Sprague-Dawley rats were purchased from SLAC Company (Shanghai, China). They were housed 2–3 per cage under constant temperature (23 ± 1°C) and light-controlled vivarium (12-h light/12-h dark cycle). Rats housed in the same cage underwent the same manipulations. Food and water were available ad libitum. All animal procedures were approved by the University Committee on Animal Care of Soochow University and performed according to the NIH Guide for the Care and Use of Laboratory Animals. All efforts were made to minimize animal suffering and to reduce the number of animals. Rats weighing 270–290 g were subjected to 2-h MCAO using the intraluminal suture occlusion model, as we described previously (Liu et al., 2017). Prior to reperfusion, all rats included in this study displayed typical neurologic deficit of MCAO, circling to the left (non-ischemic side). Detailed animal usage for each experiment was listed in figure legends. Successful MCAO was further confirmed by 2,3,5-triphenyltetrazolium chloride (TTC) staining.

To establish a causal role of HIF-1α in early ischemic BBB damage, the HIF-1α inhibitor YC-1 (Cayman Chemical Company, Ann Arbor, MI, USA dissolved in a solution of 1% dimethyl sulfoxide, DMSO) or vehicle was administered at 2 mg/kg body weight through femoral vein at 24 h and 30 min prior to the onset of ischemia. To inhibit HIF-1α, researchers have used various dosages of YC-1 ranging from 1 mg/kg to 30 mg/kg body weight (Hsiao et al., 2004; Yeh et al., 2007). We chose the dose based on previous publications (Fischer et al., 2002; Schoch et al., 2002; Yan et al., 2011).

To block interaction of NE with β2-AR, β2-AR antagonist ICI 118551 (30 nmol/μL, 0.5 mL, Sigma) or vehicle was injected into the ventral striatum of the ischemia hemisphere with a Hamilton syringe at a rate of 0.25 μL/min 10 min before MCAO. The coordinates for ventral striatum are taken from the atlas of Paxinos and Watson (1986) (AP, −1.0 mm; ML, −2.0 mm; DV, −7.0 mm).

Immediately after 2-h MCAO, Evan’s blue dye (EB; Sigma, St. Louis, MO, USA, 2% wt/vol in PBS) was intravenously administered (3 mL/kg) via the left femoral vein. All rats were reperfused for 10 min to allow sufficient circulation of EB to the ischemic brain, but minimize the impact of reperfusion on BBB integrity. At the end of reperfusion, the rat was transcardially perfused with ice-cold PBS and then the brain was quickly taken out.

Immediately after 2-h MCAO and 10 min reperfusion, the rat was perfused with ice-cold PBS followed by 4% PFA. The 20-μm-thick cryosection was used for immunostaining analysis for occludin, HIF-1α and VEGF as we described previously (Wang et al., 2017). In brief, non-specific binding sites were blocked by pre-incubating tissue for 1 h at room temperature in PBS containing 0.1% Triton X-100, 1% BSA, and 5% goat serum. Sections were then incubated overnight with anti-HIF-1α (1:200 dilution, Novus), anti-VEGF antibody (1:200 dilution, Abcam), anti-NeuN (1:200 dilution, Millipore), anti-GFAP (1:2000 dilution, Millipore), anti-RECA-1 (1:100 dilution, Abcam) or anti-occludin (1:150 dilution, Invitrogen) primary antibody at 4°C. The latter was followed by incubation with Cy3 conjugated secondary antibody (anti-mouse, 1:800 dilution) or 488-conjugated secondary antibody (anti-rabbit, 1:800 dilution) for 2 h at room temperature. Immunostain was visualized under LSM 700 confocal laser-scanning microscope (Zeiss), and images were taken from the ischemic region and the mirrored region on the NI hemisphere.

