Identification of Residues Involved in Neurotensin Binding and Modeling of the Agonist Binding Site in Neurotensin Receptor 1*

The neurotensin receptor 1 (NTR1) subtype belongs to the family of G protein-coupled receptors and mediates most of the known effects of the neuropeptide including modulation of central dopaminergic transmission. This suggested that nonpeptide agonist mimetics acting at the NTR1 might be helpful in the treatment of Parkinson's disease and schizophrenia. Here, we attempted to define the molecular interactions between neurotensin-(8–13), the pharmacophore of neurotensin, and the rat NTR1. Mutagenesis of the NTR1 identified residues that interact with neurotensin. Structure-activity studies with neurotensin-(8–13) analogs identified the peptide residues that interact with the mutated amino acids in the receptor. By taking these data into account, computer-assisted modeling techniques were used to build a tridimensional model of the neurotensin-(8–13)-binding site in which the N-terminal tetrapeptide of neurotensin-(8–13) fits in the third extracellular loop and the C-terminal dipeptide binds to residues at the junction between the extracellular and transmembrane domains of the receptor. Interestingly, the agonist binding site lies on top of the previously described NTR1-binding site for the nonpeptide neurotensin antagonist SR 48692. Our data provide a basis for understanding at the molecular level the agonist and antagonist binding modes and may help design nonpeptide agonist mimetics of the NTR1.

Most neuropeptides and peptide hormones exert their effects through binding to receptors that belong to the family of G protein-coupled receptors (GPCR) 1 with seven transmembrane (TMs) helices. In general, several GPCR subtypes have been identified for a given neuropeptide. Over the past decade, a number of nonpeptide antagonist ligands of neuropeptide GPCRs have been discovered, most often through random screening of large numbers of compounds (1,2). As a rule, nonpeptide antagonists show receptor subtype selectivity and cross the blood-brain barrier, making them of great value to explore the physiopathological roles of their cognate receptor (2). A number of recent studies have been devoted to the mapping and tridimensional representation, through mutagenesis and computer-assisted molecular modeling, of binding sites for peptide and nonpeptide ligands of GPCRs (3)(4)(5)(6). Such approaches have been useful for understanding the molecular basis of subtype or species selectivity of GPCRs for agonist and antagonist ligands. They have shown that most of the time peptide agonist and nonpeptide antagonist binding sites for a given receptor are topologically distinct (7)(8)(9)(10) and sometimes have provided indications as to the molecular mechanisms by which an agonist may activate its receptor (11)(12)(13)(14)(15). Finally, they may assist in the rational design of selective nonpeptide ligands with agonist or antagonist properties.
Neurotensin (NT) is a 13-amino acid peptide that exerts neuromodulatory functions in the central nervous system and endocrine/paracrine actions in the periphery. Three NT receptors, termed NTR1, NTR2, and NTR3 according to the order in which they were cloned, have been identified so far (16 -21). The NTR1 and NTR2 are GPCRs and share 60% homology, whereas the NTR3 belongs to an entirely different family of proteins (21). All three receptors bind NT through its C-terminal hexapeptide sequence -Arg-Arg-Pro-Tyr-Ile-Leu-OH (22). The NTR1 has high affinity for NT, whereas the NTR2 has lower affinity for the peptide and is selectively recognized by the anti-histamine H1 receptor antagonist levocabastine. The nonpeptide NT antagonist SR 48692 preferentially binds to the NTR1. Many of the known central and peripheral effects of NT are blocked by SR 48692 and can therefore be attributed to the NTR1 (22)(23)(24). Recently, we provided evidence that the NTR2 mediates the SR 48692-insensitive central antinocisponsive effect of NT (25). The functions associated with NT binding to the NTR3 have yet to be elucidated.
In a recent study, using mutagenesis approaches combined with computer-assisted molecular modeling, we established a tridimensional model of the SR 48692-binding site in the rat NTR1 (26). Mutational analysis identified several residues in the receptor TMs that interact with the nonpeptide antagonist as follows: Met 208 in TM4, Tyr 324 , Arg 327 , and Phe 331 in TM6, and Tyr 351 , Thr 354 , Phe 358 , and Tyr 359 in TM7 (Fig. 1). A model of the rNTR1 was constructed using rhodopsin as a template. SR 48692 was then docked in the receptor model, taking into account the mutagenesis data. The antagonist binding site was found to lie within the first two helical turns of the TMs, facing the extracellular side of the membrane (26). These studies were facilitated by the fact that SR 48692 has a rather rigid structure in solution that has been elucidated by x-ray crystallography (27). For many purposes, it would be interesting to identify the agonist binding site in the NTR1 at the molecular level. This is rendered somewhat more difficult by the fact that the structure of NT- (8 -13) is highly flexible and can adopt many conformations in solution (28,29).
