The activation of peroxisome proliferator-activated receptor (PPAR)γ is known to inhibit angiogenesis. As a potential mechanism for this, we aimed at examining the effects of PPARγ agonists on the pro-angiogenic enzyme cyclooxygenase (COX)-2 in human endothelium.
Cultured endothelial cells were pre-incubated with the PPARγ agonists rosiglitazone (RSG) or GW1929 before stimulation with vascular endothelial growth factor (VEGF) or phorbol myristate acetate (PMA). RSG and GW1929 attenuated VEGF- and PMA-stimulated COX-2 activity, as well as protein and mRNA expression. This effect was abolished by the PPARγ antagonists bisphenol A diglycidyl ether and GW9662 as well as by PPARγ small-interfering RNAs (siRNAs). Transient transfection experiments revealed that the induction of COX-2 promoter was significantly inhibited by RSG through an interference with the cAMP response element (CRE) site. COX-2 downregulation after siRNA targeting CRE-binding protein (CREB) confirmed the role of CREB in mediating COX-2 transcription. Correspondingly, PPARγ agonists attenuated CREB activation. As both protein kinase C (PKC)α and β are involved in VEGF-induced COX-2 expression and CREB activation, we investigated which isoform(s) of PKC was affected by RSG. RSG only reduced VEGF- and PMA-stimulated PKCα membrane translocation.
VEGF induces CREB-mediated COX-2 expression through a PKCα-dependent pathway in human endothelium. The anti-angiogenic effect of PPARγ agonists is due, at least in part, to an interference with the VEGF-stimulated PKCα-mediated activation of CREB and the related expression of COX-2.
Angiogenesis plays a pathophysiological role in a number of disease conditions, including cancer, diabetic retinopathy, and atherosclerosis.1 A pro-angiogenic microenvironment resulting from hypoxia, with a predominant role of vascular endothelial growth factor (VEGF) in combination with pro-inflammatory mediators, activates endothelial cells leading to extracellular matrix remodelling, endothelial cell migration, proliferation, and differentiation into new capillaries.
The molecular mechanisms regulating angiogenesis are incompletely understood, but involve pro-inflammatory prostaglandins, acting at multiple sites.2 Rate-limiting enzymes in the prostaglandin biosynthesis are cyclooxygenases (COX)-1 and -2.3 The inducible and pro-inflammatory isoform COX-2 has been shown to orchestrate angiogenesis, by promoting endothelial survival, proliferation, migration, and tubular sprouting, as well as by increasing VEGF production.2 COX-2 is overexpressed in the angiogenic neovasculature within tumours,4 atherosclerotic plaques,5 and diabetic retinas,6 where the selective inhibition of COX-2 activity leads to a suppression of angiogenesis and inhibition of disease progression.
Peroxisome proliferator-activated receptor (PPAR)γ is a ligand-activated transcription factor belonging to the nuclear receptor superfamily.7 PPARγ activators include the insulin-sensitizing thiazolidinediones (TZDs) rosiglitazone (RSG) and pioglitazone, now clinically approved for the treatment of type 2 diabetes. Beyond their metabolic actions, PPARγ agonists exert direct vasculoprotective effects, being able to inhibit vascular inflammation and atherosclerosis progression in animal models and in humans with or without diabetes.7 Importantly, they also modulate angiogenesis, both in vitro and in vivo.8,9 Underlying anti-angiogenic mechanisms targeting endothelial cells include the suppression of Akt phosphorylation, VEGF receptors-1 (Flt-1) and 2 (Flk/KDR), and urokinase plasminogen activator expression, the induction of plasminogen activator inhibitor-1 and the thrombospondin receptor CD36, as well as the disorganization of the actin cytoskeleton.10 However, despite the recognized pro-angiogenic role for COX-2, little is known about the effects of PPARγ agonists on the endothelial COX-2 expression and activity.
Aim of this study was therefore to investigate the direct role of PPARγ activation on COX-2 expression and activity induced by VEGF in endothelial cells, and to explore underlying molecular mechanisms.
