Docking of Linear Peptide Antagonists into the Human V1a Vasopressin Receptor

A novel photoactivatable linear peptide antagonist selective for the V1a vasopressin receptor, [125I][Lys(3N3 Phpa)8]HO-LVA, was synthesized, characterized, and used to photolabel the human receptor expressed in Chinese hamster ovary cells. Two specific glycosylated protein species at 85–90 and 46 kDa were covalently labeled, a result identical to that obtained with a previous photosensitive ligand, [125I]3N3Phpa-LVA (Phalipou, S., Cotte, N., Carnazzi, E., Seyer, R., Mahe, E., Jard, S., Barberis, C., and Mouillac, B. (1997) J. Biol. Chem.272, 26536–26544). To identify contact sites between the new photoreactive analogue and the V1a receptor, the labeled receptors were digested with Lys-C or Asp-N endoproteinases and chemically cleaved with CNBr. Fragmentation with CNBr, Lyc-C, and Asp-N used alone or in combination, led to the identification of a restricted receptor region spanning the first extracellular loop. The results established that sequence Asp112–Pro120 could be considered as the smallest covalently labeled fragment with [125I][Lys(3N3Phpa)8]HO-LVA. Based on the present experimental result and on previous photoaffinity labeling data obtained with [125I]3N3Phpa-LVA (covalent attachment to transmembrane domain VII), three-dimensional models of the antagonist-bound receptors were constructed and then verified by site-directed mutagenesis studies. Strikingly, these two linear peptide antagonists, when bound to the V1a receptor, could adopt a pseudocyclic conformation similar to that of the cyclic agonists. Despite divergent functional properties, these peptide antagonists could interact with a transmembrane-binding site significantly overlapping that of the natural hormone vasopressin.

Over the past few years, interest in locating ligand-binding sites in G protein-coupled receptors has increased exponentially. Indeed, identification of these binding sites is of prime importance both for a better understanding of the structure and the function of the G protein-coupled receptor superfamily and for facilitating rational design of potential therapeutic agents. Extensive mutational analysis and receptor three-dimensional molecular modeling have led to valuable information concerning "small ligand" and peptide/protein ligand re-ceptor-binding sites (for review, see Refs. [1][2][3][4][5][6]. In 1995, we published the mapping of arginine-vasopressin (AVP) 1 -binding site in the V 1a receptor subtype and described a major localization within transmembrane regions (TMR) in a position equivalent to that defined for the cationic neurotransmitters (7). Because all receptor residues potentially interacting with AVP are conserved in the different members of the AVP/oxytocin (OT) receptor family, we proposed that the binding pocket identified in the V 1a might be common to V 2 , V 1b , and OT receptor subtypes. Extracellular residues responsible for receptor-selective and species-selective binding have also been identified (8 -10). Unfortunately, these first analyses of AVP receptor structure/function relationships did not provide much information on AVP receptor antagonist-binding domains (Refs. 11 and 12, and for review see Ref. 13). The photoaffinity labeling technique is an essential complement to modeling and mutagenesis approaches and allows direct unambiguous identification of the contact regions between a receptor and its specific photoactivatable ligands (for review see Ref. 14). At the present time, very few photoaffinity labeling studies have led to the direct determination of labeled amino acid residues in peptide G protein-coupled receptor; remarkable results with bovine V 2 receptor (15), human NK1 tachykinin receptor (16), and rat type A cholecystokinin receptor (17) allowed identification of covalently labeled residues with photoactivatable agonist analogues of AVP, substance P, and cholecystokinin, respectively.
Very recently, a first radioiodinated photoreactive linear peptide antagonist has been used in our laboratory to photolabel the human and rat V 1a receptors (12,18,19). Our results have clearly indicated that covalent attachment of the [ 125 I]3N 3 Phpa-LVA occurs in a restricted domain of the human receptor including TMR VII. Based both on this photolabeling result and on the hypothetical three-dimensional model of the human V 1a receptor, residues potentially involved in binding and affinity of the antagonist ligand have been targeted. An interaction between the hydrophobic N terminus of the [ 125 I]3N 3 Phpa-LVA ligand and an aromatic cluster of residues in the TMR VI has thus been experimentally verified. However, because of the lack of structural and conformational data for this family of V 1a -selective compounds and because of their peptidic nature and highly variable linear structure, the determination of a single contact point between the peptide antagonist and the receptor does not provide enough information to propose a docking mode of the ligand into the receptor.
