Hypoxia Induces Phosphorylation of the Cyclic AMP Response Element-binding Protein by a Novel Signaling Mechanism*

To investigate signaling mechanisms by which hypoxia regulates gene expression, we examined the effect of hypoxia on the cyclic AMP response element-binding protein (CREB) in PC12 cells. Exposure to physiological levels of hypoxia (5% O2, ∼50 mm Hg) rapidly induced a persistent phosphorylation of CREB on Ser133, an event that is required for CREB-mediated transcriptional activation. Hypoxia-induced phosphorylation of CREB was more robust than that induced by any other stimulus tested, including forskolin, depolarization, and osmotic stress. Furthermore, this effect was not mediated by any of the previously known signaling pathways that lead to phosphorylation of CREB, including protein kinase A, calcium/calmodulin-dependent protein kinase, protein kinase C, ribosomal S6 kinase-2, and mitogen-activated protein kinase-activated protein kinase-2. Hypoxic activation of a CRE-containing reporter (derived from the 5′-flanking region of the tyrosine hydroxylase gene) was attenuated markedly by mutation of the CRE. Thus, a physiological reduction in O2 levels induces a functional phosphorylation of CREB at Ser133 via a novel signaling pathway.

To investigate signaling mechanisms by which hypoxia regulates gene expression, we examined the effect of hypoxia on the cyclic AMP response element-binding protein (CREB) in PC12 cells. Exposure to physiological levels of hypoxia (5% O 2 , ϳ50 mm Hg) rapidly induced a persistent phosphorylation of CREB on Ser 133 , an event that is required for CREB-mediated transcriptional activation. Hypoxia-induced phosphorylation of CREB was more robust than that induced by any other stimulus tested, including forskolin, depolarization, and osmotic stress. Furthermore, this effect was not mediated by any of the previously known signaling pathways that lead to phosphorylation of CREB, including protein kinase A, calcium/calmodulin-dependent protein kinase, protein kinase C, ribosomal S6 kinase-2, and mitogenactivated protein kinase-activated protein kinase-2. Hypoxic activation of a CRE-containing reporter (derived from the 5-flanking region of the tyrosine hydroxylase gene) was attenuated markedly by mutation of the CRE.

Thus, a physiological reduction in O 2 levels induces a functional phosphorylation of CREB at Ser 133 via a novel signaling pathway.
Hypoxic/ischemic trauma is a primary factor in the pathology of many disease states. Even brief periods of localized oxygen deprivation can result in severe cellular and tissue damage, such as that produced by cerebral or myocardial infarction. Severe hypoxia results in depletion of cellular ATP levels and cessation of oxidative phosphorylation, which results in profound deficiencies in cellular function (1). It is therefore not surprising that sophisticated mechanisms have evolved which allow the cell to adapt to moderate levels of hypoxia long before ATP depletion occurs (2,3). In recent years, the mechanisms underlying adaptation of mammalian cells to hypoxia have begun to be elucidated. A major component of this adaptation is regulation of gene expression. A number of genes have been identified, including erythropoietin, vascular endothelial growth factor, hypoxia-inducible factor-1, and tyrosine hydroxylase (TH), 1 which are involved in the adaptive response to hypoxia (4 -7). However, the primary mechanism(s) by which cells sense changes in oxygen levels and transduce this signal into the molecular events associated with changes in gene expression remain unknown.
Rat pheochromocytoma (PC12) cells are a catecholaminergic cell line that has proven to be a useful system to study hypoxiaregulated gene expression. An acute reduction in oxygen tension triggers a variety of cellular responses in PC12 cells, including depolarization (8) and inhibition of an oxygen-sensitive K ϩ channel conductance (9). Prolonged (Ͼ3 h) exposure to hypoxia also leads to an induction of TH gene expression and mRNA stability (10) and a stimulation of the immediate early genes c-fos and junB (6) in PC12 cells. The specific intracellular signaling pathways by which reduced oxygen stimulates expression of these genes are not understood.