Human umbilical vein endothelial cells (HUVEC) were harvested, characterized, and maintained as described,11 conforming with the principles outlined in the Declaration of Helsinki (see Cardiovasc Res 1997;35:2–3). Cells were obtained from discarded umbilical veins and treated anonymously. As such, approval from the University ethics review board was not necessary. Cells were used up to the fifth passage from primary culture. The human microvascular endothelial cell line (HMEC-1), obtained from Dr Thomas J. Lawley, was cultured as described.12 All experiments, when not otherwise specified, were performed in HUVEC, but control experiments were also done in HMEC-1, with identical results. For treatment with PPARγ agonists, subconfluent endothelial cells were pre-incubated with RSG or GW1929 for 24 h in 10% foetal calf serum (FCS) medium. Cells were then shifted to low-serum medium (2.5% FCS) for 16 h and subsequently stimulated with 20 ng/mL VEGF or 10 nmol/L PMA for 0–12 h. Cellular toxicity by RSG (±stimuli used) was checked for concentrations up to 10 µmol/L through a variety of techniques including cell count, morphology, Amido-black, and 3-(4, 5-dimethylthiazolyl-2)-2, 5-diphenyltetrazolium bromide assays.
For COX-2 and PPARγ immunocytochemistry, see Supplementary material online.
Measurement of COX activity
The concentration of 6-keto-PGF1α in the cell medium was determined with a competitive enzyme immunoassay from Cayman. Detailed protocols are provided in Supplementary material online.
The formation of vascular-like structures by HUVEC was assessed on the basement membrane matrix, growth factor-reduced Matrigel. Detailed protocols are provided in Supplementary material online.
Cell migration assay
Endothelial cell migration was assessed by performing a scratch wound healing assay. Detailed protocols are provided in Supplementary material online.
Cell lysis and immunoblotting
Total cellular proteins were isolated and immunoblotted as previously described,13 using antibodies against COX-2 (Cayman), β-actin (Sigma), CRE-binding protein (CREB), and phosphoCREBSer133 (p-CREB) (Upstate Cell Signaling Solution, Lake Placid, NY, USA), COX-1, protein kinase C (PKC)α, PKCβ1, PPARγ, ERK 1/2, and the Na+/K+-ATPase α1 subunit (Santa Cruz Biotechnology, Santa Cruz, CA, USA).
Assessment of PKC translocation
After exposure to RSG and subsequent stimulation with PMA or VEGF for 10 min, HUVEC were harvested and processed as previously described.13
Isolation of RNA and northern analysis
Northern blots were performed as previously published.14
Targeting silencing of PPARγ and CREB by small-interfering RNAs
For PPARγ and CREB gene silencing by RNA interference, detailed protocols are provided in Supplementary material online.
Transient transfections and luciferase assay
The previously described COX-2 promoter luciferase reporter constructs13 were kindly provided by Prof. Aida Habib (Department of Biochemistry and Internal Medicine, American University of Beirut, Beirut, Lebanon) and transfected. Detailed protocols are provided in Supplementary material online.
Results are expressed as means ± SD of three independent experiments performed in triplicate. The Student's t-test was used for comparing means between control groups and compound-treated group. Multiple comparisons were performed by one-way ANOVA. A P< 0.05 was considered statistically significant.
RSG and NS-398 inhibit angiogenic responses in endothelial cells
We first confirmed the role of PPARγ activation and documented the involvement of COX-2 activity in angiogenesis by performing two in vitro angiogenic assays: (i) a scratch wound healing assay, which specifically evaluates the stimulated migratory capacity of endothelial cells and (ii) a Matrigel assay, which evaluates the angiogenic factor-stimulated organization of endothelial cells into tube-like structures. A 24 h pretreatment with 10 µmol/L RSG (a concentration close to serum levels achievable in humans15) inhibited the stimulated locomotion of endothelial cells (width of denuded area: 130 ± 9 vs. 84 ± 8 µm, for RSG vs. VEGF, respectively) (Figure 1A). As the activation of COX-2 is critically involved in orchestrating specific steps of the angiogenic cascade,2 we tested the effects of the COX-2 selective inhibitor NS-398 on the endothelial angiogenic response. Exposure to 5 µmol/L NS-398 for 30 min before growth factor stimulation resulted in a marked inhibition of cell migration (155 ± 4 vs. 84 ± 8 µm, for NS-398 vs. VEGF, respectively) (Figure 1A), thus confirming endothelial COX-2 as a pro-angiogenic mediator. Correspondingly, both RSG and NS-398 significantly reduced the stimulated tube-like differentiation of HUVEC in Matrigel, as indicated by 30% and 60% inhibition of cellular branching points, respectively (Figure 1B). Similar results were obtained using the phorbol ester PMA16 as a stimulus (data not shown). At the concentrations and times used in our angiogenesis assays, we observed no cytotoxic effects by RSG (data not shown).
