Biased suppression of TP homodimerization and signaling through disruption of a TM GxxxGxxxL helical interaction motif.

Thromboxane A2 (TXA2) contributes to cardiovascular disease (CVD) by activating platelets and vascular constriction and proliferation. Despite their preclinical efficacy, pharmacological antagonists of the TXA2 receptor (TP), a G protein-coupled receptor, have not been clinically successful, raising interest in novel approaches to modifying TP function. We determined that disruption of a GxxxGxxxL helical interaction motif in the human TP's (α isoform) fifth transmembrane (TM) domain suppressed TP agonist-induced Gq signaling and TPα homodimerization, but not its cell surface expression, ligand affinity, or Gq association. Heterodimerization of TPα with the functionally opposing prostacyclin receptor (IP) shifts TPα to signal via the IP-Gs cascade contributing to prostacyclin's restraint of TXA2 function. Interestingly, disruption of the TPα-TM5 GxxxGxxxL motif did not modify either IP-TPα heterodimerization or its Gs-cAMP signaling. Our study indicates that distinct regions of the TPα receptor direct its homo- and heterodimerization and that homodimerization is necessary for normal TPα-Gq activation. Targeting the TPα-TM5 GxxxGxxxL domain may allow development of biased TPα homodimer antagonists that avoid suppression of IP-TPα heterodimer function. Such novel therapeutics may prove superior in CVD compared with nonselective suppression of all TP functions with TXA2 biosynthesis inhibitors or TP antagonists.

Thromboxane A 2 (TXA 2 ) is generated by thromboxane synthase metabolism of prostaglandin H 2 , the immediate product of cyclooxygenase (COX) action on arachidonic acid (1)(2)(3). Platelet COX-1, the only COX isoform expressed in mature platelets, is the dominant source of TXA 2 synthesis under normal conditions ( 4 ). Other cells, including macrophages and monocytes, contribute to TXA 2 generation via both COX-1 and COX-2 with the latter isozyme being particularly relevant during infl ammation ( 2,5 ). TXA 2 acts as a local autocrine or paracrine mediator to mediate a range of physiological and pathophysiological responses that include platelet activation, vasoconstriction, and smooth muscle cell proliferation ( 3,(6)(7)(8)(9)(10). These processes are of particular relevance to cardiovascular disease (CVD) in which TXA 2 generation is markedly elevated and expression of its receptor, the TXA 2 receptor (TP), is increased (11)(12)(13). In humans, inhibition of platelet COX-1 with low-dose aspirin is widely used for prevention of heart attack and stroke (14)(15)(16)(17), while in mouse models of atherogenesis and injury-induced vascular proliferation or remodeling, disease severity was blunted by antagonism or deletion of the TP ( 8,18,19 ). Interestingly, in hyperlipidemic mice the TP antagonist was more effective in reducing atherogenesis than COX inhibition ( 20 ). This may reflect antagonism of COX-independent TP ligands, such as the isoprostanes, free radical-derived metabolites of arachidonic acid that can activate the TP in vivo ( 21 ). These, and other studies, have placed signifi cant emphasis on the TP as a therapeutic target in CVD ( 8,20,22 ). Despite their potential, however, pharmacological antagonists of the TP have been Abstract Thromboxane A 2 (TXA 2 ) contributes to cardiovascular disease (CVD) by activating platelets and vascular constriction and proliferation. Despite their preclinical effi cacy, pharmacological antagonists of the TXA 2 receptor (TP), a G protein-coupled receptor, have not been clinically successful, raising interest in novel approaches to modifying TP function. We determined that disruption of a GxxxGxxxL helical interaction motif in the human TP's ( ␣ isoform) fi fth transmembrane (TM) domain suppressed TP agonist-induced Gq signaling and TP ␣ homodimerization, but not its cell surface expression, ligand affi nity, or Gq association. Heterodimerization of TP ␣ with the functionally opposing prostacyclin receptor (IP) shifts TP ␣ to signal via the IP-Gs cascade contributing to prostacyclin's restraint of TXA 2 function. Interestingly, disruption of the TP ␣ -TM5 GxxxGxxxL motif did not modify either IP-TP ␣ heterodimerization or its Gs-cAMP signaling. Our study indicates that distinct regions of the TP ␣ receptor direct its homo-and heterodimerization and that homodimerization is necessary for normal TP ␣ -Gq activation. Targeting  Consequently, there is signifi cant interest targeting TM domains to modulate the function of membrane-spanning proteins, including GPCRs ( 39,41,(54)(55)(56)(57). Among TM domains, the GxxxG motif, in which two glycines are separated by any three other residues, is strongly over-represented ( 58 ), highly conserved across species ( 59,60 ), and can direct homologous or heterologous helical interactions ( 52,53,61 ). Neighboring residues, especially the large aliphatic residues isoleucine, valine, and leucine, appear critical to GxxxG-mediated helix-helix interactions ( 59,60 ). In a number of proteins (62)(63)(64)(65), including at least two GPCRs ( 42,66 ), placement of a leucine three residues after the second glycine to create a GxxxGxxxL motif, directs for protein-protein interaction and function. We identifi ed a GxxxGxxxL motif within the fi fth transmembrane of the TP ␣ and examined its relevance for TP function. We determined that mutation of the three critical amino acids G 205 , G 209 , and L 213 dramatically reduced TP ␣ signaling via the Gq-InosP cascade without altering the receptor's cell surface expression, ligand recognition, or Gq association. TP ␣ homodimerization was, however, signifi cantly impaired, suggesting the normal homodimerization of the TP ␣ is necessary for signal transduction. Strikingly, neither TP ␣ -IP heterodimerization nor signaling of the heterodimer to Gs-cAMP in response to TP agonists was impacted by mutation of the TP-TM5 GxxxGxxxL motif indicating the specifi city of this motif for TP ␣ homodimerization. We suggest that targeting the TP-TM5 GxxxGxxxL motif may allow selective suppression of the TP ␣ homodimer signal without altering TP ␣ -IP heterodimer function.

