Identification of Key Residues for Interaction of Vasoactive Intestinal Peptide with Human VPAC1 and VPAC2Receptors and Development of a Highly Selective VPAC1Receptor Agonist

The widespread neuropeptide vasoactive intestinal peptide (VIP) has two receptors VPAC1 and VPAC2. Solid-phase syntheses of VIP analogs in which each amino acid has been changed to alanine (Ala scan) or glycine was achieved and each analog was tested for: (i) three-dimensional structure by ab initio molecular modeling; (ii) ability to inhibit 125I-VIP binding (K i ) and to stimulate adenylyl cyclase activity (EC50) in membranes from cell clones stably expressing human recombinant VPAC1or VPAC2 receptor. The data show that substituting residues at 14 positions out of 28 in VIP resulted in a >10-fold increase ofK i or EC50 at the VPAC1receptor. Modeling of the three-dimensional structure of native VIP (central α-helice from Val5 to Asn24 with random coiled N and C terminus) and analogs shows that substitutions of His1, Val5, Arg14, Lys15, Lys21, Leu23, and Ile26 decreased biological activity without altering the predicted structure, supporting that those residues directly interact with VPAC1 receptor. The interaction of the analogs with human VPAC2 receptor is similar to that observed with VPAC1 receptor, with three remarkable exceptions: substitution of Thr11 and Asn28 by alanine increased K i for binding to VPAC2receptor; substitution of Tyr22 by alanine increased EC50 for stimulating adenylyl cyclase activity through interaction with the VPAC2 receptor. By combining 3 mutations at positions 11, 22, and 28, we developed the [Ala11,22,28]VIP analog which constitutes the first highly selective (>1,000-fold) human VPAC1 receptor agonist derived from VIP ever described.

The vasoactive intestinal peptide (VIP) 1 is a prominent neuropeptide with wide distribution in both peripheral and central nervous systems and a large spectrum of biological actions in mammals (1,2). VIP-containing nerves and VIP effects have been described in digestive tract, cardiovascular system, airways, reproductive system, immune system, endocrine glands, and brain (1). Besides its short-term actions on exocrine secretions, hormone release, muscle relaxation, and metabolism (1,2), VIP has been also characterized as a growth regulator for fetuses and tumor cells and during embryonic brain development (3). There are recent evidences for an important role of VIP in the perception of pain (4) and suppression of inflammation (5). Finally, VIP has been involved in diseases such as the watery diarrhea syndrome and clinical applications of VIP have been already suggested in impotence, asthma, lung injury, a variety of tumors and neurodegenerative diseases (1)(2)(3).
VIP belongs to a large family of structurally related peptides (2, 6, 7) that comprises VIP, pituitary adenylate cyclase-activating peptide PACAP-27, and its C-terminal extended form PACAP-38, secretin, glucagon, and glucagon-like peptides-1 and -2, gastric inhibitory polypeptide, peptide histidine methionine amide, growth hormone-releasing factor (GRF), and peptides isolated from the venom of the Gila Monster. VIP and PACAP are the most closely related peptides in terms of structure and function (2,6). They share two common receptors, VPAC 1 and VPAC 2 , which display high affinity for both VIP and PACAP (2,8). These receptors together with receptors for VIP-related peptides (see above) clearly constitute an original subfamily within the superfamily of G protein-coupled receptors (2,9,10). This subfamily referred to as class II (2) also comprises receptors for parathyroid hormone, calcitonin, corticotropin-releasing factor, and the so called EGF-TM7 receptors (11). Class II family of receptors for peptides display several common properties including large N-terminal extracellular domains containing highly conserved cystein residues, N-terminal leader sequences, and complex gene organization with many introns (2).
Although the structure-function relationship of VIP receptors, including VPAC 1 and VPAC 2 , has been recently documented (2,9,(12)(13)(14)(15)(16)(17)(18)(19)(20), the structure-function relationship of VIP itself is still poorly understood. Some old studies carried out before the characterization and cloning of VIP receptor subtypes (21)(22)(23) indicated that: (i) the entire sequence of VIP is required for full biological function (21,22); (ii) VIP-related peptides having significant sequence homologies with VIP such as peptide histidine methionine amide, secretin, GRF, and helodermin behave as low potency VIP agonists (23); (iii) there are important differences between species in the pharmacology of VIP receptors, especially between rodents and humans (2,15). In this context, the present study explores the structurefunction relationship of VIP for interacting with the two human VIP receptor subtypes VPAC 1 and VPAC 2 . The contribution of each side chain of VIP was investigated by the systematic single exchange of each residue of VIP 1-28 by alanine or glycine. All modified VIP peptides were then analyzed for their binding affinities and abilities to stimulate adenylyl cyclase in CHO cell clones stably expressing human recombinant VPAC 1 or VPAC 2 receptor. Moreover, the contribution of each residue was also analyzed after molecular modeling of all VIP analogs. These studies allowed us to determine key residues for interaction with human VPAC 1 and VPAC 2 receptors and also resulted in the development of the most highly selective VPAC 1 receptor agonist ever described.