In a mutational study of the NT-NTR1 interaction in which all the charged residues in the extracellular domains and the TMs of the rNTR1 were substituted for glycyl residues, it was proposed that Asp 139 in E1 might interact with the positive charges on the side chains of Arg 8 and Arg 9 in the NT sequence and that Arg 143 in the upper part of TM3 might make an ionic link with the C-terminal carboxylate of NT (30). However, in the course of studying the SR 48692/rNTR1-binding site, we found that mutating Arg 143 in Gln or Met did not alter the affinity of NT for the NTR1, precluding an ionic interaction of this residue with the C terminus of NT. We further observed that some of the residues that participate in the antagonistreceptor interaction (Met 208 , Arg 327 , and Phe 331 ) were also involved in NT binding. In addition, Tyr 347 in the third extracellular loop (E3) connecting TM6 and TM7 appeared to be essential for NT binding (26). Others proposed that the binding site of NT in the rNTR1 entirely lies in E3, based on computerassisted modeling of both the receptor and the ligand (31). A number of residues in E3, mainly aromatic (Phe 331 , Trp 339 , Phe 344 , Phe 346 , and Tyr 349 ), were depicted as interacting with the C-terminal hexapeptide sequence of NT. However, there were no mutagenesis data to support these findings. Actually, our previous work showed that mutating Phe 346 or Tyr 349 did not affect the affinity of NT for the rNTR1 and, conversely, that mutation of Tyr 347 greatly reduced the peptide affinity (26).
In the present study, we combined mutational analysis of the rNTR1 and structure-activity studies with NT- (8 -13) analogs in order to identify which residues in the NT pharmacophore might interact with the receptor residues that were found by mutagenesis to be important for NT binding. Thus, the pharmacological properties of NT on rNTR1 bearing mutations on residues Met 208 (TM4), Arg 327 (TM6), Phe 331 (TM6), and Tyr 347 (E3) as well as other residues in E3 (Trp 339 , Thr 341 , and Phe 344 ) were analyzed (Fig. 1). The binding or biological potency of NT-(8 -13) analogs substituted on each of the 6 residues in the hexapeptide sequence or amidated at the C terminus was determined on the mutant receptors that showed decreased affinity for NT. These data were then used to dock NT- (8 -13) in a model of the rNTR1 constructed as described previously (26) with the additional representation of extracellular loops. Our model predicts that the C-terminal dipeptide of NT- (8 -13) interacts with residues in TMs 4 and 6 that lie at or near the junction with the extracellular domain of the rNTR1 and that the rest of the molecule interacts with extracellular residues in E3. Interestingly, the NT-binding site rests on top of the SR 48692 binding pocket that penetrates deeper in the TM core, and both ligands share common points of anchorage at the junction between the TMs and the extracellular domain of the receptor.
Site-directed Mutagenesis-The rNTR1 HindIII-NotI fragment, 1.45kilobase pair corresponding to the total reading frame plus the 5Јnoncoding end, was used as a template for oligonucleotide site-directed mutagenesis as described previously (26). Correct sequences of the mutant receptors cDNA were verified by ABI Prism Dye Terminator Cycle Sequencing Ready Reaction Kit following the manufacturer's protocol. The HindIII-NotI fragments were then subcloned into pcDNA3 eucaryotic vector (Invitrogen). Restriction and modification enzymes were from Promega.
Cell Culture and Transfection-COS M6 cells were grown in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) containing 8% fetal bovine serum (Dutcher) and 50 g/ml gentamicin (Sigma). For transient transfection, 100-mm cell culture dishes seeded with 10 6 cells the day before were washed twice with Tris-buffered saline (Tris 200 mM, NaCl 137 mM, CaCl 2 2.3 mM, MgCl 2 0.5 mM, Na 2 HPO 4 0.4 mM, pH 7.4) and incubated for 30 min with 1 g of recombinant pcDNA3 plasmid in the presence of DEAE-dextran (0.5 mg/ml) at room temperature. After 3 h in culture medium supplemented with 100 M chloroquine, cells were washed twice with Tris-buffered saline and cultured for 48 -72 h.