Constructs
Hemagglutinin (HA)-tagged human IP and TP ␣ cloned into the mammalian expression vector pcDNA3 (Invitrogen, CA) were as described previously ( 48 ). QuikChange site-directed mutagenesis (Stragagene, CA) was used to replace G 205 and G 209 with leucines, a small-to-large replacement that disrupts helix-helix interaction ( 62,64,65 ). We replaced L 213 with a tyrosine based on the studies the GxxxGxxxL motif in the ␤ 2-adrenergic receptor ( 42 ). The resulting mutant was termed TP L205,L209,Y213 . HA-tagged IP, TP ␣ , and TP L205,L209,Y213 were fused at their C termini to either Renilla luciferase (rLuc) or yellow fl uorescent protein (YFP), previously described ( 67 ). Briefl y, the stop codon was removed by PCR and each stopless construct cloned into prLuc-N3(h) (Perkin Elmer, MA) and pEYFP-N1 (Clontech, CA) plasmids in frame with the fusion protein start site. All sequences were verifi ed by DNA sequencing.

Cell culture and transfection
Cell lines were from the American Type Tissue Culture Collection (Rockville, MD). HEK 293 cells were maintained as described previously ( 48 ); Meg-01 cells were grown in RPMI-1640 (Invitrogen) containing 10% fetal bovine serum and 1% penicillin-streptomycin. Transient transfections were performed using FuGENE 6 (Roche Applied Science, IN) for HEK 293 cells (2 g total DNA), as previously described ( 48 ) or for Meg-01s (3 g total DNA) by nucleofection using an Amaxa Nucleofector TM II and Nucleofector TM Kit C (Lonza, NJ) per the manufacturer's instructions.
clinically disappointing compared with low-dose aspirin, in large part because none replicate aspirin's irreversible inhibitory effect on platelets (22)(23)(24)(25). Most recently, the TP receptor antagonist terutroban showed comparable, but not superior, effi cacy with aspirin in preventing recurrent ischemic stroke in clinical trials ( 26 ).
The TP is a cell surface G protein-coupled receptor (GPCR) that is expressed in a wide variety of tissues and cells including platelets, smooth muscle cells, endothelial cells, lungs, kidneys, heart, thymus, and spleen (27)(28)(29). In humans, but not in other species, there are two splice variants, the TP ␣ and TP ␤ , which, despite reported differences in their upstream promoter use, posttranslational modifi cations, interacting proteins, and agonist-induced regulation, do not display signifi cant physiological or pathophysiological divergence ( 22 ). A number of tissues express both variants ( 30 ), although TP ␣ is the only isoform expressed in platelets ( 31 ). Research from our group and others has defi ned the TP's functional and regulatory pathways (31)(32)(33)(34)(35)(36). Signaling via the TP can be transduced through multiple G proteins with Gq and G12/13, which stimulate respectively the phospholipase-C pathway of inositol phosphate (InosP)/ intracellular calcium elevation and RhoA, appearing most relevant its biological actions ( 37,38 ). Agonist activation of the TP leads to its internalization and degradation, although sustained agonist activation can increase TP stability and surface expression driving TP responsiveness ( 32 ).
