Genetic Analysis of Receptor-Gαq Coupling Selectivity

Many different G protein-linked receptors are preferentially coupled to G proteins of the Gq/11family. To elucidate the molecular basis underlying this selectivity, different Gq/11-coupled receptors (m3 muscarinic, V1a vasopressin, and gastrin-releasing peptide receptor) were coexpressed (in COS-7 cells) with mutant αs subunits in which residues present at the C terminus of αs were replaced with the corresponding αq/11 residues. Remarkably, whereas none of the receptors was able to interact with wild type αs to a significant extent, all three receptors gained the ability to productively couple to a mutant αs subunit containing a single Glu → Asn point mutation at position −3. Moreover, the m3 muscarinic and the V1a vasopressin receptors but not the GRP receptor also gained the ability to interact with a mutant αs subunit containing a single Gln → Glu point mutation at position −5, indicating that the αq/11 residues present in these mutant G protein constructs play key roles in determining the selectivity of receptor recognition. To identify the site(s) on Gq/11-coupled receptors that can functionally interact with the C terminus of αq/11subunits, we next analyzed the ability of a series of hybrid m2/m3 muscarinic receptors to interact with a mutant αs subunit (sq5) in which the last five amino acids of αs were replaced with the corresponding αq/11 sequence. Similar to the wild type m2 and m3 muscarinic receptors, none of the investigated hybrid receptors was able to efficiently interact with wild type αs. Interestingly, however, three mutant m2 receptors in which different segments of the second and third intracellular loops were replaced with the corresponding m3 receptor sequences were identified, which, in contrast to the Gi/o-coupled wild type m2 receptor, gained the ability to efficiently activate the sq5 subunit. This observation suggests that multiple intracellular receptor domains form a binding pocket for the C terminus of G protein αq/11subunits.

Upon binding of extracellular ligands, G protein-coupled receptors (GPCRs) 1 undergo conformational changes that enable the receptor proteins to interact with specific classes of heterotrimeric G proteins consisting of ␣, ␤, and ␥ subunits (1-7). The activated G protein subunits (␣-GTP and/or free ␤␥) are then able to bind to and modulate the activity of downstream effector enzymes and/or ion channels. Typically, an individual GPCR can interact with only a distinct subset of the many structurally similar G proteins that are expressed within a cell (1)(2)(3)(4)(5)(6)(7). How this selectivity is achieved at a molecular level is not well understood at present, particularly since the molecular architecture of the receptor⅐G protein complex has not been elucidated.
Numerous mutagenesis and biochemical studies (1-7) have shown that multiple intracellular GPCR domains, primarily including the second intracellular loop (i2) as well as the N-and C-terminal portions of the third intracellular loop (i3), play key roles in determining selective G protein recognition. Specific residues contained within these domains are predicted to form a binding pocket for the G protein heterotrimer (7), in analogy to the ligand binding domain present on the extracellular surface of GPCRs (1,2,4,6).
Accumulating evidence (8 -11) suggests that the C-terminal portion of the G protein ␣ subunits, via binding to the intracellular receptor surface (12,13), is of fundamental importance for dictating the specificity of receptor-G protein interactions. Several recent studies (8 -11) have shown, for example, that substitution of ␣ i or ␣ s residues into the C-terminal segment of ␣ q subunits results in mutant G proteins that can be activated by G i -or G s -coupled receptors, respectively. By taking advantage of this observation, the C-terminal amino acids in ␣ i/o subunits that are critical for proper receptor recognition have been mapped in great detail (11). In contrast, the functional roles of individual amino acids present at the C termini of other classes of G␣ subunits have not been studied systematically to date.
