Different single receptor domains determine the distinct G protein coupling profiles of members of the vasopressin receptor family.

The vasopressin receptor family is unique among all classes of peptide receptors in that its individual members couple to different subsets of G proteins. The V vasopressin receptor, for example, is preferentially linked to G proteins of the G class (biochemical response: stimulation of phosphatidylinositol hydrolysis), whereas the V vasopressin receptor is selectively coupled to G (biochemical response: stimulation of adenylyl cyclase). To elucidate the structural basis underlying this functional heterogeneity, we have systematically exchanged different intracellular domains between the V and V receptors. Transient expression of the resulting hybrid receptors in COS-7 cells showed that all mutant receptors containing V receptor sequence in the second intracellular loop were able to activate the phosphatidylinositol pathway with high efficiency. On the other hand, only those hybrid receptors containing V receptor sequence in the third intracellular loop were capable of efficiently stimulating cAMP production. These findings suggest that the differential G protein coupling profiles of individual members of a structurally closely related receptor subfamily can be determined by different single intracellular receptor domains.

An extraordinarily large number of neurotransmitters, peptide hormones, neuromodulators, and autocrine and paracrine factors exert their physiological actions via binding to specific plasma membrane receptors that are coupled to distinct classes of heterotrimeric G proteins (G protein-coupled receptors (GPCRs)). 1 During the past decade, several hundred members of this receptor superfamily have been cloned and sequenced (Watson and Arkinstall, 1994). Characteristically, each GPCR interacts only with specific subclasses of the many structurally similar G proteins found within a cell (Hedin et al., 1993;Conklin and Bourne, 1993). To understand how this selectivity is achieved at a molecular level has become the research focus of an ever increasing number of laboratories.
Mutagenesis studies as well as experiments with short synthetic receptor peptides that can mimic or inhibit receptor-G protein interactions have shown that multiple intracellular receptor domains are involved in G protein coupling (Strosberg, 1991;Savarese and Fraser, 1992;Hedin et al., 1993). Consid-erable insight into the structural basis of receptor-G protein coupling selectivity has been provided by detailed mutational analysis of biogenic amine receptors such as the muscarinic acetylcholine (Wess, 1993) and adrenergic receptors (Dohlman et al., 1991;Strader et al., 1994).
In contrast, little is known about the structural elements involved in G protein recognition by GPCRs that bind peptide ligands. However, such receptors form one of the largest subclasses of GPCRs, and more than sixty different peptide receptors have been cloned to date (Watson and Arkinstall, 1994). These receptors (including, for example, those for melanocortins, cholecystokinin, endothelins, neurokinins, bombesin-like peptides, opioids, or somatostatin) are known to play key roles in the regulation of a multitude of fundamental physiological processes. Investigations into the structural basis of the G protein coupling selectivity displayed by these receptors have been hampered by the fact that the individual members of virtually all peptide receptor subfamilies couple to similar G proteins. All cholecystokinin, endothelin, neurokinin, and bombesin receptors, for example, are preferentially coupled to G proteins of the G q/11 family, whereas the various opioid and somatostatin receptors are all selectively linked to G proteins of the G i/o class (Watson and Arkinstall, 1994). This pattern has precluded the use of hybrid peptide receptors (in which distinct domains are exchanged between functionally different members of a receptor subfamily) to study the structural basis underlying the selectivity of protein recognition displayed by these receptors.
In contrast to all other peptide receptor subfamilies, the group of vasopressin receptors is exceptional in that its members clearly differ in their G protein coupling profiles. The vasopressin receptor family is formed by three distinct subtypes, V 1a , V 1b , and V 2 , which share a high degree (40 -50%) of sequence identity (Birnbaumer et al., 1992;Lolait et al., 1992;Morel et al., 1992;Sugimoto et al., 1994;de Keyzer et al., 1994). However, the V 1a and V 1b receptors are selectively coupled to G proteins of the G q/11 family (Laszlo et al., 1991), which mediate the activation of distinct isoforms of phospholipase C␤, resulting in the breakdown of phosphoinositide lipids (PI hydrolysis). The V 2 receptor, on the other hand, preferentially activates the G protein G s (Laszlo et al., 1991), resulting in the activation of adenylyl cyclase(s).
The individual vasopressin receptors mediate numerous important physiological effects including hepatic glycogenolysis, contraction of vascular smooth muscle and mesangial cells, aggregation of platelets, and antidiuresis in the kidney (Laszlo et al., 1991). Moreover, recent studies have shown that mutations in the V 2 receptor gene are responsible for the X-linked form of nephrogenic diabetes insipidus (for recent reviews see Birnbaumer (1995) and Spiegel (1996)).