Across the GPCR superfamily, there is substantial evidence for receptor dimerization ( 39,40 ) and a signifi cant contribution therein to receptor traffi cking and ligand recognition, signaling, and regulation (41)(42)(43)(44). We reported that the TP forms dimeric or oligomeric receptor complexes (45)(46)(47)(48). In addition to homodimerization, TP ␣ can heterodimerize with TP ␤ leading to enhanced isoprostane responsiveness ( 47 ). Further, we observed equal propensity for TP ␣ to heterodimerize with the receptor for prostacyclin (PGI 2 ) ( 48 ). A predominantly COX-2-derived mediator, PGI 2 acts via its receptor, the prostacyclin receptor (IP), to activate the Gs-adenylyl cyclase signaling pathway causing vasodilation and inhibition of platelet activation ( 49 ). In mice, the restraint placed by the PGI 2 -IP system on TXA 2 -TP function limits the proliferative and platelet response to vascular injury ( 18 ) demonstrating the in vivo relevance of this interplay. Further, the elevated cardiovascular hazard in patients treated with COX-2 inhibitors can be explained by selective suppression of COX-2-derived PGI 2 without alteration of COX-1-derived TXA 2 levels ( 50 ). We determined that heterodimerization of the TP with the IP contributes to the PGI 2 -TXA 2 interplay; dimerization with the IP dramatically shifts TP function from a lipid raft-excluded Gq-coupled receptor to a raft-associated Gs-coupled receptor that yields a robust Gs-cAMP response, concomitant with reduced Gq-InosP signaling to TP agonists ( 45,48 ). Loss of this shift in TP function in individuals heterozygous for a signaling-defi cient IP mutant, IP R212C , may contribute to their accelerated CVD ( 51 ).
The importance of transmembrane (TM) helical interactions to protein structure and function is evident across multiple diverse integral membrane protein families ( 52,53 ). anti-rabbit IgG (Jackson ImmunoResearch, PA; 1:10,000) and visualized by enhanced chemiluminescence (ECL Plus, GE Healthcare/ Amersham, NJ). Quantifi cation of proteins was by densitometry.

Receptor modeling
The human hTP ␣ sequence was aligned with solved crystal structures, bovine rhodopsin (OPSD, UniProt P02699) and the human ␤ 2-adrenergic receptor (ADRB2, UniProt P0755) in ClustalW. Both the PAM250 and BLOSUM algorithms indicated closer alignment of hTP ␣ to align more closely with OPSD (similarity score 30.16) than with ADRB2 (33.48). Each bundle of seven transmembrane ␣ -helices was therefore based on a 2.8 Å crystallographic bovine rhodopsin template (1HZX) ( 68 ) to using the internet-based protein-modeling server, SWISS-MODEL (GlaxoSmithKline, Geneva, Switzerland), and energy minimized using the Gromos96 force fi eld in DeepView. Extracellular and cytoplasmic loop regions were manually constructed, built according to JPred consensus, and energy minimized using the NAMD molecular dynamics simulator.

Statistical analysis
Data were analyzed using GraphPad Prism software. Comparisons were made using a one-sample t -test or by ANOVA suitable post hoc multiple comparison testing as appropriate.

A GxxxGxxxL motif is located in the fi fth transmembrane of the TP ␣
Analysis of the TP ␣ amino acid sequence revealed a GxxxGxxxL motif in TM5, G 205 LSVG 209 LSFL 213 ( Fig. 1A ). DNA levels were equalized in all transfections using empty pcDNA3 vector. Assays were performed 48 h after transfection.

