Dimerization of the Human Receptors for Prostacyclin and Thromboxane Facilitates Thromboxane Receptor-mediated cAMP Generation*

Prostacyclin (PGI2) and thromboxane (TxA2) are biological opposites; PGI2, a vasodilator and inhibitor of platelet aggregation, limits the deleterious actions of TxA2, a vasoconstrictor and platelet activator. The molecular mechanisms involved in the counterregulation of PGI2/TxA2 signaling are unclear. We examined the interaction of the receptors for PGI2 (IP) and TxA2 (TPα). IP-induced cAMP and TP-induced inositol phosphate generation were unaltered when the receptors were co-expressed in HEK 293 cells (IP/TPα-HEK). TP-cAMP generation, in response to TP agonists or a TP-dependent isoprostane, iPE2III, was evident in IP/TPα-HEK and in aortic smooth muscle cells, but not in cells expressing either receptor alone, or in IP-deficient aortic smooth muscle cells. Augmentation of TP-induced cAMP generation, with the IP agonist cicaprost, was ablated in IP-deficient cells and was independent of direct IP signaling. IP/TPα heterodimers were formed constitutively when the receptors were co-expressed, with no overt changes in ligand binding to the individual receptor sites. However, despite inefficient binding of iPE2III to either the IP or TPα, expressed alone or in combination, robust cAMP generation was evident in IP/TPα-HEK, suggesting the formation of an alternative receptor site. Thus, IP/TPα dimerization was coincident with TP-cAMP generation, promoting a “PGI2-like” cellular response to TP activation. This represents a previously unknown mechanism by which IP may limit the cellular effects of TP.

PGI 2 1 and TxA 2 are the predominant products of cyclooxygenase (COX) metabolism of arachidonic acid formed in the macrovascular endothelium and platelets, respectively (1,2). These two mediators are biological opposites. PGI 2 , a potent vasodilator, inhibitor of platelet aggregation (3) and smooth muscle cell (SMC) growth in vitro (4), demonstrates antithrombotic and anti-platelet actions in vivo (5). In contrast, TxA 2 is a potent vasoconstrictor (6), stimulates platelet aggregation (7), amplifies the activity of other platelet agonists (7), and stimulates proliferation of SMC (8). The hypothesis that PGI 2 modulates cardiovascular homeostasis and disease gained support from the association of a selective COX-2 inhibitor, rofecoxib, which depresses PGI 2 levels without affecting platelet TxA 2 biosynthesis, with a higher risk of myocardial infarction in humans, compared with a nonselective COX-1/ COX-2 inhibitor naproxen (VIGOR trial) (9). Recent work, using mice genetically deficient in the receptors for PGI 2 (the IP) or TxA 2 (the TP), demonstrated that the proliferative and platelet response to vascular injury was TP-mediated and was limited specifically by PGI 2 (5). In addition, delivery of PGI 2 synthase in vivo prevents proliferation and migration of SMC, key features of restenosis and atherosclerosis (10,11), whereas the antioxidant and antiplatelet actions of PGI 2 delayed atherogenesis and may underlie the protection from cardiovascular disease afforded by female gender (12). Maintenance of the PGI 2 /TxA 2 balance appears to be a critical regulator of vascular disease; however, the molecular mechanisms underlying the counterregulation of PGI 2 /TxA 2 signaling have not been fully elucidated.
A single gene encoding a G protein-coupled receptor (GPCR) has been reported for both mediators (13,14), although in contrast to IP, where splice variants have not been described, two variants of TP, termed TP␣ and TP␤, have been identified (15). IP is coupled to at least two signaling systems, namely the generation of intracellular cAMP and activation of PLC (16). Both TP␣ and TP␤ are coupled to PLC, whereas the former may activate and the latter may inhibit AC activity (16). There is substantial evidence for reciprocal regulation between IP and TP. TP␣, but not TP␤, is a target for IP-mediated, PKAdependent phosphorylation, resulting in TP␣ desensitization (17). Similarly, U46619-mediated activation of TP enhances IP-mediated cAMP generation in human platelets (18), where only TP␣ is expressed (19).
The interaction between IP and TP may not, however, be limited to events occurring secondary to activation of their respective second messenger systems. GPCRs have long been considered to exist and function as independent monomeric units. However, GPCRs from both closely related and distinct subfamilies are capable of interacting physically with one another to form heterodimers (20,21). Far from being a benign association, GPCR heterodimerization can substantially modify receptor function (20,21). Signaling may change as a result of altered agonist affinity for the receptors, altered affinity of the receptors for their respective G proteins or signaling via alternate pathways (21). Heterodimerization of the ␦and -opioid receptors, for example, creates a "new" receptor binding site that has a reduced affinity for individual ␦orselective ligands but that can ligate cooperatively selective agonists to induce synergistic functional responses (22).