Mouse brain microvascular endothelial cells bEND3 (American Type Culture Collection) were grown as a monolayer in DMEM with 10% fetal bovine serum (FBS), 100 U/ml penicillin and 100 μg/mL streptomycin at 37°C in a humidified incubator with 5% CO₂ and 95% room air. The cells were subcultured into 60 mm dishes coated with type I collagen (Nakamuta et al., 2005) and allowed to grow to confluence before exposure to OGD for 2 h. After OGD treatment (3-(4,5-dimethylthiazole-2-yl)-2,5-diphenyl tetrazolium, MTT; Sigma) assay was used to assess the cytotoxicity.

SH-SY5Y cells (neurons) were purchased from American Type Culture and was cultured in DMEM/1640 (1:1; Gibico, Life technologies) containing 10% FBS, 100U/mL penicillin and 100 μg/mL streptomycin, respectively, at 37°C in a humidified incubators with 5% CO₂ and 95% room air. The cells were subcultured into 6-well cell culture cluster and allowed to grow to 80%–90% confluence before exposure to OGD for 2 h.

Primary astrocyte cultures were prepared from 2-day-old neonatal C57BL/6J mice. In brief, dissociated cortical cells were suspended in DMEM/1640 (Gibico, Life Technology) containing 25 mM glucose, 4 mM glutamine, 1 mM sodium pyruvate, and 10% FBS and plated on uncoated 25 cm² flasks at a density of 6 × 10⁵ cells cm⁻². Monolayers of type 1 astrocytes were obtained 12–14 days after plating. Non-astrocytic cells were detached from the flasks by shaking and removed by changing the medium. Astrocytes were dissociated by trypsinization and then reseeded on uncoated T75 flasks. These cells were allowed to grow to 80%–90% confluence before exposure to 2-h OGD treatment.

To mimic acute ischemia-like conditions in vitro, SH-SY5Y cells were exposed to OGD for 2 h as we described previously (Liu J. et al., 2016). In brief, confluent SH-SY5Y cells were subjected to an ischemic injury by transferring cultures to glucose free medium (DMEM/1640 without glucose) pre-equilibrated with 95% N₂ and 5% CO₂. Cells were then incubated in a humidified airtight chamber (MIC-101, Billups-Rothberg Inc., Del Mar, CA, USA) equipped with an air lock and flushed with 5% CO₂/95% N₂ for 15 min. The chamber was then sealed and kept at 37°C for another 105 min. Control cultures were incubated with normal DMEM/1640 medium without FBS for 2 h at 37°C in 5% CO₂/95% air. Immediately after OGD treatment, the conditioned media (CM) and cells were collected separately for further analyses.

Cells grown on six-well plates at 80%–90% confluence were incubated with the corresponding medium containing HIF-1α inhibitor YC-1 (10 μmol/L) at 2 h before OGD or β2-AR antagonist ICI 118551 (1 μmol/L) at 10 min before OGD.

At 24 h after seeding, the media were replaced with conditioned media collected from OGD-treated neurons (OGD-neuron CM) or media from control neurons without OGD treatment (Neuron media). Endothelial cells that were maintained in regular endothelial cell media (Media) served as controls. To investigate whether the β2-AR antagonist ICI 118551 suppressed occludin degradation, endothelial cells exposed to OGD-neuron CM were co-treated with vehicle or β2-AR antagonist ICI118 551 at the concentrations of 1 μmol/L.

SH-SY5Y cells at 60%–70% confluence were transfected with 6 μL HIF-1α siRNA (Santa Cruz, sc-35561) or control siRNA-A (Santa Cruz, sc-37007) which was diluted with the same volume of transfection reagent (Santa Cruz, sc-29528) according to manufacturer’s instruction. Specific silencing was confirmed by western blot.

Tissue were homogenized in MMP lysis buffer (50 mM Tris-HCl PH 7.6, 150 mM NaCl, 5 mM CaCl₂•2H₂O, 0.05% Brij-35, 0.02% NaN₃, 1% Triton X-100) and MMP-2/9 levels in homogenates were assessed by gel gelatin zymography as we described previously (Shu et al., 2015).