To study the structural elements responsible for the functional diversity found within the vasopressin receptor family, we have created a series of V 1a /V 2 hybrid receptors in which distinct intracellular domains were exchanged between the two wild type receptors (Fig. 1). Functional characterization of the resulting hybrid receptors in transfected COS-7 cells led to the novel observation that different single receptor segments determine the differential G protein binding profiles of two structurally closely related peptide receptors.

EXPERIMENTAL PROCEDURES
Construction of Hybrid Receptors-Chimeric rat V 1a /human V 2 vasopressin receptor genes were constructed by using standard polymerase chain reaction mutagenesis techniques (Higuchi, 1989). The V 1a expression plasmid, V1pcD-SP6/T7, has been described previously . The V 2 expression plasmid, V2pcD-PS, was prepared as follows. A 1.7-kilobase pair EcoRI-XbaI fragment was cut out from hV2-pcDNAI/Amp (a plasmid containing the genomic human V 2 vasopressin receptor sequence; kindly provided by Dr. Allen Spiegel, NIH) and subcloned into the pcD-PS mammalian expression vector (Bonner et al., 1988) using the EcoRI and SpeI sites present in the pcD-PS polylinker sequence. The two introns interrupting the V 2 receptor coding sequence were removed, and a stretch of nucleotides coding for a 9-amino acid epitope derived from the influenza virus hemagglutinin protein (YPYDVPDYA; Kolodziej and Young (1991)) was inserted after the initiating Met codon by employing standard polymerase chain reaction mutagenesis techniques (Higuchi, 1989). The ligand binding and G protein coupling properties of the epitope-tagged wild type V 2 receptor did not differ significantly from those found with the nontagged version. 2 The composition of the individual V 1a /V 2 hybrid receptors is given in the legend to Fig. 1. To create a mutant m2 muscarinic receptor (human; m2␤i3) in which the i3 loop (amino acids 206 -390) was replaced with the corresponding human ␤ 2 -adrenergic receptor sequence (amino acids 220 -277), the Hm2pcD (Bonner et al., 1987) and ␤2pSVL (Fraser, 1989) expression plasmids were used for polymerase chain reaction-based mutagenesis. The correctness of all polymerase chain reaction-derived sequences was confirmed by dideoxy sequencing of the mutant plasmids (Sanger et al., 1977).

Transient Expression of Wild Type and Mutant
Receptors-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, 2 ϫ 10 6 cells were seeded into 100-mm dishes. About 24 h later, cells were transfected with the various vasopressin receptor constructs (4 g of plasmid DNA/dish) by a DEAE-dextran method (Cullen, 1987). In G␣ 16 (human) coexpression experiments, COS-7 cells were cotransfected with 4 g of individual receptor constructs and 1 g of the pcisG16 expression plasmid (Amatruda et al., 1991).
Radioligand Binding Assays-For radioligand binding studies, COS-7 cells were harvested approximately 72 h after transfections, and membrane homogenates were prepared as described previously by Dörje et al. (1991). Binding buffer consisted of 50 mM Tris (pH 7.4), 3 mM MgCl 2 , 1 mM EDTA, 0.1% bovine serum albumin, and 0.1 mg/ml bacitracin. Incubations were carried out for 1 h at 22°C in a 0.5 ml volume with increasing concentrations of the radioligand, [ 3 H]arginine vasopressin ([ 3 H]AVP, 81 Ci/mmol; DuPont NEN). Six different concentrations of the radioligand (0.05-5 nM) were used. Nonspecific binding was assessed in the presence of 5 M AVP. Protein concentrations were determined according to Bradford (1976). Binding data were analyzed by a nonlinear least squares curve-fitting procedure using the computer program, Kaleidagraph (Synergy Software).
