Oxidative damage of cardiomyocytes is limited by extracellular regulated kinases 1/2-mediated induction of cyclooxygenase-2.

Oxidative stress causes cardiac damage following ischemia/reperfusion and in response to anthracyclines. Extracellular signal-regulated kinases (ERK) 1/2 are activated by oxidative stress in cardiac myocytes and protect cardiac myocytes from apoptosis. Prostaglandins (PG) also protect cells from injury in a number of tissues, including the cardiomyocyte. Cyclooxygenase (COX) the rate-limiting enzyme in PG biosynthesis has two isoforms, the constitutive COX-1 and an inducible COX-2. Here, we examined the effects of two oxidative stresses, hydrogen peroxide (H2O2) and the anthracycline doxorubicin on the activity of ERK1/2 and the expression of COX isoforms and PG formation in neonatal rat primary cardiomyocytes. These cells expressed COX-1 at rest and both COX isoforms on treatment with phorbol 12-myristate 13-acetate. Exposure to 50 microM H2O2 for 10 min or doxorubicin at 10 and 100 micrograms/ml caused expression of COX-2 that was prevented by free radical scavengers. COX-2 induction was associated with activation of ERK1/2 and the specific ERK-inhibitor PD098059 abolished COX-2 expression. Treatment of cells with decoy oligonucleotides corresponding to COX-2 promoter elements implicated the AP-1 and NF-kappaB-2 but not the NF-kappaB-1 in the transcription of COX-2. Induction of COX-2 mRNA and protein was accompanied by increased prostacyclin formation, which was abolished by the selective COX-2 inhibitor, NS-398, and PD098059. H2O2 and doxorubicin enhanced the release of lactate dehydrogenase and free radical scavengers prevented this. NS-398 enhanced the release of lactate dehydrogenase in response to H2O2 and doxorubicin, whereas the injury was prevented by iloprost, a stable prostacyclin analogue. In cardiomyocytes cell injury by H2O2 and doxorubicin is limited by an increase in prostacyclin formation that reflects induction of COX-2 mediated by ERK1/2 activation.

Reperfusion of ischemic myocardium results in an abrupt aggravation of cardiomyocyte injury, demonstrated experimentally by the re-introduction of oxygen into hypoxic myocardium (1). This injury is due in part to the generation of reactive oxygen species (2,3), and the injury is limited by antioxidants and free radical scavengers (4). Potential sources of reactive oxygen species generation during ischemia/reperfusion of vascular tissue include superoxide (O 2 . ) via neutrophil NADPH oxidase (5,6) or from leakage of electrons from the electron transport chain in the mitochondria (7). The O 2 . produced is converted to hydrogen peroxide (H 2 O 2 ) by superoxide dismutase, a weak free radical that reacts with O 2 . to generate the highly active hydroxyl radical ( ⅐ OH). Anthracyclines, such as doxorubicin induce cardiac injury in a similar manner, in that they are metabolized to the corresponding semiquinone free radical by broadly distributed enzymes, flavin reductases (7,8).
In addition to causing tissue injury, free radicals induce genes that are protective, such as hemoxygenase-1, which generates the antioxidant, biliverdin (9). Cyclooxygenase is likewise cytoprotective in several tissues (10) and plays a role in preventing apoptosis (11). Moreover, cyclooxygenase activity is known to be sensitive to free radicals (12). In this study, we explored the expression of cyclooxygenase in response to free radical-induced injury of primary cardiomyocytes using H 2 O 2 and doxorubicin, both of which induce cardiac prostaglandin (PG) 1 formation (13). Cyclooxygenase (COX), the rate-limiting enzyme in prostaglandin synthesis, exists as two isoforms (14,15). COX-1 is present in most cells and is responsible for constitutive PG formation, whereas COX-2 is largely absent but is induced by cytokines, growth factors, and hormones (16). A recent study has reported the expression of COX-2 in ischemic human myocardium and in dilated cardiomyopathy, but not in normal cardiomyocytes (17). How COX-2 is induced in the heart and its function in these conditions is unknown. Cardiac fibrosis has been reported in mice where the COX-2 isoform has been disrupted (18), suggesting that COX-2 expression may be protective.
