Molecular Basis of V2 Vasopressin Receptor/GsCoupling Selectivity*

The molecular mechanisms governing the coupling selectivity of G protein-coupled receptors activated by peptide ligands are not well understood. To shed light on this issue, we have used the Gq/11-linked V1a and the Gs-coupled V2 vasopressin peptide receptors as model systems. To explore the structural basis underlying the ability of the V2 receptor to selectively recognize Gs, we systematically substituted distinct V2 receptor segments (or single amino acids) into the V1a receptor and studied whether the resulting hybrid receptors gained the ability to mediate hormone-dependent cAMP production. This strategy appeared particularly attractive since hormone stimulation of the V1a receptor has virtually no effect on intracellular cAMP levels. Functional analysis of a large number of mutant receptors transiently expressed in COS-7 cells indicated that the presence of V2 receptor sequence at the N terminus of the third intracellular loop is critical for efficient activation of Gs. More detailed mutational analysis of this receptor region showed that two polar V2 receptor residues, Gln225 and Glu231, play key roles in Gs recognition. In addition, a short sequence at the N terminus of the cytoplasmic tail was found to make an important contribution to V2 receptor/Gs coupling selectivity. We also made the novel observation that the efficiency of V2 receptor/Gs coupling can be modulated by the length of the central portion of the third intracellular loop (rather than the specific amino acid sequence within this domain). These findings provide novel insights into the molecular mechanisms regulating peptide receptor/G protein coupling selectivity.

G protein-coupled receptors (GPCRs) 1 form one of the largest protein families found in nature (1). Structurally, all GPCRs are characterized by the presence of seven transmembrane helices (TM I-VII) connected by three extracellular and three intracellular loops (i1-i3). By binding to the extracellular surface of GPCRs, ligands induce conformational changes in the receptor proteins, enabling them to interact with a distinct set of heterotrimeric G proteins that are attached to the intracellular side of the plasma membrane (2)(3)(4)(5)(6). GPCR ligands display a remarkable degree of structural diversity, including, for example, small biogenic amines, large glycoprotein hormones, lipids, calcium ions, and polypeptides (7). GPCRs activated by peptide ligands are particularly abundant, reflecting their involvement in an extraordinarily large number of important physiological processes.
In most cases, ligand-activated GPCRs interact with only a limited number of potential G protein coupling partners (2)(3)(4)(5)(6). The structural basis underlying this coupling selectivity is currently being investigated by many different laboratories. Structure-function analysis of different classes of biogenic amine GPCRs including the muscarinic acetylcholine (8) and adrenergic receptors (2)(3)(4)(5) has led to considerable insight into the structural elements that control the G protein-coupling selectivity of this class of receptors. Such studies have shown that multiple intracellular receptor regions including the i2 loop, the N-and C-terminal portions of the i3 domain, and the membrane-proximal segment of the C-terminal tail (i4) are involved in determining the specificity of receptor/G protein coupling. Particularly in the case of the muscarinic receptor family (m1-m5), the amino acids dictating the signaling selectivity of the different receptor subtypes have been mapped in great detail (8 -12).
In contrast, the molecular mechanisms governing the coupling selectivity of GPCRs activated by peptide ligands are not well understood at present. Studies to explore these mechanisms are hampered by the fact that the individual members of a given peptide receptor family usually couple to similar sets of G proteins. All members of the neurokinin, cholecystokinin, endothelin, and bombesin receptor families, for example, are preferentially coupled to G proteins of the G q/11 family, whereas the various opioid and somatostatin receptor subtypes are all selectively linked to G proteins of the G i/ G o class (1). Thus, this pattern has severely limited the use of hybrid receptor approaches (which have been instrumental in mapping functionally critical domains in non-peptide GPCRs) to identify peptide receptor sites that are critical for G protein recognition.
However, the vasopressin peptide receptor family (which consists of three subtypes, V1a, V1b, and V2) offers an attractive model system to study mechanisms involved in peptide receptor-coupling selectivity, since the individual members of this class of receptors clearly differ in their G protein preference. Whereas the V1a and V1b receptors are selectively coupled to G proteins of the G q/11 family, which mediate the breakdown of phosphatidylinositol (PI) lipids, the V2 receptor preferentially activates the G protein G s , resulting in the activation of adenylyl cyclase (13)(14)(15)(16)(17)(18).
The V1a and V1b receptors are involved in a number of important central and peripheral processes, whereas the V2 receptor is required for water reabsorption in the renal collecting duct system (19). Several laboratories have shown that mutations in the V2 receptor gene are responsible for the X-linked form of nephrogenic diabetes insipidus (reviewed in Refs. 20 and 21).
In an initial study (22), we have recently employed a hybrid receptor strategy to broadly map V1a and V2 receptor domains determining the coupling properties of these two receptor sub-types. We found that the relative functional importance of individual intracellular loops differs among the two vasopressin receptor subtypes. Efficient coupling to G q/11 by the wild type V1a or hybrid V1a/V2 receptors was found to be critically dependent on the presence of V1a receptor sequence in the i2 loop, whereas efficient recognition of G s by the wild type V2 receptor or V1a/V2 mutant receptors required the presence of V2 receptor sequence in the i3 loop (22).