Bioluminescence resonance energy transfer assay
Dimerization of rLuc and YFP fused receptors was examined by measuring bioluminescence resonance energy transfer (BRET) from a donor (rLuc-fused) receptor to an acceptor (YFP-fused) receptor following addition of substrate for rLuc (coelenterazine H; Molecular Probes, Life Technologies, NY). In BRET saturation experiments cells were transfected with a fi xed amount of rLuc receptor (0.25 g) together with increasing amounts of YFP receptor (0.125-1.75 g). In BRET competition assays increasing amounts of HA-tagged competitor receptor were cotransfected together with a fi xed ratio (1:7) of receptor-rLuc + receptor-YFP. BRET measurements were performed essentially as described previously ( 45 ). Briefl y, cells were harvested (phenol red-free Hank's Balanced Salt Solution containing 0.02% EDTA), redistributed in 96-well plates (black, clear; 100,000 cells/well) and maintained at 37°C. Total YFP (Ex485 nm, Em555 nm) was fi rst collected using a luminescence multi-plate reader (VICTOR3, Perkin Elmer) and calculated as fold over basal. Using the same plate, donor (485 nm) and acceptor (555 nm) emissions were gathered sequentially from each well, following addition of coelenterazine H (5 M in Ca 2+ -free phosphate-buffered saline rested for 30 min at room temperature before use). Milli BRET units (mBu) were calculated as the ratio of Em555 over Em485 nm corrected for cells expressing the rLuc receptor alone, and multiplied by 1,000.

Cell surface expression of the TP
HEK 293 and Meg-01 cells were transfected with HA-tagged wild-type TP (TP WT ) or TP L205,L209,Y213 . Cells were harvested into ice-cold FACS buffer (DPBS containing 1% BSA and 0.1% sodium azide). Cell suspensions were stained with anti-HA mouse IgG1 (monoclonal 16B12) conjugated to Alexa Fluor® 488 (Invitrogen, CA) for 30 min prior to washing. Median fl uorescence intensity (MFI) was collected using a fl ow cytometer as a measure of cell surface expression.

Measurement of second messenger generation
Measurement of intracellular IP1 or cAMP was performed using the IP-One Tb kit (Cisbio Bioassays, MA) or LANCE cAMP 384 kit (Perkin Elmer, MA), respectively, according to the manufacturer's instructions. Cells were stimulated with or without the TP agonist U46619 (Cayman Chemicals, MI) for 1 h.

Immunoprecipitation and immunoblotting
HA-tagged TP WT or TP L205,L209,Y213 were immunoprecipitated from transfected HEK 293 cells using the Pierce HA Tag IP/ Co-IP kit (Thermo, IL), according the manufacturer's instructions. Eluted proteins were resolved via NuPAGE electrophoresis (Invitrogen, CA) under reducing conditions. HA-tagged TP WT or TP L205,L209,Y213 were visualized with anti-TP (Cayman Chemicals, MI; 1:100) while immunoprecipitated Gq␣ was visualized with anti-Gq/11␣ (Millipore, CA; 1:1000). Antigen-antibody complexes were revealed using horseradish peroxidase-conjugated Additional GxxxG motifs were identifi ed in N terminus (G 5 SSLG 9 ), fi rst intracellular loop (G 51 ARQG 55 ), and second extracellular loop (G 188 AESG 192 ). Given that a TM GxxxGxxxL motif has been implicated in the function of at least two GPCRs ( 42,66 ), we chose to focus further on the G 205 LSVG 209 LSFL 213 domain. Three-dimensional homology modeling of the TP revealed an outward-facing orientation of G 205 , G 209 , and L 213 ( Fig. 1B ) in TM5 indicating that this domain is appropriately positioned for protein-protein interaction within the membrane. To defi ne the functional relevance of the TM5 GxxxGxxxL motif in the TP we employed site-directed mutagenesis to replace G 205 and G 209 with leucines and L 213 with a tyrosine to generate TP L205,L209,Y213 .

Disruption of the TM5 GGL motif suppressed TP function and but did not alter receptor surface expression
We fi rst measured the ability of the TP WT and TP L205,L209,Y213 to transduce a signal via the phospholipase C/InosP pathway in response to the thromboxane mimetic U46619. In transiently transfected HEK 293 cells, the maximal signaling capacity of TP L205,L209,Y213 was signifi cantly reduced by ‫ف‬ 25% compared with TP WT transfected cells, although there was no signifi cant change in EC 50 ( Fig. 2A ). Depressed signaling via the TP L205,L209,Y213 was also evident in Meg-01 cells ( Fig. 2B ), which are platelet-like cells that serve as a closer approximation of the TP's normal physiological environment, with an ‫ف‬ 50% reduction in InosP generation and a signifi cant rightward EC 50 shift. Thus, disruption of the TM5 GxxxGxxxL motif markedly suppressed TP response to agonist.
We examined whether this loss of receptor responsiveness refl ected simply reduced cell surface expression of the mutant receptor. Cell surface expression of the TP WT or TP L205,L209,Y213 , both tagged at their N terminus with the HA epitope tag, was examined by fl ow cytometry in transfected HEK 293 or Meg-01 cells. No signifi cant difference in cell surface receptor levels, as measured by median surface HA fluorescence intensity, was apparent between TP WT and TP L205,L209,Y213 transfectants in either cell type ( Fig. 3 ). Thus, disruption of the TM5 GxxxGxxxL motif did not substantially modify receptor processing to the surface, indicating that the signaling defi cit we observed for agonist ( 69,70 ). In displacement analyses, we detected no change in the K i for either of the TP agonists U46619 or IBOP, arguing against dissociation of the TP L205,L209,Y213 from Gq. Further, comparable levels of Gq coimmunoprecipitated could not be explained by quantitative changes in the receptor population on the plasma membrane.

