Blood–Brain Barrier Rescue by Roflumilast After Transient Global Cerebral Ischemia in Rats

Phosphodiesterase 4 inhibitors (PDE4-I), which selectively increase cyclic adenosine monophosphate (cAMP) levels, have shown neuroprotective effects after several neurological injuries inducing blood–brain barrier (BBB) damage including local/focal cerebral ischemia. The present investigated whether roflumilast confers BBB neuroprotection in the hippocampus after transient global cerebral ischemia (TGCI) in rats. TGCI resulted in whole BBB disruption as measured by the increase of Evans blue (EB) and IgG extravasation, neurodegeneration, and downregulation of claudin-5 and endothelial nitric oxide synthase (eNOS) levels in the CA1 hippocampal subfield of ischemic rats. Roflumilast attenuated BBB disruption and restored the levels of eNOS in the CA1 hippocampal area. Moreover, roflumilast increased the levels of B2 cell lymphoma (BcL-2) and neuron-glial antigen-2 (NG2) in the CA1 subfield after global ischemia in rats. The protective effects of roflumilast against TGCI-induced BBB breakdown might involve preservation of BBB integrity, vascularization and angiogenesis, and myelin repair.


Introduction
Transient global cerebral ischemia (TGCI) may occur in a multitude of clinical settings, including cardiac arrest (Frisch et al. 2017), severe hypotension and shock (Wen et al. 2018), craniocerebral trauma (Veenith et al. 2016), and serious cerebrovascular events (Johnston et al. 2019). Damage that is caused by TGCI can result in extensive neurodegeneration in vulnerable brain areas, such as CA1 pyramidal neurons of the hippocampus, in humans (Petito et al. 1987;Stamenova et al. 2018;Haglund et al. 2019) and animals (Kirino 1982;Pulsinelli et al. 1982). Hippocampal CA1 neurodegeneration results in long-lasting cognitive and emotional impairments (Langdon et al. 2008;Geri et al. 2014) that negatively affect patients' quality of life (Moulaert et al. 2010).
Excessive glutamate release, oxidative stress, neuroinflammation, blood-brain barrier (BBB) disruption, and microcirculation dysfunction play important roles in the pathogenesis of TGCI (Ju et al. 2018;Guo et al. 2021). Glutamate release causes postsynaptic calcium influx via N-methyl-D-aspartate (NMDA) receptors, triggering nitric oxide synthase (NOS) activation and nitric oxide (NO) production (Garthwaite et al. 1991). Damage to the BBB increases endothelial cell permeability, leading to an increase in macromolecule and serum protein transport (e.g., immunoglobulin G (IgG) extravasation). After TGCI, significant microvascular changes occur, characterized by the downregulation of BBB tight junction proteins (e.g., occludin, claudin-3, claudin-5, and zonula occludens-1; Anuncibay-Soto et al. 2018;Jin et al. 2019) and a decrease in 1 3 endothelial NOS (eNOS) expression (Mun et al. 2010;Yang et al. 2019). In the BBB, claudins serve two main purposes: channel forming and sealing. In particular, claudin-5 is the predominant claudin expressed in endothelial tight junction and has a pivotal role in the maintenance of BBB (Scalise et al. 2021). TGCI also damages white matter affecting myelinated axons and oligodendrocytes precursor cells (OPCs) (Kubo et al. 2009;Yoshioka et al. 2011). All these events contribute to the worsening of neuroinflammation, brain edema formation, and poor clinical prognosis (Lorberboym et al. 2003;Brouns et al. 2011). Therefore, preventing BBB disruption may be a new avenue for treating TGCI-related injuries.
Phosphodiesterase-4 (PDE4) inhibitors selectively increase the cyclic adenosine monophosphate (cAMP)/cAMP response element binding protein (CREB) signaling pathway and have been shown to have beneficial effects against BBB damage in animals with cerebral ischemia. Belayev et al. (1998) reported that the prototype PDE4 inhibitor BBB022 (rolipram) protected BBB integrity against focal ischemia that was induced by transient middle cerebral artery occlusion (MCAo) in rats. Rolipram also decreased cerebral edema, reduced neuroinflammation, and preserved the expression of occludin and claudin-5 in the brain in MCAo mice (Kraft et al. 2013). Recently Cai et al. (2022) have found that the selective PDE4 inhibitor roflumilast improved neurobehavioral outcomes and reversed BBB disruption following MCAo in rats. These findings indicate that roflumilast may exert a protective effect against BBB injury after focal ischemia. Remaining to be determined, however, are the effects of roflumilast on BBB damage that is caused by TGCI.
The present study investigated whether roflumilast confers BBB neuroprotection after TGCI in rats. Blood-brain barrier permeability was evaluated by Evans blue (EB) staining and IgG extravasation. Markers of BBB integrity were measured in the CA1 subfield of the hippocampus, a brain region that is highly vulnerable to TGCI. Immunohistochemistry and western blot were performed to detect the expression of tight junction proteins (claudin-3 and claudin-5) and neurodegeneration, reflected by neuronal nuclei (NeuN) and B2 cell lymphoma (BcL-2) expression. FluroJade C was employed to detect degenerating neurons. Endothelial nitric oxide synthase and NG2 cells are involved in BBB protection after cerebral damage; thus, their expression was also investigated in ischemic animals after roflumilast treatment.

