BI 1015550

Roflumilast prevents ischemic stroke-induced neuronal damage by restricting GSK3β-mediated oXidative stress and IRE1α/TRAF2/ JNK pathway

A B S T R A C T
Inhibition of phosphodiesterase 4 (PDE4) protects against neuronal apoptosis induced by cerebral ischemia. However, the exact mechanisms responsible for the protection of PDE4 inhibition have not been completely clarified. Roflumilast (Roflu) is an FDA-approved PDE4 inhibitor for the treatment of chronic obstructive pul- monary disease. The potential protective role of Roflu against ischemic stroke-associated neuronal injury remains unexplored. In this study, we investigated the effect and mechanism of Roflu against ischemic stroke using in vitro oXygen-glucose deprivation reperfusion (OGD/R) and in vivo rat middle cerebral artery occlusion (MCAO) models. We demonstrated that Roflu significantly reduced the apoptosis of HT-22 cells exposed to OGD/R, enhanced the nuclear translocation of nuclear factor erythroid 2-related factor 2 (Nrf-2), and reduced oXidative stress. Treatment with Roflu increased the phosphorylation of protein kinase B (Akt) and glycogen synthase kinase 3β (GSK3β) but decreased the level of phosphorylated inositol requiring enzyme 1α (IRE1α). Interestingly, constitutively active GSK3β (S9A) mutation abolished the effects of Roflu on oXidative stress and IRE1α phos- phorylation. Moreover, Roflu decreased the binding of IRE1α to tumor necrosis factor receptor-associated factor 2 (TRAF2) and attenuated the phosphorylation of c-Jun N-terminal kinase (JNK). We also found that PDE4B knockdown reduced the phosphorylation of both IRE1α and JNK, while overexpression of PDE4B antagonized the role of PDE4B knockdown on the activation of IRE1α and JNK. Besides, the inhibition of PDE4 by Roflu produced similar effects in primary cultured neurons. Finally, Roflu ameliorated MCAO-induced cerebral injury by decreasing infarct volume, restoring neurological score, and reducing the phosphorylation of IRE1α and JNK. Collectively, these data suggest that Roflu protects neurons from cerebral ischemia reperfusion-mediated injury via the activation of GSK3β/Nrf-2 signaling and suppression of the IRE1α/TRAF2/JNK pathway. Roflu has the potential as a protective drug for the treatment of cerebral ischemia.

1.Introduction
Cerebral ischemia is a devastating neurological disorder that affects millions of people each year worldwide [1]. Ischemic stroke is the most common type of cerebral ischemia, and it occurs under the condition that the blood flow to the brain is suddenly reduced or stopped. Intra- venous injection of recombinant tissue plasminogen activator is the only Food and Drug Administration (FDA)-approved treatment currently for patients suffering from acute ischemic stroke [2]. However, the application of rt-PA is restricted by its limitations and side effects, such as narrow therapeutic windows and high risk of bleeding [3]. Hence, further studies are necessary to identify novel targets and therapeutic candidates against ischemic brain injury.Phosphodiesterase 4 (PDE4) is highly expressed in neuronal cells [4]. Inhibition of PDE4 enhances the level of intracellular cAMP, which subsequently activates the downstream molecules protein kinase A (PKA) and exchange protein directly activated by cAMP (Epac) [5]. Accumulation of intracellular cAMP is also supposed to promote the phosphorylation of protein kinase B (AKT) and glycogen synthase kinase 3β (GSK3β) [6–8]. PDE4 inhibition produces strong anti-inflammatory and cognition-enhancing effects through acting on the above signaling pathways [9,10]. Among the 11 isoforms of the phosphodiesterase (PDE) superfamily (PDE1-11), PDE4 is the major isoform which is highly expressed in the mammalian brain, and thus PDE4 has been viewed as a promising target for the intervention of neurological disorders [9]. Our previous studies revealed that the inhibition of PDE4 protected against cerebral ischemia by inhibiting the excessive activation of astrocytes and microglial cells [11]. Consistently, we verified that the PDE4 in- hibitor FCPR03 was effective in decreasing the level of reactive oXygen species (ROS), attenuating neuronal apoptosis, and reducing infarct volume in rats subjected to ischemia/reperfusion injury [6].

Besides, our recent studies showed that inhibition of PDE4 attenuated oXidative stress and protected against over-activation of endoplasmic reticulum (ER) stress-induced by ischemia/reperfusion [12]. These findings attracted our attention to study the involvement of ER stress in the protective effects of PDE4 inhibition. Physiologically, appropriate ER stress is beneficial for the removal of intracellular misfolded/aggregated proteins and ER homeostasis [13]. Under the pathological conditions of and evolutionarily conserved signaling molecule during ER stress [21], Binding of IRE1α to tumor necrosis factor receptor-associated factor 2 (TRAF2) promotes cellular death through activating c-Jun N-terminal kinase (JNK). However, the signaling pathway linking PDE4 inhibition, ER stress, IRE1α, and neuronal apoptosis remains unexplored. The study aimed to evaluate the protective role of Roflu against ischemic injury and to identify the underlying signaling pathways. We hypothesize that Roflu would protect neurons from cerebral ischemia by increasing cAMP and activating downstream molecules, such as AKT and GSK3β, and finally regulating ER stress-mediated apoptotic signal pathways. To be noted, current PDE4 inhibitors are limited by severe side effects, such as emesis and vomiting. Roflu is an FDA-approved PDE4 inhibitor for the treatment of COPD exacerbations. Compared with the first-generation PDE4 inhibitors, Roflu yields only mild emetic side effects. The canon- ical PDE4 inhibitor Rolipram shows a strong effect on an emetic-like effect with a dose of 0.3 mg/kg, while Roflu shows an emetic poten- tial only at a dose of 3 mg/kg [22]. Hence, Roflu offers a more favorable window for the treatment of neurological diseases compared to Roli- pram [22]. In this study, we report that Roflu rescues neuronal cells from OGD/R and cerebral ischemia-reperfusion by restricting GSK3β-me-severe or prolonged ER stress, the biological functions of the ER are diated oXidative stress and suppressing ER stress-mediated IRE1-impaired and ultimately leads to cellular apoptosis [14]. Our previous study links PDE4 inhibition to neuroprotection in neuronal cells following ischemic stroke-induced injury [6]. However, the signaling pathways mediating the suppression of ER stress by PDE4 inhibition under the ischemic condition are less understood. And importantly, the involved ER stress pathway conferring the protective effects of PDE4 inhibition is worthy of further study.

Roflumilast (Roflu) is a selective PDE4 inhibitor with an IC50 of 0.8 nM. The PDE4 enzyme contains two narrow but deep hydrophobic pockets (Q1 and Q2) and a metal binding pocket (M pocket). The difluromethoXy group of Roflu binds at the Q1 pocket and makes more hydrophobic interactions with residues inside this pocket. While the cyclopropylmethylether group of Roflu binds at the Q2 pocket, the dichloropyridyl group extends to the M pocket and forms one hydrogen bond to a water molecule [15]. As a result, Roflu has a high potency toward PDE4. Roflu displays robust anti-inflammatory and immuno- modulatory effects, and it has been approved by the FDA for the treat- ment of severe chronic obstructive pulmonary disease (COPD) [10]. In recent years, the potential therapeutic role of Roflu in neurological disorders has aroused widespread concerns. In the central nervous sys- tem, Roflu has been initially reported to ameliorate hypertension-induced cognitive deficits through upregulating cAMP/cAMP-response element-binding protein (CREB) signaling and enhancing the expression of brain-derived neurotrophic factor in the rat hippocampus [16]. Roflu treatment has also been reported to restore chronic cerebral hypoperfusion-induced cognitive impairments in aged rats [17]. These findings are consistent with the observation that Roflu restores hippocampal long term potentiation after blast-induced trau-matic brain injury [18]. Recently, Wang et al., showed that Roflu not only improved learning and memory but also attenuated depression-like behavior in APP/PS1 transgenic mice [4]. As Roflu improves cognition in multiple models, it has been proposed as an excellent candidate for the treatment of dementia or brain injury [19]. However, whether Roflu has neuroprotective effects in a cerebral ischemia-reperfusion model is unclear.Inspired by the protective effects of Roflu against cognitive deficits, we are interested to test the role of Roflu in cerebral ischemia. The ER stress is viewed as an essential step during the pathology of cerebral ischemia/reperfusion injury, and prolonged periods of ER stress are deleterious to neurons [20]. Our previous studies indicated that PDE4 knockdown inhibited oXygen-glucose deprivation (OGD)-triggered ERα/TRAF2/JNK pathway, suggesting that PDE4 inhibition is a promising strategy for the treatment of cerebral ischemia, and Roflu is a potential drug to reverse brain damage and ER stress caused by cerebral ischemia.

