A cluster of aromatic amino acids in the i2 loop plays a key role for Gs coupling in prostaglandin EP2 and EP3 receptors.

To assess the structural requirements for G(s) coupling by prostaglandin E receptors (EPs), the G(s)-coupled EP2 and G(i)-coupled EP3beta receptors were used to generate hybrid receptors. Interchanging of the whole i2 loop and its N-terminal half (i2N) had no effect on the binding of both receptors expressed in HEK293 cells. Agonist-induced cAMP formation was observed in wild type EP2 but not in the i2 loop- or i2N-substituted EP2. Wild type EP3beta left cAMP levels unaffected, whereas i2 loop- and i2N-substituted EP3 gained agonist-induced adenylyl cyclase stimulation. In EP2, the ability to stimulate cAMP formation was lost by mutation of Tyr(143) into Ala but retained by mutations into Phe, Trp, and Leu. Consistent with this observation, substitution of the equivalent His(140) enabled EP3beta to stimulate cAMP formation with the rank order of Phe > Tyr > Trp > Leu. The point mutation of His(140) into Phe was effective in another EP3 variant in which its C-terminal tail is different or lacking. Simultaneous mutation of the adjacent Trp(141) to Ala but not at the following Tyr(142) weakened the acquired ability to stimulate cAMP levels in the EP3 mutant. Mutation of EP2 at adjacent Phe(144) to Ala but not at Tyr(145) reduced the efficiency of agonist-induced cAMP formation. In Chinese hamster ovary cells stably expressing G(s)-acquired EP3 mutant, an agonist-dependent cAMP formation was observed, and pertussis toxin markedly augmented cAMP formation. These results suggest that a cluster of hydrophobic aromatic amino acids in the i2 loop plays a key role for G(s) coupling.

Individual members of the superfamily of G protein-coupled receptors (GPCRs) 1 efficiently interact only with a subset of the many structurally similar G protein heterotrimers (1)(2)(3). The spectrum of cellular responses triggered by activation of a specific GPCR is determined by the type of G proteins recognized by the activated receptor. It is therefore very important to elucidate the molecular basis governing the selectivity of receptor/G protein interaction for understanding cellular signal transduction.
Accumulating evidence indicates that multiple receptor regions of GPCRs are involved in G protein coupling and determining the selectivity of G protein recognition. Numerous studies have shown that the second intracellular loop (i2 loop), the membrane-proximal portions of the third intracellular loop (i3 loop), and the N-terminal segment of the cytoplasmic tail all contain amino acids predicted to play roles in regulating selectivity of receptor/G protein interactions (4,5). Traditional mutagenesis approaches, including the use of hybrid receptors and alanine-scanning mutagenesis techniques, have led to important insights into the structural basis underlying the selectivity of receptor/G protein interactions (6). For example, intracellular loop 1 (i1 loop) is less important in determination of G protein selectivity but may indirectly contribute to G protein recognition. The i2 loop and i3 loop are of critical importance in determining the selectivity of receptor/G protein coupling and the efficiency of G-protein activation. The C-terminal tail plays a role in constraining basal activity, by preventing access of the G-protein to the receptor surface. Despite such information, it still remains controversial which receptor elements are critical for G protein selectivity and activation, and thus it is still difficult to predict whether a particular receptor can couple to a G protein.
Prostaglandin E 2 (PGE 2 ), one of the best known arachidonate metabolites, exhibits a broad range of biological actions in diverse tissues through their binding to specific receptors on the plasma membrane (7). We and other groups have revealed the primary structures of eight types of prostanoid receptors, including four subtypes of PGE receptors (EP1, EP2, EP3, and EP4), and demonstrated that they belong to the subfamily of rhodopsin-type (class I) GPCRs (8,9). Prostanoid receptors thus have several unique features specific to prostanoid receptors in addition to those in common with other rhodopsin-type receptors; for example, they contain fewer basic or acidic amino acids throughout their putative transmembrane domains (10). To assess the roles of such unique structural features, we have investigated the properties of receptors with mutations within such unique regions and demonstrated that the arginine residue within the putative seventh transmembrane domain conserved in all prostanoid receptors is important not only for interaction with the carboxylic acid group of agonists but also for particular signal activation (11)(12)(13)(14). Furthermore, we found that the aspartate residue within the seventh transmembrane domain of the EP3 receptor plays a key role in governing G protein association and activation (15). On the other hand, multiple EP3 receptor isoforms exist, which are different only in their C-terminal structures (16,17). We found that these isoforms are different in their constitutive G i activities and thus concluded that the C-terminal tail plays a role in constraining the basal activity, by preventing access of the G i to the receptor surface (18 -21). Thus, structurally close members of the GPCR subfamily such as the prostanoid receptors are useful not only for understanding prostanoid receptor-specific events but also for elucidating the general molecular basis of the structure and function relationship of GPCRs, including G protein selectivity.
To gain new insight into the mechanisms governing receptor/G protein coupling selectivity, here we designed a series of experiments using two members of the prostanoid receptors, aiming to identify structural requirements for selective G s coupling. We first constructed G s -coupled EP2 and G i -coupled EP3 hybrid receptors with the i1, i2, or i3 loops interchanged and examined possible functional interchanges in G s coupling in these mutant receptors. Second, we searched for the functional amino acids critical for G s coupling.