To identify the site(s) on GPCRs that can functionally interact with the C terminus of ␣ i/o subunits, we have recently used a novel experimental strategy involving the coexpression of hybrid receptors with hybrid G␣ subunits. We initially showed that the wild type m2 muscarinic receptor, a prototypical G i/ocoupled receptor (14 -16), does not efficiently interact with wild type ␣ q but can productively couple to mutant ␣ q subunits in which the last five amino acids of ␣ q were replaced with the corresponding ␣ i or ␣ o sequences (9). We then demonstrated, by analyzing a large number of mutant m2 receptors, that the ability of the m2 receptor to interact with such hybrid ␣ q subunits was critically dependent on the structural integrity of a four-amino acid motif (VTIL; Val 385 , Thr 386 , Ile 389 , and Leu 390 ) predicted to be located at the C terminus of the i3 loop (9,11). In addition, gain-of-function studies showed that substitution of the VTIL motif into mutant m3 muscarinic receptors (which were unable to couple to wild type ␣ q ) could confer onto these receptors the ability to efficiently couple to mutant ␣ q containing ␣ i or ␣ o sequences at their C terminus (9). These data therefore suggested that the VTIL motif can functionally interact with the C terminus of ␣ i/o subunits and that this contact is intimately involved in determining coupling selectivity and triggering G protein activation.
To investigate whether the results obtained with the G i/o -coupled m2 muscarinic receptor are generally applicable, we recently have extended these studies to other functional classes of GPCRs and G␣ subunits. In the present study, we first wanted to examine which specific amino acids present at the C terminus of ␣ q/11 subunits are of particular importance for selective receptor recognition. Toward this goal, we systematically substituted distinct ␣ q/11 residues into the C terminus of ␣ s (a G protein subunit that mediates the activation of adenylyl cyclase) and studied whether the resulting mutant ␣ subunits gained the ability to be activated by different G q/11 -coupled receptors. Remarkably, two ␣ q/11 residues were identified which, when substituted into wild type ␣ s , resulted in ␣ s single point mutants that, in contrast to wild type ␣ s , could be recognized by G q/11 -coupled receptors.
The second goal of this study was to identify the site(s) on G q/11 -coupled receptors that can functionally interact with the C terminus of ␣ q/11 subunits. To address this issue, we took advantage of the finding that the G i/o -coupled m2 muscarinic receptor, in contrast to the G q/11 -coupled m3 muscarinic receptor (15,16), cannot functionally interact with a mutant ␣ s subunit containing ␣ q sequence at its C terminus. Based on this observation, mutant m2 muscarinic receptors in which distinct intracellular domains were replaced with the corresponding m3 receptor sequences were studied for their ability to gain coupling to this mutant ␣ subunit. These coexpression experiments led to the novel finding that multiple m3 receptor regions, including residues within the i2 loop and the N-and C-terminal portions of the i3 domain, are critical for selective recognition of the C terminus of ␣ q/11 subunits.

EXPERIMENTAL PROCEDURES
DNA Constructs-A rat ␣ s cDNA (17) cloned into the pcDNAI expression vector (10) was used as a template for PCR mutagenesis. All wild type and mutant G␣ s subunits contained an internal hemagglutinin epitope tag (DVPDYAS; Refs. 18 and 19). The presence of the epitope tag that replaced ␣ s residues 76 -82 did not affect the receptor and effector coupling properties of wild type ␣ s (18,19). The construction of sq5 (which codes for a mutant ␣ s subunit in which the last five amino acids of ␣ s were replaced with the corresponding ␣ q sequence) has been described previously (10). To introduce mutations into the C-terminal segment of ␣ s , a 42-base pair NsiI-XbaI fragment was removed from the wild type plasmid and replaced with PCR fragments containing the desired mutations. The correctness of all PCR-derived sequences was verified by dideoxy sequencing of the mutant plasmids (20).
The construction of the hybrid m2/m3 muscarinic receptors used in this study has been described previously (Refs. 21-23; for precise amino acid compositions, see legend to Fig. 4).
