INTRODUCTION

Top Abstract Introduction Methods Results & Discussion Conclusions References

Three different types of opioid receptors (, µ, and ), identified based on their pharmacological properties, have recently been cloned (see reviews: Reisine, 1995; Dhawan et al., 1996; Zaki et al., 1996) and assigned to the large superfamily of rhodopsin-like G protein-coupled receptors (GPCRs). This superfamily of GPCRs consists of integral membrane proteins that transduce optical and chemical signals across the cellular membrane (Watson and Arkinstall, 1994) and share a common 3D structure. The seven-helical structure of the transmembrane domain has recently been demonstrated by electron cryomicroscopy (EM) studies of bovine, frog, and squid rhodopsins with a resolution of 6-9 Å (Schertler et al., 1993; Unger and Schertler, 1995; Unger et al., 1997; Davies et al., 1996). Many members of the GPCR family, especially rhodopsin, have been extensively studied by site-directed mutagenesis and a variety of physicochemical methods. These experimental data and the analysis of variability and hydrophobicity patterns in amino acid sequences of GPCRs have made it possible to assign the transmembrane helices of GPCRs to the peaks in the rhodopsin EM maps (Baldwin, 1993), and to construct a number of different approximate GPCR models (see reviews: Ballesteros and Weinstein, 1995; Donnelly et al., 1994). Some of these models have been built from the structure of the nonhomologous 7--bundle membrane protein, bacteriorhodopsin (Henderson et al., 1990), whereas others have used the low-resolution rhodopsin EM maps and a few experimentally derived constraints to pack together seven "ideal" helices with arbitrarily chosen side-chain conformers (for example, Baldwin, 1997; Donnelly et al., 1994; Herzyk and Hubbard, 1995). The calculation of a more precise, atomic-level structure requires refinement of the spatial positions of entire helices, determination of their precise geometry, as helices are never "ideal" in proteins (Barlow and Thornton, 1988), and careful attention to side-chain packing.

To refine the structure of the transmembrane domain, we have developed and recently described a novel modeling approach that is based on the presence of numerous polar residues in the hydrophobic, lipid-embedded -helices of GPCRs (Pogozheva et al., 1997). It is known that water-inaccessible polar groups of proteins have a strong tendency to form H-bonds (McDonald and Thornton, 1994). In transmembrane -helices, peptide backbone groups are already paired, whereas the polar side chains must interact with each other to form intra- or interhelical H-bonds. The candidate H-bonding pairs can be identified from the analysis of sequence alignments as polar residues in intramembrane segments that appear and disappear simultaneously in various GPCRs. The corresponding H-bonds can then be used as constraints for packing the seven -helical fragments by distance geometry calculations. Moreover, the side-chain H-bonds from many different GPCRs can be combined to increase the number of simultaneously applied constraints and to calculate an "average" 7--bundle structure. The computational procedure was organized as an iterative refinement with evolving constraints that begins from an initial model of the -bundle and continues until each buried polar side chain of each of the 410 GPCRs considered can participate in at least one hydrogen bond in the final structure (the root mean square deviation, r.m.s.d., between the initial and final structures was ~4 Å) (Pogozheva et al., 1997). This "saturation of hydrogen bonding potential" (McDonald and Thornton, 1994) criterion was very sensitive to structural mistakes during the refinement procedure. The transmembrane segments of individual GPCRs are hydrophobic and contain less than 30% polar residues, but when 410 different amino acid sequences are simultaneously considered, all interhelical contacts within the -bundle are "labeled" by polar side chains forming intramolecular H-bonds. Displacement of any -helix from its correct position breaks some H-bonds, producing unpaired polar side chains within the lipid bilayer in tens or hundreds of GPCRs.

The "average" atomic structure of the -bundle has been tested by using it as a template to calculate the transmembrane domains of specific GPCRs whose H-bonds and close packing of nonpolar side chains must be compatible with the same common structure. The models of 28 different GPCRs (including vertebrate and invertebrate rhodopsins and a number of opioid, chemokine, glycoprotein, cationic amine, melatonin, and purine receptors) were generated by distance geometry, using H-bonds specific to each receptor, while using the "average" model to restrain the spatial positions of the helices. Analysis of the GPCR models reveals many features that are responsible for structural stability of the transmembrane -bundle, such as the formation of extensive networks of interhelical H-bonds, aromatic and sulfur-aromatic clusters that are spatially organized as "polarity gradients," close packing of side chains throughout the transmembrane domain, and the formation of interhelical disulfide bonds in many GPCRs (Lomize et al., 1998). Some other features of the models are related to biological function and evolution of GPCRs, such as the formation of a spatially continuous "minicore" of 43 evolutionarily conserved residues, a multitude of correlated replacements of residues buried within the core, a Na⁺ binding site, and complementarity of receptor binding pockets to many structurally dissimilar, conformationally constrained ligands (Lomize et al., 1998).

