Introduction

Top Abstract Introduction Experimental Procedures Results Discussion References

The neurohypophysial hormone arginine vasopressin (AVP) is a cyclic nonapeptide (Fig. 1) whose actions are mediated by stimulation of specific G protein-coupled receptors (GPCRs) currently classified into V₁ vascular (V₁R), V₂ renal (V₂R), and V₃ pituitary (V₃R) AVP receptors (Thibonnier et al., 1998b). AVP is involved in numerous physiological functions, including the regulation of body fluid osmolality, blood volume, vascular tone, and blood pressure (Thibonnier, 1993). In addition, AVP belongs to the family of vasoactive and mitogenic peptides involved in physiological and pathological cell growth and differentiation (Van Biesen et al., 1996).

View larger version (10K): [in this window] [in a new window]   Fig. 1.   Chemical structure of AVP, the peptide V1R antagonist d(CH2)5 [Tyr(Me)2]AVP, and the nonpeptide V1R antagonists OPC-21268 and SR-49059.

The various members of the family of human and animal AVP-oxytocin (OT) receptors have been cloned (Birnbaumer et al., 1992; Kimura et al., 1992; Lolait et al., 1992; Morel et al., 1992; Gorbulev et al., 1993; De Keyser et al., 1994; Mahlmann et al., 1994; Sugimoto et al., 1994; Thibonnier et al., 1994; Hutchins et al., 1995). Stable expression of these cloned receptors in immortalized cell lines now allows the detailed characterization of the ligand-binding pocket and the signal transduction pathways coupled to a given AVP-OT receptor subtype, without the possible interference from other receptor subtypes and endogenously bound hormone. Such characterization will facilitate the rational design of potent and selective therapeutic agents.

The combination of receptor three-dimensional modeling and site-directed mutagenesis experiments has suggested that the AVP agonist binding domain is made of a 15- to 20-Å-deep central cavity defined by the transmembrane helices and surrounded by the extracellular loops of the receptor (Mouillac et al., 1995; Hibert et al., 1999). As shown for other families of GPCRs, residues that are critical for peptide agonist binding are not involved in antagonist binding to the AVP-OT receptors. Furthermore, the determinants of nonpeptide AVP receptor antagonist binding were unknown before this work.

The first nonpeptide AVP V₁R antagonist found by random screening and optimization of chemical entities (Yamamura et al., 1991), OPC-21268 (Fig. 1), has an excellent affinity for the rat V₁R (25 nM) but a poor affinity for the human V₁R (8800 nM) (Thibonnier et al., 1998a). The human and rat V₁Rs share a high degree of structural homology with 96% sequence identity. The differing residues are presumably involved in species-related variations in antagonist binding. Comparison of the human and rat V₁R sequences revealed that only 20 amino acid differences are present in the extracellular loops and the upper portions of the transmembrane segments (TMSs; see Fig. 3). We reasoned that these interspecies differences in amino acid sequence modulate the receptor affinity for nonpeptide compounds. In this work, we produced a series of reverse mutations in which corresponding rat amino acids were introduced by site-directed mutagenesis into the human V₁R sequence. The influence of these interspecies amino acid differences on nonpeptide antagonist binding was subsequently tested. A single amino acid substitution in the seventh TMS produced a 27-fold increase in the affinity toward OPC-21268. To gain information about the location of the OPC-21268 binding site, a model of this compound was docked onto a homology-built three-dimensional model of the human V₁R. The hydrophobic moieties of this nonpeptide antagonist were found to be located deep within the transmembrane region, whereas the polar part is on the extracellular surface. This model of the ligand-receptor complex is consistent with the mutagenesis results and provides an explanation for the increased affinity of the mutants tested in this study.

