Cyclooxygenase-2 Is Induced in Monocytes by Peroxisome Proliferator Activated Receptor γ and Oxidized Alkyl Phospholipids from Oxidized Low Density Lipoprotein*

Low density lipoprotein (LDL) oxidation and monocyte infiltration of the vessel wall underlie atherogenesis. These cells express cyclooxygenase-2, but the way oxidized LDL stimulates cyclooxygenase-2 transcription is unknown. Oxidized LDL, oxidatively fragmented phospholipids isolated from oxidized LDL, a synthetic oxidized alkylphospholipid (azPC) that is a potent peroxisome proliferator activated receptor (PPAR) γ agonist, or the PPARγ agonist rosiglitazone all induced cyclooxygenase-2 expression and enhanced prostaglandin E2 (PGE2) secretion in primary human monocytes. The cyclooxygenase-2 inhibitor NS398 blocked PPARγ-induced PGE2 secretion. Phospholipase A1 and A2 digestion shows that oxidized alkylphospholipids, and not oxidized fatty acids, were the relevant agonists. The upstream PPAR-responsive element (PPRE) of cyclooxygenase-2 was required for induction of a luciferase reporter by oxidized phospholipids, azPC, and rosiglitazone, and a (COX-2 PPRE)3-luciferase reporter was responsive to these PPARγ agonists. Circulating human monocytes do not contain PPARγ, but PPARγ was induced rapidly (<4 h) in monocytes upon ligation of surface ICAM-3, but not P-selectin glycoprotein-1 even though both interactions prime cytokine secretion. Cyclooxygenase-2 induction by oxidized phospholipids only occurred in monocytes containing PPARγ. Thus PPARγ was induced rapidly in primary monocytes by appropriate outside-in signaling, sensitizing them to previously undetectable agonists in oxidized LDL. Cyclooxygenase-2 and PGE2secretion are induced, not inhibited, by selective PPARγ agonists that include oxidatively fragmented phospholipids in oxidized LDL.

idation of LDL particles to proatherogenic ones not only generates modified particles that lead to inappropriate accumulation of intracellular lipid deposits, it generates a series of lipid inflammatory mediators ranging from fragmented phospholipids that potently activate inflammatory cells through the platelet-activating factor receptor (2,3) to lysophosphatidic acid that regulates cell function through a family of Edg receptors (4). Recently we identified a new class of agonists in oxidized LDL, arising from the oxidation of a minor class of phospholipids, that are highly potent and selective agonists for the nuclear hormone receptor and lipidactivated transcription factor PPAR␥ (5). One of these proved to be as potent and selective as the pharmacologic agent rosiglitazone, currently in widespread clinical use.
Cyclooxygenases-1 and -2 initiate the conversion of arachidonate to all of the prostaglandins and thromboxanes through dual cyclooxygenase and peroxidase activities. Cyclooxygenase-2 is normally not present in the endothelium of major vessels, but it is present in endothelium, smooth muscle, and particularly the infiltrating monocytes/macrophages of atherosclerotic lesions (6,7). This abnormal cyclooxygenase-2 expression contributes to the excessive production of circulating prostaglandin I 2 in patients with these lesions (8). Cyclooxygenase products are vasoactive and have a role in pathologic vascular remodeling and plaque rupture (9). Selective inhibition of cyclooxygenase-2 decreases atherogenesis in apoE-deficient mice (10), and it is the macrophage expression of this enzyme that proves to be particularly detrimental (11). Similarly, cyclooxygenase-2 is expressed abnormally in colons by tumor-infiltrating macrophages (12)(13)(14). Cyclooxygenase inhibition decreases neointimal formation in hyperlipidemic animals (15), angiogenesis (16), and it suppresses tumorigenesis (17).
Expression of cyclooxygenase-2 is controlled at the transcriptional level by proximal 5Ј-elements (e.g. nuclear factor-IL6, E-box, and cAMP response elements (18 -20)), and it is also subject to post-transcriptional control (21)(22)(23)(24). In addition, we identified a distal peroxisomal proliferator response element (PPRE) that confers sensitivity to non-steroidal anti-inflammatory drugs and fatty acids (25). There are three known isoforms of this nuclear hormone receptor and ligand-activated transcription factor, and non-steroidal anti-inflammatory drugs and fatty acids and certain of their oxidized products are low affinity agonists for two of these, PPAR␥ and PPAR␣ (26,27).
Oxidized LDL is present in atherosclerotic lesions (28), and oxidized LDL induces cyclooxygenase-2 expression in a murine macrophage-like cell line (29). Oxidized LDL contains PPAR␥ agonists (30), one of which we identified recently as hexadecyl azelaoyl phosphatidylcholine (azPC) (5). This is a prominent oxidation product of the minor pool of phospholipase A 1 -resistant alkylphosphatidylcholines in LDL (31). Because PPAR␥ (32,33) and cyclooxygenase-2 (6,7) are both found in atherosclerotic lesions, we questioned whether there was a link between them. We asked whether oxidized LDL would induce cyclooxygenase-2 expression in human monocytes, and what events would be required for any such induction.
