Peroxisome Proliferator-activated Receptor γ Ligands Suppress the Transcriptional Activation of Cyclooxygenase-2

We investigated whether peroxisome proliferator-activated receptor γ (PPARγ) ligands (ciglitazone, troglitazone, and 15-deoxy-Δ12,14prostaglandin J2) inhibited cyclooxygenase-2 (COX-2) induction in human epithelial cells. Ligands of PPARγ inhibited phorbol ester (phorbol 12-myristate 13-acetate, PMA)-mediated induction of COX-2 and prostaglandin E2 synthesis. Nuclear run-offs revealed increased rates of COX-2 transcription after treatment with PMA, an effect that was inhibited by PPARγ ligands. PMA-mediated induction of COX-2 promoter activity was inhibited by PPARγ ligands; this suppressive effect was prevented by overexpressing a dominant negative form of PPARγ or a PPAR response element decoy oligonucleotide. The stimulatory effects of PMA were mediated by a cyclic AMP response element in the COX-2promoter. Treatment with PMA increased activator protein-1 (AP-1) activity and the binding of c-Jun, c-Fos, and ATF-2 to the cyclic AMP response element, effects that were blocked by PPARγ ligands. These findings raised questions about the mechanism underlying the anti-AP-1 effect of PPARγ ligands. The induction of c-Jun by PMA was blocked by PPARγ ligands. Overexpression of either c-Jun or CREB-binding protein/p300 partially relieved the suppressive effect of PPARγ ligands. When CREB-binding protein and c-Jun were overexpressed together, the ability of PPARγ ligands to suppress PMA-mediated induction of COX-2 promoter activity was essentially abrogated. Bisphenol A diglycidyl ether, a compound that binds to PPARγ but lacks the ability to activate transcription, also inhibited PMA-mediated induction of AP-1 activity and COX-2. Taken together, these findings are likely to be important for understanding the anti-inflammatory and anti-cancer properties of PPARγ ligands.

COX-2 is an important therapeutic target for preventing or treating arthritis and cancer (10 -12). Selective COX-2 inhibitors decrease inflammation and are widely used to treat arthritis (13). COX-2 is overexpressed in transformed cells (8,14,15) and in malignant tumors (16 -20). COX-2 knockout mice are protected against both intestinal (21) and skin tumors (22). Moreover, selective COX-2 inhibitors suppress the formation and growth of tumors in experimental animals (23)(24)(25)(26)(27) and decrease the number of colorectal polyps in patients with familial adenomatous polyposis (28). Because targeted inhibition of COX-2 is a promising approach to treating inflammation and preventing cancer, it is important to elucidate the signaling mechanisms that regulate COX-2 expression.
In the current study, we show that PPAR␥ ligands inhibited AP-1-mediated transcriptional activation of COX-2 in human epithelial cells. The anti-AP-1 activity of PPAR␥ ligands was a consequence of inhibition of c-Jun expression and competition for limiting amounts of the general coactivator CREB-binding protein (CBP). These results may help to explain the ability of PPAR␥ ligands to suppress carcinogenesis and arthritis.
Western Blotting-Cell lysates were prepared by treating cells with lysis buffer (150 mM NaCl, 100 mM Tris (pH 8.0), 1% Tween 20, 50 mM diethyldithiocarbamate, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, 10 g/ml trypsin inhibitor, and 10 g/ml leupeptin). Lysates were sonicated for 20 s on ice and centrifuged at 10,000 ϫ g for 10 min to sediment the particulate material. The protein concentration of the supernatant was measured by the method of Lowry et al. (50). SDS-polyacrylamide gel electrophoresis was performed under reducing conditions on 10% polyacrylamide gels as described by Laemmli (51). The resolved proteins were transferred onto nitrocellulose sheets as detailed by Towbin et al. (52). The nitrocellulose membrane was then incubated with primary antisera. Secondary antibody to IgG conjugated to horseradish peroxidase was used. The blots were probed with Renaissance Western blot detection system according to the manufacturer's instructions.
