Hyperosmolar Mannitol Stimulates Expression of Aquaporins 4 and 9 through a p38 Mitogen-activated Protein Kinase-dependent Pathway in Rat Astrocytes*

The membrane pore proteins, aquaporins (AQPs), facilitate the osmotically driven passage of water and, in some instances, small solutes. Under hyperosmotic conditions, the expression of some AQPs changes, and some studies have shown that the expression of AQP1 and AQP5 is regulated by MAPKs. However, the mechanisms regulating the expression of AQP4 and AQP9 induced by hyperosmotic stress are poorly understood. In this study, we observed that hyperosmotic stress induced by mannitol increased the expression of AQP4 and AQP9 in cultured rat astrocytes, and intraperitoneal infusion of mannitol increased AQP4 and AQP9 in the rat brain cortex. In addition, a p38 MAPK inhibitor, but not ERK and JNK inhibitors, suppressed their expression in cultured astrocytes. AQPs play important roles in maintaining brain homeostasis. The expression of AQP4 and AQP9 in astrocytes changes after brain ischemia or traumatic injury, and some studies have shown that p38 MAPK in astrocytes is activated under similar conditions. Since mannitol is commonly used to reduce brain edema, understanding the regulation of AQPs and p38 MAPK in astrocytes under hyperosmotic conditions induced with mannitol may lead to a control of water movements and a new treatment for brain edema.

Water passes across cell membranes substantially faster than can be explained by simple diffusion. This is necessary for the active regulation of water homeostasis under conditions of osmotic stress in the kidneys, lungs, brain, etc., and to meet these needs, a family of membrane channel proteins evolved. These proteins, termed "aquaporins" (AQPs), 1 are found in all life forms, including archaea, eubacteria, fungi, plants, and all animal phyla (1)(2)(3). AQPs are widely distributed, and more than one AQP may be present in the same cell. All AQPs seem to have six transmembrane domains with five connecting loops and their amino and carboxyl termini in the cytoplasm (4 -6). AQPs are synthesized as monomers, but there is evidence that they exist as tetramers in the membrane and have four water pores (7). The expression of some AQPs is affected by various factors, including vasopressin (8,9), hypoxia and reoxygenation (10), growth factors (11), and retinoic acid (12). In addition, hyperosmotic stress increases the expression of AQPs 1, 2, 3, and 5 (13)(14)(15)(16)(17)(18)(19).
Many types of cells respond to changes in osmotic pressure with adaptive mechanisms that allow them to re-establish homeostasis of those aspects of cell structure and function that have been disturbed osmotically (20). Osmotic stress may damage DNA and proteins, resulting in impairment of cell function and the induction of repair processes and protection systems (21). These relatively nonspecific responses to cell damage may be an important aspect of cellular adaptation to osmotic stress. Mitogen-activated protein kinases (MAPKs), specifically extracellular signal-regulated kinase (ERK), c-Jun N-terminal protein kinase (JNK), and p38 MAPK, are important intracellular signal transduction pathways that are activated in response to changes in osmolality (22)(23)(24)(25). Thus, osmoregulation of AQPs might be mediated by MAPKs. There are some reports that the expression of AQP1 and AQP5 under hyperosmotic conditions is regulated by MAPKs (19,26), but the role of MAPKs in the expression of AQP4 and AQP9 under hyperosmotic stress is unclear.
Regulation of tissue water content and brain volume is critical for normal functioning of the central nervous system, which is highly sensitive to any increase in intracranial pressure. Brain edema may rapidly become life-threatening because of the rigid encasement of the brain. AQPs were suggested to have a role in maintaining the homeostasis of the brain (27)(28)(29). Among them, numerous studies indicated that the expression of AQP1 (30,31), AQP4 (32)(33)(34), and AQP9 (35) is sensitive to brain injury, swelling, and other experimental interventions. AQP4 in astrocytes of the brain cortex decreased in the acute phase of brain injury (32,33) and increased in the later phase (32,34). AQP4-deficient mice survive much longer and have a better neurological outcome from experimentally induced brain edema than wild-type mice (36). AQP4 in astrocytes is tightly associated with the blood-brain barrier (37). In the brain cortex, astrocyte-specific AQP9 appears in the infarct border zone after transient ischemic stroke (35).
Hyperosmotic solutions of mannitol are used commonly to reduce brain edema, so it is important to elucidate the relationship between the expression of AQPs and hyperosmolarity induced with mannitol. This study was designed to assess the effects of hyperosmotic stress induced with mannitol on the expression of AQP4 and AQP9 and to identify the signal transduction pathways regulating the response in rat astrocytes in which AQP4 and AQP9 are localized (10). We found that the expression of AQP4 and AQP9 is induced in cultured rat astrocytes and in rat brain cortex by hyperosmotic stress. In addition, we demonstrated that this induction requires selective activation and involvement of the p38 MAPK pathway in cultured rat astrocytes. We believe this study provides the first example of osmotic regulation of AQP4 and AQP9 and the first evidence that the expression of AQP4 and AQP9 is regulated by the p38 MAPK pathway.

EXPERIMENTAL PROCEDURES
Cell Culture-Cultures of cortical astrocytes were prepared from rat postnatal cortices (P2) according to previously described methods (38). Trypsinized and dissociated cortical cells were cultured in 75-cm 2 culture flasks (Corning Glass) containing low glucose (1000 mg/liter) Dulbecco's modified Eagle's medium (l-DMEM; Invitrogen) supplemented with 10% fetal bovine serum (Sigma). After incubation for 5-7 days, the cells were trypsinized and subcultured in 60-mm diameter culture dishes (Falcon; PGC Scientific, Frederick, MD). The cell population consisted of over 95% astrocytes as determined by immunocytochemical staining with anti-glial fibrillary acidic protein (GFAP) antibody. In neural cells, AQP4 and AQP9 are expressed in astrocytes but not in neurons, oligodendrocytes, and microglia (10).