Homogenate aliquots (30 μg of total protein) were boiled and then electrophoresed in 10% SDS-PAGE acrylamide gels, transferred onto nitrocellulose membranes (Bio-Rad) and incubated for 1 h in Tris-buffered saline and 0.1% Tween 20 (TBS-T) containing 5% nonfat milk. Membranes were then incubated overnight at 4°C with primary antibodies against occludin (1:300, Invitrogen), HIF-1α (1:300, Novus), VEGF (1:500, Abcam), washed in TBS-T, and then incubated for 2 h at room temperature with corresponding HRP-conjugated anti-rabbit or anti-mouse antibodies (1:3000, Boster). The membranes were developed with the SuperSignal West Pico HRP substrate kit (Pierce) and photographed on a Gel DOC^TM XR⁺ image station (Bio-rad). Protein band intensities were quantitated after normalization to β-actin or total protein stained by Ponceau S.

Total cellular RNA was isolated using Trizol reagents (Invitrogen) according to manufacturer’s protocol as we described previously (Jin et al., 2015). RNA (0.5 μg) was reverse-transcribed (RT) with random primers in a 20 μL final reaction volume using TaqMan^® Reverse Transcription Kits (Applied Biosystems). 0.5 μL RT products were amplified with the 7900HT Fast Real-Time PCR System (Applied Biosystems) in a 10 μL final reaction volume using SYBR^® Green PCR Master Mix (Applied Biosystems) under the following conditions: 2 min at 50°C and 10 min at 95°C, followed by a total of 40 cycles of two temperature cycles (15 s at 95°C and 1 min at 60°C). Primers (Integrated DNA Technologies) for VEGF and glyceraldehydes 3-phosphate dehydrogenase (GAPDH) were designed against known mouse sequences: VEGF forward: 5′-AGAAAGCCCATGAAGT GGTG-3′, reverse: 5′-ACTCCAGGGCTTCATCATTG-3′; β-actin forward: 5′-ACTATCGGCAATGAGCGGTTCC-3′, reverse: 5′-AGCACTGTGTTGGCATAGAGGTC-3′. The fluorescence threshold value (C_t value) was calculated using the SDS Enterprise Database software (Applied Biosystems). The relative value of mRNA expression was calculated by the comparative ΔΔDC_t method described in our previous publication (Liu et al., 2007). In brief, mean C_t values were normalized to the internal control GAPDH and the difference was defined as ΔΔC_t. The difference between the mean ΔC_t values of treated and untreated cells was calculated and defined as ΔΔDC_t. The comparative mRNA expression level was expressed as 2^−ΔΔDC_t.

VEGF contents in cell culture supernatants were assessed using an ELISA kit from Boster (Wuhan, China). Aliquots of 100 μL of cell conditioned media were added to the appropriate microtiter wells provided by the kit. For the standard curve, 100 μL of different concentration VEGF standards and the blank standard were added to the appropriate microtiter wells. The plate was then incubated for 1.5 h at 37°C. After discarding the liquid of each well (no wash), 100 μL of biotinylated anti-human VEGF antibody solution was added into each well except the chromogen blanks and incubated for 1 h at 37°C. After washing three times with the wash buffer, 100 μL avidin-peroxidase complex working solution was added to each well, except the chromogen blanks, and incubated for 30 min at 37°C. Lastly, 100 μL of stabilized chromogen was added to each well, and incubated for 30 min at 37°C in the dark before the addition of 100 μL of stop solution. The absorbance at 450 nm was recorded on a plate reader. A standard curve was generated and the concentrations for unknown samples were obtained from the standard curve.

HIF-1α is a key mediator of the adaptive cellular response to hypoxia and VEGF has been implicated in BBB permeability increase (Schoch et al., 2002; Yan et al., 2011). To determine the roles of HIF-1α and VEGF in BBB damage after 2-h MCAO, western blot was used to compare HIF-1α and VEGF levels in the regions of interest (ROI 1, tissue damage area; ROI 2, BBB damage area; Jin et al., 2012). As shown in Figure 1, 2-h MCAO induced a nearly 1.5-fold increase of HIF-1α protein in ROI 2 (P < 0.05), but not ROI 1 (P > 0.05, Figure 1C). Pretreatment with HIF-1α inhibitor YC-1 at a dose of 2 mg/kg body weight significantly prevented this increase (P < 0.05, Figure 1C). Surprisingly, 2-h MCAO did not change the protein levels of VEGF (Figure 1D), but significantly increased VEGF mRNA expression, and YC-1 significantly inhibited VEGF upregulation (Figure 1E). The results suggested that 2-h MCAO induced VEGF mRNA expression in a HIF-1α-dependent manner.