The present study was designed to study the structural basis underlying the ability of the V2 receptor to selectively couple to G s in greater molecular detail, by using a gain-of-function mutagenesis approach. Specifically, we systematically substituted distinct segments/single amino acids of the V2 receptor sequence into the V1a receptor subtype, and then studied the resulting hybrid receptors for their ability to mediate increases in intracellular cAMP levels. This strategy appeared particularly attractive since the wild type V1a receptor essentially offers a "null background," i.e. ligand stimulation of this receptor subtype has virtually no effect on intracellular cAMP levels (22).
In this report, we describe the identification of residues at the TM V/i3 loop and TM VII/i4 domain junctions that are intimately involved in V2 receptor/G s coupling selectivity. Moreover, we made the novel observation that receptor/G s coupling selectivity can be regulated simply by the length (rather than the presence of a distinct amino acid sequence) of the central portion of the i3 loop.

EXPERIMENTAL PROCEDURES
DNA Constructs-To generate hybrid V1a/V2 vasopressin receptor constructs, expression plasmids coding for the rat V1a receptor, V1pcD-SP6/T7 (14), and the human V2 receptor, V2pcD-PS (22), were used. The wild type V2 receptor construct contained a nine-amino acid hemagglutinin epitope tag (YPYDVPDYA) inserted after the initiating methionine codon. Previous studies showed that the ligand-binding and G protein-coupling properties of the epitope-tagged wild type V2 recep-tor did not differ significantly from those found with the nontagged version (22,23). To facilitate the construction of V1a/V2 mutant receptors, the following silent restriction sites were introduced into the V1a receptor expression plasmid: SpeI (codons 173/174), BspEI (codons 224/ 225), KpnI (codons 235/236), and Eco47III (codons 304/305). Chimeric V1a/V2 receptor genes were constructed by using standard polymerase chain reaction mutagenesis techniques (24). The precise amino acid composition of the individual V1a/V2 hybrid receptors is given in Table  I (see also Fig. 1). The correctness of all polymerase chain reactionderived sequences was confirmed by dideoxy sequencing of the mutant plasmids.
Transient Expression of Receptors-COS-7 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine 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, cells were transfected with the various vasopressin receptor constructs (4 g of plasmid DNA per dish) by a DEAE-dextran method (25). Cells were incubated with 900 l of the DNA/DEAE-dextran (10 mg/ml) mixture for 3 h.
Radioligand Binding Assays-For radioligand binding studies, COS-7 cells were harvested approximately 70 -72 h after transfections, and membrane homogenates were prepared as described previously (26 (27). Binding data were analyzed by a nonlinear least squares curvefitting procedure, using the computer program LIGAND (28).
response curves, seven different concentrations of AVP (ranging from 10 Ϫ13 to 10 Ϫ6 M) were used. Incubations were terminated by aspiration of 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 previously (29,30). PI Assays-About 20 -24 h after transfections, cells were split into six-well dishes (approximately 0.3-0.4 ϫ 10 6 cells/well) and labeled with 3 Ci/ml myo-[ 3 H]inositol (20 Ci/mmol; American Radiolabeled Chemicals Inc.). After a 20 -24-h labeling period, cells were preincubated for 20 min at room temperature with 2 ml of Hanks' balanced salt solution containing 20 mM HEPES and 10 mM LiCl. Cells were then stimulated with 1 M AVP for 1 h at 37°C (total volume per well, 1 ml), and increases in intracellular inositol monophosphate (IP 1 ) levels were determined by anion exchange chromatography as described elsewhere (9,31).

RESULTS
When transiently expressed in COS-7 cells, the wild type V2 vasopressin receptor, upon stimulation with the hormone AVP, produced a pronounced increase in intracellular cAMP levels (11 Ϯ 1-fold above basal; EC 50 ϭ 0.56 Ϯ 0.14 nM), whereas the V1a vasopressin receptor was unable to mediate an appreciable increase in adenylyl cyclase activity (Figs. 2 and 3, Tables II and III). To elucidate the structural basis underlying this selectivity, we employed a gain-of-function mutagenesis approach. Distinct segments/amino acids of the V2 receptor sequence were substituted into the V1a receptor subtype, and the resulting hybrid receptors were then examined for their ability to mediate increases in intracellular cAMP levels. All studies were carried out with COS-7 cells transiently expressing the different wild type and mutant vasopressin receptors.
[ 3 H]AVP binding studies showed that most of the mutant receptors were expressed at levels comparable to those found with the two wild type receptors (Table II). In only a few rare cases (e.g. CR2), B max values were found to be strongly reduced (see below for details).