Hyperosmotic Stress of Cultured Rat Astrocytes-When astrocyte cultures became confluent, equal volumes of medium with or without various concentrations of mannitol (Sigma) were added, and the cultures were incubated for the indicated times. Osmolality of the medium was measured by the freezing point depression method (Osmostat TM OM-6040; ARKREY, Kyoto, Japan). All cultures were maintained at 37°C in an atmosphere of 5% CO 2 , 95% air.
Semiquantification of mRNA-The PCR products were visualized by ethidium bromide staining following separation on 1.5% agarose gels and quantified by laser-induced fluorescence-linked capillary gel electrophoresis. Details of laser-induced fluorescence-linked capillary gel electrophoresis have been reported elsewhere (10,42,43). Briefly, 40 l of PCR products was applied to a P/ACE system 5010 (Beckman Instruments). Samples were injected at a pressure of 0.5 p.s.i.g. for 20 s. Separation was carried out at 9.4 kV for 35 min in a capillary (47 cm long, 100-m diameter) filled with electrophoresis gel buffer containing DNA fluorescence stain (LIFluor dsDNA 1000 kit; Beckman). Laserinduced fluorescence detection was preformed using an argon-ion laser source with excitation at 488 nm and emission at 520 nm. The chromatograms were visualized and analyzed automatically using GOLD Software (Beckman). The peaks were expressed in relative fluorescence units, and the retention time was expressed in minutes. AQP4, AQP9, and ␤-actin PCR products were compared by integrating each peak area.
Western Blot Analysis-Cells cultured under the specified conditions were harvested with Ca 2ϩ -and Mg 2ϩ -free phosphate-buffered saline (PBS (-)) and centrifuged for 10 min at 800 ϫ g at 4°C. For quantitation of AQP4 and AQP9 protein, cell pellets were suspended in 100 l of Tris-buffered saline (TBS) containing 200 mM phenylmethanesulfonyl fluoride, 10 M pepstatin A, 10 M leupeptin, 2 mM EDTA, and 0.5% Nonidet P-40 and then sonicated on ice at 8-watt output for 10 s. Samples were kept at room temperature for 15 min with 5 l of SDS sample buffer. For quantitation of MAPKs and phosphorylated MAPKs, the cell pellets were suspended in 150 l of buffer (50 mM Tris-HCl, pH 7.4, 1 mM phenylmethanesulfonyl fluoride, 10 M pepstatin A, 10 M leupeptin) and sonicated on ice at 8-watt output for 10 s. The cell lysates were centrifuged for 30 min at 15,000 ϫ g at 4°C, and the supernatants were collected and boiled with 5 l of SDS sample buffer for 5 min. For GFAP, cell pellets were suspended and sonicated as for AQPs and boiled with 5 l of SDS sample buffer for 5 min. Total protein was measured using a BCA protein assay reagent kit (Pierce). 3 g of protein for AQP4 and GFAP, 20 g for AQP9, and 30 g for MAPKs and phosphorylated MAPKs were separated by 12.5% SDS-polyacrylamide gel electrophoresis (ATTO, Tokyo, Japan) and transferred to a Clear Blot Membrane-P (ATTO), which was then blocked for 1 h with 5% skim milk in TBS-T (20 mM Tris, pH 7.6, 137 mM NaCl, 0.1% Tween 20). The membranes were incubated overnight with anti-AQP4 antibody (Chemicon International, Inc., Temecula, CA) diluted 1:750, anti-AQP9 antibody (Chemicon) diluted 1:500, anti-p38 MAPK antibody (Santa Cruz Biotechnology, Santa Cruz, CA) diluted 1:500, anti-phosphorylated p38 MAPK antibody (Cell Signaling Technology, Inc., Beverly, MA) diluted 1:1000, anti-ERK antibody (Promega) diluted 1:5000, anti-phosphorylated ERK antibody (Cell Signaling Technology) diluted 1:1000, anti-JNK antibody (Santa Cruz Biotechnology) diluted 1:500, anti-phosphorylated JNK antibody (Cell Signaling Technology) diluted 1:1000, or anti-GFAP antibody (Chemicon) diluted 1:1000 all in TBS-T containing 5% skim milk. After washing the membranes four times for 15 min each with TBS-T, they were incubated for 1 h with 1:1000 diluted (for AQP4, MAPKs, and GFAP) or 1:500 diluted (for AQP9) horseradish peroxidase-conjugated secondary antibody in TBS-T buffer containing 5% skim milk. Protein was visualized using an ECL detection system (Amersham Biosciences). Relative band intensities were determined by densitometry using Kodak Digital Science 1D version 2.0 (Eastman Kodak Co.).
Isolation of the 5Ј-Flanking Regions of Human AQP4 and AQP9 Genes-The 5Ј-flanking region of the human AQP4 gene was isolated as described previously (44). The AQP4 gene is regulated by alternative promoters upstream of exons 0 and 1. The 5Ј-flanking regions of the exon 0 promoter (bp Ϫ1930 to ϩ29) (nucleotide ϩ1 corresponds to the initiation codon of exon 0) and the exon 1 promoter (bp Ϫ297 to ϩ113) (nucleotide ϩ1 corresponds to the initiation codon of exon 1) were subcloned into the vector pCR2.1 (Invitrogen), and 5Ј deletion mutants were constructed. Deletion mutants corresponding to bp Ϫ1930 to ϩ29 in the exon 0 promoter (eight constructs) and from bp Ϫ297 to ϩ133 in the exon 1 promoter were fused to the promoterless vector PGV-B (Toyo-ink, Tokyo, Japan) containing the luciferase reporter gene.