As has previously been discussed (Pogozheva et al., 1997; Lomize et al., 1998), the GPCR models obtained are consistent with a large body of experimental data that were not used in deriving the models and that therefore can serve as an independent control. The model of rhodopsin, for example (1boj and 1bok Protein Data Bank files), is in agreement with the arrangement of -helices in the low-resolution 3D EM maps; mapping of water- and lipid-accessible rhodopsin residues by chemical probes; identification of residues surrounding retinal by site-directed mutagenesis and cross-linking; the orientations of all-trans and 11-cis retinal relative to the membrane plane and the distances from the ligand to the intra- and extracellular surfaces, determined by linear dichroism and fluorescence quenching; reconstitution studies of opsin with synthetic retinal analogs; the conformation and environment of the protonated retinal Schiff base, studied by Raman, Fourier transform infrared, and ¹³C solid-state NMR spectroscopies; cross-linking studies; the compensatory replacements of Glu¹¹³ (III:3) by Asp⁹⁰ (II:21) or Asp¹¹⁷ (II:7); and many other data (Pogozheva et al., 1997). (Superscript residue numbers correspond to the particular receptor sequences. Numbers in parentheses indicate the helix number (Roman numerals) and the residue position in 26-residue transmembrane segments, identified by Baldwin (1993) (Arabic numerals) and shown in Fig. 1.) The "average" model of the -bundle is also in agreement with constraints experimentally derived by site-directed mutagenesis for other GPCRs, such as the proximity of Asp³⁹⁷ (II:28) and Lys⁵⁸³ (VII:3) in the lutropin/choriogonadotropin hormone receptor (Fernandez and Puett, 1996), Asn⁸⁷ (II:14) and Asn³¹⁸ (VII:17) in the gonadotropin-releasing hormone receptor (Zhou et al., 1994), Asp¹²⁰ (II:14) and Asn³⁹⁶ (VII:17) in the 5-HT_2A receptor (Sealfon et al., 1995), Asp¹²⁵(III:7) and Lys³³¹(VII:4) in _1B-adrenergic receptors (Porter et al., 1996), and the formation of an artificial Zn²⁺-binding site by histidine residues incorporated in positions V:1, V:3, and VI:27 in mutant NK-1 and  opioid receptors (Elling et al., 1995; Thirstrup et al., 1996). The models of cationic amine receptors (Lomize et al., 1998) are consistent with accessibilities of residues from helices III, V, and VII to water-soluble probes (Javitch et al., 1995; Fu et al., 1996) and with a vast sample of site-directed mutagenesis data demonstrating, for example, the interaction of AspIII:7 with the protonated amine of ligands (Fraser et al., 1989; Javitch et al., 1995; Ho et al., 1992; Mansour et al., 1992, 1997; Porter et al., 1996; Savarese and Fraser, 1992; Strader et al., 1987, 1988; Wang et al., 1991, 1993), the involvement of SerV:6 of -adrenoreceptors and SerV:7 of -adrenoreceptors in H-bond formation with catechol ligands, the importance of SerV:10 for ligand binding and activation (Strader et al., 1989; Wang et al., 1991; Hwa et al., 1997), and the proximity of the indole rings of Trp¹⁰⁹ (III:3) and Trp³³⁰ (VII:8) of the ₂-adrenoreceptor to the azido group of iodoazidopindolol, an affinity label for -adrenergic receptors (Wong et al., 1988).

View larger version (39K): [in this window] [in a new window]   FIGURE 1   Sequence alignment of transmembrane helices (TMH I-TMH VII) and extracellular loops (EL-1, EL-2, EL-3) of human , µ, and  receptors. Asterisks above the sequences for each helix indicate the 26-residue transmembrane segments, identified by Baldwin (1993) and used for identification of GPCR residues as the number of helix (Roman numerals):number of residue in the 26-residue fragment (Arabic numerals). For example, Asp128 in the -receptor sequence is denoted as III:7. Numbering of the µ receptor is that of the rat receptor for consistency with mutagenesis data.

In the present paper, we discuss in detail the 3D structures of , µ, and  opioid receptors calculated from the previously developed "average" model of the transmembrane domain. This is an especially interesting case for verification of the receptor models by ligand docking, because the three different opioid receptor types have a number of structurally distinct, conformationally constrained ligands, from small, rigid alkaloids to larger cyclic peptides, with well-studied structure-activity relationships (SARs). In addition, we have included in the models the tentative structures of the three extracellular loops, which were calculated by distance geometry. Although the ligand-binding pocket consists mainly of residues from the transmembrane -bundle, the extracellular loops of opioid receptors have also been shown to be important for interactions with many ligands (Chen et al., 1995; Fukuda et al., 1995; Hjorth et al., 1995; Meng et al., 1995, 1996; Minami et al., 1996; Onogi et al., 1995; Pepin et al., 1997; Varga et al., 1996; Valiquette et al., 1996; Wang et al., 1994, 1995; Xue et al., 1994, 1995; Zhu et al., 1996a,b), whereas the extracellular N-terminus can be deleted in µ and  receptors (Kong et al., 1994; Surratt et al., 1994) or exchanged between receptor subtypes (Meng et al., 1996) without affecting the ligand binding.

	METHODS

Top Abstract Introduction Methods Results & Discussion Conclusions References

The modeling described here was done in three stages: 1) distance geometry calculations of transmembrane domains of , µ, and  opioid receptors from the previously determined "average" transmembrane -bundle structure; 2) modeling of the extracellular loops of the opioid receptors; and 3) incorporation of various opioid ligands into the calculated receptor structures.

Distance geometry calculations of transmembrane -bundles for , µ, and  receptors

The transmembrane 7--bundles of , µ, and  opioid receptors were calculated using their own specific H-bonds, while using the "average" GPCR model to restrain the spatial positions of the helices, as previously described for bovine rhodopsin (Pogozheva et al., 1997). The positions of the helices were restrained by incorporating C... C distances from the "average" model as the upper limits in calculations with the distance geometry program DIANA (Güntert et al., 1991). These C... C limits were increased by 1 Å (0.5 Å for distances of the more loosely packed helix I) to allow some relaxation of the specific receptor structures relative to the "average" model, i.e., small shifts of helices that are necessary to adopt the replacements of side chains in the "core" of the -bundle.

To examine possible H-bonds and to determine conformers of side chains in opioid receptors, we applied an iterative distance geometry refinement approach, which we have previously described (Pogozheva et al., 1997). Each iteration of the refinement included 1) examination of the structures calculated in the previous iteration for new potential H-bonding partners (spatially proximate polar groups that did not form H-bonds in the previous iteration of the model), for correlations in sequence alignments and for structural flaws (violations of constraints, appearance of hindrances or holes produced by incorrectly packed side chains, helices that are multiply curved by contradictory constraints or are loosely packed because of insufficient constraints); 2) modification of distance and angle constraints (H-bonds and conformers of side chains) to increase the number of simultaneously formed H-bonds, and to correct discovered flaws; and 3) distance geometry calculations with the modified constraints. The analysis of calculated structures (step 1) was performed using the program ADJUST (Pogozheva et al., 1997) and the molecular modeling software QUANTA (Molecular Simulations). The constraints and the corresponding -bundle structure evolved simultaneously during the refinement. During the refinement, conformers of most side chains were unequivocally determined. Final systems of H-bonds are shown in Table 1.

In calculations with DIANA, the -helix geometry was restrained by backbone H-bonds (upper limits for NH_i... . O==C_i4 distances = 1.9 Å, except those broken by Pro residues) and by dihedral angle constraints ( = 70° to 50°,  = 50° to 30°). Because the program requires a single chain, the loops connecting -helices were approximated by Gly_n fragments, with the number of Gly residues corresponding to the length of each loop in the -opioid receptor. In the later iterations of the calculations, glycine residues in the extracellular loops were replaced with the amino acids corresponding to the opioid receptor sequences (see below). The standard target function minimization strategy (Güntert et al., 1991) was used for calculations. The weighting factors for upper and lower distance limits and van der Waals and angle constraints initially were 1, 1, 0.6, and 20, respectively, and 1, 1, 2.0, and 5 by the final two iterations. The HisVI:20 and HisVII:4 side chains were considered to be uncharged, and all other His, Asp (including AspII:14), Glu, Lys, and Arg side chains were considered charged.

