Mouse bEnd.3 were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Gibco, C11995500BT) supplemented with 10% fetal bovine serum (FBS, Gibco) at 37°C in 5% CO_2. The Claudin-5^112Δ5-mutated bEnd.3 cell line was generated by the clustered regularly interspaced short palindromic repeats/CRISPR-associated protein 9 (CRISPR/Cas9) method using sgRNA (ccaCAACATCGTGACGGCGCAGA; Yang et al., 2021), and the Claudin-5^297Δ4-mutated bEnd.3 cell line was generated by the CRISPR/Cas9 method using sgRNA (cctTGACCGGCGCTCAGTGCACC). The cells were stimulated with hypoxia as previously reported (Yang et al., 2019b, 2020). In brief, the cells were treated with 1% O₂/94% N₂/5% CO₂. Then, the cells were incubated at 37°C in a constant temperature incubator for 2, 4, or 6h. Cell viability was examined by Cell Counting Kit-8 (CCK-8) assays.

Endothelial bEnd.3 cells at a density of 80% were transfected with RFP-GFP-LC3 adenovirus diluted in serum-free DMEM. Then, we changed the medium to DMEM with 10% FBS after 6–8h. After 3days, 2μg/ml puromycin was added to the adenovirus transfection and control wells. After 5days, monoclonal screening of transfected cells was performed.

The trans-endothelial electrical resistance (TEER) and paracellular permeability were measured to reflect the barrier property of the endothelial monolayer as previously described (Yang et al., 2020). In brief, wild-type or Claudin-5^112Δ5-mutated bEnd.3 cells were cultured on the trans-well plate of 0.4mm pore size (Millipore, United States) for a 5days. Then, the resistance of inserts was monitored by CellZscopeR-System (NanoAnalytics GmbH, Muenster, Germany). TEER value was calculated as Ω*cm². For paracellular permeability measurement, the tightness of wild-type or Claudin-5^112Δ5-mutated bEnd.3 monolayer cell was determined by the infiltration of FITC-labeled dextran (10kDa, 0.5mg/L, Thermo-Fisher, United States). After applying the dextran inside the well, the dye in the lower chamber was measured with a spectrophotometer-computer interfaced system (BioTek Epoch, United States) at a wave length of 594nm. For the measurements of TEER and permeation of tracers, three repeats of each measurement were investigated for each line.

Claudin-5^112Δ5-mutated bEnd.3 cells at a density of 80% were transfected with pIRES-eGFP or pIRES-eGFP-Claudin-5 according to the instructions of the Lipofectamine 2000 DNA transfection reagent (Invitrogen, 11668027). Then, we changed the medium to DMEM with 10% FBS after 6–8h. After 48h, the cells were stimulated with hypoxia for 4h, and the protein expression levels of Claudin-5, LC3 or inducible nitric oxide synthase (iNOS) were evaluated by immunostaining.

Brain microvascular endothelial cells were cultured to form a confluent monolayer in a 10cm dish. Then the cells were exposed to hypoxic conditions for 4 and 6h, and we used a cell membrane and cytoplasm extraction kit (Beyotime, P0033). In brief, the cells were washed two times with cold PBS, and 1ml of membrane protein extraction reagent A was added. PMSF was added within a few minutes before use to a final concentration of 1mM. The cells were collected in a 1.5ml EP tube and placed on ice for 10–15min. Ultrasonic crushing was performed twice at 0.6Hz for 4s/time. The samples were centrifuged at 700g for 10min at 4°C, and the supernatant was collected and centrifuged at 14,000g for 30min at 4°C. The supernatant contained the cytoplasmic proteins, which were stored at −80°C. The supernatant contained 30–50μl of supernatant residue to avoid contact with the precipitate. The samples were centrifuged at 4°C and 14,000g for 10s. Then, 40μl of membrane protein extraction reagent B was added to the precipitate, vortexed for 5s, and placed in an ice bath for 5–10min, and the previous vortex and ice bath incubation steps were repeated 1–2 times. The samples were centrifuged at 14,000g for 5min at 4°C, and the supernatant was collected to obtain the cell membrane protein fraction.