EXPERIMENTAL PROCEDURES
Cell Culture-All tissue culture reagents were obtained from Life Technologies, Inc. PC12 cells, obtained from the American Type Culture Collection, were cultured in Dulbecco's modified Eagle's medium and Ham's F-12 medium supplemented with 15 mM HEPES, pH 7.4, 10% fetal bovine serum, and with penicillin (100 units/ml) and streptomycin (100 g/ml). PKA-deficient (123.7) PC12 cells, kindly provided by Dr. J. A. Wagner (Cornell University Medical College, New York), were grown in high glucose Dulbecco's modified Eagle's medium supplemented with 15 mM HEPES, pH 7.4, 10% fetal bovine serum, 5% heat-inactivated horse serum, and 0.1 mg/ml G418. To compare wildtype PC12 cells with 123.7 cells, PC12 cells were grown in the same media in the absence of G418. When cells reached 85-90% confluence in 35-mm tissue culture dishes (Corning), they were exposed either to continued normoxia or placed in an oxygen-regulated incubator (Forma Scientific, Marietta, OH) in an environment of 5% O 2 , 5% CO 2 , balanced with N 2 , for various times. In previous studies, we have shown that the partial pressure of oxygen in the media of cells exposed to 5% O 2 is in the range of 30 -50 mm Hg (10). For experiments in which cells were pretreated with drugs or vehicle, cells were switched to serum-free Dulbecco's modified Eagle's medium and Ham's F-12 medium containing either drugs at the indicated concentrations, or the corresponding vehicle, for the indicated times before the start of hypoxia.
Immunoblotting-For immunoblotting analysis, cells were harvested by scraping in 0.25 ml of 1% SDS and were sonicated briefly with a microtip ultrasonic cell disruptor (Kontes, Vineland, NJ). In some experiments, samples were solubilized in an ice-cold nondenaturing lysis buffer containing 10 mM Tris, pH 7.4, 1% Triton X-100, 0.2 mM sodium vanadate, 10 mM sodium fluoride, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 2 g/ml leupeptin, and 2 g/ml aprotinin. In these experiments, lysates were centrifuged 10 min at 14,000 ϫ g at 4°C to remove any Triton-insoluble material. Both solubilization protocols gave similar experimental results. Samples containing 50 -100 g of protein were then run on 9% SDS-polyacrylamide gels (Protogel, National Diagnostics, Atlanta, GA) and transferred to nitrocellulose membranes (Schleicher & Schuell) using standard electrophoresis and electroblotting procedures. Nitrocellulose membranes were blocked with 3% nonfat dry milk in a buffer containing 10 mM sodium phosphate, pH 7.2, 140 mM NaCl, and 0.1% Tween 20. Blots were then imunolabeled overnight at 4°C with antibodies specific for either Ser 133 phospho-CREB (1:1,000, New England Biolabs, Beverly, MA) or which recognize equally phospho-and dephospho-CREB (1:1,000, New England Biolabs). Immunolabeling was detected by enhanced chemiluminescence (ECL, Amersham Pharmacia Biotech) according to the manufacturer's recommended conditions. In some cases, blots were stripped and reprobed with another antibody. Blots were stripped by incubation for 1 h at 50°C in a solution containing 62.5 mM Tris-HCl, pH 6.7, 2% SDS, and 0.7% ␤-mercaptoethanol. Blots were then washed for 1 h in several changes of phosphate-buffered saline/Tween 20 at room temperature and probed with ECL to confirm that antibodies had been removed completely. Blots were then reblocked and immunolabeled as described above. Immunoreactivity was quantified using densitometric analysis with an ImagePro digital analysis system (Media Cybernetics, Silver Spring, MD). Immunoreactivity for CREB was found to be linear over a 5-fold range of protein concentrations.