Modeling the extracellular loops

The extracellular domain of the opioid receptors consists of three loops (EL-1, EL-2, and EL-3), whose tentative structures are modeled here, and an N-terminus that was not considered (Fig. 1). It is apparent from the sequence homology of the loops among the , µ, and  receptors, that essentially the same structure can be expected in the different receptor subtypes. EL-1 and EL-3 are rather short (four or five residues) (Fig. 1), whereas EL-2 is longer (20 residues in  and µ receptors and 23 residues in the  receptor) and can interact directly with all opioid ligands, because it partially covers the binding cavity between helices III and VII in the model of the transmembrane -bundle. Initially, only this longer EL-2 was added to the transmembrane -bundle for distance geometry calculations. EL-2 connects transmembrane helices (TMHs) IV and V and is attached to TMH III by a conserved disulfide bond (Watson and Arkinstall, 1994), giving this loop a U-like shape (the peptide chain comes from TMH IV toward TMH III and returns back to TMH V, as shown in Figs. 4 and 5). Both branches of the U-like EL-2 are too short to form any additional -helices in the calculated models of the transmembrane domain, and the geometrical constraints imposed by their attachment to TMH III, IV, and V force them to adopt extended structures. This extended character of the peptide chain is also consistent with the general (i, i + 2) pattern of alternate polar and nonpolar side chains around the disulfide bond in amino acid sequences of opioid receptors and rhodopsins, for example. The pattern is of the form p-n-p-Cys-p-n-p-Ar, where p, n, and Ar denote polar, nonpolar, and aromatic residues, respectively. We suggest that the two extended antiparallel stretches of EL-2 near the conserved disulfide bond are paired in a -hairpin (residues 195-203 in the , 214-222 in the µ, 207-215 in the  receptor), and the two remaining fragments of EL-2, which connect the -hairpin to helices IV and V, adopt a nonregular structure. These connections contain Pro, Gly, and polar residues and are highly variable in families of opioid receptors and other GPCRs. The characteristic Pro²⁰³-Ser²⁰⁴-Pro²⁰⁵-Ser²⁰⁶ sequence in the  opioid receptor, for example, is an excellent breaker of both -helix and -structure. The nonregular structure of these connections can be also suggested based on insertions in this region arising in many different GPCRs, such as insertions of Ser²²⁰ and Val²⁰⁵-Asp²⁰⁶ residues in the  receptor (Fig. 1).

The hypothesized -hairpin formation is supported by several observations. First, the -hairpin provides the formation of many H-bonds between residues that appear and disappear in a correlated manner in amino acid sequences of opioid receptors, such as Asp²¹⁶... Thr²²⁰, Lys¹⁴¹... Asp²¹⁶, Ser²¹⁴... Gln³¹⁴, and His²²³... Glu³¹⁰ (present only in µ receptors), Glu¹¹⁸... Gln²⁰¹, Glu¹¹⁸... Lys¹²², and Ser²⁰⁴... Arg²⁹¹ (present only in  receptors), and Lys¹³²... Glu²⁰⁹ and Asp²¹⁶... His³⁰⁴ (present only in  receptors). Two insertions in EL-2 of the  receptor are also correlated: they provide simultaneous lengthening of both nonregular connections between the -hairpin and transmembrane helices IV and V, thus allowing the -hairpin to stay in the same spatial position. Second, the -hairpin can readily be inserted in the cavity between helices III and VII, without the appearance of interatomic hindrances, and it forms numerous hydrophobic contacts and several hydrogen bonds with the transmembrane -bundle. Third, the structure of the -hairpin itself is stabilized by hydrophobic contacts of several interacting nonpolar residues (Val¹⁹⁶, Leu²⁰⁰, and Phe²⁰² in the  receptor; Ile²¹⁵, Leu²¹⁹, and Phe²²¹ in the µ receptor; or Ile²⁰⁸, Leu²¹², and Phe²¹⁴ in the  receptor; see Fig. 5). At the same time, several polar residues (Gln²⁰¹ in ; Asp²¹⁶, Thr²¹⁸, Thr²²¹ in µ; Glu²⁰⁹, Thr²¹¹, Gln²¹³ in  receptors) are arranged on the opposite face of the -hairpin and form H-bonds with each other and with polar residues from helix III (Glu/Thr/AspIII:3 and LysIII:1). Fourth, the presence of several Ser and Thr residues with high -sheet propensities in this region (positions 214, 218, 220, and 222 in the µ opioid receptor, for example) is also consistent with the hypothesized formation of the -hairpin.

The probable conformation of the -turn in the -hairpin can also be readily identified. Because the -turn consists of an odd number of residues (residues 198-200 in , 217-219 in µ, 210-212 in  receptor), the only allowed standard type is the type I with a G1 -bulge, i.e., the _RRL motif (Sibanda and Thornton, 1991). This motif is very common in protein -hairpins (Sibanda and Thornton, 1991) and has been shown to be independently stable in aqueous solution (deAlba et al., 1996), because, unlike the "standard" type I and II -turns, the _RRL turn is consistent with the direction of twist in -structure (Richardson and Richardson, 1989). In the structure of the µ opioid receptor, this turn is further stabilized by H-bonds formed by the COO group of Asp²¹⁶ with the main-chain NH group of Thr²¹⁸, and between the side chains of Thr²¹⁸ and Thr²²⁰ (Fig. 5). The consistency of the -hairpin with the entire system of distance constraints for the -bundle was further verified by distance geometry calculations for , µ, and  receptors (the H bonds of the -hairpin are shown in Table 1).