To allow a more complete location of the binding sites for this family of V 1a linear peptide antagonists and to generate meaningful information on receptor-antagonist interactions, we thus decided to label the human V 1a receptor with a second radioiodinated photoreactive antagonist. In the present study, properties of [ 125 I][Lys(3N 3 Phpa) 8 ]HO-LVA, an antagonist containing an azido group at a position (side chain of lysine residue 8) likely to covalently bind another domain of the receptor are described. Combining photolabeling with this new ligand of the human V 1a receptor, cyanogen bromide cleavage and endoproteinase digestions of the receptor, a restricted photolabeled domain has been identified which spans the first extracellular loop from Asp 112 to Pro 120 . Taking into account the present and previous photolabeling data, three-dimensional models of antagonist-bound receptors have been constructed and experimentally verified. Residues possibly involved in the interactions with the photoactivatable antagonists have been mutated, and affinities of the mutant receptors for the ligands have been estimated.
Cell Culture and Membrane Preparation-The human V 1a receptor cDNA was a generous gift of Dr. M. Thibonnier (24), and CHO cells were transfected as described before (12). CHO cells stably expressing the human V 1a , V 1b , V 2 , or OT receptor were maintained in culture in Petri dishes in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10% fetal calf serum, 4 mM L-glutamine, 500 units/ml penicillin and streptomycin, 0.25 g/ml amphotericin B in an atmosphere of 95% air and 5% CO 2 at 37°C. Depending on the experiment to be conducted, cells were treated overnight with 5 mM sodium butyrate to increase receptor expression (25,26). Cells were harvested, washed two times in PBS without Ca 2ϩ and Mg 2ϩ , polytron-homogenized in lysis buffer (15 mM Tris-HCl, pH 7.4, 2 mM MgCl 2 , 0.3 mM EDTA) and centrifuged at 100 ϫ g for 5 min at 4°C. Supernatants were recovered and centrifuged at 44,000 ϫ g for 20 min at 4°C. Pellets were washed in Buffer A (50 mM Tris-HCl, pH 7.4, 5 mM MgCl 2 ) and centrifuged at 44,000 ϫ g for 20 min at 4°C. Membranes were suspended in a small volume of Buffer A, and protein contents were determined. Aliquots of membranes were used immediately in binding assays and photolabeling experiments or stored at Ϫ80°C.
Radioligand Binding Assays-As described in previous papers (7,12) I]HO-LVA (50 -100 pM) as the radioligands. The concentrations of the unlabeled ligands varied from 1 pM to 10 M. AVP (10 M) and HO-LVA (400 nM) were used to determine nonspecific binding, respectively. Bound and free radioactivity were separated by filtration over bovine serum albumin presoaked Whatman GF/C filters. The ligand binding data were analyzed by nonlinear least squares regression using the computer program Ligand (27).
Inositol Phosphate Assays-Inositol phosphate (IP) accumulation was determined as described (28). Briefly, CHO cells expressing the human V 1a receptor were plated and grown in 6-well dishes for 48 h in Dulbecco's modified Eagle's medium-supplemented medium and then labeled for 24 h with myo-[2-3 H]inositol (10 -20 Ci/mmol; NEN Life Science Products) at a final concentration of 1 Ci/ml in a serum-free, inositol-free medium (Life Technologies, Inc.). Cells were washed twice with PBS medium, equilibrated at 37°C in PBS for 1 h, and then incubated for 10 min in PBS supplemented with 10 mM LiCl in the presence or absence of increasing concentrations of [Lys(3N 3 Phpa) 8 ]-HO-LVA or 3N 3 Phpa-LVA (from 10 Ϫ12 to 10 Ϫ6 M). CHO cells were stimulated for 15 min with 10 Ϫ9 M AVP (a concentration close to the K act value determined in CHO cells). After stopping the reaction with perchloric acid, total IPs were extracted and purified by anion exchange chromatography column (Dowex AG1x8, formate form, 200 -400 mesh; Bio-Rad). For each sample, a fraction containing total IPs was collected and counted. K inact constants were calculated as K inact ϭ IC 50 /(1 ϩ [AVP]/K act ), in which IC 50 is the concentration of antagonist leading to 50% inhibition, [AVP] ϭ 1 nM and K act is the concentration of AVP inducing half-maximal accumulation of IP (K act ϭ 0.32 nM in CHO cells expressing the wild-type human V 1a receptor (12)).
Photoaffinity Labeling Experiments-These experiments were conducted as described before for the previous photoactivatable antagonist [ 125 I]3N 3 Phpa-LVA (12). Briefly, membranes (500 g) were resuspended in 4 ml of binding Buffer A containing bovine serum albumin (0.5 mg/ml) and were incubated for 1 or 3 h in the dark in the presence of [ 125 I][Lys(3N 3 Phpa) 8 ]HO-LVA (1-2 nM) with or without vasopressin (10 M) to define specific labeling. Membranes were separated from unbound ligand by two subsequent centrifugations (20 min, 44,000 ϫ g, 4°C) and washed with Buffer A. The final pellet was resuspended in 1 ml of Buffer A and irradiated with UV light (254 nm) for 1 min on ice. After photolysis, membranes were washed twice (2 ϫ 1 ml of Buffer A) and finally resuspended in Laemmli buffer (29). Samples were subjected to SDS-polyacrylamide gel electrophoresis (PAGE) using 1-mmthick 12% cross-linked gels. Gels were fixed in glacial acetic acid: methanol:Me 2 SO:water (16:40:2:42), dried, and exposed to Kodak XAR-5 film at Ϫ80°C. In order to evaluate the yield of covalent binding, dried gels were also cut into slices, and radioactivity contents in the slices were determined. Covalent binding was calculated as percentage of the total number of receptor pmol expressed in the membrane preparation.
