Abstract
Classically, the prostaglandin E2 (PGE2) receptor EP4 has been classified as coupling to the Gαs subunit, leading to intracellular cAMP increases. However, EP4 signaling has been revealed to be more complex and also involves coupling to pertussis toxin-sensitive Gαi proteins and β-arrestin-mediated effects. There are now many examples of selective activation of independent pathways by G protein-coupled receptor (GPCR) ligands, a concept referred to as functional selectivity. Because most EP4 ligands had thus far only been functionally characterized by their ability to stimulate cAMP production, we systematically determined the potencies and efficacies of a panel of EP4 ligands for activation of Gαs, Gαi, and β-arrestin relative to the endogenous ligand PGE2. For this purpose, we adapted three bioluminescence resonance energy transfer (BRET) assays to evaluate the respective pathways in living cells. Our results suggest considerable functional selectivity among the tested, structurally related agonists. PGE2 was the most selective in activating Gαs, whereas PGF2α and PGE1 alcohol were the most biased for activating Gαi1 and β-arrestin, respectively. We observed reversal in order of potencies between β-arrestin 2 and Gαi1 functional assays comparing PGE1 alcohol and either PGF2α, PGD2, or 7-[(1R,2R)-2-[(E,3R)-3-hydroxy-4-(phenoxy)but-1-enyl]-5-oxocyclopentyl]heptanoic acid (M&B28767). Most ligands were full agonists for the three pathways tested. Our results have implications for the use of PGE2 analogs in experimental and possibly clinical settings, because their activity spectra on EP4 differ from that of the native agonist. The BRET-based methodology used for this first systematic assessment of a set of EP4 agonists should be applicable for the study of other GPCRs.
- PGE2, prostaglandin E2
- AC, adenylate cyclase
- ANOVA, analysis of variance
- BRET, bioluminescence resonance energy transfer
- GFP, green fluorescent protein
- GPCR, G protein-coupled receptor
- HEK, human embryonic kidney
- PBS, phosphate-buffered saline
- PCR, polymerase chain reaction
- PI3K, phosphatidylinositol 3-kinase
- PTX, pertussis toxin
- Rluc, R. reniformis luciferase
- RT, room temperature
- YFP, yellow fluorescent protein
- M&B28767, 7-[(1R,2R)-2-[(E,3R)-3-hydroxy-4-(phenoxy)but-1-enyl]-5-oxocyclopentyl]heptanoic acid
- L-902688, 5R-[(1E)-4,4-difluoro-3R-hydroxy-4-phenylbut-1-en-1-yl]-1-[6-(1H-tertrazol-5-yl)hexyl]pyrrolidin-2-one
- GW627368X, N-{2-[4-(4,9-diethoxy-1-oxo-1,3-dihydro-2H-benzo[f]isoindol-2-yl)phenyl]acetyl} benzene sulfonamide
- U-46619, (5Z)-7-[(1R,4S,5S,6R)-6-[(1E,3S)-3-hydroxy-1-octenyl]-2-oxabicyclo[2.2.1]hept-5-yl]-5-heptenoic acid.
Prostanoids (prostaglandins and thromboxane) are lipid hormone mediators derived from cyclooxygenase-catalyzed metabolism of arachidonic acid. Prostaglandin E2 (PGE2) is the most widely produced prostanoid in the body. Its various effects include the contraction or relaxation of smooth muscle, modulation of immune responses, and the regulation of the production of a variety of cytokines. PGE2 is also implicated in several pathologies including cancer, inflammation, and hypertension. Four G protein-coupled receptors (GPCRs) designated subtypes EP1 through EP4, which activate different G protein-dependent signaling pathways but are often expressed in the same cell types, mediate the effects of PGE2. The biology of responses to PGE2 is correspondingly complex, but evidence is increasing that EP4 is a critical determinant for the role of PGE2 in carcinogenesis and cancer progression (Fulton et al., 2006).