Cell Culture, Transient Expression, and Surface Expression of EP2based or EP3-based Mutant Receptors in HEK293 Cells-HEK293 cells were maintained in Dulbecco's modified Eagle's medium with 10% heat-inactivated fetal bovine serum under humidified air containing 5% CO 2 at 37°C. For transfection using the LipofectAMINE 2000 reagent (Invitrogen), cells in 60-mm tissue culture dishes were incubated at 37°C for 4 h with a transfection mixture composed of 3 ml of Dulbecco's modified Eagle's medium, containing 10% heat-inactivated fetal bovine serum, 1 g of DNA, and 15 l of LipofectAMINE 2000 reagent. For the cAMP assay, HEK293 cells were then trypsinized, and aliquots of recovered cells were transferred to 24-well tissue culture plates. Surface expression of receptor proteins on HEK cell membranes was confirmed by an immunofluorescence assay using antibodies against the N-terminal region of the mouse EP2 and EP3 receptors under nonpermeabilized conditions. PGE 2 -binding Assay-The harvested HEK293 cells expressing each receptor were homogenized using a Potter-Elvehjem homogenizer in 20 mM Tris-HCl (pH 7.5), containing 10 mM MgCl 2 , 1 mM EDTA, 20 M indomethacin, and 0.1 mM phenylmethylsulfonyl fluoride. After centrifugation at 250,000 ϫ g for 20 min, the pellet was washed, suspended in 20 mM Mes-NaOH (pH 6.0) containing 10 mM MgCl 2 and 1 mM EDTA, and was used for the [ 3 H]PGE 2 -binding assay. The membranes (50 g) were incubated with various concentrations of [ 3 H]PGE 2 at 30°C for 1 h, and [ 3 H]PGE 2 binding to the membranes was determined by adding a 1000-fold excess of unlabeled PGE 2 into the incubation mixture. The specific binding was calculated by subtracting the nonspecific binding from the total binding.
Measurement of cAMP Formation-Cyclic AMP levels in HEK293 cells were determined as reported previously (24). The receptor-expressing HEK293 cells cultured in 24-well plates (2 ϫ 10 5 cells/well) were washed with HEPES-buffered saline containing 140 mM NaCl, 4.7 mM KCl, 2.2 mM CaCl 2 , 1.2 mM MgCl 2 , 1.2 mM KH 2 PO 4 , 11 mM glucose, 10 M indomethacine, and 15 mM HEPES, pH 7.4, and preincubated for 10 min. Reactions were started by the addition of test reagents along with 100 M Ro-20 -1724. After incubation for 10 min at 37°C, reactions were terminated by the addition of 10% trichloroacetic acid. The content of cAMP in the cells was measured by radioimmunoassay with a cAMP assay system (Amersham Biosciences).
Stable Expression of mEP3␤, EP3-H140F, mEP2, and EP2-Y143A in the Chinese Hamster Ovary (CHO) Cells-cDNAs for mEP3␤, EP3-H140F, mEP2, and EP2-Y143A were transfected into CHO cells using the LipofectAMINE PLUS system according to the manufacturer's instructions, and stable transformants were cloned as described previously (16). CHO cells expressing each receptor (5 ϫ 10 5 cells) were pretreated with or without PT (20 ng/ml) for 7 h before the addition of the agonist. The cells expressing EP3 receptors were incubated at 37°C for 10 min with or without sulprostone in the absence or presence of 10 M forskolin. The cells expressing EP2 receptors were incubated at 37°C for 10 min with or without butaprost. The cAMP contents were determined as described above.
Statistical Analysis-All data shown are expressed as means Ϯ S.E. of three independent experiments. Statistical analysis was carried out by Student's t test. p values of Ͻ0.005 were considered to indicate a significant difference.

Agonist Binding Properties in Hybrid EP2-EP3 Receptors-
Wild type and mutant EP receptors analyzed in this study were transiently expressed in HEK293 cells and assayed for their ability to mediate agonist-dependent stimulation of adenylyl cyclase (mediated by G s ). Consistent with its reported profile, the wild-type EP2 receptor (mouse, mEP2) caused a pronounced increase in intracellular cAMP levels upon stimulation with butaprost, an EP2 agonist. On the other hand, sulprostone stimulation of the wild type EP3␤ receptor (mouse, mEP3␤) left cAMP levels unaffected. To explore the structural basis underlying G s coupling, a series of hybrid EP2/EP3 receptors were created in which the intracellular domains were systematically exchanged between the two wild type receptors (Fig. 1A). EP2-i1 and EP3-i1 represent EP2 and EP3␤ with interchanged i1 loops, respectively. Moreover, we created hybrid receptors in which the N-terminal (i2N) or C-terminal halves of the i2 loops (i2C) were individually exchanged between the wild type receptors as described below. For every mutant receptor used in this study, the expression of receptor proteins in HEK293 cells was examined by immunofluorescent analysis using antibodies against the N-terminal region of the mouse EP2 and EP3 receptors under nonpermeabilized conditions, and membrane surface expression and the expression levels of each mutant receptor were found to be comparable with those of wild-type receptors ( Fig. 1A and data not shown).