A Sequence at the N Terminus of the i4 Domain Contributes to V2 Receptor/G s Coupling Selectivity-We demonstrated in a previous study (22) that a mutant V1a receptor (CR1) in which the i3 loop (Tyr 238 -Phe 300 ; Fig. 1) was replaced with the corresponding V2 receptor sequence (Gln 225 -Leu 274 ; Fig. 1) gained the ability to stimulate adenylyl cyclase with high efficacy (Figs. 2 and 3; Table III). This observation suggested that the i3 loop of the V2 receptor contains major structural elements determining the coupling selectivity of this receptor subtype. However, we also noted that the maximum cAMP response mediated by CR1 was about 25% smaller than that observed with the wild type V2 receptor (Figs. 2 and 3; Table III), suggesting that other V2 receptor domains, besides the i3 loop, are also critical for optimum G s coupling efficiency.
Molecular genetic and biochemical studies with biogenic amine receptors and other classes of GPCRs have shown that residues within the i2 loop and portions of the cytoplasmic tail (i4) also contribute to G protein coupling (2-6). We therefore initially created two additional mutant receptors, CR2 and CR3 (Fig. 2), by substituting V2 receptor sequences into the i2 loop (CR2) or the i4 domain (CR3) of CR1. Fig. 2 shows that maximum cAMP responses (E max ) mediated by CR2 were smaller than those observed with CR1 (E max ϭ 56 Ϯ 7% compared with wild type V2). However, [ 3 H]AVP radioligand binding studies showed that CR2 was poorly expressed (B max ϭ 40 fmol/mg; Table II), making the loss-of-function data difficult to interpret. In contrast to CR2, CR3 (which is expressed at similar levels as CR1; Table II) gained the ability to stimulate adenylyl cyclase with the same efficacy as the wild type V2 receptor, indicating that the last 52 amino acids of the V2 vasopressin receptor (Thr 320 -Ser 371 ; Fig. 1) make a critical contribution to V2 receptor/G s coupling selectivity. Consistent with this finding, substitution of this C-terminal V2 receptor sequence directly into the wild type V1a receptor resulted in a hybrid receptor, CR4, that gained the ability to stimulate cAMP production to a significant extent (E max ϭ 28 Ϯ 4%; Fig. 2).
We next created two additional mutant V1a receptors, CR5 and CR6, in which the N-terminal portion of the i4 domain (including several residues predicted to be located at the cytoplasmic end of TM VII) and the distal portion of the i4 region (the region C-terminal of the conserved Cys-Cys pair) were replaced with the corresponding V2 receptor sequences (Thr 320 -Leu 340 and Leu 339 -Ser 371 , respectively; Fig. 1). Functional studies showed that CR5 was capable of mediating liganddependent cAMP accumulation in a fashion similar to CR4 (E max ϭ 29 Ϯ 1%; Fig. 2). In contrast, CR6 displayed only residual functional activity, similar to the wild type V1a receptor (Fig. 2). When the V2 receptor sequence present in CR5 (Thr 320 -Leu 340 ) was substituted into CR1 (which contains V2 receptor sequence in the i3 loop), the resulting chimeric receptor, CR7, showed a functional profile very similar to the wild type V2 receptor (Figs. 2 and 3). As shown in Fig. 3, AVP concentrationresponse curves generated for CR7 and the wild type V2 receptor were almost superimposable (for E max and EC 50 values, see Table III), further highlighting the functional importance of the Thr 320 -Leu 340 V2 receptor segment.
The N-terminal Portion of the i3 Loop Plays a Key Role in Determining V2 Receptor/G s Coupling Selectivity-We next wanted to examine which specific regions/amino acids within the i3 loop of the V2 vasopressin receptor are of primary importance for proper recognition of G s . Toward this goal, we initially created a series of mutant V1a receptors (CR8 -CR12; Fig. 4) in which distinct segments of the i3 loop were replaced with the corresponding V2 receptor sequences. As shown in Fig. 4, CR9 and CR10 in which C-terminal portions of the i3 loop contained the V2 receptor sequence (residues Gly 239 -Leu 274 and Pro 256 -Leu 274 , respectively) were unable to mediate stimulation of adenylyl cyclase to a significant extent. In agreement with this observation, CR14 in which the entire C-terminal third of the receptor protein (excluding the N-terminal portion of the i3 loop) was derived from the V2 receptor (residues Pro 256 -Ser 371 ) did not display an increase in functional efficacy (E max ϭ 26 Ϯ 5%; Fig. 4), as compared with CR4 or CR5, which contain smaller substitutions within their i4 domains (Fig. 2).
On the other hand, CR8, which contains 30 amino acids of V2 receptor sequence (Gln 225 -Gly 254 ) at the beginning of the i3 loop (including several residues predicted to be located at the cytoplasmic end of TM V) gained the ability to efficiently stimulate ligand-dependent cAMP accumulation (E max ϭ 71 Ϯ 2%; Fig. 4). Comparison of complete AVP concentration-response curves showed that CR8 behaved functionally very similarly to CR1 in which the entire i3 loop consists of V2 receptor sequence (Fig. 5, Table III). Consistent with results described in the previous section, replacement of the i4 domain of CR8 with the corresponding V2 receptor sequence (Thr 320 -Ser 371 ; yielding CR13) led to a further improvement in G s coupling efficacy (E max ϭ 88 Ϯ 5%; Fig. 4).