Northern Blotting-Total cellular RNA was isolated from cell monolayers using an RNA isolation kit from Qiagen Inc. 10 g of total cellular RNA per lane were electrophoresed in a formaldehyde-containing 1.2% agarose gel and transferred to nylon-supported membranes. After baking, membranes were prehybridized overnight in a solution containing 50% formamide, 5ϫ sodium chloride/sodium phosphate/ EDTA buffer (SSPE), 5ϫ Denhardt's solution, 0.1% SDS, and 100 g/ml single-stranded salmon sperm DNA and then hybridized for 12 h at 42°C with radiolabeled cDNA probes for human COX-2 and 18 S rRNA. COX-2 and 18 S rRNA probes were labeled with [ 32 P]CTP by random priming. After hybridization, membranes were washed twice for 20 min at room temperature in 2ϫ SSPE, 0.1% SDS, twice for 20 min in the same solution at 55°C, and twice for 20 min in 0.1 ϫ SSPE, 0.1% SDS at 55°C. Washed membranes were then subjected to autoradiography.
Nuclear Run-off Assay-2.5 ϫ 10 5 cells were plated in four T150 dishes for each condition. Cells were grown in growth medium until ϳ60% confluent. Nuclei were isolated and stored in liquid nitrogen. For the transcription assay, nuclei (1.0 ϫ 10 7 ) were thawed and incubated in reaction buffer (10 mM Tris (pH 8), 5 mM MgCl 2 , and 0.3 M KCl) containing 100 Ci of uridine 5Ј-[ 32 P]triphosphate and 1 mM unlabeled nucleotides. After 30 min, labeled nascent RNA transcripts were isolated. The human COX-2 and 18 S rRNA cDNAs were immobilized onto nitrocellulose and prehybridized overnight in hybridization buffer. Hybridization was carried out at 42°C for 24 h using equal cpm/ml of labeled nascent RNA transcripts for each treatment group. The membranes were washed twice with 2ϫ SSC buffer for 1 h at 55°C and then treated with 10 mg/ml RNase A in 2ϫ SSC at 37°C for 30 min, dried, and autoradiographed.
Transient Transfection Assays-184B5/HER cells were seeded at a density of 5 ϫ 10 4 cells/well in 6-well dishes and grown to 50 -60% confluence. For each well, 2 g of plasmid DNA were introduced into cells using 8 g of LipofectAMINE as per the manufacturer's instructions. After 7 h of incubation, the medium was replaced with basal medium. The activities of luciferase and ␤-galactosidase were measured in cellular extract as described previously (55).
Electrophoretic Mobility Shift Assay-Cells were harvested, and nuclear extracts were prepared. For binding studies, an oligonucleotide containing the CRE of the COX-2 promoter was used. The complementary oligonucleotides were annealed in 20 mM Tris (pH 7.6), 50 mM NaCl, 10 mM MgCl 2 , and 1 mM dithiothreitol. The annealed oligonucleotide was phosphorylated at the 5Ј-end with [␥-32 P]ATP and T4 polynucleotide kinase. The binding reaction was performed by incubating 5 g of nuclear protein in 20 mM HEPES (pH 7.9), 10% glycerol, 300 g of bovine serum albumin, and 1 g of poly(dI⅐dC) in a final volume of 10 l for 10 min at 25°C. The labeled oligonucleotide was added to the reaction mixture and allowed to incubate for an additional 20 min at 25°C. The samples were electrophoresed on a 4% nondenaturing polyacrylamide gel. The gel was then dried and subjected to autoradiography at Ϫ80°C.
Statistics-Comparisons between groups were made with the Student's t test. A difference between groups of p Ͻ 0.05 was considered significant.

PPAR␥ Ligands Inhibit the Induction of COX-2 in Human
Epithelial Cells-We determined the expression of PPAR␥ in human breast and oral epithelial cells. Western blotting analysis revealed that PPAR␥ was expressed in 184B5, 184B5/HER (Fig. 1A), and premalignant oral epithelial cells (data not shown). The receptor was also detected in human breast cancer (Fig. 1B). To investigate if the PPAR␥ receptor expressed in cell lines was transcriptionally active, 184B5/HER and MSK Leuk1 cells were transfected with a PPAR response element cloned upstream of luciferase (PPRE3-tk-luciferase). Treatment of 184B5/HER (Fig. 1C) or MSK Leuk1 cells (data not shown) with PPAR␥ ligands (ciglitazone, 15d-PGJ 2 ) caused a dose-dependent increase in promoter activity. Similar effects were observed with troglitazone.