Saturation binding studies showed that among the EP2based hybrid receptors, EP2-i2, EP2-i2N, and EP2-i2C retained the ability to bind to the agonist [ 3 H]PGE 2 with high affinity, but EP2-i1 and EP2-i3 failed to bind to the agonist ( Table I). The EP2-i2, EP2-i2N, and EP2-i2C hybrid receptors exhibited K d values close to that obtained for the wild type EP2 receptor (Table I). [ 3 H]PGE 2 binding to these mutants was displaced by the addition of butaprost with K i values similar to that of the wild type EP2 receptor (K i for butaprost, 1.7-3.0 M). These three hybrid receptors were expressed at levels similar to that found for the wild-type EP2 receptor (B max ϭ 803-1110 fmol/ mg; Table I). On the other hand, all EP3-based hybrid receptors except for EP3-i1 retained the ability to bind to [ 3 H]PGE 2 . These hybrid receptors exhibited K d values close to that ob- The part of the receptors derived from mEP2 is shown in black, and that from mEP3␤ is shown in gray. The amino acid sequences of the i1-i3 loops of EP2 and EP3 are shown below the diagrams, and the region interchanged between the two receptors is boxed. Extracellular N-terminal sequences were detected using corresponding antibodies on nonpermeabilized transfected HEK293 cells. The surface expression was visualized using secondary antibodies labeled by fluorescence. Background was compared using cells transfected with empty vector, pcDNA3 (Mock). B and C, agonist-dependent cAMP formation in HEK293 cells expressing mEP2 and EP2-based mutant receptors (B) and in HEK 293 cells expressing mEP3␤ and EP3-based mutant receptors (C). HEK293 cells expressing each receptor or pcDNA3-transfected HEK293 cells were seeded and cultured for 24 h before the assay (2 ϫ 10 5 cells/well). For the mEP2 and EP2-based mutant receptors, the cells were stimulated for 10 min by adding media with the indicated concentrations of butaprost, an EP2-selective agonist (B). For the mEP3␤ and EP3-based mutant receptors, the cells were stimulated for 10 min by adding media with the indicated concentrations of sulprostone, an EP3-selective agonist (C). Amino acid tained for the wild type EP3 receptor (Table I). [ 3 H]PGE 2 binding to EP3-i2, EP3-i3, EP3-i2N, and EP3-i2C was displaced by the addition of sulprostone with K i values similar to that of the wild type EP3 receptor (K i for sulprostone, 1.1-4.3 nM). These four EP3 hybrids were expressed at levels similar to that found for the wild-type EP3 receptor (B max ϭ 1503-1920 fmol/mg; Table I). Consistent with the previous reports (16,25), wild type EP3␤ showed an ϳ10-fold higher affinity to [ 3 H]PGE 2 than wild type EP2. In this transient expression system in HEK293 cells, the expression levels of wild type EP3 were also about 1.5-fold higher than wild type EP2.
Agonist-dependent Stimulation of Adenylyl Cyclase by Hybrid EP2-EP3 Receptors-The hybrid receptors showing considerable binding affinities for PGE 2 (EP2-i2, EP2-i2N, EP2-i2C, EP3-i2, EP3-i2N, EP3-i2C, and EP3-i3) were then subjected to cAMP formation analysis. Wild-type mEP2 mediated a butaprost-dependent increase in cAMP. In contrast, the mutant EP2 receptor (EP2-i2) containing the EP3 receptor sequence in the i2 loop almost completely lost the ability to mediate agonist-dependent stimulation of adenylyl cyclase; butaprost failed to elicit a significant increase in cAMP production over the background level (Fig. 1B). These results suggested that the i2 loop of EP2 might be essential for G s coupling. On the other hand, substitution of the i3 loop of the EP3 receptor with the EP2 receptor resulted in a mutant receptor (EP3-i3) that was similar to the wild type EP3␤ receptor and lacked the ability to mediate stimulation of adelylyl cyclase. However, the mutant EP3 receptor (EP3-i2) in which the i2 loop was replaced with the corresponding EP2 receptor sequence gained the ability to stimulate cAMP production with high efficacy (9.8 Ϯ 0.91 pmol/well) and high sulprostone potency (EC 50 ϭ 21 Ϯ 1.9 nM) (Fig. 1C). Furthermore, a mutant EP2 receptor (EP2-i2C) in which the C-terminal half region of the i2 loop was replaced with the corresponding EP3 receptor sequence was able to stimulate cAMP formation in a fashion similar to the wild-type EP2 receptor. In contrast, a mutant EP2 receptor (EP2-i2N) containing the EP3 receptor sequence in the N-terminal half of the i2 loop again lost the ability to mediate agonist-dependent stimulation of cAMP accumulation. Consistent with these results, substitution of the C-terminal half region of the i2 loop of the EP3 receptor with the EP2 sequence resulted in a mutant receptor (EP3-i3C) that lacked the ability to mediate stimulation of adenylyl cyclase. However, a mutant EP3 receptor (EP3-i3N) in which the i2N region was replaced with the homologous EP2 receptor sequence gained the ability to stimulate cAMP production with high sulprostone potency (EC 50 ϭ 23 Ϯ 2.1 nM) and high efficacy (22.3 Ϯ 1.9 pmol of cAMP/well). It should be noted that the maximal response by EP3-i2N was significantly higher than EP3-i2. These results suggested that the N-terminal half of the i2 loop in the EP2 receptor (8 amino acids shown in Fig. 1C) is required for G s coupling, and/or the corresponding region of the EP3 inhibits G s coupling. From these results, we speculated that the i2N region of the EP2 receptor may contain a key amino acid residue required for selective G s coupling.