Ligand affi nity and Gq association are not modifi ed by mutation of the TM5 GxxxGxxxL motif
We considered whether suppressed agonist-induced signal transduction in TP L205,L209,Y213 refl ected a change in ligand binding leading to reduced agonist affi nity. Intact HEK 293 cells expressing either TP WT or TP L205,L209,Y213 were labeled with a single concentration of 3 H-SQ 29,548 and displacement examined for two TP agonists, U46619 (K i = 90 nM for TP WT vs. 52 nM for TP L205,L209,Y213 ) and IBOP (K i = 1.8 nM for TP WT vs. 2.5 nM for TP L205,L209,Y213 ), or by unlabeled SQ 29,548 (K i = 4 nM for both TP WT and TP L205,L209,Y213 ) as a reference. No signifi cant difference in displacement was evident between the WT and mutant receptors. We also examined an isoprostane, iPE 2 III (K i = 334 nM for TP WT vs. 403 nM for TP L205,L209,Y213 ), a free radical-generated metabolite of arachidonic acid that can activate the TP in vivo ( 21 ), and again saw no difference in radioligand displacement ( Fig 4 ). Thus, disruption of the TM5 GxxxGxxxL motif did not alter the receptor ligand binding properties.
We considered also whether disruption of the TM5 GxxxGxxxL motif interferes with the association of the TP ␣ to its effector, Gq, leading to suppressed signaling. As for other GPCRs, association of the G protein with the TP in the inactive conformation provides a high affi nity state  homodimerization plays in TP ␣ expression and function also remains unclear. However across GPCR studies, one or more TMs have been frequently implicated in dimer formation and function ( 54,55,60 ). Given the outwardfacing orientation of the TP-TM5 GxxxGxxxL motif, thus positioned for intermolecular protein interaction, we examined whether homodimerization was modifi ed in the TP L205,L209,Y213 mutant. TP WT and TP L205,L209,Y213 were fused at their C termini to either rLuc (energy donor) or YFP (energy acceptor) and energy transfer quantifi ed as a measure of dimerization. In saturation experiments, expression of the donor-tagged receptor is held steady and expression with either HA-tagged TP WT or HA-tagged TP L205,L209,Y213 in HEK 293 transfectants ( Fig 5 ). Taken together these analyses indicate that mutation of the TM5 GxxxGxxxL motif in TP ␣ allows normal formation of the high affi nity receptor-Gq complex at the cell surface.

Disruption of the TM5 GxxxGxxxL motif modifi es TP homodimerization
We reported previously that, similar to other GPCRs, the TP physically associates to form homodimers ( 32,47,48 ). The molecular determinants of TP ␣ homodimerization have not been defi ned; similarly, the precise role or function. These data support the concept that distinct molecular interactions drive the physical association of the TP ␣ -TP ␣ and IP-TP ␣ dimers and their downstream signaling.

A TM GxxxGxxxL motif is found in numerous class A GPCRs
Given that a TM GxxxGxxxL motif was functionally relevant in at least two other GPCRs, the ␤ 2-adrenoreceptor and the ␣ -factor yeast receptor ( 42,66 ), we searched the SwissProt database (http://prosite.expasy.org/scanprosite/) for human GxxxGxxxL-containing GPCRs. Sixtynine receptors were identifi ed of which, after removal of olfactory (24 hits), taste (2 hits), and orphan (9 hits) receptors, 22 GPCRs were identifi ed that contain one or more TM GxxxGxxxL motifs ( Table 1 ). Interestingly, all but one of these 22 were class A GPCRs suggesting a particular prevalence of this motif among rhodopsin-like GPCRs.