Stimulation of PI Hydrolysis-About 20 -24 h after transfections, cells were transferred into 6-well plates, and 3 Ci/ml of [ 3 H]myoinositol (20 Ci/mmol, American Radiolabeled Chemicals Inc.) was added to the growth medium. After a 40 -48-h labeling period, cells were incubated with increasing concentrations of AVP for 1 h at 37°C. Increases in intracellular inositol monophosphate (IP 1 ) levels were determined by anion exchange chromatography as described (Berridge et al., 1983;Blin et al., 1995). Studies with cells expressing the wild type V 1a receptor showed that AVP-induced IP 1 accumulation was approximately linear over the 1-h assay interval.
cAMP Assays-Approximately 20 -24 h after transfections, cells were transferred into six-well plates, and 2 Ci/ml of [ 3 H]adenine (15 Ci/ mmol; American Radiolabeled Chemicals Inc.) was added to the growth medium. After a 40 -48-h labeling period, cells were preincubated in Hanks' balanced salt solution containing 20 mM Hepes and 1 mM 3-isobutyl-1-methylxanthine for 15 min (37°C) and then stimulated with increasing concentrations of AVP for 1 h 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 by Salomon et al. (1974). In a subset of experiments, cells were incubated with pertussis toxin (PTX; 500 ng/ml) for the last 24 h of culture. Studies with cells expressing the wild type V 2 receptor showed that AVP-induced cAMP accumulation was approximately linear over the 1-h assay interval.
Drugs-PTX was purchased from List. Unless otherwise noted, all other drugs were obtained from Sigma.

Functional Profile of Wild Type Vasopressin Receptors and Ligand Binding Properties of Wild Type and Mutant
Receptors-All wild type and mutant vasopressin receptors analyzed in this study were transiently expressed in COS-7 cells and assayed for their ability to mediate AVP-dependent stimulation of adenylyl cyclase (mediated by G s ) and PTX-insensitive stimulation of PI hydrolysis (mediated by G proteins of the G q/11 class; Smrcka et al. (1991); Berstein et al. (1992)). Consistent with its reported functional profile, the wild type V 1a receptor (rat) mediated a pronounced increase in inositol phosphate production (7.6 Ϯ 0.5-fold above basal) but left intracellular cAMP levels unaffected (Table I and Figs. 2 and 3). On the other hand, AVP stimulation of the wild type V 2 receptor (human) led to a marked increase in cAMP production (13.2 Ϯ 2.3-fold above basal) but resulted in only a rather weak stimulation of the PI pathway (Table I and Figs. 2 and 3).
To explore the structural basis underlying this selectivity, a series of hybrid V 1a /V 2 receptors ( Fig. 1) were created in which distinct intracellular domains were systematically exchanged between the two wild type receptors (note, however, that V2i1 contains V 1a receptor sequence not only in the first intracellular loop (i1) but also in the extracellular N-terminal domain and the first transmembrane segment). Saturation binding studies showed that all mutant receptors, when transiently expressed in COS-7 cells, retained the ability to bind the agonist radioligand [ 3 H]AVP with high affinity (Fig. 1). Moreover, all hybrid receptors were expressed at levels similar to those found with the two wild type receptors (B max ϭ 480 -690 fmol/ mg; Fig. 1).
Whereas the wild type V 2 receptor bound [ 3 H]AVP with 2-3-fold higher affinity than the wild type V 1a receptor (p Ͻ 2 T. Schöneberg and J. Wess, unpublished results.
a nd, not determinable with sufficient accuracy. b no significant increase in cAMP above basal levels.
0.05), this affinity pattern was reversed when the i3 loop was exchanged between the two wild type receptors (Fig. 1). In agreement with previous studies using hybrid m2/m3 muscarinic (Wess et al., 1990) and hybrid D 2 /D 3 dopamine receptors (Robinson et al., 1994), this finding may indicate that the i3 loop can exert indirect conformational effects on the configuration of the AVP binding site predicted to be formed by amino acids located on the extracellular receptor surface (Chini et al., 1995). Stimulation of Adenylyl Cyclase-Initially, we examined the ability of the different mutant receptors to mediate AVP-induced increases in intracellular cAMP levels. Mutant V 2 receptors in which the i1 loop (together with the N-terminal domain and the first transmembrane domain), the second intracellular loop (i2), or the C-terminal tail (i4) were replaced with the corresponding V 1a receptor sequences were able to stimulate cAMP production in a fashion similar to the wild type V 2 receptor (Table I and Fig. 2). In contrast, a mutant V 2 receptor (V2i3) containing V 1a receptor sequence in the third intracellular loop (i3) almost completely lost the ability to mediate agonist-dependent stimulation of adenylyl cyclase (Table I and  Consistent with these results, substitution of the i1, i2, or i4 domain of the V 2 receptor into the V 1a receptor resulted in mutant receptors that, similar to the wild type V 1a receptor, lacked the ability to mediate stimulation of adenylyl cyclase (Table I and Fig. 2). However, a mutant V 1a receptor in which the i3 domain was replaced with the homologous V 2 receptor sequence (V1i3) gained the ability to stimulate cAMP production with high efficacy (9.5 Ϯ 1.4-fold increase in cAMP above basal) and high AVP potency (EC 50 ϭ 0.88 Ϯ 0.12 nM) (Table I and Fig. 2).