Effects of Point Mutations at Tyr 143 on G s Coupling of the EP2 Receptor-Among the 8 amino acids in the i2N region, 3 amino acids were identical between mEP2 and mEP3␤, which were candidates for key amino acids ( Fig. 2A). In addition, the rat EP2 receptor contains an Ala residue at position 138 instead of Ser, indicating that Ser 138 is less important for G s coupling. We therefore constructed four mutant receptors with Ala mutations at each of the four candidate positions (EP2-Y136A, EP2-G140A, EP2-Y141A, and EP2-Y143A). Among these mutants, EP2-Y136A showed cAMP formation in an agonist dose-dependent manner similar to wild type EP2, whereas EP2-G140A and EP2-Y141A showed high efficacies of cAMP production similar to that of the wild type receptor, although they showed rightward shifted butaprost dose-response curves. In contrast, EP2-Y143A failed to increase cAMP formation above background levels (Fig. 2B). The binding properties of EP2-Y143A was similar to those of the wild-type receptor ( Table I), suggesting that loss of cAMP producing activity is due to a loss of G s coupling and that Tyr 143 in EP2 sequences within the i2 loop of mEP2 and its hybrid receptors (B) and those of mEP3␤ and its hybrid receptors (C) are shown above the graphs. The EP3-derived sequences are boxed, and amino acids common in mEP2 and mEP3␤ receptors are presented in boldface letters. The cAMP contents were determined as described under "Experimental Procedures." The results shown are the means Ϯ S.E. of triplicate determinations. *, plays a critical role for G s coupling. We further examined the effects of various amino acid substitutions of Tyr 143 of EP2 on agonist-induced cAMP accumulation. All mutant EP2 receptors with single amino acid substitutions showed binding properties similar to the wild-type EP2 receptor (Table I and data not shown). Substitution of Tyr 143 with Phe (EP2-Y143F) resulted in a receptor stimulating cAMP production with an efficiency higher than that of the wild-type EP2 receptor (Fig. 2D). Agonistdependent cAMP accumulation was observed in EP2-Y143W and EP2-Y143L, but their agonist dose dependence was lower than that of the wild type EP2 receptor. Substitution with other residues resulted in a great loss in the ability to stimulate the cAMP response (Fig. 2C). The potency order of mutants in butaprost-induced cAMP producing ability was as follows: EP2-Y143F Ͼ wild type Ͼ EP2-Y143W, EP2-Y143L Ͼ Ͼ EP2-Y143N, EP2-Y143D, EP2-Y143R, EP2-Y143P, EP2-Y143I, EP2-Y143A ϭ 0. These results suggested that the aromatic ring nature of tyrosine at this position in the EP2 receptor appears to be required for G s coupling with high efficiency. Substitution of His 140 with an Uncharged Aromatic Residue Is Sufficient to Confer G s Coupling on the EP3 Receptor-In order to explore whether a single or a few amino acid mutations can confer G s coupling on mEP3␤, we constructed three mutant EP3 receptors, EP3-H140Y, EP3-R137G/H140Y, and EP3-A133Y/H140Y, all of which include conversion of His 140 into Tyr (Fig. 3A). Surprisingly, all three mutant EP3 receptors exerted sulprostone-dependent cAMP formation in a fashion similar to that of the mutant EP3-i2N receptor (Fig. 3B). This finding indicated that the single amino acid substitution of His 140 into Tyr is sufficient to confer G s coupling on EP3␤. We further constructed mutant EP3 receptors with His 140 replaced with various amino acids (Fig. 3A). All mutant EP3 receptors with single amino acid substitutions showed [ 3 H]PGE 2 binding properties similar to the wild-type EP3 receptor (Table I and data not shown). Substitution of His 140 with Phe resulted in a mutant EP3 receptor (EP3-H140F) with the most potent ability to stimulate cAMP production; its maximal cAMP production was 2-fold that of the EP3-H140Y receptor (Fig. 3D). Moreover, the mutant receptors with His 140 replaced with Trp and Leu (EP3-H140W and EP3-H140L) exerted moderate and slight increases in cAMP accumulation upon sulprostone stimulation, respectively. The EC 50 values for sulprostone of these four mutant receptors were similar (ϳ8.5-20 nM). In contrast, the mutant EP3 receptors with substitution of His 140 into other amino acids elicited no significant increase in cAMP levels (Fig.  3C). The potency order of mutants for sulprostone-induced cAMP-producing activity was as follows; EP3-H140F Ͼ EP3-H140Y Ͼ EP3-H140W Ͼ EP3-H140L Ͼ Ͼ EP3-H140D, EP3- H140N, EP3-H140R, EP3-H140A, EP3-H140P, EP3-H140I, wild-type mEP3␤ ϭ 0. The binding affinities of EP3 mutants for PGE 2 and sulprostone were similar to that of the wild-type receptor (Table I and data not shown), suggesting that the difference in the cAMP response was not caused by an altered binding affinity for the agonist. These results indicate that substitution of His 140 into a noncharged aromatic residue is sufficient to confer G s coupling on the EP3 receptor. Moreover, the preference of aromatic residues in the efficiency of G s coupling at the equivalent positions in both EP3 and EP2 receptors suggested that this amino acid contributes to G s coupling in similar mechanisms for both EP2 and EP3 receptors.