The cells were cultured to form a confluent monolayer in a 24-well plate. The cells were washed twice with cold PBS, and cold acetone was added and incubated for 10min. Then, the cells were washed once in cold 95% ethanol, once in cold 75% ethanol, and twice with PBS, after which the cells were blocked with blocking solution at room temperature for 1h (blocking solution: 1% BSA+0.5% Tween-20 in PBS). Then, the cells were incubated with the following primary antibodies at 4°C overnight (1:200, diluted with blocking solution): Claudin-5 (1:200; Invitrogen, 35–2,500), LC3A/B (1:200; Cell Signaling Technology, 4,108), Caveolin-1 (1:200; Santa Cruz Biotechnology, sc-894), hypoxia inducible factor 1 subunit alpha (HIF-1a; 1:200; ProteinTech, 20960-1-AP), and iNOS (1:100; ProteinTech, 18985-1-AP). The cells were then incubated with the following secondary antibodies at room temperature for 2h: Alexa Fluor 488-conjugated goat anti-mouse IgG (1:400; Thermo Fisher Scientific, A-11029) and Alexa Fluor 647-conjugated goat anti-rabbit IgG (1:400; Jackson ImmunoResearch, 111-605-003). DAPI (1:500; Sigma Aldrich, D9542) staining was also performed. For zebrafish tissue immunostaining, the brain sections were incubated with primary antibodies against Claudin-5 (Invitrogen, 35–2,500) and LC3B (Sigma-Aldrich, L7543), Alexa Fluor 647-conjugated goat anti-mouse IgG and Cy3-conjugated goat anti-rabbit IgG secondary antibodies and DAPI. Images were obtained with an Olympus FV3000 confocal laser microscope. Fluorescence quantitative analysis was performed using Image-Pro Plus software.

Western blotting was performed as described previously (Yu et al., 2019). In brief, protein samples were separated in 10–15% SDS-PAGE acrylamide gels, transferred onto PVDF membranes (Millipore, 0.22μm and blocked with 5% skim milk for 2h). The blots were then incubated at 4°C overnight with the following primary antibodies: Claudin-5 (1:1,000; Invitrogen, 34–1,600), LC3 (1:1,000; ProteinTech, 14600-1-AP), Caveolin-1 (1:200; Santa Cruz Biotechnology, sc-894), HIF-1a (1:1000; ProteinTech, 20960-1-AP), iNOS (1:1,000; ProteinTech, 18985-1-AP), LAMP-1 (1:1000; ProteinTech, 21997-1-AP), ATP1A1 (1:1,000, ProteinTech, 14418-1-AP), or β-actin (1:1,000; Servicebio, GB12001). Then, the blots were incubated at room temperature for 2h with HRP conjugated anti-rabbit or anti-mouse IgG (H+L; 1:1,000; Servicebio) secondary antibodies. Western blot bands were analyzed by adding ECL advance Western blotting detection reagents (Thermo, 34,580) and imaged using a SmartChemi-500 imaging system (Sage Creation Science, China). The immunoblot bands were quantitatively analyzed by ImageJ software.

Intracellular ROS were measured using 2',7'-dichlorofluorescein diacetate (DCFH-DA, Beyotime, S0033S). DCFH-DA was diluted with serum-free medium to a final concentration of 10μM. The cells were harvested in a 15ml centrifuge tube, and 1ml of diluted DCFH-DA was added to the cells and incubated for 20min at 37°C. The cells were washed three times with serum-free DMEM. Then, detection was performed by flow cytometry. Cell apoptosis was detected by Annexin V-FITC cell apoptosis detection kit (Beyotime, C1062M). Briefly, cells were harvested and stained with annexin V-FITC and propidium iodide for 20min at room temperature and examined using a flow cytometer.

For statistical analysis, GraphPad Prism 5 software was used. All values are presented as the means±SD, and two-tailed Student’s t-test or one-way ANOVA was applied to determine statistical significance (^*p<0.05, ^**p<0.01 and ^***p<0.005).

