The human VPAC1 receptor: three-dimensional model and mutagenesis of the N-terminal domain.

The human VPAC(1) receptor for vasoactive intestinal peptide (VIP) and pituitary adenylate cyclase activating peptide belongs to the class II family of G-protein-coupled receptors with seven transmembrane segments. Like for all class II receptors, the extracellular N-terminal domain of the human VPAC(1) receptor plays a predominant role in peptide ligand recognition. To determine the three-dimensional structure of this N-terminal domain (residues 1-144), the Protein Data Bank (PDB) was screened for a homologous protein. A subdomain of yeast lipase B was found to have 27% sequence identity and 50% sequence homology with the N-terminal domain (8) of the VPAC(1) receptor together with a good alignment of the hydrophobic clusters. A model of the N-terminal domain of VPAC(1) receptor was thus constructed by homology. It indicated the presence of a putative signal sequence in the N-terminal extremity. Moreover, residues (Glu(36), Trp(67), Asp(68), Trp(73), and Gly(109)) which were shown to be crucial for VIP binding are gathered around a groove that is essentially negatively charged. New putatively important residues for VIP binding were suggested from the model analysis. Site-directed mutagenesis and stable transfection of mutants in CHO cells indicated that Pro(74), Pro(87), Phe(90), and Trp(110) are indeed important for VIP binding and activation of adenylyl cyclase activation. Combination of molecular modeling and directed mutagenesis provided the first partial three-dimensional structure of a VIP-binding domain, constituted of an electronegative groove with an outspanning tryptophan shell at one end, in the N-terminal extracellular region of the human VPAC(1) receptor.


VPAC 1 receptor.
The VPAC 1 receptor for the neuropeptides vasoactive intestinal peptide (VIP) 1 and pituitary adenylate cyclase activating peptide is a class II G protein-coupled receptor (1). Together with the VPAC 2 receptor subtype, they mediate a large array of VIP or pituitary adenylate cyclase activating peptide actions on exocrine secretions, release of hormones, relaxation of muscles, metabolism, growth control of fetuses, and tumor cells and embryonic brain development (2,3). Class II receptors for peptides have low sequence homologies with other members of the superfamily of G protein-coupled receptors (1,4). They share several specific properties, the most important of which is the presence of large N-terminal extracellular domains which contain 10 highly conserved amino acids including six cysteines, putative N-terminal leader sequences and several potential N-glycosylation sites (1, 4 -6). A complex gene organization with many introns is also common to all class II receptors (5). We know from mutagenesis studies that the N-terminal extracellular domain plays a dominant, although not exclusive, role in determining the peptide ligand binding affinity (1, 4 -6). However, no structure of the large N-terminal extracellular domain of the class II receptors is yet available.
The human VPAC 1 receptor has been extensively characterized by site-directed mutagenesis and construction of receptor chimeras (7)(8)(9)(10)(11)(12)(13)(14). The data demonstrate an important role of the N-terminal extracellular domain constituted of 144 amino acid residues: in this fragment, Glu 36 , Trp 67 , Asp 68 , Tryp 73 , Gly 109 , and Lys 143 are crucial for VIP binding affinity (7,10,11,13); six cysteines are required to ensure VIP binding (8); two (Asn 58 or Asn 69 ) out of the three N-glycosylation sites play a mandatory role for correct delivery of the receptor to the plasma membrane (9). Other domains that are functional for VIP binding (8,14) or peptide selectivity (13) have been mapped in extracellular loops and in the third transmembrane domain of human VPAC 1 receptor. Recently, constitutively active mutants of human VPAC 1 receptor have been produced after point mutagenesis of Arg (15) or Thr (16) residues of the second and fourth transmembrane domain, respectively.