Accordingly, studies using mice lacking the EP4 receptor [EP4(−/−)] and EP4-selective ligands have revealed a role for this receptor in the progression of colon carcinogenesis, in addition to closure of the ductus arteriosus at birth, bone formation, protection against inflammatory bowel disease, and progression of rheumatoid arthritis (Regan, 2003; Sugimoto and Narumiya, 2007). These studies suggest that EP4 agonism or antagonism could find clinical applications, for instance, in the treatment of cancers. The EP4 receptor was initially shown to couple to Gαs, leading to stimulation of adenylate cyclase and increases in intracellular cAMP concentrations (Coleman et al., 1994). More recently, it was demonstrated that EP4 could also couple to a PTX-sensitive inhibitory G protein (Gαi/o family), resulting in activation of phosphatidylinositol 3-kinase (PI3K) (Fujino and Regan, 2006). EP4-mediated PI3K signaling then leads to extracellular signal-regulated kinase-dependent induction of functional expression of early growth response factor-1, which may play a role in cancer, as well as phosphorylation of cAMP-response element-binding protein and inhibition of protein kinase A activity (Fujino et al., 2003, 2005). Moreover, activation of EP4 by PGE2 leads to rapid desensitization (Nishigaki et al., 1996), as well as recruitment of β-arrestins 1 and 2 and internalization of the receptor (Desai and Ashby, 2001). Accumulating evidence indicates that β-arrestins can also serve as scaffolds to activate signaling cascades independently of G protein coupling for many GPCRs (DeWire et al., 2007). This was shown to be the case for EP4 in colorectal cancer cells, where β-arrestin 1-dependent c-Src activation, and not G protein activation, is responsible for PGE2/EP4-mediated increased cellular migration and metastasis (Buchanan et al., 2006). Therefore, specific subsets of EP4 signaling are involved in the clinically relevant effects of PGE2.
Growing evidence suggests that GPCRs can exist in multiple active conformations (Kenakin, 2003). Different ligands can stabilize distinct active receptor conformations that are only permissive to a subset of the receptor's complete repertoire of behaviors, or activate distinct pathways with different potency and intrinsic activity (efficacy). Thus, individual agonists lead to differential and independent coupling of the receptor to different G proteins or intracellular effectors, a concept referred to as functional selectivity (Galandrin et al., 2007; Urban et al., 2007). To date, prostanoid receptor ligands are mainly characterized by their relative binding affinity to the different receptor subtypes as determined by radioligand competition binding assays (Abramovitz et al., 2000; Davis and Sharif, 2000). The evaluation of the functional pharmacology of EP4 receptor ligands is often limited to their ability to induce cAMP production (Wilson et al., 2004), because the development of these drugs occurred before other EP4 signaling pathways were characterized. Therefore, information about activation of other EP4 pathways by these compounds is still lacking, and systematic investigation of the effect of these ligands on other EP4 signaling pathways is warranted. Knowledge about biased agonism of these ligands will be relevant for their use as pharmacological tools, with potential clinical relevance (Buchanan et al., 2006).
In this study, we use three different BRET assays for functional pharmacological characterization of various EP4 receptor ligands in living HEK293 cells by evaluating the relative potency and intrinsic activity of these ligands on three distinct signaling pathways relevant for EP4 biology, namely, the activation of Gαi and Gαs subunits, and the recruitment of β-arrestin.
Materials and Methods
Reagents.
M&B28767 was a gift from sanofi-aventis (Dagenham Essex, UK); L-902688 was a gift from Merck Frosst (Pointe-Claire, QC, Canada); all other EP4 ligands were purchased from Cayman Chemical (Ann Arbor, MI).
Cell Culture and Transfections.
Human embryonic kidney (HEK) 293E cells (passage number ≈10–30; Invitrogen, Carlsbad, CA) were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (WisEnt, Rocklin, CA), 100 units/ml penicillin/streptomycin and 2 mM l-glutamine (Invitrogen), and 200 μg/ml G418. Transient transfections were performed in six-well or 10-cm dishes using the polyethylenimine (Polysciences, Warrington, PA) method (Boussif et al., 1995).
Plasmids.