Cell Membrane Preparation-Transfected cells were washed twice with phosphate-buffered saline and collected in ice-cold Tris/HCl 5 mM, pH 8. After homogenization by repeated passages through a syringe needle and centrifugation at 4°C for 30 min at 100,000 ϫ g, cell membranes were resuspended in 300 l per dish of Tris/HCl 5 mM, pH 7.5, and stored at Ϫ20°C. Membrane protein concentration was determined by the Bio-Rad Protein Assay.
Inositol Phosphate Determination-24 h after transfection with the wild type rNTR1 or mutant receptor, cells were trypsinized and grown for 18 h in 12-well plates in culture medium in the presence of 0. Binding Experiments-Binding experiments for the two radioligands were carried out with 1-100 g of cell membrane proteins in a final volume of 250 l of 50 mM Tris/HCl, pH 7.5, containing 0.1% bovine serum albumin and 0.8 mM 1,10-phenanthroline, for 20 min at room temperature. The reaction was stopped by addition of 2 ml of ice-cold buffer and filtration on cellulose acetate filter (0.2 m, Sartorius) followed by 2 washes of the tube and filter with 2 ml of the same buffer. Nonspecific binding was determined in the presence of 1 M unlabeled ligand. For saturation experiments, concentrations of radioligand ranging from 0.01 to 2 nM for 125 I-NT or from 0.1 to 10 nM for [ 3 H]SR 48692 were tested. For competitive inhibition experiments, increasing concentrations of unlabeled ligands were incubated with 0.1 nM 125 I-NT or 2 nM [ 3 H]SR 48692. Saturation and competition data were analyzed by the LIGAND program (33). For structure-activity relationships study, IC 50 values of neurotensin analogs were derived from the inhibition curves.
Computer-assisted Modeling-We have previously described a tridimensional model of the seven TMs of rNTR1 (26). Here, without modifying TMs orientation, we added to this model the amino acid sequence of the external loops, and a new model was built by means of the Sybyl program using structural homology data bases, molecular dynamics, and energy minimization. The NT-(8 -13) sequence was manually docked in the model. Molecular dynamics and global energy minimization were then effected on the agonist-receptor complex by the Powel method using Kollman all-atom force field while freezing the TM helical backbone and assigning relative interaction forces (20 -200 kcal) between rNTR1 and NT-(8 -13) residues according to the mutagenesis and structure-activity relationship data.  (Table I). All receptors were well expressed in COS M6 cells with SR 48692 B max values ranging from 3 to 20 pmol/mg protein. None of the mutations affected antagonist affinity, consistent with previous data showing that the SR 48692 binding pocket lies below E3 in the TMs of the rNTR1 (26). In contrast, the W339A and F344A mutations resulted in a 10-and 5-fold decrease, respectively, in agonist affinity as compared with the wild type rNTR1.

W339A and F344A Mutations in E3-Trp
Comparison of the B max values obtained with either radiolabeled ligand shows for the wild type rNTR1 a 2-3-fold higher value for SR 48692 as compared with NT. We have previously shown that in membranes from cells transfected with the NTR1, three populations of binding site with high (K d ϳ0.1-0.2 nM), low (K d ϳ50 -100 nM), and very low (K d Ͼ10 M) affinities for NT are present (34). [ 3 H]SR 48692 binds to all three sites with the same affinity, whereas 125 I-NT, over the concentration range used, measurably binds only to the high affinity site (34). In contrast to the wild type receptor, both the W339A and F344A mutants showed much higher (Ͼ10-fold) SR 48692 than NT B max values (Table I). Competition experiments were performed for the three receptors using [ 3 H]SR 48692 as the labeled ligand and NT as the competitor ( Fig. 2A). A similar portion of binding (20 -25%) could not be competed for by NT, indicating that the proportion of very low affinity NT-binding site was the same for the wild type and mutant receptors. IC 50 values for NT on the three receptors are given in Table II. The values for the mutant receptors were 50 to 150 times lower than that for the wild type receptor. These data can be interpreted as indicating that both the W339A and F344A mutations increased the proportion of low affinity NT-binding site in addition to decreasing agonist affinity for the high affinity binding site.
This suggests that mutations in the E3 loop modified the conformational equilibrium of the NTR1. Therefore, it was of interest to compare the ability of NT to stimulate IP production in COS cells transfected with the wild type and mutant receptors (Fig. 2B). Wild type NTR1-transfected cells responded to NT with an EC 50 value of 0.77 Ϯ 0.18 nM (mean Ϯ S.E. from three independent experiments) that was closed to its high affinity K d value (Table I). The W339A and the F344A mutants receptors also responded to NT ( Fig. 2B) with EC 50 values that were 28.0 Ϯ 5.0 and 47.0 Ϯ 4.0 nM, respectively (means Ϯ S.E. from three independent experiments). These values were 1-2 orders of magnitude higher than corresponding high affinity K d values (Table I). Thus, in addition to affecting the conformational equilibrium of the NTR1, the W339A and F344A mutations appear to modify the coupling of the receptor high affinity state to G protein(s).
