All animals were housed in the controlled animal facilities with a 12 h light/dark cycle at Yonsei University. Mice were housed with 1–5 mice in the cage with free access to food and water. Animal experiment was followed in the Guide to the Care and Use of Laboratory Animals approved by the Association for Assessment and Accreditation of Laboratory Animal Care and the National Institutes of Health guideline. Ethical Committee of the Yonsei University approved this study. Adult male C57BL/6 mice (Orient, Seongnam, GyeongGi-Do, Korea; aged 8–12 weeks) were used in this experiment. The group was divided into three groups; normal control group (control), ischemia/reperfusion (I/R) group, and I/R+si-ASK1 group.

Isoflurane vaporizer was used for anesthesia of mice with 2.0% isoflurane in 30% oxygen and 70% nitrous oxide. Mice were placed on homeothermic blanket and maintained rectal temperature at 37 ± 0.5°C. The MCA was blocked by inserting 6-0 nylon suture into the left external carotid artery (ECA) to induce transient focal cerebral ischemia (tFCI) for 60 min. The blood flow was restored by removing the nylon suture. Mice were housed the cages and monitored during experiments. At 24 h after reperfusion, mice were anesthetized by intraperitoneal injection of Zoletil mixture (30 mg/kg; VirvacLaboratories, Carros, France), and performed cardiac perfusion with normal saline.

Mouse brain endothelial cells (bEnd.3, ATCC, Manassas, VA, USA) were cultured with Dulbecco’s Modified Eagle’s Medium high glucose (DMEM, Hyclone^TM, GE Healthcare Life Sciences, Logan, UT, USA) adding fetal bovine serum (10%; FBS, GE Healthcare Life Sciences) and penicillin-streptomycin solution (1%; Thermo Scientific, Waltham, MA, USA). The endothelial cells were incubated in the CO₂ chamber.

Murine neuronal cell line, Neuro-2A, was cultured in DMEM high glucose cultured media, supplemented with 10% FBS (GE Healthcare Life Sciences) and 1% penicillin-streptomycin solution (Thermo Scientific). The neuronal cells were incubated at 37°C in a humid atmosphere in the presence of 5% CO₂. To examine effects of endothelial cell-released factor on neuronal cells, neuronal cells were incubated with EC-CM which were collected after H/R injury.

Before OGD, the culture medium was discarded, washed with phosphate buffer saline (PBS) and changed with deoxygenated glucose-free balanced salt solution (BSS) under the anaerobic chamber (Forma Scientific, Inc., Marietta, GA, USA). Endothelial cells were exposed to OGD condition for 6 h. After hypoxia, BSS solution was discarded and changed into DMEM cultured media and cells were transferred to a CO₂ incubator for 24 h. After reperfusion, cells and EC-CM were collected and stored at -80°C. NQDI-1 (600 nM, Tocris Bioscience, Bristol, UK), an inhibitor of ASK1, was treated in cultured media from 1 h before hypoxia to 6 h during hypoxia.

ASK1 gene was silenced by using siRNA (Ambion, Austin, TX, USA, sense, GCUCGUAAUUUAUACACUGtt; antisense, CAGUGUAUAAAUUACGAGCtt; 5 μM). A 100 μl solution of siPORTNeoFX (Ambion) and siRNA for ASK1 was injected into the lateral ventricle of mice (mediolateral 1.0 mm; anteroposterior 0.2 mm; dorsoventral 3.1 mm) with an osmotic pump (Alzet, Cupertino, CA, USA) for 3 days before ischemic injury.

At 24 h after reperfusion, brain of mice was isolated after perfusion with saline. Each brain was fixed with 3.7% formaldehyde and stored at -80°C. Samples were sectioned coronally by a cryotome. Thickness of sections was 20 μm. Sections were stained with cresyl violet (Sigma-Aldrich, St. Louis, MO, USA) for 3 min and washed with distilled water for three times. Samples were mounted by using mounting medium (Vector Laboratories, Inc., Burlingame, CA, USA) and observed under microscope (Olympus, Tokyo, Japan).

Frozen brain sections were sliced by a microtome with 20 μm thickness. Sections were onto coating slide glass (MUTO Pure Chemicals Co., Ltd, Tokyo, Japan) and dried at room temperature. Frozen mouse brain coronal sections were washed with PBS for one times, and dipped into -20°C ethyl alcohol for 20 min, followed by 0.3% Triton X-100 diluted in PBS for 1 h. After washing with PBS, samples were blocked in 5% bovine serum albumin (BSA) diluted in PBS for 1 h and incubated with rabbit anti-NeuN antibody (Abcam, Cambridge, UK) at 4°C overnight. After washing with PBS, terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay for measurement for DNA fragmentation was performed followed by manufacture’s instruction (RocheDiagnostics, Indianapolis, IN, USA). Hoechst 33342 (ThermoFisher Scientific) or Propidium iodide (PI; Sigma-Aldrich) was used for counterstaining. After washing with PBS, stained brain were mounted with Vectashield (Vector Laboratories, Inc.). After then, sections were observed by using an LSM 700 confocal laser scanning microscope (Carl Zeiss, Thornwood, NY, USA).