Progressive shortening of the V2 receptor sequence present  Table I and Fig. 1). The positions of the seven TM domains (I-VII) and the four intracellular receptor domains (i1-i4) are marked. The number of V2 receptor residues present in each construct is indicated above the filled bars. The fact that the two receptors differ in the size of their i3 loops is ignored in this scheme. B, COS-7 cells transiently expressing the various wild type and mutant receptors were stimulated with AVP (1 M), and the resulting increases in intracellular cAMP levels (fold stimulation above basal) were determined as described under "Experimental Procedures." Basal cAMP levels (no ligand added) were similar for the different receptor constructs (data not shown). In the case of the wild type V2 receptor, basal levels amounted to 780 Ϯ 60 cpm/well. In each individual experiment, the wild type V2 receptor response was set equal to 100%. Data are expressed as means Ϯ S.E. of three to nine independent experiments, each carried out in triplicate. in CR8 resulted in CR11 and CR12 (Fig. 4), which contained only 18 (Gln 225 -Glu 242 ) and 12 (Gln 225 -Leu 236 ) V2 receptor residues, respectively. Fig. 4 indicates that both hybrid receptors retained the ability to activate the G s /adenylyl cyclase system (E max ϭ 27-29%), although maximum cAMP responses were reduced compared with CR8 (for a complete functional characterization of CR12, see Table III).
Two Polar Residues at the TM V/i3 Loop Junction Are Critical for Efficient V2 Receptor/G s Coupling-As outlined in the previous section, CR12, which contains only 12 amino acids of V2 receptor sequence (Gln 225 -Leu 236 ), gained significant coupling to G s (Fig. 4). This short V2 receptor sequence is shown enlarged in Fig. 6, aligned with the corresponding V1a receptor segment (both rat and human sequences are shown). Fig. 6 indicates that there are three groups of V2 receptor receptor residues (Gln 225 /Val 226 , Phe 229 -Glu 231 , and His 233 -Ser 235 ; boxed in Fig. 6), which clearly differ in their physicochemical properties from the corresponding amino acids present in the V1a receptor.
To test the potential importance of these V2 receptor residues for efficient G s coupling, we initially substituted these three groups of amino acids directly into the wild type V1a receptor, creating hybrid receptors CR15-17 (Fig. 7A). However, none of these mutant receptors gained significant coupling to the G s /adenylyl cyclase system (Fig. 7B). On the other hand, when we introduced the same substitutions into CR9, which also contained V2 receptor sequence in the central and C-terminal portions of the i3 loop (yielding CR18 -CR20), we identified two mutant receptors, CR18 and CR19, which showed clearly improved cAMP responses (Fig. 7). As shown in Fig. 7B, CR18 (containing V2-Gln 225 and V2-Val 226 ) and CR19 (containing the V2 receptor residues Phe 229 /Arg 230 /Glu 231 ) displayed 2-and 4-fold increases in E max values, respectively, as compared with CR9. No such increase in maximum cAMP responses was observed for CR20 which contains the V2 receptor triplet, His 233 /Ala 234 /Ser 235 (Fig. 7B).
We next used a loss-of-function mutagenesis approach to evaluate the potential functional importance of Gln 225 , Val 226 , Phe 229 , Arg 230 , and Glu 231 in more detail. Starting from the CR8 mutant receptor, which is able to activate G s with high efficacy (Fig. 4), we created five mutant receptors (CR22-CR26) in which these five amino acids were individually replaced with the corresponding V1a receptor residues (Fig. 8A). Three of these mutant receptors (CR23, CR24, and CR25, containing Val 226 3 Gly, Phe 229 3 Cys, and Arg 230 3 Tyr point mutations, respectively) behaved functionally very similarly to CR8 (Fig. 8B). In contrast, two of the introduced point mutations, Gln 225 3 Tyr and Glu 231 3 His, yielded hybrid receptors (CR22 and CR26, respectively) that showed strong reductions (by about 60 -70% as compared with CR8) in receptor-mediated cAMP responses (Fig. 8B). Fig. 4, CR8 (which contains 30 amino acids of V2 receptor i3 loop sequence) displayed an E max that was approximately 2.5fold greater than that observed with CR12 (which contains only 12 amino acids of the V2 receptor i3 loop sequence). This observation suggested that the central portion of the i3 loop of the V2 receptor, in addition to residues at the TM V/i3 loop junction, might also be important for efficient coupling to G s . To further test this hypothesis, we generated a hybrid receptor, CR27 (Fig. 9A), in which the central portion of the i3 loop of CR12 (V1a receptor residues His 258 -Val 278 ) was replaced with V2 receptor residues Gly 245 -Gly 254 (note that the introduced V2 receptor segment is 11 amino acids shorter than the replaced V1a receptor segment, reflecting differences in i3 loop sizes between the two receptor subtypes; Fig. 1). Interestingly, CR27 gained the ability to stimulate cAMP production with markedly increased efficacy (E max ϭ 73 Ϯ 3%), in a fashion similar to CR8 (Fig. 9B).