CAT Assays-The TH-CAT reporter plasmids (Ϫ272TH)CAT and (Ϫ272CRE Ϫ )CAT (26), were generously provided by Dr. Dona Chikaraishi (Duke University, Durham, NC). PC12 cells were transfected in 35-mm dishes with 2 g of either (Ϫ272TH)CAT or (Ϫ272CRE Ϫ )CAT using LipofectAMINE, using the manufacturer's recommended conditions (Life Technologies, Inc.). Beginning 24 h after transfection, cells were exposed for 24 h to either hypoxia (5% O 2 ) or normoxia (21% O 2 ). Cells were then lysed in 0.5 ml, and samples containing 120 g of protein were analyzed in duplicate for CAT levels by enzyme-linked immunosorbent assay (Boehringer Mannheim, GmBH). Data are normalized per total g of protein in each sample. In each experiment, cells were transfected in quadruplicate dishes, and duplicate samples were analyzed for CAT activity. In each experiment, the variance in the range of CAT values was less than 15% from the mean value.
PKA Enzyme Assays-After exposure to hypoxia or normoxia, cells were assayed for PKA enzyme activity exactly as described previously (27). PKA specific activity was calculated as the difference in 8-bromo-cAMP-stimulated phosphorylation of Leu-Arg-Arg-Ala-Ser-Leu-Gly (Kemptide) and phosphorylation found in the presence of PKI 6 -22 , a specific inhibitor of PKA. 8-Bromo-cAMP-stimulated activity averaged 20-fold higher than that observed in the presence of PKI 6 -22 . In each experiment, PKA activity levels were normalized per g of protein. PKA enzyme activity was found to be linear over a 30-fold range of protein concentrations (between 0.5 and 15 g of protein/assay).
MAPKAP Kinase-2 Enzyme Assays-PC12 cells were grown to 90% confluence in 100-mm dishes. The medium was replaced with serumfree medium, and cells were exposed for various times, between 20 min and 6 h, to hypoxia (5% O 2 ) or 300 mM sorbitol. In some experiments, cells were pretreated for 1 h with 20 M SB203580 (kindly provided by Dr. John C. Lee, Smith Kline Beecham, King of Prussia, PA) before exposure to sorbitol. Cells were harvested by scraping in 0.4 ml of an ice-cold nondenaturing lysis buffer containing 10 mM Tris, pH 7.4, 1% Triton X-100, 0.2 mM sodium vanadate, 10 mM sodium fluoride, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 2 g/ml leupeptin, and 2 g/ml aprotinin. Lysates were incubated 20 min at 4°C and then centrifuged 10 min at 14,000 ϫ g to remove Triton-insoluble material. Aliquots containing 1 mg of total protein were immunoprecipitated for 2 h at 4°C with 5 g of an anti-MAPKAP kinase-2 polyclonal antibody (Upstate Biotechnology Inc., Lake Placid, NY) coupled to protein Gagarose beads (Santa Cruz Biotechnology, Santa Cruz, CA). MAPKAP kinase-2 enzyme activity was then assayed in an immunocomplex kinase assay, using 10 g of a specific MAPKAP kinase-2 substrate peptide from a MAPKAP kinase-2 immunoprecipitation kinase kit, exactly as described in the manufacturer's recommended protocol (Upstate Biotechnology).

RESULTS
Effect of Hypoxia on CREB Phosphorylation-PC12 cells were exposed to hypoxia (5% O 2 ) for various times between 0 and 24 h, as shown in Fig. 1. Whole cell extracts were immunolabeled with an antibody specific for Ser 133 -phosphorylated CREB. Hypoxia induced a dramatic increase in phospho-CREB immunoreactivity (Fig. 1A). The effect peaked at 6 h but persisted strongly at 24 h of exposure to hypoxia. This effect was not caused by changes in the total amount of CREB, as shown in Fig. 1B, where cell extracts were immunolabeled with an antibody that recognizes phospho-and dephospho-CREB equally. The effect of osmotic stress on CREB phosphorylation was also examined. In Fig. 1C, it can be seen that exposure of PC12 cells to 300 mM sorbitol also rapidly induces Ser 133 phosphorylation of CREB. Unlike hypoxia, which required several hours to produce a maximal effect, sorbitol-stimulated CREB phosphorylation was maximal at the earliest time point examined (20 min). Exposure of SK-N-MC cells to arsenite, another stressor, has been reported to induce CREB phosphorylation (18).