The promoter region of AQP9 (bp Ϫ1125 to ϩ55) (nucleotide ϩ1 corresponds to the initiation codon) was also isolated by a similar method. Briefly, the region was isolated using a Promoter Finder DNA Walking kit (Clontech, Palo Alto, CA), which contains adapter-linked human genomic DNA digested with EcoRV, ScaI, DraI, PvuII, and SspI. A two-step PCR was performed according to the manufacturer's instructions. The first PCR consisted of 94°C for 1 min followed by 30 cycles of 94°C for 15 s, 60°C for 30 s, and 68°C for 4 min and was done in a final volume of 50 l with 10 pmol of a specific primer, 5Ј-GCTCTTCAAGAC-CAGTCTCTGCTTGA-3Ј (nucleotides ϩ60 to ϩ46). The second PCR was done with the same protocol using 1 l of the first PCR product and a specific primer, 5Ј-CTCTGAGGACTCCTGTTTCTACCAAT-3Ј (nucleotides Ϫ11 to Ϫ35). The PCR products were subcloned into pCR2.1, and both strands were sequenced completely. BLAST searches were performed on the NCBI sites. The sequence was identical to GenBank TM accession number AC025431. Deletion mutants corresponding to bp Ϫ1125 to ϩ55 in the AQP9 promoter (four constructs) were fused to the promoterless vector PGV-B.
Cell Transfection and Luciferase Reporter Gene Assay-Reporter assays were done on AQP4 and AQP9 gene fragments fused upstream of the firefly luciferase reporter in PGV-B using the Dual-Luciferase re-porter assay system (Promega), with which the pRL-thymidine kinase (TK) plasmid (Promega) containing the Renilla luciferase gene under control of the TK promoter was co-transfected as an internal control. For transient transfection, astrocytes in l-DMEM supplemented with 10% fetal bovine serum were plated at 1 ϫ 10 5 cells/well in a 48-well plate, incubated for 18 h, and transfected in serum-free medium with 0.4 g of the test plasmid and 0.01 g of pRL-TK using 1 l of Lipofect-AMINE 2000 (Invitrogen). After 4 h, 150 l of l-DMEM containing 20% fetal bovine serum was added. 48 h after transfection, equal volumes of l-DMEM or 3% mannitol in l-DMEM were added. After 24 h, the astrocytes were washed and harvested. The activities of control Renilla luciferase and firefly luciferase were measured in triplicate.
Induction of Hyperosmolarity in Rats-All animal studies were undertaken after the protocols were approved by the Animals Care and Use Committee of Nagoya City University Graduate School of Medical Sciences. Male Wistar rats (150 g) were used throughout. Animals were anesthetized with pentobarbital (Schering-Plough Animal Health, Kenilworth, NJ) (3.2 mg/100 g, body weight) and placed under a heating lamp. Intraperitoneal infusion was performed using a modification of a technique described previously (45). A polyurethane catheter (20 gauge) was inserted into the peritoneum by puncture and tightly fixed to avoid leakage of the experimental solution. The animals were then allowed to recover under a heating lamp. An isosmotic solution (121 mmol of NaCl, 5 mmol of KCl, 1 mmol of MgCl 2 , 1 mmol of NaH 2 PO 4 , 2 mmol of CaCl 2 , 18 of mmol NaHCO 3 , 22 mmol of glucose) or a hyperosmotic mannitol solution (1.1 mol of mannitol added to the isosmotic solution) was administered by intraperitoneal infusion using an infusion pump. The infusion rate was 1.9 ml/h, and the duration of the infusion was 1, 3, or 6 h. Control animals were anesthetized and cannulated but were not administered any solution. At the conclusion of the experiments, the animals were disconnected from the infusion pump and rapidly decapitated. Samples for Western blot analyses, done as described above, were obtained from the superficial region of the brain cortex, and cerebral sections (5 mm thick) were obtained for immunohistochemical staining. Blood samples were removed, and plasma osmotic pressure was measured by the freezing point depression method as described above.
Immunohistochemical Staining-Coronal sections from the brain were embedded in paraffin by the AMeX method (46). Briefly, tissues were fixed in acetone at 4°C for 3 days, cleared in methyl benzoate and xylene, and then embedded in paraffin. Sections 3 m thick were prepared and deparaffinized with xylene. Immunoreactivity was visualized using the streptavidin/biotin method (Histofine SAB-PO Kit, Nichirei, Tokyo, Japan). After washing the sections with PBS (Ϫ), they were treated with 0.3% (v/v) hydrogen peroxide in methanol for 30 min to inactivate endogenous peroxidase. They were then immersed in a 1:20 dilution of nonimmune goat serum for 10 min to block nonspecific binding, and after blotting to remove excess serum they were incubated at room temperature for 1 h with anti-AQP4 antibody (Chemicon) diluted 1:400, anti-AQP9 antibody (Chemicon) diluted 1:100, or anti-GFAP antibody (Chemicon) diluted 1:50 in PBS (Ϫ) containing 1% bovine serum albumin. Control sections were treated with nonimmune rabbit immunoglobulins (MBL, Nagoya, Japan). The sections were rinsed three times with PBS (Ϫ), incubated for 30 min each with the secondary antibody, rinsed three times, and then incubated with a streptavidin-biotin-peroxidase complex for 15 min. After washing the sections with PBS (Ϫ), the peroxidase reaction was developed by incubating them in 0.02% (w/v) 3,3Ј-diaminobenzidine tetrahydrochloride (Sigma) solution containing 0.003% (v/v) hydrogen peroxide and 10 mM sodium azide. The sections were counterstained with hematoxylin.