DISCUSSION
Protein-protein interactions are ubiquitous to biological processes and are vital for signaling complex assembly. Compared with soluble protein regions, relatively little is known about the interaction of membrane embedded proteins within lipid bilayers, although there is substantial and increasing interest in therapeutic targeting of TM interactions ( 61 ). GPCRs are characterized by their seven transmembrane spanning regions, which are capable of intra-as well as intermolecular interactions that defi ne tertiary and quaternary receptor structure and function. The GxxxG interaction motif, fi rst described in homodimerization of the single TM sialoglycoprotein glycophorin A (GpA), has been identifi ed as a dominant TM motif across diverse protein families ( 52,58 ). In GpA, as in other transmembrane proteins, residues that neighbor the GxxxG domain appear critical and are thought to provide a three-dimensional structure within the helix creating the protein-protein interface. In one particular subclass, termed "glycine zip pers," a small residue (glycine, alanine, or serine) is located three positions before or after the GxxxG motif ( 59 ). More generally, large residues (isoleucine, valine, or leucine) are commonly found within one or two positions of the GxxxG pair ( 58 ), forming a groove (the glycines) and ridges (the large residues) arrangement. In the case of the TP ␣ -TM5 GxxxGxxxL motif we determined a similar arrangement with a groove created by S 201 G 205 G 209 and a ridge created by leucines 203, 206, 210, and 213 ( Fig. 8 ). The positioning of L 213 three residues after the GxxxG pair serves to align the GGL triplet along the same ␣ -helix face ( Figs. 1B, 8 ) and was observed in multiple other class A human GPCRs ( Table 1 ) as well as ␣ integrins ( 65 ).
To defi ne its contribution to TP ␣ function, we performed the following triple mutation: G 205 → L 205 , G 209 → L 209 , and L 213 → Y 213 . The choice of glycine-to-leucine and leucine-to-tyrosine was based on studies of other GxxxG motifs of the acceptor-tagged receptor (which is quantifi ed independently as fold over basal YFP emission) is gradually increased. A saturable BRET curve indicates a specifi c interaction of the two protomers to form a dimer while the concentration of acceptor at which the BRET signal reaches 50%, the BRET 50 , refl ects the affi nity of individual promoters for each other ( 71 ) ( Fig 6A-D ). We determined that although TP L205,L209,Y213 retained the capacity to dimerize, the BRET 50 for TP L205,L209,Y213 homodimerization was significantly right shifted (BRET 50 = 1.83 ± 0.1, n = 5) compared with that of TP WT homodimerization (BRET 50 = 1.4 ± 0.08, n = 4), indicating reduced effi ciency in formation of the homodimer when the TM5 GxxxGxxxL motif was disrupted ( Fig. 6E ). To confi rm impaired homodimerization of the mutant receptor, BRET was measured in HEK 293 cells expressing a fi xed ratio of rLuc-TP WT + YFP-TP WT (1:7) and competition by unfused TP WT or TP L205,L209,Y213 examined. As expected, TP WT effi ciently competed for the interaction of rLuc-TP WT and YFP-TP WT reducing the BRET signal in a concentration-dependent manner. TP L51,L54 , in which the TP-IC1 GxxxG motif G 51 ARQG 55 was mutated, was as effi cient as the TP WT in competition for rLuc-TP WT -YFP-TP WT interaction while, in contrast, TP L205,L209,Y213 did not alter the BRET signal confi rming its relative defi ciency for dimer formation ( Fig 6F ). Together these data indicate the importance of TP-TM5 GxxxGxxxL for effi cient TP ␣ homodimerization.