The residues (Glu 36 , Trp 67 , Asp 68 , Trp 73 , Gly 109 , and Lys 143 ) which are crucial for VIP binding affinity are dispersed along the primary sequence of the N-terminal extracellular region of the human VPAC 1 receptor. We question whether their spatial distribution is responsible for their functional properties. In this paper, we have developed the first threee-dimensional model of a large part of the N-terminal domain of the human VPAC 1 receptor by homology modeling. Although the already available data from site-directed mutagenesis experiments (7,11,13) were not used as modeling constraints, the structure obtained does accommodate them and generates new hypotheses regarding the possible role of several amino acid residues. Hypotheses were experimentally tested by mutagenesis. Altogether, molecular modeling and mutagenesis suggest that an electronegative groove topped at one end by a tryptophan shell could constitute a partial VIP-binding domain in the human VPAC 1 receptor.

EXPERIMENTAL PROCEDURES
Materials-Enzymes and vectors were obtained from Promega (Charbonnière, France) and culture medium from Life Technologies, Inc. (Cergy-Pontoise, France). Synthetic VIP was from Neosystem (Strasbourg, France). 125 I-VIP was prepared and purified as described (17). Other highly purified chemicals used were from Sigma (Saint-Quentin-Fallavier, France).
Molecular Modeling-A three-dimensional model of the 1-144 Nterminal domain of the human VPAC 1 receptor was calculated using the Modeller 4.0 software (18,19). This method uses sequence homology between the receptor protein domain and a protein whose three-dimensional structure is known to predict a three-dimensional model. To find an homologous sequence to the VPAC 1 receptor 1-144 sequence, we screened the Protein Data Bank using the FASTA algorithm (20). The 152-262 region of a yeast lipase was selected (Protein Data Bank code: 1LBS), carrying 27% sequence identity and 50% sequence homology with residues 8 -117 of the VPAC 1 receptor. To further check the homology, the HCA (hydrophobic cluster analysis) plots of both sequences were compared using the Visualfasta program (Patrick Durand, Paris, France). Briefly, the HCA method is based on a two-dimensional helical plot of the sequence, allowing detection of hydrophobic clusters in proteins (21). Alignment of hydrophobic clusters mostly supports homologous folds. The quality of alignment is evaluated by the HCA score. In this case, the score is 60%. The resulting alignment is used as input for Modeller 4.0. The three-dimensional structure d1 calculated by Modeller4 is then minimized using the Hyperchem 5.0 software (Hypercube Inc.). The energy minimization is carried out using the conjugate gradient method with the AMBER forcefield. The stereochemical quality of the minimized three-dimensional model is finally checked using Procheck (22). The model has 88% of the ⌽,⌿ angle pairs in the allowed regions of the Ramachandran plot, indicating a correct stereochemistry. Electrostatic potentials were calculated and drawn using CHIME (MDL Information system Inc.). Molecular views were drawn with the WinMGM program (23).
Site-directed Mutagenesis Experiments-Oligonucleotide-directed mutagenesis was performed as previously described (4). Identification of the desired mutations was obtained by direct double-stranded sequencing of the mutated region (7). To control the correct expression of the wild-type and mutated VPAC 1 receptor in transfected cells, each construction was fused in the C-terminal position with the green fluorescent protein (GFP) in the eukaryote expression vector pEGFP-N2 (CLONTECH, Palo Alto, CA) as described (11). We previously demonstrated that the presence of GFP at the C terminus of VPAC 1 receptor does not modify its phenotype with respect to VIP binding and activation of adenylyl cyclase (11). For mutants with null phenotypes, new constructs were developed in which a Flag sequence was inserted after the leader peptide between Ala 30 and Ala 31 as described (15). It was previously shown that this insertion does not modify the dissociation constant for VIP or the dose response of VIP in stimulating cAMP production as compared with the native VPAC 1 receptor (15). This extracellular Flag made it possible to directly assess the cell surface expression of mutated receptors (15) by immunofluorescence and anti-Flag antibody binding experiments (see below).