2.Materials and methods
2.1.Reagents
Roflu (#S2131) was obtained from Selleck Chemicals (Houston, TX, USA). N-acetyl-L-cysteine (NAC, #T0875) was the product of TargetMol (Wellesley, MA, USA). Antibodies against caspase 3 (#9662), phos- phorylated AKT (p-AKT, Ser473, #4060), AKT (#2920), phosphorylated JNK (p-JNK, Thr183/Tyr185, #4668), JNK (#9252), phosphorylated GSK3β (p-GSK3β, Ser9, #5558), GSK3β (#12456), TRAF2 (#4724),
IRE1α (#3294), Nrf-2 (#12721), Histone H3 (#4499), and IgG (#3900)were purchased from Cell Signaling Technology Inc. (Massachusetts, USA). Antibodies against neuron-specific nuclear protein (NeuN, #mab377), Immoblilon PVDF membranes (#ISEQ00010), and Immo- bilon Western Chemiluminescent HRP Substrate (#WBKLS0100) were purchased from Merck (Darmstadt, Germany). Mounting medium with DAPI (#ab104139), antibodies against phosphorylated IRE1α (p-IRE1α, Ser724, #ab48187), and microtubule-associated protein 2 (MAP2, #ab11268) were obtained from Abcam (Cambridge, UK). Goat anti- mouse IgG DyLight 488 (#A23210), Goat anti-Rabbit IgG DyLight 488 (#A23220), and Goat anti-Rabbit IgG Dylight 549 (#A23320) were obtained from Abbkine Scientific (Wuhan, Hubei, China). Opti-MEM™ Reduced Serum Medium (#31985070), High glucose DMEM (#11965092), glucose-free DMEM (#11966025), a Pierce™ Rapid Gold BCA Protein Assay Kit (#A53225), Pierce™ Protein A/G Magnetic Beads (#88803), Neurobasal (#21103049), B27 (#17504001), L-Glutamine(#25030149), Lipofectamine 2000 Reagent (#11668027), Lipofect- amine 3000 Reagent (#L3000008), Fetal bovine serum (#A3161001C), and CellROX™ Deep Red Reagent (#C10422) were obtained from Thermo Fisher Scientific (Waltham, MA, USA). The Cell Counting Kit-8 (CCK-8) was purchased from Dojindo Molecular Technologies (#CK04; Tokyo, Japan). Small interfering RNAs (siRNAs) targeted PDE4B were obtained from GenePharma (Shanghai, China). The PDE4B and GSK3β (S9A) plasmids were obtained from Shuangquan Biotechnology (Guangzhou, Guangdong, China). A nuclear and cytoplasmic Protein EXtraction Kit (#P0027), a lactate dehydrogenase (LDH) CytotoXicity Assay Kit (#C0017), a Catalase Assay Kit (#S0051), and a Hydrogen stress in HT-22 cells, as evidenced by decreased expression of the 78-kDa PeroXide Assay Kit (#S0038) were from Beyotime Biotechnology glucose-regulated protein (GRP78) and phosphorylated eukaryotic translation-initiation factor 2α (eIF2α) in the cells after knocking down PDE4 [12]. Inositol requiring enzyme 1α (IRE1α) is the most prominent(Shanghai, China). Antibodies against GAPDH (#FD0063) and β-tubulin (#FD0064), Goat anti-rabbit IgG(H + L)-HRP(#FDR007), Goat anti- mouse IgG(H + L)-HRP (#FDM007), IPKine mouse anti-rabbit IgGlight chain secondary antibody, HRP labeled (#FD0115, for co-IP), and RIPA buffer (#FD009, for IB; #FD011, for co-IP) were purchased from FDBIO SCIENCE (Hangzhou, Zhejiang, China). A cAMP Parameter Assay kit (# KGE012) was obtained from R&D Systems (Minnesota, USA). Dihydroethidium (DHE, #S0063) and Total SuperoXide Dismutase Assay Kit with WST-8 (#S0101 M) were purchase from Beyotime Biotech- nology (Shanghai, China).

2.2.Animals
Male Sprague-Dawley (SD) rats weighing 260─300 g were used in the experiment. Rats were raised in a 12 h light/dark cycle and free access to food and water. All of the animal experiments were approved by the Laboratory Animal Ethics Committee of Southern Medical Uni- versity (Guangzhou, Guangdong, China) and conducted following the ethical guidelines of the NIH Guide for the Care and Use of Laboratory Animals (NIH, revised 1996). Animal studies are reported in compliance with the ARRIVE guidelines [23]. Power calculations of animal sample size were conducted similar to our previous studies [6]. Power calcu- lation analysis showed that a sample of eight rats per group was considered sufficient to detect a difference with 95% confidence and 80% power. A total of 74 rats were enrolled in this study. 18 rats were assigned to the sham-operated group. From a total of 56 rats submitted to MCAO, 11 rats were excluded from the analysis due to no neurologic abnormalities (7) or death (4). Data were reported on 63 rats.

2.3.Drug preparation
For in vivo experiments, Roflu was dissolved in 0.9% saline con- taining 5% Tween-80 and 0.5% dimethyl sulfoXide (DMSO). For in vitro experiments, Roflu was dissolved in DMSO at the stock concentration of 10 mM. Roflu was diluted to the desired final concentration before each experiment, and the final concentration of DMSO is 0.1%.

2.4.Cell culture
The HT-22 cell line was obtained from Merck (Darmstadt, Germany). The cells used in this study were from passage 5–20. Cells were cultured in DMEM containing 10% FBS. Cells were cultured at 37 ◦C in a humidified environment containing 5% CO2 and 95% air.The culture of primary cortical neurons was conducted as previously described with minor modifications [12]. Briefly, the cortices of em- bryonic day 18 fetal SD rats were dissected by micromanipulation, and the dissected cortex was digested with 0.125% trypsin for 5 min at 37 ◦C. Then, the digestion was terminated with DMEM containing 5% FBS, and cells were dispersed with a Pasteur pipette and seeded onto plates or glass-bottom dishes pre-coated with poly-D-lysine. At 4–6 h after seed- ing, the medium was changed to Neurobasal medium supplemented with 2% B27 and 1% glutamine. The medium was half-changed every 3 days and culture was maintained 8–14 days for the experiment.

2.5.cAMP assay
HT-22 neuronal cells were seeded in a 6 cm dish at the density of 80–90% confluence. Cells were treated with various concentrations of Roflu (0.1, 0.3, and 1 μM) for 6 h. Then, the intracellular cAMP level was detected using a cAMP Parameter Assay kit according to the manufac- ture’s instruction.

2.6.Oxygen-glucose deprivation and re-oxygenation
The oXygen-glucose deprivation and re-oXygenation (OGD/R) model was established as previously reported [12]. Briefly, HT-22 cells or neurons were washed with PBS twice, and the medium was replaced with glucose-free DMEM. The cells were then transferred to a hypoXia chamber (Billups-Rothenberg, Del Mar, CA, USA) flushed with 95% N2 and 5% CO2. The control groups were cultured with DMEM with glucose in an aerobic environment. For reoXygenation, the medium was replaced with high-glucose DMEM. For HT-22 cells, cells were exposed to OGD for 6 h. For primary cortical neurons, cells were exposed to OGD for 1 h.