Animals
Male Wistar rats (90-100 days old) were acquired from the local vivarium of the State University of Maringá (Paraná, Brazil). The animals were housed in groups of 3 animals per cage in a temperature-controlled room (22 °C ± 1 °C) with a normal 12-h/12-h light/dark cycle (lights on at 7:00 a.m., lights off at 7:00 p.m.). The animals had free access to tap water and a standard commercial chow diet (Nutrilab-CR1; Nuvital Nutrients, Curitiba, PR, Brazil) "ad libitum" during the experiments. The experimental procedures conformed to the ethics of this article and are protected by copyright. All rights reserved principles of the Brazilian College of Animal Experimentation (COBEA) were approved by the local Ethics Committee on Animal Experimentation of the State University of Maringá (CEUA No. 7995230420). All efforts were made to minimize the number of animals used and reduce their suffering.

Transient Global Cerebral Ischemia
Transient global cerebral ischemia was induced using the four-vessel occlusion (4-VO) model (Bacarin et al. 2013). The animals were first anesthetized in a chamber that contained halothane (Isoforine, Cristália, SP, Brazil), and the dorsal and ventral regions of the neck were trichotomized. The rats were then fixed in a stereotaxic frame, and a halothane/oxygen mixture was delivered (0.5 L/min) through a face mask that was affixed to the rat's snout. After bilateral exposure of the alar foramen of the first cervical vertebrae, the vertebral arteries were permanently electrocoagulation (unipolar current, 3-4 mA). The common carotid arteries were exposed through an incision on the ventral neck and loosely tied with silk thread. Four to 5 h later, when the animals completely recovered from anesthesia, the silk thread was carefully tightened for 15 min. During this time, such signs as loss of the righting reflex, mydriasis, tonic stretching of the paws, and the absence of responses to touch were considered indicative of effective ischemia. After reperfusion, the animals were maintained in a warming box (37 °C ± 1 °C) for 1 h to avoid ischemia-induced cerebral hypothermia. Signs such as loss of righting reflex, mydriasis, tonic stretching of the paws, and absence of touch responses were considered indicative of effective ischemia. Sham animals underwent the same surgical procedure but without occlusion of the arteries.

Drugs
Vehicle (0.5% carboxymethylcellulose) or roflumilast (provided by Dr. Jos Prickaerts, Maastricht University, The Netherlands) was dissolved in the vehicle and 2% Tween 80 and administered intraperitoneally (i.p.) once daily until of the sacrifice. The roflumilast dose (0.003 mg/kg) was based on previous studies showing functional neuroprotection in rats with TGCI (Bonato et al. 2021) or chronic cerebral hypoperfusion (Santiago et al. 2018).

Experiment 1
To assess the time course of BBB disruption after TGCI, the rats were submitted to sham or TGCI surgeries. One hour after reperfusion, the vehicle was administered i.p. 24 h, 72 h, or for 7 days, once/day (Fig. 1A). The animals' brains were removed and processed for spectrophotometric analysis to evaluate EB extravasation.