Causal role of PPARγ in RSG-induced inhibition of VEGF- and PMA-stimulated COX-2 activity and protein expression
We investigated the possibility that RSG inhibits PMA- or VEGF-mediated COX-2 activity by measuring the production of the COX-derived prostaglandin 6-keto-PGF1α, the major non-enzymatic by-product of PGI2. RSG (10 µmol/L, 24 h) significantly suppressed both PMA- and VEGF-induced PGI2 production (see Supplementary material online, Figure S1).
We examined RSG effects directly on COX-2 protein expression by western analysis and immunocytochemistry. COX-2 expression was highly induced by VEGF (20 ng/mL, 6 h), but RSG and GW1929 caused a significant decrease in VEGF-mediated COX-2 induction without affecting the constitutive expression of COX-1 (Figure 2A). Similar results were obtained using PMA as a stimulus (Figure 2B). Immunocytochemistry confirmed western blot results (Figure 2C). To test whether COX-2 downregulation by RSG was truly mediated by PPARγ, we used two pharmacological antagonists of PPARγ, GW9662 and BADGE. Pretreatment with GW9662 (2 µmol/L) or BADGE (50 µmol/L) before RSG completely prevented the suppressive effect of RSG on the stimulated COX-2 protein expression (Figure 3A and B). To further confirm a PPARγ-dependent signalling, HUVEC were transfected with PPARγ-specific small-interfering RNAs (siRNA) and then assayed for PPARγ and COX-2 expression. PPARγ protein expression was specifically and significantly reduced by the cognate PPARγ siRNA compared with mock transfections (no siRNA) or transfection with a control scrambled siRNA the sequence of which is unrelated to PPARγ (Figure 3C). Interestingly, compared with the mock condition, PPARγ siRNA significantly reverted RSG-mediated COX-2 inhibition (Figure 3D). Taken together, these results show that proangiogenic agents exert at least part of their action on COX-2 expression through a mechanism negatively affected by PPARγ agonists.
PPARγ agonists inhibit PMA- and VEGF-stimulated COX-2 at a transcriptional level
Consistent with reduced COX-2 protein expression, PPARγ agonists also decreased COX-2 mRNA steady-state levels at northern blot (see Supplementary material online, Figure S2).