EXPERIMENTAL PROCEDURES
Materials-Enzymes and vectors for cloning were obtained from Promega (Charbonnière, France). The recombinant human VPAC 1 receptor was stably expressed in CHO cells as described (24 11,22,28 ]VIP analogs were obtained by custom peptide synthesis from Neosystem (Strasbourg, France). 125 I-VIP was prepared and purified as described (26). Other highly purified chemicals used were from Sigma (Saint-Quentin-Fallavier, France).
A double coupling was performed when Fmoc-Asn(Trt)-OH, Fmoc-Val-OH, Fmoc-Gln(Xan)-OH, Fmoc-Arg(Pmc)-OH, Fmoc-Thr(But)-OH, and Fmoc-Tyr(But)-OH were involved in coupling. Fmoc deprotection was performed with a solution of piperidine in dimethylformamide in a 2:8 (v/v) ratio. Final deprotection of the peptides from the resin was performed in a mixture containing trifluoroacetic acid, water, thioanisole, ethylenedithiol, phenol in a 82.5/5/5/2.5/5 (v/v) ratio (30) for 2 h. Peptides were then precipitated by addition of cold diethyl ether and dissolved in a mixture of 0.1% trifluoroacetic acid in water/acetonitrile and lyophilized. They were purified by high performance liquid chromatography on a Waters instrument equipped with a Waters Delta Packs C18 column (340 ϫ 100 mm) using a gradient of solvents composed of water, 0.1% trifluoroacetic acid, and acetonitrile containing 0.1% trifluoroacetic acid. Purity of all peptides was checked by analytical high performance liquid chromatography on a Waters instrument using a C18 column (Novarpack, 5 m, 100 Å, 180 ϫ 3.5 mm) and all were at least 97% pure (UV detection at 214 and 254 nm). They were characterized by electrospray mass spectrometry using a Jeol LX 2000 spectrometer.
Molecular Modeling of VIP and VIP Analogs-The conformations of native VIP and VIP analogs were calculated by molecular modeling using the OSIRIS software which is an ab initio approach allowing the calculation of three-dimensional structure of soluble proteins of less than 150 residues (31). The first step of OSIRIS is used in the present study since it is sufficient to correctly determine most secondary structures of different ␣-helix proteins (32). Briefly, the primary sequence of the peptide is introduced and 179 pairs of ⌽ and angles are succes- sively attributed to all amino acids of the sequence. For each rotation axis, the energy is calculated using the following equation, where E(k) is the energetic component associated to the torsion axis k, between atoms w and wϩ1, with E(k) tor as torsional energy, E vdw ij as Van der Waals energy, E elec ij as electrostatic energy, E sol.in ij as internal solvation energy between atoms i and j, and E sol.out ij as solvent solvation energy, and q as an energy quantum equal to the interaction energy between atoms i and j, divided by the number of torsion axes between atoms i and j. A structure is then obtained from the linear combination of each residue energy minimum. The energy of this construct is calculated and the residues having the lowest energy are frozen. Then the calculation is started again, keeping the structure of frozen residues constant and testing the 179 pairs of ⌽, angles on the others. At each step of calculation, residues are frozen, varying between 5 and 28. Increasing the value allows to test the variability of the calculated VIP or VIP analog structures. Calculations were performed on RAM-SES (Rapid Analysis Master/Slaves Extensible System), a parallel hardware of 21 tracer Europa Pentium Pro PC cadenced at 180 Mhz connected by a 100 Mbytes Network and controlled by a HP Vectra VA Pentium Pro cadenced at 200 Mhz. Molecular views were drawn with WinMGM 1.0 (Ab Initio Technology, Obernai, France) as described (33).
Statistical analysis of the predicted structures of VIP and VIP analogs was made using the SICLA program (32). This procedure starts from n ϫ n distance tables calculated from the local root mean square distances of all pairs of structures obtained using different E mvi () values. The root mean square distances was calculated by fitting the backbone of 5 amino acids of a structure to the 5 corresponding residues of the reference structure (here, the predicted structure obtained with ϭ 5). The root mean square distances of the fragment is attributed to the central amino acid, the window is then moved by 1 residue along the sequence and calculations are repeated. The clustering procedure used in SICLA is an automatic nonhierarchical classification that compares pairs of structures (34,35). This procedure is described in detail elsewhere (32). From the clustering of structures calculated with different values, a mean distance between the central structure and the others is obtained. This distance is representative of the peptide structural variability.