The mitogen-activated protein kinases (MAPKs) are serine/ threonine protein kinases and 4 subfamilies have been described. One of the subfamilies, extracellular regulated kinases (ERKs)1/2, is activated by a variety of growth factors, cytokines, and phorbol esters, and regulates cellular proliferation and differentiation (19,20). In cardiac myocytes activation of ERK1/2 regulates gene expression and is implicated in the development of cellular hypertrophy (21,22). Oxygen-derived free radicals induce activation of ERK1/2 and a second MAP kinase, p38 in cardiac myocytes and other cells (23,24). H 2 O 2induced activation of ERK1/2 is mediated through the Ras/Raf-1/Mek pathway (23)(24)(25). PD098059, a selective inhibitor of ERK1/2 activation, aggravates H 2 O 2 -induced apoptosis of cardiomyocytes (23). The mechanism of the protective role played by ERK1/2 activation is unknown. We investigated a possible link between ERK1/2 activation and COX-2 expression in car-diomyocytes injured by oxygen-free radicals and used a transcription factor decoy approach to decipher the role of the promoter elements NF-B and AP-1 in COX-2 induction.
Cell Culture-Primary cultures of neonatal rat cardiomyocytes were prepared by a modification of the method originally described by Simpson et al. (26). Briefly, the hearts from 1 to 3-day-old Wistar rats were minced and dissociated with approximately 80 units/ml type II collagenase and 0.06% pancreatin. Dispersed cells were incubated in 25-cm 2 flasks for 30 min at 37°C in a CO 2 incubator. Nonattached viable cells were collected and seeded into fresh flasks at a density of 1.5 ϫ 10 5 cells/ml. To obtain a near-pure cardiomyocyte preparation, cells were incubated in Dulbecco's modified Eagle's medium with Ham's F-12 supplemented with 1.2 g/liter sodium hydrogen carbonate, 10% (v/v) NBS containing 1 mM pyruvic acid, 1 g/ml transferrin, 10 g/ml insulin, 10 ng/ml sodium selenite, 250 M vitamin C, 100 M 5-bromo-2-deoxyuridine, 100 units/ml penicillin G, and 100 g/ml streptomycin. This medium was replaced with medium containing no NBS at 48 h. The cells were incubated for 24 h before addition of the test compounds.
Treatment of Cells with PMA-On day 4 following the isolation, cardiomyocytes grown in 25-cm 2 flasks were rinsed twice with PBS. PMA was added to medium without NBS and incubated with cells for 6 h. For negative controls, cardiomyocytes were incubated with medium alone. Following treatment the medium was removed and cells were washed with PBS and medium containing COX inhibitors were then added: 200 M aspirin for 30 min or 1 M NS-398 for 45 min. All cells were then washed twice with PBS and 2 ml of HBSS was added containing 50 M arachidonic acid (peroxide free) for 10 min to measure the maximum amount of prostaglandin formed. The supernatants were stored at Ϫ70°C for PG analysis.
Treatment of Cells with Hydrogen Peroxide and Doxorubicin-On day 4 following isolation, the cardiomyocytes were rinsed twice with PBS. H 2 O 2 , 50 M and 0.5 mM, or 100 g/ml DX (172 M) was added to medium without NBS and incubated with cells for 10 or 80 min, respectively. For negative controls cardiomyocytes were incubated with medium alone for equivalent amounts of time. Oxygen-free radical scavengers PEG-CAT (200 units/ml) and PEG-superoxide dismutase (200 units/ml) were added to cells 1 h before treating with DX. The medium was then removed, the cells washed with PBS and fresh medium added to allow cells recover for 10 min, 3 h, and 10 h. At these time points medium was removed and stored at Ϫ70°C for LDH analysis. The cells were then washed with PBS and medium containing COX inhibitors were added: 200 M aspirin for 30 min or 1 M NS-398 for 45 min. All cells were then washed twice with PBS and 2 ml of HBSS was added containing 50 M arachidonic acid (peroxide-free) for 10 min to measure the maximum amount of PG formed. This HBSS was stored at Ϫ70°C for analysis. The cells were then washed twice with PBS and either lysed in protein 1% Triton X-100 lysis buffer and stored at Ϫ20°C for Western blot analysis or scraped into Tri-Reagent and stored at Ϫ70°C for RNA isolation. In further experiments cells were treated with 1 M NS-398 alone, or 10 g/ml DX (17.2 M) with or without 1 M NS-398, or with 10 g/ml DX in the presence of 200 units/ml PEG-CAT and 200 units/ml PEG-superoxide dismutase for 24 h. As above the medium was removed and stored at Ϫ70°C for LDH analysis, the cells washed with PBS and 2 ml of HBSS was added containing 50 M arachidonic acid for 10 min. The HBSS was stored at Ϫ70°C for PG analysis. The cells were washed and scraped into Tri-Reagent and stored at Ϫ70°C for mRNA analysis.