Here we show that oxidized LDL induces cyclooxygenase-2 expression in primary human monocytes, and that this induction results from ligand activation of PPAR␥ by oxidized phosphatidylcholines and the distal PPRE in the cyclooxygenase-2 5Ј-regulatory region. This mechanism is possible because PPAR␥ itself can be induced rapidly in human monocytes in response to signals (34,35) initiated by engagement of the surface receptor and adhesion molecule ICAM-3. Cyclooxygenase-2 expression is said (36 -38) to be inhibited by PPAR␥ agonists, but this is not true when PPAR␥-selective agonists are examined directly. We establish a molecular link from the oxidation of LDL to the expression of cyclooxygenase-2 through PPAR␥ and phospholipid oxidation products.
Oxidation and Analysis of LDL and Synthetic Phospholipids-LDL was oxidized overnight at 37°C with 20 M CuSO 4 . The resulting lipid products were purified by reverse phase HPLC (39), and the fractions eluting from the column were collected at 1-min intervals, dried, and reconstituted for assay as before (2,5). Material causing cyclooxygenase-2 induction generally eluted in fractions 5 and 6, and either fraction alone or combined was considered as the relevant oxidized phospholipid of oxidized LDL. azPC was synthesized as described (5) from hexadecyl lysophosphatidylcholine after mild alkaline hydrolysis (0.5 N NaOH in methanol, 4 h at 24°C) to remove contaminating biologically active platelet-activating factor-like phospholipids (40). After neutralization 2 mg of purified lipid was reacted with 10 mg of azelaic anhydride (University of Utah Chemical Synthesis facility) in the presence of 1 mg of 4-dimethylaminopyridine in CHCl 3 :pyridine (4:1) for 36 h before purification by reverse phase HPLC.
Cell Preparation-Primary human monocytes were isolated in an unactivated state by countercurrent elutriation from freshly drawn blood (41). The purified cells were resuspended (1 ϫ 10 6 /ml) in HBSS containing 0.5% HSA and 10 g/ml polymyxin B. Monocytes were added to plates coated with 10 g/ml CAL3.10 anti-ICAM-3 monoclonal antibody (34) (unless otherwise stated) and allowed to adhere for 1 h. They were washed and then stimulated with the stated agonists for 7 h before collecting the medium or analyzing cellular material by Western blotting. The murine macrophage cell line RAW264.7 was obtained and grown as suggested by ATCC.
Plasmids-The 7.2-kb (Ϫ7273 to ϩ1) and 3.9 (Ϫ3966 to ϩ1) cyclooxygenase-2-luciferase reporter plasmids were described previously (25). The 3.5-kb cyclooxygenase-2 reporter construct was created by deletion of the Ϫ6900 to Ϫ3180 sequence from the 7.2-kb construct (which removes the Ϫ3721 to Ϫ3707 PPRE) by BglII digestion. The Ϫ966 luciferase reporter was generated by SacI digestion of the 7.2-kb reporter construct. The acyl-CoA oxidase-luciferase plasmid (25) and the SV40-␤-galactosidase reporter were as described previously (5). Plas-mids were transformed into TOP10FЈ E. coli strain using the TA cloning kit. Plasmids from log phase cells were isolated using a Bigger Prep kit (5 Prime 3 3 Prime; Boulder, CO) and purified in CsCl gradients. The cyclooxygenase-2 PPRE-luciferase reporter (COX-2 PPRE) 3 -luc was created by ligating a synthetic oligonucleotide between the SacI and BglII sites of pGL3basic. The ligated oligonucleotide contained a 5Ј-SacI sequence, a synthetic COX-2 PPRE, a BamHI spacer, a second synthetic COX-2 PPRE, a KpnI spacer, and a third synthetic COX-2 PPRE followed by a BglII spacer. The synthetic COX-2 PPRE was AGGCGA-CAGGTCA based on the human Ϫ3721 PPRE. All plasmids were sequence verified.
Transfection of Cultured Cells, Reporter Assays, and mRNA Estimation-The murine macrophage-like cell line RAW264.7 (ATCC) was transfected using 1 g of appropriate reporter plasmid, 0.1 g of SV40-␤-galactosidase plasmid, and 10 l of LipofectAMINE/ml Opti-MEM. Cells were washed once with PBS before the transfection solution was added for a period of 4 h. This was removed, and the cells were incubated in normal medium (e.g. Dulbecco's modified Eagle's medium and 10% fetal calf serum for RAW264.7 cells) containing any stated agonist for 18 -20 h before collection and assay. When PPAR expression plasmids were cotransfected, 0.25 g of the relevant plasmid was combined with 0.55 g of pGL3-PPRE and 0.1 g of SV40-␤-galactosidase plasmid. Agonist was then added to each well and incubated for 18 -24 h. Cyclooxygenase-2 primers (forward, 5Ј-CTG GTG CCT GGT CTG ATG ATG-3Ј; reverse, 5Ј-GTC CTT TCA AGG AGA ATG GTG C) were synthesized by the DNA synthesis facility (University of Utah). Cellular RNA was extracted with TRIzol (Invitrogen), and total RNA was reverse transcribed with Moloney murine leukemia virus transcriptase (Promega) and amplified with Taq polymerase using 35 cycles of (90°C denature, 55°C anneal, 72°C elongate, each for 1 min).