Cell Transfection-Wild-type and mutated VPAC 1 receptors were stably transfected into CHO cells as previously described (25) using 3 l of Fugen-6 (Life Technologies, Cergy-Pontoise, France) and 2 g of DNA construct. After transfection and 48 h of culture, GFP fluorescence was observed to estimate receptor expression as previously reported (11). CHO cells were selected in the presence of geneticin (G418) at a final concentration of 0.8 g/liter for 4 days, then they were grown for 3-4 days without G418. After a second round of G418 selection (0.8 g/liter) for 4 days, CHO cells were grown without G418 and used for fluorescence studies (see below) or for membrane preparation as described (7).
Ligand Binding Assay and Measurement of Adenylyl Cyclase Activity-The binding properties of the wild-type and mutated VPAC 1 recep-tors were analyzed by competitive inhibition of 125 I-VIP binding to transfected cell membranes by unlabeled VIP as previously described (17). Binding parameters (K d and B max ) were determined using the LIGAND computer program (26). Adenylyl cyclase activity was assayed as previously reported (27). Each experiment consisted of a full concentration-response curve and the concentration of VIP which induced half-maximal response (EC 50 ) was determined. Protein content in membrane preparations was evaluated by the procedure of Bradford (28) using bovine serum albumin as standard.
Fluorescence Studies-Microscopy immunofluorescence studies on transfected CHO cells were performed on nonpermeabilized cells using anti-Flag antibodies as described (15). Briefly, transfected cell suspensions are incubated with the mouse monoclonal anti-Flag antibodies diluted in phosphate-buffered saline containing 1% (w/v) bovine serum albumin, washed, and then incubated with anti-mouse immunoglobulin G conjugated to rhodamine. Cells are then collected directly on microscope slides by centrifugation at 700 rpm for 10 min (Cytospin3, Shandon Pittsburgh PA) and selected fields are observed using a Leica DM IRB microscope. The fluorescence of GFP in stably transfected living cells was observed directly on a Leica DM IRB microscope. All images were obtained and treated with the Archimed Pro software (Micromécanique, Evry, France).
Assessment of Cell Surface Expression of Mutated Receptors-Cell surface expression of mutated receptors with null phenotypes was assessed using the mouse monoclonal anti-Flag antibodies as described (15) with minor modifications. Briefly, transfected cells were incubated with anti-Flag antibodies diluted in phosphate-buffered saline containing 1% (w/v) bovine serum albumin, then washed and incubated in presence of the radiolabeled second antibodies ( 125 I-labeled goat antimouse whole IgG). The radioactivity was determined in cell lysates. Nonspecific binding was determined with cells only incubated with radiolabeled second antibodies. Binding of anti-Flag antibodies to epitope-tagged mutant receptors was given as a percentage of specific anti-Flag antibodies binding to epitope-tagged wild type receptor (15).

RESULTS AND DISCUSSION
The N-terminal domain of the VPAC 1 receptor is isolated from the rest of the protein in the extracellular medium. It has a predicted length of 144 residues (24) and thus is large enough to be folded as an independent domain (29). We looked for a protein sharing sequence homology with the N-terminal sequence of the VPAC 1 receptor in the Protein Data Bank that contains proteins with known three-dimensional structures. Using the FASTA algorithm (20), the 152-262 fragment of Candida antarctica yeast lipase B (PDB code 1LBS) was selected. This region is a subdomain of the lipase with a self fold. It shares 27% sequence identity and 50% sequence homology with the 8 -117 region of the VPAC 1 receptor (Fig. 1A) which will be referred to as VPAC 1 -(8 -117), the sequence homology and identity, the good alignment of hydrophobic clusters, and the similar distribution of proline and glycine residues validate the calculation of a three-dimensional model of the 8 -117 region of the VPAC 1 receptor based on the crystal structure of the C. antarctica yeast lipase B.