Treatment of Cells with Iloprost, or Fluprostenol, followed by 0.5 mM H 2 O 2 or 100 g/ml Doxorubicin-Cells were treated with 30 nM iloprost alone, or with 1-100 nM iloprost, or medium alone (positive control) for 24 h, before treating with 0.5 mM H 2 O 2 or 100 g/ml DX, and LDH released into medium was measured. Cells were also treated both with the PGF 2␣ analog, 30 M fluprostenol alone, or 30 M fluprostenol with 0.5 mM H 2 O 2 or 100 g/ml DX, and LDH released into medium was measured. The cells were subsequently scraped into Tri-Reagent for COX-2 mRNA analysis.
Treatment of Cells with H 2 O 2 or Doxorubicin and the ERK Inhibitor PD098059 or the p38 MAP Kinase Inhibitor SB203580 -Cells were pretreated with 50 M PD098059 or 20 M SB203580 for 1 h, before being exposed to 50 M H 2 O 2 or 100 g/ml DX with PD098059 or SB203580 for 10 or 80 min, respectively. Following exposure cells were allowed to recover in serum-free medium with 50 M PD098059 or 20 M SB203580 for 3 h. As a positive control cardiomyocytes were incubated with 50 M H 2 O 2 or 100 g/ml DX alone, for 10 or 80 min, respectively, followed by a 3-h incubation of cells in serum-free medium. As discussed above, all cells were washed with PBS and 2 ml of HBSS was added containing 50 M arachidonic acid for 10 min. The HBSS was stored at Ϫ70°C for PG analysis. The cells were washed and either lysed in protein lysis buffer or scraped into Tri-Reagent for mRNA analysis.
ERK1/2 Activation by H 2 O 2 and Doxorubicin-Cells were treated with 50 M H 2 O 2 or 100 g/ml DX and 50 M PD098059 or 20 M SB203580 with the same positive and negative controls. All cells were then washed with PBS and ERK1/2 activity was measured using p44/ p42 MAP kinase assay kit.
Transcription Factor Decoys and COX-2 Expression-Transcription factor decoy oligomers were synthesized by a method previously described by Schmedtje et al. (27) in vitro by annealing the complementary strands in 1 ϫ annealing buffer (20 M Tris-HCl, pH 7.5, 20 M MgCl 2 , and 50 M NaCl). The mixture was heated to 80°C and allowed to cool to room temperature slowly over 3 h. The oligomers were designed from two NF-B binding regions and one AP-1-like binding region on the rat COX-2 promoter. The NF-B sequences used were, the distal site to the COX-2 promoter, (NF-B-1) Ϫ413 5Ј-GAGAG-GCAAGGGGATTCCCTTAGTTAG-3Ј Ϫ386 and the proximal site to the COX-2 promoter, (NF-B-2) Ϫ98 5Ј-GGGGGTGGGGAAAGCCGAG-GCGGAAA-3Ј Ϫ72. The AP-1-like sequence was Ϫ173 5Ј-TCATTT-GCGTGAGTAAAGCCTGCC-3Ј Ϫ149. A double-stranded DNA with a scrambled sequence (5Ј-CAGGAGAGTATCCTGCGATGCATCTGCT-3Ј) in the place of the NF-B and AP-1-like elements was used as a control in the decoy experiments. Decoy naked double-stranded DNA was placed in the medium at a concentration of 20 M 1 h before exposing cells to 50 M H 2 O 2 for 10 min, or 100 g/ml DX for 80 min. For the 3-h recovery fresh decoy naked double-stranded DNA was added to medium at 20 M, enabling sequestration of the transcription factor in the cytoplasm. As discussed under "Treatment of Cells with Hydrogen Peroxide and Doxorubicin," all cells were washed with PBS and 2 ml of HBSS was added containing 50 M arachidonic acid for 10 min. The HBSS was stored at Ϫ70°C for PG analysis. The cells were washed and TABLE I Prostaglandin generation (pg/mg of protein/10 min) in neonatal rat cardiomyocytes The main prostaglandins generated by cardiomyocytes were 6-keto-PGF 1␣ and PGE 2 , which were PMA-induced. This induction was blocked by NS-398, a specific COX-2 inhibitor. PGF 2␣ or TXB 2 formation was not significantly enhanced by PMA stimulation. Cells that were PMAtreated with aspirin show that PGF 2␣ is nonenzymatically generated and TXB 2 production was mostly COX-1-dependent.
