin-mediated pinocytosis is a common example of micropinocytosis that is formed in the epithelium of the blood vessels. Although enhanced endothelial endocytosis is known to play a critical role in early BBB injury after stroke, 20,26,27 the type and cause of increased endocytic vesicles remain unknown. To date, the role of endocytosis in hyperglycemia-elicited vascular endothelial cell dysfunction is poorly understood, nor are the ultrastructure changes at the BBB interface thoroughly characterized in the poststroke brain under HG condition.
In this study, we used two-and three-dimensional reconstruction by electron microscopy to elucidate the morphological and ultrastructural changes at the BBB interface after mild ischemia/reperfusion (I/R) brain injury, aiming to provide reliable structural evidence to understand the potential mechanisms underlying HG-exacerbated HT.

| Animals and experimental models
Healthy male ICR mice (8-12 weeks old, 20-22 g) were obtained from the Experimental Animal Center of Nantong University. Mice were housed with a 12/12-hour light/dark cycle and bred with ad libitum access to food and water. All the processes of animal experiments were approved by the Institutional Animal Care and Use Committee of Nantong University.
Mice were randomly assigned to three groups: sham control group, middle cerebral artery occlusion (MCAO) group with normoglycemia (MCAO-NG), and MCAO group with hyperglycemia (MCAO-HG). Mice in the MCAO-HG group received intraperitoneal injection of 50% glucose (0.06 mL/10 g) 15 min before brain ischemia to induce acute hyperglycemia. 28,29 Mice were anesthetized by isoflurane, and a skin incision was made to expose the right common carotid artery (CCA), external carotid artery (ECA), and internal carotid artery (ICA). A 6-0 monofilament with silicon-coated tip (6023PK5Re, Doccol Corporation) was introduced from the ECA to the ICA until it reached the beginning of the MCA.

| Assessment of BBB permeability
After 3 h of post-MCAO reperfusion, 2% Evans Blue (0.1 mL/10 g) was injected through tail vein. One hour later, mice were anesthetized with intraperitoneal injection of 2.5% avertin (2,2,2-tribromoethanol, Sigma-Aldrich, 400 mg/kg body weight) and transcardially perfused with 0.9% NaCl and 4% paraformaldehyde. Brains were removed, and coronal sections of the brain were cut. After photographs of the sliced brains were captured, brain tissues were fixed in paraformaldehyde again and cryoprotected in 30% sucrose in PBS.
Frozen serial coronal brain sections (25 μm thick) were prepared, mounted, and imaged by Leica TSC SP8 confocal microscope (Leica, Germany). The severity of BBB leakage was evaluated by the fluorescence intensity of Evans blue dye in the brain, as measured by the ImageJ software (version 1.43b, NIH, USA).

| Immunofluorescence staining
Immunofluorescence staining was performed on 25-μm-thick freefloating brain sections as previously described. 30 Briefly, brain sections were permeabilized, blocked, and incubated with rabbit anti-microtubule-associated protein 2 (MAP2, Cell Signaling Technology, USA) overnight at 4°C. Sections were then washed by PBS and incubated with fluorochrome-conjugated secondary antibodies. Images were acquired by Nikon Eclipse Ni-E microscope (Nikon, Japan), and infarct areas were measured on six equally spaced coronal sections (25 μm, take 6 slices for every 10 slices) within the MCA territory per brain by ImageJ software (version 1.43b, NIH, USA).

| Assessment of neurological deficits
Double-blind standard was conducted throughout the test.
Neurological deficits were assessed 4 h after reperfusion according to the Clark's scoring system, 31 including a general neurological scale (0-28) and a focal neurological scale (0-28). Six areas were assessed for the general score, and seven areas were assessed for the focal score to evaluate the severity of neurological deficits after ischemic stroke.

| Transmission electron microscopy (TEM)
According to the results of MAP2 immunostaining, tissue blocks

| Quantitative analysis of TEM images
TEM micrographs of brain capillaries were collected from the ipsilateral (ischemic) striatum and quantitatively analyzed using ImageJ software (version 1.43b, NIH, USA). Lumen circularity was calculated using the circularity function in ImageJ software (version 1.43b, NIH, USA), whereby a value of 1 indicates a perfect circle. 21,32 The length of a tight junction was measured from its starting point to endpoint.
Endothelial vesicles were counted and expressed as the number of vesicles per mm 2 of the total endothelial cell area. To visualize the different sizes of endothelial vesicles, frequency distribution histograms on vesicle diameters were plotted.
Resin blocks were carefully trimmed using a Leica EM trimming knife until the surface of black tissue in the block was visible. To select the area of interest, we used a scanning electron microscope (Teneo VS, Thermo Fisher Scientific, USA) with one ultramicrotome in its specimen chamber, which allowed us to trim the resin blocks and acquire the EM images of the sample simultaneously. After

| Immunoblotting
Protein was extracted from the striatum and subjected to Western blotting analysis as previously described. 33 The following primary

| Statistical analysis
Data are expressed as mean ± standard error (mean ± SEM), and statistical analysis was performed by the SPSS 20.0 software. Paired t-test was used for comparison of blood glucose levels before and after glucose injection. Differences in means among multiple groups were analyzed by one-way ANOVA, followed by the Fisher's LSD test. Kruskal-wallis test was used to analyze data that do not conform to normal distribution. A P value less than 0.05 was considered statistically significant.