Determination of BBB Leakage
BBB permeability was measured by EB dye extravasation as previously reported (Wu et al. 2017;Jin et al. 2019), with slight modifications. After isoflurane inhalation anesthesia (Biochimico, Rio de Janeiro, RJ, Brazil), the rats received 2% EB (Sigma-Aldrich, Steinheim, Germany; 4 ml/kg) by intrapenile injection. Three hours later, the animals were anesthetized with an overdose of thiopental (50 mg/kg; Thiopentax, Cristália, SP, Brazil). The animals underwent transcardiac perfusion with 110 mL of 0.01 M phosphatebuffered saline (PBS) (5 min; 22 mL/min) to remove the remaining intravascular EB dye. After, the whole brain (except the cerebellum and brain stem) was weighed and homogenized with an equal volume of trichloroacetic acid in ethanol (1:3) and incubated for 24 h. The material was then centrifuged at 12,000 rpm for 10 min, and the supernatant was separated. The supernatant and the standards (0, 0.08, 0.16, 0.32 0.65, 1.25, 2.50, and 5.00 µg/mL) were detected at 620 nm by an Asys Expert Plus (Biochrom, Berlin, Germany) microplate reader. The content of each group was calculated according to the standard curve and represented as EB extravasation (µg/g brain tissue).

Immunohistochemistry and Immunofluorescence
Five to six animals from each experimental group were anesthetized with an overdose of sodium thiopental (50 mg/kg; Thiopentax, Cristália, SP, Brazil) for transcardiac perfusion with 88 mL of 0.01 M PBS (4 min; 22 mL/min), followed by 132 mL of 4% paraformaldehyde (PFA) (pH = 7.4) in 0.2 M phosphate buffer (PB) (6 min; 22 mL/min). The brains were removed and post-fixed in a 4% PFA solution in 0.2 M PB for 24 h. After this period, the brains were kept in a 30% sucrose solution in 0.01 M PBS at 4 °C for cryoprotection for 120 h. After snap frozen, the brains were sectioned using a cryostat (Criocut 1800, Reichert-Jung, Heidelberg, Germany). Thirty 30 µm coronal sections of the hippocampus were collected in sixth replicate tubes containing PBS plus 0.1% sodium azide. The sections corresponded to the stereotaxic coordinates (− 3.60 to − 4.52 mm from Bregma), according to the atlas of Paxinos and Watson (1997).
For IgG detection, sections were washed in 0.01 M PBS and then incubated in 0.1 M PBS containing 3% H 2 O 2 (Merck, Darmstadt, Germany). After four washes with 0.01 M PBS, the sections were incubated in a solution containing 2% bovine serum albumin (BSA; Sigma-Aldrich, Steinheim, Germany) in 0.01 M PBS, plus 0.3% Triton-X 100 (PBS + ; Sigma-Aldrich, Steinheim, Germany) for 1 h to block nonspecific reactions. Sections were subsequently Fig. 1 Experimental design. A Experiment 1: Determining the BBB permeability over time in rats with TGCI. Rats underwent sham or transient global cerebral ischemia (TGCI) surgeries. Animals were sacrificed at 24 h, 72 h, or 7 days after reperfusion. Three hours before sacrifice, the animals received an intravenous injection (i.v.) of Evans blue (EB). The rats had their brains removed and the extravasation of EB was determined. B Experiment 2: Effects of roflumilast in BBB permeability. One hour after reperfusion, animals received vehicle or roflumilast (0.003 mg/kg) intraperitoneal (i.p.) and then once a day for 3 days. Three hours before sacrifice, the animals received EB i.v.. Seventy-two hours after reperfusion, the rats were sacrificed, and their brains were processed for molecular and histological analysis incubated with rabbit anti-IgG antibody (1:200, Cat# BA-4000, RRID: AB_2336206, Vector Laboratories, California, USA) (Table 1) in PBS + solution. The sections were kept overnight under constant stirring at 4 °C. The sections were washed and then incubated with the secondary biotinylated mouse anti-rabbit antibody (1:500, Cat# sc-2491, Santa Cruz Biotechnology, California, USA) and kept under constant stirring at room temperature for 2 h. Then, they were incubated with the avidin-biotin complex (ABC; Kit, Vector Laboratories, California, USA), at room temperature, for another 2 h. For the development, a 0.025% solution of 3,3,9-diaminobenzidine hydrochloride (DAB; Sigma-Aldrich, Steinheim, Germany) 0.05% H2O2 was used. The sections were properly washed in 0.01 M PBS and mounted on previously gelatinized slides. Dehydration and diaphanization processes with xylene were performed previously to cover the slides with Permount ® (Fisher Scientific, New Jersey, USA) and coverslips.
Quantitative analyses were conducted under blinded conditions. Analysis was performed using an Olympus BX41 microscope (Tokyo, Japan) coupled with a high-performance device color camera (QColor3, Ontario, Canada) or fluorescence microscope (Leica DM 2500, Wetzlar, Germany) coupled with a camera (Leica DF345 FX, Wetzlar, Germany) when appropriate. Lighting conditions and magnifications were kept constant during image capture to avoid signal saturation in immunohistochemistry. To avoid bias in the immunofluorescence image analysis, quantification was prevented by keeping constant the detector gain and pixel dwell. The images were quantified in pre-fixed digital microscopic areas of CA1 (NeuN, NG2; 0.32 mm 2 ). IgG and claudin-5 expression were evaluated in the total area of the hippocampus. ImageJ software (National Institutes of Health, Bethesda, MD, USA) was used to obtain the integrated optical density (IOD) (Haley and Lawrence 2017;Kho et al. 2018). The selected images were converted to 16-bit grayscale and the background was subtracted. Results are expressed as the mean ± SEM of the IOD/area.