The expression vectors containing human EP4 receptor and Gγ2 were obtained from the Missouri S&T cDNA Resource Center (Rolla, MO; www.cdna.org). EP4-YFP was constructed by ligating the PCR-amplified coding sequence of EP4 (lacking its stop codon) into the pGFP-N1-Topaze backbone of CXCR4-YFP (Issafras et al., 2002) at HindIII/AgeI sites. The mutations A55T, C124A, and M185V were inserted in Rluc to generate Rluc3, a variation of the Rluc8 mutant described by Loening et al. (2006). The GFP10-Epac-Rluc3 BRET2-cAMP biosensor was constructed in three steps. First, the green fluorescent protein variant GFP10 (Hamdan et al., 2006), the linker GSAGT-(Acc65I/HindIII)-KLPAT, and Rluc3 were inserted in pcDNA3.1/Zeo (Invitrogen). Part of the human Epac1 (residues 144–881) was then amplified by PCR, digested by Acc65I-HindIII, and inserted using the same restriction sites between GFP10-GSAGT and KLPAT-Rluc3 in pCDNA3.1/Zeo. Finally, the substitutions T781A and F782A were introduced by PCR to remove the Rap1 binding site of Epac1. YFP-Gβ1 expression vector was obtained from GFP10-Gβ1 (Galés et al., 2005) by replacing GFP10 with the coding sequence of eYFP (pIRES-eYFP; Clontech, Mountain View, CA). Plasmids encoding Gαi1-91Rluc and Rluc-β-arrestin 2 have been described previously (Perroy et al., 2003; Galés et al., 2006). All generated constructs were confirmed by sequencing.
BRET Measurement.
Transiently transfected HEK293 cells were seeded in 96-well, white, clear-bottom microplates (ViewPlate; PerkinElmer Life and Analytical Sciences, Waltham, MA) coated with poly(d-lysine) and left in culture for 24 h. Cells were washed once with PBS and the Rluc substrates coelenterazine h (for BRET1 experiments; NanoLight Technology, Pinetop, AZ) or coelenterazine 400A (for BRET2 experiments; Biotium, Hayward, CA) added at a final concentration of 5 μM to BRET buffer (PBS, 0.5 mM MgCl2, 0.1% glucose). BRET readings were collected using a Mithras LB940 plate reader (Berthold Technologies, Bad Wildbad, Germany) and MicroWin2000 software. BRET1 measurement between Rluc and YFP was obtained by sequential integration of the signals detected in the 460 to 500 nm (luciferase) and 510 to 550 nm (YFP) windows, whereas BRET2 readings between Rluc3 and GFP10 were collected by sequential integration of the signals detected in the 365 to 435 nm (Rluc3) and 505 to 525 nm (GFP10) windows. The BRET signal was calculated as the ratio of light emitted by acceptor (YFP or GFP10) over the light emitted by donor (Rluc or Rluc3). The values were corrected to net BRET by subtracting the background BRET signal obtained in cells transfected with Rluc (BRET1) or Rluc3 (BRET2) constructs alone. Ligands were incubated at room temperature for 3 (Gαi1), 10 (Epac), or 25 (β-arrestin) min, with the exception of kinetics studies. For experiments with GW627368X or PTX, cells were treated for 10 min at room temperature or 16 h at 37°C, respectively, with the inhibitor before agonist exposure. For kinetic analysis, ligands were injected using the Mithras LB940 injector 10 min after coelenterazine addition and BRET readings were collected at 1- (Gαi1), 17- (Epac), or 50-s (β-arrestin) intervals.
Radioligand Saturation Binding.
Transiently transfected HEK293 cells were incubated for 90 min at 4°C in PBS/0.5% bovine serum albumin (w/v) buffer with different concentrations of [3H]PGE2, in the presence or absence of 1000-fold excess unlabeled PGE2 to determine specific binding. Cells were washed three times with PBS/0.5% bovine serum albumin and lysed with 0.1 N NaOH/0.1% Triton X-100. Bound radioactivity was measured on cell lysates with a liquid scintillation counter.
Data Analysis.
Data from BRET assays were the mean of independent experiments each of which was performed in triplicate. Curve-fitting and statistical analysis was conducted by use of GraphPad Prism 4 software (GraphPad Software Inc., San Diego, CA). Statistical significance of the differences between more than two groups was calculated by one-way ANOVA, followed by Tukey's post-test.
Results
BRET Assay Allows Monitoring of EP4-Dependent Gαs Activation.
