Cyclooxygenase-2 suppresses hypoxia-induced apoptosis via a combination of direct and indirect inhibition of p53 activity in a human prostate cancer cell line.

Although p53-inactivating mutations have been described in the majority of human cancers, their role in prostate cancer is controversial as mutations are uncommon, particularly in early lesions. p53 is activated by hypoxia and other stressors and is primarily regulated by the Mdm2 protein. Cyclooxygenase (COX)-2, an inducible enzyme that catalyzes the conversion of arachidonic acid to prostaglandins and other eicosanoids, is also induced by hypoxia. COX-2 and resultant prostaglandins increase tumor cell proliferation, resistance to apoptosis, and angiogenesis. Previous reports indicate a complex, reciprocal relationship between p53 and COX-2. To elucidate the effects of COX-2 on p53 in response to hypoxia, we transfected the COX-2 gene into the p53-positive, COX-2-negative MDA-PCa-2b human prostate cancer cell line. The expression of functional p53 and Mdm2 was compared in COX-2+ versus COX-2- cells under normoxic and hypoxic conditions. Our results demonstrated that hypoxia increases both COX-2 protein levels and p53 transcriptional activity in these cells. Forced expression of COX-2 increased tumor cell viability and decreased apoptosis in response to hypoxia. COX-2+ cells had increased Mdm2 phosphorylation in either normoxic or hypoxic conditions. Overexpression of COX-2 abrogated hypoxia-induced p53 phosphorylation and promoted the binding of p53 to Mdm2 protein in hypoxic cells. In addition, COX-2-expressing cells exhibited decreased hypoxia-induced nuclear accumulation of p53 protein. Finally, forced expression of COX-2 suppressed both basal and hypoxia-induced p53 transcriptional activity, and this effect was mimicked by the addition of PGE2 to wild-type cells. These results demonstrated a role for COX-2 in the suppression of hypoxia-induced p53 activity via both direct effects and indirect modulation of Mdm2 activity. These data imply that COX-2-positive prostate cancer cells can have impaired p53 function even in the presence of wild-type p53 and that p53 activity can be restored in these cells via inhibition of COX-2 activity.

The p53 tumor suppressor protein plays a critical role in the regulation of cell growth, the protection of normal cells from malignant transformation, and the response of tumor cells to radiation and chemotherapy (1,2). Although alterations in the p53 gene are found in more than half of all human cancers, the role of p53 mutations in human prostate cancer is controversial, as such mutations are uncommon, particularly in early lesions (3,4). After exposure of cells to a variety of stressors, including genotoxic DNA damage, nucleotide depletion and hypoxia, wild-type p53 protein is activated (1). Recent reports demonstrate alterations in p53 activation and stabilization in human premalignant and malignant lesions that retain the expression of functional wild-type p53 (5)(6)(7). This wild-type p53 inactivation may result from either abnormal sequestration in the cytoplasm where it is functionally muted (6,7) or increased binding of the protein to its critical negative regulator, Mdm2, which prevents its activation and promotes its degradation via ubiquitination (8).
Cyclooxygenase-2 (COX-2), 1 an inducible enzyme that catalyzes the conversion of arachidonic acid to prostaglandins (PGs) and other eicosanoids, is also induced in response to hypoxia and other stressors (9 -12). COX-2 and resultant PGs increase tumor cell proliferation, resistance to apoptosis, and angiogenesis in colon, prostate, and other cancers (13)(14)(15)(16)(17). Our group demonstrated that COX-2 expression and PGE 2 production are increased in prostatic inflammation, prostatic intraepithelial neoplasia, and some primary prostate cancer cells (18). We also reported that COX-2 and PGE 2 are key mediators of cellular responses observed after hypoxia resulting from induction of vascular endothelial cell factor (12,19) and the master oxygen sensor, hypoxia-inducible factor (HIF)-1␣ (20).