Fluoro-Jade C Staining
FJC staining is an established detection technique for degenerating neurons (Schmued and Hopkins 2000). FJC (TR-100-FJ, biosensis, SA, Australia) staining was performed on sections adjacent to those used for immunohistochemistry. The gelatinized slides containing sections were immersed in a basic alcohol solution consisting of sodium hydroxide in 80% ethanol for 5 min. Next, they were rinsed with 70% ethanol followed by distilled water. The slides were then incubated in potassium permanganate solution for 10 min. After washing in distilled water, the slides were transferred to a solution of FJC and kept in the dark for 40 min. Thereafter, slides were washed in distilled water, air dried, and coverslipped with DPX (Sigma-Aldrich, St. Louis, United States). Quantitative analyses were conducted under blinded conditions. Analysis was performed using a fluorescence microscope (Leica DM 2500, Wetzlar, Germany) coupled with a camera (Leica DF345 FX, Wetzlar, Germany). The images were quantified in pre-fixed digital microscopic areas of CA1 (0.32 mm 2 ). ImageJ software (National Institutes of Health, Bethesda, MD, USA) was used to obtain the IOD (Haley and Lawrence 2017;Kho et al. 2018). The selected images were converted to 16-bit grayscale and the background was subtracted. Results are expressed as the mean ± SEM of the IOD/area.

Western Blot
For Western blotting analysis, 5-6 animals from each experimental group were randomly taken and anesthetized with isoflurane (Biochimico, Rio de Janeiro, RJ, Brazil). The CA1 subfield of the hippocampus was carefully removed and dissected using spatulas, tweezers, and a . All blots were stripped with harsh stripping buffer (20% SDS 10%, 12.5% Tris HCl 0.5 M, and 0.8% ß-mercaptoethanol in H2O), to assess the protein control glyceraldehyde-3-phosphate dehydrogenase (GAPDH). The revelation was performed with an ECLplus ® chemiluminescence kit (Invitrogen, Carlsbad, CA, USA), and the bands were visualized with the aid of Chemi-doc (Bio-Rad Laboratories Inc., Hercules, USA). Specific band intensities were quantified using ImageJ (NIH, Bethesda, MD, USA) and normalized to GAPDH protein levels. Results are expressed as the mean ± SEM of the protein level.