Our previous study has shown the activation of autophagy in endothelial cells in response to hypoxia (Yang et al., 2019b, 2020). To confirm the alterations of autophagic flux in bEnd.3 cells in response to hypoxia, bEnd.3 cells were transfected with RFP-GFP-LC3 adenovirus for live imaging analysis. As shown in Supplementary Figure S1, hypoxia induction increased the numbers of both GFP and RFP dots in bEnd.3 cells compared with that in normoxia-treated cells, and we found that the red/yellow dot ratio increased due to hypoxia induction, indicating an increased expression of LC3, a marker of autophagy, and a formation of autophagosomes to autolysosomes in hypoxia-treated cells. Previous studies have shown that short-term hypoxia induction causes the redistribution and endocytosis of membranous Claudin-5, and autophagy is then activated to mediate the degradation of abnormally accumulated Claudin-5 in cytosol, thereby eliminating ROS and protecting the integrity of the BBB barrier from fast injury (Yang et al., 2019b, 2020).

To explore the impact of the membranous TJ protein Claudin-5 on the occurrence of autophagy in response to hypoxia, two stabilized Claudin-5-knocked out bEnd.3 mutated cell lines, Claudin-5^112Δ5 with a 5bp deletion and Claudin-5^297Δ4 with a 4bp deletion, were first constructed and generated by CRISPR/Cas9 strategy (Supplementary Figure S2A). The expression of Claudin-5 was examined by immunofluorescence staining and immunoblotting respectively, and the results indicated a total loss of Claudin-5 from cell membrane in these lines (Supplementary Figures S2B,C). To verify the endothelial barrier function of bEnd.3 cells after loss of Claudin-5, the TEER and paracellular permeability of the wild-type or Claudin-5^112Δ5 mutated bEnd.3 monolayer cells were measured. The results indicated a significant loss of the tightness of monolayer Claudin-5^112Δ5 mutated cells in comparison to that of the wild-type bEnd.3 cells (Supplementary Figures S2D,C). Since a functional alteration was observed in bEnd.3 cells after loss of Claudin-5, we additionally analyzed the cell viability in response to hypoxia by CCK-8 assays in both wild-type and mutated bEnd.3 cells. The result showed that the survival rate of both Claudin-5-knocked out cells was still above 90% till 6h post-hypoxia induction, and there was no significant variations of the cell viability among the wild-type and Claudin-5-knocked out lines under short term hypoxic induction (Supplementary Figure S3).

Next, to verify whether Claudin-5 affects the occurrence of autophagy in response to hypoxia, the expression of LC3 in Claudin-5^wt and Claudin-5^112Δ5 bEnd.3 cells were analyzed by immunofluorescence staining first. As a result, it was found that in bEnd.3 Claudin-5^wt cells, Claudin-5 and LC3 showed obvious colocalization in the cytoplasm after 4 or 6h hypoxia induction (white arrows in Figure 1A). The expression level of LC3 in cells with Claudin-5 knockout was significantly reduced (Figure 1A). Meanwhile, the expression levels of LC3 and LAMP-1, a lysosomal membrane protein and a specific marker of autolysosomes (Zhan et al., 2018), in Claudin-5^wt and Claudin-5^112Δ5 bEnd.3 cells in response to hypoxia were analyzed by immunoblotting. The protein expression levels of both LC3 and LAMP-1 in bEnd.3 Claudin-5^112Δ5 cells were significantly lower than those in Claudin-5^wt bEnd.3 cells after 4 or 6h hypoxia induction (Figure 1B). The inhibition of the autophagy activation was confirmed by another Claudin-5 mutant bEnd.3 line of Claudin-5^297Δ4, where the expression of both LC3 and LAMP-1 was sufficiently reduced due to the loss of Claudin-5 in response to hypoxia induction for 4 or 6h (Supplementary Figure S4). A rescue experiment was also performed to further verify the specificity of Claudin-5 on hypoxia-induced autophagy occurrence. The pIRES-eGFP or pIRES-eGFP-Claudin-5 plasmid was first transfected into mutated Claudin-5^112Δ5 bEnd.3 cells. The followed immunofluorescence staining showed that cells successfully transfected with the pIRES-eGFP-Claudin-5 plasmid expressed LC3 at levels higher than the vehicle in response to 4h hypoxia induction (Supplementary Figure S5). These results revealed that endothelial membranous Claudin-5 was specifically involved in the activation of autophagy during the early stage of hypoxia.

