Targeting Prostaglandin E2 Receptors as an Alternative Strategy to Block Cyclooxygenase-2-dependent Extracellular Matrix-induced Matrix Metalloproteinase-9 Expression by Macrophages*

COX-2-dependent prostaglandin (PG) E2 synthesis regulates macrophage MMP expression, which is thought to destabilize atherosclerotic plaques. However, the administration of selective COX-2 inhibitors paradoxically increases the frequency of adverse cardiovascular events potentially through the loss of anti-inflammatory prostanoids and/or disturbance in the balance of pro- and anti-thrombotic prostanoids. To avoid these collateral effects of COX-2 inhibition, a strategy to identify and block specific prostanoid-receptor interactions may be required. We previously reported that macrophage engagement of vascular extracellular matrix (ECM) triggers proteinase expression through a MAPKerk1/2-dependent increase in COX-2 expression and PGE2 synthesis. Here we demonstrate that elicited macrophages express the PGE2 receptors EP1–4. When plated on ECM, their expression of EP2 and EP4, receptors linked to PGE2-induced activation of adenylyl cyclase, is strongly stimulated. Forskolin and dibutryl cyclic-AMP stimulate macrophage matrix metalloproteinase (MMP)-9 expression in a dose-dependent manner. However, an EP2 agonist (butaprost) has no effect on MMP-9 expression, and macrophages from EP2 null mice exhibited enhanced COX-2 and MMP-9 expression when plated on ECM. In contrast, the EP4 agonist (PGE1-OH) stimulated macrophage MMP-9 expression, which was inhibited by the EP4 antagonist ONO-AE3-208. When compared with COX-2 silencing by small interfering RNA or inhibition by celecoxib, the EP4 antagonist was as effective in inhibiting ECM-induced proteinase expression. In addition, ECM-induced MMP-9 expression was blocked in macrophages in which EP4 was silenced by small interfering RNA. Thus, COX-2-dependent ECM-induced proteinase expression is effectively blocked by selective inhibition of EP4, a member of the PGE2 family of receptors.

Atherosclerosis is a chronic inflammatory disease characterized by lipid accumulation, macrophage recruitment, smooth muscle proliferation, and fibrosis (1,2). Macrophage proteinase expression compromises the structural integrity of atherosclerotic lesions by degrading components of the extracellular matrix (ECM), 2 which contributes to lesion ulceration or rupture and subsequent sequelae of thrombosis, myocardial infarction, and stroke (3)(4)(5)(6). A substantial body of evidence has identified cyclooxygenase (COX)-2 as a targetable component of the signaling pathway responsible for increased proteinase expression by macrophages in atherosclerotic lesions. COX metabolizes arachidonic acid to an unstable endoperoxide, which is then converted to the principal prostaglandins (PG) by specific synthases (7,8). COX-2 expression is elevated in atherosclerotic lesions (9 -14). PGE 2 , an important mediator of the inflammatory response, stimulates proteinase expression by a variety of cells including macrophages (15)(16)(17)(18). Both PGE synthase and matrix metalloproteinase (MMP) activities are elevated in regions of symptomatic plaques rich in macrophages and susceptible to rupture (13,19). Moreover, treatment of low density lipoprotein receptor-deficient mice with a specific COX-2 inhibitor resulted in reduced aortic atherosclerosis (20), and statin-dependent plaque stabilization was associated with decreased COX-2, PGE synthase, and MMP activities (21)(22)(23)(24). Finally, a naturally occurring polymorphism in the COX-2 promoter that is associated with reduced expression appears to protect against myocardial infarction and stroke (25).
Taken together, these data suggest that selective inhibition of COX-2 would lead to reduced PGE 2 and MMP expression, resulting in greater plaque stability. However, recent data indicate that administration of COX-2-selective inhibitors increases adverse cardiovascular events (26,27). The explanation for this effect is not fully understood; however, inhibition of COX-2 results in the loss of all downstream PGs, some of which have anti-inflammatory functions (28). Likewise, it has been suggested that selective COX-2 inhibitors block PGI 2 production by vascular endothelium, without inhibiting COX-1-dependent platelet thromboxane A2 expression, thereby supporting a pro-thrombotic state (29,30). Finally, several cyclooxygenase-independent effects of the COX-2 inhibitors have been described (31), which may contribute to adverse cardiovascular events.