Stimulation of PI Hydrolysis-To study whether or not the various mutant receptors were capable of coupling to G proteins of the G q/11 class, their ability to mediate AVP-dependent stimulation of PI hydrolysis (measured as increase in intracellular IP 1 levels in the presence of Li ϩ ) was examined. Mutant V 1a receptors in which the i1, i3, or i4 domain were replaced with the corresponding V 2 receptor sequences were able to stimulate the breakdown of PI lipids in a fashion very similar to that of the wild type V 1a receptor (Table I and Fig. 3). However, replacement of the i2 loop in the V 1a receptor with the homologous V 2 receptor sequence resulted in a mutant receptor (V1i2) that, similar to the wild type V 2 receptor, could activate the PI pathway only poorly (Table I and Fig. 3).
In agreement with these results, substitution of the i1, i3, or i4 domain of the V 1a receptor into the V 2 receptor yielded mutant receptors that were unable to efficiently stimulate PI hydrolysis (Table I and Fig. 3). Remarkably, however, replacement of the i2 loop in the V 2 receptor with the homologous V 1a receptor sequence yielded a hybrid receptor (V2i2) that gained the ability to stimulate inositol phosphate production with high efficacy (7.4 Ϯ 0.6-fold increase in IP 1 above basal) and high AVP potency (EC 50 ϭ 1.82 Ϯ 0.21 nM), in a fashion very similar to that of the wild type V 1a receptor (Table I and Fig. 3).
"Bifunctionality" of Hybrid Receptors Is Not Due to Promiscuous G Protein Coupling-As outlined above, all mutant receptors containing V 1a receptor sequence in the i2 loop were able to efficiently activate the PI pathway (mediated by G q/11 ), whereas all mutant receptors containing V 2 receptor sequence in the i3 loop were capable of productively coupling to stimulation of adenylyl cyclase (via G s ). Consequently, two mutant receptors were identified, V2i2 and V1i3 (Fig. 1), which could efficiently couple to both second messenger responses.
To exclude the possibility that the ability of V2i2 and V1i3 to couple to both stimulation of PI hydrolysis and adenylyl cyclase was due to a complete loss of G protein coupling selectivity (coupling promiscuity; Wong and Ross (1994)), we examined the ability of these two mutant receptors to mediate coupling to G i , a G protein (family) recognized by neither of the two wild type receptors. It is known that wild type or mutant GPCRs that can couple to both G s and G i can stimulate adenylyl cyclase activity with markedly increased efficacy after inacti-  Table I and Figs. 2 and 3) are summarized underneath the receptor structures (PI, stimulation of PI hydrolysis; AC, stimulation of adenylyl cyclase). The symbols are defined as the percentage of maximum PI and cAMP responses induced by the wild type V 1a and V 2 receptor, respectively: ϩϩϩϩ, 90 -100%; ϩϩϩ, 80 -90%; ϩ, 10 -30%; Ϫ, no significant response. The following sequences were exchanged between the rat V 1a  and human V 2 (Birnbaumer et al., 1992) receptor (amino acid numbers in parentheses): FIG. 2. AVP-induced cAMP accumulation mediated by wild type V 2 and hybrid V 1a /V 2 vasopressin receptors. Transfected COS-7 cells transiently expressing the different receptors were incubated in 6-well plates for 1 h at 37°C with the indicated AVP concentrations, and the resulting increases in intracellular cAMP levels were determined as described under "Experimental Procedures." The data are presented as fold increase in cAMP above basal levels in the absence of AVP. Basal cAMP levels for the wild type V 2 receptor amounted to 870 Ϯ 390 cpm/well. The basal cAMP levels observed with the different mutant receptors were not significantly different from this value. Each curve is representative of three independent experiments, each carried out in duplicate. vation of G i by PTX treatment (Liggett et al., 1991;McClue et al., 1994). This is illustrated in Fig. 4 for a mutant m2 muscarinic receptor (m2␤i3) in which the i3 loop was replaced with the corresponding ␤ 2 -adrenergic receptor sequence (a structurally homologous m2 muscarinic/␤ 1 -adrenergic mutant receptor can couple to G s , G i , G o , and G z ; Wong and Ross (1994)). PTX pretreatment (500 ng/ml; 24 h) of cells expressing this mutant receptor resulted in a marked increase in cAMP production at all agonist (carbachol) concentrations tested (Fig. 4). In contrast, PTX pretreatment had only little effect on the magnitude of the AVP-induced cAMP responses mediated by the wild type V 2 as well as the V2i2 and V1i3 mutant receptors (Fig. 4).