Protein Analysis-Western blotting for cyclooxygenase-2 and PPAR␥ was accomplished by washing adherent monocytes twice with PBS and lysing them in a buffer composed of 20 mM Tris-HCl (pH 7.5), 1 mM EDTA (pH 8), 16 mM CHAPS, 0.5 mM dithiothreitol, 1 mM benzamidine HCl, 1 g/ml leupeptin, 10 g/ml soybean trypsin inhibitor, and 1ϫ Laemmli solution. The cells were scraped in this buffer, the lysate was incubated at 95°C, and the proteins were resolved by SDS-PAGE in a 9% gel. The proteins were transferred to an Immobilon-P membrane, and the membranes were blocked with 2.5% non-fat dry milk and 0.5% human albumin in TBST (Tris-buffered saline Tween 20) buffer and then probed with E8 anti-PPAR␥ or mouse monoclonal anti-cyclooxygenase-2. Antibody was detected with horseradish peroxidase-conjugated goat polyclonal anti-mouse IgG antibody and ECL reagent using Biomax (Eastman Kodak) film. The blots were then stripped and reprobed with anti-actin monoclonal antibody to assess protein loading.
Immunohistochemistry-Permanox chamber slides (LabTek Brand Products, Naperville, IL) were coated with 10 g/ml anti-ICAM-3 antibody or P-selectin overnight at 4°C, washed, and blocked with 1.5% HSA in HBSS for 2 h at 37°C. The slides were washed three times with PBS, and monocytes (10 6 /chamber) were added and allowed to interact with the surface for 1 h before agonists or HSA/HBSS was added. This incubation was continued for 7 h at 37°C when nonadherent cells were removed, the chambers washed three times with PBS, and the cells fixed by the addition of Ϫ20°C methanol. The slides were air dried and frozen. Thawed slides were washed three times with PBS, blocked for 1 h with 1.5% goat serum in PBS, and then incubated with H-100 rabbit polyclonal anti-human PPAR␥ antibody (Santa Cruz) (5 g/ml in 1.5% goat serum in PBS) for 30 min at 24°C. The slides were washed for 5 min/wash using three changes of PBS before the primary antibody was detected with a 30-min incubation with horseradish peroxidaseconjugated goat anti-rabbit antibody (BIOSOURCE International, Camarillo, CA) (1 g/ml in 1.5% goat serum in PBS). The slides were again washed for 5 min/wash using three changes of PBS before peroxidase was detected with AEC (Vector Laboratories, Burlingame, CA) peroxidase substrate. After a 30-min incubation, the cells were washed for 5 min/wash using three changes of deionized water. Images were collected with an Olympus AX70 photomicroscope and stored electronically.
Prostaglandin Analysis and Suppression-The amount of PGE 2 and 6-keto-PGF 1␣ released from adherent monocytes into 0.5% HSA/HBSS medium was determined by ELISA. Monocytes were pipetted into wells coated with protein, which unless otherwise stated was anti-ICAM-3, and allowed to interact with the substrate for 1 h before the addition of any agonist. This incubation was continued for an additional 7 h. 20 M arachidonate was added 20 min prior to the end of the incubation, and the medium was then collected for analysis according to the manufacturer's directions. Prostaglandin standards for the ELISA were diluted into HSA/HBSS, the monocyte incubation buffer. For some experi-ments, the cyclooxygenase-2 inhibitor NS398 (10 M) was incubated with adherent monocytes for 30 min prior to the addition of agonists.

Oxidized Alkylphospholipids in Oxidized LDL Induce Cyclooxygenase-2 in Human
Monocytes-Oxidation of LDL generates a host of lipid oxidation products, some of which are biologically active. We probed for the presence of cyclooxygenase-2 protein in monocytes exposed to these products by Western analysis to find that this enzyme was not normally present in freshly isolated human monocytes. Adhesion to and spreading on a surface coated with an anti-ICAM-3 antibody (42) (a surrogate ␤ 2 -integrin ligand that induces select activation responses in human leukocytes (5,34,35)) alone did not result in cyclooxygenase-2 expression (Fig. 1a). However, these adherent monocytes expressed significant amounts of cyclooxygenase-2 protein after being exposed to oxidized, and not native unoxidized, LDL. Cyclooxygenase-2 induction by oxidized LDL, or the total lipids extracted from oxidized LDL, was equivalent to that of the positive control LPS. Lipids extracted from unoxidized LDL did not do this. Cyclooxygenase-2 protein expression resulted from the appearance of new message because unactivated monocytes contained no detectable cyclooxygenase-2 mRNA, but this message was readily detectable after stimulation with oxidized LDL and its oxidatively generated polar phospholipid fraction (Fig. 1b). Inclusion of actinomycin D completely blocked the accumulation of cyclooxygenase-2 message (not shown), establishing the importance of transcriptional modulation.