Despite the adverse cardiovascular events associated with selective COX-2 inhibition, the COX-2-PGE 2 -MMP-9 axis remains an attractive target to block macrophage proteinase expression at sites of chronic * These studies were supported by National Institutes of Health Research Grants HL073375 (to D. J. F.), CA089578 (to A. J. D.), and GM15431 (to R. M. B.) and a grant from the Center for Cancer Prevention Research. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1 To whom correspondence should be addressed: Dept inflammation. In this regard, we previously reported that macrophage engagement of vascular smooth muscle cell (SMC)-derived ECM triggers proteinase expression through a protein kinase C-dependent activation of MAPK erk1/2 , which leads to increased COX-2 expression and PGE 2 synthesis (32). Selective inhibition of macrophage COX-2 activity blocked ECM-induced proteinase expression and ECM-induced PGE 2 production, and MMP-9 expression was markedly reduced in COX-2 Ϫ/Ϫ macrophages compared with wild type macrophages (32). In studies reported here, we have determined whether selective inhibition of the PGE 2 receptor family (EP1-4) can attenuate ECM-induced MMP-9 expression by macrophages. We demonstrate that thioglycollate-elicited macrophages express EP1-4. When plated on ECM, their expression of EP2 and EP4, receptors linked to PGE 2 -induced activation of adenylyl cyclase (33), is strongly stimulated. Although forskolin and dibutryl cAMP stimulate macrophage MMP-9 expression in a dose-dependent manner, the selective EP2 agonist (butaprost) has no effect on macrophage MMP-9 expression, and macrophages from EP2 null mice exhibited enhanced COX-2 and MMP-9 expression when plated on ECM. In contrast, a selective EP4 agonist (PGE 1 -OH) stimulated macrophage MMP-9 expression, which was inhibited by the EP4 antagonist ONO-AE3-208. Finally, preincubation of macrophages with the EP4 antagonist or inhibition of their EP4 expression by siRNA was as effective in inhibiting ECM-induced MMP-9 expression as treatment with the selective COX-2 inhibitor celecoxib or inhibition of COX-2 expression by siRNA. Thus, COX-2-dependent ECM-induced proteinase expression is markedly attenuated by selective inhibition of EP4, a member of the PGE 2 family of receptors.

MATERIALS AND METHODS
Isolation of Peritoneal Macrophages-Thioglycollate-elicited peritoneal macrophages were obtained from Swiss Webster, EP2 wild type, and EP2 null mice (34) by the method of Edelson and Cohn (35) as described previously (36). Mice were injected intraperitoneally (3 ml/mouse) with 3% Brewer thioglycollate medium containing 0.3 mM thioglycollate (Difco). Four days later cells were harvested by lavage with cold DPBS. Peritoneal cells were recovered by centrifugation and resuspended in Dulbecco's modification of Eagle's medium (DMEM) supplemented with 10% Cellect Gold fetal bovine serum (FBS), penicillin (100 units/ml), streptomycin (100 g/ml), and 4 mM glutamine (Invitrogen) and plated into tissue culture flasks or multiwell plates. Cells were allowed to adhere for 4 h and then were washed free of nonadherent cells. Experiments to determine the effect of matrix on macrophage proteinase expression were carried out in DMEM supplemented with 0.1% low endotoxin bovine serum albumin.
Murine Macrophage Cell Line-RAW264.7 macrophages (37) were obtained from American Type Culture Collection and maintained as adherent cultures in DMEM-10% FBS.
Preparation of Extracellular Matrix-coated Dishes-SMC-derived matrices were prepared as previously described (38) with the following modifications. Rat aortic smooth muscle cells (SMC; VEC Technologies, Inc.) were plated into 6-, 12-, or 24-well plates in DMEM supplemented with supplied growth medium. 3-4 days after reaching confluence, the cell layer was removed by sequential exposure to 0.1% Triton X-100 in DPBS (2 min at room temperature) and 0.2 mM NH 4 OH in DPBS (2 min at room temperature). The remaining insoluble matrices were washed 3ϫ with DPBS and stored at 4°C.
Determination of Metalloproteinase Activity-The presence of metalloproteinase activity in cellular-conditioned media was determined utilizing enzyme zymography as previously described (39). Conditioned media were mixed with SDS sample buffer (without mercaptoethanol) and incubated for 30 min at 37°C. Samples and molecular weight markers were electrophoresed in a 10% polyacrylamide gel containing 0.1% gelatin. The gel was then washed (2ϫ) in 2.5% Triton X-100 to remove SDS. The gel was incubated at 37°C for 48 h in 200 mM NaCl containing 40 mM Tris-HCl and 10 mM CaCl 2 (pH 7.5) and stained with Coomassie Blue. The presence of gelatinolytic activity was identified as clear bands on a uniform blue background after destaining.