TM5 GGL domain disruption does not modify IP-TP ␣ heterodimerization or function
The studies thus far indicate that the GxxxGxxxL motif in TM5 of the TP ␣ is important for effi cient homodimerization and that its disruption suppresses receptor signaling. We have also reported the TP can interact with the IP, a Gs -cAMP coupled receptor, to form a heterodimer ( 45 ). When heterodimerized with the IP, the TP's microdomain localization, signal transduction, and regulation is markedly altered with reduced "normal" transduction of Gq-InosP signal in response to TP agonists and a concomitant switch to signal via the Gs-cAMP pathway in an IP-like manner ( 46 ). This signaling shift likely contributes to the restraint placed on the TP via the IP and to the increased risk of CVD in individuals heterozygous for signalingdefi cient IP mutants ( 51 ). We next asked whether disruption of the TP-TM5 GxxxGxxxL motif modifi es IP-TP ␣ heterodimerization and what, if any, is the function contribution to the IP-TP ␣ -Gs signaling in response to TP activation. Interestingly, disruption of the TM5 GxxxGxxxL motif did not modify heterodimerization of the TP with the IP; the BRET saturation curves and BRET 50 for IP-TP L205,L209,Y213 (1.24 ± 0.06, n = 6) was indistinguishable from the IP-TP WT (BRET 50 = 1.26 ± 0.06, n = 6). Concordantly, U46619-induced cAMP generation, the signature "switch" in TP signaling from the Gq pathway to the Gs pathways, was not different between IP-TP WT and IP-TP L206,L209,213 in transfected HEK 293 cells or MEG-01 cells ( Fig. 7 ). Thus, while the TM5 GxxxGxxxL motif was critical for effi cient TP ␣ homodimerization and Gq-signaling, this motif did not contribute to IP-TP ␣ heterodimerization Together these analyses clearly indicate no major role for the TM5 GxxxGxxL motif in processing of the TP ␣ to form a high-affi nity receptor-Gq complex at the cell surface.
Homodimerization of GPCRs appears universal across the superfamily ( 40,44,72 ). Given the established contribution of GxxxG motifs to helix-helix interactions, the extensive evidence that TMs are critical for GPCR homodimerization and the outward-facing orientation of the G 205 xxxG 209 xxxL 213 triplet in TP ␣ -TM5, we considered whether this motif contributes to TP ␣ homodimer formation. We found that while saturable BRET was achieved, the BRET 50 for TP L205,L209,Y213 homodimerization was signifi cantly right-shifted compared with TP WT . Thus, while TP L205,L209,Y213 protomers can dimerize, they do so with a reduced affi nity. Importantly, we confi rmed independently that TP L205,L209,Y213 was unable to compete for TP WT -TP WT interaction, confi rming the mutant's dimerization defi ciency. Thus, similar to the ␤ 2-AR ( 42,57 ) and yeast ␣ -factor ( 66,73 ) receptors, a TM motif GxxxGxxxL is necessary for normal effi cient TP ␣ homodimerization. Reports vary as to the contribution of homodimerization to receptor function, with substantial evidence that homodimerization is necessary for normal surface expression of the receptor ( 39,66,(74)(75)(76)(77) and that a dimeric pair coupled to a single G protein forms the basic signaling unit ( 43,78,79 ). Thus, dimerization-defi cient GPCRs often fail to traffi c normally to the cell surface, while ER retained GPCRs cause their WT counterparts to stay in the ER in a dominant negative manner ( 42 ). Indeed, in the case of ␤ 2 adrenergic receptor disruption of the TM6 GxxxGxxxL motif right shifted the BRET 50 for homodimerization coincident with reduced cell surface receptor expression ( 42 ). Our data showing normal processing, cell surface expression, and G protein association of TP L205,L209,Y213 despite impaired dimerization suggests that for the TP ␣ the two processes, homodimerization and cell surface expression, are independent. Alternatively, it may be that the level of TP L205,L209Y,213 homodimerizes suffi ciently to traffi c to the cell surface, but that the reduced protomer affi nity signifi cantly modifi es the effi ciency with which signal is transduced. Our data does not reveal how activation of Gq via TP L205,L209,Y213 is reduced, but one possibility is the formation of a suboptimal conformation of the TP L205,L209,Y213 homodimer, impacting the receptor dimer's ability to undergo the necessary conformational shift to fully activate Gq.