After incorporation of the -hairpin in the model, the extracellular ends of TMHs II, III, VI, and VII were extended by one to three residues (through residues II:28, III:3, VI:3, and VII:2), because these residues can form fragments of amphiphilic helix with nonpolar side chains facing the lipid environment. This also provides, simultaneously, many additional intra- and interhelical H-bonds and hydrophobic contacts between nonpolar side chains. The remaining extracellular loop fragments (113-117 (EL-1), 190-194 and 204-209 (EL-2), and 290-294 (EL-3) in the  receptor; 130-136 (EL-1), 209-213 and 223-228 (EL-2), and 309-312 (EL-3) in the µ receptor; 123-127 (EL-1), 200-206 and 216-222 (EL-2), and 303-306 (EL-3) in the  receptor) were simply considered as short, nonregular connections whose tentative structures were defined by distance geometry calculations based on the appearance of correlated H-bonded residues in the loops (such as Arg²⁹¹-Asp²⁹⁰ and Asp²⁸⁸-Arg²⁹²-Asp²⁹³, which simultaneously appear only in EL-3 of the  receptor), and constraints for dihedral angles  and , which fix them in the allowed areas of the Ramachandran map (the intervals of the angles were constrained similarly to that in the REDAC strategy; Güntert and Wuthrich, 1991). The final structures of the extracellular loops in all opioid receptors provide close packing of Trp and Phe residues conserved in EL-1 (positions 114 and 116 in , 133 and 135 in µ, and 124 and 126 in  receptors) and the orientation of most tryptophan, tyrosine, and phenylalanine side chains in the loops toward the lipid-water interface, where they can interact with lipid headgroups, as is characteristic for membrane proteins (Deisenhofer and Michel, 1991; Schultz, 1992; Grigorieff et al., 1996).

Final calculations of the transmembrane domains, including the three extracellular loops, were made using 64, 69, and 70 side-chain H-bonding constraints for , µ, and  receptors, respectively (Table 1); constraints for dihedral angles of the main chain in the loops and for all side chains of the transmembrane -bundle; C... C distance constraints taken from the "average" model; and restraints on the geometry of the TMHs, as described above. The constraints also included backbone H-bonds in the -hairpin of the EL-2 fragment (residues 195-203 in , 214-222 in µ, 207-215 in  receptor) and a conserved disulfide bond connecting this -hairpin to TMH III. Totals of 877, 896, and 884 angle constraints and 691, 690, and 651 distance constraints were used for calculations of , µ, and  receptors, respectively.

The calculations with DIANA yielded well-defined sets of structures for each (, µ, and ) opioid receptor (pairwise r.m.s.d. of 212 TMH C atoms was <0.7 Å for the 10 structures of each receptor with the lowest target function). The r.m.s.d. between C atoms of TMHs of different (, µ, and ) receptors was larger (~0.9 Å). All backbone angles of the models are within the allowed regions of the Ramachandran map, and all side chains have standard ¹-⁴ conformers, as is automatically provided by the dihedral angle constraints (violations of the individual angle constraints were <10°). A few violations of van der Waals constraints of ~0.5 Å were present near Pro residues in -helices; no violations of H-bond distances greater than 0.6 Å were found. The structures of receptors with the lowest target function were selected for ligand docking and energy minimization.

Ligand docking

All opioid ligands were inserted manually into the binding pockets, using the Molecular Modeling module of QUANTA to move the ligands and control hindrances and receptor-ligand H-bonds. The docking was simplified by using only rigid or conformationally constrained ligands (Fig. 2 and Tables 2 and 3) whose structures have been solved by x-ray crystallography (Bye, 1976; Klein et al., 1987; Urbanczyk-Lipkowska and Etter, 1987; Verlinde et al., 1984; Calderon et al., 1997; Doi et al., 1990; Flippen-Anderson and George, 1994; Griffin et al., 1986; Lomize et al., 1994; Collins et al., 1996) or NMR spectroscopy (Mosberg and Sobczyk-Kojiro, 1991; Collins et al., 1996). The procedure of manual ligand docking is similar to assembling a jigsaw puzzle that consists of two semirigid pieces; however, three circumstances complicated the process. First, because most of the ligands are not completely rigid, it was necessary to consider several possible conformers of their flexible elements, such as the N-cyclopropylmethyl group in morphine, the N-phenethyl group in fentanyl, or Tyr¹ in cyclic opioid peptides. Second, in a few cases, described in the Results, it was necessary to adjust conformers of several receptor side chains in the binding pocket, which were not unequivocally defined by distance geometry calculations. Third, because the DIANA-generated receptor structures were not completely identical (although the r.m.s.d of C atoms was low: ~0.7 Å), the ligand docking was performed with two or three structures with the lowest target function.

View larger version (13K): [in this window] [in a new window]   FIGURE 2   Structures of nonpeptide opioid ligands.

View this table: [in this window] [in a new window]   TABLE 3   Torsion angles of two small cyclic opioid peptides, JOM-13 (Tyr-c[D-Cys-Phe-D-Pen]) and [L-Ala3]DPDPE (Tyr-c[D-Pen-Ala-Phe-D-Pen]), in the models of  opioid receptor-ligand complexes and in published crystal structures of the peptides

The receptor-ligand H-bonds and ion pairs served as important attachment points for ligand docking. Two such attachment points are the carboxyl and imidazole groups of AspIII:7 and HisVI:20, respectively, the only polar groups situated at the bottom of the binding pocket in all three calculated opioid receptor models (, µ, and ). The importance of the AspIII:7 and HisVI:20 residues for binding opioid ligands has been clearly demonstrated by mutagenesis (Befort et al., 1996b; Mansour et al., 1997; Surratt et al., 1994). In the receptor models, these carboxyl and imidazole groups are arranged in such a way that they can interact simultaneously with the N⁺ and OH groups, respectively, of the Tyr¹ or tyramine group present in most opioid ligands. Importantly, all surrounding side chains in the bottom of the binding pockets (IleV:4, IleVI:19, CysVII:6, IleVII:7) can be tightly packed (arranged without hindrances or holes) with the tyramine fragment of the ligands. It should be mentioned that, even without consideration of receptor-ligand H-bonds, the largest ligands, such as norBNI and cyclic peptides, can be inserted in the binding pockets without hindrances in only one way, because they occupy nearly all available space within the pockets, and any shift of the ligands would produce significant overlaps with surrounding receptor atoms. However, for some smaller ligands, the mode (or modes) of docking can be determined only if key attachment points of the interacting molecules (H-bonds or ionic interactions) are assumed.