Previous reports indicate complex, functional, and reciprocal interactions between the COX-2 and p53 systems. An association between COX-2 overexpression and p53 mutations or low p53 protein levels has been reported in human gastric and breast cancers (21,22). Moreover, electrophilic PGs produced by COX-2 were shown to inhibit wild-type p53 activity by covalently binding p53, preventing its nuclear accumulation (23). It has been reported that inhibition of COX-2 in a colon cancer cell line by celecoxib, a selective COX-2 inhibitor, increases the nuclear localization of active p53 (24). Subbaramaiah et al. (25) demonstrated that p53 inhibits COX-2 expression. Recent evidence also suggests that COX-2 and p53 have reciprocal effects on cellular responses to hypoxia via modulation of HIF-1␣ activity (20,26).
We hypothesized that COX-2 expression in prostate cancer cells decreases wild-type p53 stability, nuclear accumulation, and activity after hypoxia, via both direct effects on p53 activation and indirectly by activation of its primary negative regulator, Mdm2. In the present study, we utilized a prostate cancer cell line, MDA-PCa-2b, as an in vitro model. This cell line expresses the wild-type p53 gene and does not express COX-2 (27). We tested the effects of forced expression of COX-2 on cellular proliferation, apoptosis, and the activity of p53 and Mdm2 proteins, under both normoxic and hypoxic conditions.

EXPERIMENTAL PROCEDURES
Cell Line, Cell Culture, and Reagents-The MDA-PCa-2b human prostate cancer cell line was purchased from ATCC (Gaithersburg, MD). Cells were cultured in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum under normoxic conditions (20% O 2 , 5% CO 2 , 75% N 2 ) in a humidified Napco incubator at 37°C. Hypoxic stimulation was produced with an ambient oxygen concentration of 0.1% using a controlled incubator with CO 2 /O 2 monitoring and CO 2 /N 2 gas sources. Oxygen levels were determined by an indicator. To reach a low oxygen level (0.1%), the chamber was sealed and subjected to five rounds of evacuation followed by flushing with 95% N 2 -5% CO 2 as described by Graeber et al. (28). To prevent reoxygenation of hypoxic cells, the medium or lysis buffer was pre-equilibrated to the experimental oxygen conditions overnight and added to cells on ice. Prostaglandins, i.e. PGE 2 , PGI 2 , PGF 2 ␣, and PGD 2, were purchased from Cayman Co., Ann Arbor, MI.
Transfection-The expression vector pcDNA-hCOX-2, containing the full-length cDNA encoding the human COX-2 gene, was a generous gift from Dr. T. Hla (Department of Physiology, University of Connecticut). Transient transfection was performed using Lipofectamine TM 2000 reagent according to the manufacturer's instructions (Invitrogen). Cells used as a mock control were transfected with empty pcDNA1 vector (Invitrogen). Typically, 3 g of plasmid DNA were used per 60-mm dish. All assays were performed after 40 h of transfections. Cells were treated with (c, d) or without (a, b) hypoxia for 8 h. Apoptotic cells were detected by TUNEL assays and examined using a fluorescent microscope (magnification ϫ200). The nonapoptotic cells stained red; apoptotic cells stained yellow or green. Data shown are representative of three independent assays. transfected) were cultured in 12-well cluster plates. Incubations were continued under either normoxic or hypoxic conditions for 8 h. After washing with ice-cold phosphate-buffered saline, apoptotic cells were assayed using the TUNEL method with the ApopTag in situ apoptotic detection kit (Oncor, Gaithersburg, MD) according to the manufacturer's instructions. The labeled cells were examined using a fluorescent microscope.