Intracellular cAMP accumulation after stimulation of adenylate cyclase (AC) is the best characterized EP4 signaling pathway. To monitor this Gαs-dependent pathway in living cells, we generated a BRET2 version of the CAMYEL Epac-based BRET sensor described previously (Jiang et al., 2007). The inactive cytosolic mutant form of human Epac1 was inserted between GFP10 and Rluc3, a mutant form of Renilla reniformis luciferase with increased stability and light output. Binding of cAMP to GFP10-Epac-Rluc3 induces a conformational change in Epac that results in a decrease of the BRET2 signal between the Rluc3 donor and the GFP10 acceptor. Stimulation of HEK293 cells coexpressing this Epac biosensor and human EP4 receptor with PGE2 lead to a strong concentration-dependent cAMP response (Fig. 1A), with an average EC50 of 0.3 nM (Table 1). Addition of forskolin (100 μM) to maximally stimulate AC led to a BRET2 variation greater than that obtained with PGE2, indicating that the maximal PGE2 response is within the dynamic range of the Epac sensor. Stimulation of Epac/EP4-transfected cells with the cell-permeable cAMP analog 8-bromo-cAMP also led to a sigmoidal concentration-response BRET2 variation (Fig. 1B); stimulation of cells with a saturating (1 μM) dose of PGE2 indicated that the maximal PGE2-induced response was within the linear range of the Epac sensor response. The cAMP levels induced by PGE2 stimulation rose rapidly and peaked after ∼5 to 10 min, after which they slowly decreased because of the action of phosphodiesterases and the desensitization of the receptor and cAMP production system (Fig. 1C). To ensure that the measured BRET2 signal was specific to EP4 receptor activation, we pretreated cells with the EP4-specific antagonist GW627368X; this resulted in abrogation of the PGE2-mediated cAMP level increase (Fig. 1D). A similar effect was observed when cells were transfected with Epac sensor in the absence of the EP4 receptor (Fig. 1D). To further ensure that other endogenously expressed prostanoid receptors were not activating the Epac sensor, we also treated cells that were transfected with the sensor, but not EP4, with either 1 μM PGD2 (the natural DP receptor ligand), PGF2α (the natural FP receptor ligand), carbaprostacyclin (a stable analog of PGI2 and agonist of IP receptor), or U-46619 (a TP receptor agonist). Average BRET variations of less than 3% of the basal signal were detected in all cases, as opposed to variations of 21.5% induced by PGE2 in cells cotransfected with EP4 (results not shown).
BRET Assay to Monitor Activated EP4-Mediated Gαi1β1 Rearrangement.
In addition to coupling to Gαs, EP4 can also couple to the PTX-sensitive Gαi/o family of G proteins, and we detected an increase in cAMP response by EP4 after treatment with PTX (Fig. 1D), as reported previously (Fujino and Regan, 2006). However, because this Gαi/o-dependent inhibition of AC is masked by the Gαs coupling to EP4, we used a recently developed BRET assay that allows monitoring of receptor activation-dependent structural rearrangements within the heterotrimeric Gαi1β1γ2 complex in live cells (Galés et al., 2006). Specifically, the Rluc donor was inserted within the loop connecting helices A and B of the Gαi helical domain, and the YFP acceptor was fused to the N terminus of Gβ1. Under basal conditions, a strong BRET signal between Gαi1-Rluc and YFP-Gβ1 was measured (Fig. 2A), indicative of the preassociation of Gαi1 and Gβ1. Kinetic analysis of the BRET signal upon stimulation of EP4 by PGE2, but not vehicle alone (Fig. 2A, inset), revealed a very rapid decrease in the BRET signal (within 1 s); this decrease was sustained for at least 3 min. This signal was dose-dependent, with an average EC50 of 4.08 nM (Table 1; Fig. 2B). No response was detected in cells pretreated with PTX or the EP4 antagonist GW627368X, or in cells in which empty vector was cotransfected in place of EP4 (Fig. 2C), demonstrating that the measured BRET variation specifically reflected EP4 receptor-mediated activation of Gαi1. To ascertain that other endogenously expressed prostanoid receptors were not implicated in Gαi1 activation, we also treated cells in which empty vector was cotransfected in place of EP4 with 1 μM PGD2, PGF2α, carbaprostacyclin, or U-46619. An average BRET variation of less than 2% of the basal level was detected after stimulation with either of these ligands (results not shown), as opposed to 41% variation induced by PGE2 when EP4 was cotransfected (Fig. 2C). The ligand-induced BRET variation probably reflects G protein activation based on the following observations. 1) It is completely abrogated by PTX; 2) it is conserved for a large panel of Gαi-coupled GPCRs (Galés et al., 2006); 3) BRET variations correspond to the signaling efficacy of α2-adrenergic receptor ligands (Galés et al., 2006); and 4) the potencies of the δ-opioid receptor agonist d-pen-2,5-enkephalin to induce BRET variation and to promote [35S]guanosine 5′-3-O-(thio)triphosphate (GTPγS) binding in cells transfected with δ-opioid receptors were very similar (Audet et al., 2008).