The possibility that PPAR␥ ligands inhibited PMA-mediated induction of PGE 2 synthesis was investigated. Treatment of 184B5/HER cells with PMA led to a severalfold increase in PGE 2 production. This effect was suppressed by PPAR␥ ligands in a dose-dependent manner (Fig. 2). PPAR␥ ligands also inhibited PMA-mediated induction of PGE 2 synthesis in MSK Leuk1 cells (data not shown). To determine whether the above effects on production of PGE 2 could be related to differences in amounts of COX-2, Western blotting of cell lysate protein was carried out. PMA induced COX-2 protein (Fig. 3, A-D and G). Treatment with PPAR␥ ligands (ciglitazone, Fig. 3A; 15d-PGJ 2 , Fig. 3B; troglitazone, Fig. 3, C and G) caused a dose-dependent decrease in PMA-mediated induction of COX-2. In contrast, the M metabolite of troglitazone, a compound that cannot transactivate PPAR␥, did not block the induction of COX-2 by PMA (Fig. 3D). In addition to PMA, sphingomyelinase and taxol are known to induce COX-2 (56,57). Hence, we also determined whether PPAR␥ ligands could suppress sphingomyelinase-and taxol-mediated induction of COX-2. Ciglitazone caused dosedependent suppression of the induction of COX-2 by sphingomyelinase (Fig. 3E) and taxol (Fig. 3F).
Transcriptional Activation of COX-2 Is Inhibited by PPAR␥ Ligands-To elucidate further the mechanism responsible for the changes in amounts of COX-2 protein, we determined steady state levels of COX-2 mRNA by Northern blotting. As shown in Fig. 4, A and B, treatment with PMA enhanced levels of COX-2 mRNA, an effect that was suppressed by ciglitazone or troglitazone in a concentration-dependent manner. Comparable effects were observed with 15d-PGJ 2 (data not shown). Nuclear run-off assays were performed to determine whether differences in amounts of COX-2 mRNA reflected altered rates of transcription. We detected a marked increase in rates of synthesis of nascent COX-2 mRNA after treatment with PMA consistent with the differences observed by Northern blotting (Fig. 4C). This effect was suppressed by ciglitazone (Fig. 4C) and 15d-PGJ 2 (data not shown).
Transient transfections were performed to elucidate further the effects of PMA and PPAR␥ ligands on COX-2 transcription. PMA stimulated COX-2 promoter activity, an effect that was blocked by both ciglitazone (Fig. 5A) and 15d-PGJ 2 (Fig. 5B). The suppressive effects of ciglitazone and 15d-PGJ 2 were blocked by overexpressing a dominant negative form of PPAR␥. In addition to blocking transcriptional activation by endogenous PPAR␥, the dominant negative form of PPAR␥ lacks the ability to recruit CBP (53). Additional transient transfections were performed to confirm the role of PPAR␥ in mediating the inhibitory effects of ciglitazone and 15d-PGJ 2 . We examined the ability of a PPRE decoy oligonucleotide to prevent the inhibitory effects of ciglitazone and 15d-PGJ 2 on PMA-mediated stimulation of COX-2 promoter activity. As shown in Fig.  6, the PPRE decoy oligonucleotide relieved the suppressive effects of both ciglitazone (Fig. 6A) and 15d-PGJ 2 (Fig. 6B). In contrast, neither scrambled nor missense PPRE decoy oligonucleotides had any effect.
To define the region of the COX-2 promoter (Fig. 7A) that responded to PMA and PPAR␥ ligands, transient transfections were performed with a series of human COX-2 promoter deletion constructs. As shown in Fig. 7B, PMA treatment caused nearly a 4-fold increase in COX-2 promoter (Ϫ1432/ϩ59) activity, an effect that was suppressed by ciglitazone. Both the inductive effect of PMA and the suppressive effect of ciglitazone were detected with all COX-2 promoter deletion constructs except the Ϫ52/ϩ59 construct. A CRE is present between nucleotides Ϫ59 and Ϫ53, suggesting that this element may be responsible for mediating the effects of PMA. To test this notion, transient transfections were performed using COX-2 promoter constructs in which specific enhancer elements including the CRE were mutagenized. As shown in Fig. 7C, mutagenizing the CRE site caused a decrease in basal promoter activity and a loss of responsiveness to PMA. By contrast, mutagenizing the NF-IL6 or NFB sites had little effect on COX-2 promoter function.