A Cluster of Aromatic Residues at the Center of the i2 Loop Is Required for Efficient G s Coupling-The present study suggested that the bulky aromatic amino acid at the center of the i2 loop may be one of determinants for G s coupling in prostanoid receptors. However, when we examined the sequences of the i2 loop of the prostanoid receptors, we found that the EP2 receptor has two more aromatic amino acids, Phe 144 and Tyr 145 , just after Tyr 143 . The existence of three aromatic amino acids at this position is conserved among all members of G scoupled prostanoid receptors. Interestingly, the EP3 receptors of various species also contain the latter two aromatic residues, Trp 141 and Tyr 142 , just after the key position, His 140 (Fig. 4). As shown above (Fig. 1, B and C), interchanging the i2C regions had little effect on the ability of the EP2 and EP3 receptors to stimulate adenylyl cyclase activity, suggesting less importance of the i2C region for G s coupling. However, this interchange did not alter the existence of the latter two aromatic residues in the cluster. We therefore hypothesized that the latter two residues in the cluster may have potential roles in G s coupling in the prostanoid receptors, and we examined the effects of mutations at both or either aromatic residues in the EP2 and G s couplingacquired EP3 receptors (Fig. 5). In the EP2 receptor, simultaneous alanine mutations of Phe 144 and Tyr 145 (EP2-YAA) led to a great reduction in the efficiency of agonist-induced cAMP production. A single alanine mutation at Phe 144 (EP2-YAY) resulted in a significant reduction of the butaprost-dependent cAMP response, whereas mutation of Tyr 145 to Ala (EP2-YFA) led to a slight increase in the efficiency of the agonist-induced cAMP response. The rank order of cAMP-producing activity (at 10 Ϫ6 M) of these mutants was as follows: EP2-YFA Ͼ mEP2 (YFY) Ͼ EP2-YAY Ͼ Ͼ EP2-YAA Ͼ EP2-Y143A (AFY) ϭ 0. These results suggest that Tyr 143 is the most critical for G s coupling, but Phe 144 is also required for highly efficient coupling, and Tyr 145 contributes to G s coupling only when an aromatic residue is not present at position 144. We investigated whether a similar tendency could be observed in the G s -acquired EP3 mutant. As discussed above, EP3-H140Y (YWY), which has a cluster of three aromatic residues at the center of the i2 loop, exhibited agonist-dependent adenylyl cyclase activity, whereas wild type EP3␤ (HWY) showed no response upon sulprostone treatment. Simultaneous introduction of Ala residues at positions Trp 141 and Tyr 142 led to a complete loss of the ability to stimulate cAMP formation (EP3-YAA). A single alanine mutation at Trp 141 (EP3-YAY) resulted in a receptor almost unable to stimulate cAMP production, whereas mutation of Tyr 142 to Ala left agonist-dependent cAMP levels unaffected (EP3-YWA). The rank order of these mutants in cAMP-producing activity was as follows: EP3-YWA ϭ EP3-H140Y (YWY) Ͼ Ͼ EP3-YAY Ͼ EP3-YAA, mEP3␤ (HWY) ϭ 0. Thus, similar results were obtained for the EP3 point mutants, indicating that the existence of a hydrophobic aromatic residue at position 140 is the most critical, but Trp 141 and Tyr 142 also contribute significantly and little to G s coupling, respectively. These results suggest that a cluster of aromatic residues at the center of the i2 loop plays a key role in high efficiency G s coupling of the prostanoid receptors.
A Gain-of-function Mutation Does Not Alter Intrinsic G i Activity of the EP3 Receptor-In this study, we used the mEP3␤ receptor as a prostanoid receptor that does not couple to stimulation of adenylyl cyclase and found that the point mutation at His 140 is sufficient to confer G s coupling on the EP3 receptor. Since the bulky hydrophobic amino acid equivalent to His 140 of EP3 was proposed to be important in the general interaction with G proteins, we examined whether this point mutation affects intrinsic G i activity. We established CHO cells stably expressing the G s coupling-acquired mutant EP3 receptor (CHO-EP3H140F) and compared its functional properties with those of CHO cells expressing wild-type EP3␤ (CHO-EP3␤). As observed in HEK293 cells, the two EP3 receptors showed similar binding affinities (EP3␤, K d ϭ 2.84 nM; EP3H140F, K d ϭ 3.17 nM), but the expression level of EP3H140F was lower than that of EP3␤ cells (CHO-EP3␤, B max ϭ 1240 fmol/mg; CHO-H140F, B max ϭ 367 fmol/mg). In CHO-EP3␤ cells, sulprostone did not elicit cAMP formation but inhibited forskolin-induced cAMP formation in a dose-dependent manner with an EC 50 of 3.1 nM (Fig. 6A). This inhibition by sulprostone was completely abolished by pretreatment of the cells with pertussis toxin. In contrast, in CHO-EP3H140F cells, sulprostone dose-dependently stimulated cAMP formation with an EC 50 of 22 nM, and the compound exhibited no more inhibition against forskolininduced cAMP production. However, once the cells were pre-treated with pertussis toxin, sulprostone-induced cAMP formation was significantly potentiated even in the presence of forskolin. It should be noted that the potentiating effects of pertussis toxin were significantly observed even at 10 Ϫ9 M, suggesting that this mutant receptor is capable of G i coupling with high efficiency. These results indicate that the EP3-H140F receptor still has an intrinsic G i activity. Thus, we conclude that the H140F point mutation is sufficient to confer G s coupling with high efficiency on the EP3␤ receptor without affecting intrinsic G i coupling. We further established CHO cells stably expressing the wild-type EP2 (CHO-EP2) and EP2-Y143A receptor (CHO-EP2Y143A) and examined the effects of pertussis toxin on cAMP formation. The two cell lines exhibited same order of PGE 2 binding sites, but the CHO-EP2Y143A cells did not show any cAMP responses upon butaprost treatment (Fig. 6B). Moreover, pertussis toxin failed to restore butaprost-induced cAMP response, indicating that loss of agonist-induced cAMP-producing activity in EP2-Y143A is not a result of gain of G i activity.