Western Blot Identification of MMP-9-Macrophage-conditioned media were electrophoresed in 4 -15% polyacrylamide gels, and proteins were transferred to a polyvinylidene difluoride membrane. After transfer, the membrane was placed in blocking buffer for 1 h, washed in PBS (1ϫ), and incubated for 2 h in blocking buffer containing rabbit anti-mouse MMP-9 IgG (Chemicon). The membrane was washed (2ϫ) in 25 mM Tris buffer (pH 8) containing 137 mM NaCl, 2.7 mM KCl, and 0.5% Tween (TTBS) and incubated for 1 h in blocking buffer containing 0.3 g/ml goat anti-rabbit IgG conjugated to HRP (Transduction Laboratories). The membrane was washed (3ϫ) in TTBS, and bound HRP was visualized utilizing enhanced chemiluminescence.
Western Blot Identification of Phosphorylated MAPK erk1/2 and p38-Cell lysates were electrophoresed in gradient gels, and proteins were transferred to a polyvinylidene difluoride membrane, which was then blocked. Following 1 wash in TTBS, the membrane was incubated for 1 h in blocking buffer containing 75 ng/ml rabbit anti-phosphospecific p44/p42 MAPK IgG or rabbit anti-phosphospecific p38 MAPK IgG (Cell Signaling Technology). The membranes were washed (2ϫ; TTBS) and incubated for 1 h in blocking buffer containing 0.3 g/ml goat anti-rabbit IgG conjugated to HRP. After visualization of bound HRP, membranes were stripped in 62.5 mM Tris buffer (pH 6.7) containing 100 mM ␤-mercaptoethanol and 2% SDS for 30 min at 50°C, washed, and probed for total MAPK erk1/2 or p38 (Cell Signaling Technology).
Western Blot Identification of COX-2-Cell lysates were electrophoresed in gradient gels, and proteins were transferred to a polyvinylidene difluoride membrane. After transfer, the membrane was blocked, washed in PBS (1ϫ), and incubated for 1 h in blocking buffer containing 0.5 g/ml rabbit IgG raised against a peptide containing amino acids 584 -598 of murine COX-2 (Cayman Chemical). The membrane was washed (2ϫ; TTBS) and incubated for 1 h in blocking buffer containing 0.3 g/ml goat anti-rabbit IgG conjugated to HRP.
COX-2 siRNA Transfection-RAW264.7 macrophages were transfected with COX-2 siRNA (Santa Cruz) or mock-transfected according to the manufacturer's protocol, which was modified as follows. Macrophages (T-75 flask) were washed 3ϫ with DPBS to remove serum, and medium was replaced with antibiotic-free DMEM. The cells were mechanically harvested, recovered by centrifugation, and resuspended in antibiotic-free DMEM supplemented with 10% FBS. The cells (2 ϫ 10 6 /well) were separated into aliquots in a 6-well plate and incubated overnight. The next morning, 9.1 l of 10 M COX-2 siRNA was added to 152 l of transfection medium (Santa Cruz) and incubated for 5 min at room temperature. In another tube 9.1 l of the transfection reagent (Santa Cruz) was added to the transfection medium and incubated for 5 min. The two tubes were combined and incubated for 20 min to form the siRNA transfection reagent complex. Immediately before transfection, macrophage medium was removed and replaced with 1.5 ml of antibiotic-free DMEM supplemented with 10% FBS. The plate was placed on a rocker plate in the laminar flow hood, and the transfection reagent complex was added dropwise. Cells were incubated for 30 h at 37°C.
EP4 siRNA Transfection-RAW264.7 macrophages were transfected with EP4 or control siRNA (Dharmacon RNA Technologies) according to the manufacturer's protocol, which was modified as follows. Macrophages (0.5 ϫ 10 6 /well) were separated into aliquots into a 6-well plate and incubated overnight in serum and antibiotic-free DMEM. The next morning 200 l of 1 M siRNA was added to 200 l of diluted (1:50) DharmaFECT 4 transfection reagent, incubated for 20 min, and added to 1.6 ml of DMEM containing antibiotics and 10% FBS. Immediately before transfection, macrophage medium was removed and replaced with 2 ml of the transfection medium. Cells were incubated 24 h at 37°C.

