Cyclooxygenase-2 Induction and Prostacyclin Release by Protease-activated Receptors in Endothelial Cells Require Cooperation between Mitogen-activated Protein Kinase and NF-κB Pathways*

The functional significance of protease-activated receptors (PARs) in endothelial cells is largely undefined, and the intracellular consequences of their activation are poorly understood. Here, we show that the serine protease thrombin, a PAR-1-selective peptide (TFLLRN), and SLIGKV (PAR-2-selective peptide) induce cyclooxygenase-2 (COX-2) protein and mRNA expression in human endothelial cells without modifying COX-1 expression. COX-2 induction was accompanied by sustained production of 6-keto-PGF1α, the stable hydrolysis product of prostacyclin, and this was inhibited by indomethacin and the COX-2-selective inhibitor NS398. PAR-1 and PAR-2 stimulation rapidly activated both ERK1/2 and p38MAPK, and pharmacological blockade of MEK with either PD98059 or U0126 or of p38MAPK by SB203580 or SB202190 strongly inhibited thrombin- and SLIGKV-induced COX-2 expression and 6-keto-PGF1α formation. Thrombin and peptide agonists of PAR-1 and PAR-2 increased luciferase activity in human umbilical vein endothelial cells infected with an NF-κB-dependent luciferase reporter adenovirus, and this, as well as PAR-induced 6-keto-PGF1α synthesis, was inhibited by co-infection with adenovirus encoding wild-type or mutated (Y42F) IκBα. Thrombin- and SLIGKV-induced COX-2 expression and 6-keto-PGF1α generation were markedly attenuated by the NF-κB inhibitor PG490 and partially inhibited by the proteasome pathway inhibitor MG-132. Activation of PAR-1 or PAR-2 promoted nuclear translocation and phosphorylation of p65-NF-κB, and thrombin-induced but not PAR-2-induced p65-NF-κB phosphorylation was reduced by inhibition of MEK or p38MAPK. Activation of PAR-4 by AYPGKF increased phosphorylation of ERK1/2 and p38MAPK without modifying NF-κB activation or COX-2 induction. Our data show that PAR-1 and PAR-2, but not PAR-4, are coupled with COX-2 expression and sustained endothelial production of vasculoprotective prostacyclin by mechanisms that depend on ERK1/2, p38MAPK, and IκBα-dependent NF-κB activation.

The principal mechanism through which serine proteases regulate cell behavior is by activation of a unique family of G-protein-coupled receptors, referred to as protease-activated receptors (PARs) 2 (1,2). These receptors are activated by the proteolytic activities of their ligands, which unmask a cryptic N-terminal receptor sequence that binds to and triggers receptor function while remaining tethered. Molecular cloning has identified four subtypes of the PAR family that exhibit differential tissue expression as well as selectivity in activation by serine proteases and by peptides that mimic the tethered ligand sequences. PAR-1 is a ubiquitously expressed receptor and is activated by the multifunctional serine protease, thrombin, and by the synthetic PAR-1 peptide TFLLRN. PAR-4 is a second thrombin receptor that binds thrombin with low affinity and can be activated by the polypeptide sequence AYPGKF (3,4). Signaling through PAR-2, in contrast, is preferentially stimulated by trypsin, tryptase, and membrane-type serine protease-1 (5) and can be activated independently of proteolysis by the selective PAR-2 peptide SLIGKV. Cellular responses to thrombin and to other proteases are therefore determined in part by the expression of PARs as well as by the differential recruitment of signaling pathways following ligand binding.
Endothelial cells respond to PAR-1 activation by thrombin with a number of functional responses including secretion of von Willebrand factor (6), up-regulation of IL-8 synthesis (7), and expression of adhesion molecules (8). Thrombin also causes endothelial cell proliferation, which is essential for wound healing and angiogenesis (9). However, whereas this protease clearly promotes both proinflammatory and prothrombotic events in the vascular wall, thrombin also increases endothelial expression of complement inhibitory proteins (e.g. decay-accelerating factor) (3) and enhances the release of vasodilator molecules that have vasculoprotective properties (6,11). In particular, we have shown that thrombin exposure induces rapid and sustained phases of prostacyclin (PGI 2 ) synthesis by human endothelial cells (12). PGI 2 contributes to the acute inflammatory response by promoting vasodilation and increased vascular permeability, is a potent antiplatelet agent, and has antiproliferative and antifibrotic effects mediated by paracrine actions on vascular smooth muscle and fibroblasts, respectively (13,14). The potential effects of PAR-2 activation on endothelial prostanoid produc-tion have received little attention, but there is some evidence that PAR-2 can mediate protective anti-inflammatory responses in vivo (15) as well as contributing to inflammation (16). PAR-2 is strongly induced in endothelial cells following cytokine challenge (17), and it has been suggested that activation of this receptor is functionally significant only in the context of inflammation. We have recently shown, however, that human endothelial cells respond to PAR-2 activation with increased prostanoid synthesis without the requirement for cytokine exposure (12), suggesting that PAR-2 is also a physiologically relevant receptor. Understanding how signaling through distinct PARs regulates endothelial production of PGI 2 therefore has important implications for both normal and pathophysiological control of the vasculature.
