Hypoxia-induced regulation of MAPK phosphatase-1 as identified by subtractive suppression hybridization and cDNA microarray analysis.

Subtractive suppression hybridization was used to generate a cDNA library enriched in cDNA sequences corresponding to mRNA species that are specifically up-regulated by hypoxia (6 h, 1% O(2)) in the oxygen-responsive pheochromocytoma cell line. The dual specificity protein-tyrosine phosphatase MAPK phosphatase-1 (MKP-1) was highly represented in this library. Clones were arrayed on glass slides to create a hypoxia-specific cDNA microarray chip. Microarray, northern blot, and western blot analyses confirmed that MKP-1 mRNA and protein levels were up-regulated by hypoxia by approximately 8-fold. The magnitude of the effect of hypoxia on MKP-1 was approximately equal to that induced by KCl depolarization and much larger than the effects of either epidermal growth factor or nerve growth factor on MKP-1 mRNA levels. In contrast to the calcium-dependent induction of MKP-1 by KCl depolarization, the effect of hypoxia on MKP-1 persisted under calcium-free conditions. Cobalt and deferoxamine also increased MKP-1 mRNA levels, suggesting that hypoxia-inducible factor proteins may play a role in the regulation of MKP-1 by hypoxia. Pretreatment of cells with SB203580, which inhibits p38 kinase activity, significantly reduced the hypoxia-induced increase in MKP-1 RNA levels. Thus, hypoxia robustly increases MKP-1 levels, at least in part through a p38 kinase-mediated mechanism.

Subtractive suppression hybridization was used to generate a cDNA library enriched in cDNA sequences corresponding to mRNA species that are specifically upregulated by hypoxia (6 h, 1% O 2 ) in the oxygen-responsive pheochromocytoma cell line. The dual specificity protein-tyrosine phosphatase MAPK phosphatase-1 (MKP-1) was highly represented in this library. Clones were arrayed on glass slides to create a hypoxia-specific cDNA microarray chip. Microarray, northern blot, and western blot analyses confirmed that MKP-1 mRNA and protein levels were up-regulated by hypoxia by ϳ8-fold. The magnitude of the effect of hypoxia on MKP-1 was approximately equal to that induced by KCl depolarization and much larger than the effects of either epidermal growth factor or nerve growth factor on MKP-1 mRNA levels. In contrast to the calcium-dependent induction of MKP-1 by KCl depolarization, the effect of hypoxia on MKP-1 persisted under calcium-free conditions. Cobalt and deferoxamine also increased MKP-1 mRNA levels, suggesting that hypoxia-inducible factor proteins may play a role in the regulation of MKP-1 by hypoxia. Pretreatment of cells with SB203580, which inhibits p38 kinase activity, significantly reduced the hypoxia-induced increase in MKP-1 RNA levels. Thus, hypoxia robustly increases MKP-1 levels, at least in part through a p38 kinase-mediated mechanism.
Hypoxia is a critical physiological stimulus in a variety of disease states, including ischemia, respiratory disorders, and tumorigenesis (1,2). In recent years, the mechanisms by which cells respond and adapt to decreased O 2 levels have begun to be elucidated. For example, the hypoxia-inducible factor (HIF) 1 family of proteins includes transcription factors that are stabilized and activated specifically under conditions of low O 2 (for review, see Ref. 3). Upon activation by hypoxia, the HIF-␣/␤ heterodimer can enhance expression of genes that contain the hypoxia response element motif (5Ј-RCGTG-3Ј) in their 5Јflanking regions, such as vascular endothelial growth factor, erythropoietin, glucose transporter-1, and many others (see Ref. 3). Other transcription factors that can be activated by hypoxia include the cAMP response element-binding protein, c-Fos, JunB, Elk-1, and nuclear factor-B (4 -10). The intracellular signaling pathways that are modulated by hypoxia have also begun to be characterized, and these include calcium-dependent signaling pathways, mitogen-activated protein kinase (MAPK), the p38 stress-activated protein kinases (SAPKs), Akt, and Src (6,(11)(12)(13)(14)(15).