ECM Stimulates Macrophage Expression of EP2 and EP4 Prostanoid
Receptors-The diverse physiological and pathophysiological effects of PGE 2 are mediated by engagement of the EP family of G-protein-coupled prostanoid receptors (40,41). The EP family is comprised of four subtypes (EP1-4), which exhibit varying affinities for PGE 2 and trigger distinct signaling pathways. Differential expression of PGE 2 receptors has been reported for macrophages stimulated with LPS (42,43). However, the influence of ECM on macrophage PGE 2 receptor expression is unknown. Therefore, we cultured thioglycollate-elicited peritoneal macrophages on an ECM deposited in situ by rat vascular smooth muscle cells and determined its effect on EP1-4 receptor mRNA utilizing RT-PCR (Fig. 1). Elicited macrophages cultured on tissue culture plastic expressed all PGE 2 receptors (EP1-4). When plated on vascular ECM, macrophage EP2 and EP4 expression were markedly induced, whereas EP1 and EP3 expression was unchanged. Similar results were observed utilizing the murine macrophage cell line RAW264.7 (data not shown).
Macrophage engagement of ECM triggers protein kinase C-dependent activation of MAPK erk1/2 , which leads to the stimulation of COX-2 expression and MMP-9 synthesis (32), and activation with LPS triggers the p38-dependent increase in COX-2 expression and MMP-9 synthesis (44). Thus, signaling through the MAPK cascade regulates COX-2 and MMP-9 expression by macrophages. Therefore, we determined whether the observed ECM-dependent increase in EP2 and EP4 expression ( Fig. 1) was also dependent on the MAPK cascade. As observed in Fig. 2A, levels of phosphorylated MAPK erk1/2 and p38 were increased when cells were plated on SMC-ECM ( Fig. 2A). Preincubation of cells with an inhibitor of protein kinase C (calphostin C) or MEK-1 (U0126) blocked ECM-induced activation of MAPK erk1/2 . In contrast, ECM-induced levels of phosphorylated p38 were slightly increased in cells preincubated with calphostin C and decreased in cells incubated with Thioglycollate-elicited macrophages were suspended in DMEM supplemented with 0.1% low endotoxin bovine serum albumin (LE-BSA) and separated into aliquots in 12-well control or ECM-coated plates (7.5 ϫ 10 5 cells/well). After overnight incubation, total RNA was isolated, and mRNA levels for PGE 2 receptors (EP1-4) were determined by RT-PCR as described under "Materials and Methods." The data are representative of three separate experiments. FEBRUARY 10, 2006 • VOLUME 281 • NUMBER 6 U0126. Equal protein loading was confirmed by probing for total MAPK erk1/2 and p38.

EP4-dependent Proteinase Expression by Macrophages
The causal relationship between MAPK erk1/2 activation and ECMinduced MMP-9 expression is shown in Fig. 2B. As previously reported, the expression of MMP-9 by macrophages cultured on SMC-ECM was markedly elevated as compared with that in cells cultured on plastic (32). Virtually all the MMP-9 secreted by murine macrophages was in the pro form (105 kDa), which is slightly larger than that expressed by human cells (92 kDa). High molecular weight multimers of MMP-9 are also visible in some preparations. Preincubation of macrophages with U0126 completely blocked ECM-induced MMP-9 expression; in contrast, proteinase expression was partially attenuated by exposure to a p38 inhibitor (SB202190; Calbiochem). We next determined the role of MAPK erk1/2 and p38 on ECM-induced EP2 and EP4 expression. RNA isolated from elicited macrophages plated on ECM in the presence of a MEK-1 or p38 inhibitor was probed for expression of EP2 and EP4 utilizing RT-PCR. ECM-induced expression of the PGE 2 receptors was blocked by preincubation with the MEK-1 inhibitor, whereas inhibition of p38 had no effect (Fig. 2C). Taken together with earlier studies (32), these data demonstrate that ECM-induced activation of MAPK erk1/2 after macrophage engagement on ECM is responsible for increased COX-2 and EP2 and EP4 receptor expression.
ECM-induced MMP-9 Expression Is EP4-dependent-EP2 and EP4 are coupled to G␣s and, when activated by PGE 2 , lead to an increase in intracellular cyclic AMP (33). In an effort to determine whether either of these receptors played a role in ECM-induced MMP-9 expression, macrophages were incubated with increasing concentrations of an activator of adenylyl cyclase (forskolin) or dibutryl cAMP. As seen in Fig. 3, incubation of macrophages with either forskolin or dibutryl cAMP led to a dose-dependent increase in MMP-9 expression, determined by both Western blot and enzyme zymography of conditioned media.