	Experimental Procedures

Top Abstract Introduction Experimental Procedures Results Discussion References

Materials. Standard reagents, unless stated otherwise, were purchased from Sigma Chemical Co. (St. Louis, MO). CHO-K1 cells were obtained from American Type Culture Collection (Rockville, MD). Cell culture media and geneticin were purchased from Life Technologies (Grand Island, NY). Fetal bovine serum was obtained from Hyclone (Logan, UT). Restriction and modification enzymes were obtained from Promega (Madison, WI). [³H]AVP (specific activity, 68.5 Ci/mmol) was obtained from DuPont-New England Nuclear (Wilmington, DE). The XL2-Blue Escherichia coli strain was purchased from Stratagene (La Jolla, CA). The expression vector pcDNA3.1 was purchased from Invitrogen (San Diego, CA). AVP and the peptide V₁R antagonist d(CH₂)₅Tyr(Me)AVP were purchased from Bachem California (Torrance, CA). The nonpeptide V₁R antagonist SR-49059 (batch no. MY10-075) was provided by Dr. C. Serradeil-Le Gal (Sanofi Recherché, Toulouse, France). The nonpeptide V₁R antagonist OPC-21268 (batch no. 93F92 M) was provided by Dr. J. F. Liard (Otsuka America Pharmaceutical, Inc., Rockville, MD).

Site-Directed Mutagenesis of Human V₁R. The human V₁R cDNA clone was isolated by screening a human liver cDNA library as described previously (Thibonnier et al., 1994) and inserted into the pcDNA3.1 expression vector. The mutations were introduced in the human V₁R wild-type sequence by using the Quickchange mutagenesis kit from Stratagene according to the manufacturer's recommendations. The presence of the mutations was verified by double-stranded DNA sequencing with the Taq Dye Deoxy Terminator cycle sequencing kit and a model 373A sequencer from Applied Biosystems (Foster City, CA).

Radioligand Binding Assays in Intact Cells. Transfected CHO cells were grown to confluence in 24-well dishes and washed twice with PBS plus 10 mM MgCl₂ and 0.2% BSA, pH 7.4. Saturation binding experiments of AVP receptors of transfected CHO cells were performed in 24-well dishes in duplicate with increasing concentrations of [³H]AVP with or without 1 µM unlabeled AVP (Thibonnier et al., 1994). Affinity (K_d) and capacity (B_max) values of the AVP receptors were calculated by a nonlinear least-squares analysis program. Protein concentration was measured with BCA reagent (Pierce Chemical Co., Rockford, IL) using albumin as an internal standard. Competition binding experiments were performed as described previously (Thibonnier et al., 1994, 1998b) using one fixed concentration of [³H]AVP and increasing concentrations of unlabeled peptide and nonpeptide AVP analogs (n = 3-8 for each analog) for 30 min at 30°C. IC₅₀ values were derived from nonlinear least-squares analysis, and K_i values were calculated with the equation of Cheng and Prusoff: K_i = IC₅₀/(1 + L_f/K_d).

Three-Dimensional Molecular Modeling of V₁R. Very little direct structural information is available for GPCRs, and for many years, molecular models of these receptors have been built based on the crystal structure of bacteriorhodopsin. Although bacteriorhodopsin consists of the seven transmembrane helical domains by which GPCRs are characterized, it shares very little sequence homology with any of the GPCRs. Still, the use of bacteriorhodopsin to establish the orientation of the transmembrane domains of the V₁R is the only way to build a model based on an experimentally determined high-resolution structure (Henderson et al., 1990). Coordinates of bovine rhodopsin are also available but only for the seventh TMS without any loops. As a basis for our model building of the V₁R receptor, we used a model of the seventh TMS of V₁R generated by G. Vriend with the program WHATIF (Vriend, 1990) based on the crystal structure of bacteriorhodopsin (G Protein-Coupled Receptor Data Base at http://swift.embl-heidelberg.de/7tm/htmls/consortium. html; Rodriguez et al., 1998).

Molecular Modeling of AVP and Antagonist Ligands. Using program Look v3.5, a model of 8-AVP was built based on the structure of OT obtained from the crystallographically determined structure of the neurophysin-OT complex (Fig. 1; Rose et al., 1996). This model of AVP assumed a type I -turn structure, containing a hydrogen bond between the carbonyl oxygen of Tyr² and the amide proton of Asn⁵ (Fig. 2A).

View larger version (23K): [in this window] [in a new window]   Fig. 2.   Model of 8-AVP (A) and the nonpeptide V1R antagonists OPC-21268 (B) and SR49059 (C) in ball-and-stick representation. Note the disulfide bridge between cysteines 1 and 6 of AVP.