Heterodimers of the angiotensin II AT1 receptor and the * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ To whom correspondence should be addressed: Center for Experimental Therapeutics, University of Pennsylvania, 808 BRB II/III, 421 Curie Blvd., Philadelphia, PA 19104. Tel.: 215-573-2323; Fax: 215-573-9004; E-mail: emer@spirit.gcrc.upenn.edu. 1 The abbreviations used are: PGI 2 , prostacyclin; COX, cyclooxygenase; TxA 2 , thromboxane; SMC, smooth muscle cell; TP, receptor for TxA 2 ; IP, receptor for PGI 2 ; GPCR, G protein-coupled receptor; hIP, human IP; hTP, human TP; ASMC, aortic smooth muscle cell(s); hASMC, human ASMC; mASMC, mouse ASMC; WT, wild type; IPKO, IP knockout; PKA, protein kinase A; 8-Br-cAMP, 8-bromo-cyclic AMP. bradykinin B 2 receptor demonstrate increased activation of AT1-coupled G proteins in response to angiotensin II, a phenomenon that may underlie preeclamptic hypertension (23). Thus, two opposing vascular mediators, angiotensin II, a vasoconstrictor, and bradykinin, a vasodilator, can alter the action of the other via a direct interaction of their receptors (23). In the present study, we examined whether the interaction between two similarly opposing vascular mediators, PGI 2 and TxA 2 , is also mediated via receptor interaction. We demonstrate that coexpression of IP and TP␣, either endogenously or in an overexpression cell model, facilitated TP-mediated cAMP generation. The absence of the IP, in SMCs cultured from IP knockout (IPKO) mice or in HEK 293 cells, rendered the TP largely inactive with regard to AC activity. This interaction between IP and TP␣ is not dependent on IP-cAMP signaling, but is coincident with the formation of an IP/TP␣ heterodimer.

EXPERIMENTAL PROCEDURES
Materials-Cyclic AMP radioimmunoassay kit, enhanced chemiluminescence kits, protein G-Sepharose, and all radiochemicals were purchased from Amersham Biosciences. Cell culture reagents, G418, and Albumax were obtained from Invitrogen. Complete protease inhibitor tablets were obtained from Roche Applied Science. IBOP, SQ 29548, iPE 2 III, and iPF 2␣ III were purchased from Cayman Chemical Co. (Ann Arbor, MI). H89 was obtained from Calbiochem. Isobutylmethylxanthine and deoxycholic acid were purchased from Sigma. Monoclonal anti-HA and anti-Myc were obtained from Covance (Richmond, CA). NuPAGE gels and buffers were purchased from Invitrogen. Secondary antibodies were purchased from Jackson Immunoresearch (West Grove, PA). Cicaprost was obtained from Schering AG under agreement.
Epitope Tagging of hIP and hTP-The 9-amino acid hemagglutinin epitope (HA; YPYDVPDYA) or 10-amino acid Myc epitope (EQKLI-SEEDL) was inserted between the N-terminal initiator methionine and the second amino acid of the hTP␣ or hIP to generate HAhIP, MychIP, or HAhTP␣. Generation of HAhIP was as described previously (24). To generate HAhTP␣ and MychIP, 5Ј-oligonucleotides that contained 3 miscellaneous bases, 6 bases encoding a HindIII site, the 3 miscellaneous bases immediately 5Ј of the initiator methionine, 3 bases encoding a methionine, the epitope tag coding sequence, and 21 bases encoding amino acids 2-8 were generated. 3Ј-Oligonucleotides were complementary to the receptor coding sequence downstream of a unique restriction site (an EcoN47 site for hIP or a NotI site for hTP␣). Using the hTP␣ or hIP cDNAs as templates, polymerase chain reactions were carried out to generate the 5Ј-HAhTP and 5Ј-MychIP fragments. The resulting products were cloned into PCR 2.1 (Qiagen, CA) and, following verification of the sequence, were excised using HindIII/NotI or HindIII/ EcoN47, as appropriate. Using the same enzymes, the 3Ј fragment in pcDNA 3.1 (or pcDNA3.1 Hygro for MychIP) was generated, and the two receptor pieces were ligated to each other. The integrity of the splice site was verified by sequencing.
Membrane Preparation and Radioligand Binding-Membranes were prepared from confluent 100-mm dishes as described previously (24). Radioligand binding studies were carried out using membrane proteins Reactions were allowed to continue for 30 min at 30°C and terminated by the addition of 3 ml of ice-cold wash buffer (10 mM HEPES, pH 7.4, 0.01% bovine serum albumin), followed by immediate filtration through GF/C filters that had been thoroughly soaked in the same ice-cold buffer. Following one wash with ice-cold wash buffer, radioactivity associated with the filters was quan- tified by scintillation counting. Nonspecific binding was measured in the presence of a 500-fold excess of unlabeled iloprost or SQ 29548, respectively. Saturation binding data were analyzed using GraphPad Prism 3.0 to calculate K d and B max and to compare one-and two-site curve-fitting models (partial F-test).
cAMP Measurements-Cells were grown to confluence in 12-or 6-well plates. hASMC and mASMC were pretreated overnight with 3 M indomethacin to inhibit endogenous eicosanoid generation, followed by the addition of medium containing isobutylmethylxanthine (0.01 M) 30 min prior to agonist treatment. Cells were treated, and reactions were terminated by aspiration. cAMP was extracted with ice-cold 65% ethanol for 30 min. Samples were dried under vacuum and reconstituted in assay buffer, and cAMP was quantified by radioimmunoassay, as described previously (24).