Prostaglandins are synthesized through the activities of two isoforms of cyclooxygenase (COX), designated COX-1 and COX-2 (18). COX-1 is a constitutive enzyme present in most tissues. In contrast, COX-2, the product of a related gene, is strongly up-regulated by cytokines and growth factors (19,20) and is now recognized to be constitutively expressed in several tissues (20). Regulation of COX-2 occurs at both transcriptional and post-transcriptional levels, and mitogen-activated protein kinases (MAPKs) are known to be involved in regulating COX-2 expression in response to cytokines (21)(22)(23). ERK1/2 and p38 MAPK have been implicated in both acute and chronic responses to extracellular stimuli in endothelial cells and we have shown that both of these MAPK families are capable of regulating acute thrombin-stimulated PGI 2 synthesis through their direct and indirect effects on Group IV phospholipase A 2 ␣ activation (24). We have also previously shown that COX-2 expression in primary human endothelial cells is enhanced by exposure to thrombin and the PAR-1-selective peptide TFLLRN and have provided evidence that activation of PAR-2 by SLIGKV promotes increased COX-2 expression with kinetics distinct from those evident in thrombin-stimulated cells (12). The signaling events that link PAR-1 and PAR-2 with COX-2 expression and sustained prostanoid production are not defined, but the ERK1/2 and p38 MAPK families may play important roles.
The COX-2 gene contains numerous cis-acting promoter elements, including NF-B sites (25). NF-B is a sequence-specific transcription factor that regulates expression of numerous genes and can exert protective or detrimental effects, depending upon the cellular context (26 -28). In resting cells, NF-B is complexed in the cytoplasm with its inhibitory subunit IB␣. Agonist stimulation promotes serine phosphorylation of IB␣, which triggers its proteasomal degradation and subsequently activates NF-B. NF-B can also be activated by an alternative mechanism that involves phosphorylation of IB␣ on Tyr 42 (27). In addition, optimal induction of NF-B-dependent genes may require agonist-mediated phosphorylation of NF-B proteins within their transactivation domain (29). There is some evidence that thrombin and a PAR-2-selective peptide can activate NF-B in vascular cells (30,31), suggesting that the NF-B pathway may be an important component of PAR-mediated signaling. However, whereas COX-2 expression in several cell types is at least partly dependent on NF-B activity, the molecular details of its activation in response to ligand binding to PARs are unknown, and hence its significance as a regulator of endothelial PGI 2 synthesis remains to be determined.
We have investigated the molecular signaling mechanisms underlying PAR-induced COX-2 expression and PGI 2 synthesis in human endothelial cells. We show for the first time that activation of endothelial PAR-1 and PAR-2, but not PAR-4, promotes COX-2 induction and sustained formation of the vasculoprotective mediator PGI 2 through mechanisms that require activation of ERK1/2 and p38 MAPK as well as IB␣-dependent NF-B activation. We further show that phosphoryl-ation of the p65 subunit of NF-B is a component of both PAR-1-and PAR-2-mediated endothelial activation. These findings have important implications for understanding the vascular effects of therapeutic interventions that target COX-2 and upstream activators of NF-B.

EXPERIMENTAL PROCEDURES
Reagents-Human ␣-thrombin, bovine serum albumin (BSA; fraction V), and polyvinylidene difluoride membranes (Immobilon-P TM ) were all purchased from Sigma. The PAR-1-selective peptide TFLLRN and the PAR-2 peptide SLIGKV were obtained from Bachem (St. Helens, Merseyside, UK), and 2F-LIGRLO was from Peptides International Inc. (Louisville, KY). These peptides have been extensively characterized with respect to their selectivity of action at PARs (e.g. see Ref. 32). Human recombinant IL-1␣ was from R&D Systems (Oxford, UK), and the bicinchoninic acid (BCA) protein assay was from Pierce. PG490 (triptolide), PD98059, SB203580, SB202190, and U0126 were from Calbiochem. Reagents for SDS-PAGE were purchased from Bio-Rad (Hemel Hempstead, Hertfordshire, UK) and National Diagnostics (Hessle, Hull, UK). [ 3 H]6-keto-PGF 1␣ was obtained from Amersham Biosciences. The RNeasy Mini kit and the Superscript II reverse transcriptase were from Qiagen Ltd. (Crawley, West Sussex, UK) and Invitrogen, respectively. The DyNAmo SYBR Green quantitative PCR kit was purchased from Finnzymes (Espoo, Finland). Other molecular biology reagents were all obtained from Promega (Southampton, UK). Culture media were purchased from Sigma or Invitrogen. All other reagents were obtained from Sigma or BDH (Poole, Dorset, UK) at the equivalent of AnalaR grade.
Cell Culture-Human umbilical vein endothelial cells (HUVEC) were isolated and cultured as previously described (12,24). Briefly, cells were grown in medium M199 (Sigma) supplemented with 20% (v/v) fetal calf serum, 4 mM glutamine, 100 units/ml penicillin, 100 units/ml streptomycin, and 20 mM NaHCO 3 and cultured at 37°C in a 5% CO 2 , 95% air atmosphere in 25-mm 2 tissue culture flasks (BD Biosciences) precoated with 1% (w/v) gelatin. At confluence, cells were passaged into 75-mm 2 tissue culture flasks and cultured in the presence of 20 mg/ml endothelial cell growth factor. Experiments were performed on confluent passage 2 cells grown in the appropriate gelatin-coated cell culture dishes.