Data Analysis-Data are expressed as means Ϯ S.E. of at least four independent experiments. Statistical analysis was performed by oneway factorial analysis of variance combined with Scheffe's test for all comparison pairs. Reporter gene assays were carried out three times, and the results were analyzed by Student's t test for statistical significance. Differences with p values Ͻ 0.05 were considered significant.
To analyze the time course of the mannitol-mediated induction of AQP4 and AQP9 expression, astrocytes were treated with hyperosmotic medium containing 3% mannitol for various periods of time. AQP4 mRNA increased significantly after 1, 3, 6, and 12 h of treatment with mannitol, and AQP9 mRNA increased significantly after 3, 6, and 12 h of treatment ( Fig.  2A). After 24 h, both AQP mRNAs returned to base-line levels. AQP4 and AQP9 proteins increased after 6 h of treatment with FIG. 1. Effect of mannitol on expression of AQP4 and AQP9 mRNAs and proteins. A, rat astrocytes were incubated for 6 h in isosmotic control medium or hyperosmotic medium containing the designated concentrations of mannitol. Cells were processed for RT-PCR and semiquantification of AQP4 or AQP9 mRNA. The expression of mRNA (normalized to ␤-actin) (n ϭ 8 for each group) is presented as a percentage of the control (mean Ϯ S.E.; *, p Ͻ 0.05 versus control). B, rat astrocytes were treated as described in A and processed for immunoblotting with anti-AQP4 or anti-AQP9 antibody. C, immunoblots (n ϭ 8 for each group) were analyzed by densitometry, and band densities are presented as a percentage of the control (mean Ϯ S.E.; *, p Ͻ 0.05 versus control).
p38 MAPK Regulates AQP4 and -9 under Hyperosmotic Conditions mannitol (Fig. 2, B and C). Although AQP4 mRNA returned to the base-line level after 24 h of treatment with mannitol, AQP4 protein remained at the same high level, suggesting that this gene is regulated at both the transcriptional and post-transcriptional levels. AQP9 protein peaked at 12 h but returned to the base-line level after 24 h of treatment with mannitol.
To see whether hyperosmotic induction of AQP4 and AQP9 mRNAs requires de novo protein synthesis, astrocytes incubated in hyperosmotic medium containing 3% mannitol were treated with or without the protein synthesis inhibitor, cycloheximide (25 g/ml), for 6 h. The increased expression of both AQP mRNAs induced by mannitol was not affected by cycloheximide (Fig. 3), suggesting that de novo protein synthesis is not involved.
To determine whether hyperosmotic induction of AQP4 and AQP9 is solute-specific, astrocytes were incubated in isosmotic medium or hyperosmotic medium supplemented with 0.4% NaCl (502 Ϯ 0.7 mOsM), 1% urea (501 Ϯ 2.5 mOsM), 1.5% glycerol (491 Ϯ 4.5 mOsM), 3% sorbitol (485 Ϯ 9.2 mOsM), or 3% mannitol (491 Ϯ 4.7 mOsM), for 6 h. NaCl and glycerol had little effect on the expression of AQP4 and AQP9 mRNAs and proteins, whereas they were decreased by urea and increased by mannitol and sorbitol (Fig. 4, A-C). Except for NaCl, the effects of specific solutes on hyperosmotic induction of AQP4 and AQP9 were similar to those reported for AQP1, -3, and -5 (17,19,26). Although 6 h of exposure to hyperosmotic medium supplemented with NaCl did not increase the expression of AQP4 and AQP9, other studies have demonstrated that over 12 h of treatment with NaCl is required to increase the expression of AQP1 and AQP5 (19,26). To examine the time course of NaCl-, urea-, glycerol-, or sorbitol-mediated induction of AQP4 and AQP9 expression, astrocytes were treated with hyperosmotic medium supplemented with the solutes for various periods of time before harvesting for immunoblot analysis. AQP4 and AQP9 proteins increased after 12 h of treatment with NaCl (Fig. 4D). Although AQP4 protein was at the same high level even after 24 h of treatment, AQP9 protein decreased but was still higher than in the control. After 12 or 24 h of treatment, urea and glycerol did not affect the levels of AQP4 and AQP9 proteins. The effect of sorbitol was the same as that of mannitol (data not shown).

The Role of MAPKs in the Expression of AQP4 and AQP9 in Cultured Rat Astrocytes Subjected to Hyperosmotic Stress-
Hyperosmotic stress can activate signaling through three MAPK pathways, p38 MAPK, ERK, and JNK, all three of which are activated in mouse astrocytes by sorbitol (47). To investigate whether these three MAPKs are activated in rat astrocytes by mannitol, we performed immunoblot analyses of cells incubated in hyperosmotic medium using antibodies that react with either the phosphorylated or total amounts of each of the three MAPKs. Whereas the total amount of each MAPK remained constant, p38 MAPK was activated by hyperosmotic stress after 30 min, and activation persisted even at 120 min (Fig. 5A). ERK (Fig. 5B) and JNK (Fig. 5C) activation both peaked at 30 min after hyperosmotic exposure.