Technically, docking of most ligands was performed in two steps. First, the tyramine fragment of each ligand was placed in the bottom of the binding pocket to form H-bonds with AspIII:7 and HisVI:20, while the rest of the ligand molecule was oriented toward the extracellular surface. Then the spatial position of the ligand molecule and the conformation of its flexible elements, which are connected to the tyramine fragment, were adjusted to exclude all hindrances with receptor atoms and to form additional H-bonds and hydrophobic contacts in the binding pocket. For example, the cyclic peptides considered (Table 3) have rigid, well-defined structures of their cycles (D-Cys-Phe-D-Pen and D-Pen-Ala-Phe-D-Pen), but have a considerably more flexible exocyclic Tyr¹ residue and side chain of Phe³ (Deschamps et al., 1996; Lomize et al., 1996). Only structures of the cycles that were determined by x-ray crystallography (Lomize et al., 1994; Flippen-Andersen et al., 1994; Collins et al., 1996) were examined here. The structure of the rigid cycle (D-Cys-^EPhe-D-Pen) for the µ-selective agonist JH-42 was considered to be close to the crystal structure of JOM-13 cycle on the basis of theoretical conformational analysis (Mosberg et al., 1996). However, during step 2 of the docking procedure, the  angle of Tyr¹, the  angle of the second residue, and the ¹ angles of the Phe³⁽⁴⁾ and Tyr¹ side chains in these peptides were adjusted to allow the formation of additional H-bonds and to exclude steric hindrances with the receptor. Some details of the adjustment for the peptides and other opioid ligands are described in the Results and Discussion. It should be stressed that the bound conformations obtained for all ligands were identical or very close to the crystal structures, both geometrically (Tables 2 and 3) and energetically (the energy differences between crystal and bound conformations were 0.5-2.0 kcal/mol after energy minimization of the ligands with the CHARMm force field (Brooks et al., 1983; Momany and Rone, 1992), using a dielectric constant  = 10 and the adopted-basis Newton-Raphson method). Conformational analyses of peptides from the JOM-13 and [D-Pen²,D-Pen⁵]enkephalin (DPDPE) series have been discussed elsewhere (Lomize et al., 1994, 1996; Mosberg et al., 1994a,b, 1996). The proposed docking modes were compared with available SAR, cross-linking, and mutagenesis data, as described in detail in the Results and Discussion for 16 opioid ligands.

After manual docking of the ligands with the Molecular Modeling module of QUANTA, the steric overlaps between ligand and receptor atoms did not exceed 0.5 Å. All remaining hindrances were removed during 35 subsequent iterations of unconstrained minimization of the complexes with the CHARMm force field (Brooks et al., 1983; Momany and Rone, 1992), using a dielectric constant  = 3 and the adopted-basis Newton-Raphson method. Initial approximations that yielded energies greater than 2000 kcal/mol after 35 minimization steps were rejected, because this demonstrated residual hindrances or distorted geometry of ligands or receptor. The final energies of accepted receptor-ligand complexes ranged from 2803 to 2176 kcal/mol, and the structures of these complexes were not altered after short-term minimization: the r.m.s.d. between atoms of ligand and of receptor binding site residues from the initial and minimized structures were <0.1 Å.

	RESULTS AND DISCUSSION

Top Abstract Introduction Methods Results & Discussion Conclusions References

Models of , µ, and  opioid receptors

The calculated , µ, and  opioid receptor models are nearly identical within the transmembrane domain (r.m.s.d. of 212 common C atoms of the TMHs are ~0.9 Å); however, small differences are observed in the extracellular loops (Fig. 3) because of unequal numbers of residues among the receptors (Fig. 1). All opioid receptor models have a ligand-binding cavity that is partially covered by the extracellular loops (Fig. 4). The loops create an almost continuous surface, with the -hairpin formed by EL-2 in the middle (Fig. 5), surrounded by the smaller, nonregular EL-1 and EL-3. This region is represented in 3D EM maps of frog and bovine rhodopsins by a considerable amount of electron density that does not contain -helices (Unger et al., 1997).

View larger version (45K): [in this window] [in a new window]   FIGURE 3   Superposition of structures of DIANA-calculated  (bold line), µ (thin line), and  (dashed line) receptors (stereo view). The r.m.s.d. between 212 C atoms of transmembrane helices of  and µ,  and , and µ and  receptors are 0.74, 0.80, and 0.90 Å, respectively.

View larger version (61K): [in this window] [in a new window]   FIGURE 4   Cartoon representation of transmembrane helices and extracellular loops of -opioid receptors with JOM-13, side view and top view from the extracellular surface. Helical fragments are purple, loop fragments are white, the -turn is orange, the disulfide bridge between helix III and EL-2 (residues Cys121-Cys198) is yellow, and JOM-13 is green.

View larger version (23K): [in this window] [in a new window]   FIGURE 5   Proposed structure of the -hairpin in EL-2 of the µ opioid receptor with proximal H-bonded polar residues from helices III and VII and from EL-3 and conserved disulfide bond between Cys140(III:0) and Cys217(EL-2). H-bonds are indicated by the dashed line.