Stable Expression of cDNA Encoding Human VPAC 1 and VPAC 2 Receptors in CHO Cells-Full-length VPAC 1 or VPAC 2 receptor cDNAs were ligated in the EcoRI site of pcDNA3 (Invitrogen). Both receptors were tagged at the C terminus with a marker dodecapeptide (Tag) as described (14). The recombinant plasmid encoding human VPAC 2 was transfected into CHO cells (ECACC 85050302) by electroporation. Briefly, four million CHO cells, cultured as described (14), were preincubated on ice for 5 min with 20 g of plasmid DNA and 20 g of salmon sperm DNA carrier in cold Ham's F-12 medium. After electroporation (330 V, 1,000 microfarads, infinite resistance), cells were kept on ice for 5 min and then transferred to Petri dishes containing 10 ml of culture medium (Ham's F-12, 10% (v/v) fetal calf serum, 100 IU/ml penicillin, 100 g/ml streptomycin). After 48 h, cells were selected by addition of geneticin (G418) at the final concentration of 800 g/ml for 2 weeks. Resistant cells were cloned by limiting dilution as described (24). A clone, referred to as clone 10, was isolated and selected on the basis of its binding parameters for VIP (see below). Clone 10 cells were grown in the above described culture medium containing 100 g/ml G418 in a humidified atmosphere of 95% air and 5% CO 2 at 37°C. A clone (referred to as clone 15) of CHO cells stably expressing the human VPAC 1 receptor was previously isolated and characterized (24). Both clone 10 and clone 15 cells were passaged every 7 days in 25-cm 2 plastic culture flasks and used between passages 8 and 30, under conditions for which the density of VPAC 2 and VPAC 1 receptors did not vary significantly (see below).
Membrane Preparation-Cells, either clone 10 or clone 15, were grown to confluency for 3-4 days. After removing the culture medium, attached cells were washed three times with phosphate-buffered saline, then harvested with a rubber policeman and centrifuged for 10 min at 2,000 ϫ g. The cell pellet was exposed for 30 min at 4°C to hypoosmotic 5 mM Hepes buffer, pH 7.4, and a particulate fraction, referred to as membrane preparation, was obtained as described (17). Membranes were stored at Ϫ80°C until use. Protein content was measured by the procedure of Bradford (36) with bovine serum albumin as standard.
Receptor Binding Assay-Membranes (100 g of protein/ml) were incubated at equilibrium for 60 min at 30°C with 0.05 nM 125 I-VIP and increasing concentrations of native VIP or VIP analogs. The incubation medium was 20 mM Hepes, pH 7.4, containing 2% (w/v) bovine serum albumin and 0.1% (w/v) bacitracin. The reaction was stopped by transferring the incubation medium to an excess of ice-cold buffer. Bound and free peptides were immediately separated by centrifugation at 20,000 ϫ g for 10 min, and membrane pellets were washed twice with 10% (w/v) sucrose in ice-cold 20 mM Hepes. The radioactivity was then counted with a ␥-counter. Specific binding was calculated as the difference between the amount of 125 I-VIP bound in the absence (total binding) and presence (nonspecific binding) of 10 M unlabeled native VIP. The nonspecific binding represented about 10 and 15% of total radioactivity bound to clone 15 and clone 10 cell membranes, respectively. All binding data were analyzed using the LIGAND computer program (38). The dissociation constant (K d ) and binding capacity (B max ) of VIP binding to clone 10 and clone 15 cell membranes were determined by Scatchard analysis. From the linear Scatchard plots, the K d of VPAC 1 receptor in clone 10 cell membranes and VPAC 2 receptor in clone 15 cell membranes were 0.4 and 0.7 nM, respectively. The B max were 1.6 and 1.3 pmol/mg of protein, respectively. The constants K i for the inhibition of 125 I-VIP binding by unlabeled peptides were calculated using the Cheng-Prusoff equation as described (39). All competition curves for VIP and VIP analogs fit with a monophasic binding profile for both VPAC 1 and VPAC 2 binding assays.
Adenylyl Cyclase Assay-Adenylyl cyclase activity in membranes from clone 10 or clone 15 cells was assayed in the presence of increasing concentrations of native VIP or VIP analogs as described (37). Doseresponse curves were fitted and concentration of peptide giving halfmaximal response (EC 50 ) were calculated using the Prism software suite (GraphPad Software, San Diego, CA).