Preparation of Cell Lysates and Cytosolic and Nuclear Proteins-MDA-Pca-2b cells cultured under the desired conditions were lysed as described previously (17). Briefly, cells were rinsed twice with ice-cold phosphate-buffered saline and scraped with 1.5 ml of phosphate-buffered saline containing 4 mM iodoacetate. After centrifugation, the pellets were resuspended in CHAPS extraction solution (10 mM CHAPS, 2 mM EDTA, pH 8.0, and 4 mM iodoacetate in phosphate-buffered saline) with protease inhibitors. The samples were incubated for 30 min on ice and centrifuged at 15,000 ϫ g for 10 min. The supernatants were collected and stored at Ϫ70°C. Proteins from the cytosolic and nuclear fractions were isolated using a commercial kit purchased from Pierce, according to the manufacturer's instructions. Protein content was assayed using a kit from Bio-Rad.
Immunoprecipitation and Immunoblotting-For immunoprecipitation, the lysates were precleaned with protein G-Sepharose beads for 2 h at 4°C. The lysates were then incubated with 1 g of monoclonal anti-Mdm2 antibody (clone IF2) from Calbiochem, followed by the addition of protein G-Sepharose beads. The immunoprecipitated products were then subjected to Western blot analysis. For immunoblotting, the cell lysates were electrophoresed on SDS-polyacrylamide gel, electrophoretically transferred to a polyvinylidene difluoride membrane (PerkinElmer Life Sciences), and incubated with targeting antibodies overnight at 4°C. Secondary horseradish peroxidase-linked donkey anti-mouse IgG (Amersham Biosciences) was used. Filters were developed by the enhanced chemiluminescence system (Amersham Biosciences). Antibodies against p53, phospho-p53, and phospho-Mdm2 (Ser-166) were products from Cell Signaling Technology (Beverly, MA). Endogenous Mdm2 antibodies were purchased from Calbiochem, and anti-COX-2 antibodies were obtained from Transduction Labs, Lexington, KY. ␤-Actin was used as the internal control in all Western blot analyses.
Luciferase Assay-The reporter plasmid pGL3-Luc-E1bTATA, a gift from Dr. J. Manfredi (Derald H. Ruttenberg Cancer Center, Mount Sinai School of Medicine), was used to assay p53 transcriptional activity. The plasmid contains the p21 promoter sequences and the coding region for firefly luciferase upstream of the minimal adenovirus E1b promoter. MDA-PCa-2b cells were cultured in 12-well cluster plates and co-transfected with 1 g of the reporter plasmid or empty pGL3-Luc vector and pcDNA-hCOX-2 or empty pcDNA plasmid. Internal normalization was performed by co-transfection of a ␤-galactosidase expression vector driven by the CMV promoter (Clontech). Total DNA transfected into each plate was 3 g in each 60-mm dish. After 40 h, the transfected cells were lysed by scraping into reporter buffer (Clontech), total protein concentration was determined, and luciferase and ␤-galactosidase activities were assayed and quantitated using a TD-20e luminometer. The resulting activities were normalized to protein concentrations and ␤-galactosidase activity.

Hypoxia Induces COX-2 Expression in COX-2-transfected MDA-PCa-2b
Cells-It has been reported that COX-2 expression is increased by hypoxia and cobalt chloride (a hypoxia surrogate) in human vascular endothelial cells and a metastatic prostate cancer cell line, respectively (11,12). In our initial studies, we examined the basal and hypoxic increase in COX-2 protein expression in mock-transfected (control) and COX-2-transfected MDA-PCa-2b cells. As shown in Fig. 1, mock-transfected MDA-PCa-2b cells do not express COX-2 protein under either normoxic or hypoxic conditions (Fig. 1A). When these cells were transfected with the COX-2 gene, we demonstrated that COX-2 protein expression is increased under hypoxic conditions in a time-dependent fashion (Fig. 1B). As the COX-2 construct utilized does not contain the promoter region of the gene, we have assumed that regulation of protein expression by hypoxia is occurring at the post-transcriptional level, a phenomenon that has previously been reported (29,30).