| Hyperglycemia aggravates BBB damage and enlarges brain infarct after mild ischemic stroke
To assess BBB integrity after mild I/R (20 min / 4 h) injury, we monitored the extravasation of Evans blue (EB) dye into the brain parenchyma after injection into the mouse tail vein. Similar to previous reports, [34][35][36] we observed more severe brain leakage of EB after ischemic stroke in the ipsilateral brain tissues of HG mice, compared to NG controls ( Figure 1A,B). To further quantitatively evaluate the extent of EB leakage, fluorescence intensity was used to measure the EB absorbance at 620 nm. Both NG and HG mice had dramatically elevated EB fluorescence intensity at 4 h of reperfusion after 20-min ischemia ( Figure 1D), whereby EB fluorescence intensity was approximately 1.5 times higher in the HG group than in the NG group ( Figure 1D).
The deterioration of BBB dysfunction was accompanied by enlarged brain infarct in the HG mice after MCAO, as assessed by immunostaining of MAP2, a protein predominantly expressed in the dendrites. After 4 h of reperfusion, brain infarct (MAP2-negative area) had already formed even in the NG group ( Figure 1C). However, brain injury remained mild at this time point and the infarct was To minimize individual differences among mice, we monitored the blood glucose of each mouse, and only selected those mice with blood glucose higher than 20 mmol/L for the MCAO model ( Figure 1F). MCAO/reperfusion induced prominent neurological deficits (on both the general and focal scales) in NG and HG mice, compared to the sham control group ( Figure 1G). HG mice demonstrated more severe focal deficits than NG mice, whereas no significant difference was found between HG and NG mice in general deficits after mild stroke ( Figure 1G). Together, these results suggested that hyperglycemia increased BBB permeability, and exacerbated brain injury and neurological deficits after mild ischemic stroke.

| Hyperglycemia enhances transcellular but not paracellular pathway in BMECs after mild I/R brain injury
To elucidate how hyperglycemia influenced BBB permeability and integrity, we examined the ultrastructure of the NVU by TEM imaging, 37 focusing on microvessels with a dimeter less than 8 μm. With the guidance of MAP2 immunostaining in adjacent brain sections, we were able to locate the infarct core region in the ipsilateral striatum, which exhibited the most prominent EB leakage.
In the sham control group, we observed intact and homogeneous NVU ultrastructure typical for the homeostatic brain, 19,38,39 characterized by the homogenous intercellular tight junction structures between BMECs, few endothelial vesicles inside BMECs, and surrounding pericytes and astrocyte end-feet with normal morphology ( Figure 2A

| Three-dimensional visualization of endothelial vesicles in the neurovascular unit after mild ischemic stroke
Serial block-free scanning electron microscopy (SBF-SEM) allows three-dimensional visualization of large sample volumes at ~50 nm resolution. [40][41][42] Consistent with the aforementioned two-dimensional data, compression of microvessels and swelling of surrounding astrocytes were observed in the NVU in the ipsilateral striatum after MCAO under HG condition ( Figure 4A). Furthermore, we detected a typical clathrin-mediated membrane invagination in the ipsilateral NVU, which was coated by lattice-like densities along the z-axis ( Figure 4A). Non-coated vesicles of different sizes were also observed in the endothelial cytoplasm ( Figure 4A).

| Hyperglycemia enhances CAV1-and clathrinmediated endocytosis in brain endothelial cells after mild I/R injury
Mammalian cells utilize multiple endocytic pathways at microscale, such as clathrin-mediated endocytosis, caveolae-mediated endocytosis, and other clathrin-and caveolae-independent pathways. Recent studies suggest that clathrin can interact with HIF-1 and promote angiogenesis under hypoxic conditions, for example, during tumorigenesis. 43 Furthermore, transcription of CAV1 in response to hypoxic conditions could be directly regulated by HIF-1. 44 Based on this prior knowledge, we hypothesized that HG-induced increase of endothelial vesicles are related to caveolin-or clathrin-mediated endocytosis, considering the diameter of these vesicles. To this end, we examined the expression levels of HIF-1α, CAV1 and clathrin after MCAO and 4 h of reperfusion in HG and NG mice. HIF-1α was significantly upregulated in the ipsilateral striatum after I/R injury compared to sham controls, and was further upregulated in the HG group compared to NG mice ( Figure 5A,B). Similarly, the expression levels of CAV1 and clathrin were significantly elevated in the ipsilateral striatum of the HG group compared to sham group ( Figure 5C,D). In contrast with the enhanced transcytosis signaling, the tight junction protein occluding showed no significant changes among all groups ( Figure 5A,E), indicating comparable paracellular permeability of the BBB at this injury stage. Together, these results suggest that the increased endothelial vesicles observed in the HG brain after ischemic stroke resulted from caveolin-and clathrin-mediated endocytosis pathways.