either lysed in protein lysis buffer or scraped into Tri-Reagent for mRNA analysis. Reverse Transcription Polymerase Chain Reaction (PCR)-This procedure was used to compare the expression of COX-1 and COX-2 mRNAs between untreated and treated cells. Each pair of sense and antisense primers were designed to span at least one intron of the gene to exclude contaminating genomic DNA. Primers used were: glyceraldehyde 3-phosphate dehydrogenase (GAPDH) 5Ј-AACCCATCACCAT-CTTCCAGGAGC-3Ј (sense) and 5Ј-CACAGTCTTCTGAGTGGCAGTG-AT-3Ј (antisense); COX-1 5Ј-TCACAAGAGTACAGCTATGAGCAGT-3Ј (sense) and 5Ј-TGGGCTGGCACTTCTCCAGCATCAG-3Ј (antisense); COX-2 5Ј-CCCTGCTGGTGGAAAAGCCTCGTCC-3Ј (sense) and 5Ј-TA-CTGTAGGGTTAATGTCATCTAG-3Ј (antisense). Total RNA isolated from Tri-Reagent was reverse transcribed into cDNA and this was used as template for the PCR. RNA samples of 1 g were denatured at 65°C for 10 min. 100 ng of random hexamers, 1 ϫ reverse transcription buffer, 0.1 M dithiothreitol, 25 M deoxyribonucleoside triphosphates, and 200 units of Moloney murine leukemia virus reverse transcriptase were added and incubated at 37°C overnight in a reaction volume of 20 l. The reverse transcription reaction was stopped by heating to 95°C for 5 min. A 1:10 volume of the generated cDNA reaction was used in the subsequent amplification reaction. PCR was performed in a 50-l volume with 1.5 mM magnesium chloride, 1 mM deoxynucleoside triphosphates, 0.5 M of each sense and antisense primer, and 1 unit of Taq polymerase. The reaction cycles were denaturing at 94°C for 1 min, annealing at 56°C for GAPDH and COX-2 for 1 min, annealing at 52°C for COX-1 for 1 min, and extension at 72°C for 1 min. PCR cycle profiles for GAPDH, COX-1, and COX-2 were performed. Products were run on agarose gels (1.2%) with PGEM DNA markers. Polaroid instant photographs were taken of the gels with a Polaroid DS34 direct screen instant camera. Expression of COX-1 and COX-2 were normalized with respect to GAPDH from parallel samples. In order to verify that the band produced at the correct size was rat COX-2 cDNA, this band was excised from gel, purified, cloned into PCR 2.1 vector, and sequenced (ABI Prism 310 Genetic Analyzer, Perkin-Elmer).
Western Blotting-Lysates were prepared by treating cells with lysis buffer. Protein concentration was measured using a protein Bio-Rad assay, according to the manufacturer's procedure. Lysate samples (20 g of protein/lane) were applied to 10% SDS-polyacrylamide minigels, electrophoresed, and transferred to nitrocellulose membranes overnight at 40 V constant voltage at 4°C with use of a Western blot transfer apparatus. Immunostaining for COX expression was performed with a goat polyclonal anti-COX-1 antibody and a mouse monoclonal anti-COX-2 antibody, (R6), raised against the recombinant fulllength mouse (m) COX-2 expressed by baculovirus. R6 antibody is FIG. 2. Induction of COX-2, 6-keto-PGF 1␣ generation, and cell injury by doxorubicin. Cells were exposed to 100 g/ml doxorubicin for 80 min, washed, and allowed to recover for the times shown, (n ϭ 3). A, RT-PCR for COX-1, COX-2, and GADPH. There was no induction of COX-1 but COX-2 was induced at 10 min and at 3 h. By 10 h COX-2 expression had returned to normal. B, the generation of 6-keto-PGF 1␣ was induced at 3 h after doxorubicin exposure ( †, p Ͻ 0.05 versus controls). This increase was abolished by the COX-2 inhibitor NS-398 (1 M) and by free radical scavengers, PEG-superoxide dismutase (SOD) (200 units/ml) and PEG-CAT (200 units/ml). The non-selective COX inhibitor aspirin inhibited 6-keto-PGF 1␣ formation at all time points (**, p Ͻ 0.005 versus doxorubicin-treated). C, cell injury in response to doxorubicin was measured as release of LDH and increased at both 3 and 10 h ( †, p Ͻ 0.05 versus controls). Pretreatment with PEG-superoxide dismutase and PEG-CAT prevented the increase in LDH (*, p Ͻ 0.05 versus doxorubicin).
pre-absorbed with COX-2 (21 M) but not COX-1 (25 M) (28). Immunostaining for ERK1/2 was performed with mouse monoclonal antibodies and developed with a horseradish peroxidase Western blot detection kit.
LDH Assay-Cytotoxicity was assessed by measuring LDH in the culture medium spectrophotometrically. The LDH release was standardized with a cell injury index defined at (A-B)/(C-B) ϫ 100, where A ϭ LDH activity in the test sample; B ϭ LDH activity measured in media with no cells (0% control), and C ϭ LDH activity in samples from wells in which cells were lysed with Triton X-100 (100% control).