Thus, a growing number of hypoxia-responsive genes have now been identified. However, the coordinated processes by which these genes mediate the whole cellular response to hypoxia are virtually unknown. A better understanding of the cellular and molecular mechanisms by which cells respond to and adapt to a reduction in O 2 levels would provide important insight toward developing useful therapies against hypoxiarelated disorders. To address this issue, we have used subtractive suppression hybridization (SSH) to generate a cDNA library enriched in transcripts that are specifically regulated by hypoxia in pheochromocytoma PC12 cells. Coupled with cDNA microarray analysis, this study represents the first step toward delineating the global gene expression profile that is regulated by hypoxia.
One of the genes that was most frequently represented in this SSH library was identified as MAPK phosphatase-1 (MKP-1; also termed 3CH134 and CL100) (16,17). This phosphatase is one member of a family of dual specificity phosphatases or MKPs that oppose the effects of MAPKs and SAPKs (18 -20). Phosphorylation of MAPKs and SAPKs can be induced by a wide array of cellular stimuli (for review, see Refs. 21 and 22). Upon phosphorylation in specific Thr-X-Tyr motifs, these enzymes become activated and can translocate to the nucleus and phosphorylate various transcription factors, thereby regulating gene expression. The MKP enzymes are capable of dephosphorylating both phosphothreonine and phosphotyrosine in Thr-X-Tyr motifs, such as those found in MAPKs and SAPKs. Thus, MKPs oppose the effects of MAPKs and SAPKs. Activation of MAPKs and SAPKs is frequently associated with activation of MKPs, suggesting that MKPs play a role in feedback control of MAPK signaling (23).
MKPs can be generally classified as being primarily localized either in the nucleus (MKP-1 and MKP-2) or in the cytosol (MKP-3, MKP-4, MKP-5, and M3/6) (19). The nuclear MKPs are highly inducible and are considered to be immediate-early genes. Recently, it has been shown that the physical interaction of MAPKs or SAPKs with MKPs can stimulate the cata-lytic activity of both cytosolic and nuclear MKPs (24 -27). It has also been suggested that an increase in MKP gene expression represents another level of negative feedback regulation on MAPK signaling pathways (23,28).
In previous studies, we (13,14) and others (6,29) have shown that MAPKs and certain SAPKs are activated in response to hypoxia. Here, we employ SSH coupled with cDNA microarray analysis to demonstrate that MKP-1 is strongly induced by hypoxia in PC12 cells. We also show that this occurs in a calcium-independent manner and that the p38 SAPKs are involved in mediating this effect.

EXPERIMENTAL PROCEDURES
Cell Culture-PC12 cells were cultured as described previously (4). HepG2 and Hep3B cells were grown in the same culture medium used for PC12 cells. HEK293 and COS-7 cells were grown in Dulbecco's modified Eagle's medium supplemented with 15 mM Hepes, pH 7.4, 10% fetal bovine serum, 100 units/ml penicillin, and 100 g/ml streptomycin. In some experiments, cells were switched to serum-free Dulbecco's modified Eagle's medium or serum-free Dulbecco's modified Eagle's medium formulated in the absence of calcium (Life Technologies, Inc.) and supplemented with 1 mM EGTA as described previously (4). Prior to experimentation, cells were grown to ϳ80% confluence on 100-mm plates in an environment of 21% O 2 and 5% CO 2 . Hypoxia was delivered in an O 2 -regulated incubator (Forma Science, Inc., Marietta, OH) as described previously (30).
Drug Treatments-In some experiments, cells were switched to serum-free medium and treated with various drugs (or the corresponding vehicle) for 1 h prior to the initiation of hypoxia. BAPTA-AM, cobalt chloride, and deferoxamine mesylate were obtained from Sigma. Wortmannin, PD098059, and SB203580 were all obtained from Calbiochem. Nerve growth factor (NGF) (Alomone Labs, Jerusalem, Israel) and epidermal growth factor I (EGF) (Calbiochem) were applied for 20 min under normoxic conditions at final concentrations of 50 ng/ml. When peptide growth factors were used, experiments were performed in serum-free medium supplemented with 0.1% bovine serum albumin in the presence or absence of the indicated growth factors.