Y347A, Y347M, and Y347F Mutations in E3-Our previous studies have shown that mutating Tyr 347 to alanine in the rNTR1 resulted in a loss of detectable 125 I-NT binding without modifying the affinity for SR 48692 (26). The data in Table I show in addition that the B max values for [ 3 H]SR 48692 were comparable for the wild type rNTR1 and Y347A mutant, indicating that the latter was well expressed in COS cells. In order to determine the decrease in NT affinity for the Y347A mutant, competition experiments were performed with [ 3 H]SR 48692 as the labeled ligand, and IC 50 values for NT were derived from the competition curves (Table III). The data show that the loss of NT binding potency amounted to almost 4 orders of magnitude for the Y347A mutant as compared with the wild type receptor. Such a loss indicates a major role of the Tyr 347 side chain in NT binding. In order to determine the respective contribution of the phenyl and hydroxyl moieties of Tyr 347 to NT binding, the Y347M and Y347F mutants were constructed and expressed in COS cells, and competition experiments were carried out. Loss of the phenyl ring in the Y327M mutant resulted in a 1000-fold decrease in NT potency, whereas removal of the hydroxyl group in the Y347F mutant brought about a 20-fold decrease in agonist potency (Table III), thus showing the greater contribution of the phenyl than of the hydroxyl moiety to NT binding.
M208A Mutation in TM4 and F331A Mutation in TM6 -These residues located near the junction between TMs and extracellular domains of the rNTR1 were shown previously to play a role in SR 48692 binding (26). Their mutation to alanine also resulted in a 10-fold decrease in 125 I-NT binding affinity (26). The data are shown in Table I  Neurotensin Structure-Activity Relationships with the Wild Type rNTR1-Previous studies have shown that the pharmacophore of NT resides entirely in its C-terminal hexapeptide sequence. Therefore, it can be hypothesized that residues in the NT- (8 -13) sequence are likely to interact with residues in the rNTR1 that have been found by mutagenesis to play a role in NT binding. In order to evaluate the contribution to NT binding of the side chains of each residue in the NT-(8 -13) sequence, a series of NT-(8 -13) analogs was synthesized in which Arg 8 and Arg 9 were replaced by citrulline (citrulline is isosteric but lacks the side chain positive charge of arginine), and Pro 10 , Tyr 11 , Ile 12 , and Leu 13 were substituted by an alanyl residue. NT, NT-(8 -13), and NT-(8 -13) analogs were tested in competition binding experiments using either 125 I-NT or [ 3 H]SR 48692 as the labeled ligand. IC 50 values for the unlabeled peptides were derived from these experiments (Table II). The order of potency for the peptides was comparable whatever the radioligand used. NT-(8 -13) was 5-10-fold more potent than NT. The NT-(8 -13) analogs showed decreases in binding potency that ranged from 1 to 4 orders of magnitude when compared with NT- (8 -13). The data emphasize the major contribution of Tyr 11 to NT binding, the important role of the side chain methyl groups of Ile 12 and Leu 13 , and the somewhat lesser participation of Pro 10 and of the positive charges of Arg 8 and Arg 9 (Table II).