Determination of Eicosanoids by Gas Chromatography/Mass Spectrometry (GC/MS)-GC/MS was used to examine the range of PG formed by resting and PMA-stimulated cardiomyocytes. For each sample, 200 l of supernatant was removed and spiked with deuterated internal standards. Following 15 min of equilibration, the pH was lowered to pH 3 with 10% formic acid and the samples were extracted with ethyl acetate and dried under vacuum. The methoxylamine derivative was formed by adding 50 l of a solution containing 0.5% methoxylamine in pyridine and incubating overnight at room temperature. The sample was then dried under vacuum. Derivitization to the pentafluorobenzyl ester was accomplished by adding 10 l of diidopropylethylamine and 20 l of 10% pentafluorobenzyl bromide in acetronitrile to the methoxylamine derivative of the sample which was then allowed to stand overnight at room temperature. The sample was again dried under vacuum before being resuspended in 20 l of NЈ-OЈ-bis(trimethylsilyl)-acetamide and incubated at 37°C for 30 min. The samples were redried under vacuum and brought up in 15 l of dodecane. GC/MS was performed on a Varian 3400 gas chromatograph (GC) linked to a Finnigan Incos XL mass spectrometer (MS) operated in the negative ion, chemical ionization mode, using methane as the reagent gas. Source and interface temperatures were 150 and 310°C, respectively. The gas chromatograph was equipped with a 30-m MS5 capillary column and operated in splitless mode, with an injector temperature of 250°C. Following sample injection (3 l), the GC temperature was ramped from 190 to 310°C at 20°C/min. Analytes were monitored by selected monitoring of the mass ion, using INCOS data system.
Statistical Analysis-All quantitative data are expressed as mean Ϯ S.E. The data were analyzed by parametric unpaired Student's t test. p values Ͻ 0.05 were considered statistically significant.

COX Isoform Expression and Product Formation in Rat
Neonatal Cardiomyocytes-Morphological examination on the third day following isolation showed a synchronous and rhythmically contracting monolayer of cardiomyocytes. GC/MS analysis showed that the primary prostaglandin product in cardiomyocytes is prostacyclin Ͼ Ͼ PGE 2 ϾPGF 2␣ Ͼ TXB 2 (Table  I). Similar levels of PG formation by cardiomyocytes have been reported (29). Formation of both 6-keto-PGF 1␣ and PGE 2 was significantly enhanced in response to PMA but neither PGF 2␣ nor TXB 2 were altered (Table I). Selective inhibition of COX-2 activity by 1 M NS-398 specifically abolished the PMA-induced increase in 6-keto-PGF 1␣ and PGE 2 , demonstrating that the increase in these PGs was COX-2-dependent. Inhibition of FIG. 3. Cardiomyocyte response to 10 g/ml doxorubicin for 24 h. A, RT-PCR for COX-1, COX-2, and GADPH. 10 g/ml DX over 24 h induced COX-2 transcript levels which was prevented by PEG-superoxide dismutase (SOD) and PEG-CAT. B, doxorubicin also induced 6-keto-PGF 1␣ formation at 24 h ( †, p Ͻ 0.05 versus control) that was abolished by the selective COX-2 inhibitor NS-398, aspirin, and PEG-superoxide dismutase and PEG-CAT (*p Ͻ 0.05, p Ͻ 0.005). C, cell injury in response to doxorubicin was measured as release of LDH. There was a significant rise, ( †, p Ͻ 0.05 versus control) in the amount of LDH released after 24 h, which was prevented by treatment with PEG-superoxide dismutase and PEG-CAT (**, p Ͻ 0.005). Treatment of cells with the selective COX-2 inhibitor NS-398 alone over the 24-h period did not affect LDH release, however, NS-398 enhanced significantly (*, p Ͻ 0.05) the release of LDH by doxorubicin.
COX activity by 200 M aspirin prevented TXB 2 generation, showing that its generation was COX-1-dependent. PGF 2␣ appears to be generated non-enzymatically.
H 2 O 2 -induced COX-2 Expression at 3 h-Cardiomyocytes were exposed to 50 M H 2 O 2 for 10 min which did not effect cell viability. RT-PCR analysis (Fig. 1A) showed that H 2 O 2 -induced COX-2 expression at 3 h did not affect mRNA levels of COX-1 or the housekeeping gene, GAPDH. There was a corresponding increase in COX-2 protein, whereas the expression of COX-1 did not change (Fig. 1B). H 2 O 2 also induced 6-keto-PGF 1␣ production at 3 h, with levels returning to baseline by 10 h (Fig.  1C), as did the expression of COX-2 protein and mRNA. 1 M NS-398 had no effect on product formation at baseline or at 10 min of recovery, but abolished the increase in 6-keto-PGF 1␣ at 3 h, demonstrating that this increase was COX-2-dependent. NS-398 did not effect the expression of COX-2 message and was therefore only inhibiting product activity (data not shown). Aspirin, a non-selective COX inhibitor, reduced the amount of 6-keto-PGF 1␣ formed at all time points.