EB was used as a marker to evaluate BBB integrity (Wang et al., 2017). We next examined the effect of HIF-1α inhibitor YC-1 treatment on 2-h MCAO-induced BBB disruption and occludin degradation (Figure 2A). Representative images of EB dye in wet brain tissue was shown in Figure 2B. Obvious EB leakage was observed in the ipsilateral hemisphere of brain from rats subjected to 2-h MCAO (Figure 2B). HIF-1α inhibitor YC-1 treatment dramatically reduced the area of EB leakage (Figure 2C) and quantification data showed that YC-1 decreased ~78% of EB leakage (Figure 2D). These results indicated that HIF-1α inhibition was effective in reducing BBB damage within the first 2 h after ischemia onset.

Our recent study showed that loss of occludin contributed to BBB damage after 2-h ischemia (Wang et al., 2016). To determine whether ischemia-induced HIF-1α upregulation contributed to the rapid disruption of occludin, IHC and western blot were conducted to detect occludin protein levels in ROI with BBB leakage. Our data showed that pretreatment with YC-1 significantly inhibited ischemia-induced occludin degradation in ROI 2 (EB leakage area, Figure 2F). Western blot was done to confirm the IHC results and found that 2-h MCAO induced a significant reduction in occludin protein levels in ROI 2, but not ROI 1 (Figure 2E). These results indicated that HIF-1α inhibition significantly attenuated occludin degradation within the first 2 h after ischemia onset (Figure 2E).

β2-AR has been shown to regulate HIF-1α expression in cancer cells (Shan et al., 2013). To investigate the effect of blocking β2-AR on ischemia-induced HIF-1α upregulation and BBB disruption, rats received ICI118551 (30 nmol/mL, 0.5 mL; Qu et al., 2008) 10 min before MCAO (Figure 3A). Our results showed that blocking β2-AR significantly inhibited 2-h ischemia-induced HIF-1α upregulation (Figure 3B, P < 0.05 vs. Vehicle), occludin degradation (Figure 3C, P < 0.05 vs. Vehicle) and BBB damage (Figures 3D–F, P < 0.05 vs. Vehicle), suggesting that blocking β2-AR inhibited BBB damage through regulating HIF-1α.

Cell type-specific action for HIF-1α has been defined in astrocytes and neurons in a rat neonatal stroke model (Mu et al., 2003; Vangeison et al., 2008). However, the cellular pattern of HIF-1α expression was not known in the ischemic brain at 2 h after MCAO onset. Double-immunofluorescence labeling was used to determine the cellular localization of HIF-1α, and DAPI staining was conducted to verify the subcellular localization of HIF-1α. As shown in Figure 4, HIF-1α was mainly expressed in neurons (Figure 4B, upper panel), but not in astrocytes (Figure 4B, middle panel) or endothelial cells (Figure 4B, bottom panel) in the ROI 2 at 2 h after MCAO onset. YC-1 significantly decreased the number of HIF-1α-positive neurons, but had no effect on HIF-1α expression in astrocytes or endothelial cells (Figure 4C). These results indicated that the neurons might be major cellular source of HIF-1α early after cerebral ischemia onset.

The cellular distribution and expression of VEGF protein was determined by immunofluorescence with a polyclonal VEGF antibody. VEGF-positive cells were identified to be neurons (colocalization with NeuN but not with GFAP, data not shown).