Docking of AVP and Antagonist Ligands onto V₁R. Docking of the ligands was done with the program LIGIN (Sobolev et al., 1996), based on a built-in complementarity function. This function is a sum of the surface areas of atomic contacts. These contacts are weighted according to the types of atoms in contact, and another term is included to prevent short contacts. After maximizing the complementarity function, LIGIN optimizes the lengths of possible hydrogen bonds. To take into account possible movements of the receptor on ligand binding, steric overlap between the ligand and a specified number of residues in the receptor can be allowed without energy penalty.

Molecular Modeling of Site-Directed Mutagenesis. After docking the model of OPC-21268 onto wild-type V₁R, the receptor-ligand complex was subjected to an energy refinement using program X-PLOR. Interactions between this antagonist and four residues on the receptor (Ile²²⁴, Ile³¹⁰, Glu³²⁴, and Gly³³⁷) were analyzed. Based on the mutagenesis results, mutations of these four residues to Val²²⁴, Val³¹⁰, Asp³²⁴, and Ala³³⁷ were modeled in the program O (Jones et al., 1991). Interactions between the antagonist and the mutated residues, as well as any other residues in close contact to the ligand, were analyzed.

Data Analysis. Nucleotide and amino acid sequences were analyzed with the computer package MacVector (Oxford Molecular, Oxford, UK) on a Macintosh G3 computer. Binding parameters (K_d and B_max) of AVP receptors were calculated by a nonlinear least-squares analysis program using the software package Kaleidagraph (Synergy Software, Reading PA; Thibonnier and Roberts, 1985). Data were expressed as mean ± S.E. Statistical analysis was performed with Kruskal-Wallis and Mann-Whitney nonparametric tests (StatView statistical package; Abacus Concepts, Berkeley, CA). P < .05 was considered statistically significant.

	Results

Top Abstract Introduction Experimental Procedures Results Discussion References

Radioligand Binding Characteristics of Wild-Type and Mutated Human V₁Rs. The amino acid differences between human and rat V₁Rs are presented in Fig. 3. Mutations that altered the charge or shape were produced by introducing corresponding rat amino acid residues into the wild-type human V₁R sequence. Because mutations located into the first two extracellular loops do not affect the affinity of antagonists (Mouillac et al., 1995), we centered our attention on interspecies amino acid differences present in the other components of the ligand binding pocket. An extensive ligand binding characterization of these mutated human AVP receptors was carried out in stably transfected Chinese hamster ovary (CHO) cells. As shown previously, CHO cells do not express endogenous AVP-OT receptors (Thibonnier et al., 1994), and each clone tested in our experiments expressed a single AVP receptor clone.

View larger version (47K): [in this window] [in a new window]   Fig. 3.   Amino acid differences between the human and the rat V1Rs. The primary sequence and the putative transmembrane segments of the human V1R are shown. The putative ligand binding pocket is boxed by the rectangle. Arrows pointing toward the corresponding residues of the rat sequence mark nonconserved residues between the human and rat V1R sequences.

Docking of Hormone AVP onto Human V₁R. AVP has a polar as well as a nonpolar surface. The exocyclic tripeptide Pro⁷-Arg⁸-Gly⁹ and one side of the hormone ring (Gln⁴, Asn⁵) are mainly hydrophilic, whereas the other part of the ring (Cys¹, Cys⁶, Tyr², and Phe³) is essentially hydrophobic. This dual surface property is reflected in the nature of the binding pocket that is formed by residues from TMSs 1, 3, 4, 5, 6, and 7, as well as the first extracellular loop (Fig. 4). The bottom of the cleft is mainly hydrophobic, closed by the aromatic and hydrophobic residues Met¹³⁵, Phe¹³⁶, Phe¹⁷⁹, Phe³⁰⁷, and Ile³³⁰. The entrance to the binding pocket and one side of it contain predominantly hydrophilic residues. The Arg⁸ guanido group at the entrance to the cleft forms a salt bridge with Asp¹¹² located on the first extracellular loop. Trp¹¹¹ forms van der Waals contacts with the hydrophobic part of Arg⁸. The -amino group of Lys¹²⁸ forms a hydrogen bond to the amide side chain nitrogen of Asn⁵. Other hydrogen bonds are formed between the side chain moieties of Gln¹⁸⁵ and Ser¹⁸² with Gln⁴ and of Ser²¹³ O with Tyr² OH. Another wall of the pocket is lined with the hydrophobic residues Ile⁵⁵ and Ile³³⁰.