Inositol Phosphate Production-Inositol phosphate production was assessed, as described previously (24). Briefly, cells were labeled overnight with 2 Ci/ml [ 3 H]myoinositol. Thirty minutes prior to stimulation, cells were treated with 20 mM LiCl at 37°C. After stimulation for 10 min at 37°C, the reactions were terminated by aspiration. Total inositol phosphates were extracted with formic acid for 30 min at room temperature and neutralized using 5 M ammonia. Total inositol phosphates were recovered by anion exchange using Dowex 1-X8 AG anion exchange resin.
Co-immunoprecipitation-All immunoprecipitation procedures were carried out at 4°C. HEK 293 cells stably expressing hTP␣ or hIP or both were treated with 3 mM dithiobis(succinimidylpropionate) (Pierce) for 30 min and lysed in buffer A (150 mM NaCl, 1 mM EDTA, 20 mM Tris-HCl (pH 8.0), 10% glycerol, and a mixture of protease inhibitors) for 2 h at 4°C. The resulting supernatants were precleared by a 1-h rotation with 100 l of 10% (w/v) protein G-Sepharose to each tube. Anti-Myc-or anti-HA-protein G-Sepharose was prepared by adding 9 g of anti-Myc ascites per lysate to 10% protein G-Sepharose followed by a 1-h rotation. MychIP or HAhTP␣ was immunoprecipitated from precleared lysates by adding 150 l of anti-Myc-or anti-HA-protein G-Sepharose to each lysate and rotating for 16 h. Protein G was precipitated at 14,000 rpm for 1 min, washed three times with Buffer A, and resuspended in 10 l of sample buffer (Nupage). Immunoblotting for HA or Myc was carried out as described above, using the appropriate biotinylated antibody (1:500) followed by peroxidase-labeled streptavidin.

Generation of HEK 293 Cell Lines Expressing HA-hTP␣ or
Coexpressing Both HA-hTP␣ and MychIP-We have described previously the generation of a HEK 293 cell line stably expressing HA-tagged hIP (hIP-HEK; see Ref. 24). Cell lines stably expressing HAhTP␣ (TP␣-HEK), or coexpressing both MychIP and HAhTP␣ (IP/TP␣-HEK), were generated in the present study in order to examine the effect of IP/TP␣ coexpression on TP␣ signaling.
Lysates from each cell line were resolved by SDS-PAGE and immunoblotted with an anti-HA or anti-Myc antibody to establish that the receptors were being appropriately expressed. HAhTP␣ was observed as a broad complex of between 48 and 64 kDa in lysates derived from TP␣-HEK or IP/TP␣-HEK cells ( Fig. 1). MychIP was observed as a 44 -60-kDa complex in IP/TP␣-HEK cell lysates (Fig. 1). This corresponds to the molecular weight of HAhIP previously observed in hIP-HEK cells (24). The broad molecular weight range of both hIP and hTP␣ is a result of receptor glycosylation (19,24). Stimulation of hIP-HEK (see Ref. 24) or IP/TP␣-HEK with the prostacyclin analogue cicaprost for 5 min induced a concentration-dependent increase in intracellular cAMP (EC 50 ϭ 0.05 Ϯ 0.02 nM, n ϭ 4; Fig. 2) and inositol phosphate production (EC 50 ϭ 97.1 Ϯ 33.3 nM, n ϭ 4; Fig. 2), indicating that co-expression of hTP␣ did not alter hIP-mediated activation of two signaling systems when overexpressed in HEK 293 cells (24). Stimulation of TP␣-HEK with the specific TPagonist U46619 resulted in increased inositol phosphate production (EC 50 ϭ 174.4 Ϯ 65.2 nM, n ϭ 3; Fig. 3), which was not altered significantly by co-expression of hIP (EC 50 ϭ 231.7 Ϯ 40.8 nM, n ϭ 3; Fig. 3). Much evidence supports ligation of the TP by isoprostanes, free radical-catalyzed products of arachidonic acid. Indeed, two TP-dependent iso- prostanes (27), iPE 2 III (Fig. 3), and iPF 2␣ III (data not shown) stimulated inositol phosphate generation in TP␣-HEK and IP/TP␣-HEK, albeit with significantly higher EC 50 values (3.56 Ϯ 1.12 and 4.43 Ϯ 0.63 M, respectively) compared with U46619. Thus, similar to other studies, the addition of HA or Myc tags to the N terminus of IP and TP did not alter the expression or signal transduction properties of the receptor. Furthermore, co-expression of the receptors did not alter their discrete signal transduction properties.