A three-dimensional model of VPAC 1 -(8 -117) was therefore built using Modeller 4.0 (18,19), refined by energy minimization, and validated by Procheck (22) (Fig. 2, A and B). The local root mean square distance between the VPAC 1 -(8 -117) threedimensional model obtained and the lipase subdomain is 3.8 Å, indicating a very similar fold of the two protein subdomains. This is illustrated in Fig. 2 which shows a three-dimensional image of the lipase structure versus the three-dimensional model of VPAC 1 -(8 -117). The VPAC 1 -(8 -117) structure shows an overall good compactness, except for the N-terminal extremity (residues 8 -30) which looks as a separated, mostly helical structure (Fig. 2B). This fragment was previously hypothezised to be a signal peptide (24). The model agrees with and even reinforces the suggestion. Furthermore, the hydrophobicity of the 11-23 fragment is sufficient to support an interaction with a membrane and furthermore, the structure is mobile around the loop located at residues 24 -28. The main secondary structure of the 8 -117 model is coil since the ␣ helix percentage is 29% and the ␤ structure represents 7%. This is due to the fact that the lipase subdomain taken as template is poor in ␣ and ␤ structures, with 35 and 9%, respectively.
In previous work (1), we have shown that the N-terminal domain of the human VPAC 1 receptor plays an important role in the binding of its natural ligand, VIP. Within the 8 -117 region of the receptor, many mutants were previously characterized (7,8,11,13). Five residues, all dispersed along the N-terminal sequence of the VPAC 1 receptor are crucial for VIP binding. They include Glu 36 , Trp 67 , Asp 68 , Trp 73 , and Gly 109 whose mutation into alanine (7,11,13) or glycine (7) or whose deletion (7) completely abolished VIP binding and adenylyl cyclase stimulation (see Table I). All those residues are dispersed in the primary sequence ( Fig. 1) but are clearly gathered around a groove in the three-dimensional model (Fig. 3A). Conversely, residues whose mutation did not alter the receptor phenotype (see legend to Fig. 3) are randomly mapped in the three-dimensional structure with no preferential localization in the groove (Fig. 3A). These observations strongly suggest that the groove is a VIP-binding site of the VPAC 1 receptor (Fig.  3A). Using CHIME, the electrostatic potential of the threedimensional model was calculated (Fig. 3B). The highly negatively charged nature of the groove (red potential on Fig. 3B) reinforces the hypothesis that it could be a VIP-binding site since VIP has several positively charged residues (Arg 12 , Arg 14 , Lys 15 , Lys 20 , and Lys 21 ). A previous work supported that Arg 14 , Lys 15 , and Lys 21 of VIP directly participate in the binding of the peptide to the human VPAC 1 receptor (25). Beside an interesting concentration of charged residues, we also noticed the presence of three tryptophans (Trp 67 , Trp 73 , and Trp 110 ) all gathered on top of the electronegative groove (Fig. 4). Two of those residues (Trp 67 and Trp 73 ) were previously shown to be crucial for the VIP recognition (see Table I). Another interesting feature is the proximity of two Phe residues (Phe 90 and Phe 93 ) (Fig. 4). Those amino acids could favor ⌸-⌸ interactions with their congeners in VIP (Phe 6 , Try 10 , and Tyr 22 ). Indeed, those residues were suggested to be important for VIP structure and/or VIP binding to VPAC 1 receptor (25).
To further validate the model, the mutation to alanine of several residues of the negative groove is proposed (Fig. 4). We checked by fluorescence analysis of GFP in living transfected CHO cells that all mutants were expressed in transfected cells as with the wild-type receptor (not shown). The VIP binding parameters (K d and B max ) and the stimulation of adenylyl cyclase activity (EC 50 ) were then measured in stably transfected CHO cells. Table I summarizes the experimental results.