RNA Extraction and Real Time Reverse Transcription-PCR-Quiescent HUVEC were treated as described in the legends to Figs. 2 and 5, and RNA was extracted using a Qiagen RNeasy minikit according to the manufacturer's instructions. Total HUVEC RNA was reverse transcribed by incubating 1 g of RNA with 0.5 g of oligo(dT) 15 primer and 200 units of M-Superscript reverse transcriptase for 1 h at 37°C, 15 min at 42°C, and 3 min at 98°C. 10% of the resulting cDNA was amplified using the SYBR Green quantitative PCR kit. Primers used for real time PCR were synthesized by MWG-Biotech (Milton Keynes, UK). The 5Ј to 3Ј sequences used were as follows: GCATCTTCTTTTGCGTCGCC (GAPDH-1 forward), GTCATTGATGGCAACAATATCC (GAPDH-1 reverse), GCCCAG-CACTTCACGCATCAG (COX-2 forward), and AGACCAGGCACCA-GACCAAAGACC (COX-2 reverse). The PCR was performed (Opticon II) using the following parameters: 95°C for 2 min, 30 cycles of 93°C for 1 min, and 60°C for 1 min and a final extension step at 60°C for 7 min. Each analysis contained a range of standards (known concentrations of the same target sequence). Data were analyzed using the Opticon II monitor 2 analysis software, and a standard curve was plotted that correlated cycle number with the amount of product formed after each cycle. COX-2 mRNA levels were normalized to GAPDH for each sample.
Analysis of NF-B Activation Using Luciferase Adenovirus Infection-All viruses were E1/E3 (early transcribed regions)-deficient and belonged to the Adv5 serotype. Viruses were propagated in 293 human embryonic kidney cells (American Type Culture Collection, Manassas, VA) and purified by cesium chloride ultracentrifugation. Titers of viral stocks were determined by plaque assay on 293 cells. Replication-deficient control adenoviral vector without insert (Ad0) was provided by Drs A. Byrnes and M. Wood (Oxford University). Adenovirus encoding porcine IB␣ with a cytomegalovirus promoter and a nuclear localization sequence (AdIB␣) was provided by Dr. R. de Martin (University of Vienna). The adenovirus expressing mutated human IB␣ (tyrosine 42 to phenylalanine; AdIB␣ Y42F) was from Prof. J. F. Engelhardt (University of Iowa). Efficient infection of cells (Ͼ95%) was confirmed using virus encoding bacterial ␤-galactosidase at a multiplicity of infection (MOI) of 100:1 followed by the addition of the fluorimetric substrate, fluorescein-di-␤-galactopyranoside (Sigma) (data not shown). Measurement of NF-B transactivation is based on the use of an adenovirus reporter vector, which contains a luciferase gene under the control of NF-B (AdNF-B-Luc). This adenovirus reporter was provided by Dr. P. B. McCray, Jr. (University of Iowa) and is a modification of the pNF-B-Luc reporter vector (BD Biosciences/Clontech). pNF-B-Luc contains the firefly luciferase gene from Photinus pyralis and four tandem copies of the NF-B consensus sequence fused to a TATA-like promoter from the herpes simplex virus thymidine kinase promoter. The vector backbone also contains an f1 origin for single-stranded DNA production, a pUC origin of replication, and an ampicillin resistance gene for propagation and selection in Escherichia coli. Therefore, NF-B-directed promoter activity is estimated as being proportional to the firefly luciferase activity.
For adenoviral infection, HUVEC were seeded at 80% confluence (9,600 cells) in 30-mm 2 wells and infected in serum-free medium (RPMI 1640) with Ad0 or AdNF-B-Luc (MOI of 100:1) with or without coinfection with AdIB␣ or AdIB␣ Y42F (MOI 100:1). After 1 h of infection, medium containing free adenoviruses was replaced with complete culture medium for 24 -36 h to allow expression of the gene(s) of interest. Cells were then stimulated for the time periods detailed in the legend to Fig. 8, subsequently washed with sterile phosphate-buffered saline (PBS), and lysed in 100 l of lysis buffer (0.65% Nonidet P-40, 10 mM Tris (pH 8.0), 1 mM EDTA, 150 mM NaCl) on ice for 5 min. 50 l of lysate were added to the wells of a luminometer cuvette strip containing 120 l of luciferase assay buffer (25 mM Tris-phosphate (pH 7.8), 8 mM MgCl 2 , 1 mM EDTA, 1% (v/v) Triton X-100, 1% (v/v) glycerol, 1 mM dithiothreitol, 0.5 mM ATP). Luciferase activity was measured using a LabSystems luminometer by dispensing 30 l of luciferin (Bright-Glo TM luciferase assay system; Promega)/assay point. Luciferase activity was normalized in each experiment to the amount of protein.
Immunofluorescence-HUVEC were plated on 13-mm glass coverslips in 4-well plates (Nunc-Nalgene) and grown for 24 -48 h until 80% confluent. Following the required treatments, cells were fixed in 4% paraformaldehyde, washed with 0.5% BSA-PBS containing 0.5% BSA (PBS-BSA), and permeabilized in PBS-BSA supplemented with 0.1% Triton. Coverslips were then incubated for 1 h with anti-p65NF-B antibodies (10 g/ml) in PBS-BSA at room temperature. After incubation, coverslips were washed twice in PBS-BSA and incubated with fluorescein isothiocyanate-conjugated secondary goat anti-rabbit antibodies (1:100) for 45 min. Cells were then washed (twice), and slides were mounted with either Fluorsave (Dako, Cambridgeshire, UK) or with propidium iodide-containing mounting medium (Vectashield, Burlingame, CA). Immunofluorescence was monitored by confocal microscopy using a Zeiss LSM 510 inverted microscope.
Measurement of PGI 2 Release-Confluent cultures of HUVEC in 24-well tissue culture trays were treated as described in the legends. Supernatants were assayed for 6-keto-PGF 1␣ (the stable hydrolysis product of PGI 2 ) by radioimmunoassay as previously described (12).