2.7.Cell viability
Cell viability was detected by using the CCK-8 kit. Cells were treated with various concentrations of Roflu for 2-4 h. Then, the cells were subjected to the indicated time of OGD followed by 24 h of re- oXygenation. After treatment, cells were cultured with 20 μl CCK-8 in each well at 37 ◦C for 1 h. And the absorbance at 450 nm was detected by a multi-functional microplate reader (Synergy HT; Biotech, Winooski, VT, USA). Quantitative analysis of cell viability was normalized to the OD value of the control group to exclude unwanted variation.

2.8.LDH cytotoxicity assay
The detection of cell toXicity was based on the release of LDH. HT-22 neuronal cells were subjected to 6 h of OGD insult followed by 24 h re- oXygenation. Then, the medium was transferred toward a 96-well plate. The release of LDH was detected by an LDH CytotoXicity Assay Kit. The cell toXicity was calculated as follows: Cell toXicity (absorbance of theprocessed sample – absorbance of medium control)/(absorbance of cell maximum enzyme activity – absorbance of medium control) × 100.

2.9.Transfection of small interfering RNA (siRNA)
PDE4B knocking down was conducted by transfecting siRNA specific for PDE4B. The sequences of the siRNA for PDE4B were: sense: 5′- CCUGCAAGAAGAAUCAUAUTT-3′,antisense:5′-AUAUGAUUCUU- CUUGCAGGTT-3’. siRNA with a scrambled sequence was used as a control. For transfection, the cells were seeded at a density of 80% confluence. The PDE4B siRNA and lipofectamine 2000 reagent were diluted with Opti-MEM independently and incubated at room temper- ature for 20 min. The cells were then incubated with the miXtures for 6 h. After treatment, the medium was replaced with DMEM containing 1% FBS.

2.10.Plasmid transfection
PDE4B or GSK3β (S9A) plasmid was transfected into HT-22 neuronal cells using lipofectamine 3000 reagents according to the manufacture’s instruction. Briefly, plasmids were miXed with P3000 in Opti-MEM, and the miXture was then incubated with Opti-MEM containing lipofect- amine 3000 for 10 min. The DNA-lipid miXture was then added into each well.

2.11.Immunocytochemistry
Cells were fiXed with 4% PFA at room temperature for 30 min and permeabilized with 0.5% Triton X-100. The unspecific binding sites were blocked with 5% goat serum at room temperature for 1 h. Then, the cells were incubated with antibodies against p-JNK, p-IRE1α, MAP2. Following several washes, samples were incubated with secondary an- tibodies labeled with Dylight 488 or Dylight 549 at room temperature for 1 h. Finally, the cells were counterstained with DAPI and imaging with a confocal microscope (Nikon ECLIPSE Ti, Tokyo, Japan).

2.12.Immunoprecipitation
HT-22 neuronal cells were lysed using RIPA buffer containing 1% protease inhibitor. After centrifugation, the supernatant was pre-cleared
with magnetic beads to exclude unspecific binding protein. The super- natant was then incubated with the primary antibody overnight at 4 ◦C to form an immune complex. Subsequently, 25 μl of pre-washed protein A/G magnetic beads was added to the antigen–antibody miXture and incubated for 1 h at room temperature with constant miXing. Then, the beads were collected with a magnetic stand and washed with IP buffer (pH 7.4, 0.025 M Tris, 0.15 M NaCl, 0.001 M EDTA, 1% NP40, 5% glycerol) for 5 times. Finally, the targeted antigens were eluted by Low- pH elution at room temperature with constant miXing for 10─15 min.

2.13.Extraction of cytoplasmic and nuclear fraction proteins
The cytoplasmic and nuclear proteins were isolated using a Nuclear and Cytoplasmic Protein EXtraction Kit according to the manufacture’s instruction. Briefly, cells were collected by scraping and shaking with cytoplasmic protein extraction reagent A containing 1% proteases in- hibitor for 30 min. Then, cytoplasmic protein extraction reagent B was added to each tube and the sample was vortexed for 5 s followed by centrifugation at the speed of 12000 rpm for 5 min. The supernatant was collected as the cytoplasmic protein. For the extraction of nuclear pro- tein, the pellet was vortexed with nuclear protein extraction reagent contained 1% protease inhibitor for 30 s, and incubated in an ice bath for 1 h. Finally, the nuclear fraction protein was collected by centrifugation at the speed of 12000 rpm for 10 min, and the supernatant was collected as the nuclear protein.

2.14. Detection of catalase activity
Cells were lysed with RIPA buffer followed by centrifugation at 12000 rpm for 10 min. The supernatant was collected for the detection of catalase activity. Then, the sample was diluted with catalase detection buffer to a volume of 40 μl. H2O2 (250 mM, 10 μl) was added to each sample to react for indicative time. The reaction was terminated by adding a 450 μl catalase reaction stop solution. The left H2O2 was detected by the peroXidase solution. The catalase activity was calculated as follows: Catalase Enzyme Activity H2O2 consumed by catalase of sample dilution times/(reaction minutes sample volume protein concentration).

2.15.Detection of intracellular H2O2
Intracellular H2O2 was detected according to the manufacturer’s instruction. Briefly, after treatment, cells were quickly frozen at 80 ◦C for at least 4 h. Then, cells were collected by scraping and homogenated procedures without insertion of the filament.

2.17.Evaluation of the infarct volume
To detect the infarct volume, the 2,3,5-triphenyl tetrazolium chlo- ride (TTC) staining was conducted as previously described [6]. Briefly, SD rats were euthanised by overdose of isoflurane. Then, the brains were frozen at 30 ◦C for about 15 min and cut into 2 mm thick coronal slices by brain matrices (RWD Life Science). The brain slices were incubated in 2% TTC solution (dissolved in saline) at 37 ◦C for 30 min. Finally, the slices were fiXed with 4% PFA for 30 min and scanned by a scanner(Canon, Tokyo, Japan). The infarct volume was analyzed by Image J and calculated as follows: infarct volume (contralateral hemisphere vol- ume – not infarct ipsilateral hemisphere volume)/contralateral hemi- sphere volume.

2.18.Neurological deficit measurement
Neurological deficit scores were evaluated at 24 h after the MCAO surgery as previously described [6]. “0” represents no deficit; “1” rep- resents a failure to extend contralateral forelimb upon lifting the rats by its tail; “2” represents circling to the contralateral side while walking; “3” represents falling to the contralateral side while walking; “4” pre- sents no spontaneous walking.

2.19.Western blotting
Sample (brain tissues and cells) were collected and quickly frozen at 80 ◦C. Then, tissues and cells were lysed with RIPA buffer containing
1% protease inhibitor and 1% phosphatase inhibitor on ice followed by ultrasonic lysis and centrifugation. Then, the supernatants were collected and the protein concentrations were detected by using a BCA Protein Assay Kit. After protein quantification, the supernatants were boiled at 95 ◦C for 15 min. The protein extracts were resolved by SDS- PAGE and transferred toward 0.22 μm PVDF membranes. Then, the PVDF membranes were blocked with 5% skim milk, followed by three washes with TBST. After blocking the unspecific binding antigen, the membranes were incubated with the primary antibody against the target protein overnight. The next day, the membranes were washed three times with TBST to remove unbound primary antibody and incubated with a secondary antibody labeled with HRP at room temperature for in hydrogen peroXide detection lysis solution. The lyse was cen-h. Finally, the membranes were washed three times with TBST and trifugated at the speed of 12000 rpm for 5 min. The supernatant was used for the detection of the concentration of H2O2 and the protein concentration. Finally, the intracellular H2O2 level was calculated as follows: intracellular H2O2 level H2O2 concentration/protein concentration.

