IP receptor-dependent activation of PPARgamma by stable prostacyclin analogues.

Stable prostacyclin analogues can signal through cell surface IP receptors or by ligand binding to nuclear peroxisome proliferator-activated receptors (PPARs). So far these agents have been reported to activate PPARalpha and PPARdelta but not PPARgamma. Given PPARgamma agonists and prostacyclin analogues both inhibit cell proliferation, we postulated that the IP receptor might elicit PPARgamma activation. Using a dual luciferase reporter gene assay in HEK-293 cells stably expressing the IP receptor or empty vector, we found that prostacyclin analogues only activated PPARgamma in the presence of the IP receptor. Moreover, the novel IP receptor antagonist, RO1138452, but not inhibitors of the cyclic AMP pathway, prevented activation. Likewise, the anti-proliferative effects of treprostinil observed in IP receptor expressing cells, were partially inhibited by the PPARgamma antagonist, GW9662. We conclude that PPARgamma is activated through the IP receptor via a cyclic AMP-independent mechanism and contributes to the anti-growth effects of prostacyclin analogues.

Prostacyclin (PGI2) and its stable analogues have potent vasodilatory, anti-platelet and anti-mitogenic effects in the cardiovascular system [1,2]. Classically these agents mediate their biological effects by binding to cell surface prostacyclin (IP) receptors, which couple to the stimulatory G protein (GS) to activate adenylyl cyclase and increase cyclic AMP (cAMP) [1,3]. In addition to activating plasma membrane receptors, PGI2 can also signal through peroxisome proliferator-activated receptors (PPARs), a family of nuclear transcription factors that bind to specific peroxisome proliferator response elements (PPREs) in the promoter region of target genes to regulate their expression [3-5]. The PPAR family comprises of three isoforms, α, δ (β), and γ, which are activated by a broad spectrum of ligands, including hypolipidemic agents and derivatives produced by the cyclo-oxygenase and lipoxygenase pathway to regulate biological processes such as lipid metabolism, insulin sensitivity and inflammation [6-8]. Because of the complex interaction of prostaglandins with PPARs, the function of endogenous regulation by PGI2 has been hard to elucidate. Nonetheless, PGI2 signalling through PPARδ is thought to have important roles in embryo implantation [5], tumourgenesis [9] and apoptosis [10]. It is less clear if PGI2 is a major endogenous ligand for PPARα and PPARγ.

Stable PGI2 analogues like iloprost and carbacyclin, also act as PPAR ligands, directly binding to and causing transcriptional activation of PPARα and PPARδ in vitro [4]. The latter may in part mediate the anti-proliferative effects of treprostinil in lung fibroblasts [11]. Surprisingly, no information exists about whether these analogues can regulate PPARγ. Interestingly, the relatively selective IP receptor agonist, cicaprost does not bind or promote activation of either PPARα or PPARδ [4]. Thus, cicaprost has often been used experimentally to distinguish between mechanisms involving cell surface IP receptors and PPARs [3,5]. Moreover, PPAR activation is readily observed in cells not expressing the IP receptor [4,10,12], and this has led to the assumption that the receptor does not play a major role in directly regulating PPAR function. However, PPARs are substrates for several kinases, including cAMP-dependent protein kinase (PKA), which can phosphorylate all isoforms and enhance PPAR activity both in the absence and presence of ligands [13]. This raises the possibility that the IP receptor may contribute to the regulation of PPARs.

The aim of this study was to investigate whether PGI2 analogues could regulate PPARγ. Given its broad role as a suppressor of inflammation, tissue injury and cell proliferation [6,7], PPARγ could represent a therapeutic target for PGI2 analogues. Using a cell-based reporter gene assay in HEK-293 cell lines stably expressing the IP receptor or the empty plasmid, we make the novel observation that activation of PPARγ by PGI2 analogues is dependent on the presence of the IP receptor. The mechanism of activation is unknown but does not appear to involve cAMP.

Materials and methods

Cell lines and culture. The human IP receptor was cloned into the pcDNA3.1/Zeo vector (Invitrogen, Paisley, UK) and transfected into HEK-293 cells to generate a stable line (HEK-293-IP) as previously described [14]. As a control, a stable line transfected with the pcDNA3.1Zeo vector alone was also generated (HEK-293-Zeo). Two individual zeocin resistant colonies per cell type were isolated and maintained in minimal essential medium (MEM) containing Earle’s salts and l-glutamine (Invitrogen, Paisley, UK) supplemented with 10% fetal bovine serum (FBS) (Invitrogen), 1% penicillin–streptomycin (Invitrogen) and Zeocin (400 μg/ml) (Invitrogen).