To evaluate the relative effect of hypoxia on CREB phosphorylation, we compared the ability of hypoxia to stimulate phosphorylation of CREB with that of forskolin and KCl-induced FIG. 1. Effect of hypoxia on CREB phosphorylation in PC12 cells. PC12 cells were exposed to either hypoxia (5% oxygen) or normoxia (controls, 21% oxygen) for various times, between 0 and 24 h, as indicated. Whole cell lysates were immunolabeled with an antibody that specifically recognizes Ser 133 phospho-CREB (panel A) or an antibody that equally recognizes phospho-and dephospho-CREB (panel B) as described under "Experimental Procedures." Panel C, phospho-CREB immunoreactivity in PC12 cells exposed to 300 mM sorbitol for various times between 0 and 6 h. Representative immunoblots are shown, with n ϭ 3-12 dishes of cells at each time point, performed in at least two separate experiments. depolarization. As shown in Fig. 2, hypoxia induced a robust Ser 133 phosphorylation of CREB, with an average effect greater than that produced by either forskolin-or KCl-induced depolarization, two prototypical stimuli used to activate CREB.
Multiple protein kinases are known to phosphorylate CREB at Ser 133 , including PKA (20). To determine whether the induction of CREB phosphorylation was mediated by PKA, we tested the effect of hypoxia on CREB phosphorylation in PKA-deficient PC12 cells (123.7) (28). These cells express mutant regulatory subunits of PKA which constitutively inactivate the catalytic activity of both type I and type II PKA (28). PKA enzyme assays confirmed that 123.7 cells had negligible 8-bromo-cAMP-stimulated PKA activity compared with parental PC12 cells (Table I). PKA-deficient PC12 cells were exposed to hypoxia (5% O 2 ) for 0, 1, 6, or 24 h. We found that the effect of hypoxia on CREB phosphorylation persisted in the absence of PKA (Fig. 3); that is, hypoxia induced an increase in phospho-CREB immunoreactivity (Fig. 3A) but did not alter total CREB immunoreactivity (Fig. 3B). It can also be seen in Fig. 3A that two major CREB-immunoreactive proteins were detected in 123.7 cells. The lower band corresponds to ATF-1, a structurally related CRE-binding transcription factor that cross-reacts with this antibody (18). Interestingly, phosphorylation of ATF-1 was also induced by hypoxia.
Ca 2ϩ -dependent protein kinases, including calcium/calmodulin-dependent protein kinase-I, -II, and -IV and PKC, are also known to stimulate phosphorylation of CREB (21,22,29,30). To test whether a Ca 2ϩ -dependent protein kinase mediates phosphorylation of CREB by hypoxia, PC12 cells were exposed for 6 h to either normoxia (21% O 2 ) or hypoxia (5% O 2 ) in the presence or absence of Ca 2ϩ . In Fig. 4A, PC12 cells were preloaded with 100 M BAPTA-AM, a membrane-permeable Ca 2ϩ chelator, in Ca 2ϩ -free medium supplemented with 1 mM EGTA and were then exposed to hypoxia. The hypoxia-induced phosphorylation of CREB persists in the absence of both extracellular and intracellular Ca 2ϩ (Fig. 4A). Identical results were obtained when cells were exposed to hypoxia in the absence of only extracellular Ca 2ϩ (i.e. Ca 2ϩ -free medium supplemented with 1 mM EGTA but without preloading cells with BAPTA-AM; data not shown). Pretreatment of cells with either chelerythrine chloride or Ro 31-8220, two inhibitors of PKC, also failed to inhibit hypoxia-induced phosphorylation of CREB (Fig. 4, B and C). Both of these drugs were found to inhibit PKC enzyme activity effectively in PC12 cell extracts (data not shown). Nerve growth factor-induced phosphorylation of CREB has been shown to be mediated by RSK-2 (14,24). It has been shown recently that in addition to its inhibitory