To determine the pattern of expression of mutants with null phenotypes (see Table I) in transfected CHO cells, immunofluorescence and antibody binding experiments were performed. Indeed, insertion of a Flag sequence between Ala 30 and Ala 31 in the N-terminal extracellular domain of the VPAC 1 receptor enabled us to perform indirect immunofluorescence studies and to assess cell surface expression of receptors by anti-Flag antibody binding in nonpermeabilized transfected CHO cells. Microscopy revealed that all mutants studied were delivered at the cell surface like the wild type receptor (Fig. 5). Since immunofluorescence techniques are not quantitative, we also measured cell surface expression of mutants by antibody binding to transfected cells (Table II). It appeared that mutants with null phenotypes did not exhibit any decrease in cell surface expression as compared with that of the wild type receptor. Finally, it should be stated that mutants which exhibit partial activity including P87A and F90A have normal expression since Scatchard analysis of binding data indicated binding capacities similar to that of the wild type receptor (Table I). Several categories of mutants were characterized.
(i) The mutation of the above mentioned aromatic residues including Phe 90 , Phe 93 , and Trp 110 as well as the mutation of Trp 73 that was previously changed for a glycine were tested. The mutant F90A had a decreased binding affinity for VIP and potency of VIP in stimulating cyclase activity. In contrast, the mutation of F93A did not alter significantly the receptor phenotype supporting that only one aromatic residue (Phe 90 ) would help the binding of VIP. On the other hand, Trp 73 and Trp 110 are crucial since the corresponding mutants W73A and W110A exhibit no VIP binding and no VIP-stimulated adenylyl cyclase (Table I). From these data, it can be concluded that all the tryptophan residues located above the electronegative groove are crucial for VIP binding including Trp 67 (11), Trp 73 , and Trp 110 (this paper). They appear to form a shell at one end of the groove (Fig. 6) and could serve different purposes. It could be involved in the structure of the VIP-binding site in the extracellular N-terminal domain of the VPAC 1 receptor. This would be in good agreement with the facts that individual mutation of each of the three tryptophans results in a drastic loss of function (see Table I) and that these tryptophans are highly conserved in the class II family of G protein-coupled receptors (1). Since the exposition of the three tryptophans to the water environment is unlikely, they could interact with other domains of the VPAC 1 receptor. Alternatively, it could be suggested that the aromatic environment that is formed by the tryptophan shell acts as an anchor for VIP since favorable hydrogen bonding can occur between Trp and aromatic residues (Phe, Tyr, Trp, and His) (30). In the VIP peptide, four residues were previously shown to be important for receptor binding: His 1 , Phe 6 , Tyr 10 , and to a lesser extent, Tyr 22 (25). For aromatic residues present in the N-terminal domain of VIP, especially His 1 , this interaction could represent an initial step in the receptor recognition. On this assumption, the Nterminal domain of VIP could initially bind to the groove described herein and then be recognized by a transmembrane binding pocket as suggested by recent data suggesting that Glu 3 of VIP could be in close proximity to transmembrane segment 2 of the VPAC 1 receptor (31).
(ii) The model suggests that Pro 74 and Pro 87 could be important for the groove structure (Fig. 4). The results demonstrate that the P74A receptor mutant has no detectable binding or stimulation of adenylyl cyclase (Table I). The P87A receptor has a decreased binding affinity for VIP and consequently VIP was less potent to stimulate the cyclase activity (Table I). A previous mutation of Pro 87 into glycine had no effect on the receptor phenotype (Ref. 7 and Table I). It can be argued that Pro 87 is less important than Pro 74 in generating a break in the secondary structure and that its mutation into glycine could accommodate a correct groove structure. This is in line with the fact that proline and glycine are both able to generate breaks in   secondary structures of proteins (32).