Hypoxia rapidly increased the HIF-1α as well as the mRNA levels of VEGF (Bernaudin et al., 2002), but VEGF protein increase was much later, which occurred at 4 h or 8 h during the phase of reperfusion (Mu et al., 2003). We applied oxygen-glucose deprivation (OGD) to mimic ischemia in vitro (Zhang Y. et al., 2016). Human dopaminergic neurons (SH-SY5Y cells) were subjected to 2-h OGD. YC-1 or HIF-1α siRNA was applied to inhibit HIF-1α level and non-targeting siRNA was used as a control siRNA (Ctrl siR). ELISA and western blot were recruited to detect VEGF level in cultured media (CM) and cellular extracts (CE), respectively.

Two-hour OGD induced a significant increase of VEGF mRNA expression in neurons (P < 0.05 vs. Ctrl), and pretreatment with YC-1 (Figure 5A, left panel) or HIF-1α siR (Figure 5A, right panel) significantly inhibited this effect (P < 0.05 vs. OGD). However, 2-h OGD had no effect on the protein levels of VEGF in the CE (Figure 5B), but significantly increased its levels in the CM (Figure 5C, P < 0.05 vs. Ctrl). Pretreatment with YC-1 (Figure 5C, left panel) or HIF-1α siR (Figure 5C, right panel) significantly inhibited VEGF increase in the CM (P < 0.05 vs. OGD) suggesting that VEGF secretion was mediated by HIF-1α.

We applied in vitro cultured brain microvascular endothelial monolayer to mimic the BBB as bEND3 cell line has the ability to maintain BBB characteristics over many passages, is easy to grow and can form functional barriers (Camós and Mallolas, 2010). We have shown that OGD induces MMP-2 secretion in neurons, astrocytes and endothelial cells and the secreted MMP-2 disrupts BBB by degrading occludin (Liu et al., 2012), here we explored the effect of YC-1 on MMP-2 secretion in neurons and endothelial cells. After 2-h OGD, MMP-2 levels in CM and CE were detected by gel gelatin zymography. As shown in Figure 6, OGD-induced MMP-2 increase in the conditioned media accompanied by a decrease in the CE, indicating an OGD-induced MMP-2 secretion in neurons. Pretreatment with YC-1 significantly inhibited MMP-2 secretion in neurons (Figures 6A,B, P < 0.05 vs. OGD), but not in endothelial cells (Figures 6C,D). Of note, YC-1 did not reverse OGD-induced occludin degradation in endothelial cells (Figure 6E). These results suggested that HIF-1α mediated OGD-induced MMP-2 secretion in neurons but not in endothelial cells, and further studies are warranted to define the mechanisms underlying MMP-2 secretion in endothelial cells under OGD conditions.

We next examined the effect of blocking β2-AR on HIF-1α expression in neurons and endothelial cells subjected to 2-h OGD. Neurons or endothelial cells were incubated with ICI 118551 for 10 min before exposing to 2-h OGD. As shown in Figure 7, 2-h OGD significantly upregulated HIF-1α in neurons (Figure 7A), but not in endothelial cells (Figure 7B, upper and middle panel) and astrocytes (Figure 7C, upper and middle panel). Of note, 2-h OGD did not induce VEGF secretion in astrocytes (Figure 7C, bottom panel). Although pretreatment with ICI 118551 did not inhibit OGD-induced occludin degradation in endothelial cells (Figure 7B, upper and bottom panel), it significantly inhibited OGD-induced HIF-1α upregulation in neurons (Figure 7A, P < 0.05 vs. OGD). These results suggested that blocking β2-AR inhibited OGD-induced HIF-1α upregulation in neurons.

The above results have demonstrated that OGD induced the secretion of MMP-2 and VEGF into extracellular medium through activating β2-AR-HIF-1α pathway, which mediated occludin degradation in endothelial cells. To further clarify the interaction between neurons and endothelial cells, we exposed endothelial cells to the conditioned media from OGD-treated neurons for 2 h. Western blot analysis showed that the OGD-neuron conditioned media induced occludin degradation in endothelial cells and pretreatment neurons with ICI 118551 inhibited the degradation (Figure 7D).