Doxorubicin Induces COX-2 Expression Due to Free Radical Formation-Addition of 10 or 100 g/ml doxorubicin to cells resulted in a marked and rapid reduction in their rate of contraction. Once doxorubicin was removed, however, and cells were allowed to recover for 20 min, a normal rate of contraction was restored. Addition of 100 g/ml doxorubicin to cells for 80 min, followed by a 3-h recovery, did not cause cells to apoptose appreciably. However, 10 g/ml doxorubicin over 24 h induced cell shrinkage, nuclear condensation, formation of apoptotic bodies, and blebs in the cell membrane in about 10% of cells. 100 g/ml doxorubicin induced a time-dependent change in COX-2 mRNA levels ( Fig. 2A), detectable after 10 min and peaking after 3 h recovery, without any change in COX-1 or GAPDH mRNA levels. 200 units/ml PEG-superoxide dismutase and PEG-CAT prevented COX-2 expression, demonstrating a role for free radicals in doxorubicin-mediated COX-2 induction. Doxorubicin (100 g/ml) also caused an increase in 6-keto-PGF 1␣ formation after 3 h (Fig. 2B). This increase in 6-keto-PGF 1␣ was abolished by 1 M NS-398 without preventing the expression of COX-2, demonstrating that the increase in product formation was COX-2-dependent. The increase in The release of LDH into cell medium was measured as a marker of cell injury (Fig. 2C). Following the initial 80-min exposure of cells to 100 g/ml doxorubicin, there was no detectable rise in LDH at 10 min. After 3 h, there was a significant increase indicating that cell injury had occurred. There was no further increase in LDH released from 3 to 10 h. The increase in LDH release was suppressed by pretreatment with PEGsuperoxide dismutase and PEG-CAT. PEG-superoxide dismutase alone increased the amount of LDH released, demonstrating the role of the hydroxyl radical in doxorubicinmediated cell injury.
The lower concentration of 10 g/ml doxorubicin also induced COX-2, but at a later time point. Doxorubicin (10 g/ml) induced COX-2 message at 24 h without affecting mRNA levels of COX-1 or GAPDH. Doxorubicin (10 g/ml) also caused an increase in COX-2-dependent 6-keto-PGF 1␣ formation (Fig. 3, A  and B). Again PEG-superoxide dismutase and PEG-CAT prevented COX-2 induction and product formation, suggesting a free radical-dependent mechanism. At this low drug concentration a continual accumulation of free radical metabolites of doxorubicin may be responsible for the COX-2 induction. It is also possible, however, that doxorubicin at this low concentration causes a steady increase in ceramide production by the cardiomyocyte (26) which is reported to induce apoptosis and mitochondrial H 2 O 2 production (27).
10 g/ml doxorubicin also caused the release of LDH which was prevented by PEG-superoxide dismutase and PEG-CAT (Fig. 3C). The COX-2 inhibitor NS-398 (1 M) alone did not increase the release of LDH above that of controls. The release of LDH, however, was aggravated by 1 M NS-398 (Fig. 3C), suggesting that a COX-2-dependent product protects the cells from injury.

Iloprost Protects Cardiomyocytes from Free Radical-induced
Damage-Since COX-2-dependent PGs protect cardiomyocytes from injury, we examined the effects of iloprost (a stable analog of prostacyclin) on LDH release by cells exposed to high concentrations of H 2 O 2 (as 50 M H 2 O 2 did not cause a substantial increase in LDH release following treatment). 30 nM iloprost reduced the amount of LDH released by H 2 O 2 and doxorubicininduced cell damage (Fig. 4, A and B). In contrast, an analog of PGF 2␣ , 30 M fluprostenol, did not prevent LDH release from cardiomyocytes, demonstrating that the reduction in LDH release was specific for iloprost.
Free Radicals Cause ERK Activation, Which Mediates COX-2 Expression-Incubation of cardiomyocytes with 50 M H 2 O 2 or 100 g/ml doxorubicin increased ERK1/2 activity (Fig. 5A). The increase in ERK1/2 activity was suppressed in the presence of the ERK1/2 inhibitor PD098059 (50 M) which also prevented the induction of COX-2 protein, mRNA, and product in response to H 2 O 2 or doxorubicin (Fig. 6, B, C, and D). In contrast, the p38 MAP kinase inhibitor, SB203580 (20 M), did not prevent COX-2 induction (Fig. 5, B, C, and D), suggesting that COX-2 expression was specifically ERK1/2 mediated.