BRET Assay to Monitor Recruitment of β-Arrestin to Activated EP4 Receptor.
A third EP4 receptor signaling pathway, relevant to cancer, is the recruitment of β-arrestin to EP4. This was measured by following the ligand-induced BRET signal between Rluc-β-arrestin 2 and EP4-YFP receptors in live HEK293 cells. BRET-based measurement of β-arrestin recruitment to monitor receptor activation has been used for a multitude of GPCRs, and has also been used as a high-throughput screening method for the identification of ligands (Hamdan et al., 2005). Fusion of YFP at the C terminus of EP4 had no effect on the binding affinity of [3H]PGE2 for the receptor (Kd = 5.5 ± 3.1 nM for EP4-YFP and 5.0 ± 1.9 nM for EP4). BRET kinetics showed a rapid PGE2-induced recruitment of β-arrestin 2 to EP4 (Fig. 3A) that led to a prolonged interaction (over 30 min), representative of a class B receptor-β-arrestin interaction (Oakley et al., 2000). Exposure of cells to PGE2 resulted in a significant concentration-dependent increase in the BRET signal detected over basal level (Fig. 3B), with an average EC50 of 2.01 nM (Table 1). PGE2-induced recruitment of Rluc-β-arrestin 1 to EP4-YFP showed similar kinetics and potency as for β-arrestin 2 (data not shown). Pretreatment of cells with PTX had no effect on the PGE2-induced recruitment of β-arrestin 1 or 2 to EP4 (results not shown), suggesting that this process was not Gαi/o-dependent.
Potency of EP4 Ligands for Gαs, Gαi1, and β-Arrestin 2 Signaling Pathways.
Using the BRET assays described above, we compared the potencies of various EP4 ligands to induce Gαs, Gαi1, and β-arrestin 2 responses. The chemical structure and the binding affinity reported previously (in competition assays with [3H]PGE2 to recombinant EP4 receptors) of these ligands are shown in Table 2. In principle, it is problematic to interpret differences in ligand potencies between different assays systems, assessing different functional readouts. Although different potencies might indeed reflect agonist bias, they might also be due to different coupling efficacies of the target receptor to the respective effectors, or simply to differences in parameters that determine the sensitivity of a given assay. However, the relative comparison of EC50s obtained for a set of ligands with respect to a reference ligand allows us to draw conclusions on their respective set of relative potencies. Moreover, the reconstituted systems used in this study are similar with respect to the expressed quantities of the transfected receptor, β-arrestin, and Gαi1 (Supplemental Figs. 1 and 2; Supplemental Table 1). Therefore, as a prospective limitation, our study does not address physiological or pathophysiological conditions of primary cells, which may be different from our model system due to the availability of the respective signaling molecules.
Overall, the observed potency of all ligands for cAMP generation was greater than for β-arrestin recruitment, except for PGE1 alcohol, for which no significant difference of pEC50 values was observed between these two assays (Table 1; Fig. 4A). Likewise, potency of ligands in the Gαi1 assay was equal to or greater than their potency in β-arrestin recruitment, again with the exception of PGE1 alcohol, which was significantly more potent for recruitment of β-arrestin 2 than for Gαi1β1 rearrangement. Overall, the potencies that we obtained for cAMP production were slightly lower (approximately half a log) than those previously reported in the study by Wilson and colleagues (2004), possibly because of variations in receptor density. However, relative potencies (compared with PGE2) of all agonists tested by us match those reported by Wilson et al. (2004).