Cell Culture and Transfection Conditions-COS-7 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum at 37°C in a humidified 5% CO 2 incubator. For transfections, 1 ϫ 10 6 cells were seeded into 100-mm dishes. About 24 h later, COS-7 cells were cotransfected with the indicated G protein and receptor constructs (0.8 g of DNA each) by using a DEAE-dextran procedure (24). To reduce the expression levels of the different receptor and G protein constructs, transfection mixtures were supplemented with 2.4 g of pcDNAI vector DNA. The following wild type receptor expression plasmids were used: human m2 muscarinic receptor in pcD (25), rat m3 muscarinic receptor in pcD (25), mouse gastrin-releasing peptide (GRP) receptor (26) in pcDNA3 (kindly provided by J. Battey, NIH), and rat V1a vasopressin receptor (27) in pcD-SP6/T7 (kindly provided by M. Brownstein, NIH).
cAMP Assays-Approximately 20 -24 h after transfections, cells were transferred into six-well plates, and 2 Ci/ml [ 3 H]adenine (15 Ci/mmol; American Radiolabeled Chemicals) was added to the growth medium. After a 24 -36 h labeling period, cells were preincubated in Hanks' balanced salt solution containing 20 mM Hepes and 1 mM 3-isobutyl-1methylxanthine for 20 min (room temperature) and then stimulated with the appropriate agonist ligands for 30 min at 37°C. The reaction was terminated by aspiration of the medium and addition of 1 ml of ice-cold 5% trichloroacetic acid containing 1 mM ATP and 1 mM cAMP. Increases in intracellular [ 3 H]cAMP levels were then determined by anion-exchange chromatography as described (28).
Radioligand Binding Assays-N-[ 3 H]Methylscopolamine (79.5 Ci/ mmol; NEN Life Science Products) saturation binding experiments were carried out with membrane homogenates prepared from transfected COS-7 cells essentially as described (29). Nonspecific binding was determined in the presence of 1 M atropine. Protein concentrations were determined by the method of Bradford (30). Binding data were analyzed by using the computer program Ligand (31). Western Blotting-All wild type and mutant G␣ s subunits were detected with the 12CA5 monoclonal antibody (Boehringer Mannheim) directed against the hemagglutinin epitope tag present in all G protein constructs. Samples containing 20 g of membrane protein prepared from transfected COS-7 cells were resolved by SDS-polyacrylamide gel electrophoresis (10%), electroblotted onto nitrocellulose, and probed with the 12CA5 antibody as described (9). Immunoreactive proteins were detected by incubation with horseradish peroxidase-conjugated sheep anti-mouse antibody (Amersham Life Science, Inc.) and visualized using an enhanced chemiluminescence system (Amersham).
Drugs-[Arg 8 ]Vasopressin was purchased from Sigma. All other ligands used in this study were obtained through Research Biochemicals Inc.

Functional Interaction of G q/11 -coupled Receptors with
Mutant ␣ s Subunits-All studies were carried out with COS-7 cells cotransfected with different GPCRs and G␣ constructs. Initially, the ability of the rat m3 muscarinic receptor (25), a prototypical G q/11 -coupled receptor (15,16), to interact with wild type ␣ s (s(wt)) was examined. As expected, based on its known G protein coupling preference, the m3 muscarinic receptor, when cotransfected with s(wt) (or vector DNA as a control) and stimulated with the agonist carbachol (500 M), did not efficiently stimulate cAMP production. The maximum increase in cAMP accumulation (above basal levels) mediated by the m3 receptor in the absence or presence of cotransfected s(wt) amounted to only 1.5-2-fold (see Fig. 2A). However, coexpression of the m3 muscarinic receptor with a mutant ␣ s subunit in which the last five amino acids (QYELL) were replaced with the corresponding ␣ q sequence (EYNLV; resulting in the hybrid ␣ subunit, sq5) led to a pronounced stimulation (6 -8-fold) of adenylyl cyclase activity (see Fig. 2A). Essentially similar results were obtained with two other G q/11 -coupled receptors, the rat V1a vasopressin (27) and the mouse GRP receptor (26). Like the m3 receptor, these two peptide receptors, when activated by the appropriate agonist ligands ([Arg 8 ]vasopressin (1 M) and bombesin (1 M)), did not couple efficiently to s(wt) (see Figs. 2, B and C). In contrast, both peptide receptors were able to productively interact with sq5, mediating 4 -8-fold increases in cAMP levels in the presence of this G protein subunit (see Figs. 2, B and C).