RESULTS
Alanine Scanning Analysis of VIP for Interacting with Human VPAC 1 Receptor-Twenty-six analogs of VIP were synthesized in which each side chain of VIP 1-28 was systematically replaced with alanine. Two other analogs were also synthesized in which glycine was substituted for alanine in position 4 or 18 of native VIP. The ability of each analog to interact with the recombinant human VPAC 1 receptor expressed in CHO clone 15 cells was first analyzed by competitive inhibition of 125 I-VIP binding to cell membranes. For all VIP analogs, the doseresponse curves for inhibiting 125 I-VIP binding were parallel to that of native VIP and complete inhibition could be observed at 10 M peptide concentration (not shown). All competitor curves fitted with a monophasic binding profile consistent with a single binding site. Analysis of K i (Table I) showed that substituting residues at 14 positions including positions 1, 3, 5, 6, 7, 8, 10, 12, 14, 15, 20, 21, 23, and 26 of VIP resulted in an important decrease in affinity for VPAC 1 receptors, i.e. more than one log. Fig. 1 shows the competition curves for 125 I-VIP binding to VPAC 1 receptor for some of these VIP analogs with decreased affinity e.g. H1A, F6A, K20A, and I26A VIP analogs. Substitution at other positions resulted in no change of affinity or a small decrease of affinity of less than one log. The most pronounced decreases of affinity occurred with H1A, F6A, R12A, R14A, and L23A analogs for which a 100 -200-fold decrease was observed. Other significant decreases of affinity were observed with D3A, V5A, T7A, D8A, Y10A, K15A, K20A, K21A, and I26A analogs, e.g. Ͼ10-fold. All VIP analogs were also tested for their ability to stimulate adenylyl cyclase activity in membranes of CHO clone 15 cells expressing the human VPAC 1 receptor. The dose-response curves for all analogs paralleled that of VIP and 10 M analogs achieved the same maximal stimulation of enzyme activity as native VIP (not shown). Fig. 1 shows the dose-response curves for stimulating enzyme activity for some of these analogs with decreased potency. In general, there is a good correlation between the EC 50 for stimulating enzyme activity and the K i for inhibiting 125 I-VIP binding (see Table I) (Fig. 2). All analogs behaved as VPAC 1 receptor agonists with identical or lower potencies than native VIP.
Altogether, the data were consistent with the importance of many residues along the VIP 1-28 sequence. One can notice the particular contributions of the N-terminal 1-8 sequence and also of the central part of the molecule with five crucial basic residues i.e. Arg 12 , Arg 14 , Lys 15 , Lys 20 , and Lys 21 . Two hydrophobic residues in the C-terminal part also play a role, i.e. Leu 23 and Ile 26 . A recurrent issue for explaining alanine scanning data for peptide ligand is to determine whether a given substitution by alanine disrupts a specific interaction between peptide and receptor and/or alters the global structure of the peptide. To address this important question, modeling of threedimensional structure of all VIP analogs was performed.
Molecular Modeling of VIP and VIP Analogs-The conformation of native VIP as well as those of all VIP analogs were analyzed using OSIRIS (see "Experimental Procedures"). The first step of this ab initio procedure allows to assign a secondary structure to each residue. Iterative calculations were performed and at each iteration residues are frozen up to the entire sequence, varying between 5 and 28 ( ϭ 5, 7, 9, 11, 12, 16, 20, and 28 residues). This mimics nucleation centers observed during the folding of a protein and was used here to estimate the structural variability of native VIP and each analog. Fig. 3 shows the different conformers obtained for native VIP. The conformational variability is essentially located at the N-and C-terminal extremities, the central domain from Val 5 to Asn 24 being clearly helical. A statistical analysis was then carried out to assess the structural dispersion shown in Fig. 3. A statistical mean distance (D) based on the local root mean square between all pairs of conformers was calculated using SICLA (see "Experimental Procedures"). The structural dispersion of several VIP analogs is clearly higher than that of native VIP. Indeed, D3A, F6A, T7A, D8A, Y10A, R12A, and K20A VIP analogs exhibit D values Ͼ1.5 Å versus 1.2 Å for native VIP. This indicates that the corresponding analogs have a significant change of conformation as compared with native VIP (see below). In contrast, other mutations did not affect this dispersion with D values Ͻ1.2 Å, e.g. H1A, S2A, A4G, K15A, Q16A, M17A, A18G, V19A, Y22A, L23A, S25A, I26A, and N28A VIP analogs. These latter analogs have predicted structures which are very similar to that of native VIP (for example, see the L23A analog in Fig. 3). An intermediate behavior was observed for the V5A, T11A, R14A, K21A, N24A, and L27A VIP analogs which exhibited D values between 1.2 and 1.5 Å. For these analogs, we noticed that the dispersion is mostly due to one of the eight conformers corresponding to one peculiar value as shown, for example, for the V5A VIP analog (Fig. 3). In this particular example, when ϭ 28 the predicted structure of the V5A analog does not fit as well in the N-terminal domain as for the other values. We assumed that a single deviation is not significant and we concluded that an intermediate D value is not representative of a structural change in the analog as compared with native VIP. For analogs with D values Ͼ1.5 Å, the high dispersions are clearly due to differences of several, if not all, conformers. The D3A, F6A, T7A, D8A, and Y10A mutants exhibit changes in their N-terminal domain as compared with native VIP (Fig. 3). The R12A mutant exhibits a global alteration of the predicted structure due to partial disruption of the central helical domain (Fig. 3). Finally, the K20A mutant exhibits a change in its C-terminal domain as compared with native VIP (Fig. 3).