Comparing the rank orders of the set of EP4 agonists, we found that the rank orders of potency for Gαi1 and Gαs activation were similar: L-902688 ≥ PGE1 ≥ PGE2 = 11-deoxy PGE1 ≫16,16-dimethyl PGE2 = misoprostol = 17-phenyl-trinor PGE2 = M&B28767 ≫PGE1 alcohol = PGF2α = PGD2. The rank order for the recruitment of β-arrestin 2 differed slightly: L-902688 ≥ PGE1 = PGE2 = 11-deoxy PGE1 ≫16,16-dimethyl PGE2 ≥ misoprostol = 17-phenyl-trinor PGE2 = M&B28767 > PGE1 alcohol ≫PGF2α = PGD2. Thus, unlike for the G protein rank orders of potency, PGE1 was not superior to PGE2/11-deoxy PGE1 (Gαi1) or inferior to L-902688 (Gαs), misoprostol was inferior to 16,16-dimethyl-PGE2 but was not superior to PGE1 alcohol and PGE1 alcohol was clearly more potent than PGF2α/PGD2 (Table 1). We observed a reversal in the rank order of potencies between β-arrestin 2 and Gαi1 functional assays comparing PGE1 alcohol and either PGF2α, PGD2, or M&B28767 (Fig. 4A; Table 1). Other agonists showed important differences of relative potency between signaling pathways, but without showing reversal of relative potencies. For example, pEC50 values for cAMP production and Gαi1 activation were not significantly different for 16,16-dimethyl PGE2, but differed by more than a log for PGE2 (Fig. 4A); similarly, pEC50 values for cAMP production and β-arrestin recruitment were nearly identical for PGE1 alcohol, but differed by more than a log for L-902688, M&B28767, misoprostol, and PGF2α (Fig. 4, A and B; Table 1).
When comparing relative potency ratios, a bias for Gαi1 over Gαs signaling could be observed for all ligands compared with PGE2, with 16,16-dimethyl PGE2 being the most Gαi1-biased ligand (Fig. 5A). Six of the ten ligands had a relative bias toward β-arrestin recruitment over Gαs compared with PGE2, with PGE1 alcohol as the most β-arrestin-biased agonist (Fig. 5B). PGE1 alcohol was also the only ligand that had an arrestin over Gαi1 bias (Fig. 5C).
Intrinsic Activity of EP4 Ligands for Gαs, Gαi1, and β-Arrestin 2 Signaling Pathways.
We then compared the relative intrinsic activities of the different ligands in Gαs, Gαi1, and β-arrestin 2 signaling pathways (Table 1), using PGE2 as a reference “full agonist” set at 100%. Most ligands proved to be full agonists for the three pathways tested. One exception was PGD2, which was a partial agonist for recruitment of β-arrestin 2 (intrinsic activity of 77.9 ± 4.4%; Table 1, Fig. 6), but a full agonist for activation of Gαs and Gαi1 pathways (93.2 ± 4.2 and 96.1 ± 1.6%, respectively). Inversely, L-902688 had (slightly) lower intrinsic activity in cAMP production and the Gαi1 pathway (94.4 ± 2.6 and 89.9 ± 2%, respectively), while being full agonist for β-arrestin (102.6 ± 1.1%).
Discussion
During the past years, it became increasingly clear that G protein-coupled receptors can activate independently a variety of signaling effectors, and that distinct receptor ligands can do so with different potencies and efficacies (intrinsic activities). This selective activation of independent pathways by ligands has been termed functional selectivity (Urban et al., 2007). The conceptual basis for this is that GPCRs do not have merely “inactive” and “active” conformations but that ligands can stabilize distinct receptor conformations, which are more or less potent and efficient in activating a given readout (Kenakin, 2003, 2007). The significance of functional selectivity obviously raises the question of whether previously reported properties of synthetic or natural ligands were based on the “right” (that is, clinically relevant) readout, especially for compounds that have therapeutic use. Moreover, for most clinical contexts the relevant signaling pathway of any given receptor is often still elusive (Bosier and Hermans, 2007). Experiments using synthetic ligands with known functional selectivity profiles in animal models will be required to pin down the clinically relevant receptor signaling pathways, and might also identify drug candidates or leads for further development. However, functional selectivity profiles of drugs and synthetic ligands are only beginning to be identified (Galandrin and Bouvier, 2006; Audet et al., 2008; Gao and Jacobson, 2008; Masri et al., 2008). Therefore, reassessment of ligand-induced signaling activity of known GPCR ligands is warranted, taking functional selectivity into account.