The calculated opioid receptors structures have several clusters of polar side chains that form extensive networks of interhelical hydrogen bonds (Fig. 6 and Table 1). Four such clusters consist of a "core set" of polar residues that are conserved throughout most GPCRs, augmented by more variable, peripheral polar residues that are connected to the central "core" by H-bonds (Pogozheva et al., 1997). The large, polar cluster I consists of conserved AsnI:18, AspII:14, SerIII:14, AsnVII:13, SerVII:14, and AsnVII:17 residues (Asn⁸⁶, Asp¹¹⁴, Ser¹⁵⁴, Asn³²⁸, Ser³²⁹, and Asn³³² in the µ receptor) and is supplemented, in opioid receptors, by the more variable TyrI:7, ThrII:18, AsnIII:10, Ser/CysVII:15 (Tyr⁷⁵, Thr¹¹⁸, Asn¹⁵⁰, and Cys³³⁰ in the µ receptor). Cluster I contains a cavity that can be filled by water or by a sodium ion coordinated with oxygens of the AspII:14, AsnIII:10, SerIII:14, SerVII:14, and AsnVI:17 side chains. Cluster II is formed around the conserved AsnII:9-TrpIV:11 pair by TyrIII:9, ThrIII:13, ThrIII:17, and SerIV:15 (Asn¹⁰⁸-Trp¹⁹² pair and residues Tyr¹⁴⁹, Thr¹⁵³, Thr¹⁵⁷, and Ser¹⁹⁶ in the µ receptor). Cluster III consists of the conserved AsnII:4 and TyrVII:21 (Asn¹⁰⁴ and Tyr³³⁶ in the µ receptor) and the more variable ThrI:28 and AspVII:25 (Thr⁹⁷ and Asp³⁴⁰ in the µ receptor), and cluster IV consists of the conserved triad AspIII:24-ArgIII:25-TyrIII:25 (Asp¹⁶⁴, Arg¹⁶⁵, Tyr¹⁶⁶ in the µ receptor) at the C-terminus of TMH III, TyrV:22, and ArgVI:3 (Tyr²⁵², Arg²⁸⁰ in the µ receptor) from the ends of helices V and VI and the more variable TyrII:6, ThrIII:20, SerIII:22, Lys/AsnIV:2, LysIV:4, AsnIV:7, and Asn IV:10 (Tyr¹⁰⁶, Thr¹⁶⁰, Ser¹⁶², Asn¹⁸³, Lys¹⁸⁵, Asn¹⁸⁸, Asn¹⁹¹ in the µ receptor). Clusters I and II are situated in the middle of the transmembrane domain, and III and IV are close to the intracellular surface. These clusters are present in most GPCRs because they contain many conserved polar residues. The opposite, extracellular surface of the -bundle, which includes the binding pocket, forms several smaller "variable" polar clusters that are specific for different subfamilies of GPCRs. Some of the subfamily-specific clusters are present in all opioid receptors (GlnII:24-TyrII:28-AspIII:7-TyrVII:11-His/TyrVII:4 and TyrIII:8-Asp/GluV:-1-LysV:3 (Gln¹²⁴-Tyr¹²⁸-Asp¹⁴⁷-His³¹⁰-Tyr³¹⁶ and Tyr¹⁴⁸-Glu²²⁹-Lys²³³ in the µ receptor)), whereas others are found only in µ (Thr¹³⁷-Lys¹⁴⁰-Asp²¹⁶-Thr²¹⁸-Thr²²⁰, Ser²¹⁴-Asn²³⁰-Gln³¹⁴-His²²³-Glu³¹⁰, and Tyr²⁹⁹-Lys³⁰³-Ser³¹⁷; Figs. 5 and 6),  (Glu¹¹⁸-Lys¹²²-Gln²⁰¹, Ser²⁰⁴-Arg²⁹¹-Asp²⁹⁰, and Asp²⁸⁸-Arg²⁹²-Asp²⁹³), or  (Asp¹²⁸-Lys¹³²-Glu²⁰⁹-Gln²¹³, Lys²⁰⁰-Arg²⁰²-Asp²⁰⁴-Asp²¹⁷-Glu²¹⁸-Ser²²⁰, Glu²⁹⁷-Thr^302-Ser³¹¹-Tyr³¹², and Asp²⁰⁶-Asp²¹⁶-His³⁰⁴; Table 1) subtypes. Extracellular loops 2 and 3 are connected by a His²²³... Glu³¹⁰ H-bond in the µ receptor (Asp²¹⁶... His³⁰⁴ and Ser²⁰⁴... Arg²⁹¹ in  and  receptors, respectively). This H-bond is probably structurally important, because alkylation of His²²³ by N-ethylmaleimide in the µ receptor reduces the binding affinity of several opioid ligands (Shahrestanifar et al., 1996).

View larger version (63K): [in this window] [in a new window]   FIGURE 6   H-bond network of the µ opioid receptor (stereo view). Colors of residues depicted: green, Tyr, Trp; red, Asp, Glu; blue, His, Lys; yellow, Ser, Thr, Asn, Gln. The receptor is shown with morphine (purple) in the binding site. H-bonds are indicated by the dashed line.

The positions and tilts of the helices of the transmembrane domain of our models differ from all previously published models of opioid receptors (Alkorta and Loew, 1996; Befort et al., 1996b; Cappelli et al., 1996; Knapp et al., 1995; Habibi-Nezhad et al., 1996; Metzger et al., 1996) and other GPCRs that have been deposited in the PDB (Bernstein et al., 1977), GPCRDB (http://swift.embl-heidelberg.de/7tm/) (Oliveira et al., 1993), and CORD (http://www.opioid.umn.edu) databases. The largest deviations of our  opioid receptor structure (r.m.s.d. of C atoms in the range of 4.2-6.5 Å) are observed when compared with the earliest GPCR models, which were constructed from nonhomologous bacteriorhodopsin structures, or by using 2D (projection) EM maps of rhodopsin and a few supplementary experimental constraints. The incorporation of geometric constraints derived from 3D EM maps of rhodopsins (Herzyk and Hubbard, 1995) leads to a model with a smaller (3.9 Å) deviation from our structure. Recently an improved approximation of the transmembrane domain structure has been obtained by the direct fit of two kinked and five straight helices to the 3D EM map of frog rhodopsin (Baldwin et al., 1997). This model has the lowest r.m.s.d (3.3 Å for 179 common C-atoms; Fig. 7) when compared with our structures of the transmembrane domain of the  opioid receptor or bovine rhodopsin. The 3.3-Å r.m.s.d. between this model and our model of  opioid receptor originates from the outward shifts of helices II and V, the shift of the C-terminus of helix III, and from an almost one-turn shift of helices V and VI in the direction perpendicular to the membrane plane in the model of Baldwin et al. (1997). As a result, and as discussed by the authors themselves, the model of Baldwin et al. (1997) contradicts some experimental data, such as the observed formation of a Zn²⁺ binding cluster in positions V:-1 and VI:27 and in positions V:3 and VI:27 (Elling et al., 1995; Thirstrup et al., 1996); formation of H-bonds between residues III:7, V:3, V:6, V:7 and catecholamine ligands (Strader et al., 1987, 1988, 1989; Wess et al., 1991); interaction of Asp II:14 and Asn VII:17 (Zhou et al., 1994; Sealfon et al., 1995); and the contact of Gly III:11 and Phe VI:12 in rhodopsin (Han et al., 1996a,b). All of these experimental data are simultaneously satisfied in our models (the models are compared in more detail by Lomize et al., 1998).

View larger version (50K): [in this window] [in a new window]   FIGURE 7   Comparison of the -opioid receptor model transmembrane -bundle (blue) and the EM-based model of Baldwin et al. (Baldwin, 1997) (red) (stereo view).