Structure-Activity Relationship of VIP for Interacting with Human VPAC 1 Receptor-The 28 residues of native VIP can be classified in three categories with respect to the interaction of VIP with human VPAC 1 receptor: (i) residues which can be substituted into alanine or glycine without significant alteration of their binding affinity or biological potency. This category includes Ser 2 , Ala 4 , Asn 9 , Thr 11 , Leu 13 , Gln 16 , Met 17 , Ala 18 , Val 19 , Tyr 22 , Asn 24 , Ser 25 , Leu 27 , and Asn 28 . It is quite interesting to note that, compared with native VIP, mutants at these positions do not exhibit significant change in their predicted structure; (ii) residues whose substitution into alanine is associated with a significant alteration of the binding affinity or biological potency of the corresponding analog as well as a change of the predicted analog structure. This category includes Asp 3 , Phe 6 , Thr 7 , Asp 8 , Tyr 10 Table I and legend to Fig. 1 for details. though we can speculate that the decreased binding affinity of the corresponding mutant is due to its altered structure, the possible direct involvement of the category ii residues in the VIP-VPAC 1 receptor interaction cannot be ruled out; (iii) residues whose substitution into alanine results in a decreased binding affinity or biological potency of the corresponding analog without change of the predicted analog structure. This category includes His 1 , Val 5 , Arg 14 , Lys 15 , Lys 21 , Leu 23 , and Ile 26 . These residues are likely to participate in the direct interaction between VIP and human VPAC 1 receptor. The VIP sequence with the three categories of residues is shown in Fig. 4.
It is worth noting that all analogs exhibiting alteration of the predicted peptide structure, as compared with native VIP, have a decreased binding affinity or biological potency. In this context, it was interesting to align the amino acid sequences of human VIP and human PACAP-27 (Fig. 4) since both peptides have the same high affinity for human VPAC 1 receptors (38). It clearly appears that residues belonging to categories ii and iii are identical in VIP and PACAP or highly conserved, e.g. exchanges between valine and isoleucine at positions 5 and 26.
Alanine Scanning Analysis of VIP for Interacting with Human VPAC 2 Receptor: Selectivity of VIP Analogs-The ability of each VIP analog to interact with the recombinant human VPAC 2 receptor expressed in CHO clone 10 cells was analyzed by competitive inhibition of 125 I-VIP binding to cell membranes. All competitor curves fitted with a monophasic binding profile, consistent with a single binding site (not shown). Fig. 5 shows the competition curves for 125 I-VIP binding to VPAC 2 receptor for some VIP analogs, e.g. T11A, Y22A, and N28A. Analysis of K i (Table II) showed that substituting residues at 16 positions of VIP resulted in a Ͼ1 log decrease of affinity for VPAC 2 receptors. This includes the same 14 positions as for VPAC 1 receptor and two additional spots at positions 11 and 28. Since the corresponding analogs T11A and N28A VIP do not exhibit changes in their predicted structure as compared with native VIP, it can be suggested that Thr 11 and Asn 28 participate in the direct interaction between VIP and human VPAC 2 receptor. Further differences between VPAC 1 and VPAC 2 receptors can be noticed since some VIP analogs exhibit a much higher decrease in binding affinity for the VPAC 2 receptor than for the VPAC 1 receptor, i.e. D8A, Y10A, and L23A. All VIP analogs were also tested for their ability to stimulate adenylyl cyclase activity in membranes of CHO clone 10 cells expressing the human VPAC 2 receptor. The dose-response curves for all analogs paralleled that of VIP and 10 M analogs achieved the same maximal stimulation of enzyme activity as native VIP (not shown). The dose-response curves for T11A, Y22A, and N28A analogs are shown in Fig. 5. In general, there is a good correlation between the EC 50 for stimulating enzyme activity and the K i for inhibiting 125 I-VIP binding ( Fig. 6 and Table II). However, the EC 50 of Y22A VIP for stimulating adenylyl cyclase activity is much higher than its K i for binding to VPAC 2 receptor ( Fig. 5 and Table II). This property is specific for VPAC 2 receptor since the EC 50 and K i of Y22A VIP at the VPAC 1 receptor are very similar (Table I).