A Gain of Function Is Independent of the C-terminal Structure of the EP3 Receptor-We previously reported that mouse EP3 isoforms with different C-terminal tails (EP3␣, EP3␤, and EP3␥) and C-terminal truncated form (T335) differ in their agonist-dependent G s activity (21). Since these isoforms are different only in C-terminal structure, we previously demonstrated that the C-terminal tail could play a role in G s coupling of EP3 receptor. Based on this notion, the effects of i2 loop mutations can be explained by modification of the C-terminal function in G s coupling. To explore this possibility, we examined the effects of H140F mutation on cAMP-producing activity in other EP3 isoforms (Fig. 7). We employed EP3␥ and Cterminally truncated T335, both of which increased cAMP levels in an agonist-dependent manner when expressed in CHO cells (21). In our previous report, the G s activity elicited by EP3␥ observed in CHO cells requires more than 10 Ϫ6 M of agonist, and its maximal response is not as high as EP2 or EP4 receptors, and thus the G s coupling is considered to be less efficient. Indeed, the increase in cAMP formation by wild-type EP3␥ or T335 was hard to detect even in the presence of 10 Ϫ5 M of agonist in the current expression system. On the other hand, introduction of H140F mutation into EP3␥ or T335 resulted in a receptor showing agonist-dependent cAMP-producing activity with similar EC 50 values around 10 Ϫ8 M. Moreover, a significant increase in basal cAMP levels in the absence of agonist was observed in both EP3␥-H140F and T335-H140F but not in EP3␤-H140F. The increase in basal cAMP levels by the T335-H140F was significantly higher than that by the EP3␥-H140F. Instead, the agonist-dependent increase in cAMP levels in the mutant T335 appeared lower than that in the mutant EP3␥. However, in the current system, we could hardly detect cAMP increases with any significant difference in wildtype EP3␥ and T335 even in the presence of 10 Ϫ5 M agonist. These results suggested that the effects of i2 loop mutation on G s coupling of EP3 are independent of C-terminal structure, which is likely to govern the balance of constitutive and agonist-induced G protein activation as observed in the G i activity of the EP3 isoforms.

DISCUSSION
One of the most important findings in this study is the "gain of function" of G s activity of the EP3 receptor by a point mutation; conversion of the amino acid His 140 at the center of the i2 loop into an uncharged aromatic residue is sufficient to confer G s coupling with high efficiency on the EP3 receptor (Fig. 3). The importance of the aromatic moiety of the equivalent amino acid was also demonstrated in G s coupling of the EP2 receptor (Fig. 2). Previously, the importance of bulky hydrophobic amino acid residues at the position corresponding to Tyr 143 of mEP2 in G protein coupling has been pointed out in studies on several kinds of rhodopsin-type receptors (26 -29). This site is located at the C-terminal end of a highly conserved i2 loop motif with the following most common sequence: DRYXX(V or I)XXPL, where X is any amino acid (2). The last residue in the consensus sequence, Leu, is replaced with Phe or Met in some members of G s -coupled receptors. According to the report by Moro et al. (27), Leu 131 in the human M1 muscarinic receptor, which is equivalent to Tyr 143 of the mEP2 receptor, is critical for stimulation of phosphatidylinositol turnover (G q coupling). Moreover, they showed that introduction of the equivalent point mutation F139A into the ␤ 2 adrenoreceptor caused a significant loss in isoproterenol-induced cAMP accumulation (G s coupling). Based on these findings, they concluded that the bulky hydrophobic amino acid at this position is an important amino acid that governs general coupling with any kind of G protein. However, the current finding that conversion of His 140 into Phe in the EP3 receptor failed to alter G i coupling (Fig. 6) may suggest that EP3 does not require the particular amino acid at this position for efficient G i coupling. Indeed, the importance of the hydrophobic amino acid at the corresponding position has not been reported for G i -coupled receptors. However, the His residue is completely conserved in EP3 receptors derived from various species and is quite unique to EP3 in the GPCR family (Fig. 4). Interestingly, the His residue observed at the key position of EP3 also has a positively charged imidazole structure, which is ineffective in G s coupling. This can also be interpreted to signify that the His residue participates in the G i selectivity of EP3 receptor by preventing efficient G s coupling. In this respect, G i -coupled receptors contain a nonhydrophobic amino acid at this position; the EDG2 and EDG3 receptors have a basic amino acid, and chemokine receptors, CXCR4 and CXCR6, have a threonine residue (Fig. 4). The variety of amino acids at this position in G i -coupled receptors may reflect a variety in the way to exert their G i selectivity and the existence of some other domains such as the C-terminal region of the i3 loop to be required for G i activation with high efficiency as suggested for the M2 muscarinic receptor (30,31).