Statistical Analysis
SAS 9.4 software (Institute Inc, Cary, NY, USA) was used for statistical analysis. Data were analyzed for assumptions of a normal Gaussian distribution (Shapiro-Wilk test) and homogeneity of variances (Bartlett's test). As part of the data did not fit the normal distribution and/or homoscedasticity assumptions, the generalized linear model (GLM) followed by a gamma distribution was used. Data are expressed as means ± SEM, and differences were considered significant at p < 0.05.

TGCI Causes BBB Leakage 72 h After Reperfusion Which Is Attenuated by Roflumilast
As shown in Fig. 2A, a significant difference was found in the determination of EB extravasation over time (χ 2 = 32.31; p = 0.001) among groups. The Isch + veh72h group presented an increase in the EB extravasation when compared to sham + veh and isch + veh24h groups (p < 0.001), indicating a loss of BBB integrity 72 h after reperfusion. Previous treatment with roflumilast 0.003 mg/Kg prevented the EB extravasation (Fig. 2B, χ 2 = 13.53; p = 0.001) when compared to the isch + veh group (p = 0.008). In Fig. 2D, E, there was a difference in the IgG expression in the hippocampus (χ 2 = 15.93; p = 0.001) and in the CA1 subfield of the hippocampus (χ 2 = 19.25; p = 0.001). Ischemic groups treated with vehicle or roflumilast showed an increase in IgG expression in both the total hippocampus and CA1 subfield when compared to sham animals (p = 0.001).
The levels of eNOS (Fig. 3E) presented significant difference among groups (χ 2 = 7.56; p = 0.02). Further analysis revealed that the levels of eNOS decreased in the isch + veh group compared with the sham + veh group (p = 0.005). Roflumilast reversed this ischemia-induced eNOS decrease compared with the isch + veh group (p = 0.05).

Roflumilast Fails to Prevent Hippocampal Neuronal Loss but Enhances BcL-2 Levels in the CA1 Hippocampal Subfield
No significant differences in the NeuN expression were observed among groups (Fig. 4C, χ 2 = 2.24; p = 0.33). A significant difference in Fluro-Jade expression was found among sham and ischemic groups (Fig. 4D, χ 2 = 20.47; p = 0.0001). Ischemic animals treated with vehicle (p = 0.006) or roflumilast (p = 0.01) increased the Fluoro-Jade positive cells compared to the sham + group.
However, ischemic rats that were treated with roflumilast exhibited a significant increase in BcL-2 expression compared with the sham + veh (p = 0.03) and isch + veh (p = 0.01) groups.

Fig. 2 TGCI causes BBB leakage 72 h after reperfusion which was attenuated by roflumilast.
A Evans blue (EB) extravasation (µg/g) of brain tissue as a function of reperfusion time (n = 8-10/group). B EB extravasation (µg/g) of brain tissue 72 h after reperfusion. C Representative photomicrographs of IgG extravasation in the hippocampus (above) and CA1subfiled in the hippocampus (below). D Integrated optical density (IOD) values of IgG in the hippocampus. E IOD values of IgG in the CA1 subfield of the hippocampus (n = 5-6/ group). Data are shown the means ± SEM (columns and bars) of the experimental groups. **p < 0.001 compared to sham + veh; #p < 0.05, ##p < 0.001 compared to isch + veh 24 h or isch + veh (generalized linear model followed by gamma distribution)

Roflumilast Increases NG2 Expression 72 h After Reperfusion
A significant difference was found in NG2 positive cells analysis among groups (χ 2 = 24.06; p = 0.0001) as shown in Fig. 5 in the CA1 subfield of the hippocampus. The Isch + veh group increased the NG2-positive cells when compared to the sham group (p = 0.0001). This effect of ischemia further increased after roflumilast 0.003 mg/Kg when compared to isch + veh (p = 0.05) and sham + veh groups (p = 0.0001).