Plasmid constructs. A chimeric receptor containing the yeast GAL4 DNA binding domain fused to human PPARγ was created by insertion of a GAL4 DNA binding domain encoding fragment into the mammalian expression vector pcDNA3 (Invitrogen) to generate the vector GAL4-pcDNA3. The PPARγ-LBD fragment was digested with BamHI and NotI and ligated into the vector GAL4-pcDNA3, which had been digested with the same enzymes to generate GAL4-hPPARγ-pcDNA3 [15,16]. The reporter plasmid for these GAL4 chimeric receptors (pGAL5TKpGL3) contains five repeats of the GAL4 response element upstream of a minimal thymidine kinase in the pGL3 luciferase expression vector (Promega, Southampton, UK). The control vector, pMLuc2 (Merck Biosciences, Nottingham, UK), contains the minimal thymidine kinase (TK) promoter adjacent to the Renilla luciferase gene and was used to control for transfection efficiency. Having reporter and control vectors containing the minimal TK promoter was crucial in the experimental design since treprostinil increased Renilla luciferase activity when driven by the full length TK promoter in the pRL-TK vector (Promega) (2.4-fold increase compared to untreated, P < 0.001), but only weakly under the control of the minimal-TK promoter (1.2-fold increase compared to untreated, P = NS).

Transfections and luciferase reporter gene assay. Cells were transfected in suspension using Lipofectamine 2000. The luciferase reporter construct pGAL5TKpGL3 was transiently co-transfected into cells with the Renilla control vector, pMLuc2 with either the GAL4-hPPARγ-pcDNA3 reporter construct or the GAL4-pcDNA3 control construct. Transfected cells were seeded onto 96-well plates at a density of 1 × 105 cells/ml and left for 48 h. Subsequently, cells were either left untreated or stimulated with agonist and/or antagonist for a further 24 h. Total cell lysates were prepared using 1X passive lysis buffer (Promega, Southampton, UK). Luciferase and Renilla activities were determined using a dual luciferase assay system (Promega) in a Tropix TR717 microplate luminometer (Applied Biosystems, Warrington, UK) according to manufacturer’s instructions. Background values from untransfected cells were substracted from all luciferase and Renilla readings. The luciferase values were normalised to Renilla values and expressed as mean fold increase from untreated cells.

Intracellular cyclic AMP measurement. Cells were grown to 70–80% confluence in 6-well plates and starved in MEM containing low serum (0.1%) for 48 h before being stimulated with agonist and/or antagonist for 30 min in media containing 10% FBS. Cyclic AMP was extracted and measured using a competitive enzyme immunoassay kit (Cyclic AMP ACE EIA kit, Cayman Chemical, Ann Arbor, MI) according to manufacturer’s instructions. Protein concentration was determined using the Bradford assay (Bio-Rad Laboratories, Hemel Hempstead, UK).

Reagents. Treprostinil sodium (also known as remodulin and UT-15) was kindly provided by United Therapeutics (Washington, MD) and the IP receptor antagonist, RO1138452 by Roche (Palo Alto, CA). Carbacyclin was purchased from Biomol (Exeter, UK), Rp-cAMPS from Biolog Life Science Institute (Bremen, Germany), rosiglitazone from Alexis Corporation (Lausanne, Switzerland), 2′5′-dideoxyadenosine (DDA) and GW9662 from Merck Biosciences (Nottingham, UK) while forskolin and pertussis toxin was from Sigma–Aldrich (Poole, UK). Drugs were prepared in dimethyl sulfoxide (DMSO) or water and then further diluted in media. The final concentration of DMSO did not exceed 0.01%.

Cell proliferation assays. HEK-293 cells were seeded onto 6-well plates at a density of 0.5–1 × 104 cells/ml and grown in MEM for 24 h before being growth arrested in MEM containing low serum (0.1% FBS) for 48 h. To assess effects on proliferation, cells were then incubated for 48 h in fresh media containing either low serum or 10% FBS in the absence or presence of drugs or DMSO as indicted. Cells were counted using an automated cell counter (Sysmex F-520 P; Malvern Instruments Ltd, UK).