COX-2 Expression Prevents Apoptosis in Hypoxic Prostate Cancer Cells-Hypoxia induces cell cycle arrest at the G 0 /G 1 phases and cellular apoptosis via both p53-independent (28) and p53-dependent (31-33) pathways. We examined the effect of COX-2 on cell viability and apoptosis under hypoxic conditions. Hypoxic treatment dramatically decreased cell viability and increased apoptosis in the COX-2 Ϫ , mock-transfected versus COX-2 ϩ MDA-PCa-2b cells ( Fig. 2A). TUNEL assay revealed that the decrease in viable cell numbers resulted, at least in part, from induction of apoptosis by hypoxia in the COX-2 Ϫ cells. COX-2 ϩ cells were resistant to the apoptotic effects of hypoxia (Fig. 2B).

COX-2 Modulates Mdm2 Phosphorylation under Normoxic and Hypoxic
Conditions-Mdm2 is a critical negative regulator of p53 protein expression, nuclear localization, and transcriptional activity (8). Both Mdm2 and p53 proteins are activated by phosphorylation. Hypoxic induction of p53 accumulation is partially modulated by decreased phosphorylation leading to decreased activation of Mdm2 (34,35). Fig. 3 demonstrates the effects of forced COX-2 expression on Mdm2 levels and phosphorylation under both normoxic and hypoxic conditions. Consistent with previous reports (34, 35), parental, mock-trans- In contrast, forced expression of COX-2 led to an increase in Mdm2 phosphorylation under normoxic conditions. The enhanced expression of phospho-Mdm2 protein was sustained in the COX-2 ϩ cells, even under hypoxic conditions (Fig. 3A). These data demonstrated that COX-2 not only promotes Mdm2 phosphorylation under normoxic conditions but, furthermore, prevents hypoxic inhibition of phosphorylated Mdm2 protein. In contrast, hypoxia induced a modest increase in endogenous Mdm2 levels in both control and COX-2 ϩ cells (Fig. 3B).  depicts the quantitative analysis of relative expression levels of Mdm2 and phospho-Mdm2 (inactive and active Mdm2, respectively) and demonstrates that forced expression of COX-2 results in increased active, phosphorylated Mdm2 protein expression.

COX-2 Abrogates Hypoxic Induction of Ser-20 Phosphorylation of p53 and Promotes the Binding of p53 to Mdm2 in
Response to Hypoxia-Phosphorylation is a critical modulator of p53 activity (36). It has also been proposed that phosphorylation of p53 could impede binding between p53 and Mdm2 (36,37) leading to stabilization and activation of p53 protein (8,38). We examined the levels of p53 phosphoprotein under normoxic and hypoxic conditions in the mock-transfected versus COX-2transfected cells. Among several potential phosphorylation sites of the p53 protein, Ser-15, Ser-20, and Ser-392 are reported to be involved in the regulation of p53 protein activation in response to a variety of stress signals (36,39,40). Under normoxic conditions, there was no demonstrable p53 phosphoprotein in either cell population. In response to hypoxia, p53 protein was subjected to phosphorylation at Ser-20 only in the COX-2-negative, mock-transfected MDA-PCa-2b cells (Fig. 4A). Of note, there were no detectable levels of p53 phosphoprotein at the other sites (data not shown). In contrast, when these cells were transfected with the COX-2 gene, the induction of p53 phosphorylation at Ser-20 was abolished. Hypoxia also induced endogenous p53 protein levels in mock-transfected prostate cancer cells but had no effect on the cells that were engineered to overexpress COX-2 (Fig. 4A). To determine whether hypoxic induction of p53 phosphorylation influences the binding of the protein to Mdm2 (a critical step of p53 protein degradation), we performed immunoprecipitation followed by immunoblot analysis. As demonstrated in Fig. 4B, there was no difference in the binding of p53 to Mdm2 under aerobic conditions between the mock-and COX-2-transfected cells. In response to hypoxia, Mdm2-p53 binding was decreased in the control cells. In contrast, COX-2 gene transfection led to an increase in Mdm2-p53 binding. These results suggested a negative correlation between the amount of phospho-p53 protein (Ser-20) and the ability of p53 to bind to Mdm2.