Electroelution of the Photolabeled V 1a Receptors-Photolabeled membranes were subjected to SDS-PAGE using 12% cross-linked gels. The labeled bands were excised from the preparative gel and the human V 1a vasopressin receptor was electroeluted with electroeluter model 422 (Bio-Rad) in Tris/glycine running buffer (25 mM Tris, 182 mM glycine, pH 8.3, 0.1% SDS). Samples containing "partially purified" V 1a receptors were washed and concentrated using Microcon-30 (Amicon). Deglycosylation with N-glycosidase F and fragmentation with cyanogen bromide and endoproteases were performed on these concentrated samples.
Cyanogen Bromide and Endoproteinase Digestions-The electroeluted receptors were subjected to digestion with CNBr, Lys-C protease, or Asp-N protease. Double digestions were also performed (first digestion with CNBr followed by a second one with Lys-C or Asp-N protease). CNBr (a few crystals) cleavage of the electroeluted V 1a receptors was carried out on samples in a 100-l volume of 70% (v/v) formic acid. The mixture was incubated in the dark for 24 h at room temperature under argon, and the reaction was then stopped by adding 500 l of water. Sample volume was reduced under vacuum, and solvent exchange (removing formic acid) with water was accomplished. Endoproteinase Lys-C (sequencing grade from Lysobacter enzymogenes, Roche Molecular Biochemicals) was used at 0.2 g/assay in a final 50-l volume. The digestion was performed in 25 mM Tris-HCl, pH 8.5, 1 mM EDTA, 0.1% SDS at 37°C for 16 -24 h and stopped by addition of Laemmli buffer. Endoproteinase Asp-N (sequencing grade from Pseudomonas fragi, Roche Molecular Biochemicals) was used at 0.1 g/assay for 48 h at 25°C in a final 50-l volume of 25 mM Na 2 HPO 4 , pH 7.8. To perform double fragmentations, the CNBr digests were washed several times with water, and volume sample was reduced before dilution in Lys-C or Asp-N buffer. Results of CNBr and protease digestions were analyzed by a Tricine discontinuous SDS-PAGE system (10 -16.5% cross-linked gels) applied to the separation of small molecular mass species (30). Gels were then fixed in glacial acetic acid:methanol:Me 2 SO:water (10: 50:2:38), dried, and exposed to Kodak XAR-5 film at Ϫ80°C.
Computer Three-dimensional Molecular Modeling and Docking of Photoactivatable Antagonists-The three-dimensional model of the human V 1a vasopressin receptor hosting AVP was constructed using the procedure already extensively described in a previous publication for the rat V 1a receptor (7). Briefly, the transmembrane part of the V 1a receptor was constructed by using the three-dimensional model first developed on the bacteriorhodopsin experimental structure (31,32) and refined on the bovine rhodopsin footprint (33). The extracellular regions of the receptor were then built in an acceptable conformation. The rat residues were exchanged for the corresponding human residues. Transmembrane domain VII was rotated of about 20°to bring residues Thr710 and Ala711, 2 pointing both toward the core of the binding cleft. 3 The whole receptor structure was then energy minimized in order to relax the structure and to remove unfavorable steric constraints. AVP was manually docked as reported and validated in previous publications (7,9). In the next step, [ 125 I][Lys(3N 3 Phpa) 8 ]HO-LVA was docked into the binding cleft. First, the ligand was built using Sybyl 6.3 (TRI-POS Associates, Inc.) facilities and energy minimized in a fully extended conformation. The backbone and side chain of the photoactivatable ligand have then been superimposed on the corresponding features of AVP docked into the V 1a receptor, as already published. The conformations of the ligand and of the side chains have been manually controlled to optimize the electrostatic and steric interactions with the receptor walls. The ligand-receptor complex has then been minimized without constraints (TRIPOS force field, Gasteiger-Hü ckel charges, 5,000 iterations). Alternative binding modes have been tried, but none of them was more satisfactory than the AVP-based model, on a physicochemical and energetical point of view. The same practical procedure has been followed to dock the previously published photoaffinity labeling agent, [ 125 I]3N 3 Phpa-LVA (12).