Oxidized LDL contains the potent and selective PPAR␥ agonist azPC, a prominent oxidatively fragmented alkylphosphatidylcholine that activates the PPAR␥-driven gene CD36 in human monocytes (5). Cyclooxygenase-2 contains a distal PPAR␣-responsive element (25), therefore it might be induced in monocytes through PPAR␥ activation. We treated primary human monocytes with rosiglitazone as a positive control for PPAR␥ activation and with synthetic azPC to find (Fig. 1a) that both dramatically enhanced cyclooxygenase-2 protein expression. The expression level in response to these two PPAR␥ agonists was similar to that of the LPS positive control, leading to the conclusion that PPAR␥ agonists are strong, independent stimuli for cyclooxygenase production.
PPAR␥ Agonists Induce Cyclooxygenase-2 Expression in Primary Human Monocytes-We tested whether the cyclooxygenase-inducing activity(ies) in oxidized LDL had the characteristics of the PPAR␥ ligands we had found previously in oxidized LDL in that these agonists all belonged to the small subclass of phospholipids with an ether bond at the sn-1 position that makes them resistant to phospholipase A 1 digestion. The structure(s) in oxidized LDL responsible for cyclooxygenase-2 induction was a phospholipid because this activity was recovered in the polar phospholipid fraction eluting from a reversed phase HPLC column and because induction was abolished by treatment with phospholipase A 2 (Fig. 2a). This maneuver would be expected to increase the amounts of free fatty acid oxidation products such as HETEs and HODEs (43), which weakly interact with PPAR␣ and PPAR␥ (27,43,44), so none of these hydrolysis products is likely be the active principle. The phospholipids responsible for cyclooxygenase-2 induction were de- Freshly isolated primary human monocytes were immobilized on a surface coated with an antibody to their surface ICAM-3 for a period of 1 h at 37°C. After this, the cells were incubated for a further 7 h with either no additional agonist (Neg), 5 g of E. coli LPS, 1 M rosiglitazone (Rosi) or azPC, 100 g of Cu ϩ -oxidized or native LDL, or the polar phospholipids isolated from these particles. At the end of this incubation the cells were collected, lysed, and their proteins were separated by SDS-PAGE and transferred to an Immobilon membrane for Western blotting of cyclooxygenase-2 or ␤-actin as a loading control as described under "Experimental Procedures." Panel b, reverse transcription-PCR analysis of cyclooxygenase-2 messenger RNA. Human monocytes were treated as panel a, and total mRNA was recovered with TRIzol. This was reverse transcribed and amplified with cyclooxygenase-2 or ␤-actin primers as stated under "Experimental Procedures." These data are representative of one other experiment.
FIG. 2. The PPAR␥ agonists rosiglitazone and the synthetic oxidized phospholipid azPC induce cyclooxygenase-2 expression in primary human monocytes. Panel a, cyclooxygenase-2 (Cox-2) protein expression assessed by Western blotting. Freshly isolated human monocytes were allowed to adhere and spread onto a surface coated with anti-ICAM-3 antibody for 1 h and then treated with buffer or the stated agonist for 7 h before the cells were collected for cyclooxygenase-2 analysis by Western blotting as described under "Experimental Procedures." Some cells (0 h) were lysed just as other cells were aliquoted to the coated wells. The lipids used were the total lipid extract from native or Cu ϩ -oxidized LDL, the polar phospholipid (PL) fraction isolated by reverse phase HPLC from these extracts (5), or the phospholipid fractions after pretreatment with either phospholipases (PLA) A 1 or A 2 , also as described under "Experimental Procedures." The purified ovine cyclooxygenase-2 standard is in the leftmost lane of the upper gel; the lower gel is the ␤-actin loading control. Panel b, induction of a full-length 7.2-kb human cyclooxygenase-2 reporter by oxidized lipids and PPAR␥ agonists. The monocytic cell line RAW264.7 was transfected with a full-length cyclooxygenase-2-luciferase reporter and a SV40-␤-galactosidase for normalization as described under "Experimental Procedures." These cells were untreated or treated with 5 g E. coli LPS, 1 M rosiglitazone (Rosi), or the synthetic oxidized phospholipid azPC, with 100 g/ml oxidized or unoxidized LDL, or with the lipid extracts of these particles. Some samples of the lipid extract were pretreated with bee venom phospholipase A 2 to remove sn-2 acyl residues from the phospholipids in the sample or R. arrhizus phospholipase A 1 , which attacks sn-1 ester bonds. These data are representative of one other experiment. rived from the oxidation of the minor subclass, mainly phosphatidylcholines, of phospholipids that contain an ether rather than an ester bond at the sn-1 position of the glycerol backbone. This follows because treating the extracted phospholipids with phospholipase A 1 , which cleaves the ϳ99.7% of the ester-containing phospholipids in LDL (31), was without effect on cyclooxygenase-2 induction (Fig. 2a).