TP␣-mediated cAMP Formation-Treatment with the TP agonists IBOP or U46619 (100 nM, 10 min) induced a robust increase in cAMP levels in cells coexpressing HAhTP␣ and MychIP, but not in cells individually expressing the receptors or in mixed cultures of individually expressing cells (Fig. 4A). Pretreatment with the TP antagonist SQ 29548 partially reduced signaling by the TP agonists (Fig. 4B). Interestingly, iPE 2 III, but not iPF 2␣ III, also initiated an increase in cAMP in IP/TP␣-HEK cells (Fig. 4A). In contrast to the TP agonists, the activity of iPE 2 III was insensitive to SQ 29548 (Fig. 4B), suggesting that this event was TP-independent. However, signaling was not observed in the absence of hTP␣, demonstrating that inhibition with SQ 29548 is not sufficient to determine TP dependence (Figs. 4A and 5). Generation of cAMP in response to treatment of IP/TP␣-HEK with IBOP, U46619, or iPE 2 III proved concentration-dependent, and only minor cAMP increments were observed at the highest concentration of TP agonist in cells expressing TP␣ alone (Fig. 5). These results indicate that cAMP formation in response to TP activation by TP agonists or iPE 2 III is dependent on the presence of both TP␣ and IP. We examined the biological relevance of this relationship in a cell model that endogenously expresses both IP and TP. cAMP production in response to IBOP or iPE 2 III was quantified in aortic smooth muscle cells isolated from humans or WT mice. Increased cAMP levels were observed in hASMC or WT mASMC following a 10-min treatment with IBOP or iPE 2 III (Fig. 6). SQ 29548 was partially effective against IBOP and ineffective against iPE 2 III (Fig. 6, E and F), in agreement with the HEK 293 cell data (Fig. 4B). IBOP or iPE 2 III treatment of ASMC isolated from IPKO mice resulted in minimal cAMP generation (Fig. 6, C and D). Thus, the absence of the IP in ASMC cultured from IPKO mice or in HEK 293 cells (Fig. 5) uncoupled the TP from activation of AC.

Effect of IP Coactivation on TP-mediated cAMP Generation-
Activation of IP in IP/TP␣ co-expressing cells, with a submaximal concentration of cicaprost (0.02 nM, 5 min), resulted in a synergistic enhancement of cAMP generation in response to the TP agonist U46619 and to iPE 2 III (Fig. 7). Furthermore, activation of IP synergistically enhanced iPE 2 III-induced cAMP generation in both hASMC and mASMC (Fig. 8, B and D). In contrast, whereas IP activation in IP/TP␣-HEK cells synergistically enhanced cAMP formation in response to 100 -500 nM IBOP, this enhancement became additive at 1000 nM IBOP. IP activation was similarly additive with IBOP (1000 and 5000 nM) in both hASMC and mASMC. The absence of IP in mASMC (cultured from IPKO mice) ablated both synergistic and additive effects (Fig. 8, E and F). Thus, activation of the IP enhanced TP-mediated cAMP generation beyond that seen when IP was physically present but not activated.
TP-mediated cAMP Generation, Role of IP-induced PKA-Interaction between IP and TP signaling pathways has been described previously; prostacyclin-induced desensitization of hTP␣ is mediated by PKA phosphorylation of the TP␣ C-ter-minal tail. To determine whether PKA is also involved in IP-dependent, TP-induced cAMP generation, IP/TP␣-HEK cells were pretreated with the PKA inhibitor H89 (10 M, 30 min) prior to cotreatment with cicaprost followed by U46619, IBOP, or iPE 2 III, as described above. Inhibition of PKA resulted in a slight increase in the overall cellular cAMP levels but had no impact on the synergistic interaction resulting from IP/TP coactivation (Fig. 9). The increase in basal cAMP levels in response to H89 is unsurprising, since PKA is capable of activating a number of cAMP phosphodiesterases (28). Similarly, pretreatment with H89 did not inhibit cAMP production in response to IBOP or iPE 2 III (1000 or 5000 nM, 10 min) or the potentiation of TP-induced cAMP by cicaprost (0.5 nM, 5 min) in hASMC (Fig. 10). These data indicate that PKA does not play a role in IP-dependent, TP-mediated cAMP generation.
IP-dependent Modulation of TP Signaling, Role of IP-derived cAMP-We next sought to determine the relative necessity of the physical presence of IP versus the increased cAMP tone in IP-expressing cells. ASMC from IPKO mice were pretreated with 8-Br-cAMP (500 nM, 10 min), elevating cellular cAMP to levels similar to those present in cicaprost-treated WT cells (Fig. 11). However, correction of the cAMP deficit in this manner did not restore cAMP generation in response to IBOP or iPE 2 III (1000 or 5000 nM, 10 min, Fig. 11). Similarly, when HEK 293 cells expressing TP␣ alone were treated with 8-Br-cAMP (100 nM, 10 min) to boost the intracellular cAMP to approximately the same level as IP/TP␣-HEK cells treated with cicaprost, this did not restore TP-mediated cAMP generation in response to U46619, IBOP, or iPE 2 III (Fig. 12). Thus, TP-mediated cAMP generation did not occur secondary to IPinduced cAMP formation was instead dependent on, and enhanced by, the physical presence and activation of the IP.