The current study indicated that both EP2 and EP3 require one of the following amino acid residues: Phe, Tyr, Trp, or Leu at position 143 and 140, respectively, for efficient G s coupling. However, the identity of the side chain moiety (Phe, Tyr, Trp, or Leu) affected different parameters of G s coupling between EP2 and EP3. In the EP2 receptor, the identity of aromatic moiety seems to affect the EC 50 values of cAMP production only slightly, suggesting that G s coupling of the EP2 is also governed by other domains such as the i3 loop, as suggested by previous studies (32). In contrast, in the EP3 receptor, the identity of the aromatic moiety affected the maximal cAMP response without great changes in EC 50 values. Thus, it seems that the G s activation efficiency by EP3 completely depends on the side chain moiety at this position, indicating an absolutely pivotal role of this amino acid in G s coupling of EP3. However, we cannot entirely exclude the possibility that the amino acid identity may alter the G i activation efficiency, which is usually included in outcomes in a detection system for G s activity.
Previously, we reported that three C-terminal variants, EP3␣, EP3␤, and EP3␥, and C-terminal truncated T335 differ in their agonist-dependent G s activity (EP3␥ Ͼ T335 Ͼ EP3␣ Ͼ EP3␤ ϭ 0) (21). Since these variants are different only in C-terminal sequence, we speculated that the C-terminal tail may function as a key regulator of G s coupling of EP3 receptor; ␤-tail prevents and ␥-tail allows the interaction of G s with the common structure of the EP3 receptor. However, the current study demonstrated that "G s -excitable" EP3␥ further acquired drastic G s activity, and such gain of function by the H140F mutation is reproduced in C-terminally truncated T335 (Fig.  7). Thus, the gain of G s activity is independent of C-terminal structure. In our previous report, the G s activity elicited by EP3␥ observed in CHO cells requires agonist concentrations of more than 10 Ϫ6 M, and its maximal response is still not as high as that observed for EP2 or EP4 receptors, and thus the activity is considered to be less efficient. Indeed, the agonist-dependent G s activity of wild-type EP3␥ was undetectable in the current expression system. In contrast, the acquired G s activity in the mutant receptor is comparable to EP2 and EP4 in terms of the degree of maximal activity and agonist dose dependence and is thought to be essentially different from intrinsic G s activity appearing in EP3␥. However, the common EP3 structure that allows intrinsic G s activity may serve as a premise factor for point mutation resulting in gain of G s coupling with high efficiency. Whether an introduction of a cluster of aromatic residues at the center of i2 loop enables other G i -coupled receptors to gain G s coupling is an interesting issue to be examined. We also previously reported that EP3 variants are different also in their constitutive G i activity (T335 Ͼ EP3␥ Ͼ EP3␣ Ͼ EP3␤ ϭ 0) (18,21). It is quite interesting that H140F mutants exhibited basal G s activity with rank order of T335 Ͼ EP3␥ Ͼ EP3␤ ϭ 0, which is in good accordance with the potency order of constitutive G i activity in EP3 variants. Moreover, it should be noted that the agonist-dependent G s activity in the mutant T335 appeared less than that in the mutant EP3␥; the G i activity elicited by T335 receptor has been shown to be completely constitutive. These results suggested that the effects of i2 loop mutation on G s coupling of EP3 is independent of C-terminal structure, which is likely to govern the balance of constitutive and agonist-induced G protein activation as observed in the G i activity of the EP3 isoforms. Importantly, these results suggest a general role for the C-terminal tail in G protein coupling; the FIG. 7. Effects of H140F mutation on basal and sulprostone-induced cAMP-producing activity in EP3 receptor isoforms with different Cterminal tails. The H140F mutation was introduced into three kinds of EP3 receptor forms, EP3␤, EP3␥, or T335. HEK293 cells were transfected with each EP3 cDNA or the corresponding mutant receptor cDNA. The cells (2 ϫ 10 5 cells/well) were treated with the indicated concentrations of sulprostone. The cAMP contents were determined as described under "Experimental Procedures." The results shown are the means Ϯ S.E. of triplicate determinations. *, p Ͻ 0.005 versus each wild-type EP3 form; #, p Ͻ 0.005 versus EP3␤-H140F; †, p Ͻ 0.005 versus EP3␥-H140F. C-terminal tail plays a critical role in constraining the constitutive activity irrespective of class of coupling G proteins.