To investigate the possible involvement of the three MAPK cascades in the increased expression of AQP4 in response to mannitol, the effects of the p38 MAPK inhibitor SB203580, the mitogen-activated extracellular signal-regulated kinases (MEK) 1/2 inhibitor PD98059, and the JNK inhibitor SP600125 were examined. MEK 1/2 are the upstream kinases that activate ERK. The hyperosmotic induction of AQP4 mRNA and p38 MAPK Regulates AQP4 and -9 under Hyperosmotic Conditions protein was inhibited by SB203580 (10 M) but not by PD98059 (10 M) or SP600125 (10 M) (Fig. 6, A and B), and all three inhibitors had no effect on the basal expression of AQP4. To determine whether the p38 MAPK inhibitor specifically prevented AQP4 induction, astrocytes were incubated with Me 2 SO. It had no effect on AQP4 expression, suggesting that the reduction of AQP4 expression caused by the p38 MAPK inhibitor was not a result of nonspecific effects. Thus, it appears that the increased expression of AQP4 in response to mannitol requires signaling through p38 MAPK. RT-PCR revealed that SB203580 reduced the expression of AQP4 mRNA in a concentration-dependent manner (Fig. 6C). The same results were obtained with AQP9 (Fig. 7).
To determine whether just p38 MAPK activation is sufficient for inducing AQP4 and AQP9, astrocytes were incubated with the p38 MAPK activators, hydrogen peroxide (2 mM) or anisomycin (50 M), under isosmotic conditions. The total amount of p38 MAPK remained constant, and both activators caused its phosphorylation (Fig. 8, A and B). Activation of p38 MAPK by hydrogen peroxide increased 15 min after exposure and peaked at 60 min (Fig. 8A), whereas its activation by exposure to anisomycin peaked at 30 min and was observed only weakly at 60 min (Fig. 8B). Hydrogen peroxide increased AQP4 and AQP9 mRNAs, whereas they were decreased by anisomycin (Fig. 8C). AQP4 and AQP9 proteins were increased by hydrogen peroxide but were not affected by anisomycin (Fig. 8D).

Transcriptional Regulation of AQP4 and AQP9 Genes in Cultured Rat Astrocytes under Hyperosmotic Conditions-The
promoter activity of the AQP4 gene has been demonstrated in SF-126 (glioblastoma) and Madin-Darby canine kidney cells (44), but that of the AQP9 gene has not been investigated. Luciferase promoter gene assays were performed to determine the transcriptional activity of the AQP4 and AQP9 genes in astrocytes under isosmotic and hyperosmotic conditions. Eight luciferase promoter constructs of AQP4 exon 0, one of AQP4 exon 1, or four of AQP9 were transfected into astrocytes, and luciferase activity was measured under isosmotic conditions FIG. 4. Hyperosmotic induction of AQP4 and AQP9 mRNAs and proteins by different solutes. A, rat astrocytes were incubated for 6 h in control isosmotic medium (final osmotic pressure, 311 Ϯ 1.4 mOsm (n ϭ 4 for each group)) or medium supplemented with 0.4% NaCl (502 Ϯ 0.7 mOsm), 1% urea (501 Ϯ 2.5 mOsm), 1.5% glycerol (491 Ϯ 4.5 mOsm), 3% sorbitol (485 Ϯ 9.2 mOsm), or 3% mannitol (491 Ϯ 4.7 mOsm) and then processed for RT-PCR and semiquantification of AQP4 or AQP9 mRNA. The expression of mRNA (normalized to ␤-actin) (n ϭ 4 for each group) is presented as a percentage of that in control isosmotic medium (mean Ϯ S.E.; *, p Ͻ 0.05 versus control). B, rat astrocytes were treated as described in A and processed for immunoblotting with anti-AQP4 or anti-AQP9 antibody. C, immunoblots (n ϭ 4 for each group) were analyzed by densitometry, and band densities are presented as a percentage of that in control isosmotic medium (mean Ϯ S.E.; *, p Ͻ 0.05 versus control). D, rat astrocytes were incubated in hyperosmotic medium supplemented with 0.4% NaCl. At the indicated times, cells were processed for immunoblotting with anti-AQP4 or anti-AQP9 antibody. The immunoblots (n ϭ 4 for each group) were analyzed by densitometry, and band densities are presented as a percentage of the control at time 0 (mean Ϯ S.E.; *, p Ͻ 0.05 versus time 0).

FIG. 5. Phosphorylation of MAPKs induced by hyperosmotic stress.
Rat astrocytes were incubated in hyperosmotic medium containing 3% mannitol. At the indicated times, cells were harvested and processed for immunoblotting using anti-p38 MAPK antibody or antiphosphorylated p38 MAPK (p-p38) antibody (A), anti-ERK antibody or anti-phosphorylated ERK (p-ERK) antibody (B), or anti JNK antibody or anti-phosphorylated JNK (p-JNK) antibody (C). (Fig. 9, A and B). The Ϫ297/ϩ113 exon 1 construct had the greatest activity among the AQP4 constructs. All eight constructs derived from the exon 0 promoter had weak luciferase activity. There was an increase in luciferase activity of the exon 0 Ϫ830/ϩ29 construct compared with the longer isoform, indicating the presence of a suppressor element just upstream of bp Ϫ831. Luciferase activity of the Ϫ175/ϩ55 AQP9 construct was greater than that of the other longer AQP9 constructs, indicating the presence of a suppressor element just upstream of bp Ϫ174.