Formation of IP/TP␣ Heterodimers-TP-mediated cAMP generation was dependent on the physical presence of both TP␣ and IP. Many studies have demonstrated that G-protein-coupled receptor heterodimer formation, typically a constitutive process occurring when both partners are co-expressed, often results in alterations in the signaling of constituent receptors (20). Thus, we sought to determine whether coexpression of MychIP and HAhTP␣ resulted in heterodimer formation. My-chIP was immunoprecipitated from IP/TP␣-HEK cell lysates. Immunoblotting with anti-HA revealed the presence of HAhTP␣ in lysates derived from cells co-expressing HAhTP␣ and MychIP but not in lysates from cells expressing HAhTP␣ (Fig. 13). Similarly, immunoprecipitation of HAhTP␣ resulted in the co-immunoprecipitation of MychIP (Fig. 13). In either case, the co-immunoprecipitated partner appeared primarily in the monomeric rather than in the oligomeric form (Fig. 13). This is not unexpected, since reducing conditions were used to disrupt protein complexes that were stabilized, prior to immunoprecipitation, with a cross-linker (dithiobis(succinimidylpropionate)). Furthermore, these conditions disrupt disulfide linkages, important for the formation of many GPCR dimers (22), possibly contributing further to the lower molecular weight. The presence of MychIP in HAhTP␣ immunoprecipitates and vice versa indicated that coexpression of these receptors results in the formation of an IP/TP␣ heterodimer.
Receptor-Ligand Interactions-It has been demonstrated previously that receptor heterodimerization can alter the affinity of the individual receptors for their specific ligands (22,29,30). We examined if changes in ligand binding in the IP/TP␣ heterodimer could underlie the changes we observed in TP␣ signaling. Saturation binding using the TP-specific antagonist [ 3 H]SQ 29548 (Fig. 14) revealed the presence of a single high affinity binding site in membranes from TP␣-HEK (K d ϭ 34.4 Ϯ 6.7 nM; B max ϭ 4.8 Ϯ 0.3 pmol/mg; n ϭ 4). Co-expression of hIP did not alter significantly the affinity (K d ϭ 29.2 Ϯ 6.1 nM, n ϭ 5) but did reduce the number of [ 3 H]SQ 29548 binding sites (B max ϭ 1.5 Ϯ 0.1 pmol/mg; n ϭ 5). High and low affinity binding sites for the IP agonist [ 3 H]iloprost were observed in membranes derived from IP/TP␣-HEK cells, similar to those reported previously for IP-HEK (24) (data not shown).
Similarly, coexpression of hIP and hTP␣ did not alter the affinity of TP␣ for the specific TP ligands SQ 29548, IBOP, and U46619 in displacement analysis (Table I). As expected, iPF 2␣ III (31) did not bind to the TP␣, and this was unaltered when IP was co-expressed. Interestingly, despite its ability to induce robust cAMP signaling in IP/TP␣ coexpressors (Fig. 5), iPE 2 III bound very weakly to TP␣, whether IP was co-expressed or not (Table I). The affinity of hIP for cicaprost was not altered by coexpression with TP␣, and no binding of TP-specific ligands or isoprostanes to hIP was observed. Thus, coexpression of IP and TP␣ reduced the number of TP binding sites but did not result in obvious alterations in binding affinities for the individual receptor sites. In addition, iPE 2 III activity was dependent on the presence of both TP␣ and IP, although this ligand did not bind efficiently to either of the individual receptor sites. This is consistent with the generation of a modified binding site generated through the physical association of TP␣ with IP.
Effect of SQ 29548 on TP-induced cAMP Formation-When the TP site was blocked with SQ 29548, IP-induced TP-mediated cAMP generation in IP/TP␣-HEK cells was partially reduced (Fig. 15). In contrast, iPE 2 III-induced cAMP generation, in IP-activated cells, was also prevented with SQ 29548, although this isoprostane did not displace efficiently [ 3 H]SQ 29548 binding (Table I). Thus, although iPE 2 III ligated an altered binding site in the IP/TP␣ complex, subsequent signaling events were TP-mediated, in the HEK 293 cell model. In a similar fashion, IBOP-induced cAMP generation in cicaproststimulated hASMC was reduced by antagonism of the TP (Fig.  15). However, in contrast, the iPE 2 III response was unaltered in hASMC treated with SQ 29548. This inconsistency between the HEK 293 cell model and the hASMC, with regard to the activity of iPE 2 III, suggests a greater level of complexity in the native cell model. DISCUSSION PGI 2 and TxA 2 are important regulators of vascular homeostasis, and their respective levels dictate the response to vascular injury. Recently, IP was shown to limit specifically the deleterious effects of TP activation during the response to vascular injury (5). Given the importance of the PGI 2 /TxA 2 balance for vascular function, their coincident biosynthesis in vascular disease (32,33), and the frequent co-expression of IP and TP in vascular cells (16), we sought to examine their relationship at the molecular level. We concentrated our efforts on the hTP␣, since this isoform is expressed more ubiquitously and abundantly compared with hTP␤ (34).