The Transcription Factors AP-1 and NF-B Are Responsible for ERK Activation and Expression of COX-2 in Cardiomyocytes-Cardiomyocytes were treated with double-stranded DNA oligonucleotides (all at 20 M) corresponding to COX-2 promoter elements to act as decoys for transcription factors that may be generated and so localize the elements responsible for the induction of COX-2. A decoy corresponding to the AP-1-like promoter element prevented COX-2 expression in response to H 2 O 2 or doxorubicin (Fig. 6, A-D). An oligonucleotide decoy corresponding to the proximal NF-B-2 also reduced, but did not abolish the induction of COX-2, whereas an oligonucleotide decoy corresponding to the distal NF-B-1 COX-2 promoter had no effect.  (4) or doxorubicin (5), whereas the p38 MAP kinase specific inhibitor, SB203580 (20 M), had no effect (6,7). Inhibition of ERK1/2 or 38 MAP kinase did not effect COX-1 expression. D, 6-keto-PGF 1␣ formation by cardiomyocytes. PG formation was induced by H 2 O 2 and doxorubicin (2, 3) ( ‡, p Ͻ 0.05 versus control). Pretreatment with the ERK1/2 inhibitor PD098059 abolished this increase (*, p Ͻ 0.05 versus corresponding H 2 O 2 or doxorubicin-treated cells) (4 and 5), whereas the p38 MAP kinase inhibitor SB203580 had no effect (6 and 7).

DISCUSSION
Cyclooxygenase-2 has been identified recently in human cardiomyocytes in areas of myocardial infarction and from individuals with dilated cardiomyopathy, whereas no COX-2 was found in normal hearts (17). The mechanism for the induction of COX-2 in cardiomyocytes is unknown. In addition to growth factors and cytokines, shear stress, hypoxia, calcium ionophore, bradykinin, thrombin, and angiotensin II (30 -35) induce COX-2 expression. Oxidant stress may also induce COX-2 (12). Feng et al. (12) showed that the inhibition of prostaglandin synthesis in rat mesangial cells by free radical scavengers was due to the suppression of COX-2 expression.
As in the human cardiomyocyte, COX-1 was the only isoform expressed in resting rat cardiomyocytes while COX-2 was induced by treatment with a phorbol ester. COX-2 was also induced by two oxidants, doxorubicin at a therapeutically relevant concentration, and brief exposure to a relatively low concentration of H 2 O 2 . Doxorubicin is an anthracycline used as a chemotherapeutic agent and is well recognized to induce a dilated cardiomyopathy (36). A hydroxyl radical originating from a superoxide radical has been implicated in anthracycline-mediated cardiotoxicity (7,8). Alternatively, toxicity may be due to accumulation of ceramide, a lipid formed when cells are exposed to anthracyclines and which may induce mitochondrial H 2 O 2 production (37). Doxorubicin also induced toxicity in cultured cardiomyocytes detected as release of LDH. Both the expression of COX-2 and the release of LDH were abolished by the combination of superoxide dismutase and catalase confirming a free radical-dependent mechanism.
Oxygen-free radicals have indiscriminate effects on a variety of cellular targets including lipids, proteins, and oligonucleotides, resulting in cell death. At lower concentrations, however, oxygen-free radicals activate very specific pathways that in turn regulate the expression of genes influencing cell survival. Exogenous free radicals have been shown to activate p38 MAP kinase in several cell systems and less often to activate ERK1/2. Xia et al. (38) reported that free radical-induced activation of ERKs promotes cell survival, whereas activation of JNK and p38 MAPK induce apoptosis. Similarly, ceramideinduced activation of JNK is important for the induction of apoptosis (39). These results suggest that ERKs have a protective role, while activation of p38 MAPK/JNK leads to induction of apoptotic death on exposure to free radicals. The pattern of MAP kinase activation in response to oxidant stress varies between cells even in the same species. In rat vascular smooth muscle cells free radicals activate p38 MAP kinase but not ERK1/2 (40). In contrast, in rat neonatal cardiomyocytes H 2 O 2 specifically activates ERKs through activation of the Src family of tyrosine kinases, Ras and RAF-1 (23). The activation of ERKs in these cells prevented H 2 O 2 -induced apoptosis al-  2 and 3). The scrambled decoy oligomers did not prevent COX-2 expression (lanes 4 and 5). However, the AP-1 double-stranded decoy oligomer prevented COX-2 expression in response to either H 2 O 2 or doxorubicin (lanes 6 and 7). A double-stranded decoy oligomer corresponding to the distal NF-B-1 had no effect on COX-2 expression (lanes 8 and 9). A double-stranded decoy oligomer corresponding to the proximal NF-B-2 reduced but did not abolish COX-2 induction by H 2 O 2 or doxorubicin (lanes 10 and 11). C, corresponding changes in 6-keto-PGF 1␣ formation. H 2 O 2 or doxorubicin induced 6-keto-PGF 1␣ formation (**, p Ͻ 0.005 versus controls). The increase was abolished by the AP-1 double-stranded decoy oligomer and reduced by the proximal NF-B-2 double-stranded decoy oligomer (*, p Ͻ 0.05). though the mechanism was not identified. Similarly in our studies ERK1/2 was activated by H 2 O 2 and doxorubicin and was responsible for the activation of COX-2. Thus, PD098059, which prevents MEK-dependent activation of ERK1/2, abolished the expression of COX-2 mRNA and protein and the increase in prostaglandin formation. In contrast, specific inhibition of p38 MAP kinase had no effect.