The Length of the Central Portion of the i3 Loop Plays a Role in Regulating Receptor/G s Coupling Efficiency-As shown in
The introduced V2 receptor segment, Gly 245 -Gly 254 , contains two groups of arginine residues (Fig. 9A), prompting us to  H]AVP saturation binding and cAMP assays were carried out using transfected COS-7 cells as described under "Experimental Procedures." K D values were determined using the program Ligand (28). EC 50 and E max values were obtained from AVP concentration-response curves, using the computer program Kaleidagraph (Synergy Software). Data are given as means Ϯ S.E. of two to four (binding assays) or three to nine (cAMP assays) independent experiments, each carried out in duplicate (binding assays) or triplicate (cAMP assays), respectively. speculate that these positively charged residues might be important for G s coupling. However, groupwise replacement in CR27 of these charged amino acids with alanine residues (resulting in CR28 (Arg 247 3 Ala, Arg 248 3 Ala, Arg 249 3 Ala) and CR29 (Arg 251 3 Ala, Arg 252 3 Ala) (Fig. 9A) did not lead to a reduction in cAMP responses (Fig. 9B). Similarly, simultaneous replacement of all five noncharged residues located within Gly 245 -Gly 254 (CR30) or even replacement of this entire sequence element with a string of 10 alanine residues (CR31) still yielded mutant receptors that allowed very efficient coupling to G s (E max ϭ 78 -95%; Fig. 9B). Fig. 10 shows that AVP concentration-response curves generated for CR27 and CR31 were essentially superimposable (for E max and EC 50 values, see also Table III). As shown in Fig. 1, the central portion of the i3 loop of the V1a receptor is 13 amino acids longer than the corresponding V2 receptor region. Our data therefore suggest that the ability of CR27-CR31 to couple to G s with clearly improved efficacy (as compared with CR12) is primarily due to the shortening of the central portion of the i3 loop rather than the presence of specific V2 receptor residues within this domain.
To further test this hypothesis, two additional hybrid receptors, CR32 and CR33, were constructed (Fig. 9A). CR32 differs from CR27 in that the His 258 -Val 278 (V1a sequence) 3 Gly 245 -Gly 254 (V2 sequence) substitution was introduced directly into  Table I and Fig. 1). The locations of TM I-VII and the i3 loop are indicated. The fact that the two receptors differ in the size of their i3 loops is ignored. The number of V2 receptor residues present in each construct is indicated above the filled bars. For comparison, CR1 was also included in this set of experiments. B, COS-7 cells transiently expressing the various wild type and mutant receptors were stimulated with AVP (1 M), and the resulting increases in intracellular cAMP levels (fold stimulation above basal) were determined as described under "Experimental Procedures." Basal cAMP levels (no ligand added) were similar for the different receptor constructs (data not shown). In each individual experiment, the wild type V2 receptor response was set equal to 100%. Data are given as means Ϯ S.E. of four to eight independent experiments, each carried out in triplicate.  Tables II and III. Results (means) from a representative experiment carried out in triplicate are shown. Note that the curves shown in this figure and in Fig. 3 were generated in the same set of experiments, explaining that the curves for the two wild type receptors and for CR1 are identical in these two figures. Two additional experiments gave similar results. the wild type V1a receptor (rather than into CR12). CR33 is derived from CR1 by replacing Arg 243 -Gly 254 (V2 sequence) with the longer Ser 256 -Val 278 (V1a sequence) segment. Fig. 9B shows that CR32, in contrast to CR27, did not gain efficient coupling to G s and showed only residual functional activity. Moreover, CR33 proved to be considerably less active (reduction in E max by approximately 2.5-fold) than its "parent" receptor, CR1 (Fig. 9B; for a complete AVP concentration-response curve of CR33, see Fig. 10).
Hybrid Receptors Unable to Stimulate cAMP Production Can Couple to PI Hydrolysis-Several of the hybrid V1a/V2 vasopressin receptors analyzed in this study (e.g. CR6, CR9, CR10, CR15-18, CR20, and CR32) exhibited only residual activity in the cAMP assays. To exclude the possibility that this lack of functional activity was due to improper folding of the intracellular receptor surface, these receptors were also tested for their ability to mediate AVP-dependent stimulation of PI hydrolysis.  Table I and Fig.  1). The fact that the two receptors differ in the size of their i3 loops is ignored. CR15-CR17 and CR18 -CR20 are derived from the wild type V1a receptor and CR 9, respectively. The N-terminal portion of the i3 loop is shown enlarged. Numbers next to individual residues refer to amino acid positions within the rat V1a receptor sequence. For comparison, CR9 was also included in this set of experiments. B, COS-7 cells transiently expressing the various wild type and mutant receptors were stimulated with AVP (1 M), and the resulting increases in intracellular cAMP levels (fold stimulation above basal) were determined as described under "Experimental Procedures." Basal cAMP levels (no ligand added) were similar for the different receptor constructs (data not shown). In each individual experiment, the wild type V2 receptor response was set equal to 100%. Data are given as means Ϯ S.E. of three to six independent experiments, each carried out in triplicate.