2.16.MCAO procedure
The MCAO surgery was performed as previously reported [6]. SD rats were randomly assigned to four groups: sham group, MCAO VEH group, MCAO 0.3 mg/kg Roflu group, and MCAO 1 mg/kg Roflu group. Rats were anesthetized with isoflurane by using an anesthesia ventilation system (RWD Life Science, Shenzhen, Guangdong, China).Body temperature was maintained at 37 0.5 ◦C by using a self-controlled heating blanket with an anal thermometer (Beijing Sino Technology, Beijing, China). A 0.36 mm silicone-coated filament was inserted advanced a distance of 18–22 mm into the internal carotid ar- tery (ICA) to occlude the middle cerebral artery (MCA). And the decrease of cerebral blood flow was monitored by Laser Speckle Contrast Imaging (RWD Life Science). After 2 h of focal cerebral ischemia, the suture was gently extracted for reperfusion. Rats were excluded from analysis under the conditions: (1) the neurological score was 0, (2) subarachnoid hemorrhage was found, or (3) rats were dead after MACO reperfusion. Rats in the sham group underwent the same surgical detected by using the Western Chemiluminescent HRP Substrate. Densitometric analysis was conducted by using Image J.

2.20.Immunofluorescence staining
Immunohistochemistry was conducted as previously described [24]. Briefly, at 6 h after MCAO, SD rats were perfused with pre-cold saline under deep anesthesia to exclude the blood followed by perfusion of 4% PFA. The brain was collected and immersed in 4% PFA overnight, then dehydrated with 20% and 30% sucrose. Coronal sections were cut using Cryostat Microtome (Leica CM1950, Nussloch, Germany) and per- meabilized with 0.3% Triton X-100. The sections were blocked with 5% goat serum at room temperature for 1 h, then incubated with primary antibody against p-IRE1α and NeuN followed by incubated with sec- ondary antibody conjugated with Dylight 488 and Dylight 549. Finally, the sections were attached to a glass slide and mounted with Fluo- roshield Mounting Medium with DAPI and visualized with a confocal microscope.

2.21.DHE staining
DHE staining was utilized to detect the level of intracellular super- oXide anion. Briefly, after OGD/R insults, HT-22 cells were incubated with 5 μM DHE in DMEM for 30 min. Then, the cells were washed with PBS twice. Finally, the level of intracellular superoXide anion was observed by a confocal microscope.

2.22.Detection of superoxide dismutase (SOD) activity
Intracellular SOD activity was detected using a Total SuperoXide Dismutase Assay Kit with WST-8. Briefly, after OGD/R insults, HT-22 cells were lysed with SOD sample preparation solution on the ice. Then, the samples were centrifuged at the speed of 12000 rpm for 5 min, and the supernatant was collected for the detection of SOD activity accord- ing to the manufacture’s instruction. Finally, the level of SOD activity was normalized to the protein levels.