We confirmed that oxidized phospholipids and PPAR␥ agonists stimulated cyclooxygenase-2 transcription by transfecting the murine monocytic cell line RAW264.7 with a luciferase reporter under the control of the 7.2-kb full-length human cyclooxygenase-2 5Ј-regulatory region. We found (Fig. 2b) that both the synthetic oxidized phosphatidylcholine azPC and rosiglitazone stimulated reporter expression, as did intact oxidized LDL particles. Unoxidized LDL particles did not stimulate luciferase expression, and so the relevant activity(ies) was the result of the oxidative attack. The lipids extracted from oxidized LDL induced equivalent luciferase expression from this reporter, whereas those lipids extracted from unoxidized LDL did not (not shown). The active principle(s) in oxidized LDL was a phospholipid because again it was sensitive to digestion with phospholipase A 2 , and it was an ether phospholipid because it was resistant to hydrolysis by phospholipase A 1 .
Polar Lipids in Oxidized LDL Induce PGE 2 Synthesis-We found that the PPAR␥ agonists azPC and rosiglitazone enhanced PGE 2 secretion from immobilized human monocytes and that this enhancement was equivalent to that resulting from exposure to the strong inflammatory agonist LPS (Fig. 3). Intact oxidized LDL or the unfractionated lipids extracted from oxidized LDL were also potent agonists for PGE 2 secretion. We also found (not shown) that 6-keto-PGF 1␣ release was similarly enhanced by rosiglitazone, azPC, and the polar phospholipids found in oxidized LDL. We tested whether the cyclooxygenase-2 isoform was primarily responsible for the secretion of PGE 2 by including the cylcooxygenase-2-selective antagonist NS398 during the stimulation by PPAR␥ agonists. The cylcooxygenase-2 inhibitor NS398 was completely effective in suppressing this PGE 2 synthesis, so the newly synthesized cyclooxygenase-2 was functional and accounted for PPAR␥stimulated PGE 2 secretion.
PPAR␥ Is Induced Rapidly after Monocytes Engage ICAM-3-Our experiments show that monocytes respond to PPAR␥ agonists, but circulating monocytes do not express PPAR␥ (45). Because our monocytes were immobilized with anti-ICAM-3, which initiates select responses in these cells (34), we determined whether PPAR␥ expression was one of these responses.
We probed for the presence of PPAR␥ in freshly isolated monocytes by Western blotting and, as expected, found these cells (not shown) or monocytes maintained in a suspended state (Fig. 4a) expressed no detectable PPAR␥ protein. In contrast, monocytes with their surface ICAM-3 engaged by immobilized anti-ICAM-3 antibody expressed significant amounts of PPAR␥ protein. This induction was enhanced modestly in adherent monocytes by the inclusion of either PMA or LPS. The accumulation of PPAR␥ protein was an early response because adherent monocytes in similar experiments expressed PPAR␥ after just 4 h of this interaction (not shown). Monocytes interacting with other coated surfaces, including P-selectin, which primes monocytes for a number of responses (46), did not accumulate PPAR␥.
We supported this biochemical approach with an immunohistochemical one to find (Fig. 4b) that a large subset of monocytes interacting with the anti-ICAM-3-coated surface expressed PPAR␥, whereas no cells interacting with P-selectin did so. The expression of PPAR␥ did not require the addition of Human monocytes were isolated, allowed to adhere to anti-ICAM-3 as in Fig. 1, and then treated with 5 g E. coli LPS, 1 M azPC or rosiglitazone (Rosi), oxidized LDL particles, or the polar phospholipids isolated from them after purification by reverse phase HPLC. Some monocytes were exposed to 10 M NS398 for 30 min prior to the addition of the stated agonist. The amount of PGE 2 released into the 0.5% HSA and HBSS medium in 8 h was then quantitated by ELISA. These data are representative of four other experiments.

FIG. 4. Primary monocytes express PPAR␥ after engaging anti-ICAM-3. Panel a, Western blot of PPAR␥.