One of the remarkable findings in this study is that a cluster of aromatic amino acids beginning with Tyr 143 or the corresponding residue is required for G s coupling with high efficiency in prostanoid receptors (Fig. 5). This feature, the existence of three bulky aromatic amino acids following the conserved proline residue, is unique to G s -coupled prostanoid receptors (Fig. 4). The 3 amino acids just after the proline in the four G s -coupled prostanoid receptors are YFY, FFY, or YLY, whereas the other members contain HWY, LIH, IFH, or FSR. The present study demonstrated that the existence of an uncharged aromatic residue at the first position is the most critical for G s coupling. However, the simultaneous introduction of alanine mutations at the following two residues resulted in a significant loss of efficiency in G s activity in EP2 and EP3-H140Y receptors. Moreover, the existence of an aromatic residue (Phe 144 in EP2 and Trp 141 in EP3) at the second position appears to be required for G s coupling with high efficiency. In contrast, the Tyr residue at the third position is dispensable if the first two residues are aromatic, but this residue is likely to take part in G s coupling in the absence of an aromatic residue at the second position. Based on these results, we concluded that G s coupling is controlled by the three aromatic amino acids following the conserved Pro residue with a rank order of contribution of first Ͼ second Ͼ third residue in the prostanoid receptors.
How does the aromatic residue contribute to G s coupling with high efficiency? The rank order of amino acids critical for efficient G s coupling of EP2 and EP3 receptors is as follows; Phe Ͼ Tyr Ͼ Trp Ͼ Leu Ͼ Ͼ other amino acids ϭ 0 (Figs. 2 and  3). There is no doubt that the C-terminal 5 amino acids of the G␣ subunit are important for its selective binding to the receptors; both G s and G q/11 families contain a Tyr residue at Ϫ4 from the C-terminal end, whereas the G i family contains a quite different amino acid, cysteine, at this position (33)(34)(35). Recently, Liu et al. demonstrated that the aromatic moiety of the Tyr residue conserved at Ϫ4 from the C-terminal end of the G␣ s and G␣ q plays a key role in receptor/G protein interactions with high efficiency (36). By point mutation analysis, they demonstrated that agonist-induced G␣ 11 activation is controlled by the identity of the Ϫ4 residue with the rank order of Phe Ͼ Tyr Ͼ Trp Ͼ Ͼ other amino acids. Although they did not examine the effect of the Leu mutation, the three most effective amino acids, Phe, Tyr, and Trp are completely identical to the amino acids critical for G s coupling at the key position in both EP2 and EP3 receptors. From these results, we speculate that the bulky aromatic amino acid in the i2 loop takes part in recognition of the Tyr residue conserved in G s and G q through a mechanism such as electron interactions. In such case, a cluster of aromatic residues may contribute to strengthen the interaction with or to accelerate the recognition of the tyrosine residue at the Ϫ4-position of G s . Recently, Erlenbach et al. (29) employed an yeast screening system, in which random mutations were introduced into the G s -coupled vasopressin V2 receptor, to detect amino acid mutations affecting receptor interaction with the C-terminal tail of G proteins. They found that a single amino acid substitution at Met 145 into Leu or Trp within the i2 loop equivalent to Tyr 143 of mEP2 allowed the V2 receptor to couple to both G q and G s (29). They also discussed the possibility that Met 145 is a strong candidate site for interaction with G proteins based on the analogy of the high resolution x-ray structure of bovine rhodopsin. According to the original report, the i2 loop exhibits an L-like structure when viewed parallel to the membrane plane but lacks regular secondary structure (37). Because the cytoplasmic extension of TM III and the N-terminal segment of the i2 loop show considerable sequence homology among GPCRs of the rhodopsin family, it is likely that the i2 loop of the EP2 receptor adopts a structure similar to that observed in rhodopsin. If this is correct, the cluster of aromatic residues from Tyr 143 to Tyr 145 is predicted to be located just N-terminal of the bend of the L-like structure that is a characteristic feature of the i2 loop where it is easily accessible for interactions with G proteins. Taken together, we propose a cluster of aromatic amino acids in the i2 loop as a strong candidate for an interaction site with the G s protein.
The current study demonstrated that interchanging of the i2 loop or the N-terminal or C-terminal half region of the i2 loop between EP2 and EP3 left the individual binding affinities and the specificity and expression levels of the receptors unaffected. These results may reflect the fact that the i2 loops do not directly contribute to the formation of the ligand binding pocket. On the contrary, the interchanging of the i1 loops resulted in loss of binding ability of both receptors. Although all EP receptors can recognize PGE 2 as a natural ligand, it has long been suggested that each EP receptor recognizes different functional groups of agonists (38). Since it was recently proposed that both TM I and II contribute to receptor recognition of different functional groups of prostanoid ligands (39), the i1 loop of the prostanoid receptors may be critical in the formation of subtype-specific ligand binding pockets. Interchanging of the i3 loops differently affected the binding properties of the wild type receptor; EP2 lost but EP3 retained the ability to bind to PGE 2 . This finding may reflect the fact that EP2 requires an i3 loop of appropriate length to form a binding pocket that can hold prostaglandin derivatives with a bulky structure (25).
In summary, we have demonstrated that a cluster of aromatic amino acids at the center of the i2 loop plays a key role in G s coupling, at least in the prostanoid receptors. This study will be of help to understand the molecular mechanisms of G protein coupling selectivity by the individual GPCRs.