2.23. Statistical analysis
All of the data excepted the neurological deficit socres data are expressed as the mean ± standard deviation (SD) and analyzed by One- way ANOVA (analysis of variance) followed by Tukey’s post-hoc test. The neurological deficit scores data were analyzed by a non-parametric
Kruskal-Wallis test followed by Dunn’s post-hoc test. The data were analyzed with SPSS 19.0 software (SPSS Inc., Chicago, IL, USA), and the histograms were plotted with GraphPad Prism 8.3.0 (GraphPad Software, La Jolla, CA, USA). Significance was defined as P < 0.05. For all studies, experimenters were blind to the treatment conditions. Each experiment was conducted at least three times, and each sample was repeated three times. 3.Results 3.1.Inhibition of PDE4 by Roflu reduces neuronal cell death caused by OGD/R insult To investigate whether Roflu could protect against neuronal injury caused by ischemia/reperfusion insult, we first studied the effects of Roflu on OGD/R-induced cell death. The chemical structure of Roflu is shown in Fig. 1A. HT-22 cells were exposed to OGD insult for 6 h and reoXygenation for 24 h in the presence or absence of various concen- trations of Roflu (0.1, 0.3, and 1 μM). Cell viability was then determined by CCK-8 assay. As shown in Fig. 1B, OGD/R caused a significant decrease in cell viability (p < 0.01), while Roflu (0.1–1 μM) enhanced cell viability in a dose-dependent manner. LDH assay showed the elevated LDH activity in the medium after OGD/R insult (p < 0.01). Roflu (0.1–1 μM) dose-dependently reduced LDH release into the me- dium, and a significant decrease in the release of LDH was observed at the dose of 1 μM (p < 0.01, Fig. 1C). DHE staining and SOD activity assay were performed to investigate the anti-oXidant effects of Roflu under the condition of OGD/R. We found that compared with the OGD/R group, treatment with Roflu reduced the fluorescence intensity of DHE (Fig. 1D and E). We also found that Roflu improved the activity of SOD signifi- cantly (Fig. 1F). To further verify the protective effects of Roflu, Western blotting was performed to detect the level of cleaved caspase 3. As shown in Fig. 1G and H, OGD/R insult increased the level of cleaved caspase 3. Treatment with Roflu (1 μM) reversed the increase in the level of cleaved caspase 3. These data further suggest that Roflu protects neuronal cells against OGD/R-induced apoptosis. Finally, Calcein-AM/ PI staining showed that exposure of HT-22 cells to OGD/R alone significantly enhanced the ratio of PI-positive cells (p < 0.01, Fig. 1I and J). In contrast, treatment with Roflu resulted in a marked decrease in the ratio of PI-positive cells (p < 0.01, Fig. 1I and J). These data suggest that treatment with Roflu leads to cellular protection against OGD/R insult in HT-22 cells. A Roflu concentration of 1 μM was selected for further experimentation. 3.2.Phosphorylation of GSK3β at Ser9 is essential for the Roflu-induced protection of HT-22 cells from OGD/R-induced injury The above results revealed the protective effects of Roflu against cellular injury caused by OGD/R. We then moved to investigate the involved signaling pathways. Previous studies have shown that inhibi- tion of PDE4 activated Epac/AKT pathway in multiple models [7,25]. Thus, we then investigated the involvement of AKT. Firstly, we found that inhibition of PDE4 by Roflu (0.1–1 μM) resulted in an increased intracellular level of cAMP in a dose-dependent manner (Fig. 2A). Elevation of intracellular cAMP promotes the phosphorylation of AKT. Thus, we further detected the phosphorylation of AKT under our experimental conditions. As expected, OGD/R caused a significant reduction of the phosphorylation of AKT (p < 0.01), while inhibition of PDE4 by Roflu reversed the role of OGD/R and enhanced the level of phosphorylated AKT (p < 0.05, Fig. 2 B–C). GSK3β is a canonical downstream target of AKT, and GSK3β is involved in neuronal apoptosis. Hence, we further explored the phosphorylation of GSK3β in HT-22 neuronal cells. Compared with the OGD/R group, Roflu treatment effectively increased the level of phosphorylated GSK3β (Ser9) (p < 0.05, Fig. 2 D–E), which was consistent with our previous studies. To further verify the essential role of GSK3β in the protective effects of Roflu, we transfected a mutant GSK3β (S9A) plasmid, which carries a constitutively active form of GSK3β, with a mutation of serine to alanine, into HT-22 neuronal cells. The efficiency of transfection was confirmed by Western blotting (Fig. 2F). We then found that over-expression of GSK3β (S9A) attenuated the phosphorylation of GSK3β induced by Roflu (p < 0.01, Fig. 2G and H). We also found that overexpression of GSK3β (S9A) abolished the protective effects of Roflu in HT-22 cells following OGD/R (p < 0.01, Fig. 2I). These data suggest that phosphorylation of GSK3β at Ser9 is essential for the neuroprotective effects of Roflu against OGD/R-triggered cell death. 3.3.Constitutively active GSK3β (S9A) mutation abolishes the inhibitory effect of Roflu on oxidative stress Our previous study showed that inhibition of PDE4 produced an anti- oXidant effect in neuronal cells [12]. However, whether GSK3β mediates the anti-oXidant effect of PDE4 inhibition remains unknown. Thus, we transfected GSK3β (S9A) into HT-22 neuronal cells and detected the intracellular sub-localization of Nrf-2, a key transcription factor of the antioXidant response. In line with our previous results, OGD/R pro- moted the translocation of Nrf-2 from the cytoplasm into the nucleus,while inhibition of PDE4 by Roflu further promote the translocation of Nrf-2 to nuclei (Fig. 3A–C, p < 0.01). Meanwhile, overexpression of GSK3β (S9A) reduced the nuclear fraction of Nrf-2 (Fig. 3A–B, p < 0.01). Next, we detected the production of intracellular ROS in GSK3β (S9A) over-expressing HT-22 neuronal cells. Our results indicated that over-expression of GSK3β (S9A) abolished the antioXidant effect of Roflu and increased the production of ROS under OGD/R condition (Fig. 3D and E). We also detected the activity of catalase and the level of intra- cellular H2O2 in HT-22 cells following OGD/R insult. As shown in Fig. 3 F and G, OGD/R decreased the activity of catalase (p < 0.01) and caused an increase in intracellular H2O2 (p < 0.01), while Roflu reversed these oXidation-associated indicators. Interestingly, overexpression of GSK3β (S9A) blocked the neuroprotective effects of Roflu and increased oXidative stress. These data indicate that inhibition of GSK3β mediates the anti-oXidant effects of Roflu. 3.4.GSK3β (S9A) mutation reverses the effects of Roflu on IRE1α activation EXcessive oXidative stress contributes to increased ER stress, which leads to the activation of the apoptosis pathway [26]. Previous results verified that inhibition of PDE4 attenuated the ER stress triggered by oXidative stress [12]. IRE1α is an ER stress sensor, whose activity Fig. 1. Inhibition of PDE4 by Roflu reduces neuronal cell death caused by OGD/R insult. HT-22 neuronal cells pretreated with Roflu or vehicle were subjected to 6 h of OGD insult followed by 24 h re-oXygenation. (A) The chemical structure of Roflumilast. (B) Cell viability was detected by a CCK-8 assay (n = 3). (C) Cell toXicity was detected by LDH assay (n = 4). (D and E) HT-22 neuronal cells pretreated with Roflu (1 μM) or vehicle were subjected to 6 h of OGD insult followed by 1 h re- oXygenation. The level of intracellular superoXide anion was detected by DHE staining and quantified by Image J (n = 3). (F) SOD activity was detected by a Total SuperoXide Dismutase Assay Kit (n = 4). (G and H) HT-22 neuronal cells pretreated with Roflu (1 μM) or vehicle were subjected to 6 h of OGD insult followed by 6 h re-oXygenation. The expression of cleaved caspase-3 was detected by Western blotting. The relative level of cleaved caspase 3/caspase 3 was then determined by densitometry of the blots (n = 3). (I and J) HT-22 neuronal cells pretreated with Roflu (1 μM) or vehicle were suffered from 6 h of OGD insult followed by 24 h of re- oXygenation. Cell death was visualized by Calcein-AM/PI staining. The percent of cell death was measured with Image J (n = 3). Results are expressed as the mean ± SD, *p < 0.05, **p < 0.01 versus the indicated group. Fig. 2. Mutation of GSK3β at Ser9 reverses the protective effects of Roflu against OGD/R-induced injury. (A) HT-22 neuronal cells were treated with the indicated concentration of Roflu or vehicle for 6 h. Then, the intracellular level of cAMP was detected using a cAMP Parameter Assay kit (n = 3). (B–E) HT-22 neuronal cells pretreated with Roflu (1 μM) or vehicle were subjected to 6 h of OGD insult followed by 1 h of re-oXygenation. Then, the levels of p-AKT, AKT, p-GSK3β, and GSK3β were detected by Western blotting. The relative levels of p-AKT/AKT (n = 4) and p-GSK3β/GSK3β (n = 3) were determined by densitometry of the blots. (F) HT-22 neuronal cells were transfected with GSK3β (S9A) plasmid. At 36 h after transfection, the expression of GSK3β was