Freshly isolated monocytes were layered over tissue culture wells coated with the designated proteins, and after 8 h the cells were washed, and their proteins were collected in Laemmli buffer and then resolved by SDS-PAGE. PPAR␥ was detected by Western blotting with E8 monoclonal antibody and ECL as described under "Experimental Procedures." One sample (0 h) was derived from monocytes that were lysed immediately prior to the experiment; the positive controls for PPAR␥ expression were obtained from anti-ICAM-3 adherent monocytes treated with either PMA or LPS. These results are consistent with three related experiments. Panel b, immunohistochemical detection of PPAR␥. Primary human monocytes were incubated on chamber slides coated with either P-selectin or anti-ICAM-3 antibody for a total of 8 h. The cells were treated with HSA/HBSS vehicle or 2 M PMA or 5 g/ml LPS for the last 7 h of the incubation before the level of PPAR␥ was detected as a light pink AEC peroxidase reaction product by the procedure described under "Experimental Procedures." These data are representative of one other experiment. a soluble agonist, but the presence of LPS or PMA enhanced the intensity of the staining for PPAR␥ by those cells with their ICAM-3 engaged. It is notable that LPS or PMA alone was insufficient to induce PPAR␥ expression in cells engaging Pselectin, emphasizing the specificity of the outside-in signaling process.
Generation of Cyclooxygenase-2-dependent Eicosanoids in Response to PPAR␥ Agonists Requires Engagement of ICAM-3-ICAM-3 outside-in signaling rapidly controlled monocyte PPAR␥ expression, so the induction of cyclooxygenase-2 and PGE 2 by rosiglitazone or oxidized phospholipids should similarly be dependent on this interaction. Our data (Fig. 5) show that PGE 2 was secreted in response to rosiglitazone, synthetic azPC, or HPLC-purified polar phospholipids derived from oxidized LDL only under conditions where the monocytes expressed PPAR␥; that is, oxidized phospholipids, azPC, and rosiglitazone all enhanced PGE 2 secretion from monocytes interacting with anti-ICAM-3 but not those bound to P-selectin or albumin.
The Distal PPAR-responsive Element Region Is Needed for Cyclooxygenase-2 Induction by Oxidized Phospholipids-The human 5Ј-regulatory region for cyclooxygenase-2 contains a distal PPRE (25) that weakly interacts with PPAR␣ and PPAR␦, but its use by PPAR␥ has not been explicitly defined. We tested the role of this distal region in the response to oxidized phospholipids by transiently transfecting cells with two reporter constructs (COX2 Ϫ7000 and COX2 Ϫ3900 ), which contain the PPRE at Ϫ3721 and two that do not (COX2 Ϫ3500 and COX2 Ϫ966 ). The latter two cyclooxygenase reporter constructs still retain the proximal regulatory elements required for induction by PMA and cytokines (18 -20). The two reporter constructs that retained the PPRE were induced by rosiglitazone, azPC, and HPLC-purified material from oxidized LDL (Fig. 6a), whereas the two constructs lacking this region were unresponsive to these PPAR␥ agonists. The two shorter constructs that lacked the PPRE were still fully induced by LPS (and tumor necrosis factor-␣, not shown), so it is the presence of the distal Ϫ3500 to Ϫ3900 region that conferred responsiveness to the oxidized phospholipids. We also found by electrophoretic mobility shift assay that the Ϫ3721 PPRE (25) forms a complex with PPAR␥ which was competitive with the acyl-CoA oxidase PPRE and not an SP1 oligonucleotide (not shown).
We confirmed that the cyclooxygenase-2 PPRE was functional and responsive to oxidized phospholipids by trimerizing it and placing it upstream of luciferase in the pGL3 plasmid. This (COX-2 PPRE) 3 -luc reporter was induced by azPC and rosiglitazone with equivalent concentration-response relationships such that either PPAR␥ agonist was a potent stimulus for luciferase expression through this element (Fig. 6b).
High Affinity PPAR␥ Ligands Stimulate Rather than Inhibit Cyclooxygenase-2 Induction-PPAR␥ agonists have been described as inhibitors of cyclooxygenase-2 expression in combination with a primary agonist such as LPS (37) or PMA (36), but PPAR␥ agonists alone were not tested in these reports as we have done here. To compare our data with published information, we combined PPAR␥ agonists with PMA to find that 15-deoxy-PGJ 2 , rosiglitazone, or azPC modestly enhanced luciferase expression from the (COX-2 PPRE) 3 -luc reporter in response to PMA (Fig. 7a). These PPAR␥ agonists clearly were not inhibitory. In contrast, when we investigated the effect of these agents on the induction of the full-length cyclooxygenase-2 reporter, we found that rosiglitazone and azPC, which were individually stimulatory, still modestly enhanced the effect of PMA but that 15-deoxy-PGJ 2 had became quite inhibitory (Fig. 7b). This pattern was recapitulated when endogenous cyclooxgenase-2 protein expression was probed by Western blotting: PMA, rosiglitazone, and 15-deoxy-PGJ 2 individually induced expression of this enzyme in human monocytes (Fig.  7c). The combination of PMA and rosiglitazone enhanced protein expression, whereas the combination of PMA and 15-deoxy-PGJ 2 proved to be markedly inhibitory. Thus 15-deoxy-PGJ 2 inhibits PMA induction of cyclooxygenase-2, but does so through elements other than the PPRE.