The tissues were lysate with a RIPA buffer (Biosesang, Inc., Seongnam, South Korea) with Halt^TM protease & Phosphatase inhibitor cocktail (1:100, Thermo Scientific). After centrifuged at 13000 rpm at 4°C for 20 min, the pellet was collected and protein concentration was measured by using Pierce^® BCA protein assay kit (Thermo Scientific). Sample buffer was added and boiled at 95°C for 5 min. Samples were transferred to polyvinylidene difluoride membranes (PVDF, Merck Millipore, Bedford, MA, USA). Membranes were blocked with 5% BSA and incubated in the rabbit anti-phosphatidylinositol 3-kinase (Santa Cruz Biotechnology, Santa Cruz, CA, USA), rabbit anti-pAkt (1:1000, Cell signaling, Danvers, MA, USA), rabbit anti-Akt (1:1000, Cell signaling), rabbit anti-nuclear factor erythroid 2 [NF-E2]-related factor 2 (1:1000, Santa Cruz Biotechnology), rabbit anti- HO-1 (1:1000, Abcam, Cambridge, UK), rabbit anti-Bcl-2 (1:1000, Abcam), mouse anti-Bax (1:1000, Merck Millipore), goat anti-cyclooxygenase-2 (1:500, Santa Cruz Biotechnology), rabbit anti-cleaved caspase-3 (1:1000, Santa Cruz Biotechnology), and rabbit anti-caspase-3 (1:1000, Merck Millipore) which is diluted in 2% BSA. Horseradish peroxidase-conjugated anti-rabbit, anti-goat, or anti-mouse IgG reagents (1:5000) were used as secondary antibodies (Jackson ImmunoRearch Laboratories, West Grove, PA, USA). The β-actin (1:5000, Santa Cruz Biotechnology) was used as an internal control. The bands were visualized with enhanced chemiluminescence reagents (West-Q pico Dura ECL solutions; GenDEPOT, Barker, TX, USA) under the LAS 4000 program (GE Healthcare, Pittsburgh, PA, USA).

The endothelial cells onto the coated slide were fixed with 4% paraformaldehyde (PFA) for overnight at 4°C. After washing with PBS for three times, cells were blocked with 5% BSA for 1 h at room temperature. After washing with PBS, cells were reacted by the rabbit anti-PI3K (Santa Cruz Biotechnology), rabbit anti-pAkt (1:1000, Cell signaling), rabbit anti-Nrf-2 (1:1000, Santa Cruz Biotechnology), rabbit anti-HO-1 (1:1000, Abcam) and goat anti-Cox-2 (1:500, Santa Cruz Biotechnology) respectively at 4°C for 1 day, followed by FITC or Rhodamine-conjugated secondary antibody (1:1000, Jackson ImmunoResearch) at room temperature for 2 h. After washing with PBS, stained cells were mounted with Vectashield with DAPI (Vector Lab). Slides were observed under LSM 710 confocal laser scanning microscope (Carl Zeiss).

WST assay (EZ-CYTOX, Daeillab Service, Seoul, South Korea) was used for detecting cell viability of Neuro-2A cells. One day before performing the assay, the Neuro-2A cells were seeded in 96-well plates at a density 1 × 10⁴ cells per well with 100 μl of DMEM culture medium. After incubation with EC-CM for 24 h, WST (10 μl) in cultured media (100 μl) was added in each well. After 1 h 30 min reaction time at 37°C, the resultants were red by using VERSASA max microplate reader (Molecular Devices, Sunnyvale, CA, USA) at a wavelength 450 nm.

For the assessment of activity of MMP-9 in tissue, endothelial cell, and endothelia cell media, MMP-9 Biotrak Activity Assay system (Amersham Biosciences, Piscataway, NJ, USA) was used according to the manufacturer’s instructions. Tissue lysate, cell lysate, or cultured cell media was added into MMP-9 coated wells and incubated overnight for immunoreactivity at 4°C. To measure active MMP-9, detection enzyme was added and incubated at 37°C. The resultants were read by ELISA reader at 450 nm.

For detection of apoptosis in neuronal cells, Hoechst 33342/PI double staining was performed. Neuronal cells were incubated with Hoechst 33342 (10 μg/mL; ThermoFisher scientific) was or PI (Sigma-Aldrich) at 37°C for 10 min in the dark. After staining, cells were placed on a microscope slides and observed by LSM710 confocal microscope (Carl Zeiss). Hoechst 33342 is used for live cell staining of DNA and nuclei and PI is used for dead cell staining.