Discussion
The present study investigated the effects of post-treatment with the PDE4 inhibitor roflumilast on BBB integrity in rats that were subjected to TGCI. Seventy-two hours after TGCI, we detected the disruption of BBB, degenerating neurons (FJC), and the downregulation of claudin-5 and eNOS levels in the CA1 hippocampal area. Roflumilast attenuated BBB disruption in the brain (reflected by EB extravasation) in ischemic animals but did not confer selective protection in the CA1 hippocampal area (reflected by IgG immunoreactivity). Roflumilast restored to control levels the expression of claudin-5 and eNOS and increased BcL-2 and NG2 levels in the CA1 hippocampal area in TGCI rats. These findings support a role of roflumilast in protecting the BBB against damage that is caused by global cerebral ischemia.
Blood-brain barrier injury is characterized by an increase in permeability and the degradation of structural proteins (Prakash and Carmichael 2015). We first examined the time course of EB extravasation in the whole brain after TCGI. From 24 h to 7 days, the most significant BBB disruption Fig. 3 Roflumilast impacts claudin-3 and claudin-5 expression and enhances eNOS expression in the CA1 hippocampal subfield after reperfusion. A Schematic representation of blood-brain barrier (BBB) structure (created by the author using Biorender). B Representative photomicrographs of claudin-5 in the CA1 subfield of the hippocampus. C Integrated optical density (IOD) of claudin-5 in CA1subfiled of the hippocampus. D Representative blots and their respective graphs of claudin-3 in CA1 of the hippocampus. E Representative blots and their respective graphs of endothelial nitric oxide synthase (eNOS). Data are shown in the means ± SEM (columns and bars) of the experimental groups (n = 5-6/group). *p < 0.05, ***p < 0.0001 compared to sham + veh; #p < 0.05, ###p < 0.0001 compared to isch + veh (generalized linear model followed by gamma distribution) occurred at 72 h with recovery 7 days after reperfusion. These results align with other studies that reported significant EB extravasation at 24-72 h in the brain in rats that were subjected to 4-VO ) and mice that were subjected to MCAo (Kraft et al. 2013;Bieber et al. 2021). The BBB breakdown after an ischemic insult is a dynamic process. Animal models of focal ischemia have shown that BBB breakdown may follow a biphasic course after reperfusion (Belayev et al. 1996;Huang et al. 1999;Jiang et al. 2018). An early (~ 6 h) increase in permeability is followed by a refractory period during which BBB permeability returns to baseline and a delayed second increase in permeability (Carmichael 2016;Bernardo-Castro et al. 2020). In the chronic stage, an increase of BBB restoration factors leads the barrier to start decreasing its permeability (Bernardo-Castro et al. 2020). However, several studies have demonstrated that BBB permeability remains increased up to weeks after stroke (Strbian 2008). Whether a biphasic pattern of BBB breakdown occurs in animal models of global ischemia requires further investigation.
Significant IgG extravasation was observed in the hippocampus in ischemic animals, particularly in the CA1 hippocampal area. Roflumilast attenuated EB extravasation in Data are shown in the means ± SEM (columns and bars) of the experimental groups (n = 5-6/group). *p < 0.05 compared to sham + veh; #p < 0.05 compared to isch + veh (generalized linear model followed by gamma distribution)