DISCUSSION
Reactive oxygen species and oxidized LDL are postulated to initiate and maintain an inflammatory state in the vascular wall during atherogenesis (47)(48)(49). Oxidized lipoprotein particles can be isolated from atherosclerotic plaques (50), and they are found in the circulation. These particles contain fragmented and oxidatively modified phospholipids (51)(52)(53). Some of these fragmented alkylphosphatidylcholines are selective and potent agonists of the platelet-activating factor receptor (2, FIG. 5. Monocytes secrete PGE 2 in response to rosiglitazone (Rosi), azPC, or oxidized phospholipids isolated from oxidized LDL only when PPAR␥ is present. Primary human monocytes were incubated in tissue culture wells coated with the stated proteins or were maintained in a suspended state for the duration of the experiment. These cells were then treated with the designated agonist for 8 h before the amount of PGE 2 released into the 0.5% HSA/HBSS medium was analyzed by ELISA as above.
FIG. 6. Human cyclooxygenase-2 is regulated through its PPRE. Panel a, deletion analysis of human cyclooxygenase-2 regulatory region controlling luciferase expression. RAW264.7 cells were transiently transfected with luciferase reporter constructs containing various lengths of the cyclooxygenase 5Ј-regulatory region and cotransfected with a SV40 ␤-galactosidase reporter to normalize transfection efficiencies. These cells were stimulated for 16 h with 1 M rosiglitazone (Rosi) or azPC, oxidized LDL, or the HPLC-purified polar phospholipids derived from in vitro oxidized LDL, or 5 g of E. coli LPS. The ratio of luciferase to ␤-galactosidase was quantified as described above, and these data are representative of another such experiment. Panel b, a reporter controlled by a trimer of the cyclooxygenase-2 PPRE responds to PPAR␥ agonists. The (COX-2 PPRE) 3 -luc reporter and a SV40 ␤-galactosidase reporter were transfected into RAW264.7 cells and then exposed to the designated concentration of rosiglitazone or azPC as before. These data are representative of those obtained in two other such experiments.
3), whereas others, we find (5), are high affinity ligands and agonists of PPAR␥. An abundant oxidized alkylphospholipid in oxidized LDL is azPC, and this phospholipid, and not its acyl homolog, is a high affinity and selective ligand and agonist for PPAR␥ (5).
Oxidized LDL contains PPAR␥ agonists (30), and those that induce cyclooxygenase-2 expression were oxidized alkylacylphospholipids. This follows because the bioactive lipids were not present prior to the oxidation of the LDL particles and because treatment of these newly generated lipids with phospholipase A 2 , and not phospholipase A 1 , resulted in the complete loss of this activity. The resistance to phospholipase A 1 treatment marks the active phospholipids as having an sn-1 ether bond that is resistant to this esterase, whereas sensitivity to phospholipase A 2 shows the material to be a phospholipid where the sn-2 residue is critical for the ability to stimulate cylcooxygenase-2 expression. LDL contains alkylphosphatidylcholines, amounting to 0.05% or less of the total phosphatidylcholine pool (31), and so this minor pool of phosphatidylcholines is the exclusive precursor for both oxidized phospholipids with platelet-activating factor-like activity (2) and for those that induce cyclooxygenase-2 expression through PPAR␥.
These findings also mean that phospholipids with oxidatively modified sn-2 residues, and not oxidatively modified fatty acids or lysophosphatidylcholine, were responsible for cyclooxygenase-2 induction by oxidized LDL. Although HETEs, HODEs, and isoprostanes are weak PPAR␥ and PPAR␣ ligands (44) and are present in oxidized LDL (54,55) (although in an inactive esterified state), these simple lipids do not participate effectively in PPAR␥ activation or cyclooxygenase-2 induction by oxidized LDL. Were any of these lipids the relevant agonists, then the phospholipase A 2 treatment would have greatly enhanced, and not abolished, the amount of PPAR␥ agonists present in oxidized LDL. A similar argument shows that the induction of cyclooxygenase-2 by high concentrations of lysophosphatidylcholine (56) was not relevant to our findings.
Select cells in atherosclerotic lesions (32,57), and also in Alzheimer's lesions (58) and colon cancers (59), express PPAR␥. Primarily, these cells are infiltrating macrophages. Little is known about the precise control of PPAR␥ transcription in these cells, and alternative exon usage suggests that two 5Јregulatory regions control its transcription (60). PPAR␥ is not normally present in circulating monocytes, but instead accumulates during the differentiation of monocyte to macrophages (32,45). It is also induced in monocytic cells by oxidized LDL (32,57), cytokines (61), PMA (32), and LPS (62). Of these ways to stimulate PPAR␥ expression, engagement of surface adhesion proteins (see "Results") and LPS challenge (62) induce PPAR␥ accumulation rapidly.