RNA Preparation-RNA was isolated from PC12 cells with Trireagent TM (Molecular Research Center, Inc., Cincinnati, OH), essentially according to the suggested protocol, with the addition of two extra acid phenol/chloroform extractions following the single chloroform extraction recommended by the manufacturer. This modification significantly improved the quality of RNA that was recovered. RNA was resuspended in nuclease-free water and quantified with a spectrophotometer as the average of triplicate absorbance readings at 260 nm. RNA quality was verified by visualization of 20 g on formaldehyde-containing 1% agarose gels containing SYBR ® Green II (Molecular Probes, Inc., Eugene, OR). For cDNA library construction, mRNA was isolated from total RNA using the Oligotex Direct mRNA minikit (QIAGEN Inc., Valencia, CA) according to the manufacturer's protocol. For real-time polymerase chain reaction (PCR), RNA was treated with amplification-grade DNase I (Life Technologies, Inc.) to remove genomic DNA.
cDNA Library Construction-A custom-subtracted cDNA library was constructed using the PCR-Select TM cDNA subtraction kit (CLON-TECH, Palo Alto, CA) according to the manufacturer's protocol. Briefly, PC12 cells were exposed to hypoxic or normoxic conditions for 6 h, and mRNA was isolated as described above. Double-stranded cDNA was generated and subjected to restriction digest with RsaI. Following the digest, the hypoxic (tester) sample was split into two pools. Each pool was ligated with a different adaptor (N1 or N2R). Each ligated pool was then denatured and hybridized with an excess of denatured normoxic (driver) cDNA. The hybridized pools were mixed, and a second round of hybridization was performed with an excess of denatured driver. The population of hybrid molecules that contains both adaptors (N1 and N2R) is the population that represents the differentially expressed tester sequences. The entire population of hybridized molecules was subjected to PCR to amplify these desired sequences. In this PCR, driver sequence does not get amplified because it has no adaptors. Tester/driver hybrids are only linearly amplified because they contain adaptor at only one end. Tester/tester hybrids that have the same adaptor at both ends will form hairpin loops under the conditions used and will not be amplified. The amplified fragments were then ligated into the pCR ® 2.1-TOPO ® vector (Invitrogen, Carlsbad, CA). Ligated clones were electroporated into DH10B cells. The library was amplified in LB medium containing 50 g/ml kanamycin (growth medium), and the titer was determined. A portion of the library was plated, and colonies were picked into 96-well microtiter plates containing growth medium. Copies of the library were stored in this format at Ϫ80°C as a glycerol stock. The entire library was subjected to DNA sequencing by MWG Biotech, Inc. (High Point, NC) using the M13 reverse primer.
Microarray Production-Clone inserts were amplified by bacterial PCR using either M13 forward (Ϫ40) and reverse primers or primers to the adaptor sequences from the PCR-Select TM kit. PCR products were isopropyl alcohol-precipitated and resuspended in 3ϫ SSC. The final concentration of each PCR product was 0.2-1 g/l. PCR products were spotted onto poly-L-lysine-coated slides using an OmniGrid TM robot (GeneMachines, San Carlos, CA). Poly-L-lysine slides were either prepared by the method of Brown et al. (31) or purchased from CEL Associates (Houston, TX); there was no difference in quality between the two types of slides. Slides were post-processed using the succinic anhydride method (31) and stored at room temperature in a desiccator cabinet until used.
Microarray Probe Labeling and Hybridization-Probes for cDNA microarrays were generated using 100 g of total RNA from cells exposed to normoxia (21% O 2 ) or hypoxia (1% O 2 ) in a standard reverse transcriptase reaction in which some of the dTTP was replaced with either 50 M Cy3-labeled dUTP or 75 M Cy5-labeled dUTP (Amersham Pharmacia Biotech). In some experiments, the hypoxic sample was labeled with Cy3; and in others, it was labeled with Cy5, with essentially identical results. Probes were cleaned using the QIAquick nucleotide removal kit (QIAGEN Inc.). Probes were combined and hybridized to the array overnight at 58°C in buffer containing 0.57 g/l COT-1 DNA, 0.57 g/l (dA) 40 -60 , 0.23 g/l yeast tRNA, 3.5ϫ SSC, and 0.3% SDS. Slides were washed in the following buffers at room temperature: 1) 10 min in 2ϫ SSC and 0.2% SDS; 2) 5 min in 1ϫ SSC and 0.2% SDS; 3) 1 min in 2ϫ SSC; and 4) 1 min in 0.05ϫ SSC. Slides were then dried by centrifugation at room temperature and scanned immediately.
Microarray Data Analysis-Slides were scanned with a GenePix 4000A scanner (Axon Instruments, Inc., Foster City, CA) at 532 nm (Cy3) and 635 nm (Cy5) simultaneously. The images were analyzed using GenePix Version 2.0 software. The background-subtracted median ratio value was calculated for each spot, and replicate spots on each slide were averaged.