To determine whether the regulation of stimulated COX-2 mRNA by PPARγ agonists occurred at the level of transcription, we carried out transient transfection studies. To this aim, HMEC-1 were transiently transfected with a series of human COX-2 promoter region/luciferase reporter gene constructs and treated with or without RSG for 24 h, followed by stimulation with PMA. PMA treatment caused a seven-fold increase in the ‘full-length’ COX-2 promoter (−1432/+59 bp) activity compared with unstimulated control cells, an effect that was inhibited by 40% after RSG pretreatment (Figure 4A, lane 1). The RSG suppressive effect on stimulated COX-2 promoter activity suggested that RSG acts through a transcriptional mechanism not requiring the peroxisome proliferators responsive element site located at −3721/−3707 bp in the COX-2 promoter.17 Next, we observed that PMA retained its ability to increase luciferase activity (by approximately six-fold) also after transfection of the smaller COX-2 promoter construct (−327/+59 bp), containing only the proximal NF-κB, nuclear factor-interleukin (NF-IL)6, and CRE sites (Figure 4A, lane 2 vs. lane 1). Similarly, also the suppressive effect of RSG was retained in cells transfected with the same construct. This indicates that the −327/+59 bp region of the human COX-2 gene is per se sufficient to mediate both the induction by PMA and the suppression by RSG of COX-2 transcription. To determine which cis-acting elements within the COX-2 promoter actually mediate the inhibitory effect of RSG, the promoter activity of the wild-type region (−327/+59 bp) was compared with that of the deletion or site-directed mutants of the human COX-2 promoter in their ability to respond to RSG. The inducing effect of PMA and the suppressive effect of RSG were detected with all COX-2 promoter deletion mutants (Figure 4A, lanes 3 and 4 vs. lane 2), with the exception of the −52/+59 bp construct containing only the TATA box, as well as with the CRE mutagenized construct (Figure 4A, lanes 5 and 6 vs. lanes 4 and 2, respectively). The observation that the −124/+59 bp construct, only containing the CRE site, still retains the same RSG suppressive effect on COX-2 promoter activity upon PMA (Figure 4A, lane 4) and VEGF (see Supplementary material online, Figure S3) stimulation suggested the CRE site as both necessary and sufficient for PPARγ agonist-mediated downregulation of COX-2.
Typically, the CRE sequence (−59/−53 bp) is bound by the transcription factor CREB in response to cytokines or PMA.18 To expand the promoter analysis results and further confirm the CRE/CREB role in mitogen-induced COX-2 expression, HUVEC were transfected with CREB-targeted siRNA to specifically silence CREB protein and then stimulated with PMA or VEGF. Western analysis revealed that CREB silencing halved both CREB levels (Figure 4B, inset), as well as PMA- and VEGF-induced COX-2 expression (Figure 4B, left and right panels), thus suggesting a role for CREB in COX-2 transcriptional activation. CREB knockdown by siRNA also significantly reduced VEGF-stimulated endothelial tube formation and migration (see Supplementary material online, Figure S4), indicating that VEGF-stimulated angiogenic responses of endothelial cells are dependent, at least in part, on CREB activity.
PPARγ agonists inhibit CREB activation
To evaluate whether COX-2 transcriptional inhibition by RSG is mediated by an interference with CREB activation, we tested RSG effect directly on PMA- and VEGF-stimulated CREB phosphorylation at serine 133. Both PMA and VEGF induced CREB phosphorylation after 20 min of stimulation (Figure 5A and B). RSG and GW1929 inhibited the stimulated CREB phosphorylation, without influencing total CREB protein. It appeared therefore likely that PPARγ agonists modulate COX-2 expression via an attenuation of CREB activity. For this reason, we were prompted to investigate the interference by RSG with the molecular signalling upstream of CREB phosphorylation.
RSG decreases membrane translocation of PKCα
As COX-2 expression and CREB activation induced by PMA and VEGF are regulated by PKC activity,19–21 we confirmed the involvement of PKC activation in COX-2 expression and CREB phosphorylation (see Supplementary material online, Figure S5). In particular, we found that conventional PKCα and PKCβ1 are implicated in signalling upstream of CREB phosphorylation and subsequent COX-2 induction by both PMA and VEGF (see Supplementary material online, Figure S5).
To address the possible interference of RSG with the activation of PKCα and PKCβ1, we evaluated RSG effect on PKCα and PKCβ1 membrane translocation, as indices of PKC activation. Western analysis of cytosolic and membrane fractions showed that, upon PMA or VEGF stimulation, RSG reduced the membrane translocation of PKCα, but not PKCβ1 (Figure 6).
The present study demonstrates that PPARγ agonists, at pharmacologically relevant concentration, exert an anti-angiogenic activity by inhibiting VEGF-induced endothelial COX-2 expression and prostaglandin production via the attenuation of the PKCα-mediated activation of the transcription factor CREB.