In another set of experiments, we examined whether the ability of the V2i2 mutant receptor to productively couple to PI hydrolysis was specifically dependent on the presence of V 1a receptor sequence in the i2 loop of this hybrid construct or was rather due to the "loss" of V 2 -i2-loop sequence that could at least theoretically play a role in preventing access to G q/11 proteins. To distinguish between these two possibilities, a mutant V 2 receptor was constructed in which the i2 loop was replaced with the corresponding sequence of the ss4 somatostatin receptor (Fig. 5A), which does not couple to G q/11 but to G i/o proteins . Similar to V2i2, the resulting mutant receptor (V2i2ss; B max ϭ 624 Ϯ 45 fmol/mg) retained the ability to stimulate adenylyl cyclase with high efficacy (Fig.  6). However, in contrast to V2i2, the V2i2ss mutant receptor did not gain the ability to efficiently couple to stimulation of PI hydrolysis (Fig. 6). Moreover, PTX pretreatment (500 ng/ml; 24 h) of V2i2ss-expressing cells had no significant effect on the magnitude of AVP-induced increases in cAMP levels (data not shown), indicating that the V2i2ss mutant receptor and the wild type V2 receptor share similar functional properties.
Functionally Inactive Hybrid Receptors Are Properly Folded-Consistent with the observation that the i2 loop of the V 1a receptor and the i3 loop of the V 2 receptor are required for efficient coupling to G q/11 and G s , respectively, two mutant receptors were identified, V2i3 and V1i2, that were unable to efficiently couple to either of these two classes of G proteins ( Fig. 1). To exclude the possibility that the lack of functional activity found with V2i3 and V1i2 was due to improper folding of the intracellular receptor surface, the two mutant receptors were coexpressed with G ␣16 (Amatruda et al., 1991). It has been shown that G ␣16 (which is physiologically expressed only in a subset of hematopoietic cells) can be activated by almost all GPCRs including the V 1a and V 2 vasopressin receptors (Offermanns and Simon, 1995), resulting in the breakdown of PI lipids mediated by the activation of distinct isoforms of phospholipase C␤ (Lee et al., 1992). As shown in Fig. 7, the V2i3 and V1i2 mutant receptors, similar to the wild type V 2 receptor, gained the ability to efficiently activate the PI pathway when coexpressed with G ␣16 .

DISCUSSION
The vasopressin receptor family represents an ideal model system to study the molecular basis of G protein recognition by peptide receptors, because its individual members (V 1a , V 1b , and V 2 ) clearly differ in their G protein coupling properties. In this study, we have created and functionally analyzed a series of V 1a /V 2 hybrid receptors in which distinct intracellular domains (i1-i4; Fig. 1) were systematically exchanged between the two wild type receptors. cAMP assays showed that all mutant receptors that contained V 2 receptor sequence in the i3 loop were able to stimulate adenylyl cyclase activity with high efficacy and AVP potency, whereas all mutant receptors in which the i3 loop was derived from the V 1a receptor had little or no effect on intracellular cAMP levels (Fig. 1). These data strongly suggest that the i3 loop of the V 2 receptor plays a key role in proper recognition and activation of G s .
On the other hand, all hybrid constructs in which the i2 loop consisted of V 1a receptor sequence were able to activate the PI cascade in a fashion very similar to the wild type V 1a receptor, whereas all mutant receptors that contained V 2 sequence in this receptor region displayed only residual PI activity, similar FIG. 3. AVP-induced stimulation of PI hydrolysis mediated by wild type V 1a and hybrid V 1a /V 2 vasopressin receptors. Transfected COS-7 cells transiently expressing the various receptors were incubated in 6-well plates for 1 h at 37°C with the indicated AVP concentrations, and the resulting increases in intracellular IP 1 levels were determined as described (Berridge et al., 1983;Blin et al., 1995). The data are presented as fold increase in IP 1 above basal levels in the absence of AVP. Basal IP 1 levels for the wild type V 1a receptor amounted to 1500 Ϯ 310 cpm/well. The basal IP 1 levels observed with the various mutant receptors were not significantly different from this value. Each curve is representative of three independent experiments, each carried out in duplicate.