PPAR␥ Ligands Inhibit COX-2 Expression via an Anti-AP-1 Mechanism-Electrophoretic mobility shift assays were performed to identify the transcription factor that mediated the induction of COX-2 by PMA. PMA caused increased binding to the CRE site of the COX-2 promoter, an effect that was suppressed by ciglitazone (Fig. 8A) or 15d-PGJ 2 (Fig. 8B). By contrast, PMA did not increase binding when the CRE site was mutagenized (data not shown). Supershift analyses identified c-Jun, c-Fos, and ATF-2 in the binding complex (Fig. 8C). Taken together, these results indicate that PMA stimulates the binding of the AP-1 transcription factor complex to the CRE of the COX-2 promoter; this effect was blocked by treatment with PPAR␥ ligands.
Additional experiments were done to define further the mechanism(s) by which PPAR␥ ligands inhibit PMA-mediated induction of AP-1 activity. As shown in Fig. 9A, ciglitazone caused dose-dependent suppression of PMA-mediated activation of an AP-1 reporter plasmid (2xTRE-luciferase). Similar results were obtained with 15d-PGJ 2 (data not shown). More- over, PMA induced c-Jun, a component of the AP-1 transcription factor complex; this effect was also inhibited by ciglitazone (Fig. 9B) or 15d-PGJ 2 (data not shown). To determine whether PPAR␥ ligands blocked PMA-mediated induction of COX-2 solely via effects on c-Jun, transient transfections were performed. As shown in Fig. 10A, ciglitazone blocked PMA-mediated stimulation of COX-2 promoter activity, an effect that was partially reversed by overexpressing c-Jun. In addition to suppressing the expression of c-Jun, ligands of nuclear receptors can potentially inhibit AP-1 activity by other mechanisms. There is growing evidence, for example, that CREB-binding protein (CBP/p300) is important for optimal AP-1-dependent transcription (58). Addition of a PPAR␥ ligand stimulates the interaction between PPAR␥ and CBP/p300 (59,60). Hence, PPAR␥ ligand-mediated inhibition of AP-1 activity could also be a consequence of competition for limiting amounts of CBP/ p300. To evaluate this possibility, transfection experiments KBM represents the Ϫ327/ϩ59 COX-2 promoter construct in which the NFB site was mutagenized; ILM represents the Ϫ327/ϩ59 COX-2 promoter construct in which the NF-IL6 site was mutagenized; CRM refers to the Ϫ327/ϩ59 COX-2 promoter construct in which the CRE was mutagenized; CRM-ILM represents the Ϫ327/ϩ59 COX-2 promoter construct in which both the NF-IL6 element and CRE were mutagenized. After transfection, cells were treated with vehicle (open columns), PMA (50 ng/ml, black columns), or PMA (50 ng/ml) plus ciglitazone (25 M, stippled columns). Reporter activities were measured in cellular extract 6 h later. Luciferase activity represents data that have been normalized with ␤-galactosidase activity. Columns, means; bars, S.D.; n ϭ 6. were performed with a CBP/p300 expression vector. As shown in Fig. 10A, overexpression of CBP also partially relieved the suppressive effect of ciglitazone. Interestingly, when CBP and c-Jun were overexpressed together, the inhibitory effect of ciglitazone was essentially abrogated. By contrast, overexpressing NFB, CEBP-␣, or CREB did not relieve the inhibitory effects of ciglitazone (Fig. 10B).