Site-directed Mutagenesis of the Human V 1a Receptor-The construc-  Table IV for numbering) has been reported in a previous paper (12). Point mutations Y225D, Q218A, K308A and Q413A were introduced in the human V 1a vasopressin receptor using the QuickChange site-directed mutagenesis kit (Stratagene). These substitutions were directly done on the eukaryotic expression vector pCMV (34) and verified by direct dideoxynucleotide sequencing (T7 Sequencing TM kit, Amersham Pharmacia Biotech). All the mutant receptors were transiently expressed in COS7 cells. In all cases, cell membrane preparations and radioligand binding assays were conducted as described above.

RESULTS
Chemical, Pharmacological, and Functional Properties of the Photoactivatable Linear Peptide-Considering firstly that all V 1a structurally related AVP linear peptide antagonists probably bind the V 1a receptor in the same way and secondly that introduction of an arylazido group at the N terminus of the ligand led to covalent binding into the TMR VII of the V 1a receptor (12), we designed, synthesized, and characterized a novel photoactivatable linear peptide compound ( The preparations were incubated for 60 min at 30°C in Buffer A with 1 mg/ml bovine serum albumin and increasing concentrations of the radiolabeled peptides. Nonspecific binding was determined in the presence of 250 -1000-fold excess of corresponding unlabeled peptides. Data were analyzed using the Ligand progam (27). The values represented in the graphs are taken from one experiment. They are representative of at least three independent experiments done in triplicate. B and E, selectivity of the peptides for the V 1a receptor. Competition experiments were conducted with membrane preparations from CHO cells expressing the V 1a , OT, V 1b , or V 2 receptor subtypes at 30°C for 30 min in the presence of [ 3 H]AVP and unlabeled [Lys(3N 3 Phpa) 8 ]HO-LVA (B) or 3N 3 Phpa-LVA (E) peptides. Affinities (K i ) for both photoactivatable compounds were determined by displacement of radiolabeled AVP and data (see also Table I) were analyzed with the Ligand program (27). All values in the graphs are taken from representative experiments that were done at least three times each in triplicate. C and F, antagonistic properties of the peptides. myo-[ 3 H]Inositol-prelabeled CHO cells expressing the human V 1a receptor were incubated 10 min at 37°C in PBS/10 mM LiCl medium with or without increasing concentrations of [Lys(3N 3 Phpa) 8 ]HO-LVA (C) or 3N 3 Phpa-LVA (F) ranging from 10 Ϫ12 to 10 Ϫ6 M. Cells were then stimulated with AVP 10 Ϫ9 M (a concentration close to the AVP K act in this system) for 15 min. After stopping the reaction, total inositol phosphates were extracted, counted, and expressed as dpm/well. For example, in panel C, basal activity of phospholipase C was 280 dpm/well (50,000 cells); half-maximal IP accumulation with AVP 10 Ϫ9 M was 6450 dpm/well; phospholipase C activity in the presence of 1 nM AVP and 1 M Lys(3N 3 Phpa) 8 ]HO-LVA was 293 dpm/well, phospholipase C activity in the presence of 10 Ϫ6 M Lys(3N 3 Phpa) 8 ]HO-LVA alone was 302 dpm/well. All values in each graph are taken from one experiment and are representative of three independent experiments each done in triplicate. Data were analyzed using the software Kaleidagraph TM . K inact constants were calculated as described under "Experimental Procedures." antagonist HO-LVA (23), we introduced a Lys 8 derivatized with a 3-azidophenylpropionyl group in order to photolabel the V 1a receptor. The peptidic part of the molecule was synthesized on p-methylbenzhydrylamine resin (18), and the arylazido group was attached in solution with PyBOP under a dual wavelength reverse phase HPLC monitoring (22), allowing the identification of the target compound [Lys(3N 3 Phpa) 8 ]HO-LVA (Fig. 2). This ligand was purified by semi-preparative reverse phase HPLC and characterized by mass spectroscopy (fast atom bombardment mass spectrometry, Mϩ1 ϭ 1287) and UV spectroscopy (azido characteristic peak at 250 nm, ⑀ ϭ 7000 cm Ϫ1 M Ϫ1 , destroyed by 254-nm UV irradiation). As for the HO-LVA, this compound can be iodinated on its phenolic Nterminal blocking group (4HO-Phpa) using ICl and purified by HPLC in order to obtain the nonradioactive probe as a chromatographic reference (fast atom bombardment mass spectrometry, Mϩ1 ϭ 1412). As shown in Fig. 3A Fig. 3B, displacement of [ 3 H]AVP with the photoactivatable peptide allowed the measurement of K i values of 11, 574, and Ͼ2000 nM for human OT, V 1b , and V 2 receptor subtypes, respectively. These affinities were 60 -10,000-fold lower than that measured for the human V 1a AVP receptor (0.18 nM), establishing the [Lys(3N 3 Phpa) 8 ]HO-LVA as a selective ligand. This photosensitive ligand displays competitive pharmacological properties equivalent to those of the 3N 3 Phpa-LVA antagonist (Table I and Fig. 3E for comparison of the selectivity profiles of these two peptides). As shown in Fig. 3C, the photoactivatable peptide potently inhibited the IP accumulation induced