Statistical analysis. Data are presented as means ± s.e.m. of n observations and analysed using One-way ANOVA with post-test correction for multiple comparisons. P values <0.05 were considered statistically significant.

Results

PGI2 analogues activate PPARγ

To test the functionality of the PPARγ construct containing the ligand-binding domain (LBD), we investigated the effect of the specific PPARγ agonist, rosiglitazone (1 μM). This agent significantly increased luciferase levels by 5-fold compared to untreated control, an effect essentially abolished by pre-treatment with the selective PPARγ antagonist, GW9662 (1 μM; Fig. 1A). The above effects could not be attributed to the solvent, since DMSO (0.01%) did not increase PPARγ activity (Fig. 1A). Likewise, the stable PGI2 analogue treprostinil significantly increased PPARγ activity, elevating it by ∼2-fold at 100 nM and ∼3-fold at 10 μM (Fig. 1B). In separate experiments, GW9662 largely prevented PPARγ activation by treprostinil (1 μM) and the cell permeable analogue, carbacyclin (1 μM; Fig. 1C). Furthermore, the relatively selective IP receptor agonist, cicaprost (1 μM) also significantly (P < 0.001) increased luciferase activity compared to untreated cells (Fig. 1 online supplement).

Fig. 1

Activation of PPARγ by rosiglitazone and stable prostacyclin analogues. HEK-293 cells stably expressing the IP receptor (HEK-293-IP) were transiently transfected with pGAL5TKpGL3, pMLUC-2 and either GAL4-pcDNA3 (control—white bars) or GAL4-hPPARγ-pcDNA3 (reporter—black bars). After 48 h, cells were stimulated with 10% FBS in the absence (Control) or presence of DMSO (0.01%; A), rosiglitazone (ROSI; A), treprostinil (TREP; B and C) or carbacyclin (CARBA; C). In (A and C), cells were pre-treated with the PPARγ antagonist, GW9662 for 1 h prior to the addition of agonists. Unless stated, all drug concentrations were 1 μM. Luciferase activity was determined 24 h after FBS stimulation and normalised to renilla activity. Results are expressed as mean fold increase in luciferase activity above untreated control ± s.e.m. (n = 12, 3 separate transfections). ∗∗∗P < 0.001 when compared to control or as shown. #P<0.05 when compared with TREP alone.

Activation of PPARγ by PGI2 analogues is IP receptor-dependent

The conventional activation of PPARs occurs via binding of an agonist to the LBD and this appears to account for the effects of carbacyclin and iloprost on PPARα and PPARδ [4]. To test whether PPARγ activation is mediated through the IP receptor, we examined the effect of a novel IP receptor antagonist, RO1183452 [17,18] in cells expressing the IP receptor. As expected, treprostinil greatly increased cyclic AMP in HEK-293-IP cells, an effect reduced by approximately 90% when cells were pre-treated with 1 μM RO1183452 (Fig. 2A). Moreover, in the presence of the receptor antagonist, both treprostinil and carbacyclin failed to significantly activate PPARγ in HEK-293-IP cells (Fig. 2B). Similarly, the effects of cicaprost on PPARγ activity were also prevented (Fig. 1, online). In contrast, rosiglitazone increased luciferase activity by several fold in the presence of RO1183452, the increase actually being greater (P < 0.05) than that observed without the antagonist (Fig. 2C). Furthermore, in cells stably expressing the empty vector, (HEK-293-Zeo), neither carbacyclin nor treprostinil-activated PPARγ (Fig. 2D), even at a concentration of 10 μM. This occurred despite the effectiveness of rosiglitazone at activating PPARγ (Fig. 2D). Likewise, intracellular cyclic AMP was not elevated by treprostinil in HEK-293-Zeo cells (Fig. 2A, online supplement) confirming the absence of functional IP receptors in these cells.

Fig. 2

IP receptor-dependent activation of PPARγ. (A) Cyclic AMP levels measured in HEK-293-IP cells stimulated for 30 min with 10% FBS ± either TREP (100 nM) or the IP receptor antagonist, R01138452 (IPRA; 1 μM) or a combination. Data are expressed as mean pmol of cyclic AMP per mg of total protein ± s.e.m. (in duplicate of 3–4 experiments). Luciferase activity was measured in HEK-293-IP (B and C) and HEK-293-Zeo (D) cells as described in Fig. 1. Cells were stimulated with 10% FBS alone (Control) or in the presence of either treprostinil (TREP), carbacyclin (CARBA), IPRA, rosiglitazone (ROSI) or a combination as shown. Unless otherwise stated, all drug concentrations were at 1 μM. Results are expressed as mean fold increase in luciferase activity ± s.e.m. (n = 12; 3 separate transfections). ∗∗∗P < 0.001 when compared to control or as shown and ∗P < 0.05 when compared to ROSI alone.