From the above described data, it appears that several analogs discriminate between VPAC 1 and VPAC 2 receptors. They exhibit a much higher binding affinity and/or biological potency for VPAC 1 than for VPAC 2 receptors. The reverse is not true since none of the analogs had a higher affinity for VPAC 2 than for VPAC 1 . In this context, we synthesized new analogs with the aim to develop more selective human VPAC 1 receptor agonists. For that purpose, we first combined mutations at positions 11 and 28 which individually resulted in a decreased affinity for VPAC 2 receptor without any change in the affinity for VPAC 1 receptor. The [Ala 11,28 ]VIP analog clearly discriminated between VPAC 1 and VPAC 2 receptors (Tables I and II). Indeed, this VIP analog had the same affinity as native VIP for VPAC 1 receptor whereas it displayed a 44-fold lower affinity than native VIP for VPAC 2 receptor (Fig. 7). Similar data were obtained in the adenylyl cyclase assay (Fig. 7). The [Ala 11,28 ]VIP analog was as potent as native VIP for stimulating enzyme activity via VPAC 1 receptor whereas it was 21-fold less potent than VIP for stimulating enzyme activity via VPAC 2 receptor (Fig. 7). Since we also noted that position 22 in VIP was important for discriminating between VPAC 1 and VPAC 2 receptors with respect to adenylyl cyclase activation (see above), we further synthesized a VIP analog which combines 3 mutations at positions 11, 22, and 28. Quite interestingly, the [Ala 11,22,28 ]VIP analog was highly selective for VPAC 1 receptor. Indeed, its binding affinity for VPAC 2 receptor and subsequent biological potency in stimulating adenylyl cyclase were decreased by Ͼ1,000-fold as compared with native VIP (Fig. 7). In sharp contrast, this analog with a triple mutation retained a binding affinity for VPAC 1 receptor closed to that of native VIP (Fig. 7). The same held true for biological response, the triple mutant and native VIP having identical potencies for stimulating adenylyl cyclase activity in VPAC 1 receptor-transfected cells (Fig. 7). It is interesting to note that 10 Ϫ8 M of this analog which triggers maximal response at VPAC 1 receptor is inactive at VPAC 2 receptor (Fig. 7). This makes [Ala 11,22,28 ]VIP the most highly selective human VPAC 1 receptor agonist derived from VIP ever described. DISCUSSION Although several studies have long been reported regarding the pharmacology of analogs of VIP and related peptides (40 -43), no systematic evaluation of the functional and structural role of every residues of VIP has been yet performed since the cloning of the two VIP receptor subtypes VPAC 1 and VPAC 2 (44). Moreover most previous studies were performed using VIP receptors from rodents although it is clear that there are important differences between species in the pharmacology of VIP receptors (8,15). In the present study, we have synthesized a total of 28 single alanine or glycine mutants of VIP and have analyzed both their predicted three-dimensional structure by molecular modeling and their biological properties by binding assay and adenylyl cyclase assay in CHO cell clones expressing recombinant human VPAC 1 or VPAC 2 receptor. Our data provides new information: (i) they delineate key amino acid residues of VIP which play a role in biological activity for interacting with human VIP receptor subtypes and/or contribute to the three-dimensional structure of the peptide; (ii) they point to the role of specific amino acid residues in VIP for discriminating VPAC 1 and VPAC 2 receptors; (iii) they provide a rationale for developing a new highly specific agonist of the human VPAC 1 receptor by combining three single mutations at key positions of VIP. This new information represents a significant advance over the current knowledge on this widespread neuropeptide.
Analysis of the interaction of the 28 single alanine or glycine mutants of VIP indicates that 14 residues out of 28 in native VIP could not be changed without a significant decrease in affinity for binding to human VPAC 1 receptor. These important residues are distributed along the peptide chain from position 1 to 26. This distribution is in good agreement with previous observations indicating that VIP fragments obtained by deletion of the N-terminal, central, or C-terminal domains of VIP retained very low affinity, if any, for VIP receptors (21). In a previous Ala scan of the VIP derivative Ro 23-7059 the binding activity of analogs had been tested in guinea pig or human lung membranes and emphasized the importance of only 6 positions including Asp 3 , Phe 6 , Thr 7 , Tyr 10 , Tyr 22 , and Leu 23 (22). These data are somewhat different from ours since we found many other important positions and identified Tyr 22 as an unimportant position, at least for the interaction of VIP with VPAC 1 receptor. The reasons for these discrepancies probably include several factors: (i) a VIP derivative was used, i.e. Ac-  Tables I and II. The sequence of human PACAP-27 is aligned with that of VIP to show the identities (solid bar) and homologies (double arrow).

FIG. 5. Binding assay and adenylyl cyclase assay of VIP analogs in CHO cell clone stably expressing human VPAC 2 receptor.
A, dose effects of native VIP and some VIP analogs for inhibition of 125 I-VIP binding to membranes from CHO cells expressing VPAC 2 receptor. B, stimulation of adenylyl cyclase activity by native VIP and some VIP analogs in membranes from CHO cells expressing VPAC 2 receptor. The data are expressed as the percentage of maximal stimulation above basal obtained with 10 M native VIP. All data are mean Ϯ S.E. of at least three experiments performed in duplicate. q, native VIP; E, T11A mutant; Ⅺ, Y22A mutant; ⌬, N28A mutant. Note the shift for the Y22A mutant when comparing the binding assay (A) and the adenylyl cyclase assay (B).