We next wanted to examine which specific amino acids within the C-terminal segment of ␣ q/11 are of particular importance for proper receptor recognition. Toward this goal, specific residues at the C terminus of s(wt) were replaced either alone or in combination with the corresponding ␣ q/11 residues (Fig. 1). The ability of different G q/11 -coupled receptors to gain coupling to these mutant ␣ s subunits was then examined in cotransfected COS-7 cells.
As shown in Fig. 1, the C-terminal five amino acids of ␣ s differ in only three residues from the corresponding ␣ q / 11 sequence (the leucine residue at position Ϫ2 is conserved among all mammalian ␣ subunits, and a tyrosine residue is present at position Ϫ4 in both ␣ s and ␣ q/11 ). Therefore, three ␣ s single point mutants, s(Q 3 E), s(E 3 N), and s(QL 3 EV), and all three possible ␣ s double point mutants, s(QE 3 EN), s(EL 3 NV), and s(QL 3 EV), were prepared ( Figs. 1 and 2). Western analysis using a monoclonal antibody directed against the hemagglutinin epitope tag present in all G protein constructs (18,19) showed that all mutant ␣ s subunits were expressed at similar levels as s(wt) and sq5 (Fig. 3).
Initially, the ability of the m3 muscarinic receptor to functionally interact with the different mutant ␣ s subunits was examined ( Fig. 2A). Remarkably, the m3 receptor gained the ability to activate two ␣ s single point mutants, s(Q 3 E) and s(E 3 N), to a significant extent. In the presence of carbachol (500 M) and either of these two subunits, the m3 receptor was able to mediate a marked increase in adenylyl cyclase activity ((4 -5-fold, as compared with an 6 -8-fold increase observed with sq5) (Fig. 2A). In contrast, coexpression of the m3 muscarinic receptor with s(L 3 V) did not result in a cAMP response that was significantly different from background (as determined in cells transfected with either vector DNA or s(wt)) ( Fig. 2A).
To examine whether the functional effects caused by the single amino acid substitutions were additive, three ␣ s double point mutants, s(QE 3 EN), s(EL 3 NV), and s(QL 3 EV), were constructed and functionally analyzed. When coexpressed with the m3 muscarinic receptor, s(QE 3 EN) was able to mediate a cAMP response that was similar in magnitude to that observed with sq5 ( Fig. 2A), indicating that the functional effects of the Q 3 E and E 3 N point mutations were additive. In contrast, the s(EL 3 NV) and s(QL 3 EV) double point mutants showed functional responses similar to those found with s(E 3 N) and s(Q 3 E), respectively, suggesting that the C-terminal amino acid of ␣ q/11 is not critical for receptor recognition.
To study whether the results obtained with the m3 muscarinic receptor are also applicable to other G q/11 -coupled receptors, the different G␣ s constructs were also coexpressed with the V1a vasopressin and the GRP receptor. As shown in Fig.  2B, the V1a vasopressin receptor showed a functional profile that was very similar to that found with the m3 muscarinic receptor. In contrast, the activity pattern seen with the GRP peptide receptor differed somewhat from that found with the two other receptors. Similar to the m3 muscarinic and V1a vasopressin receptors, the GRP receptor gained the ability to activate s(E 3 N) (as well as the s(QE 3 EN) and s(EL 3 NV) double point mutants) and did not efficiently couple to s(L 3 V) (Fig. 2C). However, in contrast to the two other receptors examined in this study, the GRP receptor did not interact to a significant extent with the s(Q 3 E) single point and the s(QL 3 EV) double point mutants (Fig. 2C).