We also sought in vivo evidences supporting the correlations between Claudin-5 and autophagy activation in endothelial cells under hypoxia. To this end, endothelial eGFP-specific transgenic Tg(kdrl:eGFP) zebrafish was applied to investigate the autophgy activation in endothelial cells (Yang et al., 2019b, 2020). Since zebrafish Claudin-5b is the main Claudin expressed in cerebrovascular endothelial cells during the embryonic stages (Yang et al., 2021), MO-mediated silencing of claudin-5b in zebrafish embryos was first performed. After 3days, it was found that in response to 3h hypoxia treatment, the control morphant embryo presented much severer brain injury than that of the claudin-5b morphants (red arrowheads, Figures 2A,B). Immunofluorescence staining with the embryonic brains further revealed that in control morphant zebrafish larvae endothelial cells, the expression level of LC3 was higher than that in the cerebrovascluar endothelial cells of claudin-5b morphants post-hypoxia induction (Figures 2C,D). Above in vivo and in vivo evidences confirmed that loss of Claudin-5 could partly affect the activation of endothelial autophagy caused by hypoxia induction.

Since, we have observed a higher phenotype ratio of brain injury in control embryos in comparison to that in the brain of claudin-5b morphants post-hypoxia treatment, we wondered whether lack of endothelial Claudin-5 could alleviate the cell injury or death caused by hypoxia in a certain extent. Therefore, we analyzed hypoxia-caused apoptosis of the Claudin-5^wt and Claudin-5 knocked out bEnd.3 cells. The results indicated that, after 4h-hypoxia induction, Claudin-5^wt bEnd.3 cell showed a higher apoptosis ratio than Claudin-5^112Δ5 or Claudin-5^297Δ4 bEnd.3 cells (Supplementary Figure S6).

It has been reported that in response to hypoxia induction, membranous Claudin-5 in the vascular endothelial cells is packaged by Cav-1-composed caveolae and is then endocytosed into the cytoplasm, where it is eventually degraded by autophagosomes or autolysosomes (Liu et al., 2012, 2016). Therefore, we next asked whether the expression and localization of Cav-1 in hypoxia-treated bEnd.3 cells will be altered since Claudin-5 is absent. To this end, an immunofluorescence staining against Cav-1 was first performed in Claudin-5^wt and Claudin-5^112Δ5 bEnd.3 cells after 4 or 6h hypoxia induction. The results indicated that most of the Cav-1 still localized in the membrane of Claudin-5^112Δ5 bEnd.3 cells due to the loss of Claudin-5, and that the Cav-1, which colocalized with Claudin-5 in the cytosol of Claudin-5^wt bEnd.3 cells was much more than that in Claudin-5^112Δ5 bEnd.3 cells in response to hypoxia for 4 and 6h (Figure 3A). There was no significant effect of accumulation or endocytosis of Cav-1 in bEnd.3 Claudin-5^112Δ5 cells in response to hypoxia for 4 and 6h (Figure 3A). The protein level of Cav-1 in the cell membrane and cytoplasm was also quantified by immunoblotting analyses. The results confirmed that due to lack of Claudin-5, there was almost no loss of membranous Cav-1 post-hypoxia induction for 6h, and there was no significant accumulation of Cav-1 in the cytoplasm either in comparison with that of the Claudin-5^wt bEnd.3 cells (Figure 3B). These results confirmed that Cav-1 is mainly responsible for the endocytosis of membranous Claudin-5 in response to hypoxia induction, while in the absence of Claudin-5, the delocalization of Cav-1 from endothelial membrane is sufficiently inhibited.