Under hyperosmotic conditions achieved with 3% mannitol, the luciferase activities of the Ϫ1625/ϩ29, Ϫ830/ϩ29, Ϫ515/ ϩ29, and Ϫ428/ϩ29 AQP4 exon 0 constructs were induced significantly (Fig. 9A). Luciferase activity of the Ϫ428/ϩ29 construct was increased the most. These results indicate that the AQP4 promoter is up-regulated by hyperosmotic stress and that a critical cis element involved is located between Ϫ345 and Ϫ428. On the other hand, the Ϫ765/ϩ55 and Ϫ486/ϩ55 AQP9 constructs showed significant induction of luciferase activity, but other AQP9 constructs were not affected by hyperosmolarity (Fig. 9B). These results indicate that the AQP9 promoter is also up-regulated by hyperosmotic stress and that a critical cis element is located between Ϫ486 and Ϫ176.
Induction of AQP4 and AQP9 Expression in Rat Brain Cortex by Intraperitoneal Infusion of Hyperosmotic Solution-To investigate whether AQP4 and AQP9 expression is induced in rat brain cortex under hyperosmotic conditions, an isosmotic solution (osmotic pressure was 298 Ϯ 5.9 mOsM (n ϭ 4)) or a hyperosmotic solution of mannitol (1368 Ϯ 24.8 mOsM) was infused intraperitoneally for 1, 3, and 6 h, after which samples from the superficial region of brain cortex were taken for Western blot analyses, and cerebral sections were taken for immunohistochemical staining. Animals recovered from anesthesia within 1 h after intraperitoneal cannulation. The position of the intraperitoneal catheter tip was confirmed after decapitation, at which time blood samples were obtained from the right atrium. The plasma osmotic pressure of control animals was 297 Ϯ 3.9 mOsM, and it did not increase after infusion of the isosmotic solution (data not shown). Infusion of the hyperosmotic solution resulted in an increase of the plasma osmotic To determine whether the reduction of AQPs induced by MAPK inhibitors was a result of nonspecific effects, astrocytes were incubated with Me 2 SO (DMSO; 5 M). Cells were processed for RT-PCR and semiquantification of AQP4 mRNA. The expression of mRNA (normalized to ␤-actin) (n ϭ 4 for each group) is presented as a percentage of the control (mean Ϯ S.E.; *, p Ͻ 0.05 versus control). B, rat astrocytes were treated as described in A and processed for immunoblotting with anti-AQP4 antibody. C, rat astrocytes were incubated for 6 h with or without 3% mannitol and the p38 MAPK inhibitor SB203580 at the indicated concentrations. Cells were processed for RT-PCR and semiquantification of mRNA. The expression of mRNA is presented as described in A (mean Ϯ S.E.; n ϭ 4 for each group; *, p Ͻ 0.05 versus control).

FIG. 7. Effect of MAPK inhibitors on hyperosmotic induction of AQP9 mRNA and protein.
Rat astrocytes were incubated and treated as described in the legend for Fig. 6. Using AQP9 primer and anti-AQP9 antibody, RT-PCR (A and C) and immunoblotting (B) were carried out as described in the legend for Fig. 6. DMSO, Me 2 SO.
Western blotting (Fig. 10A) showed the brain cortices of rats administered the mannitol solution had increased expression of AQP4 and AQP9. AQP4 protein increased after 3 h of infusion of the mannitol solution and was expressed even when the solution was infused for 6 h. AQP9 protein increased after 1 h of infusion of the mannitol solution and peaked at 3 h. After 6 h, AQP9 protein decreased but was still higher than in the control. GFAP remained constant whether the solution infused was isosmotic or hyperosmotic. In normal brain parenchyma, AQP4 is distributed in astrocyte foot processes, and AQP4 immunoreactivity was seen mainly around the vessels and under the pial surface (28,29), and we found a normal immunostaining pattern of AQP4 under isosmotic conditions (Fig.  10B, a). Positive staining was observed under the pial surface and around microvessels, with the area along the blood vessels stained weekly. Intraperitoneal infusion of mannitol solution for 6 h significantly increased the positive staining of AQP4 under the pial surface and the area along the blood vessels (Fig.  10B, b). AQP9 was stained weakly in normal brain cortex and was difficult to find. The staining pattern of AQP9 under isosmotic conditions is shown in Fig. 10B, c, with positive staining under the pial surface. Infusion of the hyperosmotic solution for 6 h increased the positive staining of AQP9 under the pial surface and around the blood vessels (Fig. 10B, d). GFAP was stained weakly and did not change with infusion of an isosmotic or hyperosmotic solution (data not shown).

DISCUSSION
Although hyperosmotic stress often causes a decrease of total DNA synthesis in mammalian cells, the up-regulation of the expression of a limited number of genes has been described. Recent studies have shown that the expression of AQPs is induced in mammalian cells by hyperosmotic stress. AQP1 in FIG. 8. Effect of p38 MAPK activators on the induction of AQP4 and AQP9 mRNAs and proteins. A, rat astrocytes were incubated in isosmotic medium containing hydrogen peroxide (2 mM). At the indicated times, cells were harvested and processed for immunoblotting using anti-p38 MAPK antibody or anti-phosphorylated p38 MAPK (p-p38). B, rat astrocytes were incubated in isosmotic medium containing anisomycin (50 M). The blots were processed as described in A. C, rat astrocytes were incubated for 6 h with or without hydrogen peroxide (2 mM) or anisomycin (50 M) under isosmotic conditions. To determine whether the reduction of AQPs expression by MAPK activators was a result of nonspecific effects, astrocytes were also incubated with Me 2 SO (DMSO; 5 M). Cells were processed for RT-PCR and semiquantification of AQP4 or AQP9 mRNA. The expression of mRNA (normalized to ␤-actin) (n ϭ 4 for each group) is presented as a percentage of the control (mean Ϯ S.E.; *, p Ͻ 0.05 versus control). D, rat astrocytes were incubated for 6 h with or without hydrogen peroxide (2 mM) or anisomycin (50 M) under isosmotic conditions and processed for immunoblotting with anti-AQP4 or anti-AQP9 antibody.