FIG. 8. Stimulation of cAMP production by hybrid V1a/V2 vasopressin receptors, CR22-CR26. A, structure of the CR22-CR26 mutant receptors (for exact amino acid composition, see Table I and Fig.  1). CR22-CR26 are derived from CR8, which contains 30 amino acids of V2 receptor sequence (Gln 225 -Gly 254 ) at the beginning of the i3 loop. Numbers next to individual residues refer to amino acid positions within the human V2 receptor sequence. V2 receptor residues were systematically substituted with the corresponding V1a receptor residues. As a control, CR8 was also included in this set of experiments. B, COS-7 cells transiently expressing the various wild type and mutant receptors were stimulated with AVP (1 M), and the resulting increases in intracellular cAMP levels (fold stimulation above basal) were determined as described under "Experimental Procedures." Basal cAMP levels (no ligand added) were similar for the different receptor constructs (data not shown). In each individual experiment, the wild type V2 receptor response was set equal to 100%. Data are given as means Ϯ S.E. of three independent experiments, each carried out in triplicate.
A previous study (22) demonstrated that chimeric V1a/V2 receptors can efficiently stimulate PI hydrolysis (mediated by G proteins of the G q /G 11 family) as long as their i2 loops contain V1a receptor sequence. As shown in Fig. 11, all examined mutant receptors that were inefficiently coupled to G s were still able to stimulate the accumulation of inositol phosphates with high efficacy (in the presence of 1 M AVP). In all cases, maximum PI responses were similar to those observed with the wild type V1a receptor (increase in IP 1 production, 22 Ϯ 2-fold above basal).

DISCUSSION
A gain-of-function mutagenesis strategy was employed to elucidate the molecular basis underlying the ability of the V2 vasopressin receptor to selectively stimulate the G s /adenylyl cyclase system. In contrast to the V2 receptor, which mediates a robust increase in intracellular cAMP levels, hormone stimulation of the structurally closely related V1a receptor subtype leaves intracellular cAMP levels virtually unaffected (Figs. 2  and 3). To identify V2 receptor residues determining this coupling selectivity, we systematically replaced distinct V1a receptor sequences (or single amino acids) with the corresponding V2 receptor sequences and studied whether the resulting hybrid receptors would gain the ability to mediate ligand-dependent cAMP accumulation.
Consistent with a previous study (22), we found that substitution of the i3 loop of the V2 receptor into the V1a receptor subtype yielded a mutant receptor (CR1) that gained the ability to efficiently stimulate cAMP production (E max about 75%, as compared with the wild type V2 receptor; Fig. 2). Interestingly, of the hybrid receptor genes. For comparison, CR1, CR8, and CR12 were also included in this set of experiments. B, COS-7 cells transiently expressing the various wild type and mutant receptors were stimulated with AVP (1 M), and the resulting increases in intracellular cAMP levels (fold stimulation above basal) were determined as described under "Experimental Procedures." Basal cAMP levels (no ligand added) were similar for the different receptor constructs (data not shown). In each individual experiment, the wild type V2 receptor response was set equal to 100%. Data are given as means Ϯ S.E. of three to seven independent experiments, each carried out in triplicate.  Table I and Fig. 1). Numbers next to individual residues refer to amino acid positions within the rat V1a and the human V2 receptor sequences. The central portion of the i3 loop is shown enlarged. CR27-CR31 are derived from CR12, which contains 12 amino acids of V2 receptor sequence at the beginning of the i3 loop. In CR27, the V1a receptor sequence His 258 -Val 278 (21 amino acids) was replaced with the V2 receptor segment, Gly 245 -Gly 254 (10 amino acids) (see also Fig. 1). In CR32, the same substitution was introduced into the wild type V1a receptor "background." CR33 was generated by introducing a reciprocal substitution into the CR1 hybrid receptor. Please note that the sequences that were exchanged in CR32 and CR33 slightly differ in length, due to technical reasons (presence of useful restriction sites) relating to the construction a mutant receptor (CR3) that contained V2 receptor sequence in both the i3 and i4 domains was able to activate the G s / adenylyl cyclase system with the same high efficacy as the wild type V2 receptor (E max ϭ 99 Ϯ 1%; Fig. 2), suggesting that the i4 domain of the V2 receptor (Thr 320 -Ser 371 ) also contains residues that are critical for efficient G s activation. Functional analysis of the CR7 hybrid receptor showed that a stretch of 21 amino acids of V2 receptor sequence (Thr 320 -Leu 340 ) located at the TM VII/i4 domain junction can fully mimic the effects of the "Thr 320 -Ser 371 substitution" (Figs. 2 and 3).