The nature of the outside-in signal(s) leading to the rapid induction of PPAR␥ remains undefined, but it is a selective response to the signals generated by ICAM-3 ligation (34,35). ICAM-3 is the counterreceptor for ␣ d ␤ 3 integrin (63), ␣ L ␤ 2 (LFA-1) (64), and the widely distributed (65) DC-SIGN. Surprisingly, even though engagement of monocyte PSGL-1 by P-selectin primes monocytes for cytokine synthesis (46) and stimulates expression from a subset of genes (66), it failed to stimulate PPAR␥ accumulation. One key result of this differential outside-in signaling is that these monocytes are enabled to respond to agonists that were previously undetectable. Thus we found that the newly formed polar phospholipids in oxidized LDL particles significantly enhanced cyclooxygenase-2 expression and PGE 2 secretion and that this response correlated precisely with the presence of PPAR␥.
Cyclooxygenase-2 is induced in response to cytokines or PMA through proximal 5Ј-nuclear factor-IL6, E-box, and cAMP response elements (18 -20). And indeed, we found that the first 966 nucleotides of the 5Ј-regulatory region contained all of the elements needed for responsiveness to LPS (and tumor necrosis factor-␣, not shown). The regulatory region of the human gene also contains a PPRE at Ϫ3721 to Ϫ3707 (25), and it is through this element that the gene is induced by fatty acids and nonsteroidal anti-inflammatory drugs (25,38). PPAR␣ agonists promote cyclooxygenase-2 expression (25,67,68), but the PPAR␣ agonist WY14643 alone does not result in increased PGE 2 synthesis (67,68). However, the conclusion that PPAR␣ is the relevant PPAR isotype for cyclooxygenase-2 induction is flawed because these agonists become nonselective at high concentrations, allowing WY14643 to become a PPAR␥ agonist (69 and not shown).
The literature is confused on the effect of PPAR agonists on cyclooxygenase-2 expression with results suggesting that PPAR␥ agonists suppress cyclooxygenase-2 expression after PMA, LPS, or tumor necrosis factor-␣ stimulation (36 -38); that interleukin-1-induced cyclooxygenase-2 expression is reduced in response to PPAR␣ and not PPAR␥ agonists (70); or that cyclooxygenase-2 is induced by PPAR␣ (68,71) and not PPAR␥ agonists (71). This great disparity in outcomes may have multiple causes including nonspecific effects leading to differential gene expression by currently available PPAR agonists (72) and the use of 15-deoxy-PGJ 2 as an agonist. This arachidonate metabolite is highly reactive, and commercial 15-deoxy-PGJ 2 does not contain any of this unstable material (73). This mate- FIG. 7. Stimulated cyclooxygenase-2 expression is inhibited by 15-deoxy-PGJ 2 independent of its PPRE. The trimerized cyclooxygenase-2 PPRE (panel a) or the full-length cyclooxygenase-2-luc reporter (panel b) and the SV40 ␤-galactosidase reporter were transfected into RAW264.7 cells, and the cells were then treated with 1 M PMA or not (Ctl) for 30 min. At that time, 1 M rosiglitazone (Rosi), 1 M azPC, or 5 M 15-deoxy-PGJ 2 was added to initiate a 16-h incubation before the cells were harvested, and the levels of luciferase and ␤-galactosidase were determined as above. In one instance in panel b, the order of addition was reversed, and rosiglitazone or 15-deoxy-PGJ 2 was added prior to PMA. Panel c, 15-deoxy-PGJ 2 , but not the PPAR␥ agonist rosiglitazone, inhibits cyclooxygenase-2 protein expression in human monocytes. Adherent human monocytes were treated with the stated agonists (1 M) alone or in combination before the amount of cyclooxygenase-2 protein was probed by Western blotting. rial is not a selective PPAR␥ agonist (27, 74 -76) and is toxic at micromolar concentrations (77). We find that 15-deoxy-PGJ 2 at higher concentrations inhibited PMA-induced cyclooxygenase-2 expression as reported (36), but we conclude that this is not the result of PPAR␥ activation because other more potent and selective PPAR␥ agonists remain stimulatory throughout a wide range of concentrations. Nonspecific events acting on elements other than the PPRE of cyclooxygenase-2 are responsible for this effect because the COX-2 PPRE, in contrast to the full-length reporter, was not affected by 15-deoxy-PGJ 2 .
Oxidatively fragmented phospholipids (52), PPAR␥ (32,33), and cyclooxygenase-2 (6,7,78) are all present in vascular lesions. Other vascular cells, particularly endothelial cells and smooth muscle cells (79), constitutively express PPAR␥ and therefore should respond rapidly to its ligands and agonists in oxidized LDL without the need for a priming stimulus to induce PPAR␥ expression. Indeed along with infiltrating monocytes, vascular smooth muscle and endothelial cells express cyclooxygenase-2 in spontaneous and transplant atheroscleotic lesions (7). PPAR␥ ligands therefore have the potential to control cyclooxygenase-2 expression and prostanoid secretion by vascular cells in response to diverse oxidative insults.