FIG. 4. Effect of PTX on receptor-mediated stimulation of adenylyl cyclase. COS-7 cells transiently expressing the indicated receptors were studied for their ability to mediate AVP-induced increases in intracellular cAMP levels, either in the absence or in the presence of PTX (500 ng/ml). Assays were carried out as described under "Experimental Procedures." The structures of V2i2 and V1i3 are given in Fig.  1. m2␤i3 represents a human m2 muscarinic receptor (Bonner et al., 1987) in which the i3 loop was replaced with the corresponding human ␤ 2 -adrenergic receptor sequence (Chung et al., 1987). Basal cAMP levels for the wild type V 2 receptor amounted to 984 Ϯ 243 cpm/well and remained virtually unaffected by PTX pretreatment. The basal cAMP levels observed with the different mutant receptors were not significantly different from this value. The data are given as means Ϯ S.E. and are representative of three independent experiments, each carried out in duplicate.
to the wild type V 2 receptor (Fig. 1), indicating that the i2 loop of the V 1a receptor is critically involved in selective activation of G q/11 . Consistent with this pattern, substitution of the i2 loop of the V 1a receptor into the wild type V 2 receptor resulted in a mutant receptor (V2i2) that gained the ability to efficiently couple to G q/11 but still retained the ability to productively couple to G s . Analogously, replacement of the i3 loop in the V 1a receptor with the homologous V2 receptor sequence yielded a hybrid construct (V1i3) that gained efficient coupling to G s but was still able to activate G q/11 in a fashion similar to the wild type V 1a receptor. The ability of V2i2 and V1i3 to couple to both G s and G q/11 is fully consistent with the notion that different single receptor domains determine the differential G protein coupling profiles of the V 1a and V 2 vasopressin receptors.
Interestingly, Wong and Ross (1994) recently described chimeric m2 muscarinic/␤ 1 -adrenergic receptors, which are completely nonselective among the known mammalian G proteins. These mutant receptors could activate G i and G s as well as G q/11 , which is not a target of either of the two parent receptors. Moreover, Wong and Ross (1994) found that a mutant m1 muscarinic receptor containing ␤ 1 -adrenergic receptor sequence in the i3 loop does not only activate G q/11 and G s but also G i , which is not a target of either the m1 muscarinic or the ␤ 1 -adrenergic receptor. Like this mutant receptor, the bifunctional receptors described in the present manuscript (V2i2 and V1i3) were composed of sequences derived from G q/11 -and G s -coupled receptors. Therefore, to rule out the possibility that the ability of V2i2 and V1i3 to couple to both G q/11 and G s was caused by a general loss of G protein coupling selectivity, we studied whether these two mutant receptors also gained the ability to couple to G proteins of the G i class, which are activated by neither of the two wild type receptors. Because receptor-mediated activation of G i counteracts the stimulation of adenylyl cyclase mediated by activated G s , selective inactivation of G i by PTX is known to lead to an increase in the magnitude of the cAMP responses induced by (mutant) receptors that couple to both G i and G s (Liggett et al., 1991;McClue et al., 1994). We found that PTX pretreatment had little effect on the magnitude of the cAMP responses mediated by V2i2, V1i3, or the wild type V 2 receptor (Fig. 4). In contrast, PTX pretreatment of cells expressing a mutant m2 receptor that contained ␤ 2 -adrenergic sequence in the i3 loop (a modification predicted to lead to a complete loss in G protein coupling selectivity; Wong and Ross (1994)) resulted in a pronounced FIG. 5. Comparison of the i2 and i3 loop sequences of members of the vasopressin/oxytocin (OXY) receptor family. A, comparison of i2 loop sequences. The underlined sequences of the rat V 1a and human V 2 receptor were replaced with the corresponding V 2 and V 1a sequences, respectively (see also Fig. 1). The boxed sequence of the rat ss4 somatostatin receptor  was substituted into the human V 2 receptor, yielding hybrid receptor V2i2ss (see Fig. 6). *, positions at which all vasopressin/oxytocin receptors have identical residues. #, positions at which the G q/11 -coupled vasopressin/oxytocin receptors (OXY, V 1a , and V 1b ) have identical residues that differ from those present in the V 2 receptor(s). Gaps were introduced to allow for maximum sequence identity. Sequences were taken from: Kimura et al., 1992 (human OXY); Rozen et al., 1995 (rat OXY); Gorbulev et al., 1993 (pig OXY);Sugimoto et al., 1994 (human V 1b ); Lolait et al., 1995 (rat V 1b );Thibonnier et al., 1994 (human V 1a ); Morel et al., 1992 (rat V 1a ); Birnbaumer et al., 1992 (human V 2 ); Lolait et al., 1992 (rat V 2 ); Gorbulev et al., 1993 (pig V 2 ). B, comparison of i3 loop sequences. For explanations, see description of A.