Neurotensin Structure-Activity Relationships with the W339A, F344A, and Y347A Mutant Receptors-In an attempt to determine which residue in the NT-(8 -13) sequence might interact with aromatic residues in E3, competition binding experiments on the wild type and W339A, F344A, and Y347F mutant receptors were performed with [ 3 H]SR 48692 as the labeled ligand and the six NT-(8 -13) analogs described above. The Y347F mutant had to be used for these experiments because the Y347A and Y347M mutants do not retain sufficient NT affinity. IC 50 values for NT, NT-(8 -13), and NT-(8 -13) analogs were derived from the competition experiments (Table  II). The ratio of the IC 50 value for a mutant receptor over that for the wild type receptor was then calculated for NT- (8 -13) and each of its analogs. We reasoned that if mutating an amino acid in the receptor affects the interaction of a residue in the NT-(8 -13) sequence with the peptide-binding site, then modifying the side chain of that NT-(8 -13) residue should not greatly affect the analog binding potency and the IC 50 value ratio should tend to one. Plots of the ratio values thus obtained for each mutant receptor versus the position of the substituted residue in NT- (8 -13) are represented in Fig. 3. The data show that with all three mutant receptors the potency of [Ala 11 ]NT-(8 -13) was significantly less affected than that of NT-(8 -13) by the mutations. We interpret these results as indicating that Tyr 11 in the NT-(8 -13) sequence is likely to interact through contacts with Trp 339 , Phe 344 , and Tyr 347 in the E3 loop of the rNTR1. The data further suggest that the hydroxyl group of Tyr 347 might form a hydrogen bond with the hydroxyl group of Tyr 11 in the NT sequence. It was also observed that the potency of [Ala 10 ]NT-(8 -13) was less decreased than that of NT-(8 -13) by the W339A receptor mutation, suggesting that Trp 339 is either interacting with Pro 10 or necessary for the correct positioning of this residue in the NT-binding site.
Neurotensin Structure-Activity Relationships with the M208A and F331A Mutant Receptors-Similar studies as above were conducted with the M208A and F331A mutant receptors, except that in this case competition experiments were performed with 125 I-NT as the labeled ligand because, as already mentioned, [ 3 H]SR 48692 does not retain sufficient affinity for these receptors to permit binding studies. IC 50 values are given in Table II, and plots of the IC 50 ratio values are shown in Fig. 3. Interestingly, the binding potency of [Ala 12 ]NT-(8 -13) was identical for the wild type NTR1 and the M208A mutant receptor. This result suggests that Met 208 may form hydrophobic interactions with the side chain methyl  (26). However, this mutant receptor retained the ability to respond to NT by an increase in inositol phosphate (IP) production with the same basal and NT maximal responses than the wild type rNTR1 but with a potency for NT that was decreased by 4 orders of magnitude (26). We have shown that Arg 327 is involved in an ionic interaction with the carboxylic acid group of SR 48692 (26). It is known from previous structure-activity studies that the C-terminal COOH function of NT is essential for NT binding. In order to see if the large decrease in NT potency toward the R327M mutant receptor could be due to the disruption of an ionic interaction between Arg 327 and the Cterminal acidic function of NT, we determined the potencies of NT, NT-(8 -13), and NT-(8 -13)-NH 2 for their ability to stimulate IP production in the wild type rNTR1 and the R327M mutant. NT-(8 -13)-NH 2 being amidated at the C terminus would be expected to have the same potency for the wild type and mutant receptor, should our hypothesis of an ionic link be correct. The data presented in Fig. 4 and Table IV show that NT-(8 -13) was three times more potent than NT for stimulating IP production in COS cells transfected with the wild type rNTR1. Amidating the C terminus of NT-(8 -13) resulted in an almost 1000-fold loss in potency for NT-(8 -13)-NH 2 . The potency of both NT and NT- (8 -13) was decreased by 4 orders of magnitude in cells transfected with the R327M mutant as compared with the wild type receptor. In sharp contrast, NT-(8 -13)-NH 2 retained the same potency in both transfected cell systems. Actually, the amidated analog was 10 times more potent than its parent peptide NT- (8 -13) in stimulating IP production in R327M receptor-transfected cells (Fig. 4 and Table IV). These data provide strong evidence that Arg 327 forms an ionic bond with the C-terminal carboxylate of NT.
D139A Mutation in the E1 Loop-Previous studies have shown that mutating Asp 139 to glycine in the rNTR1 resulted in a loss of NT binding, and it was proposed that Asp 139 might form ionic interactions with the guanidinium group of Arg 8 or Arg 9 in the NT sequence (30). Here, we mutated Asp 139 to alanine, expressed the mutant receptor in COS M6 cells, and found that the D139A receptor was devoid of measurable 125 I-NT and [ 3 H]SR 48692 binding. NT-stimulated IP production was then measured in D139A receptor-expressing cells. A weak stimulation (74 Ϯ 29%-fold over basal, n ϭ 6) could be   (8 -13) and its analogs were derived from competition binding experiments (Table II). The ratio of the IC 50 value for a mutant receptor over that for the wild type receptor was calculated for NT- (8 -13) and each of its analogs. Represented are the plot of the ratio values thus obtained for each mutant receptor versus the position of the substituted residue in NT- (8 -13). The values are the means Ϯ S.E. from three independent experiments. **, p Ͻ 0.01 for comparison between IC 50 ratio value for a NT-(8 -13) analog and NT- (8 -13).
observed only at very high NT concentrations (100 M). The very low potency of NT in this system precluded further structure-activity studies with NT analogs for testing the hypothesis that Asp 139 might interact with Arg 8 or Arg 9 .