Northern Blots-RNA was isolated as described above, transferred to nylon membranes (Hybond TM -N ϩ , Amersham Pharmacia Biotech), and subjected to UV cross-linking. Membranes were stained with methylene blue to ensure quantitative transfer of the RNA to the membrane. Membranes were then prehybridized in a solution containing 1.0% SDS and 0.1 M NaCl in diethyl pyrocarbonate-treated water for a minimum of 1 h at 42°C in a rotating hybridization tube.
One of the fragments identified as a partial cDNA sequence of rat MKP-1 was a 243-base pair RsaI fragment corresponding to nucleotides 1498 -1740 of this gene (99.4% identity to GenBank TM /EBI Data Bank accession number X84004). This fragment was released from the pCR2.1-TOPO vector by enzymatic digestion with RsaI and then excised and purified from a 1.2% agarose gel. The insert (25 ng) was labeled with deoxycytidine 5Ј-[ 32 P-]trisphosphate (PerkinElmer Life Sciences) by the random priming method (Prime-a-Gene, Promega, Madison, WI). Probes containing 1 ϫ 10 7 cpm were added to 10 ml of high efficiency system hybridization solution with 50% formamide (Molecular Research Center, Inc.). Blots were hybridized for a minimum of 18 h at 42°C in a rotating hybridization tube. Blots were washed three times in 15 ml of 1ϫ SSC (3.0 M sodium chloride and 0.3 M sodium citrate, pH 7.0) and 1.0% SDS in diethyl pyrocarbonate-treated aqueous solution. mRNA signals were detected and quantified using a Phosphor-Imager (Molecular Dynamics, Inc., Sunnyvale, CA).
Real-time PCR-Cells were exposed to normoxia or hypoxia, and total RNA was isolated as described above. First-strand cDNA synthesis was performed using the SuperScript TM first-strand synthesis system for real-time PCR (Life Technologies, Inc.) with oligo(dT) as the primer according to the manufacturer's directions. Real-time PCR was performed in a Smart Cycler (Cepheid, Sunnyvale, CA) using the Light-Cycler DNA Master SYBR Green I dye intercalation assay (Roche Molecular Biochemicals). Primers (forward, 5Ј-TGAACTCAGCACAT-TCGGGACC-3Ј; and reverse, 5Ј-AGGGGCGAGCAAAAAGAAACC-3Ј) were generated to human MKP-1 (GenBank TM /EBI Data Bank accession number XM_003720) and used to amplify a 113-base pair fragment. Measurements were taken at the end of the 72°C extension step in each cycle, and the second-derivative method was used to calculate threshold cycle. Melt curve analysis showed a single sharp peak for all samples.

RESULTS
The sequence data obtained from the SSH library revealed that the library contained 200 different genes that corresponded to known sequences in the current public data bases. A number of sequences in the library were found to be present in multiple copies, including JunB (14 copies), tyrosine hydroxylase (five copies), and vascular endothelial growth factor (three copies), which are all genes that are strongly regulated by hypoxia (7,30,32). Six copies of the rat homolog of MKP-1, also known as CL100 or 3CH134, were also found in this library.
As a first step toward verifying regulation of the genes in the SSH library, PCR products derived from each clone in the library were spotted onto glass slides and evaluated by cDNA microarray analysis (Fig. 1) as described under "Experimental Procedures." cDNA microarray analysis revealed that tyrosine hydroxylase, JunB, and vascular endothelial growth factor were all strongly regulated by hypoxia (6 h, 1% O 2 ), as expected. These experiments also demonstrated that MKP-1 mRNA was similarly increased by an average of ϳ5-fold.