Classically, the prostaglandin receptor EP4 had been classified as coupling to the Gαs subunit as an effector. However, in recent years, EP4 signaling has revealed more complexity and has been shown to also involve coupling to PTX-sensitive Gαi proteins and β-arrestin-mediated effects. It is noteworthy that β-arrestin recruitment to EP4 rather than Gαs-mediated signaling has been associated with colorectal cancer progression (Buchanan et al., 2006). In addition, the Gαi pathway could also be implicated in the cAMP/protein kinase A-independent role of EP4-mediated increased phosphorylation of PI3K and extracellular signal-regulated kinase 1/2 and early growth response factor-1 induction in colon cancer (Pozzi et al., 2004; Cherukuri et al., 2007). Our study set out to systematically characterize the response of a wide range of EP4 agonists for the downstream effectors Gαs, Gαi1, and β-arrestin, thereby addressing functional selectivity.
By adapting three different BRET-based assays, we monitored receptor signaling in real time in live cells. β-Arrestin recruitment was assessed by direct interaction between tagged EP4 and β-arrestin 2 proteins. Activation of Gαs and Gαi pathways was determined by measuring BRET signal variations resulting from structural rearrangements within Epac and heterotrimeric Gαi1β1γ2 proteins, respectively. These energy transfer variations were practically absent in cells treated with the EP4-specific antagonist GW627368X or in cells that were not transfected with EP4, indicating that BRET modulations were indeed specific to the EP4 receptor and did not result from the activation of endogenously expressed prostanoid receptors.
Our results suggest considerable functional selectivity among the tested, structurally related agonists. All compounds showed a relative bias toward Gαi1 over Gαs activation compared with PGE2, this effect being the strongest with 16,16-dimethyl PGE2, which activated both α subunits with almost equal potency (while PGE2 activates Gαs with 10-fold potency over Gαi1) (Figs. 4A and 5A; Table 1). Likewise, the profiles of six (PGE1 alcohol, 16,16-dimethyl PGE2, PGE1, PGD2, 17-phenyl-trinor PGE2, and 11-deoxy PGE1) of the ten analogs show relatively more potent activation of β-arrestin recruitment than Gαs responses relative to PGE2. Comparing the relative potency of a ligand to induce each of the three measured signaling pathways, PGE2 was the most selective in activating Gαs, whereas PGF2α and PGE1 alcohol were the most selective for activating Gαi1 and β-arrestin, respectively (Fig. 5). When all tested pathways were considered, the profiles of PGE1 alcohol and 16,16-dimethyl PGE2 were most distinct from PGE2 (Fig. 4B). PGE1 alcohol activated both Gαi1 and β-arrestin relatively more potently than Gαs, compared with PGE2, whereas 16,16-dimethyl PGE2 was a relatively weak activator of β-arrestin recruitment, but a relatively better activator of Gαi1 than PGE2 (Fig. 5). The order of potency for PGE1 alcohol and PGF2α, PGD2, or M&B28767 was reversed for β-arrestin recruitment and Gαi1 activation: PGE1 alcohol was more potent for the β-arrestin pathway than for the Gαi1 pathway, whereas the opposite was true for PGF2α, PGD2, and M&B28767 (Fig. 4A; Table 1). This reversal of potency ratio for these signaling pathways is incompatible with a single-receptor active state, but in line with the existence of ligand-specific receptor states that result in differential activation of signaling pathways, as suggested by the concept of functional selectivity (Urban et al., 2007).