BADGE, a synthetic ligand for PPAR␥, was recently identified (61). Although this compound binds to PPAR␥, it has no apparent transactivation function (61). In fact, unlike ciglitazone or 15d-PGJ 2 , BADGE did not stimulate PPRE3-tk-luciferase activity in 184B5/HER cells (data not shown). We wondered whether this compound would still possess anti-AP-1 activity and thereby block PMA-mediated induction of COX-2. Interestingly, BADGE caused dose-dependent suppression of PMA-mediated induction of COX-2 (Fig. 11A). To confirm that BADGE did possess anti-AP-1 activity, transient transfections were performed. As shown in Fig. 11B, BADGE caused concentration-dependent suppression of PMA-mediated activation of an AP-1 reporter plasmid. DISCUSSION PPAR␥ ligands, like ligands of other nuclear receptors, modulate gene expression by multiple mechanisms. In the current study, we showed that PPAR␥ ligands suppressed the induction of COX-2 by an anti-AP-1 mechanism. The AP-1 transcription factor complex consists of a collection of dimers of members of the Jun, Fos, and ATF cAMP-response element-binding protein bZip families. Little is known about the potential of PPAR␥ ligands to interfere with AP-1-mediated gene expression. Transient transfection analyses indicated that the CRE site of the COX-2 promoter was important for the inductive effects of PMA. Electrophoretic mobility gel shift analyses showed that treatment with PMA augmented binding to the CRE of the COX-2 promoter; c-Jun, c-Fos, and ATF-2 were identified in the DNA binding complex. These findings are consistent with previous reports in which both AP-1 and the CRE were found to be important for the induction of COX-2 in human epithelial cells (57,62,63). The results are also consistent with the observations of Xie and Herschman (64,65) who were the first to demonstrate the importance of c-Jun and the CRE site for mediating the induction of COX-2. Importantly, several different results support the idea that ligands of PPAR␥ block the induction of COX-2 by antagonizing AP-1. First, PPAR␥ ligands blocked PMA-, taxol-, and sphingomyelinase-mediated induction of COX-2 (Fig. 2); each of these inducers has been reported to stimulate AP-1-mediated induction of COX-2 transcription (56,57,62,63). Second, PPAR␥ ligands inhibited PMA-mediated increases in AP-1 binding to the CRE of the COX-2 promoter. Finally, ligands of PPAR␥ suppressed PMAmediated activation of an AP-1 reporter plasmid.
Ligands of nuclear receptors, e.g. retinoids, have been reported to antagonize AP-1-mediated transcription by a variety of mechanisms (66,67). Hence, additional experiments were performed to elucidate the mechanism(s) by which PPAR␥ ligands inhibited AP-1-mediated induction of COX-2. We found that ligands of PPAR␥ blocked PMA-mediated induction of c-Jun, a component of the AP-1 transcription factor complex. The functional significance of this effect was confirmed by the finding that overexpressing c-Jun partially relieved the suppressive effects of ciglitazone on PMA-mediated induction of COX-2 promoter activity. Recent studies also suggest that transcriptional activation by AP-1 requires the coactivators CBP/ p300 (68). Ligands of PPAR␥ stimulate the interaction between PPAR␥ and CBP (59,60). Hence, competition for limiting amounts of these proteins represents a mechanism for transrepression by nuclear receptors including PPAR␥. In fact, CBP was recently implicated in PPAR␥-dependent repression of the inducible nitric-oxide synthase gene (33). Transient transfections were performed to investigate the potential of CBP to regulate COX-2 transcription. Overexpressing CBP partially reversed the suppressive effects of PPAR␥ ligands; this suggests that PPAR␥ ligands inhibited the stimulation of COX-2 promoter activity, in part, via a squelching mechanism. In support of this idea, overexpressing a dominant negative form of PPAR␥ that cannot bind CBP prevented the suppressive effect of PPAR␥ ligands (Fig. 5). When c-Jun and CBP were overexpressed simultaneously, the inhibitory effects of PPAR␥ ligands were essentially abrogated. To our knowledge, these findings represent the first evidence that PPAR␥ ligands can antagonize AP-1-mediated gene expression by multiple mechanisms (Fig. 12). Moreover, we are unaware of any prior work demonstrating that CBP is important for regulating COX-2 gene expression. Retinoids and dexamethasone, known ligands of nuclear receptors, can block the activation of COX-2 gene expression (55,69). A potential role for CBP/p300 in mediating these suppressive effects is suggested by the findings of the current study.
Overexpressing a dominant negative form of PPAR␥ or a PPRE decoy oligonucleotide relieved the suppressive effect of PPAR␥ ligands on COX-2 expression. Both treatments suppress PPAR␥-mediated transactivation of gene expression (38,53) suggesting that PPAR␥ ligands could mediate their inhib-itory effects on COX-2 induction by modulating the transcription of an unknown PPAR␥-responsive gene. This might contribute, in turn, to the observed anti-AP-1 effect of PPAR␥ ligands. Recently, PPAR␥ ligands were found to suppress the induction of COX-2 in PPAR␥(Ϫ/Ϫ) macrophages (70); this suggested that this class of compounds could act via a PPAR␥independent mechanism. By contrast, in our epithelial cell model, overexpressing a dominant negative form of PPAR␥ blocked the inhibitory effects of PPAR␥ ligands on COX-2 expression. Thus, PPAR␥ is required for mediating the suppressive effects of PPAR␥ ligands on COX-2 expression in this cell system.