Model of the NT- (8 -13)⅐rNTR1 Complex-We have previously described a model of the rNTR1⅐SR 48692 complex (26). In this model, only the seven TMs were taken into account, and their positions relative to one another as well as their orientations in the membrane were determined using rhodopsin as a template (35). In order to construct a model of the rNTR1⅐NT-(8 -13)-binding site, the TMs orientation was kept identical to that of our previous model, and the E3 loop sequence was entered into the modeling program. The above mutagenesis and structure-activity data were taken into account for manu-ally docking the NT-(8 -13) sequence in its binding site thought to lie between the E3 loop and the extracellular side of TM4, -6, and -7. Energy minimization was then performed while constraining the TMs and assigning distances between rNTR1 and NT-(8 -13) residues according to the interactions described in the preceding sections. This led to the model of the rNTR1⅐NT- central position in the E3 loop, close to Tyr 347 and to a lesser extent to Trp 339 and Phe 344 . The side chain of Arg 9 is oriented so that the guanidinium function lies near Phe 331 . Finally, the side chain of Arg 8 stretches outside the E3 loop. DISCUSSION In the present study, we used a combination of mutagenesis, pharmacological, and molecular modeling approaches in order to provide a tridimensional representation of the NT-binding site in the rNTR1. Our strategy was similar to that previously employed for modeling the SR 48692⅐rNTR1 complex (26). It entailed determining the receptor residues that are important for ligand binding by mutagenesis, modeling the rNTR1, and docking the ligand in the receptor taking into account the mutagenesis data. The latter two steps are easier to perform for SR 48692 than for NT for two reasons as follows: (i) unlike NT or NT-(8 -13) which are highly flexible in solution (28,29), SR 48692 has a rigid structure that has been determined by x-ray crystallography (27); (ii) the SR 48692 binding site lies entirely within the TMs (26), whereas that of NT-(8 -13) is partly extracellular, which makes it more difficult to model the NTbinding site as rhodopsin provides a good template for orienting the TMs of GPCRs but cannot be used for modeling extracellular domains. For this reason, it was necessary to establish which residues in the NT-(8 -13) sequence are interacting with the receptor residues that are important for agonist binding as determined by mutagenesis. This was achieved by performing systematic structure-activity studies with NT- (8 -13) analogs and the mutant receptors that showed decreased NT potency.
This approach led us to identify two residues in the receptor, Arg 327 in TM6 and Tyr 347 in the E3 loop, that play critical roles in binding NT- (8 -13) and to determine the structural elements in the NT-(8 -13) sequence with which these residues interact. Thus, we propose that Arg 327 makes an ionic link with the C-terminal COOH group of NT- (8 -13) and that Tyr 347 lies close to Tyr 11 with which it formsinteractions and hydrogen bonding. The similar large loss (3-4 orders of magnitude) in agonist affinity observed following reciprocal modification of the interacting partners (Arg 327 to Met and Leu 13 -COOH to Leu 13 -NH 2 ; Tyr 347 to Ala and Tyr 11 to Ala) is consistent with our proposal. Furthermore, the dramatic decrease of potency (20,000-fold) of the [Ala 11 ]NT-(8 -13) analog could be accounted for by additional weakerinteractions between Tyr 11 in NT and Trp 339 and Phe 344 in the E3 loop of the receptor. We also propose that the side chains of Ile 12 and Leu 13 in the NT- (8 -13) sequence form hydrophobic interactions with Met 208 and Phe 331 , respectively. In these cases, the decrease in agonist potency (30-fold) that results from the M208A and F331A mutations is 10 -20 times smaller than that (500-fold) brought about by the corresponding I12A and L13A modifications in the NT sequence. This could mean that the side chain methyl groups of Ile 12 and Leu 13 are involved in other interactions with receptor residues not identified here or, alternatively, that they are important for correctly positioning the C-terminal -Ile-Leu-COOH sequence of NT in the receptor binding site. Finally, we suggest that the guanidinium group of Arg 9 in NT forms cation-interactions with Phe 331 , the 70-fold loss of potency of [Cit 9 ]NT- (8 -13) being consistent with this proposal.