To further confirm that MKP-1 was regulated by hypoxia, MKP-1 mRNA levels were determined by northern blot analysis. These experiments revealed that MKP-1 mRNA was upregulated by an average of 8.6-fold in response to hypoxia (6 h, 1% O 2 ) in PC12 cells ( Fig. 2A). A similar increase was observed in MKP-1 immunoreactivity on western blots (Fig. 2B). Thus, hypoxia induces MKP-1 expression at both the mRNA and protein levels in PC12 cells. To determine whether or not this effect is specific to PC12 cells, the effect of hypoxia on MKP-1 expression levels was examined in HepG2, Hep3B, HEK293,   FIG. 1. MKP-1 mRNA levels are upregulated by hypoxia in microarray analysis. PC12 cells were exposed to normoxia or hypoxia (1% O 2 ) for 6 h, and total RNA was isolated as described under "Experimental Procedures." A, comparison of control versus control gene expression pattern. Cy3-and Cy5-labeled probes were each generated from RNA derived from cells maintained in normoxia. Probes were hybridized to arrays as described under "Experimental Procedures." Background-subtracted median pixel intensities are plotted from a representative experiment. Each symbol represents one spot on the array. B, comparison of control versus hypoxia gene expression pattern. The Cy5-labeled probe was generated from RNA derived from cells maintained in normoxia, whereas the Cy3-labeled probe was generated from RNA derived from cells exposed to hypoxia. The data shown are representative of results obtained in three separate hybridization experiments. and COS-7 cell lines. The probe used for northern blotting was not sensitive enough to detect MKP-1 mRNA in these cell lines, so real-time PCR was used instead. Comparison of the threshold cycles for normoxic versus hypoxic samples showed a significant increase in MKP-1 mRNA expression in response to hypoxia in the HepG2 (20.89 Ϯ 0.22 for normoxic versus 16.73 Ϯ 0.32 for hypoxic, p ϭ 9e Ϫ5 ) and Hep3B (20.19 Ϯ 0.50 for normoxic versus 18.07 Ϯ 0.24 for hypoxic, p ϭ 0.004) cell lines. This threshold cycle difference corresponds to a 13.89-fold upregulation of MKP-1 in HepG2 cells and a 7.06-fold up-regulation of MKP-1 in Hep3B cells (Fig. 2C). There was a slight increase in MKP-1 mRNA levels in HEK293 cells and a slight decrease in MKP-1 mRNA in COS-7 cells, but these differences did not achieve statistical significance.
In other experiments, PC12 cells were exposed to hypoxia for various times between 20 min and 18 h (Fig. 3). The earliest time at which MKP-1 levels were elevated in response to hypoxia was 1 h. The maximal effect of hypoxia on MKP-1 mRNA levels occurred between 3 and 6 h of exposure to hypoxia. PC12 cells were also exposed to a range of oxygen levels between 21% (normoxia) and 1% O 2 for 4 h. These experiments showed that the effects of hypoxia occurred in a dose-dependent manner, with a modest effect at 10% and a maximal effect at 1% O 2 (Fig. 4).
To compare the effect of hypoxia on the regulation of MKP-1 with that of other stimuli, PC12 cells were treated with hypoxia, KCl, EGF, or NGF. As shown in Fig. 5A, the effects of both hypoxia and KCl on MKP-1 levels were quite robust, averaging 7.6-and 6.9-fold over basal levels, respectively (Fig.  5B). The effects of CoCl 2 and deferoxamine, two agents that mimic hypoxia, were also tested. CoCl 2 (Fig. 5C) and deferox-amine (data not shown) also increased MKP-1 mRNA levels to a similar extent compared with hypoxia. In contrast, the effects of EGF and NGF on MKP-1 levels were modest in this cell type, averaging only 28 and 47% increases over control levels, respectively.
A series of experiments designed to identify the signaling

FIG. 2. Hypoxia increases MKP-1 mRNA and protein levels.
A and B, PC12 cells were exposed to either normoxia (21% O 2 ; control (C)) or hypoxia (1% O 2 ; H) for 6 h as indicated. A, RNA was isolated and subjected to northern blot analysis as described under "Experimental Procedures." B, whole cell lysates were isolated and subjected to western blot analysis as described under "Experimental Procedures." Representative blots are shown. C, cells were exposed to hypoxia or normoxia for 4 h. RNA was isolated and subjected to real-time PCR analysis as described under "Experimental Procedures." MKP-1 RNA levels are expressed as -fold change in the hypoxic sample compared with the corresponding normoxic sample (mean Ϯ S.E., n ϭ 4).  pathways involved in the hypoxia-induced regulation of MKP-1 were then performed. To test whether the induction of MKP-1 expression is Ca 2ϩ -dependent, PC12 cells were incubated in Ca 2ϩ -free medium supplemented with 1 mM EGTA and then exposed to either normoxia (21% O 2 ) or hypoxia (6 h, 1% O 2 ). As shown in Fig. 6, the hypoxia-induced increase in MKP-1 mRNA persisted in the presence or absence of extracellular Ca 2ϩ . The possibility that MKP-1 was induced by the release of Ca 2ϩ from intracellular stores was next tested. Both extracellular and intracellular Ca 2ϩ were removed by preloading cells with BAPTA-AM, a membrane-permeable calcium chelator, and then incubating cells in Ca 2ϩ -free medium during exposure to hypoxia. As a control, cells that were depolarized with KCl were also included in this experiment. As shown in Fig. 7, the hypoxia-induced increase in MKP-1 persisted in the absence of both intracellular and extracellular calcium, whereas the effect of KCl depolarization was completely blocked under calciumfree conditions.