FIG. 9. Promoter activities of the 5-flanking regions of the AQP4 and AQP9 genes. A, AQP4 promoter luciferase plasmid constructs were transfected into rat astrocytes. The transfected cells were incubated for 24 h with or without 3% mannitol, and promoter activity was analyzed. The expression of promoter activities (normalized to Renilla luciferase activity) (n ϭ 3 for each group) is presented as a percentage of that of the AQP4 exon 0 construct (bp Ϫ1930 to ϩ29) activity under isosmotic conditions (mean Ϯ S.E.; *, p Ͻ 0.05, with versus without 3% mannitol with each construct). B, AQP9 promoter activity was analyzed as described in A. The results (normalized to the Renilla luciferase activity) are expressed as percentages of the AQP9 construct (bp Ϫ1125 to ϩ55) activity under isosmotic conditions (mean Ϯ S.E.; n ϭ 3 for each group; *, p Ͻ 0.05, with versus without 3% mannitol with each construct).
p38 MAPK Regulates AQP4 and -9 under Hyperosmotic Conditions mouse inner medullary cells (13), BALB/c fibroblasts (14), and mIMCD-3 cells (15); AQP2 in outer medullary collecting duct (OMCD) cells (16); AQP3 in human keratinocytes (17) and Madin-Darby canine kidney epithelial cells (18); and AQP5 in mouse lung epithelial (MLE-15) cells (19) have been reported to increase. Here, we show that hyperosmotic stress induced with mannitol solution increased AQP4 and AQP9 expression in cultured rat astrocytes and in rat brain cortex. The protein synthesis inhibitor, cycloheximide, did not suppress the expres-sion of AQP4 and AQP9 mRNAs induced by hyperosmotic stress, indicating that the induction of AQP4 and AQP9 does not require de novo protein synthesis and is due to direct stimulation of an intracellular signaling pathway. This is supported by the results of luciferase promoter gene assays. Since AQP4 and AQP9 of astrocytes were induced by hyperosmotic mannitol solution, which is commonly used to reduce brain edema, it is suggested that AQP4 and AQP9 play important roles in the therapy of brain edema.
Mannitol decreases intracranial pressure and brain water content and increases the expression of AQP4 and AQP9. However, it is possible that up-regulation of AQP4 in astrocytes contributes to brain edema. AQP4 expression in astrocytes occurs at sites where the blood-brain barrier is disrupted (28). In brain cortex astrocytes, AQP4 decreases within 48 h in the acute phase of brain injury (32,33), and overexpression of AQP4 is observed in the peri-contusional area after 3 days and later (32). The decrease of AQP4 in the acute phase is described as an endogenous protective mechanism to reduced glial water accumulation and cell swelling (32). The increase of AQP4 in the late phase is explained as the loss of its polarity and its redistribution throughout the astrocyte. AQP4 up-regulation is suggested to be a maladaptation reaction (28). On the other hand, astrocyte-specific AQP9 of the brain cortex appears in the infarct border zone 48 h after transient ischemic stroke (35). AQP9 might be involved in reperfusion edema associated with lactic acidosis, since it is permeable to water and lactic acid (28). Therapy with hyperosmotic mannitol solution might have an influence on the expression of AQP4 and AQP9 in astrocytes after brain injury or ischemia.
When cells were treated in hyperosmotic medium containing the various concentrations of mannitol for 6 h, AQP4 protein increased maximally after treatment with 3% mannitol, whereas AQP9 protein peaked with 1% mannitol. When cells were treated with 3% mannitol, AQP4 protein increased after 6 h and remained at the same high level after 24 h, whereas AQP9 protein peaked at 12 h and returned to the base-line level after 24 h. Differences between AQP4 and AQP9 expression were also detected in rat brain cortex by Western blot analysis and immunohistochemical staining after intraperitoneal infusion of a hyperosmotic solution. These differences might reflect different roles for AQP4 and AQP9 when astrocytes are exposed to a hyperosmotic environment.
AQP1 is increased in mIMCD-3 cells by 12-h treatments with hyperosmotic raffinose, glucose, sucrose, sorbitol, and NaCl but not with urea (26). In keratinocytes, AQP3 is increased by 8-h treatments with hyperosmotic mannitol, sorbitol, glucose, sucrose, and NaCl, whereas glycerol has little effect, and the expression of AQP3 is decreased by urea (17). AQP5 is increased in MLE-15 cells by 16 -20-h treatments with hyperosmotic mannitol, sorbitol, and NaCl, whereas urea has no effect (19). In our study, hyperosmotic mannitol, sorbitol, and NaCl increased the expression of AQP4 and AQP9, whereas glycerol had no effect, and urea decreased them. Although mannitol and sorbitol increased AQP4 and AQP9 within 6 h, treatment with NaCl required 12 h to increase them. The solute specificities of the hyperosmotic induction of AQP4 and AQP9 are similar to those reported in the above studies on AQP1, -3, and -5. A hyperosmotic gradient is required to increase the expression of AQPs. Since urea and glycerol permeate cell membranes relatively freely, they do not create osmotic gradients and do not increase the expression of AQPs (17,26). Mannitol solution increased the expression of AQP4 and AQP9 in rat brain cortex. Mannitol crosses the normal blood-brain barrier and concentrates in extracellular spaces of the brain following repeated injections (48,49). It seems reasonable that the increase of FIG. 10. Effect of intraperitoneal administration of mannitol solution on the expression of AQP4 and AQP9 proteins in rat brain cortex. A, rats were anesthetized and intraperitoneally cannulated, and an isosmotic solution or hyperosmotic solution was infused. After the indicated times, samples were removed from the superficial region of the brain cortex and processed for immunoblotting with anti-AQP4, anti-AQP9, or anti-GFAP antibody. B, rats were manipulated as described in A. After 6 h, cerebral sections were obtained, and immunohistochemical staining was performed with anti-AQP4 or anti-AQP9 antibody. a, photomicrograph showing the immunostaining pattern of AQP4 under isosmotic conditions. Positive staining is observed under the pial surface and around the microvessels. The area along the blood vessel is stained weakly (arrowheads). b, positive staining of AQP4 increased under the pial surface and the area along the blood vessels (arrowheads) under hyperosmotic conditions. c, photomicrograph showing the immunostaining pattern of AQP9 under isosmotic conditions. Positive staining is observed under the pial surface. There is no staining around the blood vessel (arrowheads). d, positive staining of AQP9 increased under the pial surface and was observed around the blood vessel (arrowheads) under hyperosmotic conditions. Original magnification was ϫ50.