EP2-and EP4-dependent actions of PGE 2 can be distinguished utilizing relatively specific receptor agonists or antagonists and genetic approaches. To determine whether ECM-induced MMP-9 is dependent on EP2 signaling, macrophages were incubated with butaprost (free acid; Cayman), a selective EP2 agonist (45). The free acid form of butaprost has a Ͼ10-fold higher affinity to EP2 than the ester form. As seen in Fig. 4A, butaprost (0.01-10 M) had no effect on macrophage MMP-9 expression. AH 6809 (Cayman) is a receptor antagonist for EP2 (as well as EP1, EP3, and DP2) (46). Preincubation of macrophages with AH 6809 had little or no effect on ECM-induced MMP-9 and COX-2 expression (Fig. 4B). To directly test the role of EP2, we next compared the effect of ECM on macrophages isolated from wild type and EP2 Ϫ/Ϫ mice (34, 47) (Fig. 4C). When plated on ECM, COX-2 and MMP-9 expression were stimulated in both wild type and EP2 null macrophages. Based on these data, the prostanoid receptor EP2 does not appear to play a central role in ECM-induced MMP-9 expression. PGE 1 -OH is a selective EP4 agonist (48). Incubation of macrophages with 0.05-10 M PGE 1 -OH resulted in a dose-dependent increase in MMP-9 expression, which was inhibited by the selective EP4 antagonist ONO-AE3-208 (Fig. 5). We next determined whether EP4 receptor blockade would inhibit ECM-induced proteinase expression and com-FIGURE 2. The causal role of MAPK erk1/2 in ECM-induced EP2 and EP4 expression. A, macrophages (2 ϫ 10 6 cells/well) were preincubated for 30 min with 10 M MEK-1 inhibitor U0126 or 50 nM protein kinase C inhibitor (Calphostin C (Calp)) before plating on ECM. Lysates were prepared 30 min later, and levels of phosphorylated (p) MAPK erk1/2 and p38 were assessed by Western blot as described under "Materials and Methods." B, macrophages were preincubated with U0126 (10 M) or an inhibitor of p38 kinase (10 M; SB202190) and separated into aliquots in 6-well control plates or plates coated with SMC-ECM (2.0 ϫ 10 6 cells/well). Conditioned media were collected the next day and assayed for MMP-9 expression by Western blot. C, macrophages were preincubated with U0126 or an inhibitor of p38 kinase and separated into aliquots in 6-well control plates or plates coated with SMC-ECM (2.0 ϫ 10 6 cells/well). After overnight incubation, total RNA was isolated, and mRNA levels for PGE 2 receptors (EP1-4) were determined by RT-PCR as described under "Materials and Methods." Data presented in panels A and B are representative of two experiments, and panel C is representative of three experiments.

EP4-dependent Proteinase Expression by Macrophages
pared the relative effectiveness of the EP4 antagonist to COX-2 silencing and inhibition of COX-2 activity with celecoxib (Fig. 6). COX-2 expression by RAW264.7 macrophages was blocked by transfecting cells with COX-2 siRNA. The ability of COX-2 siRNA to block COX-2 expression was first verified in LPS (10 ng/ml)-and TNF␣ (100 ng/ml)treated RAW264.7 macrophages. As observed in Fig. 6 (left panel), COX-2 antigen in mock-transfected macrophages was markedly increased after incubation with LPS or TNF␣. In contrast, neither LPS nor TNF␣ stimulated COX-2 expression by macrophages previously transfected with COX-2 siRNA. When conditioned media were examined for MMP-9 antigen, LPS and TNF␣ treatment led to increased proteinase expression, which was not observed in macrophages transfected with COX-2 siRNA. Next we determined the effect of COX-2 silencing on ECM-induced MMP-9 expression. As reported earlier, macrophage expression of COX-2 and MMP-9 is up-regulated when plated on ECM. Inhibition of COX-2 gene expression by siRNA blocked ECM-induced MMP-9 expression. Thus, cytokine-and ECM-induced proteinase expression is causally linked to enhanced COX-2 expression and effectively blocked by COX-2 silencing.
We next compared the abilities of the selective COX-2 inhibitor celecoxib and the EP4 antagonist to inhibit matrix-induced proteinase expression (Fig. 6, middle and right panels). ECM-induced MMP-9 expression was markedly inhibited when cells were preincubated with celecoxib (5 M). Likewise, ECM-induced proteinase expression was nearly blocked by preincubation with the EP4 antagonist (10 M).