Our previously developed model of rhodopsin considered the possible rotations of several functionally important, conserved side chains (GluIII:24, TyrV:22, TrpVI:16, LysVII:11, and TyrVII:21) that can participate in alternative systems of H-bonds, depending on their possible ¹ conformers (¹  60° or 180°) (Pogozheva et al., 1997). Analysis of physicochemical data for rhodopsin indicates that conformational rearrangements of these side chains could take place during photoactivation of rhodopsin. In opioid receptors, only gauche⁺ (¹  60°) rotamers of the corresponding TrpVI:16 and TyrVII:11 and the trans (¹  180°) rotamer of AspIII:24 have H-bond partners and/or lack hindrances with surrounding atoms. Therefore rotations of the side chains of these residues are unlikely. On the other hand, rotations of the TyrV:22 and TyrVII:21 side chains are possible, because there is space in the models for both rotamers. Consequently, distance geometry calculations were performed with two different orientations of the TyrV:22 and TyrVII:21 side chains. The two sets of structures obtained were almost identical (r.m.s.d. ~0.7 Å). Hence the precision of our calculations is insufficient to discriminate the active and inactive conformations of opioid receptors or to reproduce the shifts of transmembrane helices that probably accompany activation of GPCRs (Sakmar and Fahmy, 1995; Farrens et al., 1996; Sheikh et al., 1996; Shieh et al., 1997). Consequently, we incorporated opioid agonists and antagonists into the same receptor structures, calculated with trans rotamers of the TyrV:22 and TyrVII:21 side chains, earlier assigned to the active conformation of rhodopsin.

Ligand binding

The calculated , µ, and  receptor structures have deep binding cavities, situated in the extracellular side of the transmembrane domain between helices III, IV, V, VI, and VII. These cavities are partially covered by the extracellular loops and, especially, by the central -hairpin connecting TMHs IV and V (Fig. 4). The binding pockets consist of an inner interhelical "conserved region" that is identical in , µ, and  opioid receptors (GlnII:24, TyrII:28, CysIII:0, LysIII:1, ValIII:3, AspIII:7, TyrIII:8, MetIII:11, LysV:3, IleV:4, PheV:7, TrpVI:16, IleVI:19, HisVI:20, CysVII:6, IleVII:7, TyrVII:11, and a conserved Cys in EL-2) and a peripheral "variable region" that consists of residues from the ends of TMHs (positions III:-3, III:4, V:-1, V:0, VI:23, VI:26, VI:27, VI:31, VII:-1, VII:0, VII:3, VII:4) and from the extracellular loops (for example, positions 193, 194, 195, 196, 197, 291, 293 in EL-2 and EL-3 of the  receptor). The majority of residues in the binding pocket have fixed side-chain orientations. However, several residues can have different rotamers to accommodate either bound peptides or alkaloids: AspIII:7 has ¹  60° for peptide ligands and ¹  180° for alkaloid ligands; HisVI:20 has ²  120° or 40° when interacting with peptide or alkaloid ligands, respectively; LysV:3 assumes a ³ angle of 180° or 60°, and Asp/Glu V:-1 has a ¹ angle of 60° or 180° in complexes with cyclic peptides or complexes with other opiates, respectively. Furthermore, Tyr³¹² (VII:3) in the  opioid receptor has a ¹ angle of 60° for peptide and nonpeptide ligands, except for norbinaltorphimine (norBNI), which requires ¹  180° to provide additional space for this bulky ligand.

The structures of the binding pockets were tested for complementarity to 16 rigid or conformationally constrained opioid ligands with very different chemical structures and sizes (Fig. 2). Peptides from the DPDPE and JOM-13 series were chosen because they have small rigid cycles and have been extensively studied by x-ray crystallography (Flippen-Andersen et al., 1994; Lomize et al., 1994; Deschamps et al., 1996), NMR spectroscopy (Mosberg et al., 1990; Mosberg and Sobczyk-Kojiro, 1991; Collins et al., 1996), and theoretical methods (Froimowitz, 1990; Wilkes and Schiller, 1991). Larger linear and cyclic opioid peptides are too flexible to be useful for verification of receptor models. It was found that crystal structures of all ligands examined, except DPDPE, fit the pockets, with only a few flexible torsion angles, in some cases, needing to be adjusted (Tables 2 and 3). The largest ligands (such as DPDPE or norBNI) fill almost all of the available space within the binding cavities and interact with residues from both "conserved" and "variable " regions. Smaller alkaloids (such as morphine and naloxone), on the other hand, interact predominantly with "conserved" residues in the bottom of the binding cavity, leaving some free space around the ligand. The results of extensive structure-activity studies of JOM-13 and DPDPE analogs (Heyl and Mosberg, 1992a,b; Mosberg et al., 1994a,b; Haaseth et al., 1994) and mutagenesis data (Befort et al., 1996a,b; Hjorth et al., 1995; Meng et al., 1996; Pepin et al., 1997; Surratt et al., 1994; Valiquette et al., 1996) were used to verify the ligand docking.

The complementarity of rigid opiates and their binding pockets in the receptor models is evident from two different criteria. First there is a good geometrical fit; the ligands can be inserted in the bottom of the binding pocket without significant hindrances or holes in the area of contact. Second, and even more important, there is a spatial complementarity of groups with similar polarities, such that nearly all polar groups of the ligands form H-bonds with corresponding polar side chains within the binding pocket, whereas all ligand nonpolar (aliphatic and aromatic) groups form stabilizing hydrophobic contacts with nonpolar side chains of the receptors.

Cyclic peptides

The cyclic pentapeptide DPDPE (Tyr-c[D-Pen-Gly-Phe-D-Pen]OH) is a standard -selective ligand that is widely used in studies of opioid receptors (Mosberg et al., 1983). Its more constrained [D-Cys², des-Gly³] (JOM-13) and [L-Ala³] analogs have high  affinities and selectivities, whereas [D-Ala³]DPDPE is much less potent (Haaseth et al., 1994). JH-42 is a modified version of JOM-13 with ^EPhe³ and an amidated C-terminus, modifications that result in a shift in binding selectivity from  to µ (Ho, 1997). X-ray crystallography shows that DPDPE and [L-Ala³]-DPDPE have different, unique structures of their tetrapeptide cycles (Flippen-Anderson et al., 1994; Collins et al., 1996; Deschamps et al., 1996), whereas JOM-13 is present in the crystal in two forms (A and B) of its 11-membered ring, which are similar, except for the configuration of the disulfide bridge (S-S torsion angles of 89° and 99°, respectively) (Lomize et al., 1994). In all crystal structures of these peptides, the Phe³⁽⁴⁾ side chains have gauche⁺ (¹  60°) rotamers, but Tyr¹ orientations are varied (Deschamps et al., 1996). The crystal structures of DPDPE and related peptides are consistent with NMR spectroscopy solution data, indicating conformational rigidity of the cycles (Mosberg et al., 1990; Mosberg and Sobczyk-Koiro, 1991; Collins et al., 1996; Lomize et al., 1996). However, theoretical conformational analyses indicate the possibility of several alternative low-energy structures of the disulfide-bridged cycle in DPDPE (Froimowitz, 1990; Wilkes and Schiller, 1991). All exocyclic elements of these peptides are very flexible in aqueous solution, i.e., they have undefined angles  and ¹ of Tyr¹,  of D-Cys/Pen², and ¹ of Phe³⁽⁴⁾ (Mosberg et al., 1990; Mosberg and Sobczyk-Koiro, 1991; Lomize et al., 1994; Collins et al., 1996).