Statistical Analysis-Student's t test or ANOVA, as appropriate, were used to compare means of groups of data. p Ͻ 0.05 was considered statistically significant. 2 Synthesis in HUVEC-Peptide agonists for PARs have been extensively characterized for PAR selectivity (e.g. see Ref. 32) and are currently in wide use to interrogate the functional significance of PARs in a number of systems. We have recently shown that thrombin enhances expression of COX-2 protein and mRNA in HUVEC and that this is accompanied by prolonged PGI 2 synthesis (12). We also provided evidence that the human PAR-2-selective agonist peptide SLIGKV increased COX-2 protein in HUVEC and enhanced COX-2 mRNA and PGI 2 release after 2 and 6 h, respectively (12). To gain further insight into the regulation of COX-2 expression and PGI 2 synthesis in response to activation of PARs, we initially extended these findings by comparing, within the same HUVEC cultures, the time courses of expression of COX-2 (mRNA and protein) as well as examining the detailed time courses of PGI 2 synthesis for all three stimulants in comparison with IL-1␣. Both thrombin and SLIGKV enhanced COX-2 expression in HUVEC in a concentrationand time-dependent manner ( Fig. 1). Thrombin (1 units/ml) induced a significant increase in COX-2 protein expression after 2 h, and this was maintained and further increased after 4-and 8-h exposure. Similar induction kinetics were evident in cells challenged with the PAR-1selective peptide TFFLRN (data not shown). In contrast, SLIGKV (100 M) (Fig. 1A) or the metabolically stable PAR-2 peptide 2F-LIGRLO (10 M) (data not shown) maximally induced COX-2 expression after 2 h, and this was sustained but not further enhanced with prolonged exposure to peptide. In keeping with our previous findings (12), the expression of COX-1 was not significantly modified by either thrombin or SLIGKV (Fig. 1, A and B). We next examined whether PAR-stimulated COX-2 protein expression is accompanied by changes in COX-2 mRNA expression using real time reverse transcription-PCR analysis. Both thrombin and SLIGKV caused a 3-4-fold increase in COX-2 mRNA expression after 1-2 h (Fig. 2). However, in SLIGKV-stimulated HUVEC, mRNA levels declined rapidly after 1 h, whereas expression peaked at 2 h after thrombin exposure and was still evident after 4 h. Together, these results show that activation of either PAR-1 or PAR-2 in HUVEC promotes increased COX-2 expression but that the time courses of induction differ between agonists.

PAR Agonists Differentially Induce COX-2 Expression and PGI
We have previously shown that exposure of HUVEC to thrombin triggers a biphasic PGI 2 synthesis characterized by an initial rapid release occurring within 30 min of exposure followed by a sustained and prolonged synthesis lasting for several hours (12). Here, we have confirmed this effect and characterized the PGI 2 response to PAR-2 stimulation. As depicted in Fig. 3, a 30-min exposure to thrombin (1 unit/ml) markedly enhanced (up to 18-fold) formation of the stable PGI 2 hydrolysis product, 6-keto-PGF 1␣ , whereas little effect of either SLIGKV or IL-1␣ was evident at this time point. Increasing the exposure time to thrombin further increased release, which reached a sustained peak between 4 and 8 h. Similar results were obtained with the PAR-1 peptide TFFLRN (data not shown). In contrast, no significant 6-keto-PGF 1␣ generation was evident in SLIGKV-stimulated cells until 2-4 h after exposure. Thus, the time of onset of PGI 2 synthesis as well as the magnitude of the response differs between PAR agonists. To confirm the involvement of COX-2 activity in these responses, we used the COX-2selective inhibitor NS398. As shown in Fig. 3 (inset), 6-keto-PGF 1␣ formation in unstimulated HUVEC and in cells subject to prolonged exposure to thrombin was inhibited by ϳ80% following pretreatment with NS398 (1 M), but virtually abolished by exposure to the COX-1/2 inhibitor indomethacin. These results confirm the importance of COX-2 activity in prolonged PGI 2 release in response to thrombin and suggest that a component of thrombin-induced synthesis depends upon COX-1 activity (12,24,52).