Fig. 5 Roflumilast increases NG2 expression 72 h after reperfusion.
A Diagram illustrating a coronal brain section at the intermediate level of the hippocampus showing the CA1 subfield where the analysis was performed. B Representative photomicrographs of neuronglia antigen-2 (NG2) expression in the CA1 subfield of the hippocam-pus. C Integrated optical density (IOD) values of NG2 expression in the CA1 of the hippocampus. Data are shown in the means ± SEM (columns and bars) of the experimental groups (n = 5-6/group). ***p < 0.0001 compared to sham + veh; #p < 0.05 compared to isch + veh (generalized linear model followed by gamma distribution) 1 3 the whole brain in TGCI rats. Our data extend previous findings that PDE4 inhibitors, such as rolipram and roflumilast, decreased EB extravasation in the brain of MCAo mice (Kraft et al. 2013) and in rats that were subjected to subarachnoid hemorrhage (Wu et al. 2017). Interestingly, no effect of roflumilast on IgG extravasation was detected in the CA1 hippocampal area in TGCI rats. The reason for the selectivity of roflumilast in protecting BBB integrity in the CA1 hippocampal subfield is unclear. The present findings suggest regional differences in distinct brain areas in response to BBB breakdown and neuroprotection after cerebral ischemia. Further studies are necessary to elucidate this effect.
Tight-junction BBB proteins, including claudin-5 and claudin-3, contribute to BBB integrity and function (Abdullahi et al. 2018;Winkler et al. 2021). The highest level of expression is found for claudin-5, which is responsible for tightening the BBB for small molecules (Wrinkler et al. 2021). Decreases in claudin-5 and claudin-3 expression were found in the brain in rodents that were subjected to global (Jin et al. 2019) and focal (Kraft et al. 2013;Li and Tian 2021) cerebral ischemia. We detected a decrease in claudin-5 expression in the CA1 hippocampal subfield after TGCI in rats. Claudin-3 levels, however, were not impacted 72 h after TGCI. The present results suggest a preferential impact of claudin-5 in the hippocampus in rats that are subjected to TGCI. Nevertheless, treatment with roflumilast increased claudin-5 levels in the CA1 hippocampal subfield. The mechanisms by which the PDE4 inhibitor roflumilast increased claudin-5 levels may involve the upregulation of cAMP pathways. Indeed, claudin-5 levels were upregulated in human brain microvascular endothelial cells that were incubated with forskolin and rolipram, two agents that are known to stimulate cAMP pathways (Yang et al. 2011). Moreover, the cAMP-dependent protein kinase inhibitor H89 reversed the effects of roflumilast on the MCAo-induced loss of ZO-1, cadherin, and occludin in mice (Cai et al. 2022). PDE4 is localized in hippocampal neurons, with specificity for cAMP (Lakics et al. 2010). In a previous work, we have demonstrated that roflumilast produced an increase in the phosphorylated CREB (pCTREB) expression in the CA1, CA3, and dentate gyrus 21 days after the TGCI (Bonato et al. 2021). This effect was related to neuroprotective effects of PDE-4 inhibition. It remains to be confirmed whether the protective effects of roflumilast on BBB breakdown at 72 h after TGCI are also associated with an increased CREB expression.
Nitric oxide plays a dual role with both protective and deleterious effects on brain tissue during cerebral ischemia, depending on the amount of NO and the specific NOS isoform that is activated. The excessive production of NO via neuronal NOS (nNOS) and inducible NOS (iNOS) appears to be neurotoxic, whereas the production of low amounts of NO by eNOS is protective after focal ischemia in rats (Yang et al. 2019). The positive effects of eNOS after cerebral ischemia has been attributed to vasodilatation, which helps restore blood flow to ischemic areas. Accordingly, eNOS knockout mice exhibited the exacerbation of injury after MCAo (Huang et al. 1996). In the present study, a decrease in hippocampal eNOS expression was detected after TGCI in rats. Roflumilast prevented this effect, which might coincide with its positive effects in TGCI rats.
Transient global cerebral ischemia is known to cause delayed neuronal loss, resulting in the expansion of the injured area after reperfusion. The degeneration of CA1 hippocampal neurons begins within the first hours after reperfusion (48-72 h) and spreads throughout the CA1 region until day 7 after global ischemia (Pulsinelli et al. 1982;Petito et al. 1987). The identification of degenerating neurons has classically relied on the recognition of morphological aspects, including hypereosinophilic cytoplasm, cellular shrinkage, nuclear pyknosis, and vacuolization, using traditional stainings, such as hematoxylin/eosin and Nissl staining. We used NeuN and FJC staining to evaluate hippocampal neurodegeneration after TGCI. No significant differences in NeuN levels were observed between TGCI rats and sham-operated controls 72 h after TGCI. However, ischemic rats exhibited an increase in FJC signals in the hippocampus, indicating that neurons were degenerating or dying (Santagostino et al. 2021) in response to TGCI. Indeed, identification of the neuronal marker NeuN by immunohistochemistry may have provided suboptimal sensitivity and low accuracy to visualize neurodegeneration at this time point of injury (i.e., 72 h post-ischemia).