Identification of Receptor Sites Critical for the Recognition of the C-terminus of ␣ q/11 -The second major goal of this study was to identify the site(s) on G q/11 -coupled receptors that can functionally interact with the C terminus of ␣ q/11 subunits. To address this question, we employed a gain-of-function mutagenesis approach, analyzing the ability of a series of hybrid m2/m3 muscarinic receptors (Fig. 4) to interact with the mutant ␣ s subunit, sq5, in which the last five amino acids of ␣ s were replaced with the corresponding ␣ q sequence. As shown in Fig. 5, the m2 muscarinic receptor, a prototypical G i/o -coupled receptor, did not activate either s(wt) or sq5 to a significant extent. We thus hypothesized that substitution into the m2 receptor of the m3 receptor domain(s) capable of recognizing the C terminus of ␣ q should enable the resulting hybrid receptor(s) to interact with the sq5 subunit. The m2-i2, m2-Ni3, m2-Y, and m2-AALS hybrid receptors (Fig. 4) were included in this analysis because previous studies had shown that the m3 receptor regions/residues contained in these constructs are critical for efficient recognition of G q/11 proteins (22, 23, 32, 33) (m2-tail was included as a negative control; see Refs. 21 and 23). N-[ 3 H]Methylscopolamine saturation binding studies showed that all hybrid receptors were expressed at levels similar to those found with the m2 and m3 wild type receptors (for B max values, see legend to Fig. 5).
Interestingly, three of the investigated hybrid receptors, m2-i2, m2-Ni3, and m2-AALS, gained the ability to productively interact with the sq5 mutant subunit, mediating 4 -6-fold increases in cAMP production (Fig. 5). In contrast, none of these hybrid receptors was able to efficiently interact with s(wt). In these constructs, the i2 loop, the first 21 amino acids of the i3 domain, and a four-amino acid motif (AALS; Ala 488 , Ala 489 , Leu 492 , and Ser 493 ) at the C terminus of the i3 loop were derived from the m3 muscarinic receptor. The m2-Y mutant receptor (which contains a Ser 210 3 Tyr point mutation at the N terminus of the i3 loop) and the m2-tail construct (in which the last 66 amino acids of the m2 receptor were replaced with the corresponding m3 receptor sequence) showed a functional profile similar to that of the wild type m2 muscarinic receptor.

DISCUSSION
The ␣ subunits of heterotrimeric G proteins are known to play central roles in receptor-G protein and G protein-effector interactions (3,5). Whereas more centrally located loop and helical domains of G␣ are known to be involved in effector binding, considerable evidence suggests that the C-terminal segment of G␣ can directly contact the receptor protein (3,5). Consistent with this notion, high resolution crystal structures (34,35) suggest that the C terminus of G␣ subunits is surfaceexposed and thus easily accessible for interactions with the receptor. In addition, several recent studies (8 -11) show that the C-terminal portion of G␣ subunits also plays a key role in dictating the specificity of receptor-G protein coupling. It could be demonstrated (8,9) for example that mutant versions of ␣ q (a subunit that can activate various isoforms of phospholipase C-␤) in which the last five amino acids of ␣ q replaced with the corresponding ␣ i/o sequences allowed stimulation of phospholipase C-␤ by receptors (A1 adenosine, D2 dopamine, and m2 muscarinic receptors) that otherwise are exclusively coupled to G proteins of the G i/o class which cannot activate phospholipase C-␤ efficiently. Subsequently, by studying the ability of the G i/o -coupled m2 muscarinic receptor to activate mutant ␣ q subunits containing single C-terminal ␣ i residues, the functional roles of individual amino acids present at the C terminus of ␣ i/o subunits have been defined (11).
Whereas the structural characteristics of the C-terminal domain of subunits of the ␣ i/o protein family (including ␣-transducin) have been analyzed in great detail (36 -39), little is known about the functional roles of the corresponding amino acids present in other classes of G␣ subunits. One goal of the present study therefore was to examine which specific amino acids present at the extreme C terminus of G proteins of the ␣ q/11 family are involved in proper receptor recognition. To address this question, we employed a gain-of-function mutagenesis approach involving the coexpression of different G q/11 -coupled receptors with mutant ␣ s subunits containing distinct ␣ q/11 residues at their C termini. Consistent with the results of a previous study (10), we initially demonstrated that the m3 muscarinic, the V1a vasopressin, and the GRP receptors (which are all prototypical G q/11 -coupled receptors) do not couple efficiently to wild type ␣ s but can productively interact with a mutant ␣ s subunit (sq5) in which the last five amino acids of ␣ s were replaced with the corresponding ␣ q/11 sequence.