Neuro-2A cells were seeded in 100 mm dishes at a density 4 × 10⁶ cells and incubated with EC-CM for 24 h. Cells were washed with DPBS and collected. And then, cell suspensions were filtered through a cell strainer with a 40 μm nylon mesh. PI solution (10 μl) was added and incubated at room temperature for 10 min in the dark. Fluorescent cells (PI-positive) were determined immediately by using flow cytometer (LSR II, BD bioscience, San Hose, CA, USA). Flow cytometry data was analyzed using FlowJo version 10.

Data are expressed as the mean ± standard error of the mean (SEM). Statistical comparisons among the groups were assessed with a one way ANOVA followed by a Tukey post hoc test (Prism version 6.0 software). Statistical significance between groups was considered to be present at *p < 0.05, **p < 0.01, ***p < 0.001.

We performed cresyl violet staining to assess the morphological alterations that occur in cells after ischemic injury (Figure 1A). In the control group, round and healthy cells were noted in the cerebral cortex and striatum, whereas in the I/R group, small and thin cell bodies were observed in the damaged cortex and striatum 24 h after ischemic injury. To examine whether ischemia induces neuronal cell death, we performed immunolabeling and TUNEL assays (Figure 1B). Compared to the control group, fewer neuronal nuclei NeuN-positive cells in the I/R group, were co-localized with TUNEL-positive cells in the striatum. Moreover, in the cortex, many TUNEL-positive cells were merged with NeuN-positive cells in the I/R group. However, TUNEL-positive cells (red) were not detected in the cortex and striatum of the control group. These results indicate that I/R injury promoted neuronal cell death in the lesioned brain areas at 24 h after I/R.

The previous study demonstrated that synthetic siRNA for ASK1 efficiently suppresses ASK1 and subsequently pASK1 after I/R (Kim et al., 2011). We used this method. To suppress the ASK1 level, siRNA (sense, GCUCGUAAUUUAUACACUGtt; antisense, CAGUGUAUAAAUUACGAGCtt) was injected through the intracerebroventricular route (Supplementary Figure S1). To confirm that ASK1 could be silenced efficiently by siRNA, we performed immunohistochemistry after tFCI. Our results showed that the increased ASK1 expression after ischemia was well-silenced by siRNA (Figure 2A). Next, to determine whether I/R and ASK1 could modulate MMP-9, we performed an MMP-9 activity assay at 24 h after I/R (Figure 2B). Our results showed that the level of active MMP-9 was significantly increased in the I/R group compared to the level in the control group. After silencing ASK1 with siRNA, the levels of active MMP-9 were efficiently attenuated in the brain tissue despite the I/R injury. Thus, ASK1 may contribute to MMP-9 activation at 24 h after I/R.

The MMP-9 that is secreted from endothelial cells plays pivotal roles in BBB disruption, endothelial cell morphogenesis, and capillary formation (Dong et al., 2009; Dao Thi et al., 2012). Thus, we performed MMP-9 activity assay in cell media and cultured endothelial cells after 6 h (hypoxia)/24 h (reperfusion; Figures 3A,B). Our results showed that in EC-CM, the level of MMP-9 in the H/R group, was obviously increased compared to that in the control group. In the group treated with the ASK1 inhibitor NQDI-1, MMP-9 activity level was lower than that in the EC-CM of the H/R group (Figure 3A). However, in endothelial cells, the levels of active MMP-9 were not significantly different among the groups (Figure 3B).