AQP4 and AQP9 in our in vivo experiments is associated with the osmotic gradient of mannitol across the blood-brain barrier and/or a direct effect of mannitol.
MAPKs are important intracellular signal transduction pathways that are involved in the protective response of cells to hyperosmotic stress (22)(23)(24)(25). Under hyperosmotic conditions, AQP1 is regulated by all three MAPKs (26), and AQP5 is regulated by ERK (19). In our study, all three MAPKs were activated, as indicated by Western blotting using a phosphospecific antibody. In addition, the effects of inhibitors of p38 MAPK, ERK, and JNK were examined for their effects on AQP4 and AQP9 induction by hyperosmotic stress. Only the p38 MAPK inhibitor suppressed AQP4 and AQP9 expression, indicating that AQP4 and AQP9 expression in rat astrocytes under hyperosmotic conditions can be regulated by the p38 MAPK pathway. On the other hand, hydrogen peroxide, but not anisomycin, increased AQP4 and AQP9 expression. Although both agents are potent activators of p38 MAPK, oxidative stress like that produced by hydrogen peroxide activates intracellular pathways other than MAPKs, namely the phosphoinositide 3 kinase pathway, phospholipase C signaling, protein kinase C, and nuclear factor B (50), and anisomycin activates JNK and p38 MAPK. There are two possible explanations of our results. First, p38 MAPK activation is necessary but not sufficient for AQP4 and AQP9 induction by hyperosmotic stress. Second, p38 MAPK activation is sufficient for AQP4 and AQP9 induction, but the activation must exceed 60 min, which was achieved by treatment with hydrogen peroxide and mannitol but not with anisomycin. Results very similar to ours have been reported for AQP5 expression (19). The ERK inhibitor suppressed AQP5 expression induced by hyperosmotic stress, but the ERK activator did not activate AQP5 expression. Therefore, ERK activation is necessary but not sufficient for AQP5 induction by hyperosmotic stress.
TonE (tonicity-responsive enhancer) is a cis-acting element responsible for hypertonic regulation, and a consensus sequence of TonE mediates the stimulation of transcription of several genes in response to hypertonicity (51). Under hypertonic conditions, the AQP1 promoter activity is regulated by a novel hypertonicity response element that is different from TonE (26). Although the TonE motif was found at Ϫ1498 in the 5Ј-flanking region of AQP4 exon 0 and luciferase activity of the Ϫ1625/ϩ29 construct was significantly increased under hyperosmotic conditions, its activity was not maximal. The Ϫ428/ ϩ29 construct was maximally induced, suggesting that a critical cis element for hyperosmolarity is located between Ϫ345 and Ϫ428. On the other hand, a critical cis element of the AQP9 promoter for the hyperosmolarity response is suspected to be between Ϫ486 and Ϫ176, but no TonE sites are present in this region. The hypertonicity response element of the AQP1 promoter region is not present in the AQP4 and AQP9 promoter regions either, suggesting that under hyperosmotic conditions AQP4 and AQP9 promoter activities are regulated by another, as yet unidentified response element. In addition, the element responsible for p38 MAPK induction has not been elucidated. Further investigation is needed to clarify the elements that regulate the expression of AQP4 and AQP9 under hyperosmotic conditions.
With respect to the intracellular signal pathways regulating the expression of AQP4 and AQP9 in rat astrocytes, it has been demonstrated that signal transduction via protein kinase A and protein kinase C play important roles. Treatment of the cells with a protein kinase A activator increased the expression of AQP9 in a time-and concentration-dependent manner, but the expression of AQP4 was not changed (42). Although treatment of the cells with a protein kinase C activator caused decreases in the expression of AQP4 and AQP9 in a time-and concentration-dependent manner, prolonged treatment prevented subsequent decreases (43,52). We present a new finding that the expression of AQP4 and AQP9 is regulated by p38 MAPK.
Because AQP4 and AQP9 in astrocytes play important roles maintaining brain homeostasis, elucidating the intracellular signal pathways that regulate their expression is important. Among them, p38 MAPK is activated in astrocytes after brain ischemia (53) or injury (54,55). Other studies demonstrated that the expression of AQP4 and AQP9 in astrocytes changes under similar conditions (32)(33)(34)(35). The increase of AQP4 and AQP9 due to hyperosmotic mannitol, which is commonly administered to reduce brain edema, was regulated by p38 MAPK. Clarification of the detailed relationship between AQPs and p38 MAPK in astrocytes may lead to the control of water movements and new treatments for brain edema.