by AVP (1 nM, a concentration producing a half-maximal response) in CHO cells in a concentration-dependent manner. The average K inact calculated from experimental IC 50 values was 80 Ϯ 25 pM (n ϭ 3), a value close to that determined for 3N 3 Phpa-LVA (130 Ϯ 40 pM (n ϭ 3); see Fig. 3F for comparison). This K inact value is also in agreement with the K i determined (180 pM;  8 ]HO-LVA covalently labeled the human V 1a receptor, as demonstrated by autoradiography of SDS-polyacrylamide gels used to separate proteins after binding to CHO cell membranes and UV irradiation (Fig. 4A). The photolabeled receptor migrated as two broad bands at approximately 85-90 and 46 kDa, respectively (lane 1). The labeling of the receptor was completely suppressed with an excess of AVP (lane 2), indicating that the photolabeling of both protein bands was specific for the human V 1a receptor expressed in the CHO cell line. The covalent binding yield of the probe [ 125 I][Lys(3N 3 Phpa) 8 ]HO-LVA to the receptor was calculated, and the fraction of specifically receptor-bound radioactivity recovered in the labeled bands reached 17-18% (multiple determinations). This covalent binding yield was high and exceeded the one reported before for [ 125 I]3N 3 Phpa-LVA (13-14%). As described in a previous study (12), the 46-kDa photolabeled band likely corresponds to a proteolytic truncated form of the entire receptor migrating at an apparent 85-90 kDa molecular mass. Indeed, the relative abundance of the 46-kDa species (Ϸ50% in lane   Fig. 4B, the treatment of photolabeled membranes with N-glycosidase F before SDS-PAGE reduced apparent molecular masses of the 85-90-and 46-kDa protein bands (lane 1) to approximately 50 and 34 kDa, respectively (lane 2), indicating that they were both glycosylated and contained at least one N-glycosylated site. As shown in Fig. 5, potential N-glycosylation sites (Asn-Xaa-Ser/Thr) are located at Asn 14 , Asn 27 , and Asn 196 . 50 kDa is a molecular mass very close to the theoretical mass of the receptor core deduced from the cDNA sequence (48.2 kDa, including the 1.4-kDa antagonist mass), whereas 34 kDa is significantly smaller. This observation confirmed once again that the photolabeled glycosylated protein band at 46 kDa effectively corresponds to a proteolytic truncated form of the receptor and that the photolabeled species at 85-90 kDa likely represents the native glycosylated state of the receptor expressed in the CHO cell system. Looking at the primary sequence of the receptor (Fig. 5) and at the localization of the different N-glycosylation sites (Asn 14 , Asn 27 , and Asn 196 ), a FIG. 5. Schematic representation of the human V 1a receptor. The primary sequence of the receptor, deduced from the cDNA cloning (24), is shown as well as the possible arrangement of the protein through the cell membrane. The three potential N-glycosylation sites are shown at Asn 14 , Asn 27 , and Asn 196 . White squares indicate a consensus sequence (residues 103-108) spanning the cleavage site for a metalloproteinase (15,36). Potential cleavage sites for Lys-C (gray triangles), Asp-N (gray squares), and CNBr (gray diamonds) are also indicated. As described previously (12)  deglycosylated 34-kDa protein (or the glycosylated 46-kDa species counterpart) could only account for a large truncated receptor fragment including Asn 196 and spanning the protein to the C terminus. Taking into account both the present results of deglycosylation and sensitivity of the receptor to endogenous membrane protease degradation, we concluded that proteolytic cleavage must occur at a site located between Asn 27 and Asn 196 . As mentioned before (12), sequence Phe 103 -Gln 108 in the V 1a receptor (Fig. 5) corresponds to a potential metalloproteinase cleavage site. This enzyme has been clearly shown (36) to digest the bovine renal V 2 receptor in membrane preparations between Gln 92 and Val 93 (Gln 104 and Val 105 in the V 1a sequence). The 34-kDa molecular mass of the deglycosylated photolabeled band is consistent with that of a proteolytic fragment that could be generated from such a cleavage in this receptor region. We also concluded that photolabeling of the  8 ]HO-LVA, photolabeled receptors were partially purified from a preparative SDS-PAGE by electroelution and then subjected to fragmentation with either CNBr, endoproteinase Lys-C, or endoproteinase Asp-N. Because CNBr, Lys-C, and Asp-N cleave proteins specifically at the C terminus of methionine and lysine residues and at the N terminus of aspartic acid residues, respectively (Met, Lys, and Asp in the V 1a receptor are shaded in Fig. 5), the photoactivatable peptide ligand itself is thus expected to be protected against these different fragmentations (Lys 8 is modified and constitutes the amidated C terminus end of the antagonist). Because the 46-kDa photolabeled band derives from the 85-90-kDa photolabeled species, only the truncated receptor at 46 kDa was excised from gels, electroeluted, and subjected to the chemical cleavage and enzymatic digestions.