verified by Western blotting. (G and H) HT-22 neuronal cells transfected with an empty vector or GSK3β (S9A) plasmid were treated with Roflu (1 μM) for 1 h, then cells were subjected to 6 h of OGD insult followed by 1 h of re-oXygenation. The levels of p-GSK3β and total GSK3β were detected by Western blotting. The relative level of p-GSK3β/GSK3β was determined by densitometry of the blots (n = 3). (I) HT-22 neuronal cells transfected with GSK3β (S9A) plasmid or vector were treated with Roflu (1 μM) for 1 h, then cells were subjected to 6 h of OGD insult followed by 24 h of re-oXygenation. Cell viability was detected by a CCK-8 assay (n = 3). Results are expressed as the mean ± SD, *p < 0.05, **p < 0.01 versus the indicated group induces apoptosis [27]. Here, we are interested in exploring the role of Roflu on ER stress and the alteration of IRE1α activation. We found that OGD/R induced robust ER stress, as evidenced by an increase in the level of GRP78 (p < 0.01). EXposure of HT-22 cells to Roflu (1 μM) resulted in decreased expression of GRP78 (Fig. 4A and B, p < 0.01), indicating ER stress suppression. The level of phosphorylated IRE1α in HT-22 neuronal cells was then determined by Western blotting. Our results showed that OGD/R increased the level of p-IRE1α (Fig. 4D and E, p < 0.01), while treatment with Roflu reduced the activation of IRE1α (Fig. 4D and E, p< 0.01). Immunocytochemistry also showed that Roflu attenuated the activation of IRE1α induced by OGD/R (Fig. 4C). In order to investigate the essential role of GSK3β in the inhibitory effect of Roflu on the activation of IRE1α, we then overexpressed GSK3β (S9A) in HT-22 neuronal cells. Our results showed that overexpression of GSK3β (S9A) reversed the inhibitory effects of Roflu on the phosphorylation of IREα induced by OGD/R (Fig. 4F and G). Furthermore, we determined the role of oXidative stress on the phosphorylation of IRE1α. We found that OGD/R caused a significant increase in the phosphorylation of IRE1α. While NAC, a free radical scavenger, significantly reduced phosphorylated IRE1α (Fig. 4H and I, p < 0.01). NAC also increased cell viability in cells exposed to OGD/R (Fig. 4J, p < 0.05), indicating that oXidative stress is involved in ER stress and cellular apoptosis. These data suggest that inhibition of GSK3β is involved in the inhibitory effects of Roflu on ER stress and IRE1α activation. 3.5.Roflu disrupts the interaction between IRE1α and TRAF2 and attenuates the activation of JNK induced by OGD/R insult IRE1α triggers cell death by interacting with TNF receptor-associated factor 2 (TRAF2). TRAF2 interact and promotes the phosphorylation of JNK [28]. Sustained JNK activation is known to cause apoptosis [29,30]. Having known that Roflu suppressed the activation of IRE1α in HT-22 cells exposed to OGD/R, we then studied the effect of Roflu on the interaction between IRE1α and TRAF2. An immunoprecipitation Fig. 3. Overexpression of GSK3β (S9A) abolishes the inhibitory effect of Roflu on antioXidant stress. HT-22 neuronal cells transfected with a GSK3β (S9A) plasmid or vector were treated with Roflu (1 μM) for 1 h, then cells were subjected to 6 h of OGD insult followed by 1 h of re-oXygenation. (A–C) The change of cytosolic and nuclear levels of Nrf-2 was detected by Western blotting. The relative levels of Nrf-2/Histone H3 and Nrf-2/tubulin were determined by densitometry of the blots (n= 3). (D and E) The intracellular ROS was visualized using CellROX Deep Read Reagent. Intracellular ROS was quantified by Image J (n = 3). (F) Catalase activity was detected by a catalase assay kit (n = 3). (G) Intracellular H2O2 was determined by a hydrogen peroXide assay kit (n = 4). Results are expressed as the mean ± SD, *p < 0.05, **p < 0.01 versus the indicated groupexperiment was conducted to determine the effect of Roflu on the binding of IRE1α with TRAF2. As shown in Fig. 5A–D, OGD/R treatment caused an enhanced binding of IRE1α to TRAF2, while inhibition of PDE4 by Roflu disrupted the binding. We then moved to detect the phosphorylation of JNK. In line with the immunoprecipitation, the Western blotting analysis showed that OGD/R caused a significant increase in the level of p-JNK (p < 0.01), while Roflu reduced the acti- vation JNK (p < 0.01, Fig. 5E and F). The fluorescent signal of p-JNK in the cells was also observed with a fluorescent microscope. Consistently,OGD/R increased the fluorescence intensity, while Roflu (1 μM) treat- ment resulted in a decrease in fluorescence intensity (Fig. 5G). These results indicate that inhibition of PDE4 by Roflu disrupts the interaction between IRE1α and TRAF2, and inhibits the activation of JNK. 3.6.PDE4 knockdown attenuates the phosphorylation of IRE1α and JNK ER stress plays a causal role in neuronal death [20]. Having established that Roflu protects against apoptosis and suppresses ER stress. Fig. 4. Constitutively active GSK3β (S9A) mutation abolishes the inhibitory effect of Roflu on the phosphorylation of IRE1α. HT-22 neuronal cells pretreated with Roflu (1 μM) or vehicle were subjected to 6 h of OGD insult followed by 1 h of re-oXygenation. (A) The expression of GRP78 was detected by Western blotting. (B) The relative level of GRP78/GAPDH was determined by densitometry of the blots (n = 3). (C) The change of p-IRE1α was visualized by immunofluorescence staining. (D) The variances of p-IRE1α and IRE1α were detected by Western blotting. (E) Densitometric quantification of p-IRE1α/IRE1α in (D) (n = 3). (F) HT-22 neuronal cells transfected with GSK3β (S9A) plasmid or vector were treated with Roflu (1 μM) for 1 h, then cells were subjected to 6 h of OGD insult followed by 1 h of re- oXygenation. The changes of p-IRE1α and IRE1α were detected by Western blotting. (G) Densitometric quantification of p-IRE1α/IRE1α in (F) (n = 3). (H) HT-22 neuronal cells pretreated with NAC (2.5 mM) or vehicle were subjected to 6 h of OGD insult followed by 1 h of re-oXygenation. The variances of p-IRE1α and IRE1α were detected by Western blotting. (I) Densitometric quantification of p-IRE1α/IRE1α in (F) (n = 4). (J) HT-22 neuronal cells pretreated with NAC (2.5 mM) or vehicle were subjected to 6 h of OGD insult followed by 24 h of re-oXygenation. Cell viability was detected by CCK-8 assay (n = 3). Results are expressed as the mean ± SD, *p < 0.05, **p < 0.01 versus the indicated group further confirm the role of PDE4 in OGD/R-mediated ER stress, HT-22 neuronal cells were transfected with PDE4 siRNA and were undergone OGD/R insult. As shown in Fig. 6A, OGD/R induced the activation of IRE1α (p < 0.01), while PDE4 knockdown blocked the stimulatory ef- fects of OGD/R on the phosphorylation of IRE1α (p < 0.01, Fig. 6A and B). We also found the PDE4 siRNA significantly abolished the role of OGD/R on JNK activation (p < 0.01, Fig. 6C and D). To further confirm the role of PDE4 in the activation of IRE1α and JNK, we also performed a rescue experiment with the ectopic expression of PDE4 in the siRNA treated cells. As shown in Fig. 6E–H, overexpression of PDE4 restored the role of PDE4 siRNA on the expression of phosphorylated IRE1α (p < 0.01) and JNK (p < 0.01). These results further support the role of PDE4 in the activation of IRE1α and JNK under the condition of OGD/R. 3.7.Inhibition of PDE4 by Roflu protects primary cortical neurons against OGD/R insult The above results showed that Roflu attenuated ER stress and sup- pressed IRE1α/TRAF2/JNK signaling in HT-22 neuronal cells subjected to OGD/R insult. We then confirmed our results in primary cortical neurons. First, CCK-8 assay was applied to investigate the neuro- protective effect of Roflu. As shown in Fig. 7A, OGD/R insult caused a significant decrease in cell viability (p < 0.01), while treatment with Roflu (0.1–1 μM) increased cell viability in a dose-dependent manner. A Fig. 5. Inhibition of PDE4 by Roflu inhibits IREα/TRAF2/JNK signaling in HT-22 cells subjected to OGD/R insult. HT-22 neuronal cells pretreated with Roflu (1 μM) or vehicle were subjected to 6 h of OGD insult followed by 1 h of re-oXygenation. (A and B) The interaction between IREα and TRAF2 was detected by immuno- precipitation. (C and D) The relative value of IP: IRE1α/IP: TRAF2 (n = 3) and IP: TRAF2/IP: IRE1α (n = 3) were determined by densitometry of the blots. (E and F) HT-22 neuronal cells pretreated with Roflu (1 μM) or vehicle were subjected to 6 h of OGD insult followed by 6 h of re-oXygenation. The changes of p-JNK and JNK were determined by Western blotting. The relative level of p-JNK/JNK was determined by densitometry of the blots (n = 3). (G) The expression of p-JNK was visualized by immunofluorescence staining. Results are expressed as the mean ± SD, *p < 0.05, **p < 0.01 versus the indicated group protective effect of Roflu on the viability of primary cortical neurons was observed at 0.3–1 μM. Next, cleaved caspase 3 was chosen as a repre- sentative marker for cell apoptosis. The results of Western blotting showed that OGD/R increased the level of cleaved caspase 3 (p < 0.01), and treatment with Roflu reversed the increase in the level of cleaved caspase 3 (p < 0.01, Fig. 7B and C). Additionally, the results of cellular immunofluorescence also verified the protective effect of Roflu. The results of MAP2 staining shown that OGD/R caused a significant decrease in MAP2 fluorescence (p < 0.01), while treatment with Roflu ameliorated the damage caused by OGD/R (Fig. 7D–E, p < 0.01). Then, we verified the signaling pathway described in the above results. As expected, OGD/R caused a significant reduction in the level of phos- phorylated AKT (p < 0.01) and GSK3β (P < 0.01), while inhibition of PDE4 by Roflu normalized the levels of phosphorylated AKT (p < 0.01) and GSK3β (p < 0.05, Fig. 7F–I). Finally, we detected the phosphory- lation of IRE1α and JNK. As shown in Fig. 7J–M, OGD/R caused a sig- nificant increase in the levels of phosphorylated IRE1α (p < 0.01) and JNK (p < 0.01), and inhibition of PDE4 by Roflu reduced the over- activation of IRE1α (p < 0.05) and JNK (p < 0.01). These data were consistent with the findings observed in HT-22 cells. These results indicate that inhibition of PDE4 in primary cortical neurons exhibits a substantial neuroprotective effect, and this effect is potentially mediated by activation of the AKT/GSK3β signaling pathway, which results in a decrease in over-activation of IRE1α and JNK. Fig. 6. PDE4 knockdown inhibits the activation of IRE1α and JNK. HT-22 neuronal cells transfected with si-PDE4B or negative control (NC) were subjected to 6 h of OGD insult. (A and B) At 1 h after re-oXygenation, the levels of p-IRE1α and IRE1α were detected by Western blotting. The relative level of p-IRE1α/IRE1α was determined by densitometry of the blots (n = 4). (C) At 6 h after re-oXygenation, the levels of p-JNK and JNK were detected by Western blotting. (D) Densitometric quantification of p-JNK/JNK in (C) (n = 4). HT-22 neuronal cells transfected with PDE4B siRNA (si-PDE4B) and/or PDE4B plasmid were subjected to 6 h of OGD insult. (E and F) At 1 h after re-oXygenation, the levels of p-IRE1α and IRE1α were detected by Western blotting. The relative level of p-IRE1α/IRE1α was determined by densitometry of the blots (n = 3). (G and D) At 6 h after re-oXygenation, the levels of p-JNK and JNK were detected by Western blotting. The relative level of p- JNK/JNK was determined by densitometry of the blots (n = 3). Results are expressed as the mean ± SD, **p < 0.01 versus the indicated group. 3.8.Inhibition of PDE4 by Roflu protects SD rats against focal cerebral ischemia We show that Roflu protects against injury caused by ischemia/ reperfusion in vitro. Next, we examined the role of Roflu on neuro- protection in vivo. We established the focal cerebral ischemia model by induction of MCAO for 2 h in rats. The diagram of the in vivo experi- mental protocol is illustrated in Fig. 8A. The establishment of focal ce- rebral ischemia was verified by Laser Speckle Contrast Imaging (Fig. 8B). Roflu (0.3 and 1 mg/kg) was intraperitoneally injected 2 h after MCAO, and the doses were selected according to a previous study showing the cognition-enhancing effect of Roflu in mice [16]. At 24 h after reperfusion, the infarct volume was measured by TTC staining. As shown in Fig. 8C, MCAO for 2 h caused a significant increase in infarct volume, while administration of Roflu (0.3 and 1 mg/kg) after 2 h of MCAO decrease the infarct volume in a dose-dependent manner (Fig. 8C and D). Similarly, the administration of Roflu improved the neurological function of SD rats subjected to MCAO (Fig. 8E). Then, we detected the variance of the IRE1α/JNK signaling pathway at 6 h after MCAO in the penumbra of the ipsilateral side (Fig. 8F–H). In line with the in vitro results, ischemia/reperfusion increased the level of phosphorylation of IRE1α (p < 0.01) and JNK (p < 0.01). Consistently, administrationof Roflu (1 mg/kg) ameliorated the activation of IRE1α (p < 0.01) and JNK (p < 0.01, Fig. 8F–H). The immunofluorescence experiment was then conducted to detect the expression of phosphorylated IRE1α in neurons in the penumbra area. Immunofluorescence staining revealed increased IRE1α fluorescence in the penumbra area of the cortex in animals sub- jected to MCAO and reperfusion. Double staining of IRE1α and NeuN showed that a significant enhance in phosphorylated IRE1α in cortical neurons caused by MCAO (Fig. 8 I I and J. p < 0.01), and administration of Roflu (1 mg/kg) could reduce the overactivation of IRE1α in cortical neurons (p < 0.01, Fig. 8I and J). Taken together, these data confirm that Roflu has neuroprotective effects in injury caused by ischemia/reperfusion, and the protective effect is potentially mediated by inhibition of the IRE1α/JNK pathway. 4. Discussion The ER is involved in the maintenance of intracellular protein ho- meostasis, the post-translational modifications, and proper folding of proteins [31]. The shortage of energy during the process of cerebral ischemia triggers the accumulation of misfolded proteins in the ER lumen. And abnormal aggregates of misfolded protein eventually leads to activation of the ER stress [32]. IRE1α is a canonical ER-transmembrane effecter protein launching the activation of ER stress [33]. Our previous studies indicate that inhibition of PDE4 is beneficial to suppress the ER stress under the condition of cerebral ischemia [12]. However, no study has addressed the involvement of IRE1α in the inhibitory role of PDE4 inhibition in ER stress. Also, it is not well established which signaling pathways link PDE4 to IRE1α after transient brain ischemia. The present study showed that OGD/R caused apoptotic cell death, whereas the PDE4 inhibitor Roflu prevented cell death by the activation of the AKT/GSK3β signaling pathway and inhibition of IRE1α/JNK pathway in HT-22 neuronal cells and primary neurons. This conclusion is based on the following observations: (1) inhibition of PDE4 by Roflu reduces neuronal cell death caused by OGD/R insult and pro- motes the activation of AKT/GSK3β; (2) constitutively active GSK3β (S9A) mutation decreases cell viability, abolishes the inhibitory effect of Roflu on oXidative stress, and reversed the effects of Roflu on IRE1α activation; (3) Roflu attenuated ER stress, disrupted the interaction between IRE1α and TRAF2, and attenuated the activation of JNK induced by OGD/R insult; (4) PDE4 knockdown attenuates the phos- phorylation of IRE1α and JNK, while overexpression of PDE4 reversed the role of PDE4 siRNA on the expression of phosphorylated IRE1α; (5) inhibition of PDE4 by Roflu showed similar protective effects in primary cultured cortical neurons exposed to OGD/R injury, and (6) Fig. 7. Inhibition of PDE4 by Roflu protects primary cortical neurons against OGD/R insult. (A) Primary cortical neurons pretreated with the indicated concentration of Roflu were subjected to 1 h of OGD followed by 24 h of re-oXygenation. Then, cell viability was detected by CCK-8 assay (n = 5). Primary cortical neurons pretreated with Roflu (1 μM) were subjected to 1 h of OGD followed by 6 h of re-oXygenation. (B and C) The variances of cleaved caspase 3, and caspase 3 were detected by Western blotting. The relative level of cleaved caspase 3/caspase 3 was determined by densitometry of the blots (n = 3). (D and E) The change of synaptic morphology was visualized by immunofluorescence chemistry and quantified with Image J (n = 4). (F–K) Primary cortical neurons pretreated with Roflu (1 μM) were subjected to 1 h of OGD followed by 1 h of re-oXygenation. The variances of p-AKT, AKT, p-GSK3β, GSK3β, p-IRE1α, and IRE1α were detected by Western blotting. The relative levels of p-AKT/AKT (n = 3), p-GSK3β/GSK3β (n = 3), and p-IRE1α/IRE1α (n = 4) were determined by densitometry of the blots. (L and M) Primary cortical neurons pretreated with Roflu (1 μM) were subjected to 1 h of OGD followed by 6 h of re-oXygenation. The variances of p-JNK and JNK were detected by Western blotting. The relative level of p-JNK/JNK was determined by densitometry of the blots (n = 3). Results are expressed as the mean ± SD, *p < 0.05, **p < 0.01 versus the indicated group administration of Roflu ameliorated injury in rats subjected to MCAO/R injury. Our findings suggest a novel potential clinical application of Roflu in the central nervous disease. Moreover, PDE4 and its down- stream molecules represent new therapeutic targets in the treatment of ischemia and reperfusion injury. Summarizing all the analysis above, a mechanical model for the hypothesis is illustrated in Fig. 9. Nrf-2 is an endogenous antioXidant factor that could inhibit the occurrence of ER stress after cerebral ischemia [12]. Fig. 8. Inhibition of PDE4 by Roflu protects SD rats against focal cerebral ischemia. (A) The diagram of the in vivo experimental protocol. (B) SD rats were subjected to 2 h of focal cerebral ischemia induced by MCAO. The change of cerebral blood flood was monitored by Laser Speckle Contrast Imaging. (C─E) SD rats were subjected to 2 h of focal cerebral