(iii) Glu 36 and Asp 68 , previously identified as important for VIP binding to its receptor (7,13), are located in the groove (Fig. 4). While Glu 36 is rather accessible and available for the interaction with VIP, Asp 68 appears more deeply buried in the groove. It could be suggested that Asp 68 which is highly conserved in class II G protein-coupled receptors (7) helps to maintain the structure of the VIP-binding domain possibly through the formation of a salt bridge. The partner of Asp 68 in this putative salt bridge remains to be determined. On the other hand, one can assume that the docking of VIP should induce dynamically structural reorganization of the groove, such as its opening. Asp 68 could then be more accessible and interact with one of the positively charged residues of the VIP peptide. During this mechanism, the desolvation energy needed to render Asp 68 available for an electrostatic interaction with a VIP residue would be greatly decreased as compared with what would be required if Asp 68 was water-accessible in the groove. Two other negatively charged residues are located in the groove: Asp 38 and Glu 108 . While Asp 38 seems accessible for an interaction with VIP (Fig. 4), the model suggests that the Glu 108 residue could form a salt bridge with Lys 65 (mean distance Lys 65 -Glu 108 : 4 Å). However, no change of the receptor phenotype was observed with E108A or K65A mutants (Table  I) supporting the idea that Glu 108 and Lys 65 do not participate in the VIP recognition. These data also suggest that the putative salt bridge between Glu 108 and Lys 65 would not be important for maintaining the structure of the VIP-binding site. Neither does expression of D38A support any significant role for Asp 38 in VIP binding (Table I). This seems contradictory with the idea that the VIP binding occurs in the negative Gray wireframe represents residues whose mutation into alanine does not change the binding affinity of the VPAC 1 receptor for VIP (Gln 41 , Ile 43 , Gln 46 , His 47 , Lys 48 , Gln 49 , Glu 53 , Gln 55 , Glu 57 , Asn 58 , Glu 59 , Asn 69 , and Ser 102 ). Yellow wireframe represents amino acids whose mutation into alanine abolishes VIP binding (Glu 36 , Trp 67 , Asp 68 , Trp 73 , and Gly 109 ). See text and Table I for details. NH 2 and COOH ends are indicated. B, electrostatic potential on the molecular surface of the three-dimensional model of the VPAC 1 -(8 -117) receptor domain. Colors are: blue, positively charged surface; white, neutral surface; red, negatively charged surface. The arrows show the electronegative groove containing the important amino acids described in A.

FIG. 4. Ribbon representation of the three-dimensional model of the VPAC 1 -(8 -117) receptor domain showing amino acids (purple wireframe) whose role in VPAC 1 receptor function has been tested by site-directed mutagenesis in the present study.
Yellow wireframe represents amino acids whose mutation into alanine was previously shown to abolish VIP binding. NH 2 and COOH ends are indicated.

FIG. 5. Microscopic detection after transfection in CHO cells of the epitope-tagged wild type receptor and receptor mutants with null phenotypes.
Nonpermeabilized cells were incubated with the mouse monoclonal anti-Flag antibodies, washed, incubated with antimouse immunoglobulin G conjugated to rhodamine and pelleted on slides as described under "Experimental Procedures." Controls were carried out with untransfected CHO cells or with CHO cells expressing the epitope-tagged wild type receptor that were incubated only with the rhodamine-labeled antimouse antibody. No fluorescence could be observed in both conditions (not shown).
groove. However, a better insight into the model shows that the experimentally crucial negative residues Glu 36 and Asp 68 are in the same plane in the groove while Asp 38 is beneath this plane (Fig. 7). This suggests the existence of a preferential plane of interaction between the VIP and the N-terminal domain of the VPAC 1 receptor. Two observations further argue for the existence of this preferential plane. First, the mutation of Met 66 whose side chain extends in the groove, in the same plane as Glu 36 and Asp 68 (Fig. 7) clearly results in a complete loss of VIP binding and adenylyl cyclase activation (Table I).