effects on PKC, Ro 31-8220 is also a potent inhibitor of RSK-2 and p70 S6K (31). Neither Ro 31-8220 nor PD-098059, a selective inhibitor of activation of MEK, which is an upstream activator of RSK-2 (32), affected the ability of hypoxia to induce phosphorylation of CREB (Fig. 4, C and D), although these drugs did block nerve growth factor-stimulated CREB phosphorylation (data not shown). Similarly, hypoxia-induced phosphorylation persisted in the presence of rapamycin (Fig. 4E), an inhibitor of p70 S6K activation (33). It has also been shown that MAPKAP kinase-2, an enzyme that lies downstream of the stress-activated protein kinase p38/RK (34 -36), can phosphorylate CREB in vitro and that fibroblast growth factor or arsenite-induced CREB phosphorylation in SK-N-MC cells can be blocked by the selective p38 inhibitor SB203580 (18). However, SB203580 also did not   3. CREB phosphorylation by hypoxia persists in PKAdeficient (123.7) PC12 cells. 123.7 cells were exposed either to hypoxia (5% oxygen) or normoxia (controls, 21% oxygen) for 0, 1, 6, or 24 h, as indicated. Whole cell lysates were immunolabeled either with an antibody specific for Ser 133 phospho-CREB (panel A) or with an antibody that equally recognizes phospho-and dephospho-CREB (panel B), as described under "Experimental Procedures." Representative immunoblots are shown, with n ϭ 4 dishes of cells at each time point analyzed in each of three separate experiments. inhibit phosphorylation of CREB by hypoxia (Fig. 4F), although SB203580 did inhibit completely sorbitol-induced stimulation of MAPKAP kinase-2 activity in PC12 cells (see below). Hypoxia-induced phosphorylation of CREB was also not attenuated by either wortmannin or LY294002, two inhibitors of phosphatidylinositol 3-kinase (data not shown). Finally, there is some evidence that cGMP-dependent protein kinase (protein kinase G) can, under certain conditions, mediate CREB phosphorylation in vitro (37,38). Hypoxia-induced phosphorylation of CREB persisted strongly in the presence of each of the three inhibitors of protein kinase G, including KT-5823, methylene blue, and 8-bromo-cGMP (data not shown).
Effect of Hypoxia and Osmotic Stress on MAPKAP Kinase-2-Various stressors, including sodium arsenite, osmotic stress, UV light, and inflammatory cytokines, have been shown to activate MAPKAP kinase-2 via a p38-dependent pathway (18,19,35,36). Furthermore, at least two isoforms of p38, stress-activated protein kinase-3 (SAPK-3) and SAPK-4, have been identified which are not sensitive to inhibition by SB203580 (39, 40). Thus, we considered the possibility that MAPKAP kinase-2 could mediate hypoxia-induced phosphoryl-ation of CREB via a member of the p38 family which was not sensitive to SB203580. PC12 cells were exposed for various times between 0 and 6 h to 5% O 2 or for 20 min to 300 mM sorbitol. MAPKAP kinase-2 was immunoprecipitated, and enzyme activity was assayed in an immune complex kinase assay. As shown in Fig. 5, hypoxia had no effect on MAPKAP kinase-2 enzyme activity, whereas MAPKAP kinase-2 was strongly activated by sorbitol. Activation of MAPKAP kinase-2 by sorbitol was blocked completely by pretreatment of cells with SB203580. Taken together with the fact that hypoxia-induced phosphorylation of CREB was not inhibited by SB203580 (Fig.  4F), these results show that the effects of hypoxia on CREB are clearly not mediated by either MAPKAP kinase-2 or another p38-dependent protein kinase. Thus, it appears that a novel stress-activated signaling pathway (independent of MAPKAP kinase-2) exists and that this pathway mediates robust phosphorylation of CREB in response to hypoxia in PC12 cells.