Next, it was of interest to determine whether stimulation of MKP-1 expression by hypoxia is dependent on enzymatic activation of MAPK. PC12 cells were pretreated with vehicle or PD98059, a specific inhibitor of MEK-1, the kinase directly upstream of MAPK in the Ras/Raf/MEK/MAPK signaling cascade. As shown Fig. 8 (A and B), MKP-1 expression was not diminished by pretreatment with PD98059. In fact, there was a tendency toward a greater effect of hypoxia on MKP-1 levels in the cells that were pretreated with PD98059 compared with vehicle, although this effect did not achieve statistical significance (p Ͼ 0.1). The PD98059 pretreatment was efficacious in that it blocked the hypoxia-induced increase in phospho-MAPK immunoreactivity, as shown in Fig. 8 (C and D).
To determine whether the p38 pathway might be involved in the regulation of MKP-1, the effect of SB203580 on the hypoxia-induced increase in MKP-1 RNA levels was tested. Prior to exposure to hypoxia, PC12 cells were treated either with SB203580, an inhibitor of the p38␣ and p38␤ isoforms of the p38 family of protein kinases, or with vehicle. As shown in Fig.  9, although hypoxia still induced a significant increase in MKP-1 mRNA levels in cells that were pretreated with SB203580 (3.5-fold above control levels), this effect was significantly diminished compared with the effect of hypoxia on MKP-1 levels in cells that were pretreated with vehicle (7.5-fold).
Finally, a role for the phosphatidylinositol 3Ј-kinase (PI3K) signaling pathway in the regulation of MKP-1 by hypoxia was investigated. In a previous study by this laboratory, it was demonstrated that Akt, a protein kinase downstream of PI3K, is activated by hypoxia (15). The hypoxia-induced activation of Akt is blocked by wortmannin (15), a selective inhibitor of PI3K. Although wortmannin treatment abolished the hypoxiainduced phosphorylation of Akt (Fig. 10, C and D), the hypoxiainduced increase in MKP-1 levels persisted in the presence of wortmannin (A and B). DISCUSSION The mechanisms by which cells respond to and adapt to changes in oxygen levels are not well understood. To develop a better view of the coordinated pattern of genes that are regulated by hypoxia, we utilized SSH coupled with cDNA microarray analysis. One of the genes most frequently represented in the library and highly regulated on the cDNA microarrays corresponded to the dual specificity protein-tyrosine phosphatase MKP-1. This enzyme plays in important role in modulating both the MAPK and SAPK signaling pathways (20, 29), and increased MKP-1 expression has been implicated in PC12 cell survival (33).
There has been some debate about the signaling mechanisms by which MKP-1 gene expression is regulated. Several studies have implicated calcium as playing a critical role in the regulation of MKP-1 gene expression (28,34,35). However, other studies have suggested that the SAPKs are also involved in MKP-1 expression (36,37). Thus, it was of considerable interest to determine which, if any, of these pathways induce MKP-1 gene expression under conditions of hypoxia.
Northern and western blot analyses verified that MKP-1 mRNA and protein levels were dramatically up-regulated by hypoxia in PC12 cells. The effect of hypoxia on MKP-1 was robust (8-fold after 4 h in 1% O 2 ) and comparable in magnitude to that induced by depolarization with KCl. The effects of hypoxia and KCl were much greater than those of EGF and NGF, which have been previously shown to induce only modest increases in MKP-1 expression in PC12 cells (38). Hypoxia also increased MKP-1 mRNA levels in two hypoxia-responsive cell lines, HepG2 and Hep3B, whereas the levels in COS-7 and HEK293 cells remained unchanged. Therefore, up-regulation of MKP-1 appears to be part of a generalized response to hypoxia in certain hypoxia-responsive cell types. It is interesting to note that the effects of hypoxia occurred in a dose-dependent manner, with a modest effect at 10% and a maximal effect at 1% O 2 . This contrasts with our previous results showing that the effects of hypoxia on tyrosine hydroxylase gene expression are maximal at 5% O 2 (30). This suggests that genes may be differentially responsive to various degrees of hypoxia.