Involvement of MAPK Signaling Cascades in PAR-induced COX-2 Expression and PGI 2 Synthesis-We and others have shown that MAPK signaling pathways are important regulators of acute prostanoid synthesis by endothelium (24). However, the molecular mechanisms involved in regulating prolonged endothelial prostanoid release, particularly in response to PAR activation, remain to be determined. To define the importance of ERK1/2 and p38 MAPK as regulators of PAR-induced COX-2 expression and prostanoid release, we initially examined the effects of PAR agonists on the phosphorylation status of ERK1/2 and p38 MAPK . Thrombin (Fig. 4A) and the PAR-1-selective peptide TFLLRN (not shown) caused a rapid and robust activation of ERK1/2 that was still evident after 2 h but was significantly reduced ( p Ͻ 0.05) compared with that evident after 10 min. In contrast, stimulation with the PAR-2 pep-  tide produced a rapid increase in ERK1/2 activation that did not significantly decline even after 4 h of exposure. SLIGKV also activated p38 MAPK with similar levels of activation evident at both early and late time points (Fig. 4B). The PAR-4-selective peptide AYPGKF promoted activation of ERK1/2 and p38 MAPK in HUVEC that was evident after 10 min of stimulation (Fig. 4C), which contrasts with our previous demonstration that ERK1/2 activation by thrombin or PAR-1 peptide occurs more rapidly (1-5 min of exposure) (33). However, in contrast to PAR-1 or PAR-2 stimulation, MAPK activation in PAR-4 peptide-stimulated cells was not associated with enhanced COX-2 expression (Fig. 4C) or 6-keto PGF 1␣ formation (not shown). To explore the role of ERK1/2 and p38 MAPK activation in PAR-induced COX-2 expression, we next determined the effects of pharmacological blockade of MEK1/2 or p38 MAPK on COX-2 protein and mRNA levels in HUVEC challenged with thrombin and PAR-1-and PAR-2-selective peptides. We have shown previously that these pharmacological inhibitors block agonist-induced MAPK activation in HUVEC in a concentration-dependent manner (24). As shown in Fig. 5, the MEK inhibitors PD98059 (1-10 M) and U0126 (0.3-3 M) and the p38 MAPK inhibitor SB203580 (1-10 M) dose-dependently reduced thrombin-induced, TFLLRN-induced (not shown), and SLIGKV-induced COX-2 protein expression. A combination of PD98059 and SB203580 caused a greater reduction in expression than either inhibitor alone (Fig. 5A). U0126 and SB203580 at maximally effective concentrations also reduced basal COX-2 mRNA and abolished COX-2 mRNA induction in both thrombin-and SLIGKV-stimulated endothelial cells (Fig. 5C). Blocking COX-2 activity with NS398, however, did not affect COX-2 protein expression in thrombin-, TFFLRN-, or SLIGKV-stimulated HUVEC (Fig. 7A). These data show that, in contrast to COX-2 induction by IL-1␣, which is minimally affected by blockade of the MEK-ERK pathway, 3 COX-2 expression in HUVEC challenged with PAR agonists is strongly dependent upon both ERK1/2 and p38 MAPK activation. They also show that COX-2 induction by these stimuli is independent of COX-2 activity. To define the MAPK dependence of COX-2-derived PGI 2 synthesis, we examined the effects of U0126 and SB202190 on thrombin-, SLIGKV-, and 2F-LIGRLO (not shown)-induced 6-keto-PGF 1␣ formation (Fig. 5D). Preexposure to either inhibitor or to the NF-B inhibitor PG490 markedly inhibited 3 R. A. Houliston and C. P. D. Wheeler-Jones, manuscript in preparation. 6-keto-PGF 1␣ generation in response to thrombin or the PAR-2-activating peptides.
Role of NF-B in PAR-mediated Signaling-COX-2 expression in endothelial cells exposed to a number of proinflammatory or mitogenic stimuli is known to depend in part upon the activation of NF-B (34 -36). To begin to define the involvement of NF-B-dependent signaling events in regulating prostanoid production in response to PAR activation, we initially assessed nuclear translocation of p65-NF-B by confocal microscopy. As shown in Fig. 6, a 2-h exposure to either thrombin or IL-1␣, but not the PAR-4 peptide (data not shown), caused a marked translocation of the p65 subunit of NF-B from the cytoplasm to the nucleus. Notably, the PAR-2 peptide 2F-LIGRLO at 10 M also strongly induced the nuclear translocation of p65-NF-B. We next examined the consequences of suppression of NF-B activity on PAR-induced COX-2 expression. Pretreatment with the NF-B inhibitor PG490 (25 ng/ml) abrogated thrombin-stimulated, TFFLRN-stimulated (not shown), and SLIGKV-stimulated COX-2 protein expression by 70 -80% (Fig. 7B) without modifying basal COX-2 protein or basal COX-2 mRNA levels Densitometric analyses of p-ERK1/2 and p-p38 MAPK immunoblots from three separate experiments are given as mean Ϯ S.D. *, p Ͻ 0.05 versus unstimulated control cells. C, HUVEC were exposed to the PAR-4-selective peptide AYPGKF (50 M) for the times indicated, and the resulting blots were probed for phosphorylated p38 MAPK , phosphorylated ERK1/2, or COX-2. Data are from a single experiment representative of two with similar results. (Fig. 5C). PG490 treatment also reduced PAR-stimulated COX-2 mRNA expression by between 50 and 80% (Fig. 5C) without affecting expression of the COX-1 isoenzyme (not shown). The proteasome inhibitor MG-132 also partially inhibited PAR-2-selective peptide-induced as well as thrombin-induced COX-2 expression (Fig. 7C). In these experiments, 6-keto-PGF 1␣ generation was also partly reduced by MG-132 treatment (28 Ϯ 5 and 32 Ϯ 6% inhibition of thrombin-and SLIGKV-induced release, respectively; mean Ϯ S.E., n ϭ 2-4 experiments). Further parallel studies showed that sustained 6-keto-PGF 1␣ formation in response to thrombin, SLIGKV, or 2F-LIGRLO (not shown) was also inhibited by blockade of NF-B with PG490 (Fig. 5D). Together, these results suggest that NF-B-dependent signaling events are important for regulating COX-2 expression and PGI 2 synthesis mediated by PAR-1 and PAR-2 activation.