View larger version (35K): [in this window] [in a new window]   Fig. 4.   Docking of the hormone 8-AVP onto the model of the human V1R. AVP and interacting receptor residues are shown in ball-and-stick representation. Residue numbers for AVP are in superscript. A salt bridge between Arg8 of AVP and Asp112 of the receptor is shown by a broken line. Two hydrogen bonds, one between Asn5 of AVP and Lys128 and the other between Gln4 of AVP and Gln185 of the receptor, are shown by broken lines. Distances (in angstroms) are indicated near the broken lines. The secondary structure assignment of the interacting receptor residues is shown by small tube representations of helical segments H1, H4, H6, and H7, as well as extracellular loop 1 (el1). Met135, Phe136, Ser182, and Ser213 also interact with AVP. They are not shown here for reasons of clarity. The former two residues are within van der Waals contact of Phe3, whereas the latter two form hydrogen bonds with the amide side chain nitrogen atom of Gln4 and the hydroxyl group of Tyr2, respectively.

Docking of Nonpeptide Antagonist OPC-21268 onto Human V₁R. The location of the bound antagonist OPC-21268 is distinct from the AVP-binding pocket with only partial overlap near the extracellular surface (Fig. 5). The hydrophobic part is embedded in the transmembrane region far deeper than AVP, whereas the polar part is located on the surface of the extracellular side. The binding pocket is formed by residues from TMSs 4, 5, 6, and 7, as well as the third extracellular loop (Fig. 6). The 27-fold increase in the affinity of the G337A mutant is explained by the formation of two van der Waals contacts of the methyl carbon with carbon atoms C22 and C28 of the bicyclic ring structure of OPC-21268 at the bottom of the cleft (Fig. 7). The E324D mutant has an indirect effect. It enables the formation of a hydrogen bond of the carboxylate side chain with the amide side chain atom of Gln³¹¹. This causes a polarization of this amide nitrogen atom and enables it in turn to form another hydrogen bond to the N57 nitrogen atom of OPC-21268 (Fig. 7). The I310V mutant reduces the hydrophobicity in the vicinity of the polar oxygen atom of the antagonist. The I224V mutant relieves overcrowding in a hydrophobic binding site involving the aromatic residues Trp¹⁷⁵, Phe¹⁷⁹, Phe³⁰⁷, and Trp³⁰⁴. The smaller valine side chain allows for better positioning of the aromatic residues to interact with the bicyclic ring structure of OPC-21268 (Fig. 8). Finally, the I310V mutant reduces the hydrophobicity in the vicinity of the polar oxygen atom of the antagonist. Thus, the model explains all of the mutations that significantly increase the affinity toward OPC-21268.

View larger version (42K): [in this window] [in a new window]   Fig. 5.   Superposition of the models of AVP and the nonpeptide antagonist OPC-21268 as bound to the human V1R. AVP and OPC-21268 in ball-and-stick representation (A) and with the receptor shown in ribbons (B). The loops are labeled il1, il2, and il3 for the intracellular loops and el1, el2, and el3 for the extracellular loops. The different binding modes of agonist and antagonist are clearly shown.

View larger version (43K): [in this window] [in a new window]   Fig. 6.   Docking of the nonpeptide antagonist OPC-21268 onto the model of the human V1-vascular AVP receptor. A, side view. B, top view. The receptor and the antagonist are shown in ribbon and ball-and-stick representation, respectively.

View larger version (26K): [in this window] [in a new window]   Fig. 7.   The stabilizing effect of the G337A and E324D mutations on antagonist binding. Portions of helices 6 and 7 as well as of the extracellular loop 3 (el3) are shown in tube representation; the antagonist as well as residues 311, 324, and 337 of the receptor are shown in ball-and-stick representation. Distances are indicated in angstroms. The 27-fold increase in affinity of this glycine-to-alanine mutant can be explained by the formation of two new van der Waals contacts of the methyl group with the antagonist. The glutamic acid-to-aspratic acid muation at position 324 has an indirect effect. It enables the formation of a hydrogen-bond between an oxygen atom of the carboxylate and the amide side chain nitrogen atom of Gln311. This has a polarizing effect on this nitrogen atom that in turn stabilizes another hydrogen bond of this atom with atom N57 of OPC-21268.