COX-2 Suppresses Basal and Hypoxic Induction of p53
Nuclear Localization-p53 protein translocates from the cytoplasm to the nucleus upon stress stimulation, and this nuclear targeting is essential for its activity as a transcriptional regulator (5). A recent study reported that celecoxib, a selective COX-2 inhibitor, inhibits active p53 nuclear localization in a human colon cancer cell line (24). We next examined the effect of COX-2 expression on the subcellular distribution of p53 protein under both normoxic and hypoxic conditions. As shown in Fig. 5, Western blotting using samples derived from the cytosolic and nuclear fractions revealed that forced expression of COX-2 resulted in cytosolic retention and decreased nuclear accumulation of p53 protein under normoxic conditions. In response to hypoxia, p53 protein translocated to the nucleus in the COX-2 Ϫ , mock-transfected cells. Forced expression of COX-2 prevented the nuclear accumulation of p53 in response to hypoxia (Fig. 5). These results revealed a novel role for COX-2 in the suppression of hypoxia-induced p53 protein nuclear localization, thereby inhibiting p53 activation in response to hypoxic signals.

COX-2 Inhibits Basal and Hypoxia-induced p53
Transcriptional Activities-We next investigated the effect of COX-2 expression on p53 transcriptional activity using a luciferase assay. COX-2 Ϫ and COX-2 ϩ MDA-Ca-2b cells were co-transfected with a plasmid containing the p21 promoter sequence and the coding region for firefly luciferase upstream of the minimal adenovirus E1b promoter. Fig. 6 demonstrates that COX-2-expressing cells had decreased p53 transcriptional activity as compared with COX-2 Ϫ cells in normoxic conditions. COX-2 expression in these cells also prevented the hypoxic increase in p53 transcriptional activity seen in the parental, mock-transfected cells (Fig. 6). To determine whether the effects of COX-2 are the direct result of its enzymatic activity, four major prostaglandin products of the COX-catalyzed pathway, i.e. PGE 2 , PGI 2 , PGF 2 ␣, and PGD 2 , were tested for their possible effects on p53 activity under normoxic and hypoxic conditions. As shown in Fig. 6, although only a modest effect was seen under normoxic conditions, PGE 2 exhibited a significant inhibitory effect on p53 transcriptional activity under hypoxic conditions, accounting for 90% of the effect induced by COX-2 overexpression in the same cell line. We also compared PGE 2 production in COX-2-transfected and mock-transfected cells under normoxic and hypoxic conditions. Parental cells that were mock-transfected had undetectable levels of PGE 2 . Forced expression of COX-2 in these cells increased PGE 2 production, and a time-dependent further increase was observed when the cells were cultured under hypoxic conditions (data not shown). In comparison, PGF 2 ␣ modestly inhibited p53 transcriptional activity in both normoxic and hypoxic conditions. Neither PGI 2 nor PGD 2 had any effect on these parameters. DISCUSSION It is well established that the p53 tumor suppressor protein plays a critical role in cancer development and progression. There are also accumulating data implicating COX-2 overexpression and increased secretion of prostaglandins (notably PGE 2 ) in a variety of neoplasms. Several reports confirm a reciprocal relationship between the p53 and COX-2 systems. Wild-type p53 has been shown to decrease COX-2 expression and activity (22,25). Prostaglandins derived from the COX-2catalyzed pathway inactivate wild-type p53 in colon cancer cell lines (23). Furthermore, inhibitors of COX-2 increase the activity of wild-type p53 by promoting its nuclear localization in colon cancer cells (24).
Although p53 is commonly mutated in solid tumors, the rate of such mutations in prostate cancer, particularly in early lesions, is not high (3,4). Our group and others, however, have demonstrated increased COX-2 expression in preneoplastic and cancerous prostate lesions (18). We have further shown that COX-2 inhibitors induce prostate cancer cell apoptosis in vitro and in vivo, decrease tumor angiogenesis, and decrease the levels and activity of the master oxygen sensor, HIF-1␣ (12,17,19,20). Loss of p53 activity in tumor cells enhances the levels of HIF-1␣ and augments HIF-1-dependent activation of gene transcription (i.e. vascular endothelial cell factor) in response to hypoxia (26). In addition, loss of p53 function is associated with reduced cellular apoptotic potential in response to hypoxia (31). Therefore, we hypothesized that COX-2 may inhibit wild-type p53 activity, particularly under hypoxic conditions.