As an alternative strategy to test the role of EP4 in ECM-induced MMP-9 expression, we utilized EP4 siRNA to knock down macrophage EP4 expression. Levels of EP4 mRNA were markedly increased when either control or nonspecific siRNA transfected RAW264.7 macrophages were plated on ECM (Fig. 7A). In contrast, EP4 expression was not detected in macrophages transfected with EP4 siRNA. As expected, the level of MMP-9 mRNA was increased when either control or nonspecific siRNA transfected cells were plated on ECM. When EP4 knockdown macrophages were plated on ECM, the level of MMP-9 mRNA was reduced to that observed in controls.
COX-2 expression is reportedly stimulated by PGE 2 via a positive feedback loop (49,50). Our observation that celecoxib and the EP4 antagonist partially inhibited ECM-induced COX-2 expression (Figs. 6) is consistent with a feedback loop and identifies EP4 as the receptor responsible for PGE 2 effect. To test this hypothesis, EP4 "knockdown" macrophages were plated on ECM, and COX-2 expression was determined utilizing RT-PCR (Fig. 7A). The expression of COX-2 was markedly increased when either control or nonspecific siRNA-transfected RAW264.7 macrophages were plated on ECM, whereas ECM-induced COX-2 expression was partially inhibited in EP4 knockdown macrophages.
We next determined the effects of EP4 knocked down on ECM-induced MMP-9 and COX-2 expression utilizing zymography and Western blotting (Fig. 7B). Levels of MMP-9 activity and antigen were markedly increased when either control or nonspecific siRNA transfected macrophages were plated on ECM. Likewise, levels of COX-2 antigen were increased when either control or nonspecific siRNA-transfected macrophages were plated on ECM. When EP4 knockdown macrophages were plated on ECM, levels MMP-9 and COX-2 expression were reduced to that observed in controls. Taken together, these data demonstrate that targeting a single prostanoid receptor achieves the same level of proteinase inhibition as gene silencing or pharmacological inhibition of COX-2.
Finally, the ability of an EP4 agonist to restore COX-2 and MMP-9 expression by macrophages pretreated with selective COX-2 inhibitors was determined. As seen in Fig. 8A, COX-2 and MMP-9 expression were stimulated by exposure to LPS. Co-incubation of macrophages with the selective COX-2 inhibitors NS398 and celecoxib blocked LPSinduced COX-2 and MMP-9 expression, which was restored by the addition of the EP4 ligand, PGE 1 -OH.
To corroborate these results, we knocked down EP4 expression with siRNA and subsequently treated macrophages with LPS (Fig. 8B). Neither COX-2 nor MMP-9 antigen levels were up-regulated by LPS in cells transfected with EP4 siRNA. In contrast, COX-2 and MMP-9 expression was strongly up-regulated in cells transfected with nonspecific siRNA. As predicted, exogenous PGE 1 -OH was unable to restore FIGURE 4. ECM-induced MMP-9 expression by macrophages is EP2-independent. A, RAW264.7 macrophages were suspended in DMEM containing 10% FCS and separated into aliquots in 24-well plates (3 ϫ 10 5 cells/well). After 4 h of adherence, cells were washed 3ϫ with DPBS, and media were replaced with DMEM, 0.1% LE-BSA containing the indicated concentrations of butaprost (free acid). Cells were incubated overnight, conditioned media were collected, and levels of MMP-9 were determined by Western blot. B, cells were suspended in DMEM, 0.1% LE-BSA, preincubated with the indicated concentrations of AH 6809 (30 min) and separated into aliquots into a control 24-well plate or an SMC-ECM-coated plate. Cells were incubated overnight, conditioned media were collected, and lysates were prepared. Levels of MMP-9 in conditioned media and COX-2 in lysates were determined by Western blot. C, thioglycollate-elicited macrophages isolated from EP2 ϩ/ϩ or EP2 Ϫ/Ϫ mice were suspended in DMEM, 0.1% LE-BSA and separated into aliquots in 12-well control or ECM-coated plates (4.0 ϫ 10 5 cells/well). After overnight incubation, conditioned media were collected, and cell lysates were prepared. Levels of MMP-9 and COX-2 were determined by Western blot as described under "Materials and Methods." Dose response data presented in panels A and B are representative of three experiments. Data in panel C are derived from duplicate wells and are representative of two experiments. WT, wild type; GAPDH, glyceraldehyde-3-phosphate dehydrogenase.  FEBRUARY 10, 2006 • VOLUME 281 • NUMBER 6 COX-2 and MMP-9 expression by EP4 knock-down cells. Thus, PGE 2dependent COX-2 expression appears to be mediated by the EP4 receptor.