We reported previously that in addition to homodimerization, the TP ␣ forms high affi nity heterodimers with the IP, which is distinct in sequence, membrane microdomain localization, regulation, and effector signaling (45)(46)(47)(48). TP agonists signal through the IP-TP ␣ heterodimer in an IP-like manner cAMP generation, and coincidentally suppress InosP generation ( 4,48 ). Interestingly, in the current study, mutation of the G 205 xxxG 209 xxxL 213 motif did not impact either heterodimerization with the IP or TP agonist-induced cAMP generation through the heterodimer. Thus, it appears that the TP ␣ -TM5 GxxxGxxxL motif contributes selectively to homodimerization and ( 62,64,65 ), and one study of a TM GxxxGxxxL motif in the ␤ 2-AR receptor ( 42 ), and was designed to disrupt the small-small-big triplet alignment along the outer side of TM5 (see Figs. 1B,8 ). Signaling of the TP L205,L209,Y213 via the Gq-InosP cascade was markedly reduced in both transfected HEK 293 cells and Meg-01 platelet-like cells. Given the role of GxxxG motifs in helical packing ( 58 ) we considered that this loss of function might be due to improper processing of the correctly folded receptor at the cell surface. However, comparable cell surface expression of the WT and TM5 GxxxGxxxL mutant receptor was evident by fl ow cytometry in both in vitro models, and no alteration in processing of the fully glycosylated receptor was evident by Western blotting ( Fig. 5 ). Further, displacement analysis using a range of TP ligands revealed no difference in the ligand binding properties of TP L205,L209,213 compared with TP WT , and both the mutant and WT receptor displayed high agonist affi nity, consistent with normal G protein association and the comparable levels of Gq that accompanied the WT or mutant receptor in coimmunoprecipitation experiments . two receptors couple to one G protein ( 78,79 ). In heterodimers, one promoter typically dominates the downstream signal transduced, and hence the biological outcome ( 79 ). For example, in heterodimers of the B2 receptor for the vasorelaxant bradykinin and the AT1 receptor for the vasoconstrictor angiotensin II, the latter dominates leading to enhanced AT1-Gq signaling and vasoconstriction ( 74,80 ). It remains unclear whether ligation of one or both protomers is optimal and to what extent G protein activation is symmetrical (the agonist activates the protomer that is directly associated with the G protein) or asymmetrical (the agonist indirectly activates the G protein through the non-G protein associated protomer) ( 78,81 ). In the case of the serotonin type 4 receptor homodimer, evidence supports asymmetrical G protein activation through one ligand binding to its protomer but activating signaling via the companion protomer ( 78 ). For the IP-TP ␣ , we established that the IP dominates the heterodimer's signaling through the Gs-cAMP cascade but that agonists for either protomer could activate the complex ( 48 ). Our observations that the TM5 GxxxGxxxL mutant did not support normal Gq-InosP signaling in the homodimer but was fully capable of propagating a normal cAMP response to the TP agonist in the IPTP ␣ heterodimer, provides further support for the 2-receptors-1-G-protein model and for asymmetrical G protein activation through one protomer in a dimeric complex (in this case agonism of the TP ␣ led to activation of the IP-associated Gs in the IP-TP ␣ heterodimer).
We reported that the shift in TP ␣ function to Gs signaling when dimerized with the IP likely contributes to the restraint placed by the PGI 2 -IP system and the TXA 2 -TP system in vivo ( 42,44,50 ). It is, therefore, very promising to uncover a molecular region that selectively reduces TP that distinct receptor regions direct formation of the TP ␣ IP heterodimer .
It has been over a decade since the GPCR dimerization was fi rst reported. Since that time much has been learned about the molecular mechanisms of GPCR dimerization and the biological relevance for receptor function. The most well established model of GPCR dimerization holds that Fig. 8. Modeling of the TP ␣ -TM5 highlighting the leucines that neighbor G 205 and G 209 and L 213 . By analogy with glycophorin A, the small residues, S201, G205, and G209 align to create a groove (green), while the large residues L 203 , L 206 , L 210 , and L 213 form an adjacent ridge (yellow). homodimer function without altering activation and signaling of IP-TP ␣ heterodimer . Efforts to antagonize the TP have proved clinically disappointing ( 22,26 ), perhaps because TP antagonists block activation of the TP in both its TP-TP homodimeric (Gq-coupled) and IP-TP heterodimeric (Gs-coupled) complexes ( 44 ). Our work opens novel avenues to biased interference with the TPTP homodimer while sparing the function of the IPTP heterodimer and its benefi cial cardiovascular biological effects.
Arguably, such an approach should be superior to inhibitors of thromboxane synthase, and even selective inhibition of platelet COX-1 with low-dose aspirin, because the endogenous ligand acting at the IPTP heterodimer would be spared. Recently, computationally designed peptides directed at the GxxxGxxxL motif that mediates interaction of the ␣ IIb ␤ 3 integrins were reported to modify integrin function in platelets ( 54,65 ). We propose that such a peptide targeted at TP-TM5 GxxxGxxxL domain may provide a novel approach to biased TP antagonism. Conceivably, such selective targeting of the TP homodimer would allow us to modify signaling in cell types with high TP-TP expression, such as platelets, while largely preserving the function of cells that have a higher IP-TP dimer population, such as macrophages ( 51 ).