A considerable body of evidence supports vasculoprotective effects of PPARγ agonists, independent of their metabolic control. In particular, they attenuate inflammation and improve several surrogate markers of atherosclerosis in humans.7 In addition, experimental and clinical data indicate that PPARγ activation inhibits pathological neovascularization.8,9,22
The effects of PPARγ on angiogenesis is not totally univocal in the literature, with one study particularly reporting an induction of angiogenesis by PPARγ agonists, possibly through an induction of VEGF in non-endothelial cells.23 Such discrepancy may be attributed to the pleiotropic actions of PPARγ ligands, times and doses of treatment, cell/tissue types and the experimental model used, and the complex balance between angiogenic inducers and suppressors. Independent of a possible VEGF induction, however, PPARγ ligands have been shown to inhibit the endothelial expression of VEGF receptors and the subsequent activation of downstream signalling pathways.8,9 Overall, the observed inhibition of tumour,24 retinal,8 and atherosclerotic plaque22 neovascularization strongly support the prevalence of anti-angiogenic effects of PPARγ agonists. The present study investigates such anti-angiogenic action of PPARγ agonists at the molecular level.
COX-2 has recently emerged as a key regulator of inflammatory angiogenesis.2 Although COX-2 produces the vasodilatory and antithrombotic PGI2, it has been also implicated in the pathogenesis of atherosclerosis,25 cancer,4 and proliferative retinopathy.6 VEGF is a crucial angiogenic factor in both physiological and pathological processes and has been recently shown to promote atherosclerotic plaque neovascularization and progression.26 As endothelial COX-2 expression is induced by VEGF,21 and as PPARγ agonists inhibit VEGF-regulated neovascularization, we examined whether PPARγ activation inhibits VEGF-mediated angiogenesis through the modulation of the stimulated COX-2 expression and activity in endothelial cells. Using two in vitro angiogenesis assays, the scratch wound healing and the Matrigel assay, we first confirmed that COX-2 activity is critically involved in VEGF-induced angiogenic response, as demonstrated by the impairment of VEGF-induced endothelial migration and tubular differentiation after treatment with the selective COX-2 inhibitor NS-398. In agreement with the previous studies,8,9 here we found that the PPARγ agonist RSG mimics NS-398 anti-angiogenic effect, suggesting that at least part of the anti-angiogenic effect by RSG is due to an interference with endothelial COX-2 gene expression. RSG indeed decreased COX-2 protein expression and prostaglandin production upon PMA and VEGF stimulation. As COX-2-derived prostaglandins are able to upregulate VEGF expression,27 the inhibition of VEGF-stimulated COX-2 expression by PPARγ agonists might interfere with this positive regulatory loop and provides a plausible explanation for PPARγ beneficial effects in angiogenic disorders.
Although PPARγ agonists have been reported to exert their effects through PPARγ-dependent and -independent mechanisms,28 the observed inhibition of COX-2 found here is likely to be mediated by the activation of PPARγ, as demonstrated by several lines of evidence. First, we confirmed previous data9 that endothelial cells highly express PPARγ receptor in the nucleus and, to a lesser extent, in the cytosol (see Supplementary material online, Figure S6). Secondly, the effect of the TZD RSG on COX-2 was mimicked by a different synthetic non-TZD PPARγ agonist, the tyrosine derivative GW1929, with equivalent inhibitory potency. Finally, we tested specific pharmacological (BADGE and GW9662) and genic (siRNA) inhibitors of PPARγ on COX-2 suppression by RSG and GW1929. The finding that PPARγ inhibition by BADGE and GW9662, as well as siRNA-mediated PPARγ knockdown, all overcome COX-2 suppression by RSG and GW1929 demonstrates an essential involvement of PPARγ in their effect.