As seen in Fig. 6A, CNBr cleavage of the photoaffinitylabeled V 1a receptor yielded a major labeled fragment migrating at an apparent mass Ϸ5.5 kDa and a higher minor labeled band at Ϸ7.5 kDa. A similar cleavage pattern was observed when electroeluted 46-kDa fragment was first deglycosylated with N-glycosidase F before CNBr treatment (data not shown), indicating that radiolabeled fragments at Ϸ5.5 and 7.5 kDa do not contain N-glycosylation sites. This result eliminates fragments Arg 2 -Met 86 (a fragment that could also be eliminated on the basis that this sequence is not included in the truncated 46-kDa photolabeled receptor) and As shown in Fig. 6B, digestion of the photolabeled 46-kDa receptor with Lys-C endoproteinase yielded a major fragment at Ϸ5 kDa and two minor higher labeled bands at Ϸ8 and 14. As seen in Fig. 6C, digestion of the radioiodinated antagonist-bound 46-kDa photolabeled species with Asp-N endoproteinase yielded an intense major Ϸ2-2.5-kDa labeled fragment and very minor higher labeled bands at Ϸ6, 12.5, and 18 kDa. Only one possible Asp-N fragment of the V 1a receptor could account for a labeled band with an electrophoretic migration at 2-2.5 kDa; this fragment, Asp 112 -Pro 120 , spans the first extracellular loop of the V 1a receptor. The localization of this short receptor domain, as the site of covalent binding site for [ 125 I][Lys(3N 3 Phpa) 8 ]HO-LVA, was consistent with the results of CNBr and Lys-C cleavages described above. In order to confirm this localization, CNBr cleavage was also used in combination with either Lys-C or Asp-N protease digestion. Successive treatment of the 46-kDa photolabeled receptor with CNBr and Lys-C (Fig. 7A, lane 2) generated a new fragment slightly smaller than the 5 kDa obtained with Lys-C alone (lane 1) or the 5.5 kDa produced with CNBr alone (lane 3). This labeled receptor fragment with a molecular mass of Ϸ4.5 kDa could correspond to Cys 110 -Lys 128 sequence. Successive treatment of the 46-kDa labeled species with CNBr and Asp-N endoproteinase (Fig. 7B, lane 2) demonstrated that fragments obtained with CNBr alone (lane 1) were converted into smaller ones with Asp-N protease to yield the smallest fragment with an apparent mass (2-2.5 kDa) equivalent to that generated with Asp-N protease alone (lane 3). The results of double fragmentations were consistent with those of CNBr cleavage and contact with TMR VII that has been found to be photolabeled (12). According to the present model, the residues that might be covalently bound are Thr710, Ala711, Gly714, Ser715, and possibly Asn717 and Ser718 (Fig. 8B). Phe616 is also found in the direct neighborhood in agreement with site-directed mutagenesis data (12). As in AVP (7), only Arg 8 side chain protrudes toward the extracellular loop 1 of the human V 1a receptor in order to potentially interact with Tyr 225 .
In the model, the 4-hydroxy-3-[ 125 I]phenylpropionyl group of [ 125 I][Lys(3N 3 Phpa) 8 ]HO-LVA is found in contact with residues of TMR VII, in a position equivalent to that proposed for the 3-azidophenylpropionic moiety of [ 125 I]3N 3 Phpa-LVA. In contrast, the ninth residue is not present in [ 125 I][Lys(3N 3 Phpa) 8 ]HO-LVA, and Arg 8 has been replaced by the 3-azidophenylpropionylated lysyl. The backbone constraints are such that the modified lysyl side chain has to point toward the extracellular part of the receptor. Consequently, the azidophenyl moiety is found directly in contact with the first extracellular loop (Fig. 8D), in the neighborhood of residues Tyr225, Arg226, Phe227, and Arg228, in full agreement with the labeling data reported here.