| D ISCUSS I ON
CNS endothelial cells are distinguished from peripheral vascular endothelial cells by two characteristics: 1) A unique tightly connected structure between endothelial cells that prevents watersoluble molecules from entering the brain parenchyma from the circulatory system; and 2) extremely low levels of intracellular vesicles that is often thought to limit transcellular transport of substances across the endothelium. 45  Blood-brain barrier disruption is the key pathophysiological basis for HT after stroke, 7,17 in the present study, two-and three-dimensional electron microscopy revealed that hyperglycemia increased intracellular vesicles and enhanced endocytosis pathways in BMECs, contributing to increased BBB permeability after mild cerebral I/R injury, and concurrently, the risk of HT becomes more significant. The increase of endothelial vesicle numbers was the most prominent in the NVU in the post-MCAO striatum, which harbored the most severe BBB breakdown. No alteration of tight junction structure was detected by electron microscopy after MCAO in either NG or HG group. Notably, we also observed the compression of microvessels and swelling of surrounding astrocytes in the ipsilateral NVU under HG condition. I/R brain injury with the comorbid condition of high blood glucose may lead to a vicious cycle of ischemia-edema and exacerbate ischemia-edema in local tissues. A further examination on the related endocytosis pathways revealed the involvement of HIF-1α. Therefore, brain I/R under HG condition may trigger tissue edema, which further aggravates local hypoxia and upregulates HIF-1α, subsequently increasing endocytosis. However, the specific mechanisms warrant future explorations.
We also confirmed that the increased vesicles may result from both caveolin-and clathrin-mediated endocytosis pathways. Diabetes elevates the expression of HIF-1α and vascular endothelial growth factor (VEGF) in the ischemic cerebral microvasculature, 38,50,51 which is accompanied by the disruption of the BBB, increased infarct volume, severe edema formation, and exacerbated neurological deficits. Specific inhibition of endothelial HIF-1α could partially reverse the damaging effects of diabetes on cerebrovascular injury, suggesting that HIF-1α activation is one important mechanism underlying increased BBB permeability. 38 Our results support the idea that high blood glucose further upregulates HIF-1α expression to exacerbate BBB impairment F I G U R E 4 Three-dimensional reconstruction of the neurovascular unit ultrastructure in the post-stroke striatum. (A) A representative image slice of the NVU in the hyperglycemia group after MCAO, taken by FIB-SEM. Red arrows in the right panels indicate a membrane invagination at different z-axis levels (0, 72, 96, 120 nm). Scale bars: 1 μm (left panel) or 250 nm (middle and right panels). (B-D) Endothelial vesicles were reconstructed and segmented from a 5-μm-thick sample block in each group. Purple denotes the population of vesicles inside endothelial cells of microvessels with a diameter less than 8 μm. Scale bar: 1 μm [Colour figure can be viewed at wileyonlinelibrary.com] after ischemia, possibly by promoting clathrin-and caveolin-mediated endocytosis and transcellular transport in BMECs. However, more evidences of relationship between HIF-1α and CAV1 or clathrin need to be further studied in the future.
Endothelial cells rapidly respond at the early stage of ischemic stroke by altering their transcellular ratio, and peripheral cells maintain BBB stability by secreting inhibitory signals that reduce the number of vesicles and the rate of transcellular transport in BMECs. [52][53][54][55] Post-stroke pericyte and astrocyte injury may unlock this important inhibitory mechanism, 56 which is necessary to maintain low transcytosis within the endothelium. However, whether or not increased transcellular transport provides a beneficial signal to the NVU during the initial stages of stroke injury progression remains to be explored experimentally.
In conclusion, our study not only identified HG-enhanced endothelial endocytosis after ischemic stroke, but also clarified that such endocytosis was likely mediated by clathrin and caveolin, presumably also involving regulation by HIF-1α. The ultrastructural evidence revealed the role of endothelial transcytosis in HG-exacerbated BBB impairment ahead of physical breakdown of junctional structures between BMECs, which may represent a promising therapeutic target to preserve BBB integrity in stroke patients with HG comorbidities.

ACK N OWLED G M ENTS
We are grateful to Dr. Jiansheng Guo (Center of Cryo-electron microscopy, Zhejiang University) for data collection from Teneo and data processing using Amira software. This work was sup-

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.

AUTH O R CO NTR I B UTI O N S
X.L. and Y.Z. designed the project and supervised the project. X.X. initiated the project. X.X., L.Z., and J.L. performed the EM experiments and analyzed the data. X.X., K.X., and J.W. performed the animal model and neurological score. X.L., Y.Z., J.G., G.W., and X.X.
interpreted the results and commented on the manuscript. X.L. and X.X. wrote the manuscript.

DATA AVA I L A B I L I T Y
All original data and materials are available upon request.