Substitution of Thr 320 -Leu 340 or Thr 320 -Ser 371 directly into the wild type V1a receptor led to mutant receptors (CR4 and CR5, respectively) that also gained the ability to activate the cAMP cascade (E max about 30%; Fig. 2). The ability of CR4 (previously referred to as V1i4) (22) to mediate a small but significant cAMP response was not noticed in our initial analysis of V1a/V2 hybrid receptors (22). One possible explanation for this discrepancy is that a modified transfection procedure was used in the present study (lower cell plating density, 1 ϫ 10 6 versus 2 ϫ 10 6 cells/plate, prolonged transfection time, 3 h versus 1 h).
The Thr 320 -Leu 340 V2 receptor sequence is followed by a pair of cysteine residues which are conserved among all vasopressin receptor subtypes (Fig. 1). Similarly, most GPCRs contain one or more highly conserved cysteine residues within their Cterminal i4 domains. Studies with a great number of different GPCRs, including the V2 vasopressin receptor (32), have shown that these cysteines are modified by covalent attachment of palmitic acid (for reviews, see Refs. 33 and 34), which may provide a lipophilic membrane anchor to create a fourth intracellular loop (i4 loop). Our data therefore suggest that this putative i4 loop makes an important contribution to V2 receptor/G s coupling selectivity.
Studies with hybrid ␣ 2A /␤ 2 -adrenergic (35) and hybrid mGluR1/mGluR3 glutamate receptors (36,37) also suggest that the membrane-proximal portion of the i4 domain plays a role in regulating the selectivity of G protein recognition in other GPCR families. In both cases, however, it was noted that substitution of this region alone of a donor receptor into a functionally different recipient receptor was not sufficient to confer a novel coupling profile, but required additional substitutions involving i2 or i3 loop sequences. To the best of our knowledge, the present study therefore provides the first example that exchange of the i4 loop alone between two functionally distinct receptor subtypes can lead to mutant receptors with qualitatively different G protein coupling profiles.
Another goal of this study was to examine which subdomains/specific amino acids within the i3 loop of the V2 receptor (Gln 225 -Leu 274 ) are of primary importance for efficient G s recognition and activation. To address this issue, we created and analyzed a series of mutant V1a receptors in which distinct segments of the i3 loop were systematically replaced with the corresponding V2 receptor sequences. These studies showed that the C-terminal segments of the i3 loop of the V2 receptor do not make a significant contribution to G s coupling selectivity and efficiency (see, for example, CR9 and CR10; Fig. 4). The relative lack of G s coupling seen with CR9 and CR10 is unlikely to be due to improper folding of the intracellular receptor surface, since both receptors retained the ability to mediate the stimulation of PI hydrolysis with high efficacy (Fig. 11).
In contrast, substitution into the V1a receptor of V2 receptor sequences located at the N terminus of the i3 loop (Ni3) allowed the resulting hybrid receptors to activate the G s /adenylyl cyclase system. A mutant V1a receptor containing 30 amino acids of V2 receptor Ni3 sequence (Gln 225 -Gly 254 ; CR8) quantitatively mimicked the cAMP responses mediated by CR1 (E max ϭ 71-76%; AVP EC 50 ϭ 1.1-1.2 nM; Figs. 4 and 5; Table III), suggesting that this short receptor segment contains the key structural elements of the V2 receptor i3 loop that are critical for G s coupling.
CR12, which contains only 12 amino acids of V2 receptor Ni3 sequence (Gln 225 -Leu 236 ), still retained the ability to activate the cAMP pathway, although with reduced efficacy (E max about 30%; Fig. 4; Table III). A sequence comparison showed (Fig. 6) that this sequence element contains three groups of amino acids (Gln 225 /Val 226 , Phe 229 -Glu 231 , and His 233 -Ser 235 ), which clearly differ in their physicochemical properties from the corresponding residues present in the V1a receptor. Groupwise substitution of these amino acid pairs/triplets into CR9 (which contains the V2 receptor sequences in the central and C-terminal portions of the i3 loop) suggested that residues contained within the Gln 225 /Val 226 pair and the Phe 229 /Arg 230 /Glu 231 triplet make an important contribution to G s coupling efficiency (Fig. 7). Using the CR8 mutant receptor (which contains 30 amino acids of V2 receptor Ni3 sequence) as a "background" for single amino acid substitutions, loss-of-function mutagenesis studies showed that only two of these five residues, Gln 225 and Glu 231 , are required for efficient activation of G s (Fig. 8).