In the present study, we demonstrated that hypoxia increases both COX-2 protein levels and p53 transcriptional activity in a human prostate cancer cell line. Forced expression of COX-2 conferred resistance to hypoxia-induced apoptosis in these prostate cancer cells. Forced expression of COX-2 also promoted Mdm2 phosphorylation under normoxic conditions and prevented hypoxia-induced decreased phosphorylation of Mdm2.
The Mdm2 protein is an important inhibitor of p53 function (8). Mdm2 both represses p53 transcriptional activity and promotes the degradation of p53 by the ubiquitin-proteasome system. It has previously been demonstrated that phosphorylation of Mdm2 is essential for its translocation from the cytoplasm to the nucleus, where it binds and inactivates p53 (41)(42)(43). It has also been reported that hypoxia-induced p53 accumulation and stabilization is through hypophosphorylation of Mdm2 and down-regulation of Mdm2 protein expression (34,35). In addition, Mdm2 has been linked to the regulation of p53 nucleocytoplasmic trafficking (5,42,43). Our findings demonstrated that at least one mechanism whereby COX-2 inhibits hypoxic activation of p53 is via phosphorylation and activation of Mdm2.
One of the most obvious ways to protect p53 from Mdm2mediated degradation is to prevent physical interactions between the two proteins. The structural requirements for p53/ Mdm2 interactions indicate that phosphorylation at sites within the N terminus of p53 (including serine 15, 20, and 33) may impede binding of the two proteins (36,37). We have reported that forced expression of COX-2 directly modulated p53 activity via effects on p53 protein phosphorylation, nuclear localization, and transcriptional activity. In COX-2-null MDA-PCa-2B cells, hypoxia increased p53 phosphorylation at Ser-20, a critical phosphorylation site involved in its nuclear translocation, activation, and protection from degradation (36,44). Overexpression of COX-2 inhibited hypoxia-induced phosphorylation at Ser-20, thereby enhancing the binding between p53 and Mdm2, leading to p53 protein ubiquitination and proteasomal degradation. In addition, we provided evidence that overexpression of COX-2 suppressed basal and hypoxia-induced nuclear translocation of p53 protein. Finally, COX-2expressing cells had decreased p53 transcriptional activity as compared with COX-2-null cells, under both normoxic and hypoxic conditions. Exogenous addition of PGE 2 , a major product of the COX-2-catalyzed pathway, mimicked the effect of COX-2 overexpression on the repression of p53 transcriptional activity in response to hypoxia. Both PGE 2 and PGF 2 ␣ addition partially mimicked the effect of COX-2 overexpression on p53 transcriptional activity under normoxic conditions. Although these results implied that effects on p53 activity are primarily mediated by COX-2-dependent pathways, we cannot rule out some COX-2-independent effects in this model system.
Our results demonstrated that COX-2 can inhibit wild-type p53 activation, particularly in response to hypoxic signals, via direct modulation of p53 protein as well as through the activation of the p53 inhibitor Mdm2. Somatic mutation of p53 is probably a late event in prostate oncogenesis. However, epigenetic modifications of p53 with resultant inactivation of its function have been reported in other tumors (23,24). Our results provided a possible link between overexpression of COX-2 and inactivation of wild-type p53 function in prostate cancer cells. These effects were most evident under hypoxic conditions and may help explain the relative chemo-and radiation-resistance of hypoxic prostate cancer cells, even those expressing wild-type p53. These data implied that inhibition of COX-2 can restore normal p53 function in response to intratumoral hypoxia and thereby render the cells more sensitive to therapeutic regimens.