Studies have demonstrated that, whereas both TP␣ and TP␤ are similarly coupled to G q and activation of PLC, they oppositely regulate adenylate cyclase; TP␣ increases cAMP formation via G s , whereas TP␤ couples to G i and inhibition of cAMP formation in CHO cells overexpressing the individual receptors (35). Similar to previous studies using TP␣ transfected HEK 293 cells (36), we observed an approximate 2-fold increase in cAMP upon treatment of TP␣-HEK cells with the TP agonists U46619 or IBOP. Strikingly, when IP was co-expressed, the same treatments elicited a 15-fold (U46619) and 30-fold (IBOP) increase in cAMP generation (Fig. 4). No induction of cAMP generation was observed in HEK 293 cells expressing IP alone or in mixed cultures of singly expressing cells, following treatment with IBOP or U46619 under identical conditions, demonstrating that this effect was not mediated by nonspecific agonist interaction (Fig. 4). Thus, the physical presence of IP dramatically enhanced cAMP generation in response to TP␣ activation in HEK 293 cells. We addressed the biological relevance of this phenomenon using primary smooth muscle cells that express both IP and TP␣ endogenously. TP activation led to cAMP formation in both human and mouse ASMC (Fig. 6). However, TP activation had a negligible effect on cAMP production in ASMC isolated from IPKO mice (Fig. 6, C and D). Thus, in agreement with the HEK 293 cell data, activation of The theoretical additive effect of IP and TP activation was also calculated (gray bar; IP activation alone ϩ TP activation alone Ϫ basal). cAMP was quantified as described under "Experimental Procedures." Data are presented as the mean pmol of cAMP/well Ϯ S.E. from 3-6 experiments, each performed in duplicate. *, p Ͻ 0.05 relative to theoretical additive. ***, p Ͻ 0.001 relative to theoretical additive. cAMP formation via the TP␣ required the presence of the IP.
Intriguingly, this phenomenon extended to the isoprostane, iPE 2 III. Isoprostanes are free radical catalyzed products of arachidonic acid that are increased in syndromes of vascular disease and act via TP in vivo (27). Isoprostane levels are elevated within developing atherosclerotic lesions in mice, and normalization of isoprostane levels correlates with disease regression (37). Furthermore, antagonism of TP, but not inhibition of TxA 2 synthesis with aspirin, reduced atherosclerosis in mice, suggesting that mediators other than TxA 2 , possibly isoprostanes, act at the TP to propagate the disease (38). Thus, the isoprostanes are both a marker and a mediator of disease.
We used a well characterized TP antagonist, SQ 29548, to block the TP receptor. The activity of IBOP was partially inhibited by SQ 29548 in IP/TP␣-HEK and hASMC (Figs. 4B and 6E). However, iPE 2 III activity was unaffected in both cell models. At first glance, this suggests that iPE 2 III is not acting through the TP. However, iPE 2 III did not signal in the absence of the TP in both our study (Fig. 5) and in mice (27), demonstrating its TP-dependent action. Indeed, our data suggest that sensitivity to SQ 29548 does not reliably define TP dependence.
Recently, the importance of IP as a specific limit on the deleterious effects of TP following vascular injury was demonstrated in vivo (5). Deletion of the IP exacerbated TP-dependent SMC proliferation following arterial injury in mice (5). In this setting, TxA 2 and isoprostane biosynthesis is increased as a result of platelet activation and increased oxidant stress. The ability of TP agonists and iPE 2 III to induce TP-dependent cAMP formation only when IP is physically present may represent a previously unappreciated mechanism by which IP can regulate TP activity. Thus, IP may co-opt TP to generate cAMP, increasing the cell's PGI 2 -like response.
Having established a relationship between IP-and TP-induced cAMP formation, we next wanted to examine whether activation of IP might modify this effect. Stimulation of the IP receptor resulted in a synergistic potentiation of TP-mediated cAMP generation in IP/TP␣-coexpressing HEK 293 cells treated with U46619 or iPE 2 III (Fig. 7). Synergistic enhancement of iPE 2 III-induced cAMP generation, upon IP activation, was also observed in mASMC and hASMC, demonstrating that this effect was not an artifact of receptor overexpression (Fig.  8). In contrast, whereas low concentrations of IBOP were syn- ergistically potentiated by IP activation, the interaction became additive at 1000 nM IBOP (Fig. 7). Similarly, IP activation was additive with IBOP-induced cAMP formation in both mASMC and hASMC (Fig. 8). A previous study in DAMI and CHRF cell lines, representing middle and late stage megakaryocyte maturation, also observed a synergistic induction of cAMP formation following coactivation of IP and TP (39). Thus, the net generation of cAMP in response to TP activation was maximized when IP was present and activated.