The regulation of COX-2 expression is both transcriptional and post-transcriptional.29 Our results from northern analysis and promoter transfection experiments, without ruling out a role of post-transcriptional mechanisms, indicate a transcriptional regulation of COX-2 by PPARγ agonists. The 5′-flanking promoter region of human COX-2 contains a canonical TATA box and multiple regulatory elements, including two putative NF-κB binding sites, one NF-IL6 site and one CRE site, which govern COX-2 expression in a cell type- and stimulus-specific manner.29 Although other binding sites, such as NF-κB and NF-IL6 sites, participate in optimizing COX-2 promoter activity, we here confirm the essential role of the CRE site in the activation of COX-2 promoter.18 The transcription factor CREB can bind to CRE in the COX-2 promoter, and thus favour increased transcription.18 In our cellular system, the role of CREB in COX-2 induction was confirmed by siRNA experiments, showing that CREB knockdown caused a significant decrease in the stimulated COX-2 expression. Therefore, CREB represents a potential molecular target responsible for PPARγ-mediated inhibition of COX-2. Interestingly, endothelial migration and tube formation were also reduced by CREB knockdown, suggesting that CREB plays a significant role in the endothelial angiogenic responses, at least in part by regulating COX-2 expression and other CREB-dependent proangiogenic genes.30 We found that PPARγ agonists reduce PMA- and VEGF-stimulated CREB phosphorylation, an index of activation, without affecting total CREB protein expression. Only two studies in general have reported an interference of PPARγ agonists with CREB activation/phosphorylation,31,32 and our report is the first evidence that PPARγ agonists inhibit CREB activity in endothelial cells.
Several studies have also demonstrated a critical role for PKC isoforms in mediating VEGF-dependent angiogenic responses by the vascular endothelium.33,34 As it has been reported that VEGF-induced CREB phosphorylation and COX-2 expression in endothelial cells are both PKC-dependent,20,21 and since PMA, an exogenous activator of conventional and novel PKC isoforms, mimicked VEGF effects on COX-2 and CREB, we explored the possibility that PPARγ could inhibit CREB-mediated COX-2 induction by suppressing PKC activity. We found that conventional PKCα and β1 are involved in the signal transduction upstream of CREB and COX-2. Both PKC isoforms are indeed activated by VEGF and PMA, as indicated by their translocation to membrane fractions, but PPARγ agonists only inhibited the translocation of PKCα. Therefore, we conclude that PPARγ agonists reduce CREB-mediated COX-2 expression interfering with the proximal switch of PKCα activation in the VEGF signalling pathway, thus demonstrating a novel proinflammatory role for PKCα in endothelial cells. Potential mechanism(s) responsible for the inhibition of PKCα by PPARγ agonists include the reduction of diacyglycerol (DAG) cytosolic concentrations through increased DAG-kinase activity;35 the inhibitory protein–protein interaction of PPARγ and PKCα, as recently demonstrated in monocytes;36 or some enhancement of phosphatase activity responsible for PKC inactivation.37 On this background, further studies are likely warranted to precisely establish the level of PPARγ interference in the signalling upstream of PKC activation in our experimental conditions.
Inhibition of angiogenesis by PPARγ agonists might not always be favourable, because it might negatively impact on diabetes-associated impaired vascular collateralization after myocardial infarction and peripheral arterial ischaemia. Therefore, the precise scope of PPARγ activation in angiogenesis in the clinical setting of diabetes is complex and needs to be established clinically, also in the broader context of other potential systemic adverse effects of PPARγ agonists, as a class or as individual agents. Of note, pioglitazone demonstrated beneficial effects with respect to macrovascular events in patients with type 2 diabetes,38 whereas the effect of RSG on cardiovascular outcomes still remains controversial.39
Finally, our findings highlight a potential role of CREB as a therapeutic target in angiogenesis. As CREB has been recently implicated in carcinogenesis, insulin resistance, and neointima formation40 and in mediating VEGF signalling and expression,20,41 as well as the cytokine-induced COX-2 expression in endothelial cells,18 its observed inhibition by PPARγ agonists is the prototype of other types of interventions on this new endothelial target.
This work was supported by a grant from GlaxoSmithKline (through Dr Stephen A. Smith) and a grant from the Fondazione ‘G. d'Annunzio’—Chieti (Italy).
We are grateful to the Division of Obstetrics and Gynecology at the Vito Fazzi Hospital in Lecce (Italy) for providing umbilical cords; to Prof. Aida Habib (American University of Beirut, Beirut, Lebanon) for supplying the COX-2 promoter constructs and to Dr Fabrizio Damiano (University of Salento, Lecce, Italy) for helpful technical assistance on COX-2 promoter activity experiments.
Conflict of interest: none declared.