To further investigate the functional significance of NF-B activation in response to PAR stimulation, we examined the effect of thrombin and PAR-selective peptides on the transcriptional activation of NF-B in HUVEC infected with an adenoviral reporter containing four NF-B consensus sequences upstream of luciferase (37). As shown in Fig. 8A, NF-B-directed promoter activity was increased by exposure to thrombin (1 unit/ml) in a biphasic manner with significant activation observed at both early (30 min) and late (4 -6 h) exposure times. In cells infected with a control adenovirus, no luciferase activity was observed (data not shown). Thrombin therefore induces a biphasic activation of NF-B in HUVEC. Thrombin-induced NF-B activation was also concentrationdependent (Fig. 8B), with activation detected at 0.01 units/ml and maximum at 1 unit/ml. We have previously shown that thrombin-driven COX-2 expression and PGI 2 synthesis show similar concentration dependences in thrombin-stimulated HUVEC (12). We next examined the effects of PAR-1-, PAR-2-, and PAR-4-selective peptides on luciferase activity in infected endothelial cells (Fig. 8C). The PAR-1 peptide TFLLRN strongly increased luciferase activity, with a peak evident at 6 h after exposure. SLIGKV, the PAR-2 peptide, also enhanced NF-B activity, which was smaller in magnitude than that evoked by thrombin but was sustained over a longer time period (Fig. 8C). Consistent with its lack of effect on COX-2 expression and PGI 2 formation (Fig. 4C), the PAR-4 peptide did not modify luciferase activity (Fig. 8C). To define the potential involvement of IB␣ in mediating PAR-induced NF-B acti-vation, luciferase activity induced by thrombin was further analyzed following co-infection with either a native IB␣-expressing adenovirus (AdIB␣) or with an adenovirus overexpressing IB␣ in which tyrosine residue 42 was mutated to phenylalanine (AdIB␣ Y42F). As depicted in Fig. 8A, biphasic NF-B activation by thrombin was inhibited by coexpression of native IB␣, as was the late phase of activation at 6 h (Fig.  8D). Overexpression of the mutated IB␣ also resulted in inhibition of thrombin-induced NF-B activation (Fig. 8D). In keeping with the latter finding, and with the partial inhibition of SLIGKV-and thrombin-induced COX-2 expression by MG-132 (Fig. 7C), we found that expression of IB␣ protein was only marginally reduced in HUVEC exposed to either thrombin or the PAR-2 peptide SLIGKV (Fig. 9C). Co-infection of HUVEC with AdNF-B-Luc and with either of the IB␣ viruses significantly inhibited 6-keto-PGF 1␣ generation (Fig. 8D). Collectively, these results suggest that PAR-1-and PAR-2-stimulated NF-B activation occurs in an IB␣-dependent manner and that such pathways are required for PAR-induced COX-2-derived PGI 2 synthesis.
Growing evidence suggests that posttranslational modification of p65-NF-B is also required for efficient transactivation of NF-B-dependent genes (29,38,39). We next examined whether phosphorylation of p65-NF-B on serine 536 is triggered by activation of PARs on endothelial cells. Incubation with thrombin or the PAR-2 peptide SLIGKV caused a rapid and sustained increase in p65 phosphorylation without modifying total p65-NF-B expression (Fig. 9A). To determine the relationship between PAR-induced ERK/p38 MAPK activation and p65 phosphorylation, HUVEC were preexposed to the NF-B inhibitor PG490, and the phosphorylation state of ERK1/2, p38 MAPK , and  p65-NF-B was assessed using phosphospecific antibodies. As shown in Fig. 9D, blockade of NF-B activity did not affect thrombin-or SLIGKV-induced activation of either ERK1/2 or p38 MAPK . Similarly, inhibition of MEK with U0126 or of p38 MAPK with SB202190 did not modify PAR-2-induced phosphorylation of NF-B on serine 536 but partially reduced thrombin-stimulated p65-NF-B phosphorylation (Fig. 9E). These results suggest that phosphorylation of p65-NF-B on serine 536 in HUVEC exposed to PAR-2 activators, but not thrombin, occurs independently of MEK-ERK-or p38 MAPK -mediated signaling.

DISCUSSION
In the present study, we have demonstrated that in human endothelial cells, thrombin-and PAR-2 agonist peptide-driven COX-2 expression and sustained PGI 2 synthesis require the activation of ERK1/2 and p38 MAPK signaling pathways as well as NF-B activity. We further show that PAR-mediated NF-B activation is accompanied by p65 phosphorylation and is dependent upon IB␣. This is the first report to demonstrate a functional link between PARs and COX-2-derived endothelial PGI 2 production involving MAPKs and NF-B pathways.
Thrombin is known to exert its cellular effects through activation of PAR-1. However, there is some evidence of transactivation of PAR-2 through cleaved PAR-1 (40). In the present study, we observed that thrombin and PAR-2-selective peptides induced COX-2 expression. Although it is possible that the up-regulatory effects of thrombin on COX-2 and on MAPK activation may be partially mediated through PAR-2 transactivation, the PAR-2-selective peptides employed in the present study do not activate PAR-1, thus indicating that PAR-2-peptide-induced effects occur predominantly through PAR-2 activation. In addition, COX-2 up-regulation, as well as MAPK activation and PGI 2 synthesis, was marked by differences in kinetics and magnitude in HUVEC exposed to thrombin versus PAR-2-selective peptides. Whereas these differences may well be explained by a difference in potency of PAR-1 versus PAR-2 agonists at their receptors, the possibility that these differences could have functional significance is suggested by the fact that the PAR-2-selective peptide, unlike thrombin or the PAR-1-selective peptide, does not promote acute prostanoid synthesis. This may reflect an inability of PAR-2 agonists to activate acute calciumdependent pathways in HUVEC that are required for rapid prostanoid production (24). Although the ability of thrombin to promote prostanoid release in several tissues is well documented, the potential roles of proteases and their receptors in regulating sustained endothelial cell PGI 2 synthesis have received little attention, and few reports document the ability of thrombin to promote prolonged prostanoid production via COX-2 induction in these cells (12,39). However, our results are consistent with those of a recent study showing that activated protein C, as well as thrombin, can induce COX-2 expression in HUVEC through activation of PAR-1-mediated signaling (42).