Receptor phosphorylation is an important mechanism via which the signaling of one receptor can modify that of another. Indeed, TP␣ desensitization, in response to IP activation, occurs via PKA-mediated phosphorylation of serine 329 on the C-terminal tail of TP␣ (17). In addition, GPCR phosphorylation by PKA is capable of altering the affinity of receptors for their respective G-protein (40). PKA-mediated phosphorylation of the ␤ 2 -adrenergic receptor decreases the affinity of the receptor for G s while increasing its affinity for G i (40). We hypothesized that in the presence of IP, PKA activity and the subsequent phosphorylation of TP␣ might result in increased affinity of TP␣ for G s . However, H89, a selective inhibitor of PKA, did not alter the synergistic interaction between IP and TP in IP/TP␣-HEK cells (Fig. 9) or in hASMC (Fig. 10). Furthermore, receptor independent activation of PKA with 8-Br-cAMP did not induce the TP-cAMP response. Thus, unlike IP-mediated TP␣ desensitization, IP-mediated potentiation of TP␣-cAMP formation is independent of PKA phosphorylation.
We next examined whether increased cellular cAMP, which occurred as a consequence of IP expression and activation, facilitated IP-dependent, TP-mediated cAMP generation. Cellular cAMP levels in TP␣-HEK cells, in which TP did not couple efficiently to adenylyl cyclase (Fig. 5), were elevated using 8-Br-cAMP to approximately the same level as those present in cicaprost-treated IP/TP␣-HEK cells. However, TP-induced cAMP generation was not reconstituted (Fig. 12). Similarly, ASMC isolated from IPKO mice were treated with 8-Br-cAMP, elevating cellular cAMP to the levels found in cicaprost-treated WT ASMC. Despite the correction of the cAMP deficiency in IPKO mASMC, IBOP-or iPE 2 III-induced cAMP formation was not restored (Fig. 11). It appears, therefore, that although activation of IP enhanced TP-induced cAMP, this was independent of signal transduction events resulting directly from IP activation. It may be argued that 8-Br-cAMP, a cell-permeable analogue of cAMP, does not perfectly reproduce the level of complexity, with respect to spatial and temporal distribution, inherent in IP-expressing cells. However, 8-Br-cAMP has been shown to activate the same signal transduction pathways as cAMP (41) and as such is routinely used to mimic the effects of endogenously generated cAMP (41,42). Furthermore, the internal consistency between the H89 and 8-Br-cAMP data strengthens the argument that a novel IP-dependent event mediates the change in TP signaling.
We were struck by the fact that the physical presence of IP, whether activated or not, allowed TP-mediated cAMP generation. Reports now abound on the formation of GPCR heterodimers, which are formed upon receptor co-expression. Moreover, it has recently been demonstrated that hIP is capable of homodimerization (43). Thus, we examined the possibility that IP and TP␣ might heterodimerize with consequent alterations in TP␣ signaling. Differential co-immunoprecipitation of lysates, derived from cells that coexpress both MychIP and HAhTP␣, revealed the presence of an IP/TP␣ complex (Fig.  13). Prostanoid receptor heterodimers have not been described previously, yet their existence is unsurprising given the number of GPCRs from widely divergent families that have been shown to heterodimerize (20). Whereas the presence of IP/TP␣ heterodimers in IP/TP␣-HEK 293 cells does not constitute definitive proof of their involvement in the alterations in TP␣ signaling, the dependence of TP␣-mediated cAMP formation on the physical presence of IP, in overexpressing and native cell models, strongly suggests biologically relevant heterodimerization between these two receptors. In addition, data indicating the formation of an alternative ligand binding site (see below), a frequent consequence of GPCR dimerization (22,29,30), is consistent with the formation of an IP/TP␣ complex, as is the synergistic augmentation of TP-cAMP generation following preactivation of IP (22).
The mechanism through which IP/TP␣ heterodimers couple to cAMP formation is unclear. Dimerization may alter the conformation of TP␣, thereby increasing its affinity for Gs. Indeed, formation of a heterodimer between the angiotensin AT1 receptor and the bradykinin receptor results in increased activation of AT1-associated G-proteins (23). Alternatively, TP␣ may "borrow" the signal transduction machinery of its dimeric partner IP, thereby allowing it to more efficiently signal via G s . This is not without precedent; an elegant study by Rocheville et al. (44) demonstrated that heterodimerization of the dopamine D1 receptor, with a C-terminal mutant of the somatostatin sst5 receptor, which was unable to couple to AC, restored the ability of the mutant sst5 receptors to activate the cyclase.