Based on biochemical, molecular genetic, and biophysical studies with other classes of GPCRs, Gln 225 and Glu 231 are thought to be located in a receptor region (TM V/i3 loop junction) predicted to be ␣-helically arranged (4,6,38). Guided by an improved low resolution electron density map of frog rhodopsin (39) and structural information gathered from the analysis of approximately 500 different GPCRs, Baldwin et al. (40) recently proposed an updated model for the ␣-carbon positions in the seven TM helices of CPCRs of the rhodopsin family (which includes the vasopressin receptors). In this model, Gln 225 and Glu 231 are predicted to project into a cavity formed by TM III, V, and VI. Another functionally critical residue projecting into this cavity is Arg 137 (located at the cytoplasmic end of TM III), which is conserved in most GPCRs of the rhodopsin family and plays a key role in triggering G protein activation (2-6). Rosenthal Table I and Figs. 2, 4, 7, and 9. COS-7 cells transiently expressing the wild type V1a and the indicated mutant receptors were stimulated with AVP (1 M), and the resulting increases in intracellular IP 1 levels (fold stimulation above basal) were determined as described under "Experimental Procedures." Basal IP 1 levels (no ligand added) were similar for the different receptor constructs (data not shown). In the case of the wild type V1a receptor, basal IP 1 levels amounted to 302 Ϯ 39 cpm/well. In each individual experiment, the IP 1 response mediated by the wild type V1a receptor was set equal to 100%. Data are given as means Ϯ S.E. and are representative of three independent experiments, each carried out in triplicate. that mutational modification of this residue in the V2 vasopressin receptor completely abolishes V2 receptor/G s coupling. Arg 137 , Gln 225 , and Glu 231 are therefore likely to define a distinct site on the intracellular surface of the V2 receptor that is critical for G s recognition and activation.
In agreement with the data presented here, studies with other classes of GPCRs including the muscarinic (reviewed in Wess (8)) and adrenergic receptors (reviewed in Refs. [2][3][4][5] have also demonstrated that residues at the TM V/i3 loop junction are critically involved in regulating receptor/G protein-coupling selectivity. These studies have shown that the residues that are of primary importance for proper G protein recognition are hydrophobic or noncharged. Interestingly, the functionally critical V2 receptor residues, Gln 225 and Glu 231 , are highly polar and charged, respectively, indicating that the contribution of the Ni3 region to receptor/G protein coupling selectivity is not limited to hydrophobic contacts. Comparison of the functional properties of CR8 (containing 30 amino acids of V2 receptor sequence; E max ϭ 71%) and CR12 (containing 12 amino acids of V2 receptor sequence; E max ϭ 29%) suggested that the central portion of the i3 loop also makes a contribution to V2 receptor/G s coupling efficiency. Consistent with this notion, substitution of Gly 245 -Gly 254 (V2 receptor sequence) into CR12 yielded a hybrid receptor (CR27) that gained the ability to stimulate intracellular cAMP levels with the same high efficacy as CR8 ( Fig. 9; Table III). This sequence element (Gly 245 -Gly 254 ) is particularly rich in arginine and glycine residues (Fig. 9A). However, systematic alanine substitution mutagenesis showed that these residues are not critical for the ability of CR27 to efficiently mediate G s activation. Most strikingly, a CR27-derived mutant receptor in which the Gly 245 -Gly 254 sequence was replaced with a string of 10 alanine residues (CR31) was able to mediate AVP-dependent cAMP production with the same high efficacy and AVP potency as CR27 (Figs. 9 and 10; Table III). Fig. 1 indicates that the central portion of the i3 loop of the V1a receptor is 13 amino acids longer than the corresponding V2 receptor region. Thus, during the construction of CR27 (as well as of CR28 -CR31; Fig. 9A), 21 amino acids of the V1a receptor sequence (His 258 -Val 278 ) were replaced with only 10 V2 receptor residues (Gly 245 -Gly 254 ). This observation, in addition to the observed functional profiles of CR27-CR31, prompted us to speculate that the relative length of the central portion of the i3 loop (rather than the specific amino acid sequence of this region) may play a role in regulating receptor/G s coupling selectivity. To further test this hypothesis, two additional mutant V1a/V2 receptors, CR32 and CR33, were constructed and functionally analyzed (Fig. 9). CR32, which contains the His 258 -Val 278 3 Gly 245 -Gly 254 substitution in the wild type V1a receptor background, did not gain the ability to couple to the G s /adenylyl cyclase system to a significant extent. Since CR27 (which was highly active in the cAMP assays) differs from CR32 only in the presence of 12 amino acids of V2 receptor sequence at the beginning of the i3 loop, this observation suggests that shortening of the central portion of the i3 loop by itself is not sufficient to allow efficient G s coupling but requires the simultaneous presence of V2 receptor residues at the TM V/i3 loop junction (see above). However, when the central portion of the i3 loop of CR1 (E max ϭ 76%) was extended in length by replacing Arg 243 -Gly 254 (V2 receptor sequence) with Ser 256 -Val 278 (V1a receptor sequence), the resulting hybrid receptor (CR33) displayed a pronounced loss of activity in the cAMP assays (E max ϭ 33%) (Fig. 9). Taken together, these results are consistent with the novel concept that the central portion of the i3 loop, although predicted not to be directly involved in receptor/G protein interactions, can modulate recep-tor/G protein coupling selectivity by regulating G protein access (e.g. via steric hindrance) to functionally important recognition sites on the receptor protein (such as the TM V/i3 loop junction).
In conclusion, the structural elements determining the ability of the V2 vasopressin peptide receptor to selectively activate G s were studied in molecular detail. Our findings indicate that V2 receptor coupling selectivity depends on several different structural features, some of which have not been observed previously in studies using other classes of GPCRs. Our work highlights the diversity of mechanisms by which receptor/G protein coupling selectivity can be achieved.