Endothelial PGI 2 production in response to thrombin or PAR-2 peptides occurred with different time courses and to different extents. As previously reported (12) PGI 2 release in thrombin-stimulated cells was characterized by an initial rapid increase (principally COX-1-derived) and a later prolonged and sustained phase lasting several hours. The extent of inhibition of thrombin-driven PGI 2 release by COX-2 blockade with NS-398 versus that with indomethacin (COX-1/2 inhibitor) confirms the COX-2 dependence of chronic PAR-induced prostanoid synthesis as well as the contribution from COX-1 to overall synthesis. In marked contrast, our present studies with the PAR-2-selective peptide have shown that SLIGKV does not trigger an acute release of PGI 2 synthesis but that release was only detectable 2-4 h after peptide exposure and was sustained thereafter. These differential effects of PAR-1 versus PAR-2 activation were concomitant with increased COX-2 protein levels at the same time points and were not accompanied by changes in COX-1 expression.
Our present data report an up-regulation of COX-2 and COX-2derived PGI 2 synthesis in HUVEC in response to PAR-1 and PAR-2 activation. The observation that PARs are capable of stimulating COX-2-driven PGI 2 release emphasizes the pathophysiological significance of PAR-1 and PAR-2 activation by circulating proteases and suggests that they contribute to the early inflammatory response by transiently increasing COX-2 expression and PGI 2 synthesis. Although the endogenous ligand(s) responsible for driving COX-2 induction in HUVEC through PAR-2 activation remain to be defined, our preliminary data indicate that ligands previously identified as PAR-2- selective (e.g. tryptase; trypsin) are capable of enhancing COX-2 expression in HUVEC.
COX-2 induction in the inflammatory setting has largely been considered deleterious, but relatively less is known about its role under normal physiological conditions. Our current data, along with our previous studies (12,24), have shown that PGI 2 synthesis occurs in resting unperturbed endothelial cells. Our data also indicate that, under our experimental culture conditions, unstimulated HUVEC cultures exhibit variable COX-2 expression. Whereas this could be due at least in part to serum-stimulated induction, inhibition of basal PGI 2 synthesis by the COX-2-selective inhibitor NS-398 would support the suggestion that a component of basal release is COX-2-derived under the conditions of our experiments. Although there is evidence that human and mouse arteries express COX-2 ex vivo (43), further studies are required to unequivocally demonstrate that COX-2 is present in normal endothelium. PAR-2 expression in endothelium and other cells is known to be strongly induced by proinflammatory stimuli (17), and PAR-2-mediated events can then exacerbate inflammation-based disease (44). Importantly, our results clearly show that prostanoid production in response to PAR-2 activation occurs without the requirement for pretreatment with a proinflammatory stimulus, suggesting that PAR-2 activation and its associated signaling events are physiologically relevant components of transient (and hence resolved) inflammation. The significance of these findings and the potential protective role of the PAR-COX-2-PGI 2 axis in the vasculature are highlighted by the observations that production of circulating PGI 2 is COX-2-dependent (45) and that COX-2derived PGI 2 reduces the detrimental vascular remodeling associated with hemodynamic stress (46). Protective roles for COX-2-derived mediators have also been reported in lung (47) and during ischemic preconditioning, where COX-2-derived PGI 2 synthesis is up-regulated and exerts cardioprotective effects (48).
Ligand binding to G-protein-coupled receptors leads to the activation of MAPK signaling cascades, and PAR-mediated signaling in a number of cell types is known to involve activation of one or more MAPK cascades (1,49). In endothelial cells, we and others have shown that activation of ERK1/2 and p38 MAPK is triggered by a number of cytokines including tumor necrosis factor-␣ and IL-1␣ (50, 51) as well as growth factors such as vascular endothelial growth factor (52). However, there are few reports of MAPK signaling in response to PAR activation in human endothelial cells. In the present work, we investigated the potential role of MAPKs in PAR-induced COX-2 expression. We demonstrate that both ERK1/2 and p38 MAPK pathways are required for the induction of COX-2 in response to thrombin and PAR-2 activation. Pharmacological inhibition of either ERK1/2 or p38 MAPK abrogated PAR-induced COX-2 protein and mRNA expression as well as PGI 2 synthesis. However, PD98059 and SB203580 are known to block COX-1 and COX-2 activities in a nonspecific manner (24,53); hence, we do not rule out the possibility that the substantial decreases in PGI 2 levels observed in the presence of these inhibitors could partly result from inhibition of COX enzyme activities. We have shown, however, that U0126 has no significant effect on COX activity (24). In addition, thrombin-induced COX-2 expression was not modified by preexposure to either indomethacin (COX-1/-2 inhibitor) or the COX-2-selective inhibitor NS-398, suggesting that inhibition of COX-2 expression by PD98059 and SB202190 results from kinase blockade. These results indicate that the ERK1/2 and p38 MAPK families are activated by stimulation of endothelial PAR-1 and PAR-2 receptors and that these pathways are critical regulators of PAR-induced COX-2 expression and PGI 2 synthesis. Whereas it is clear that PAR-1-and PAR-2-induced MAPK activation is functionally linked to prostanoid generation, the PAR-4-selective peptide failed to promote either COX-2 expression or PGI 2 release, despite its ability to induce both ERK1/2 and p38 MAPK phosphorylation. It is possible that PAR-4 contributes to COX-2 induction at higher thrombin concentrations (49) and/or that this receptor does not normally participate in regulating endothelial prostanoid production (4).