FIG. 6. Functional profile of wild type and mutant V 2 vasopressin receptors. COS-7 cells transiently expressing the indicated receptors were studied for their ability to mediate AVP (1 M)-induced increases in intracellular cAMP and IP 1 levels. Functional assays were carried out as described under "Experimental Procedures." V2i2 and V2i2ss represent mutant V 2 receptors in which the i2 loop was replaced with V 1a or somatostatin receptor sequence (rat ss4 subtype; O'Carroll et al. (1992)), respectively (see Fig. 5A for replaced sequences). Basal cAMP and IP 1 levels for the wild type V 2 receptor amounted to 944 Ϯ 162 and 1865 Ϯ 296 cpm/well, respectively. The corresponding levels observed with the two mutant receptors were not significantly different from these values. The data are given as means Ϯ S.E. and are representative of three independent experiments, each carried out in duplicate.
FIG. 7. Functional interaction of hybrid V 1a /V 2 vasopressin receptors with G 16 . COS-7 cells were cotransfected with G␣ 16 (Amatruda et al., 1991) and the indicated wild type or mutant vasopressin receptors (Control, transfection with receptor DNA alone). AVP (1 M)induced increases in intracellular IP 1 levels were determined as described under "Experimental Procedures." The structures of hybrid receptors V1i2 and V2i3 are shown in Fig. 1. Basal IP 1 levels for the wild type V 1a receptor in the absence or the presence of G␣ 16 amounted to 1638 Ϯ 214 and 2034 Ϯ 355 cpm/well, respectively. Similar values were found with the wild type V 2 and the two mutant receptors. The data are given as means Ϯ S.E. and are representative of three independent experiments, each carried out in duplicate.
increase in agonist-induced adenylyl cyclase activity. Thus, the functional properties of the V2i2 and V1i3 mutant receptors clearly differ from those of the generally "promiscuous" hybrid receptors described by Wong and Ross (1994), indicating that the ability of V2i2 and V1i3 to couple to both G s and G q/11 is not due to a general loss of G protein coupling selectivity.
It might also be argued that the i2 loop of the V 2 receptor plays a specific role in preventing access to G q/11 proteins. Analogously, the i3 loop of the V 1a receptor may be involved in preventing interactions with G s . To exclude such a mechanism as a possible cause of receptor-G protein coupling selectivity, an additional hybrid receptor (V2i2ss) was created in which the i2 loop of the V 2 receptor was replaced with the corresponding segment of the G i/o -coupled ss4 somatostatin receptor . The i2 loop of the ss4 somatostatin receptor shares considerably more sequence homology with the corresponding region of the V 1a receptor (35-40% sequence identity) than with that of the V 2 receptor (20 -25%; Fig. 5A). In contrast to the mutant V 2 receptor containing V 1a receptor sequence in the i2 loop (V2i2), the V2i2ss hybrid receptor did not gain the ability to efficiently couple to G q/11 , indicating that the bifunctionality of V2i2 is not due to the ability of the i2 loop of the V 2 receptor to prevent access to G q/11 proteins or to a loss of specific interactions between the i2 and i3 loop (in the wild type V 2 receptor) that constrain G protein coupling selectivity. Taken together, these data suggest that the i2 loop of the V 1a receptor is in fact directly involved in G q/11 recognition and activation.
Consistent with the proposed roles of the i2 loop of the V 1a receptor and the i3 loop of the V 2 receptor in selective recognition of G q/11 and G s , respectively, two mutant receptors, V2i3 and V1i2, were identified that showed only residual or no functional activity at all (Fig. 1). However, we could demonstrate that both mutant receptors retained the ability to productively couple to G 16 (upon coexpression with G␣ 16 ), a G protein known to be activated by most GPCRs (Offermanns and Simon, 1995). This observation strongly suggests that the inability of the V2i3 and V1i2 mutant receptors to interact with G s and G q/11 is not caused by a generalized misfolding of the intracellular receptor surface.