Site-directed Mutagenesis of the Human V 1a Receptor-In order to verify the proposed docking of both linear peptide antagonists into the V 1a receptor, the role of some receptor residues (Tables II and III) potentially interacting with the ligands, was investigated. These residues were mutated, and binding properties of the mutant receptors transiently expressed in COS7 cells were studied (Table IV). Because we have already demonstrated important aromatic/aromatic interactions between [ 125 I]3N 3 Phpa-LVA and the aromatic residue cluster (particularly Phe616) of TMR VI (12), affinities of [Lys(3N 3 Phpa) 8 ]HO-LVA for mutant receptors W613A, F616V, and F617L were measured. Mutation of Phe616 into a valine led to a dramatic loss in binding affinity (1266-fold reduction, 228 nM compared with 0.18 nM for the wild-type receptor). This result confirmed the crucial role played by Phe616 in defining high affinity of the V 1a -selective linear peptide antagonists. Consequences of Trp613 and Phe617 mutations in binding affinity were not significant. Conserved hydrophilic residues Gln218, Lys308, and Gln413 that have been demonstrated to control the binding of AVP in the rat V 1a receptor (7) were substituted with alanine residues. Binding properties of the mutants were measured using [ 125 I]HO-LVA as the radioligand (Table IV); a significant decrease in the affinity of [Lys(3N 3 Phpa) 8 ]HO-LVA and 3N 3 Phpa-LVA for the three mutants was observed (5-28-fold reduction in K i values). Finally, we decided to investigate a potential role for Tyr225, located in the first extracellular loop (corresponding to Tyr 115 in the primary sequence), in the affinity and selectivity of the photoactivatable antagonist ligands because (i) this residue has been shown (9) to control receptor subtype selectivity and participate in agonist high affinity binding by interacting with hor-  (Table IV). This small reduction in binding affinity was emphasized for 3N 3 Phpa-LVA, the previous photoactivatable antagonist ligand: in this case, K i shifted from 0.24 to 4.9 nM (20-fold reduction in affinity). In conclusion, Tyr225 could participate in the binding of both [Lys(3N 3 Phpa) 8 ]HO-LVA and 3N 3 Phpa-LVA peptides, but its role in the receptor-selective binding of these two ligands is only minor. Taken together, these mutagenesis data suggest that residues Gln218, Tyr225, Lys308, Gln413, and Phe616 contribute to the binding of   3 Phpa LVA/human V 1a receptor interactions as observed in the model In the human V 1a receptor column, the numbering of the residues is taken from the G protein-coupled receptor alignment and does not correspond to the numbering of the amino acids in the primary sequence. The first digit corresponds to the helical TMR, and the next two digits indicate the rank of the residue in the considered helix. Tyr225 is numbered according to the same rule but is no longer in the second TMR. It is located in the first extracellular loop between TMR II and TMR III (residue 115 in the primary sequence  3 Phpa-LVA, the previous photoactivatable antagonist used for mapping peptide-binding domains of the rat and human V 1a vasopressin receptors (18,19,12), this novel antagonist analogue also combines high affinity, selectivity, possibility of radioiodination and high covalent binding yield. The novel ligand allowed us to photolabel the receptor which migrated on SDS-PAGE as two protein bands with apparent molecular masses of 85-90 and 46 kDa, respectively. This result is comparable with that previously observed with [ 125 I]3N 3 Phpa-LVA (12), thus confirming the identity of the receptor and the specificity of the signal. Once again, the human V 1a vasopressin receptor was degraded during incubation with the ligand. As demonstrated in our previous study, the 46-kDa species deriving from the larger protein at 85-90 kDa is cleaved by endogenous proteases present in the CHO membrane preparations. Deglycosylation of the photolabeled receptors with N-glycosidase F confirmed this observation. Fragmentation of the photolabeled V 1a vasopressin receptor with a combination of chemical CNBr cleavage and Lys-C and Asp-N protease digestions led to the identification of a restricted receptor region that likely spans the first extracellular loop. Results of double fragmentations (CNBr followed by Lys-C or Asp-N proteases) were all consistent with that of single digestions or chemical cleavages and confirmed the identity of the Asp 112 -Pro 120 sequence as the smallest receptor fragment photolabeled with [ 125 I][Lys(3N 3 Phpa) 8 ]HO-LVA.