Clearly, PPAR␥ can induce transcriptional activation through specific DNA sites or inhibit the transcription factor AP-1. A pharmacological approach was used to determine whether these two types of receptor actions were mechanistically distinct. As noted above, BADGE is a synthetic ligand that binds to the receptor but is unable to transactivate genes via PPAR␥ (61). We investigated whether this functionally restricted PPAR␥ ligand blocked the induction of COX-2 or AP-1 activity like other PPAR␥ ligands. Importantly, although BADGE did not activate PPAR␥, it suppressed PMA-mediated induction of AP-1 activity and COX-2 expression. This finding suggests that it may be feasible to develop a class of PPAR␥ ligands that selectively inhibit AP-1 activity without stimulating transcription. There is precedent for this idea. AP-1-selective retinoids have been developed (71); these retinoids inhibit AP-1 activity but are unable to stimulate transcription (71). AP-1-selective PPAR␥ ligands would be anticipated to have different therapeutic properties and toxicity than traditional PPAR␥ ligands.
Selective COX-2 inhibitors possess both chemopreventive and anti-inflammatory properties. Compounds that interfere with the signaling mechanisms that stimulate COX-2 transcription should also inhibit carcinogenesis and decrease inflammation. In support of this idea, PPAR␥ ligands can inhibit carcinogenesis (44 -46) and reduce inflammation (32,41). Several of the known anti-neoplastic properties of PPAR␥ ligands may be explained, in part, by their ability to inhibit COX-2 expression and PG biosynthesis. For example, overexpression of COX-2 promotes angiogenesis (72) and inhibits apoptosis (73), whereas PPAR␥ ligands inhibit both of these effects (38 -40, 42). Both selective COX-2 inhibitors and PPAR␥ ligands protect against breast cancer in experimental animals (27,46). Lane 1 represents a COX-2 standard. Cellular lysate protein (25 g/ lane) was loaded onto a 10% SDS-polyacrylamide gel, electrophoresed, and subsequently transferred onto nitrocellulose. The immunoblot was probed with antibody for COX-2. B, 184B5/HER cells were cotransfected with 1.8 g of 2xTRE-luciferase and 0.2 g of pSV␤gal. Following transfection, cells were treated with vehicle (control), PMA (50 ng/ml), or PMA plus BADGE (0 -500 M) for 7 h. Luciferase activity represents data that have been normalized with ␤-galactosidase activity. Columns, means; bars, S.D.; n ϭ 6.
FIG. 12. Schematic of proposed mechanism by which PPAR␥ ligands inhibit AP-1-mediated activation of COX-2 transcription. CBP/p300 links AP-1 with components of the basal transcription machinery. TBP, TATA-box-binding protein; TFIIB, transcription factor IIB; RNA Pol II, RNA polymerase II. Treatment with PMA increases the binding of AP-1 to the CRE site of the COX-2 promoter thereby enhancing transcription. This stimulatory effect of PMA is blocked by cotreatment with a PPAR␥ ligand. PPAR␥ ligands inhibit PMA-mediated induction of COX-2 by two mechanisms as follows: 1) induction of c-Jun, a component of the AP-1 transcription factor complex, is blocked; 2) binding of a PPAR␥ ligand to its receptor enhances the interaction between CBP/p300 and PPAR␥. This results in less CBP/p300 being available for AP-1-mediated activation of COX-2, a process known as squelching.
Local production of estrogen in breast adipose tissue, a reaction catalyzed by aromatase, has been implicated in the development of breast cancer. Interestingly, the synthesis of aromatase is stimulated by PGE 2 (74) and inhibited by PPAR␥ ligands (75). Our finding that PPAR␥ ligands block the induction of COX-2 and PGE 2 synthesis may be important, therefore, for understanding how PPAR␥ ligands inhibit mammary carcinogenesis (46).
Finally, the results of this study may provide additional insights into the mechanisms underlying the anti-diabetic effects of PPAR␥ ligands. COX-2 is constitutively expressed in pancreatic islet cells (76). Prostaglandin E 2 negatively modulates glucose-induced insulin secretion, an effect that can be blocked by inhibitors of COX (54). The discovery that PPAR␥ ligands inhibit the production of COX-2-derived PGE 2 may help to explain the hypoglycemic effects of this class of agents.