EP4-dependent Proteinase Expression by Macrophages
In summary, ECM-induced proteinase expression by macrophages is COX-2-dependent and blocked by a selective inhibitor of COX-2 or an EP4 receptor antagonist. Thus, targeting the EP4 prostanoid receptor appears to be an alternative strategy to block COX-2-dependent MMP-9 expression by macrophages.

DISCUSSION
Pathologic, pharmacologic, and genetic evidence indicate that COX-2-dependent PGE 2 synthesis regulates macrophage MMP expression, which is thought to destabilize atherosclerotic plaques and contribute to the formation of aneurysms (9 -14, 20, 21, 25, 51). PGE 2 and the other prostanoid family members (PGF2␣, PGD 2 , PGI 2 , and thromboxane A 2 ) are synthesized from the common endoperoxide intermediate PGH 2 , which is generated by the action of COX-1 and COX-2 on arachidonic acid (7,8). COX-1 expression is constitutive, whereas COX-2 expression is induced in response to a variety of inflammatory mediators (52). Consequently, inhibition of COX-2 would appear to be a rational therapeutic target for the stabilization of vascular lesions. Paradoxically, the administration of selective COX-2 inhibitors leads to increased frequency of adverse cardiovascular events (26,27). It has been suggested that the loss of anti-inflammatory prostanoids and/or disturbance in the balance of pro-and anti-thrombotic prostanoids is responsible for the adverse cardiovascular effects of COX-2 inhibitors (29,30). To avoid these collateral effects of COX-2 inhibition, a strategy to block a specific prostanoid receptor(s) would be necessary. In studies reported here, we have determined whether selective inhibition of the PGE 2 receptor family (EP1-4) can block ECM-induced proteinase expression by macrophages. We demonstrate that macrophage engagement of ECM selectively stimulates expression of EP2 and EP4 prostanoid receptors, and signaling through EP4 is responsible for increased MMP-9 expression. Moreover, inhibiting EP4 expression by siRNA or blocking PGE 2 binding to EP4 inhibits ECM-induced proteinase expression as effectively as celecoxib inhibition of COX-2 activity or inhibition of COX-2 expression by siRNA.
The pleiotropic effects of prostanoids are mediated by a family of G-protein-coupled receptors (53). Each prostanoid is recognized by its own receptor or receptor subtypes. Four subtypes of the EP receptor for PGE 2 have been described (EP1-4) and cloned (54). The regulation of EP receptor expression as well as cellular functions triggered by their engagement is an area of intense investigation (55)(56)(57). In this regard, several studies have demonstrated differential expression of macrophage EP receptors in response to tissue injury and inflammatory stim- After adherence, the indicated wells received LPS (100 ng/ml) or TNF␣ (10 ng/ml). Cells were incubated overnight, conditioned media was collected, and lysates were prepared. Levels of MMP-9 in conditioned media and COX-2 in lysates were determined by Western blot. Data for each condition are derived from a single well. The entire experiment was repeated twice with similar results. Ctrl, control. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. Middle and right panels, cells were suspended in DMEM, 0.1% LE-BSA preincubated with celecoxib (5 M) or EP4 antagonist (10 M) and separated into aliquots in a control 24-well plate (plastic (Plst)) or SMC-ECM-coated plate. Cells were incubated overnight, conditioned media were collected, and lysates were prepared. Levels of MMP-9 in conditioned media and COX-2 in lysates were determined by Western blot. The data are representative of two experiments. FIGURE 7. EP4 silencing blocks ECM-induced MMP-9 expression. A, RAW264.7 macrophages transfected with EP4 siRNA or nonspecific (NS) siRNA were separated into aliquots in plastic (Plst) or SMC-ECM-coated 6-well plates (1 ϫ 10 6 /well) in DMEM, 0.1% LE-BSA, and the cells were incubated overnight. The next day total RNA was isolated, and mRNA levels for EP4, MMP-9, and COX-2 were determined by RT-PCR as described under "Materials and Methods." B, EP4 siRNA or nonspecific siRNA-transfected macrophages were separated into aliquots in plastic or SMC-ECM-coated 12-well plates (5 ϫ 10 5 /well) in DMEM, 0.1% LE-BSA, and the cells were incubated overnight. The next day media were recovered, and lysates were prepared. Levels of MMP-9 expression were determined by both zymography and Western blot. Levels of COX-2 and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were determined by Western blot as described under "Materials and Methods."