Role of cyclic AMP/PKA pathway

Having established that PGI2 analogues can activate PPARγ through the IP receptor, we wished to assess whether activation involved the cAMP pathway. Forskolin, an adenylyl cyclase activator failed to have any significant effect on luciferase activity in either HEK-293-Zeo or HEK-293 IP cells (Fig. 3A and B), even though it dramatically increased the levels of intracellular cAMP in both cell types (Fig. 3C and D). This suggests that cAMP alone cannot directly activate PPARγ. In addition, neither the adenylyl cyclase inhibitor, 2′5′-dideoxyadenosine (DDA) nor the PKA antagonist Rp-cAMPS (100 μM) significantly reduced activation of PPARγ by treprostinil in HEK-293 IP cells (Fig. 3A), although DDA did significantly reduce the cAMP elevating effects of treprostinil (Fig. 3C).

Fig. 3

IP receptor activation of PPARγ is independent of cyclic AMP and PKA. HEK-293-IP (A) and HEK-293-Zeo (B) cells were transfected with constructs as described in Fig. 1. After 48 h, cells were stimulated with 10% FBS alone (control) or in the presence of forskolin (10 μM), treprostinil (TREP; 1 μM), 2′5′-dideoxyadenosine (DDA; 100 μM), the PKA antagonist, Rp-cAMPS (100 μM) or in combination as shown. Results are expressed as mean fold increase in luciferase activity from untreated control ± s.e.m. (n = 12, 3 separate transfections). Cyclic AMP was measured in HEK-293-IP (C) or HEK-293-Zeo (D) cells stimulated with 10% FBS ± either forskolin (10 μM), DDA (100 μM) or TREP (100 nM) or a combination as shown. Results are expressed as mean pmol of cAMP per mg of total protein ± s.e.m. (measurements in duplicate of three experiments). ∗∗∗P < 0.001 when compared to control or as shown.

PPARγ antagonist partially reverses the antiproliferative effects of treprostinil

To determine if PPARγ might underlie the anti-proliferative effects of treprostinil, HEK-293-IP cells were pre-treated with GW9662 (1 μM) for 1 h prior to the addition of media containing 10% FBS and the analogue. Under these conditions, GW9662 caused a significant (P < 0.05) reversal of the antiproliferative effects of 100 nM treprostinil (Fig. 4A), while having no effect on serum-induced growth when given alone. In contrast, neither treprostinil (1 μM) nor carbacyclin (1 μM) had any effect on growth in HEK-293 cells not expressing the IP receptor (Fig. 2B online).

Fig. 4

PPARγ mediates in part the antiproliferative effects of treprostinil. Effect of treprostinil on FBS-induced proliferation in HEK-293-IP cells. Growth-arrested cells were stimulated with MEM + 10% FBS in the absence and presence of GW9662 (1 μM) treprostinil (TREP) or a combination. Cells were counted 48 hr later and data expressed as % cell proliferation relative to response mediated by 10% FBS alone and shown as mean ± s.e.m. (n = 6). ∗P < 0.05.

Discussion

The results presented here are to our knowledge the first demonstration that PGI2 analogues can directly activate PPARγ. Such, activation was however dependent on the presence of the IP receptor and could be reversed with RO1138452, a potent antagonist with a high selectivity for the IP receptor over EP and other members of the prostanoid receptor family [17,18]. This contrasts earlier studies where PGI2 analogues-induced activation of PPARα and δ occurred in cells not expressing the IP receptor [4,12] but is consistent with a recent study reporting the absence of treprostinil effects on PPARγ activity in wild type HEK-293 cells [11]. Thus, we provide good evidence for a novel receptor-mediated regulation of PPARγ, which appears distinct from direct ligand binding, but surprisingly does not appear to involve cyclic AMP. We also show that PPARγ may in part mediate the antiproliferative effects of treprostinil that were only observed in HEK-cells expressing the IP receptor.