Since the C-terminal five amino acids of ␣ s differ in only three residues from the corresponding ␣ q/11 sequence (Fig. 1), three ␣ s single point mutants (s(Q 3 E), s(E 3 N), and s(L 3 V)) were prepared and functionally analyzed. Whereas none of the tested G q/11 -coupled receptors was able to activate the s(L 3 V) subunit, all of them gained the ability to productively interact with the s(E 3 N) single point mutant. Two of the receptors (m3 muscarinic and V1a vasopressin) were also capable of activating the s(Q 3 E) subunit. These two receptors were able to interact with the s(QE 3 EN) double point mutant with greater efficacy than observed with the s(Q 3 E) and s(E 3 N) single point mutants, mimicking quantitatively the functional activity of the sq5 subunit. Taken together, these gainof-function experiments strongly suggest that two C-terminal ␣ q/11 residues, an asparagine at position Ϫ3 and a glutamate at position Ϫ5, play key roles in determining the receptor selectivity of G protein ␣ q/11 subunits. Consistent with this notion, FIG. 2. Functional interaction of different G q/11 -coupled receptors with C-terminal-modified mutant ␣ s subunits. COS-7 cells were cotransfected with expression plasmids coding for the wild type m3 muscarinic (A), V1a vasopressin (B), or GRP (C) receptor and the indicated G protein constructs (or pCDNAI vector DNA as a control). The structure of the different mutant ␣ s subunits, in which one or more amino acids at the C terminus of s(wt) were replaced with the corresponding ␣ q/11 residues, is shown in Fig. 1. In sq5, the last five amino acids of s(wt) were replaced with the corresponding ␣ q sequence. Transfected cells were incubated for 30 min (at 37°C) in the absence or presence of the appropriate agonist ligands. Basal cAMP levels (no ligand added) were similar for the different wild type and mutant G protein constructs (data not shown). The resulting increases in intracellular cAMP levels (fold stimulation above basal) were determined as described under "Experimental Procedures." In each experiment, the cAMP response mediated by sq5 was set equal to 100%. Data are given as means Ϯ S.E. of three to eight independent experiments, each carried out in triplicate. The following ligands were used: A, carbachol (500 M); B, [Arg 8 ]vasopressin (1 M); C, bombesin (1 M).

FIG. 3. Immunoblot analysis of wild type and mutant G␣ s subunits expressed in COS-7 cells.
Equal amounts of membrane protein (20 g) prepared from transfected COS-7 cells were analyzed by SDSpolyacrylamide gel electrophoresis (10%) and Western blotting using the 12CA5 monoclonal antibody as described under "Experimental Procedures." Two additional blots gave similar results. these two residues are found, with no exception, in all members of the ␣ q/11 protein family (␣ q , ␣ 11 , ␣ 14 , ␣ 15 , and ␣ 16 ; Fig. 1).
Interestingly, molecular genetic and biochemical studies (11, 36 -38) show that the C-terminal residues of subunits of the ␣ i/o protein family, which dictate the specificity of receptor recognition, are present at positions Ϫ4 (cysteine), Ϫ3 (glycine), and Ϫ1 (phenylalanine/tyrosine) (Fig. 1). Thus, the precise positions of the C-terminal amino acids that are critical for determining the specificity of receptor-G protein interactions differ between different functional classes of G␣ subunits. However, in both ␣ i/o and ␣ q/11 subunits, the residue present at the Ϫ3 position is of critical importance for receptor-G protein coupling selectivity. As shown in Fig. 1, the residues present at this position are perfectly conserved within individual G␣ subfamilies and can correctly predict the coupling profile of a given G␣ subunit. It is therefore likely that the corresponding residues in ␣ s and ␣ 12/13 (glutamate and methionine, respectively) have a similar functional importance as the glycine and asparagine residues present at Ϫ3 position of ␣ i/o and ␣ q/11 subunits, respectively. This, however, remains to be tested experimentally.