[Lys 12 ,Nle 17 ,Val 26 ,Thr 28 ]VIP (22); (ii) the authors used lung membranes as a source of VIP receptors whereas we used recombinant receptor subtypes in this study. In this context, it has been shown since the initial study of O'Donnell et al. (22) that lung is endowed with both VPAC 1 and VPAC 2 receptors (45); (iii) the authors utilized guinea pig tissues in their studies (22). It is known that guinea pig peptide hormones or neuropeptides have divergent amino acid sequences as compared with most mammals (46). In the case of VIP, the sequence is unchanged in all mammals with the notable exception of guinea pig whose VIP has four changes (47); (iv) the bioassay used in their study (22) consisting of relaxation of guinea pig trachea appears to be less sensitive than the adenylyl cyclase assay used in our study.
In the present study, we identified His 1 , Asp 3 , Val 5 , Phe 6 , Thr 7 , Asp 8 , Tyr 10 , Arg 12 , Arg 14 , Lys 15 , Lys 20 , Lys 21 , Leu 23 , and Ile 26 as being important for binding to human VPAC 1 receptor and subsequent stimulation of adenylyl cyclase activity (see Table I). It is quite interesting to note that all these amino acids are conserved in PACAP (see Fig. 4) which has the same affinity as VIP for human VPAC 1 receptor (39). Indeed, the amino acids are either identical or are highly homologous with Val 5 and Ile 26 in VIP being replaced by Ile 5 and Val 26 in PACAP, respectively (see Fig. 4). This conservation strongly validates the present data. Futher analysis of the 14 important amino acid residues in VIP may be done by comparing the amino acid sequences of VIP and VIP-related peptides which do (PACAP, secretin, peptide histidine methionine amide, and GRF) or do not (glucagon, glucagon-like peptides and GIP) behave as VIP agonists at the VPAC 1 receptor (12,39,44). This comparison supports that His 1 , Asp 3 , Phe 6 , Arg 12 , Lys 20 , Lys 21 , Leu 23 , and Ile 26 which emerged from the present Ala scan as important residues in VIP for interacting with human VPAC 1 receptor, are strictly conserved or replaced by homologous residues in VIP-related peptides which behave as VIP agonists. This suggests that these residues may play a similar role in allowing the binding of VIP, PACAP, peptide histidine methionine amide, GRF, and secretin to human VPAC 1 receptor but  50 for stimulating adenylyl cyclase activity is much higher than its K i for binding. altogether are not sufficient to ensure high affinity binding. To further substantiate this issue, it is worth pointing out that (i) we could not find well conserved amino acid residues in VIPrelated peptides which do not play a role in the activity of VIP; (ii) conversely, we did identify nonconserved residues which are important for VIP activity at the human VPAC 1 receptor.
At this stage of the Ala scan study of VIP and more generally of other similar studies with any bioactive peptide, the question arises of whether one given residue is important because it directly interacts with the receptor and/or it plays a role in maintaining the three-dimensional structure of the peptide, an issue which is not often addressed. In the present study, the conformations of native VIP and VIP analogs were calculated by molecular modeling using an ab initio approach (31). When this approach was applied to native VIP, the predicted threedimensional structure of VIP was consistent with a central ␣-helical domain from Val 5 to Asn 24 and conformational variability essentially located at the N-and C-terminal domains (see Fig. 3). This is in good agreement with previous experimental studies of the VIP structure in methanol/water solutions using nuclear magnetic resonance and circular dichroism which also indicated that the structure of VIP is mostly helical (48) with the existence of a central well defined ␣-helix, the remaining residues being not ordered (49). Similarly, PACAP38 was shown to have a disordered N-terminal domain followed by a central ␣-helical structure (50). Interestingly, helodermin, a peptide isolated from the lizzard Gila Monster (51), which is highly homologous to VIP (51) and behaves as a VIP agonist (52), has also been shown to exhibit a central ␣-helice with random coiled N and C terminus (53). With respect to three-dimensional structure, VIP analogs in which amino acid substitution into alanine was associated with alteration of binding affinity and biological potency could be classified into two categories: (i) analogs which exhibited significant change in predicted peptide structure as compared with native VIP. This includes D3A, F6A, T7A, D8A, Y10A, R12A, and K20A mutants. For these analogs, it can be suggested that altered activity is related to altered structure, although we cannot strictly rule out the possibility that the corresponding residues may directly participate in the interaction of VIP with the VPAC 1 receptor; (ii) analogs which exhibited no change in predicted peptide structure as compared with native VIP. This includes H1A, V5A, R14A, K15A, K21A, L23A, and I26A mu-tants. It is tempting to speculate that the corresponding amino acid residues in native VIP (see Fig. 8) participate in the interaction of the neuropeptide with VPAC 1 receptor. This is in line with other observations: (i) His 1 in VIP (22) as well as His 1 in some VIP-related peptides (54 -56) is known to play an important role in biological activity; (ii) the fact that half of the residues falling into the category of candidates for direct interaction with VPAC 1 receptors are basic residues including Arg 14 , Lys 15 , and Lys 21 , is probably significant in view of the fact that site-directed mutagenesis of the human VPAC 1 receptor previously identified several important acidic residues in the receptor for VIP binding, including Glu 36 (16), Asp 68 (12), and Asp 196 (57). This supports the idea that the very basic VIP with an isoelectric point Ͼ11 (6) may use some of its basic residues for an electrostatic interaction with acidic residues of the receptor; (iii) finally, the remaining important residues of VIP including Val 5 , Leu 23 , and Ile 26 are highly hydrophobic in consonance with the old idea that VIP, or at least a binding region of VIP, has an hydrophobic environment within the receptor (58). Recent observations support the idea that important residues in the VPAC 1 receptor for binding VIP are indeed hydrophobic including Trp 67 , Trp 73 , and Trp 110 (12). 2 Finally, it is worth pointing out that all analogs which have the same affinity as native VIP for VPAC 1 receptor exhibit no alteration of predicted three-dimensional structure, further validating our molecular modeling approach.