Role of the CRE in Hypoxia-stimulated Gene Expression-Phosphorylation of CREB on Ser 133 is necessary, but not sufficient, to activate CREB-mediated transcription (12,20,30,41). Therefore, we next asked whether the Ser 133 phosphorylation of CREB resulted in functional activation of a hypoxiaregulated gene. Expression of the TH gene has been shown to be activated in response to physiological levels of hypoxia both in vivo (42) and in vitro (10). The rat TH gene contains a consensus CRE localized at Ϫ45 to Ϫ38 nucleotides relative to the transcriptional start site (43). To test for a possible role for CREB in activation of the TH gene by hypoxia, we compared the effect of hypoxia on a TH reporter plasmid (Ϫ272TH)CAT and a similar construct in which the CRE had been mutated, (Ϫ272CRE Ϫ )CAT (26,43). Hypoxia activated wild-type TH-CAT activity by approximately 3-fold, an effect similar to that of forskolin or KCl on this construct (25,43,44). As others have reported previously (25,44), we found that mutation of the CRE resulted in a significant inhibition (ϳ3-fold) of basal levels of TH-CAT reporter activity (Fig. 6). We also found that the stimulation of TH-CAT reporter activity by hypoxia was attenuated, although not ablated, in (Ϫ272CRE Ϫ )CAT compared with wild-type (Ϫ272TH)CAT, as shown in Fig. 6. Thus, a CREB-CRE interaction is involved in stimulation of TH gene expression by hypoxia. The fact that some transcriptional activation persisted in (Ϫ272CRE Ϫ )CAT is consistent with pre- After 40 min, the external Ca 2ϩ -containing (Ϫ) or Ca 2ϩ -free (ϩ) medium was replaced (not including vehicle or BAPTA-AM), and cells were exposed to either normoxia or hypoxia as described. Panel B, cells were pretreated for 40 min in serum-free medium with either (Ϫ) vehicle or (ϩ) 20 M chelerythrine chloride (CHL) and exposed to either normoxia or hypoxia as described. Panel C, cells were pretreated for 40 min in serum-free medium with either (Ϫ) vehicle or (ϩ) 0.3 M Ro 31-8220 and exposed to either normoxia or hypoxia as described. Panel D, cells were pretreated for 40 min in serum-free medium with either (Ϫ) vehicle or (ϩ) 50 M PD-098059 and exposed to either normoxia or hypoxia as described. Panel E, cells were pretreated for 40 min in serum-free medium with either (Ϫ) vehicle or (ϩ) 10 nM rapamycin and exposed to either normoxia or hypoxia as described. Panel F, cells were pretreated for 1 h in serum-free medium with either (Ϫ) vehicle or (ϩ) 20 M SB 203580 and exposed to either normoxia or hypoxia as described. In all of the experiments shown above, hypoxia did not alter total levels of CREB, as determined by immunolabeling samples with an antibody that equally recognizes phospho-and dephospho-CREB (data not shown). Each experiment was performed at least two times, and the results shown are representative of those obtained in n ϭ 6 -12 dishes in each group. Note that the apparent differences in basal levels of phospho-CREB in the various panels result from differences in exposure time of the blots to film.

FIG. 5. Lack of effect of hypoxia on MAPKAP kinase-2 activity.
PC12 cells were pretreated for 1 h with either 20 M SB203580 or vehicle (0.1% dimethyl sulfoxide) in serum-free medium. Cells were then exposed to either hypoxia (5% O 2 ) for various times, as indicated, or 300 mM sorbitol for 20 min, in the presence or absence of 20 M SB203580. MAPKAP kinase-2 was then immunoprecipitated from whole cell lysates, and MAPKAP kinase-2 enzyme activity was measured in an immunocomplex kinase assay as described under "Experimental Procedures." The data shown were obtained in one experiment, with similar results obtained in a separate experiment.
vious findings, which showed that an AP1 element located at Ϫ199/205 also plays a critical role in mediating hypoxia-and depolarization-induced activation of TH gene expression (6,25), as discussed further below. DISCUSSION The intracellular pathways involved in cellular responses and adaptation to hypoxia are only poorly understood. PC12 cells are an oxygen-sensitive cell line that has been shown to respond to hypoxia with depolarization, Ca 2ϩ influx, and dopamine release (8,45). A primary mediator of depolarization and Ca 2ϩ -regulated gene expression is the transcription factor CREB (12). Interestingly, the CREB coactivator CBP/p300 has recently been shown to interact with hypoxia-inducible factor-1␣ and participate in transcriptional regulation of hypoxiaresponsive genes (46). These studies were undertaken to evaluate the role of CREB in cellular signaling mechanisms during hypoxia. The level of hypoxia used in these studies is moderate (5% O 2 , ϳ50 mm Hg) and within the range of physiological blood levels of O 2 in rats exposed to hypoxia in vivo (10,42).