To elucidate the mechanisms underlying the protective effects of ASK1 inhibition against MMP-9 in endothelial cells during H/R, we performed Western blot analysis (Figure 4) and immunocytochemistry (Figure 5). A previous study demonstrated that exposure to oxidative stress induces PI3K/Akt signaling, which activates pro-survival reactions after stimuli (Wang et al., 2007). Based on this previous study, we assessed the PI3K protein levels after oxidative stress in endothelial cells. However, the PI3K protein level of these cells was not altered after exposure to H/R. Interestingly, the PI3K level in NQDI-1-treated cells was significantly increased at 24 h after H/R (Figures 4A,B). Using Western blots, we also measured the level of Akt protein, which is downstream of PI3K. Although the quantitative graph showed that the level of phosphorylated-Akt (pAkt) in the H/R group was upregulated, compared to the level in the control group, after treatment with NQDI-1, the pAkt level increased further despite the H/R injury (Figures 4A,C). Next, we hypothesized that PI3K/Akt as upstream signaling molecules signals would modulate the Nrf-2/HO-1 pathway during hypoxia injury (Martin et al., 2004; Wu et al., 2013; Meng et al., 2014; Xu X.H. et al., 2015). The results of our Western blot analysis revealed that the, Nrf-2 protein level was also significantly upregulated, compared to the level in the H/R group after blocking ASK1 with NQDI-1 (Figures 4A,D). The HO-1 protein levels in the H/R group were not increased compared to the levels in the control group; however, blocking ASK1 increased the HO-1 levels regardless of the H/R injury (Figures 4A,E). It has been recognized that Cox-2 is an important mediator of ischemic injury development in the brain (Nogawa et al., 1997). Our data indicated that following H/R injury, the Cox-2 protein levels were reduced after treatment with NQDI-1 (Figures 4A,F). Next, we performed immunocytochemistry. Our data demonstrated that PI3K-positive cells (red) showed markedly decreased immunoreactivity in the H/R group. On the other hand, PI3K expression in the H/R+NQDI-1 group was enhanced, compared to the expression in the control and H/R groups (Figure 5A). Additionally, pAkt (red) was highly expressed in a cytosol-like pattern in the H/R+NQDI-1 group, compared to in the H/R group (Figure 5B). We also observed lower Nrf-2 immunoreactivity in the H/R group. After blocking the ASK1 signal, the expression of Nrf-2 in the H/R+NQDI-1 group was increased compared to that in the H/R group (Figure 5C). No HO-1 expression was observed in the H/R group, but intense HO-1 expression was noted in the H/R+NQDI-1 group (Figure 5D). In contrast, Cox-2 expression was upregulated in H/R-injured endothelial cells, but was downregulated after ASK1 inhibition despite the H/R injury (Figure 5E). Collectively, our data suggest that ASK1 inhibition reduces Cox-2 signals but increases PI3K/Akt/Nrf-2/HO-1 signaling after H/R injury in endothelial cells.

To elucidate whether ASK1 and MMP-9 modulate neuronal cell fate, we treated neuronal cells with EC-CM and performed Western blot analyses and cell viability assays. We measured the protein levels of B-cell lymphoma 2 (Bcl-2) an anti-apoptotic marker, and Bcl-2-associated X (Bax), a pro-apoptotic marker, and caspase-3, an apoptotic molecule, in the neuronal cell culture (Figure 6). The quantitative analysis demonstrated that the Bcl-2 level was decreased in neuronal cells incubated with H/R-injured EC-CM. However, neuronal cells incubated with H/R+NQDI-1-treated EC-CM showed efficiently increased Bcl-2 levels (Figure 6B). In contrast to the results for Bcl-2, the Bax expression level was obviously increased in neuronal cells exposed to H/R-injured EC-CM. After attenuating of ASK1 in the EC-CM, the Bax protein level was downregulated in neuronal cells (Figure 6C). Cleaved caspase-3 level in neuronal cells was elevated in neuronal cells incubated with H/R-injured EC-CM, this level was reversed after H/R+NQDI-1-treated EC-CM (Figure 6E). To investigate the neuronal cell viability, viability assays were conducted at 24 h after EC-CM incubation (Figure 6F). After normalizing the cell viability with the control values, the cell viability (%) of neuronal cells incubated with the H/R group was 71.32%; however, the cell viability increased to 83.42% after incubation with ASK1 inhibitor-treated EC-CM. These results indicated that apoptotic factors were downregulated and anti-apoptotic factor was upregulated in neuronal cells after inhibition of ASK1.

To confirm that the apoptosis-related molecules regulated by ASK1 lead to neuronal cell death, we performed Hoechst/PI dual staining and FACS analysis (Figure 7). Confocal images showed that PI-positive cells (red) were found easily in neuronal cells which incubated with EC-CM after H/R, however, neuronal cells with ASK1 inhibited EC-CM showed lower PI-positive cells and in the control group, we cannot observe PI-positive cells (Figure 7A). Single cell suspensions were labeled with PI and analyzed by flow cytometry after incubation with EC-CM. A primary gating was based on forward and side light scatter respectively and cell debris were excluded (Figure 7B). Sorted cells were measured by PI fluorescence laser. The dot plot showed that neuronal cells in the H/R group showed a much higher amount of PI-positive cells (cell percentage; 32.7%) compared to those in the H/R+NQDI-1 group (cell percentage; 23.9%; Figure 7C). In addition, merged histogram exhibited increased PI-positive cell counts in the H/R group, compared to those in the H/R+NQDI-1 group (Figure 7D). In the ischemic brain tissue, after silencing ASK1, a marked reduction in apoptotic cell death was detected by using a TUNEL assay in vivo (Supplementary Figure S2). These data suggest that neuronal cell death was ameliorated by inhibiting ASK1.