Taken together, these data provide compelling evidence that different single receptor domains are responsible for the functional diversity found within the vasopressin receptor family. The possibility therefore exists that the G protein coupling selectivity of other classes of peptide receptors is also determined by a clearly delineated intracellular receptor region. In agreement with this view, it has been demonstrated that the G protein coupling properties of a series of splice variants of the pituitary adenylyl cyclase-activating polypeptide receptor critically depend on the sequence present at the C terminus of the i3 loop (Spengler et al., 1993).
It should be of interest to investigate the functional effects of substituting the i2 loop of the V 1a receptor or the i3 loop of the V2 receptor into other GPCRs (nonvasopressin receptors). Such experiments could provide information as to whether these specific loop sequences are sufficient for proper recognition and activation of G q/11 and G s , respectively. Alternatively, this question could be addressed by randomizing intracellular vasopressin receptor sequences while leaving the i2 loop of the V 1a receptor or the i3 loop of the V 2 receptor intact.
A sequence comparison (Fig. 5) shows that the i2 and i3 loops of the V 1a and V 1b vasopressin receptors and the oxytocin receptor (which is structurally closely related to the vasopressin receptors and, like the V 1a and V 1b receptors, is selectively coupled to G q/11 ; Kimura et al. (1992)) are quite similar to each other but substantially differ from the corresponding V 2 recep-tor sequences. Each of the two loops contains a number of residues that are conserved only within the two functional receptor subclasses. It is therefore likely that these amino acids play key roles in determining the distinct G protein coupling profiles of the different vasopressin/oxytocin receptors.
In contrast to the findings reported here for different members of a peptide receptor family, multiple intracellular domains are known to be involved in determining the G protein coupling properties of receptors activated by biogenic amine ligands such as the adrenergic or muscarinic acetylcholine receptors (Wong et al., 1990;Liggett et al., 1991;Wong and Ross, 1994;Blin et al., 1995). Such regions have been shown to include the i2 loop, the N-and C-terminal segments of the i3 loop, and the membrane-proximal portion of the C-terminal tail (i4). It could be demonstrated that these regions act in a cooperative fashion to select and activate the proper set of G proteins (Wong et al., 1990;Liggett et al., 1991;Wong and Ross, 1994;Blin et al., 1995).
Interestingly, several peptide receptors (including, for example, the receptors for calcitonin, glucagon, vasoactive intestinal polypeptide, or secretin) have recently been identified that, similar to two of the hybrid receptors examined in this study (V2i2 and V1i3), can couple to both G q/11 and G s (Chabre et al., 1992;Abou-Samra et al., 1992). This property is also shared by the receptors that are activated by the glycoprotein hormones follicle-stimulating hormone, luteinizing hormone, and thyrotropin (Kosugi et al., 1993a;Allgeier et al., 1994). Loss-offunction mutagenesis studies showed, for example, that mutational modification of the N-and C-terminal segments of the i3 loop of the thyrotropin receptor virtually abolished coupling to G q/11 but had little effect on efficient activation of G s (Kosugi et al., 1993a(Kosugi et al., , 1993b. In agreement with the results of the present study, these data suggest that the region(s) in the thyrotropin receptor critical for activation of G q/11 differ(s) from that (those) required for productive coupling to G s .
Despite the existence of distinct vasopressin receptor sequences dictating specificity of G protein recognition, it is likely that most (if not all) intracellular receptor regions are generally required for the proper formation of the receptor-G protein complex. This notion is based on a large number of biochemical and molecular genetic studies with other GPCRs (Strosberg, 1991;Dohlman et al., 1991;Savarese and Fraser, 1992;Strader et al., 1994) demonstrating that there are multiple receptor-G protein contact sites involving at least three domains on the G␣ subunits (Rens-Domiano and Hamm, 1995). Because all GPCRs and all G proteins are predicted to share a similar three-dimensional structure, the molecular architecture of the receptor-G protein interface may be generally conserved.
In conclusion, this study introduces the novel concept that the differential G protein coupling profiles of individual members of a peptide receptor subfamily are determined by different single intracellular receptor domains. All our experimental data are consistent with the notion that these domains are directly involved in G protein binding and/or activation. The identification of the site(s) on the G protein(s) involved in these interactions should eventually lead to the delineation of threedimensional models of the receptor-G protein complex and provide novel insights into the molecular basis of peptide receptormediated G protein activation.