We observed that most agonists tested had full intrinsic activity in all pathways, as reported earlier for cAMP production (Wilson et al., 2004). An exception was L-902688 that had slightly but significantly reduced efficacy in activation of both tested G proteins. PGD2 was also less efficacious in β-arrestin recruitment as detected by BRET. It is important to consider that maximal responses from BRET assays may relate to either more interacting molecules or to closer donor-acceptor distances. In this case, the lower intrinsic activity for β-arrestin/EP4 BRET induced by PGD2 compared with the reference ligand PGE2 thus suggests that either less arrestin molecules are recruited to the receptor-fused energy acceptor or the energy transfer is less efficient in the receptor/arrestin complex induced by PGD2 because of distinct conformational changes. Indeed, recently reported results obtained with the angiotensin receptor AT1aR, the β2-adrenergic receptor, and the parathyroid hormone receptor type 1, revealed different arrestin conformations depending on the ligand that was applied (Shukla et al., 2008). Accordingly, it might be expected that such different arrestin conformations also translate into different BRET signals resulting from the receptor-arrestin interaction. We interpret the lower efficacy of L-902688 in the Gαi1/Gβ1 BRET as a distinct conformational rearrangement in the Gαi1β1γ2 heterotrimer. Indeed, the PGE2-induced BRET decrease between Gαi1-Rluc and YFP-β1 could reflect dissociation of αi1 from the β1γ2 dimer, but overall results obtained by Galés et al. (2006) strongly suggest that this BRET variation rather corresponds to Gαi1β1γ2 structural rearrangements within a stable heterotrimer that results in G protein activation. This is not true for the cAMP Epac BRET sensor, however, where signal intensity only relates to cAMP levels and not to different conformations of the sensor.
The paucity of differences in intrinsic activity between ligands is in line with previous reports on the efficacies of these ligand in cAMP production (Wilson et al., 2004), but somewhat surprising when considering results obtained in other receptor systems. Indeed, large differences in β-arrestin recruitment efficacy (also measured by BRET) were observed in studies with synthetic agonists of the dopamine D2 receptor (Klewe et al., 2008) or natural peptide agonists of the glucagon-like peptide receptor (Jorgensen et al., 2007). We speculate that the relative stability in intrinsic efficacy that we observed might be linked to the structural similarity of the ligands used here (Table 2). In principle, differences in intrinsic ligand activity might also be masked by receptor overexpression, leading to increased receptor reserve and underestimation of the maximal response. This clearly is not the case for the Epac-based cAMP assay where the maximal response lies within the dynamic response range as shown with a saturating concentration of forskolin. In the intermolecular BRET systems between Gαi1β1 or EP4-β-arrestin, in turn, maximal response cannot be evaluated, because they may depend on both quantitative and qualitative differences (see above).
The activation of the Gαi1and β-arrestin recruitment pathways by EP4 has been reported much later than its responses in the classical Gαs pathway. Accordingly, our study is the first to systematically report these pathways for a series of EP4 agonists. When these compounds are used as tools in experiments designed to further dissect the biological roles of EP4, for example, in cancer genesis and progression, the finding that a substantial number of the tested ligands are relatively stronger activators of these noncanonical EP4 pathways than the natural ligand PGE2 must be taken into account.
Taken together, our study is the first to systematically characterize the response of a set of EP4 agonists for the downstream effectors β-arrestin and Gαi1, using BRET-based methodology that should be applicable for the study of other GPCRs. We find significant functional selectivity among the studied ligands. Although more work is required to examine the bearing of our observations in a more complex native context, including the presence of other PGE2 receptors, our EP4-limited study is the first step for assessing the consequences of functional selectivity in physiology and drug treatment.
Acknowledgments
We thank Hendrika Fernandez for expert technical assistance.
Footnotes
-
↵ The online version of this article (available at http://jpet.aspetjournals.org) contains supplemental material.
-
This work was supported by the Canadian Institutes of Health Research [Grant CTP-79848].
-
M.L. was supported by studentships from the Heart and Stroke Foundation of Canada and Fondation de l’Hôpital Sainte-Justine.
-
S.C. and M.B. hold Canada Research Chairs in Perinatology and in Signal Transduction and Molecular Pharmacology, respectively.
-
N.H. is a Canadian Institutes of Health Research New Investigator.
-
M.L., B.B., C.L.G., M.B., S.C., and N.H. are members of the Canadian Institutes of Health Research Team in GPCR Allosteric Regulation.
-
Article, publication date, and citation information can be found at http://jpet.aspetjournals.org.
-
ABBREVIATIONS:
-
↵1 Current affiliation: Institut National de la Santé et de la Recherche Médicale, Toulouse, France.
- Received May 19, 2009.
- Accepted July 6, 2009.
- © 2009 by the American Society for Pharmacology and Experimental Therapeutics