EP4-dependent Proteinase Expression by Macrophages
uli (42, 58 -61). At sites of inflammation, monocytes bind to endothelium and transmigrate and engage the ECM, where they differentiate into macrophages. Engagement of the ECM initiates signaling pathways that culminate in alterations of monocyte/macrophage gene expression (62)(63)(64). In studies reported here, elicited peritoneal macrophages were shown to express mRNA for all four PGE 2 receptor subtypes. EP2 and EP4 mRNA levels were markedly increased after adhesion to ECM. Similar to what was observed for ECM-induced COX-2 and MMP-9 expression (32), enhanced EP expression was blocked by inhibition of MAPK erk1/2 activation. Thus, macrophage engagement of ECM triggers activation of the MAPK cascade, which plays a causal role in the up-regulation of COX-2 expression, PGE 2 synthesis, and expression of EP2 and EP4 prostanoid receptors.
Identification of the specific PGE 2 receptor responsible for induction of proteinase expression would in principle allow for more directed therapeutic targeting than complete inhibition of COX activity. PGE 2 receptors are coupled to distinct signaling pathways including activation of phospholipase C (EP1), activation of adenylyl cyclase (EP2 and EP4), and inhibition of adenylyl cyclase (EP3) (33). In studies reported here, we demonstrate that stimulation of adenylyl cyclase by forskolin or the addition of dibutryl cAMP stimulated a dose-dependent increase in macrophage MMP-9 expression. These data confirm earlier reports that demonstrated the activation of adenylyl cyclase was associated with increased MMP expression (16,65) and suggest that PGE 2 -induced proteinase expression is mediated by PGE 2 binding to EP2 and/or EP4. Despite their shared ability to activate adenylyl cyclase, it has recently become clear that EP2 and EP4 modulate cellular functions through divergent signaling pathways, which can be distinguished utilizing pharmacologic and genetic approaches (33). PGE 2 stimulation of EP2 modulates transcription primarily through protein kinase A, whereas EP4 can lead to phosphatidylinositol 3-kinase-mediated activation of MAPK erk1/2 followed by induction of early growth response factor-1 (66). We have concluded that PGE 2 binding to EP2 does not play a role in ECM-induced proteinase expression since the EP2 agonist butaprost did not stimulate macrophage MMP-9 expression, and the EP2 antagonist AH 6809 did not affect ECM-induced MMP-9 expression. Moreover, macrophages derived from EP2 null mice exhibited increased COX-2 and MMP-9 expression when plated on ECM. In contrast, engagement of EP4 appears to responsible for ECM-induced proteinase expression, since a selective EP4 agonist PGE 1 -OH stimulated MMP-9 expression, and the EP4 antagonist ONO-AE3-208 inhibited ECM-induced MMP-9 as effectively as celecoxib or COX-2 silencing. When the role of EP4 was directly examined utilizing EP4 siRNA, ECM failed to induce MMP-9 expression by EP4 knockdown macrophages. Taken together, these data have important implications for targeting macrophagedependent proteinase expression and tissue destruction in chronic inflammatory diseases. For example, EP2 and EP4 are up-regulated in peritoneal macrophages isolated from a murine arthritis/lupus model (67). EP4 null mice exhibited reduced bone destruction, and loss of proteoglycans and type II collagen in a collagen antibody induced model of rheumatoid arthritis (57). More recently, it was reported that EP4 is the main PGE 2 receptor expressed by macrophages in symptomatic atherosclerotic plaques, and PGE 2 -induced MMP expression by monocytes was blocked by the EP4 antagonist L-161 982 (68).
It has been reported that PGE 2 stimulates COX-2 expression via activation of the Ras-MAPK signaling cascade (50). A similar feedback loop appears to play a role in ECM-induced COX-2 expression by macrophages. ECM-induced COX-2 expression is MAPK erk1/2 -dependent (32), and inhibition of PGE 2 synthesis by celecoxib, inhibition of PGE 2 binding to EP4, or knockdown of EP4 expression resulted in attenuated COX-2 expression. Consequently, the strategy of blocking EP4 receptors, as a mechanism to attenuate macrophage proteinase expression, is complicated by the inhibition of a positive feedback loop that stimulates COX-2 expression. Nonetheless, it is important to note that although COX-2 expression is attenuated by EP4 knockdown or receptor antagonism, neither intervention blocks COX-2 activity. In contrast, celecoxib blocks COX-2 activity, PGE 2 synthesis, and COX-2 expression. Thus, selective EP receptor targeting remains a cogent strategy to regulate PGE 2 -dependent functions in macrophages.