Although [Cit 8 ]NT-(8 -13) exhibited a 10 -30-fold decrease in affinity for the rNTR1, we could not assign in our model of the NT-(8 -13)⅐rNTR1 complex a residue in the receptor that would interact with the side chain of Arg 8 . Previous studies have shown that mutating Asp 139 to glycine led to a loss of NT binding, and it was proposed that this residue might form ionic interactions with Arg 8 (30). Here, we show that the D139A mutant receptor is devoid of measurable NT and SR 48692 binding. It can be stimulated to produce IPs at very high NT concentrations (100 M), which represents at least a 5 order of magnitude loss of potency as compared with the wild type rNTR1. However, it seems unlikely that Asp 139 could form an ionic link with Arg 8 because [Cit 8 ]NT- (8 -13) would be expected to exhibit a greater loss of potency than that observed here. Rather, we think it more likely that the D139A mutation produces a major change in receptor conformation that affects both agonist and antagonist binding. In particular, Asp 139 lies close to Cys 142 which, by analogy with other GPCRs (36,37), is thought to make a disulfide bridge with Cys 225 in the E2 loop. Mutating Asp 139 might possibly affect the formation of the disulfide bond that has been shown for other GPCRs (38,39) to be essential for maintaining receptor conformation.
By using quite a different approach than ours for modeling the NT-(8 -13)⅐rNTR1 complex, Pang et al. (31) proposed that the NT-(8 -13)-binding site lies entirely in the E3 loop. A number of residues in E3 were predicted to interact with NT- (8 -13). However, these predictions were not directly tested by mutagenesis of the rNTR1 (31). Our previous data (26) and those presented here do confirm some of the interactions proposed by Pang et al. (31). However, they do not support others and predict points of interaction for NT- (8 -13) outside the E3 loop that were not described in the model of Pang et al. (31). Thus, we agree with Pang et al. (31) on the proposedinteractions between Phe 344 and Trp 339 in the E3 loop of the rNTR1 and Tyr 11 in NT- (8 -13). However, we provide strong evidence that the major anchoring point for Tyr 11 in the E3 loop is Tyr 347 throughinteractions and hydrogen bonding. Phe 331 was reported to make cation-interactions with Arg 9 and hydrophobic interactions with Ile 12 (31). Our data support the former proposal but not the latter. Rather, they are consistent with Phe 331 interacting with the side chain of Leu 13 and Met 208 in TM4 with that of Ile 12 . The model of Pang et al. (31) suggested that Phe 346 and Tyr 349 interact with Arg 8 and Arg 9 . This is unlikely in view of our previous findings showing that both the F346A and Y349A mutant receptors retained the same affinity for NT as the wild type rNTR1 (26). Finally, we provide evidence for a strong ionic link between Arg 327 in TM6 and the C-terminal carboxylic group of NT- (8 -13). This crucial interaction was not accounted for in the model of Pang et al. (31). Thus, although Pang et al. (31) correctly predicted by means of conformational studies that the E3 loop of the rNTR1 is an important part of the NT-binding site, the present combined mutational, structure-activity, and modeling approach has allowed us to define the residues, both in the NT and E3 loop sequences, that are involved in agonist binding and to describe other points of NT-receptor interaction outside the E3 loop.
Comparison between the NT-(8 -13)-binding site in the rNTR1 as described here and that of SR 48692 as previously reported (26) provides interesting information (Fig. 5C). Both ligands share common points of anchorage in the receptor that are Met 208 at the junction between TM4 and the E2 loop, Phe 331 at the junction between TM6 and the E3 loop, and Arg 327 located in TM6 one helical turn from the E3 loop. Met 208 and Phe 331 form hydrophobic interactions with the adamantane cage of the antagonist or with the side chain methyl groups of Ile 12 and Leu 13 in the agonist. Arg 327 forms an ionic link with the carboxylic function attached to the carbon atom that also bears the adamantane moiety in SR 48692 or with the C-terminal COOH group of NT. Thus, the (adamantane)-CH(COOH)-structure of SR 48692 occupies the same position in the receptor as the C-terminal dipeptide, -Ile-Leu-COOH, of NT. The N-terminal 8 -11 sequence of NT-(8 -13) fits in the E3 loop that connects TMs 6 and 7 with Tyr 11 occupying a central position and interacting closely with Tyr 347 in the receptor, whereas the rest of the SR 48692 molecule points toward the intracellular side of the membrane, making interactions with residues that are located within the first two helical turns of TMs 6 and 7 (Fig. 5C). Therefore, the agonist and antagonist binding sites are distinct but partially overlap at the junction between TMs and extracellular domains, sharing strong anchoring points in the receptor such as Arg 327 . This would clearly account for the competitive antagonist behavior observed for SR 48692.