ischemia induced by MCAO. Then, rats were treated with indicated concentration Roflu or vehicle. At 24 h after reperfusion, the infarct volume was detected by TTC staining and neurological deficit scores were recorded in a double-blind manner (Sham group: n = 10; vehicle group: n = 9; 0.3 mg/kg group: n = 10; 1 mg/kg group: n = 10). (F─H) SD rats were subjected to 2 h of focal cerebral ischemia caused by MCAO. Then, rats were treated with Roflu (1 mg/kg) or vehicle. At 6 h after reperfusion, the penumbra of the ipsilateral brain was collected, and levels of p-IRE1α, IRE1α, p-JNK, and JNK were detected by Western blotting. The relative levels of p-IRE1α/IRE1α and p-JNK/JNK were determined by densitometry of the blots (n = 4). (I─J) SD rats were subjected to 2 h of focal cerebral ischemia caused by MCAO. Then, rats were treated with Roflu or a vehicle. At 6 h after reperfusion, the brains were fiXed in 4% PFA followed by dehydration. Finally, the variance of p-IRE1α inside neurons of the penumbra was visualized by Immunofluorescence staining and quantified with Image J (n = 4). Results are expressed as the mean ± SD, **p < 0.01 versus the indicated group. n represents the number of rats activated by phosphorylation of Try216 reside and inactivated by phosphorylation of Ser9 residue [39]. In the present study, we mainly focused on the phosphorylation of GSK3β at Ser9. Whether Roflu exerts a protective effect against ischemic stroke-induced neuronal damage through promoting the phosphorylation of GSK3β at Tyr216 deserves further study. Additionally, point mutation at Tyr216 (such as Y216F) should be took into consideration when investigating the essential involvement of phosphorylation of GSK3β at Tyr126.ER stress is one of the key factors leading to neuronal damage, and it is also considered as a potential target for intervention in cerebral ischemic injury [14]. Various stimuli, such as over-accumulation of misfolded proteins and excessive production of ROS caused by ischemia-reperfusion, trigger ER stress and enhance cellular apoptosis [20]. IRE1α is the most conserved signal molecule during the process of ER stress [21]. Fig. 9. Roflu protects neurons against ER stress-induced apoptosis via activa- tion of the AKT/GSK3β/Nrf-2 signaling pathway. EXcessive ROS are generated during cerebral ischemia/reperfusion. Overproduced oXidative stress triggers ER stress, and activates IRE1α/TRAF2/JNK pathway. Roflu enhances the nu- clear translocation of Nrf-2 and reduces oXidative stress via activating the cAMP/AKT/GSK3β pathway. Reduced ROS attenuates ER stress response. Attenuation of ER stress by Roflu reduces the binding of IRE1α to TRAF2, thereby reducing the phosphorylation of JNK, and the apoptosis of neurons exposed to ischemia/reperfusion auto-phosphorylation and then binds to TRAF2, which subsequently activates JNK [21,40]. It is well known that over-activation of JNK causes neuronal apoptosis in many neurological diseases, including ischemic stroke. In the present study, we found that Roflu treatment reduced the production of ROS and ER stress, as evidenced by decreased levels of GRP78 and phosphorylated IRE1α. Interestingly, GSK3β (S9A) mutation abolished the inhibitory effect of Roflu on the production of ROS and ER stress, while the ROS scavenger, NAC, effectively sup- pressed the phosphorylation of IRE1α and enhanced cell viability. These findings indicate that GSK3β and ROS production are the upstream of oXidative stress, Nrf-2 is uncoupled from Kelch-like oXy-events of ER stress. In other words, Roflu inactivates GSK3β and reduced chloropropane-related protein 1 (Keap1) [34]. Activated Nrf-2 trans- locates into the nucleus to initiate transcription of downstream genes (such as heme oXygenase 1 and catalase), and thus play a protective role [35]. Inhibiting Keap1 was once considered a very potent strategy to activate Nrf-2 [36]. However, it was later discovered that direct inhi- bition of Keap1 leads to serious cardiovascular events [37]. Therefore, the non-Keap1-dependent Nrf-2 activation pathway has attracted widespread attention. Inhibition of PDE4 in neurons has a protective effect on ischemic brain tissues, but the mechanism is still unclear. Our previous studies showed that inhibition or knockdown of PDE4 sup- presses ER stress and promotes neuronal survival [12]. Simultaneously, we observed that inhibition of PDE4 increased the level of Nrf-2 in the nucleus, while inhibition of Nrf-2 blocked the protective effects of PDE4 inhibition on ER stress [12]. Thus, we are eager to explore how PDE4 inhibition affects Nrf-2 and ER stress. It has recently been discovered that in addition to Keap1, GSK3β activation promotes the degradation of Nrf-2 [38]. Phosphorylation of GSK3β promotes Nrf-2 nuclear accumu- lation and the expression of the antioXidant enzyme [38]. And we pre- viously found that PDE4 inhibition could activate the cAMP/Epac/Akt/GSK3β pathway [6]. In the present study, we found that Roflu activated AKT and inhibited the activation of GSK3β. Importantly, constitutively activation of GSK3β blocked the pro-survival role of Roflu and reduced the level of Nrf-2 in the nucleus. Therefore, we proposed that PDE4 inhibition by Roflu reduced the degradation of Nrf-2 through inactivating GSK3β. We then found that GSK3β (S9A) blocked the inhibitory role of Roflu on the production of ROS and the levels of catalase and H2O2, which are the canonical target of Nrf-2. Hence, we also increased Nrf-2 in the nucleus up-regulates antioXidant genes, and reduces the production of ROS. We would like to point out that GSK3β is the level of ROS, which in turn suppresses ER stress. Studies have identified a pro-apoptotic IRE1α/TRAF2/JNK pathway that can be activated by prolonged ER stress [41]. In the present study, we found that OGD/R promotes the binding of IRE1α to TRAF2, while Roflu effectively reduced the formation of IRE1α-TRAF2 complex. We also found that Roflu treatment decreased the phosphorylation/activation of JNK. Furthermore, we determined that the PDE4 knockdown reduced the activation of both IRE1α and JNK. Our previous evidence has demonstrated that PDE4 knockdown ameliorates OGD-triggered ER stress in neuronal cells, and thus antagonizing cellular apoptosis. Our current findings are consistent with these reports. Besides IRE1α, protein kinase R-like ER kinase and activating transcription factor 6 can activate ER stress as well. In this study, we found that Roflu inhibited the phosphorylation of IRE1α and its downstream molecules TRAF2 and JNK. We would like to point out that the roles of Roflu on the phos- phorylation and expression of protein kinase R-like ER kinase and activating transcription factor 6 are not investigated in the present study. It is desirable in the future to test the roles of Roflu on the ER homeostasis systematically. To verify that the neuroprotective effect of Roflu is not limited to HT-22 cells, we also performed primary cortical neuron cultures and confirmed its neuroprotective effects against OGD/R-induced injury. We found that Roflu conferred neuroprotection of cultured cortical neurons towards OGD/R-induced insult in a concentration-dependent manner, consistent with its potent effect in the HT-22 cell model. Most impor- tantly, all these in vitro data are consistent with the results obtained from the MCAO animal model. The MCAO model is a widely accepted animal model mimicking human ischemic stroke. The advantages of this model have been illustrated by several reviews and research articles [42–44].We found that Roflu effectively reduced infarct volume and enhanced neurological scores in rats subjected to MCAO and reperfusion. We also found that Roflu suppressed ER stress in this animal model. Recently, researchers recommended the permanent MCAO model in the preclini- cal study, as this model mimics large vessel occlusion patients without reperfusion, and 88.7–97.5% of the large vessel occlusion patients do not have vessel recanalization [44]. Hence, the permanent MCAO model represents the majority of patients with large vessel occlusion [44]. In the present study, we mainly focused on the protective role of Roflu in the transient MCAO and reperfusion animal model. Whether Roflu would produce a similar effect in a permanent MCAO model deserves further investigation. 5.Conclusion In summary, our study is the first to report the protective effects of Roflu against the neuronal injury induced by cerebral ischemia. We also explored the possibly involved signaling pathways and identified that the activation of AKT/GSK3β signaling and inhibition of ER stress- induced IRE1α/TRAF2/JNK inhibition were involved in the neuro- protection of Roflu. Roflu is an FDA-approved drug for the treatment of COPD. The present study suggests a novel potential clinical application of Roflu. Besides, the development of therapeutic strategies to specif- ically target the PDE4-mediated signaling pathway is supposed BI 1015550 to be beneficial for the treatment of cerebral ischemia.