One of the earliest known responses to hypoxia in PC12 cells is depolarization and an elevation of intracellular calcium levels (11). Because agents that elevate intracellular calcium levels have been shown to increase MKP-1 expression and because several previous studies have reported that MKP-1 is a Ca 2ϩinduced protein phosphatase (28,34,35), we investigated whether calcium is critical for the hypoxia-induced regulation of MKP-1. Interestingly, the increase in MKP-1 mRNA induced by hypoxia was completely unaffected by the removal of extracellular calcium or both extracellular and intracellular calcium. This is in striking contrast to the effects of KCl-induced depolarization on MKP-1, which were abolished in the absence of calcium. Thus, although calcium can regulate MKP-1 gene expression under certain conditions (28,34,35), hypoxia clearly induces MKP-1 mRNA in PC12 cells in a calciumindependent manner. It has been suggested that induction of MKP-1 gene expression occurs as a compensatory response to activation of MAPK (18,20). Furthermore, previous studies in our laboratory have shown that hypoxia activates MAPK in PC12 cells with a time course similar to that of the regulation of MKP-1 mRNA by hypoxia (13,14). To determine whether activation of MAPK is a prerequisite for up-regulation of MKP-1 mRNA, PC12 cells were pretreated with PD98059, a specific inhibitor of MEK, the upstream activa- (ϩ) and exposed to either normoxia or hypoxia as described under "Experimental Procedures." A, a representative MKP-1 northern blot is shown. B, MKP-1 RNA levels are expressed as average -fold induction Ϯ S.E. (n ϭ 5-6 in each group). *, significantly different from the control (p Ͻ 0.01 by independent t test); #, not significantly different from hypoxia in the absence of PD98059 (p Ͼ 0.1 by independent t test). C and D, separate dishes of cells were identically pretreated with either vehicle (Ϫ) or 50 M PD098059 (ϩ) and exposed to either normoxia or hypoxia. Whole cell lysates were harvested and subjected to immunoblotting with either an antibody that specifically recognizes phospho-p42/p44 MAPK (C) or an antibody that equally recognizes phospho-and dephospho-MAPK (total MAPK) (D). tor of MAPK. Although PD98059 blocked hypoxia-induced phosphorylation of MAPK, it did not alter the effect of hypoxia on MKP-1. Therefore, the hypoxia-induced activation of MAPK is not essential for the increase in MKP-1 RNA levels. These findings are consistent with those of other studies, in which inhibition of MEK was also insufficient to prevent the induction of MKP-1 gene expression by various stimuli (38,39). Another intracellular signaling system that is stimulated by hypoxia is the PI3K/Akt pathway. Akt (also termed protein kinase B) is a cytosolic serine/threonine protein kinase that has been shown to be critical for cell survival under adverse conditions (40,41). In an earlier study, we showed that Akt is activated by hypoxia in PC12 cells and that this effect is blocked by wortmannin (an inhibitor of PI3K, an upstream activator of Akt) (15). However, the effect of hypoxia on MKP-1 persisted in the presence of wortmannin, indicating that this effect is independent of the PI3K/Akt signaling pathway.
A significant finding in this study was that the effects of hypoxia on MKP-1 were markedly attenuated by SB203580, a specific inhibitor of the p38 family of SAPKs. Increasing evidence suggests that activation of p38 may have both apoptosispromoting and cell-protective functions, depending on the particular cellular context (42)(43)(44)(45)(46)(47). To date, there are five known members of the p38 family of SAPKs (for review, see Ref. 22). The p38␣ and p38␤ subtypes are sensitive to inhibition by SB203580, but the p38␥ and p38␦ subtypes are not (47)(48)(49)(50). Importantly, we have previously shown that both the p38␣ and p38␥ subtypes are activated by hypoxia (13). Of these, the SB203580-insensitive p38␥ isoform is most strongly activated in response to hypoxia in PC12 cells (13). Therefore, it is possible that the remaining ϳ3.5-fold increase in MKP-1 mRNA (i.e. that which is not inhibited by SB203580) is mediated by the SB203580-insensitive p38␥ isoform. Indeed, the time course of activation of p38␥ by hypoxia in PC12 cells closely parallels that of induction of MKP-1 mRNA (see Ref. 13 and Fig. 3A). However, at present, in the absence pharmacological inhibitors of the p38␥ isoforms, we are unable to test this hypothesis directly. Alternatively, other as yet unidentified (non-p38) signaling pathways may also play a role in the regulation of MKP-1 by hypoxia.