In our system, we used a chimeric construct containing the Gal4 DNA binding domain (DBD) fused to the C terminal part of the PPARγ containing the D (hinge region), E (the ligand binding domain; LBD) and F domains [15]. Thus, for a cell permeable analogue like carbacyclin, one might predict PPARγ activation in control HEK-293 cells if the primary mechanism was indeed direct binding. Clearly this was not the case, with concentrations as high as 10 μM failing to significantly activate PPARγ, despite previous data showing near maximal binding and activation of PPAR α and δ with respect to their selective ligand activators [4]. Interestingly, cicaprost, which does not to bind or activate either α or δ isoforms was also an effective activator of our PPARγ construct, supporting the notion that receptor expression is crucial. However, we cannot exclude the possibility that the IP receptor is required to promote ligand binding by these agents, at least for this PPAR isoform.

There is a growing body of evidence indicating that PPARs can be activated via phosphorylation [19,20] and involve cell surface receptor activation. In rat adipocytes, both PPARα and PPARγ are phosphorylated following treatment with insulin, an effect associated with enhanced transcriptional activity [21,22]. This insulin-dependent phosphorylation has been attributed to activation of the mitogen-activated protein kinase (MAPK) pathway, involving serine residues contained within the N-terminal ligand-independent activation domain (AF-1) of PPARα [23] and PPARγ [22]. Likewise, the eicosanoid, prostaglandin F2α (PGF2α), phosphorylates PPARγ through a mechanism dependent on the FP receptor and the MAPK pathway [24]. Such phosphorylation by the FP receptor leads to an inhibition of adipogenesis in 3T3-L1 cells through suppression of PPARγ activity [24].

Most of the PPAR phosphorylation sites described above have been located in the N-terminal (A/B domain/AF-1). Since our construct lacks this domain, our data suggests there could be a phosphorylation site or a ligand independent activation site located in the C terminal (D, E or F domain) to promote ligand-independent transcriptional activity. Alternatively, IP receptor activation could simply enhance endogenous ligand binding to the LBD, as is the case with phosphorylation of the A/B domain in PPARγ [25] or promote recruitment or dissociation of co-factors. This may occur either via phosphorylation of PPARγ or the co-activators themselves [19].

Given the established coupling of the IP receptor to Gs [1] and the ability of PKA to activate PPARγ, both in the absence and the presence of exogenous ligands [13], we presumed that cAMP would play a major role in the IP receptor-dependent activation of PPARγ. However, forskolin failed to activate PPARγ and both the adenylyl cyclase antagonist, DDA, and the specific PKA antagonist, Rp-cAMPS failed to significantly reverse the PPARγ-activating effect of treprostinil. The lack of cAMP involvement may be due to the fact that we used a chimaeric construct containing only the LBD domain. Whilst this domain does contain important PKA phosphorylation sites, the DBD domain is strongly phosphorylated by PKA activators in PPARα [13]. Thus it is possible that several domains of PPAR need to be phosphorylated in order for PKA to activate PPARγ. Moreover, we cannot exclude the possibility that PKA might mediate activation of endogenous PPARγ that contributes to the anti-proliferative effects of treprostinil in HEK-293-IP cells. Nonetheless, previous data from this and other laboratories have described a cAMP-independent pathway mediating a substantial part of the vasorelaxant responses of PGI2 analogues in blood vessels [26-28].

These findings are important in furthering our understanding of the interaction of prostacyclin and PPARs. The idea that the IP receptor is crucial for PPARγ activation by PGI2 has physiological and therapeutic implications. PPARγ is an important regulator of cell differentiation and growth, particularly in the lung [6], and therefore loss or dysfunctionality of the IP receptor could alter the control of these processes. Evidence shows that there is a down-regulation of PPARγ expression in the lungs of patients with pulmonary hypertension [29]. Thus an attractive hypothesis is that treatment with PGI2 analogues and subsequent activation of PPARγ, may overcome this basal shortage and contribute to the beneficial effects of PGI2 therapy in this remodelling disease [2,30].

In summary, we describe a novel IP receptor-dependent activation of PPARγ by PGI2 analogues that appears to be distinct from the known ligand binding effects of these analogues to the other two PPAR isoforms. This regulation may previously have been missed due to PPAR function often being studied in cells not expressing the receptor. Further work is required to elucidate the mechanism of PPAR activation, which does not appear to involve cyclic AMP. Given that PPARγ suppresses proliferation of many cell types, suggests that this nuclear receptor may be a molecular target mediating some of the therapeutic effects of PGI2 analogues when given to treat pulmonary hypertension and peripheral vascular disease.