NF-B is known to be involved in the transcriptional regulation of several immune and inflammatory markers, including the COX-2 gene (35). In the current work, we investigated the involvement of this transcription factor in PAR-induced COX-2 up-regulation. Thrombin, the PAR-1-selective peptide, and the PAR-2-selective peptide all stimulated the transcriptional activity of NF-B. Our data showing a failure of the PAR-4 peptide to promote NF-B activation in AdNF-B-Luc-infected HUVEC corroborate the lack of effect of PAR-4 activation on COX-2 expression. NF-B activation by PARs was associated with translocation of p65-NF-B to the nucleus and coincided temporally with PARinduced COX-2 expression and sustained PGI 2 formation. The NF-B inhibitor PG490 also attenuated PAR-induced COX-2 expression as well as PGI 2 synthesis, providing further support for the hypothesis that signaling to NF-B plays a critical role in COX-2 induction and endothelial prostacyclin release.
NF-B activation in response to proinflammatory stimuli as well as growth factors involves phosphorylation of the IB␣ protein. Recruitment and activation of the classical IB-kinase (IKK) complex (including IKK␣ and IKK␤) leads to serine phosphorylation of IB and subsequent degradation via the proteasome pathway. An alternative mechanism for NF-B activation involves tyrosine (Tyr 42 ) phosphorylation of IB␣, and this pathway does not require IB␣ degradation (reviewed in Refs. 29 and 38). We used recombinant adenoviruses encoding wild-type IB␣ and mutated IB␣ (Y42F) to begin to probe the potential involvement of IB␣-dependent pathways in PAR-induced NF-B activation and hence in COX-2 expression and PGI 2 synthesis. We found that NF-B activation was inhibited by co-infection with either AdIB␣ or AdIB␣Y42F, suggesting that IB␣-dependent activation of NF-B occurs in endothelial cells exposed to PAR agonists. Moreover, PAR-driven PGI 2 formation was also significantly abrogated in cells overexpressing either IB␣ or IB␣Y42F, providing evidence for a role for IB␣ in regulating PGI 2 synthesis in response to PARs activation. The inhibition of NF-B activation by infection with AdIB␣Y42F raises the possibility that tyrosine phosphorylation of IB␣ participates in regulating PAR-mediated responses in HUVEC. Indeed, recent studies using AdIB␣Y42F have indicated that tyrosine phosphorylation of IB␣ contributes to regulation of vascular endothelial growth factorinduced NF-B activation in HUVEC, 4 a growth factor that we have previously shown induces COX-2 expression and PGI 2 synthesis in human endothelial cells (12). The kinase(s) responsible for triggering tyrosine phosphorylation of IB␣ in various cell types are not defined, but recent evidence suggests that c-Src is a possible candidate mediator of this event in macrophages (54). In this respect, we have previously shown that c-Src activity is required for PAR-stimulated ERK1/2 activation in HUVEC, suggesting that activation of c-Src is a component of PAR signaling (55). Our observations that marginal IB␣ degradation occurs following PAR-1 or PAR-2 activation and that the proteosome inhibitor MG-132 only partially reduces PAR-induced COX-2 expression and PGI 2 formation lend support to the possibility that PAR-mediated NF-B activation and its downstream events in HUVEC are regulated principally through degradation-independent pathways.
Increasing evidence suggests that post-translational modifications of NF-B proteins are required for efficient gene transcription. In particular, phosphorylation of p65-NF-B on Ser 536 regulates its transactivating ability following binding to its consensus sequence (10,56). We determined whether PAR agonists modify phosphorylation of p65-NF-B on Ser 536 . We found that thrombin and SLIGKV, as well as IL-1␣, stimulated a rapid and sustained increase in the phosphorylation state of p65-NF-B. Since the time course of NF-B phosphorylation in response to either thrombin or PAR-2 peptide coincided with the kinetics of ERK1/2 and p38 MAPK phosphorylation, we examined the potential for interaction between these pathways to regulate PAR-induced p65 phosphorylation. Blockade of NF-B activity did not affect PAR-induced ERK1/2 or p38 MAPK activation, confirming that MAPK activation does not require NF-B activity. However, inhibition of either MEK or p38 MAPK partially inhibited thrombin-induced p65 phosphorylation without modifying the response to PAR-2 activation. The identity of the kinase(s) that phosphorylates p65-NF-B in response to PAR-1 and PAR-2 activation is unknown, but our data suggest that p65 phosphorylation on Ser 536 in thrombin-stimulated but not PAR-2 peptide-stimulated endothelial cells is at least partially dependent upon ERK1/2 and p38 MAPK activities. This would be consistent with a role for p65 phosphorylation in regulating thrombin-induced COX-2 expression and PGI 2 release. Our findings do not rule out the possibility that PAR activation may be accompanied by changes in the phosphorylation state of p65 on sites other than Ser 536 or that PAR-stimulated kinases other than the MAPKs are involved in regulating Ser 536 phosphorylation. The mechanisms mediating the early phosphorylation of p65-NF-B and its significance for PAR-mediated COX-2 induction remain to be elucidated. Potential candidates involved in this regulation, including IKK␣/␤ and the Akt pathway, are currently under examination in our laboratory.
In conclusion, this study has demonstrated that ERK1/2 and p38 MAPK as well as the NF-B pathway participate in the PAR-mediated regulation of COX-2 expression and hence regulate sustained PGI 2 synthesis by human endothelial cells exposed to PAR activators. These findings provide new insights into the molecular mechanisms used by proteases and PARs to regulate generation of the cytoprotective molecule PGI 2 through changes in endothelial COX-2 expression.