This highlights the importance of the methionine side chain. Second, a previous study strongly suggested that Arg 14 , Lys 15 , and Lys 21 of VIP are directly participating in the binding of the peptide to the human VPAC 1 receptor (25). In the modeled structure of the peptide (25), those residues are almost on the same side of the helix (especially Arg 14 and Lys 21 ) and their lateral chains point in the same direction. This could suggest that these residues can interact with Glu 36 and Asp 68 in the groove. In agreement with this hypothesis is the distance between the C␣ of Arg 14 (and to a lesser extent, Lys 15 ) and the C␣ of Lys 21 in VIP that is approximately 11 Å, equal to the distance between the C␣ of Glu 36 and the C␣ of Asp 68 (Asp Ϸ 12 Å) in the VPAC 1 receptor.
(iv) Three basic residues Arg 78 , Lys 91 , and Arg 99 are located on the edges of the groove (Lys 91 and Arg 99 ) or behind the tryptophan shell (Arg 78 ; see Fig. 4). They appear to extend their side chains outside the groove. According to the model they should not be involved in the VIP binding and indeed the mutation of these basic residues did not modify VIP binding (Table I).
(v) Asp 68 and Gly 109 (Fig. 4), which are highly conserved in class II G protein-coupled receptors (7), were previously shown to be very important residues for VIP binding on the basis of their deletion or mutation into glycine (7). As a control, D68A and G108A were expressed in this study (Table I). The phenotype of the new mutants confirmed the previous data since a complete loss of VIP binding and adenylyl cyclase activation was observed.
The very good correspondence between the model and the data is especially convincing since the structural model of the VPAC 1 -(8 -117) region was initially generated on a pure sequence homology basis disregarding any of the available experimental data. The model not only clustered all the functionally important residues dispersed in the primary sequence within a groove, but it also pointed out new putatively important residues. We tested those new targets which proved to be involved either in the structure and/or binding function of the VPAC 1 receptor. Analysis of the experimental data obtained by sitedirected mutagenesis using the model as frame provides the

null phenotypes after stable transfection of cDNAs into CHO cells
Nonpermeabilized transfected cells were incubated with anti-Flag antibodies and then exposed to the radiolabeled second antibodies. Cells were rinsed, and the radioactivity of cell lysates was counted. Nonspecific binding was determined with cells that were incubated only with the 125 I-labeled second antibody. It represented 0.1% of total radioactivity added. Binding of anti-Flag antibodies to epitope-tagged mutant receptors is given as a percentage of anti-Flag antibodies binding to epitope-tagged wild type receptor (mean Ϯ S.E. of three experiments).

Constructs
Surface  6. Ribbon representation of the three-dimensional model of the VPAC 1 -(8 -117) receptor domain highlighting all amino acids whose mutation alters VIP binding to the VPAC 1 receptor. Yellow wireframe represents amino acids whose mutation abolishes VIP binding and pale yellow wireframe represents amino acids whose mutation decreases the affinity for VIP (see Table I  Other amino acids whose mutation does not alter VIP binding are shown (gray wireframe) including Asp 38 which is below the plane and Glu 108 (potential salt bridge with Lys 65 ) which is above the plane. NH 2 and COOH ends are indicated.
first comprehensive view of a VIP-binding site of the human VPAC 1 receptor and more generally of a peptide-binding domain in the N-terminal extracellular region of the class II family of G protein-coupled receptors. Overall, the data identified a putative binding site (Fig. 6) made of an electronegative groove ending on a tryptophan shell constituted of Trp 67 , Trp 73 , and Trp 110 . Along the groove two acid residues Glu 36 and Asp 68 are important. Glu 36 is not conserved in class II G proteincoupled receptors (13) and could therefore be specific of VIP interaction. Several important basic residues of VIP are candidates for an interaction with Glu 36 including Arg 14 , Lys 15 , and Lys 21 (25). In contrast, Asp 68 appears to be more deeply buried in the groove and should be less accessible for a direct interaction with VIP in the current model. A structural role of Asp 68 by formation of a salt bridge is possible. However, a structural reorganization of the groove in response to the docking of VIP could bring out Asp 68 for an interaction with the above cited basic residues of VIP.