View larger version (43K): [in this window] [in a new window]   Fig. 8.   The stabilizing effect of the G337A, I224V, and I310V mutations on antagonist binding. Portions of helices 4, 5, 6, and 7 are shown in tube representation; the antagonist as well as selected residues of the receptor are shown in ball-and-stick representation. Distances are indicated in angstroms. The isoleucine-to-valine substitution at position 310 puts a less hydrophobic residue in the vicinity of the polar O47 oxygen atom of the OPC-21268 antagonist compound. This should have a slightly stabilizing effect, and it could explain the 2.3-fold increase in affinity. Residue 224 is located deep in the transmembrane region in a very crowded environment, which constitutes a hydrophobic binding pocket for the antagonist. The pocket consists of residues Trp175, Phe179, Phe307, and Trp304 in addition to residue 224. Substitution of the smaller valine for the larger isoleucine relieves overcrowding and allows for better positioning of the aromatic residues to interact with the bicyclic ring structure of OPC-21268. This could explain the 7.1-fold increase in affinity for this mutant.

	Discussion

Top Abstract Introduction Experimental Procedures Results Discussion References

AVP receptors represent a logical target for drug development in several therapeutic fields. As a new class of therapeutic agents, orally active AVP analogs could be used in several pathophysiological conditions. V₂R agonists increase the reabsorption of free water in central diabetes insipidus. V₁R antagonists could reduce the systemic vascular resistances noted in arterial hypertension, congestive heart failure, and peripheral arteriopathy. V₂R antagonists could reverse the hyponatremia of Schwartz-Bartter syndromes, congestive heart failure, and liver cirrhosis. Mixed V₁/V₂R antagonists may prevent thromboembolic events in surgical patients. V₃R agonists and antagonists could be valuable additions to the diagnosis, imaging, localization, and medical treatment of adrenocorticotropic hormone-secreting tumors. Finally, OT receptor antagonists could be used in the treatment of primary dysmenorrhea and premature labor (Thibonnier, 1998).

Three different strategies can be contemplated to develop ligands with high affinity and selectivity for a given AVP receptor subtype: 1) the systematic or rationale alterations of the ligand structure, implemented by Maurice Manning and collaborators who designed numerous peptide AVP and OT analogs (Manning et al., 1995); 2) the random screening for new chemical compounds, developed by pharmaceutical companies who isolated the first nonpeptide V₁R and V₂R antagonists (Yamamura et al., 1991, 1992; Serradeil-Le Gal et al., 1993, 1996); and 3) structure-based drug design, requiring the knowledge of the three-dimensional structure of both the ligand and receptor. The AVP-OT receptors crystallographic structure has yet to be established. However, modeling by analogy based on the structure of bacteriorhodopsin has been done for the seven TMSs of many GPCRs and has yielded useful information (Ji et al., 1998).

These three strategies are complementary. For instance, conformational energy calculations carried out on three nonpeptide AVP-OT antagonists (OPC-21268, OPC-31260, and penicilide) found that the affinity of these compounds and their selectivity for AVP and OT receptors are probably connected with mimicking the aromatic rings of the Tyr² and the Ile³ OT residues or with mimicking the aromatic rings of the Tyr² and Phe³ AVP residues (Oldziej et al., 1995). Similarly, this study illustrates that strategies 2 and 3 are indeed complementary.

By random screening and subsequent optimization of chemical entities, nonpeptide compounds were recently shown to antagonize AVP receptors (Yamamura et al., 1991, 1992; Serradeil-Le Gal et al., 1993, 1996). They specifically antagonize the V₁R or the V₂R and have different chemical structures. The first AVP receptor antagonist OPC-21268 was found to be a potent entity in rat models but was subsequently found to display a poor affinity for human AVP receptors (Thibonnier et al., 1998b). To expand our understanding of the molecular characteristics of the ligand-binding pocket of AVP receptors, we used the amino acid differences among mammalian species to search for the rat versus human molecular determinants of nonpeptide V₁R binding.