Previous studies have shown that iNOS-derived NO is the main cause of Cav-1-mediated Claudin-5 ectopic activity in response to hypoxia (Liu et al., 2016; Yang et al., 2019a). HIF-1a is a transcriptional regulator produced by the body during hypoxia, and HIF-1a is one of the main factors that regulates iNOS transcription (Robinson et al., 2011). Therefore, to ask whether the loss of Claudin-5 could conversely affect the production of iNOS and its upstream regulator of HIF-1a, we next analyzed the their expression levels in both wild-type and Claudin-5 mutant bEnd.3 cells after 4 or 6h hypoxia induction by immunofluorescence staining and immunoblotting. We found that the expressions of both iNOS and HIF-1a in Claudin-5 mutant bEnd.3 cells were significantly lower than that in Claudin-5^wt cells in response to hypoxia induction for 4 and 6h (Figures 4A,B). To confirm the specificity of membranous Claudin-5 on endothelial iNOS generation under hypoxia, rescue experiments by transfecting pIRES-eGFP-Claudin-5 plasmid in Claudin-5 mutant bEnd.3 cells were performed. It was found that restored expression of Claudin-5 in bEnd.3 Claudin-5^112Δ5 mutant cells could recover the generation of iNOS in response to hypoxia treatment for 4h (Supplementary Figure S7). Immunoblotting analyses also revealed that the protein expression levels of HIF-1a and iNOS in wild-type bEnd.3 cells after 4 and 6h of hypoxia were higher than that in Claudin-5^112Δ5 and Claudin-5^297Δ4 bEnd.3 cells (Figure 4C), indicating that lack of membranous Claudin-5 could partly inhibited the production of endothelial HIF-1a and iNOS under short term hypoxia induction.

ROS are produced by various sources in the CNS, including mitochondria, NADPH oxidase, and NOS, especially under hypoxia conditions (Yamamoto et al., 2006; Kuthati et al., 2019). We used DCFH-DA to measure intracellular ROS and then analyzed these levels by flow cytometry. It was found that the ROS level in Claudin-5^wt bEnd.3 cells was higher than those in both Claudin-5^112Δ5 and Claudin-5^297Δ4 mutant bEnd.3 cells in response to hypoxia induction for 4 and 6h, respectively (Figure 5A). This also suggests that the production of cellular ROS may be directly related to NOS in response to hypoxia. Previous studies have shown that under hypoxia induction, ROS could function on driving the expression of Bnip3 (Bcl-2 and adenovirus E1B 19-kDa interacting protein 3), which is a single transmembrane protein mediating autophagy to maintain cell survival (Zhang et al., 2019). Hence, immunoblotting analysis was performed to detect the protein expression level of Bnip3 in Claudin-5 mutated bEnd.3 cells. The results revealed that in comparison with the expression in wild-type control bEnd.3 cells, the expression level of Bnip3 in Claudin-5^112Δ5 bEnd.3 cells was significantly lower after 4 or 6h hypoxia induction (Figure 5B). Above results indicated that the existence of membranous Claudin-5 affects the production of ROS and the downstream Bnip3 expression in response to early hypoxia induction.

In summary, based on our previous finding that under early hypoxia injury, autophagy protects the BBB from fast breakdown by regulating the dynamics of Claudin-5, we further revealed the existence of membranous Claudin-5 could conversely affect the occurrence of autophagy in endothelial cells by both in vivo and in vitro evidences (Figure 5C). This is probably due to the attenuated endocytosis of caveolae-packaged Claudin-5 and reduced autophagic degradation of abnormally accumulated proteins in cytosol during the early hypoxia induction stages (as early as 6h). Meanwhile, our results indicate that lack of endothelial Claudin-5 caused a decreased sensitivity and production of HIF-1a and iNOS to early hypoxia induction, which together with the downstream ROS and Bnip3, affect the endothelial autophagy activation.