The Src family of tyrosine kinases has been implicated in the induction of COX-2 in response to cytokines and growth factors, but not specifically the ERK kinase (41). The ERK1/2 kinase pathway, however, has been identified as the common signaling pathway of serotonin or phorbol ester-induced COX-2 expression in rat mesangial cells (42) and both ERK kinase and MAPK p38 have been shown to play a role in lysophosphatidic acid-mediated COX-2 expression by these cells (43). To identify how ERK1/2 induced COX-2 in cardiomyocytes we examined three putative sites on the promoter of the rat COX-2 gene. Two of the sites are homologous to the NF-B element, which is also found in the promoters of human inflammatory genes, including the human COX-2 gene. The third, the AP-1-like site is similar to a promoter element common to early response genes. In high concentration, H 2 O 2 induces translocation of NF-B to the nucleus and subsequent gene expression in an endothelial cell line (27). This is by no means a ubiquitous event and in some cells there is gene induction in response to H 2 O 2 without NF-B translocation (27). Here, we show that AP-1 is the major promoter element involved, as a decoy promoter AP-1 sequence abolished COX-2 induction. ERK1/2 may induce c-Fos expression which complexes in a heterodimer with c-Jun forming the transcription factor that acts through AP-1. This is thought to explain interleukin-1␤ induction of nitric oxide in rat pancreatic islet cells, which is also ERK1/2 dependent (44). Moreover, increased activation of ERKs and of the AP-1 promoter has been shown in proliferating smooth muscle cells (45,46). A double-stranded oligonucleotide decoy identical to the proximal NF-B site showed that this site was also partially active, whereas the distal NF-B site was inactive. These findings implicate both the proximal NF-B and the AP-1 site, but suggest that the latter is the critical promoter element. It is worth noting that both AP-1 and NF-B are activated by H 2 O 2 in neonatal cardiomyocytes (47) and that intracellular oxidants may be the ultimate messengers common to all signals that activate AP-1 and NF-B (48). H 2 O 2 and DX induced the release of LDH from cardiomyocytes, and the selective COX-2 inhibitor NS-398 potentiated this. In contrast, treating the cells with NS-398 alone did not affect the basal levels of LDH release. The injury response was prevented by iloprost, a stable analogue of the principle cyclooxygenase product of these cells, prostacyclin. The response to iloprost appeared to be specific as it was dose dependent and no protection was seen with fluprostenol, a stable analogue of PGF 2␣ , which induces hypertrophy of rat cardiomyocytes (49). These data suggest that COX-2-dependent prostaglandins protect the cardiomyocyte from oxidant injury. Prostaglandins are cytoprotective in several tissues, including stomach (50,51) and heart (52). COX-2 expression also protects against apoptosis in a number of cancer cell lines (53) and against nitric oxide-mediated apoptosis (54). Similarly, rats injected with lipopolysaccharide (a known inducer of COX-2) are protected from ischemia/reperfusion injury (55). Prostaglandins, particularly metabolites of PGJ 2 , interact with the peroxisomal proliferator activator receptor-␥, a nuclear envelope receptor that acts as a transcription factor for apoptosis regulating genes (56). Prostacyclin has not been shown, however, to react with this receptor, so the mechanism for COX-2 dependent protection of the cardiomyocyte seen in our experiments is unknown.
Adaptive responses of bacteria to the potential toxic effects of partially reduced oxygen metabolites (H 2 O 2 , ⅐ OH, and O 2 . ) include the transcriptional regulation of oxidative stress genes (57). Similar mechanisms exist in eukaryocytic cells where reactive oxygen species cause expression of several enzymes involved in oxidant metabolism such as the human glutathione peroxidase gene (58,59). The induction of COX-2 in cardiomyocytes may represent an additional example of an adaptive response that protects the cell from oxidant stress.