Previous studies have shown that MKP-1 and other MKPs can dephosphorylate and thereby inactivate MAPKs and SAPKs (20,25,27). It has recently been shown that MKP-1 associates directly with p38 and that this interaction enhances the catalytic activity of MKP-1 (25). It has also been suggested that MKP-1 dephosphorylates the p38 kinase in cardiomyocytes (51). Although we have not measured this directly in our study, it is reasonable to assume that MKP-1 also dephosphorylates p38 in response to hypoxia as a negative feedback response. This hypothesis is consistent with the finding that that the hypoxia-induced phosphorylation of p38␣ and p38␥ peaks at 6 h and tapers off thereafter (13), 2 whereas MKP-1 levels remain elevated during at least 18 h of exposure to FIG. 9. Effects of hypoxia on MKP-1 are at least partially mediated by a p38-dependent mechanism. PC12 cells were pretreated for 1 h in serum-free medium with either vehicle (Ϫ) or 2 M SB203580 (ϩ) and exposed to either normoxia or hypoxia as described under "Experimental Procedures." A, a representative MKP-1 northern blot is shown. B, MKP-1 RNA levels are expressed as average -fold induction Ϯ S.E. (n ϭ 5-10 in each group). *, significantly different from the control (p Ͻ 0.01 by independent t test); †, significantly different from hypoxia in the absence of SB203580 (p Ͻ 0.01 by independent t test).
FIG. 10. Induction of MKP-1 is not dependent on the PI3K signaling pathway. PC12 cells were pretreated for 1 h in serum-free medium with either vehicle (Ϫ) or 100 nM wortmannin (ϩ) and exposed to either normoxia (control (C)) or hypoxia (H) as described under "Experimental Procedures." A, a representative MKP-1 northern blot is shown. B, MKP-1 RNA levels are expressed as average -fold induction Ϯ S.E. (n ϭ 6 in each group). *, significantly different from the control (p Ͻ 0.01 by independent t test); #, not significantly different from hypoxia in the absence of wortmannin (p Ͼ 0.1 by independent t test). C and D, separate dishes of cells were identically pretreated with either vehicle (Ϫ) or 100 nM wortmannin (ϩ) and exposed to either normoxia or hypoxia. Whole cell lysates were harvested and subjected to immunoblotting with either an antibody that specifically recognizes phospho-Akt (C) or an antibody that equally recognizes phospho-and dephospho-Akt (total Akt) (D). hypoxia. Our findings suggest, for the first time, that the p38 family of protein kinases can reciprocally up-regulate MKP-1 gene expression in response to hypoxia. This is consistent with a previous report showing that insulin-induced stimulation of MKP-1 mRNA in vascular smooth muscle cells can be inhibited by SB203580 (52), although this study reported that PD98059 also blocked the insulin-induced increase in MKP-1 levels.
In addition, we demonstrated that cobalt chloride and deferoxamine increase MKP-1 mRNA levels to a similar extent compared with hypoxia. Both agents have been shown to mimic the effects of hypoxia (53)(54)(55), in part by increasing HIF-1␣ binding activity and protein levels. This suggests that the hypoxia-induced increase in MKP-1 mRNA is dependent upon HIF activation. In addition, HIF-1␣ has recently been reported to be activated by phosphorylation of its regulatory domain by p38 (56). Thus, although there may be cell-specific differences in the mechanism of regulation, HIF-1␣, the p38 protein kinases, and MKP-1 appear to be linked in a complex pattern of co-regulation. Our study illustrates that the combined use of SSH libraries and cDNA microarray analysis provides a powerful approach to delineate the patterns of gene expression that are regulated by hypoxia and other environmental stimuli.