Exposure of PC12 cells to hypoxia was found to induce phosphorylation of CREB at Ser 133 , an event that is required for CREB-mediated transcriptional activation (15,20). To our knowledge, this is the first evidence for regulation of CREB by hypoxia. This response appeared to be biphasic. A modest level of CREB phosphorylation was first detected after a 20-min exposure to hypoxia, and the effect peaked at 6 h, although high levels of phospho-CREB immunoreactivity persisted up to at least a 24-h exposure to hypoxia. The relatively slow onset of this effect is distinct from that produced by sorbitol (Fig. 1) and other stimuli known to induce CREB phosphorylation (e.g. forskolin, depolarization, and growth factors), which typically elicit a rapid response that then declines (15). This suggests that the phosphorylation of CREB may be one mechanism by which PC12 cells adapt to prolonged exposure to hypoxia, although hypoxia may also induce CREB-dependent immediate early effects. The biphasic nature of this response raises the issue of whether either phase of the hypoxia-induced CREB phosphorylation requires new protein synthesis. However, interpretation of this experiment is confounded because inhibition of protein synthesis itself is a stressor that induces phos-phorylation of CREB via activation of the p38 and downstream MAPKAP kinase families of enzymes (47). Thus, clarification of the mechanism by which hypoxia induces both rapid and prolonged phosphorylation of CREB will require identification of the specific signaling pathway by which this occurs.
Strikingly, the effect of hypoxia on CREB phosphorylation was equally or more robust than that produced by the prototypical stimuli used to activate CREB, including forskolin, depolarization, or sorbitol. The hypoxia-induced phosphorylation of CREB was not mediated by any of the previously known pathways that activate CREB, including PKA-and Ca 2ϩ -dependent protein kinases, because phosphorylation of CREB by hypoxia persisted completely in PKA-deficient PC12 cells and in the absence of both extracellular and intracellular Ca 2ϩ . Hypoxia-induced phosphorylation of CREB was also not attenuated by preincubation of cells with either Ro 31-8220 or chelerythrine chloride, which inhibits both Ca 2ϩ -and non-Ca 2ϩ -dependent isoforms of PKC (48). Interestingly, unlike phosphorylation of CREB, hypoxia-induced activation of TH gene expression is blocked completely by either removal of Ca 2ϩ or by preincubation with chelerythrine chloride in PC12 cells (45). Taken together, these results demonstrate that even though a reduction in O 2 levels induces depolarization in PC12 cells (8), hypoxia can regulate gene expression via both Ca 2ϩdependent and Ca 2ϩ -independent signaling pathways.
The pp90 rsk family of kinases, including RSK-1, RSK-2, and RSK-3, are growth factor-activated serine/threonine protein kinases that are phosphorylated and activated by p42/p44 mitogen-activated protein kinase (49,50). Several laboratories have reported that pp90 rsk can mediate phosphorylation of CREB (17,24,51,52). The RSK-2 isoform specifically has been shown to phosphorylate CREB in response to nerve growth factor stimulation (12,24). Several lines of evidence strongly suggest that the effect of hypoxia is not mediated by RSK-2 or other members of the pp90 rsk family of kinases. First, hypoxiainduced phosphorylation of CREB persisted in the presence of Ro 31-8220, which has recently been shown to be as equally potent an inhibitor of RSK-2 as PKC (31). Second, in contrast to the study of Pende et al. (17), who found that growth factor-and Ca 2ϩ -induced phosphorylation of CREB was blocked by PD-098059, the effect of hypoxia on CREB persisted in the presence of PD-098059. PD-098059 is a selective inhibitor of MEK activation, which thereby also inhibits the activation of downstream effectors of MEK, including MAPK and the pp90 rsk enzymes (32,50). Finally, exposure to the same level of hypoxia used in this study strongly inhibits p42/p44 MAPK activity in PC12 cells. 2 Thus, this effect does not appear to be mediated by RSK-2 or any other MAPK-dependent pp90 rsk isoforms.