Roflumilast did not protect hippocampal neurodegeneration that was induced by global ischemia in rats. These findings are consistent with previous studies in rats that were subjected to chronic cerebral hypoperfusion (Santiago et al. 2018) and also TGCI (Bonato et al. 2021). Other studies, however, reported neurohistological protection that was conferred by the PDE4 inhibitor rolipram, following global brain ischemia in rats and mice (Block et al. 1997;Li et al. 2011;Soares et al. 2016). The PDE4 inhibitors FCPR16 (Chen et al. 2017) and FCPR03 (Xu et al. 2019) also promoted histological neuroprotection in rats that were subjected to MCAo. The reason for the absence of a histological protective effect of roflumilast following TGCI that was observed in the present study is still unclear. One explanation may be related to different post-ischemic survival times and different doses of PDE inhibitors that were used in various studies.
BcL-2 is functionally characterized as an apoptosis suppression factor (Hwang et al. 2013), and its overexpression protects against neuronal loss that is induced by several cerebral insults (Zhao et al. 2003). Guo et al. (2019) found a decrease in BcL-2 levels in the hippocampus in ischemic rats 72 h after reperfusion . In contrast, cerebral ischemia that was induced by 4-VO did not affect BcL-2 levels in the hippocampus in rats 1, 3, 24, or 72 h after reperfusion (Pilchova et al. 2015). In agreement, we found no difference in BCL-2 levels 1 3 in the hippocampus between TGCI rats and controls. However, roflumilast increased BCL-2 levels in the hippocampus in TGCI rats, indicating a neuroprotective effect of the treatment. Wang et al. (2020) reported that APP/PS1 transgenic mice that were treated with roflumilast exhibited an increase in BcL-2 levels in the cerebral cortex. Moreover, under the influence of the neurotoxic anesthetic sevoflurane, hippocampal neurons that were treated with roflumilast also showed an upregulation of BcL-2 expression (Peng et al. 2018). Interestingly, roflumilast (1 µM) treatment alone had no impact on the viability or apoptosis of hippocampal neurons.
NG2 cells are often equated with OPCs because of their ability to generate myelinating and non-myelinating oligodendrocytes. Rats that were subjected to the 4-VO model exhibited an increase in NG2 cell expression in the CA1 hippocampal subfield 72 h after reperfusion (Jin et al. 2020). Similar results were found in the present study. TGCI rats exhibited an increase in NG2 expression in the hippocampal CA1 subfield 72 h after TGCI, which was further increased by roflumilast. Several studies also found that PDE4 inhibition promotes OPCs differentiation and enhances central nervous system remyelination (Whitaker et al. 2008;Syed et al. 2013). However, not all NG2-positive cells differentiate into oligodendrocytes. NG2 cells may be expressed by multiple cell lineages in the adult central nervous system (Mallon et al. 2002). Therefore, further studies are needed to clarify whether NG2-positive cells differentiate into mature oligodendrocytes in ischemic animals treated with roflumilast.
The clinical use of PDE4 inhibitors (e.g., rolipram) has been hampered because of nausea and emesis in patients. Roflumilast was the first selective PDE4 inhibitor that was approved by regulatory agencies for the treatment of chronic obstructive pulmonary disease, with relatively weak potency for inducing gastric adverse effects (Luo et al. 2018). Preclinical evidence has demonstrated the neuroprotective effects of roflumilast after brain injury. The present findings provide important information about the neuroprotective mechanism of action of roflumilast in transient cerebral ischemic conditions. The positive effects of roflumilast may involve protection against BBB damage that is caused by TGCI, including the preservation of tight-junction proteins (claudin-5), the promotion of vascularization (eNOS), and myelin repair (NG2).
Author Contribution JM Bonato conducted the animals' surgeries and biochemical assays. She wrote the first draft of the manuscript. BA de Mattos and DV Oliveira helped to perform the animals' surgeries. H Milani analyzed the data and helped with statistical analysis. J Prickaerts and RMW Oliveira conceived the experimental design, planned the experiments, and performed the data workup. They also contributed to writing the manuscript. All authors discussed the results and commented on the manuscript.

Availability of Data and Materials
The experimental data will be available when requested.

Declarations
Ethical Approval The experimental procedures conformed to the ethics of this article and are protected by copyright. All rights reserved principles of the Brazilian College of Animal Experimentation (COBEA) and were approved by the local Ethics Committee on Animal Experimentation of the State University of Maringá (CEUA No. 7995230420).