In the crystal structures of two G protein heterotrimers, the C-terminal segment of G␣ was disordered and not visible (34,35). However, NMR studies on a C-terminal ␣-transducin peptide suggest that the last four amino acids of ␣ i/o subunits form a type IIЈ ␤-turn (which depends on the presence of the conserved glycine residue at Ϫ3 position) that is broken upon interaction with the ligand-occupied receptor (39). Dratz et al. (39) therefore speculated that different types of ␤-turn at the extreme C terminus of G␣ subunits are critical for determining the selectivity of receptor recognition. It is possible that the asparagine side chain present at Ϫ3 position in ␣ q/11 subunits is able to hydrogen bond back to the main chain, thus forming a type VIII ␤-turn characterized by a spatial arrangement of the C-terminal amino acids which differs from that found with ␣ i/o subunits (39).
To gain deeper insight into the molecular mechanisms governing receptor-G protein interactions and G protein activation, the receptor site that interacts with the extreme C terminus of G␣ needs to be identified. By using a coexpression strategy analogous to that described here, we could previously show that the C terminus of ␣ i/o subunits can contact a fouramino acid motif on the m2 muscarinic receptor (VTIL; Val 385 , Thr 386 , Ile 389 , and Leu 390 ) predicted to be located at the C terminus of the i3 loop (9,11). Since these residues are located in a region that is thought to be ␣-helically arranged (23, 40 -42), the VTIL motif is predicted to form a hydrophobic surface that is critical for G protein recognition.
To examine whether this receptor-G␣ contact site is also conserved among other functional classes of GPCRs and G proteins, the second major goal of the present study was to map the site(s) on G q/11 -coupled receptors that is (are) recognized by the C terminus of ␣ q/11 subunits. We initially showed that neither the wild type m2 nor the wild type m3 muscarinic receptor, consistent with their known G protein coupling preference for G i/o and G q/11 proteins, respectively, can efficiently interact with wild type ␣ s . Whereas the m3 muscarinic receptor gained the ability to interact with the chimeric ␣ s /␣ q subunit, sq5, the m2 muscarinic receptor was unable to interact with this mutant subunit. We therefore speculated that substitution into the m2 muscarinic receptor of the m3 receptor region(s) which can recognize the C terminus of ␣ q/11 subunits should confer on the resulting hybrid receptor(s) the ability to activate sq5. Consistent with this concept, we found that a mutant m2 muscarinic receptor, m2-AALS (Figs. 4 and 5), in which the VTIL motif was replaced with the corresponding m3 receptor residues (Ala 488 , Ala 489 , Leu 492 , and Ser 493 , respectively) gained the ability to functionally interact with the sq5 subunit. This observation suggests that the VTIL/AALS residues define a conserved GPCR site that is generally involved in the recognition of the C terminus of G␣ subunits.
Interestingly, however, functional analysis of additional m2/m3 hybrid receptors showed that the m2-AALS construct was not the only mutant receptor capable of activating sq5. Two additional m2 mutant receptors, m2-i2 and m2-Ni3 (Figs. 4, 5), were identified that showed a functional profile very similar to that of m2-AALS. In these two mutant receptors, the i2 loop and the N-terminal portion of the i3 loop, respectively, were replaced with the corresponding m3 receptor sequences. Previous studies have shown that these two domains, together with the AALS motif present at the C terminus of the i3 loop, are critical for selective recognition of G q/11 proteins by the m3 muscarinic receptor (23,33).