The data obtained from the analysis of the interaction of the 28 single alanine or glycine mutants of VIP with human VPAC 2 receptor are very similar to those obtained with VPAC 1 receptors with some specific differences. The great similarities indicate that VIP binding to VPAC 1 and VPAC 2 receptor is mostly ensured by a common set of crucial amino acids along the VIP peptide sequence. However, two important differences can be noted: (i) some VIP analogs exhibit a much lower binding affinity for the VPAC 2 than for the VPAC 1 including the D8A, Y10A, and L23A mutants. Two other analogs have no alteration of binding to VPAC 1 receptor whereas they exhibit a significant decrease of affinity for VPAC 2 receptor, i.e. T11A and N28A analogs (see Fig. 5 and Table II)  the resulting peptide behaves as a VIP agonist which clearly discriminates between human VPAC 1 and VPAC 2 receptors (see Fig.  7). (ii) The Y22A VIP analog exhibits an increased EC 50 for stimulating adenylyl cyclase activity through interaction with the human VPAC 2 receptor whereas no change was observed at the VPAC 1 receptor (see Tables I and II). Altogether, these data indicate that single amino acid substitutions in VIP either alter similarly peptide activity at the two VIP receptor subtypes or evoke a higher decrease of activity at the VPAC 2 receptor than at the VPAC 1 receptor. In other words, no selective VPAC 2 receptor ligand was observed in the series of analogs obtained by the Ala scan. Conversely, we took advantage of the selectivity observed toward the VPAC 1 receptor (see above) to develop a VIP analog which combined 3 mutations at positions 11, 22, and 28 which were shown to be individually important for discriminating human VPAC 1 and VPAC 2 receptors. The resulting [Ala 11,22,28 ]VIP analog is an agonist which exhibits an exquisite high selectivity for VPAC 1 receptor with EC 50 Ͻ1 nM and Ͼ1,000 nM in stimulating adenylyl cyclase activity through interaction with human VPAC 1 and VPAC 2 receptors, respectively (see Fig. 7). Two VPAC 1 receptor selective agonists derived from secretin and GRF were previously described (59,60). Chicken [Arg 16 ]secretin, although it discriminates between rat VPAC 1 and VPAC 2 receptors, has a high affinity for rat secretin receptor (59). Moreover, it has a rather weak affinity for the human VPAC 1 receptor (60), in line with the important differences between species in the pharmacology of VPAC 1 receptors (15,44). The chimeric, substituted peptide [Lys 15 ,Arg 16 ,Leu 27 ]VIP (1-7)/GRF (8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27) has an important selectivity for VPAC 1 receptor but its possible interaction with GRF receptors was not directly evaluated (61). In this context, the highly selective human VPAC 1 receptor agonist developed in this study by rationale combination of three mutations in VIP itself constitutes henceforth the most operative pharmacological tool derived from VIP for characterizing VPAC 1 receptor-mediated events.
In conclusion, the present activity data on the series of VIP analogs in combination with molecular modeling provide the first characterization of the functional properties of the VIP molecule for interaction with the two human VIP receptor subtypes. Key residues for interaction of VIP with human VPAC 1 or VPAC 2 receptor are highlighted in the ribbon representation of the VIP molecule (Fig. 8). In view of considerable interest of this widespread neuropeptide in health and diseases, this work already provides useful pharmacological tools and paves the way for further development in the design of new analogs and mimetic versions of VIP.