It is possible that the formation of an IP/TP␣ heterodimer generates a new receptor with signaling and binding characteristics distinct from the individual partners. Indeed, a number of studies investigating the signaling properties of newly discovered GPCR heterodimers describe an altered receptorligand-effector profile (22,29,30). Using SQ 29548 to label the TP␣ and iloprost to label the IP, we found no alterations in the binding affinities of specific IP or TP ligands to their individual receptor sites, suggesting preservation of the binding characteristics (Table I). It is interesting that expression of IP reduced the number of TP binding sites (Fig. 13), suggesting that, similar to other GPCRs (20,45,46), the interaction of these receptors may impact on membrane expression and receptor trafficking, considerations that are now under investigation. Despite the apparent status quo for ligand binding, concentra-FIG. 13. Detection of IP/TP␣ heterodimers by co-immunoprecipitation. IP/TP␣-HEK or TP␣-HEK cell lysates were subjected to co-immunoprecipitation using an anti-Myc or anti-HA antibody as described under "Experimental Procedures." Immunoprecipitates were resolved by 10% reducing SDS-PAGE and co-immunoprecipitated HAhTP␣ or MychIP was detected using an anti-HA or an anti-Myc antibody, respectively. Co-immunoprecipitation was only observed when MychIP and HAhTP␣ were coexpressed. Molecular masses are in kDa. The arrows indicate co-immunoprecipitated monomer and oligomers. Western blots are representative of three independent experiments. tions greater than 1000 nM iPE 2 III were required to displace SQ 29548 from TP␣, yet treatment of IP/TP␣-HEK 293 cells with 100 nM iPE 2 III induced a robust increase in cAMP, consistent with ligation of iPE 2 III at an alternative binding site following IP/TP␣ association. Furthermore, SQ 29548 did not inhibit either iPE 2 III-induced cAMP generation in IP/TP␣-HEK (Fig. 4B), or the synergistic interaction between cicaprost and iPE 2 III in hASMC (Fig. 15). These observations suggest that a distinct, albeit related, binding site was generated following the association of IP with TP␣. Intriguingly, the coactivation of cAMP generation by cicaprost and iPE 2 III in IP/TP␣ co-expressing HEK cells was inhibited with SQ 29548, suggesting greater complexity in the native cells following IP/TP␣ association (Fig. 15). Indeed, it is likely that in native cells, complex associations between GPCRs and other interacting proteins may further refine receptor pharmacology in a cellspecific manner.
A definitive model for the domains involved in TP␣ receptorligand binding does not exist. Previous studies have variously espoused the critical importance of transmembrane regions I, III, IV, V, VI, and VII as well as cysteine residues in extracellular domains 2 and 3 (47)(48)(49). Thus, it appears that a large portion of the extracellular surface of the receptor may be involved in forming the binding pocket, whereas the ligands themselves are no more than 15-25 Å in size. Indeed, subtle changes in TP binding have been reported previously. Funk et al. (50) generated a mutant form of TP␣, which did not recognize SQ 29548, but bound two other TP receptor agonists with unaltered affinity. In addition, Chiang and Tai (51) reported that mutations in TM5 and TM6 of TP␣ resulted in altered receptor affinity for IBOP but had no effect on the binding of SQ 29548. Both theoretical analysis and experimental evidence suggest a domain-swapped model of dimerization, which would result in alterations in the conformation of both receptors with subsequent changes in ligand binding. Thus, it is likely that IP/TP␣ heterodimer formation results in an alteration of receptor conformation and the generation of a distinct but related binding site. This model of eicosanoid receptor heterodimerization may explain how isoprostanes act in a manner that appears TP-dependent (27) without obvious binding to the TP site (Table I) and (31) and inconsistent responses to TP antagonists ( Fig. 5) (15).
In summary, our findings reveal a previously unknown level of interaction between IP and TP␣. The presence of IP facilitates TP␣-mediated generation of cAMP in a manner independent of IP-induced cAMP formation, and subsequent PKA activation, but coincident with the formation of an IP/TP␣ heterodimer. Thus, a novel mechanism for the specific limit IP imposes on the deleterious effects of TP␣ emerges, namely that IP promotes a "PGI 2 -like" cellular response to TxA 2 and isoprostanes. Indeed, a recent study in apolipoprotein E knockout mice demonstrated that COX-2 inhibition had no effect on atherosclerotic lesion formation despite selective reduction in PGI 2 generation (52), whereas IPKO mice exhibit increased lesion formation when compared with their wild type counterparts (12). Thus, in agreement with the present study, it appears that the physical presence of IP plays a distinct role in limiting the injurious actions of TxA 2 in vivo. Biosynthesis of PGI 2 , TxA 2 , and isoprostanes are increased during vascular disease. Our study demonstrates that the association of IP with TP␣ would maximize the "PGI 2 -like" response to TP activation under these conditions, thus limiting the deleterious effects of TxA 2 and coincident TP ligands.