One possible explanation for the observed ability of multiple m2/m3 hybrid receptors to activate the chimeric ␣ s /␣ q subunit, sq5, is that the C terminus of ␣ q/11 subunits can be contacted by multiple m3 receptor regions. Based on detailed site-directed mutagenesis studies (33) and several theoretical considerations (43), we recently proposed a model of the intracellular m3 receptor surface in which the functionally critical receptor residues present in the i2 loop and at the N and C terminus of the i3 domain all project into the interior of the receptor protein, thus forming a well defined G protein binding pocket (Fig. 6), analogous to the ligand binding site present on the extracellular receptor surface. It is therefore conceivable that the Cterminal portion of the G␣ subunits, by binding to this intracellular receptor "cavity," can interact with multiple receptor sites that lie adjacent to each other in the properly folded receptor protein.
As indicated above, we previously identified only one major receptor contact site for the C terminus of ␣ i/o subunits by using a coexpression strategy analogous to that described here (9). One possible reason for the discrepancy between this finding and the results of the present study is that the relative functional importance of the i2 loop and the N-and C-terminal segments of the i3 domain may differ among different receptor-G␣ pairs. Alternatively, it is possible that the assay system employed in this study was more sensitive than the one used previously. In the present study, neither of the two wild type nor the different m2/m3 hybrid receptors were able to couple to wild type ␣ s , thus allowing for a straightforward interpretation of the coexpression experiments involving the chimeric G␣ subunit, sq5. In contrast, the interpretation of our previous mutagenesis data was complicated by the fact that several of the analyzed m2/m3 hybrid receptors gained some degree of coupling to wild type ␣ q (which served as template for introducing C-terminal ␣ i residues) and were therefore able to activate phospholipase C-␤ even in the absence of coexpressed hybrid ␣ q /␣ i subunits (9).
The conclusion drawn from these gain-of-function mutagenesis experiments that the C terminus of G␣ subunits can be contacted by several different receptor sites is also supported by a recent loss-of-function study (44) examining the ability of an 11-amino acid peptide derived from the C terminus of ␣-transducin to bind to wild type and mutant versions of rhodopsin. This peptide, via direct binding to rhodopsin (12,39), can stabilize the active metarhodopsin II state and can uncouple rhodopsin-transducin interactions. However, these effects were no longer observed when the peptide was incubated with mutant rhodopsins containing multiple mutations in the i2 loop and at the C terminus of the i3 domain, suggesting that the C terminus of ␣-transducin is recognized by at least two different receptor sites (44).
Although the present study again highlights the key role of the C terminus of G␣ subunits in proper receptor recognition, considerable evidence suggests that other domains on the G␣ subunits as well as the G protein ␤␥ complex can also modulate the selectivity of receptor-G protein interactions, most likely by directly contacting the receptor protein (5,45). The GPCR sites involved in these interactions remain to be identified.
In conclusion, we have shown for the first time that the coupling specificity of G␣ s can be changed by single amino acid substitutions (␣ s 3 ␣ q ), leading to the identification of two C-terminal ␣ q/11 residues that are critical for proper receptor recognition. Moreover, our gain-of-function mutagenesis data predict that multiple intracellular receptor domains form a binding pocket for the C-terminal segment of ␣ q/11 subunits. By using a coexpression strategy similar to that described here, it should be possible to map other functionally relevant receptor-G protein contact sites. The information gathered from such studies should eventually lead to a refined model of the receptor-G protein interface. FIG. 6. Model of the intracellular m3 muscarinic receptor surface depicting residues critical for selective recognition of G q/11 proteins. The view is from the intracellular side of the membrane. The position and relative orientation of the seven transmembrane helices (I-VII) is based on the model proposed by Baldwin (43). The highlighted amino acids are predicted to determine the G protein coupling selectivity of the m3 muscarinic receptor (22,23,32,33). Numbers refer to amino acid positions in the rat m3 muscarinic receptor sequence (25). For clarity, only the membrane-proximal portions of the i3 loop that are thought to form ␣-helical extensions of transmembrane helices V and VI (23, 40 -42) are depicted. The highlighted residues at the N and C terminus of the i3 loop are predicted to form two hydrophobic patches, which, together with several hydrophilic residues in the i2 loop, are critical for proper G protein recognition (modified according to Ref. 7).