Two recent studies have shown that activation of the stressactivated protein kinase p38 by cellular stressors (18,19) and fibroblast growth factor (18) can mediate Ser 133 phosphorylation of CREB and that this can be blocked by the highly specific p38 inhibitor SB203580 (36). CREB is not phosphorylated by p38 itself, but it can be phosphorylated in vitro by the downstream substrate of p38, MAPKAP kinase-2 (18). In contrast, in our studies, hypoxia-induced phosphorylation of CREB was not diminished by SB203580, although the drug effectively blocked sorbitol-induced activation of MAPKAP kinase-2. Furthermore, hypoxia did not stimulate MAPKAP kinase-2 enzyme activity. We propose that a novel hypoxia-activated protein kinase mediates phosphorylation of CREB in response to a reduction in O 2 levels. Recent studies have identified at least two isoforms of p38 which are insensitive to SB203580: SAPK-3 FIG. 6. CRE is involved in activation of TH gene expression by hypoxia. PC12 cells were transfected with 2 g of either (Ϫ272TH) CAT or (Ϫ272CRE Ϫ )CAT, as described under "Experimental Procedures." Beginning 24 h after transfection, cells were exposed for 24 h to either hypoxia (5% O 2 , shaded bars) or normoxia (21% O 2 , black bars). CAT levels were analyzed by enzyme-linked immunosorbent assay in duplicate samples each containing 120 g of total protein, as described. The data shown are from a single experiment and are expressed as the average Ϯ S.E. and represent n ϭ 4 dishes in each group. Similar results were obtained in three separate experiments. and SAPK-4 (39,40). Like p38, the major known downstream effector of these kinases is MAPKAP kinase-2, suggesting that SAPK-3 and SAPK-4 also do not mediate the effects of hypoxia on CREB. However, we cannot exclude the possibility that other downstream targets (e.g. MAPKAP kinase-3) could mediate hypoxia-induced phosphorylation of CREB.
It has been established previously that TH gene expression is induced by hypoxia both in vivo (42) and in PC12 cells (10). The TH gene contains a CRE that has been shown to be critical for cAMP-and Ca 2ϩ -induced activation of gene expression (25,44,53,54). We found that the hypoxia-induced activation of the TH gene was attenuated significantly in a TH-CAT reporter plasmid in which the CRE had been mutated. The TH gene contains several elements that participate in hypoxia-induced activation of gene expression, including AP-1 and hypoxiainducible factor-1 elements located in the region between Ϫ284 to Ϫ190 nucleotides relative to the transcription initiation site (6). Thus, one or more of these upstream regulatory elements presumably mediates the residual activation of the TH gene in the absence of the CRE, and CREB per se is likely to be insufficient to mediate this entire effect. Our results suggest that hypoxia may induce activation of other CRE-containing genes, including c-fos, which is known to be stimulated by hypoxia (6,55).
In summary, hypoxia induces a robust and persistent phosphorylation of CREB in PC12 cells. Hypoxia-induced phosphorylation of CREB is involved in functional activation of TH, a gene that is known to be induced in response to hypoxia (10,42). Unlike other stimuli, which induce peak CREB phosphorylation relatively rapidly and then decline (15), the effects of hypoxia on CREB phosphorylation are maximal with a 3-6-h exposure to hypoxia and persist for at least 24 h. This suggests that phosphorylation of CREB and activation of a subset of hypoxia-responsive genes may be an adaptive cellular response to a reduction in O 2 levels. Interestingly, it has been reported that phospho-CREB immunoreactivity is increased in rat dentate granule cells and neocortex 1-2 days after hypoxic-ischemic induced brain damage (56). In this study, phospho-CREB immunoreactivity was induced selectively in ischemia-resistant cells, but not CA1 pyramidal cells, which undergo neuronal death after hypoxia-ischemia in vivo, suggesting that phospho-CREB may be involved in the process of neuroprotection. Current studies are under way to identify the signaling pathway by which CREB phosphorylation is induced during hypoxia. Such studies will further our understanding of the neuronal response to oxygen deprivation and, in turn, shed light on molecular mechanisms underlying the pathology of hypoxic and/or ischemic trauma.