Other residues could participate in the structure of the VIP binding groove including Pro 74 , Pro 87 , and Gly 109 . The latter residues are highly conserved in class II G protein-coupled receptors (7). Our data also indicate that the side chain of Met 66 should be accessible in the groove (Figs. 6 and 7). This side chain should be part of a VIP preferential plane of interaction which could be made of the side chains of Glu 36 , Met 66 , and potentially Asp 68 . Indeed the activity of the E36A, D68A, and M66A mutants and the distance between Glu 36 and Asp 68 on one hand, and that of Arg 14 (or Lys 15 ) and Lys 21 in the VIP peptide, on the other hand, could argue for the existence of such a plane.
Finally, it is important to consider the three consensus Nglycosylation sites (Asn 58 , Asn 69 , and Asn 100 ). We previously demontrated by site-directed mutagenesis and biochemical experiments that the two sites occupied by a 9-kDa N-linked carbohydrate on Asn 58 or Asn 69 play a mandatory role for the delivery of the VPAC 1 receptor to the plasma membrane (9). In the model, the three asparagines are located at the surface of the structure. The two functionally important glycosylation sites on Asn 58 and Asn 69 are clearly accessible to anchor carbohydrates.
The VPAC 1 receptor has important cystein residues (1,8). Six of the eight cysteins present in the N-terminal domain of the receptor are highly conserved in the class II of G proteincoupled receptors. Their mutation abolished VIP binding (8). Five of those cysteins are in the fragment that we modelized. Our model supports that Cys 72 -Cys 86 could form a disulfide bridge in vivo. Indeed, both residues are mapped at the bottom of the binding groove, facing each other even if they are 12 Å distant (S atom center to S atom center). This distance is longer than the S-S length of a disulfide bridge (2.1 Å). However, it is interesting to note that forcing Cys 72 and Cys 86 to reach a distance compatible with a S-S bond (not shown) does not alter the overall structure of the putative VIP binding groove (root mean square deviation between the two models ϭ 1 Å). The other putative S-S bonds cannot be predicted since all functionally important cysteins are not present in the model. For example, the 118 -144 domain contains one crucial Cys (Cys 122 ), and the extracellular loops of the transmembrane domain, that are not modelized here, also contain important Cys residues (8,14).
In line with these restrictions, one must notice that Lys 143 in the 118 -144 domain 2 and Asp 196 in the first extracellular loop (10) are also clearly important for the VIP binding to the VPAC 1 receptor. Therefore, the VIP binding groove we have located in the 8 -117 sequence of the receptor by modelization and mutagenesis experiments is most probably a part of the whole VPAC 1 receptor-binding site. This is in line with recent data regarding another class II G protein-coupled receptor, the PTH/PTHrP receptor, which highlight the existence of multiple binding subdomains for PTH in the receptor, including an important role for the N-terminal extracellular region of the receptor (33).
In conclusion, structural analysis of the partial three-dimensional model of the VIP-binding site and site-directed mutagenesis reveal that: (i) the N-terminal extremity of the VPAC 1 receptor is mostly helical and could correspond to a signal peptide. (ii) The crucial residues for VIP binding and adenylyl cyclase activation by VIP are around a groove containing several negatively charged residues. This groove could be part of the binding site for the VIP which has several positively charged residues. (iii) New residues implied in the VIP recognition and/or structure of the binding groove, suggested from the model (Pro 74 , Pro 87 , Phe 90 , and Trp 110 ) were experimentally validated by site-directed mutagenesis. While the recent report of the three-dimensional crystal structure of rhodopsin at 2.8 Å (34) offers a template for most other G protein-coupled receptors, it does not get the clue to the structure and crucial function of large N-terminal extracellular domains of class II receptors. In this respect, the present work provides the first model of a partial three-dimensional structure of the N-terminal domain of the human VPAC 1 receptor and should help to better understand the original structure/function relationship of G-coupled receptors of the class II family.