Interestingly, chemical CNBr cleavage and endoproteinase Lys-C digestion of the photolabeled V 1a receptor yielded labeled fragments probably corresponding to Cys 110 -Met 135 and Val 105 -Lys 128 , respectively. Both receptor fragments contain Cys 124 , a residue located in the TMR III and possibly involved in a disulfide bond with Cys 203 in the second extracellular loop (for review see Refs. 13, 31, and 32). This observation prompted us to verify the presence of such a disulfide bridge in the human vasopressin V 1a receptor. Thus, preparation of samples for CNBr cleavage as well as migration on SDS-PAGE were then performed in nonreducing conditions in order to possibly generate a larger photolabeled fragment including both sequence Cys 110 -Met 135 and Ile 192 -Met 220 (fragment containing Cys 203 ). The result of CNBr cleavage in this case was equivalent to that obtained in reducing conditions: only photolabeled fragments at 5.5 and 7.5 kDa were produced (data not shown), suggesting that disulfide bond between Cys 124 and Cys 203 would be absent in the V 1a receptor. This conclusion is in agreement with those of several previous reports. Indeed, it has been demonstrated that affinity of [ 3 H]AVP to V 1a receptors on rat liver membranes was not significantly reduced after treatment with sulfhydryl reagents such as N-ethylmaleimide (37,38). This result suggests either that the disulfide bridge is missing or that it does not play any role in AVP binding. A different result has been obtained for the V 2 receptor. For this receptor subtype, it has been shown that one or more cysteine residues are essential for hormone binding (38).  8 ]HO-LVA, spans the first extracellular loop of the V 1a receptor. This region has already been shown to constitute the site of covalent attachment in the bovine renal V 2 receptor with a photoactivatable agonist (15). In this case, the AVP analogue [ 3 H]1-deamino[Lys 8 ]vasopressin contained the photoreactive arylazido group at the side chain of Lys 8 . Determination of the first extracellular loop of the V 2 receptor as the site of interaction with AVP analogue residue 8 led to the identification of Asp 103 as the residue responsible for agonist binding specificity (8). Independently and at the same time, equivalent result was demonstrated in our laboratory for the V 1a receptor: Tyr 115 (homologue to Asp 103 in the V 2 ) was shown to behave as a crucial residue for agonist high affinity binding and also for receptor selectivity (9). The position of the photoreactive moiety (azidophenyl) in the [ 125 I][Lys(3N 3 Phpa) 8 ]HO-LVA antagonist used in the present study is again at the side chain of peptide residue 8. Interestingly, both peptide agonist and antagonist photoaffinity ligands allow covalent attachment of the first extracellular loop in the V 2 and V 1a vasopressin receptors respectively when containing the photoactivatable group at the side chain of Lys 8 . However, residues in this extracellular region of vasopressin receptors responsible for agonist receptor subtype selectivity only play a minor role for antagonists. Other binding selectivity determinants for these compounds have to be discovered elsewhere in the receptor. Photolabeling of the human V 1a receptor with two different selective linear photoactivatable and iodinatable peptide antagonists of AVP led to the identification of two receptor regions in close proximity to the bound photoligands (Ref. 12 and the present study). Taking into account these photoaffinity labeling results, three-dimensional models for the antagonist peptide-binding sites of the V 1a vasopressin receptor were then proposed and verified experimentally. The two photoactivatable ligands are linear molecules. They are both very flexible and can adopt a quasi-infinite number of conformations within an acceptable energy window. Therefore, looking for the minimum energy conformers would not provide any clue regarding the receptor-bound conformations. An alternative way consists in examining the physicochemical features of both the ligands and the binding cleft and looking for complementarity. However, it is interesting to observe that [ 125 I][Lys(3N 3 -Phpa) 8 ]HO-LVA and [ 125 I]3N 3 -Phpa-LVA present a striking homology with AVP whose binding mode had already been probed (7). Most of the receptor residues putatively involved in the binding of these two linear photoactivatable peptide antagonists are those already demonstrated to interact with AVP. These two peptides differ from AVP in positions 1, 2, 6, and 8 and at the C terminus but retain the same residues in positions 3, 4, 5, and 7. In the models, the side chains of the 4 conserved residues bind in a similar way compared with homologous features in AVP. Interestingly, the docking procedures lead these antagonists to adopt a pseudocyclic conformation similar to that of the cyclic AVP. Based on these models, the linear peptide antagonists could enter the transmembrane-binding pocket like their agonist counterparts and establish their own network of molecular interactions. This is consistent with the marked hydrophobic nature of these ligands and that of the bottom of the receptorbinding cleft. Interestingly, as confirmed from the mutagenesis results, aromatic/aromatic contacts represent the most important interactions for antagonists, whereas hydrogen bonds with conserved hydrophilic receptor residues seem to represent the most crucial interactions for agonists like AVP (7).
It is now commonly accepted that peptide ligands, agonists or antagonists, partly bind to transmembrane domains of their receptors (for review see Ref. 6). It has been demonstrated with different approaches, such as site-directed mutagenesis, photoaffinity labeling, or spectrofluorimetric methods. However, very few studies have led to the identification of binding sites for peptide antagonists. In the NK-2 neurokinin receptor systems, the N-termini of agonists and of antagonists of similar molecular size have distinct binding sites (39). In the angiotensin AT 1 and the endothelin ET A receptors, minimal overlapping binding sites between peptide agonists and antagonists have been demonstrated, respectively (40,41). Very recently, it has been observed that structurally related peptide agonist, partial agonist, and antagonist occupy a similar binding pocket within the rat cholecystokinin CCK-A receptor (42). Similarly, we describe in the present study major transmembrane overlapping binding sites in the V 1a vasopressin receptor for both peptide agonist and two linear peptide antagonists.  (27).
b Values taken from ref. 12. All values in this table are expressed as the means Ϯ S.E. calculated from three independent experiments, each done in triplicate. The numbering of the mutant receptors is taken from the G protein-coupled receptor alignment and does not correspond to the numbering of the amino acids in the primary sequence. The first digit corresponds to the helical TMR, and the next two digits indicate the rank of the residue in the considered helix (see Tables II and III). Y225 is numbered according to the same rule but is no longer in the second TMR. It is located in the first extracellular loop between TMR II and TMR III (residue 115 in the primary sequence).