The rNTR1 has previously been shown to exist in a high and a low affinity state for NT, both states being recognized with the same affinity by SR 48692 (34). The present findings that the W339A and F344A mutations markedly increase the proportion of the low affinity state suggest that the E3 loop may have flexibility and adopt conformations that confer either high or low affinity to the agonist binding site. The antagonist binding site lying below the E3 loop within the TMs would not be affected by the conformational change of the loop, and this would explain the observation that SR 48692 binds with the same affinity to the high and low affinity NT receptor states. The location of the NT-binding site is particularly interesting in view of the fact that agonist-induced G protein coupling and internalization have been shown to involve the third intracellular (I3) loop (40) and the C-terminal tail (41) of the rNTR1, respectively. The I3 loop and C-terminal domain both connect to the E3 loop through TMs 6 and 7, respectively. It may be suggested that upon binding of NT to the rNTR1, conformational changes of the E3 loop may transconform the I3 and C-terminal domains through connecting TMs 6 and 7, thereby promoting G protein coupling and internalization. SR 48692 that binds below the E3 loop to residues in TMs 6 and 7 would prevent transconformation of the intracellular domains and lock the receptor in an inactive state. If these hypotheses are correct, it might be predicted that mutations in the E3 loop or connecting TM domains could affect the transduction properties of the rNTR1. As shown here this is so for the W339A and F344A mutations, and we have obtained preliminary evidence that this is also the case for other mutations in the E3 loop and TM7. 2 A number of studies have attempted to delineate small neuropeptide-binding sites for GPCRs through receptor mutagenesis or chimeric construction approaches (3,12,(42)(43)(44)(45). However, few studies have provided molecular models of peptide agonist-receptor complex by combining mutagenesis, structure-activity studies, and computer-assisted modeling. To our knowledge, this was done for the thyrotropin-releasing hormone, the somatostatin SSTR2, and the vasopressin V1a receptors (5,6,46). The binding pocket for these three peptides lie mainly within the core of the TMs, and it can be noted that in each case a number of residues in TMs 6 and 7 are involved in peptide-receptor interactions (5,6,46). Our own model of the NT⅐rNTR1 complex shows the peptide-binding site to lie between the surface of the membrane and the E3 loop. It differs in this respect with the TRH, somatostatin, and vasopressin receptor models, suggesting a different binding mode for these peptides on the one hand and NT on the other hand. Although the ␦ agonist-␦ opioid receptor complex has not been modeled, studies of chimeric ␦and mutant ␦ opioid receptors indicated that the binding site for ␦ agonists might be comprised of residues in the E3 loop and at the top of TMs 6 and 7 (47). This suggests that NT and ␦ opioid agonists may share common mechanisms of receptor binding and activation. Interestingly, sequence homology analysis indicates that, in the GPCR fam-ily, the opioid receptors are among the most closely related to the NTR1. Mutational analysis of neurokinin receptors have shown that residues that are important for peptide agonist binding are scattered throughout extracellular domains (42). Therefore, it appears that neuropeptide-binding sites in GPCRs may be located entirely inside or outside or partly inside and outside the TMs. This variability is in contrast with GPCRs of small neurotransmitter ligands like the biogenic amines for which the binding site has always been found to reside in a structurally conserved region inside the TMs (3,4). It also contrasts with the observation that the binding pocket of nonpeptide antagonists of neuropeptide GPCRs always lie within the core of the seven TMs (10, 48 -50).
The amino acid sequences of the E3 loop and TMs 6 and 7 are highly conserved in the rNTR1 and hNTR1. In particular, all the residues shown here to play a role in NT binding are conserved with the exception of Phe 344 in the rNTR1 which is replaced by a Tyr in the hNTR1. It is therefore likely that the NT-binding site in the hNTR1 will be quite similar to that described here for the rNTR1. This conclusion was also reached by Pang et al. (31). Our proposed model of the NT-binding site might be of help for the design of nonpeptide agonist mimetics of the NTR1. Such compounds might be useful for the treatment of brain disorders such as Parkinson's disease or schizophrenia (51). A nonpeptide agonist of the NTR1 should include among other features a COOH group linked to an aliphatic structure in order to mimic the -Ile-Leu-COOH sequence of NT, a phenol ring for fulfilling the role of Tyr 11 in NT, and positively charged groups for mimicking the side chains of Arg 9 and Arg 8 . The choice of the relative spatial disposition of these chemical moieties could then be guided by our tridimensional model of the NT-binding site.