Our data confirm that the molecular determinants of agonist and antagonist binding as well as peptide versus nonpeptide compounds are distinct (Mouillac et al., 1995). Amino acid residues that are important for peptide agonist binding are not critical determinants in binding of the cyclic peptide d(CH₂)₅Tyr(Me)AVP, of the linear peptide antagonist phenylacetyl¹-D-Tyr(Me)²-Phe³-Gln⁴-Asn⁵-Arg⁶-Pro⁷-Arg⁸-NH₂, and of the nonpeptide V₁R antagonist SR49059 (Mouillac et al., 1995). Similarly, the molecular determinants of peptide antagonist binding to the OT receptor are different from those involved in peptide agonist binding; they are TMSs 1, 2, and especially 7. The introduction of just seven amino acids of the upper part of the seventh TMS of the OT receptor into the V₂R sequence is sufficient to introduce high-affinity binding for an OT peptide antagonist into the V₂R.

All point mutations affecting peptide agonist binding to AVP receptors were found to have no or little effect on peptide antagonist binding, thus suggesting that peptide agonist and antagonist binding requirements are physically distinct (Phalipou et al., 1997). So far, there is little information about the molecular determinants of nonpeptide antagonists binding to AVP-OT receptors besides the fact that they are different from those involved in peptide agonist and antagonist binding. Cotte et al. (1998) found that residues 202 in the second extracellular loop and 304 in the seventh TMS of the V₂R which modulated species selectivity of cyclic peptide antagonists containing a D-isoleucyl at position 2, did not contribute to binding of nonpeptide antagonists OPC-31260 and SR-121463A.

The combination of site-directed mutagenesis and three-dimensional modeling in our study identified key residues involved in binding of the nonpeptide antagonist OPC-21268 to the V₁R. Our data clearly identified a single residue in the seventh TMS, explaining the different affinities of the human and rat V₁Rs for OPC-21268. The docking model developed for this study confirmed the importance of this single residue: Ala³³⁷. Furthermore, the model predicts that a serine residue at this position should cause an even tighter binding due to the formation of a hydrogen bong between the serine O atom with the quinoline oxygen atom of OPC-21268, in addition to the van der Waals interaction of the serine -carbon with carbon atoms 22 and 28 of this antagonist. This study also suggests modifications to the antagonist to increase the affinity for the receptor. For example, elimination of the quinoline oxygen atom should stabilize the interactions with the hydrophobic pocket deep inside the transmembrane region. However, this may cause adverse solubility problems. A similar situation exists for residue 310 of the receptor and oxygen 47 of the antagonist. A hydrophobic residue in the vicinity of this polar atom is clearly unfavorable. A valine at this position, as found in the human sequence, is better than an isoleucine, the corresponding rat residue, but a threonine would be even better. Alternatively, replacement of oxygen 47 of the antagonist with a carbon atom should also increase the affinity. With respect to residue 224, a valine at this position seems to be optimal. This residue is located in a rather crowded hydrophobic environment into which a valine seems to fit better than the bulkier isoleucine.

Combination of the three mutations in positions 224, 324, and 337 did not further improve the affinity of the V₁R for OPC-21268 compared with the two double mutations, thus suggesting that alterations of the structure of the nonpeptide antagonist will be required to further increase the affinity of this compound.

The field of GPCRs has a lack of experimentally determined structures. Therefore, molecular modeling is a very useful tool to derive structural information for the V₁R. It provides a framework to design and test new drugs, as well as site-specific mutations, in a rational way. However, one must keep in mind the limitations of molecular modeling. The approach is based on the assumption that the seven TMSs are similar in structure to bacteriorhodopsin. The Achilles' heels of this approach are the loops connecting the helical regions as well as the N- and C-terminal nonhelical segments. The former were built by sequence similarity to known protein segments from a database within the program LOOK, whereas the N- and C-terminal stretches were left out altogether from the model because they are not involved in ligand binding. The validity of the model is supported by the experimentally determined affinities for the drugs. The model explains very well all of our findings. It does not prove that the model is correct, but the model is certainly consistent with the data, and it provides a tool for designing new drugs and mutants.

In conclusion, this study provides for the first time the structural basis of species-selective binding of a nonpeptide antagonist to the V₁R. These findings should generate new ideas for drug development of nonpeptide AVP receptor antagonists and for optimizing drug-receptor interactions.