JOM-13

The cyclic tetrapeptide JOM-13 was positioned in the  receptor-binding pocket, using crystal structures A and B of its 11-membered cycle, but with adjusted torsion angles for the exocyclic Tyr¹ residue and ¹ of Phe³. First, Tyr¹ was positioned in the bottom of the pocket to simultaneously form H-bonds between its N⁺ and OH groups with O¹ of Asp¹²⁸ (III:7) and N² of His²⁷⁸ (VI:20), respectively (the corresponding N... O distances are 2.7 and 3.1 Å, respectively, in the final model of the receptor-ligand complex). This can be done only in the trans (¹  180°) orientation of the Tyr¹ side chain. Next, the gauche⁺ orientation of Phe³ (¹ = 60°) was chosen based on SAR for JOM-13 analogs (Mosberg et al., 1994b, 1996). Then the spatial position of the disulfide-bridged 11-membered ring relative to the fixed Tyr¹ was adjusted by rotating torsion angles  of Tyr¹ and  of D-Cys². The ring position was adjusted simply to remove all significant hindrances between the ring and surrounding receptor residues. The  receptor-bound conformation thus determined is very close to crystal structure B of JOM-13, except for the configuration of the disulfide bridge (S-S torsion angle  90°), which corresponds to crystal structure A (Table 3).

The bound conformation of JOM-13 geometrically fits the binding pocket of the  receptor and forms a number of complex-stabilizing H-bonds and hydrophobic contacts with surrounding receptor residues (Fig. 8). The binding pocket can be arbitrary divided into subsites that are complementary to smaller structural elements of JOM-13, i.e., its Tyr¹ residue, D-Cys²-D-Pen⁴ disulfide bridge, Phe³ side chain, and C-terminal COO group. The positively charged nitrogen of Tyr¹ is located in a relatively polar binding subsite formed by several H-bonded residues from helices II, III, and VII (Gln¹⁰⁵(II:24), Asp¹²⁸(III:7), Tyr¹²⁹(III:8), and His³⁰¹(VII:4)). The aromatic ring of Tyr¹ occupies the bottom of the cavity between Tyr¹²⁹(III:8), Met¹³²(III:11), Ile²¹⁵(V:4), Trp²⁷⁴(VI:16), His²⁷⁸(VI:20), Val²⁸¹(VI:23), Leu³⁰⁰(VII:3), Cys³⁰³(VII:6), and Ile³⁰⁴(VII:7). There are a few small empty spaces around Tyr¹ in the pocket, which can accommodate methyl groups in the 2' and 6' positions of Tyr¹ and the extra ring of trans-Hpp¹ (trans-3-(4'-hydroxy)-phenylproline), consistent with the high affinities observed for the corresponding JOM-13 analogs (Mosberg et al., 1994a). On the other hand, the CH-atom of Tyr¹ is in close contact with Tyr¹²⁹(III:8), and an additional C-methyl group incorporated here would experience steric hindrance with the aromatic ring of Tyr¹²⁹(III:8), consistent with the decreased affinities of MeTyr¹, Hai¹ (6-hydroxy-2-aminoindan-2-carboxylic acid), and Hat¹ (6-hydroxy-2-aminotetralin-2-carboxylic acid) analogs of JOM-13 (Mosberg et al., 1994a). The Tyr¹²⁹(III:8) side chain also forms an O1/4HN H-bond with the first peptide group of JOM-13, thus explaining the low affinity of [NMe-D-Cys²]JOM-13 (Heyl, 1991). Replacements of Tyr¹ by D-Tyr¹ and Tic¹ residues, which have entirely different orientations of the tyrosine ring within the pocket, produce numerous overlaps with surrounding receptor atoms, which correlates with the observed low binding affinities of D-Tyr¹ and Tic¹ analogs of JOM-13 (unpublished observations).

View larger version (28K): [in this window] [in a new window]   FIGURE 8   JOM-13 (bold line) inside the binding pocket of the -opioid receptor (stereo view). Conserved (thin solid line) and variable (thin dashed line) residues of the binding pocket (within 4.5 Å of the ligand) are also shown.

The disulfide bonded D-Cys²-D-Pen⁴ pair of JOM-13 interacts primarily with the side chains of Thr²¹¹(V:0), Thr²⁸⁵(VI:27), Ile²⁸⁹(VI:31), and Leu³⁰⁰(VII:3). -Methylation of Cys² is expected to decrease binding, because the Me group would overlap with Leu³⁰⁰(VII:3) in the model. This is in agreement with the reduced binding affinity found for [Me-D-Cys²] JOM-13 (Heyl, 1991). The presence of empty spaces near the C-hydrogens of D-Cys² in the model is consistent with the comparable affinity observed upon replacement of D-Cys² by D-Pen² in analogs of JOM-13 (Mosberg et al., 1988).

The Phe³ side chain of JOM-13 (conformer with ¹ = 60°) occupies the bottom of a rather large nonpolar cavity that is extended toward the extracellular side of the -bundle and is covered by a -hairpin formed by EL-2 (the aromatic ring of Phe³ is located below the conserved Cys¹²¹(III:0)-Cys¹⁹⁸(EL-2) disulfide bond and interacts with Gln¹⁰⁵(II:24), Leu¹²⁵(III:4), Val²⁹⁷(VII:1), and His³⁰¹(VII:4); see Fig. 11). The presence of significant empty space in this cavity might allow a reorientation of the Phe³ side chain from ¹ = 60° to ¹ = 180°. In this case, the aromatic ring of Phe³ would occupy an alternate position, above the disulfide bond, and would interact primarily with residues from EL-2 (Val¹⁹⁷, Cys¹⁹⁸) and the extracellular terminus of TMH III (Glu¹¹⁸(III:-3), Cys¹²¹(III:0), Lys¹²²(III:1), Leu¹²⁵(III:4)). However, in the model, this would require a shift of the tripeptide ring system of JOM-13, which creates steric hindrances between Tyr¹ of the peptide and Tyr¹²⁹(III:8) of the receptor. Therefore, the preferred orientation of Phe in the  receptor is gauche (¹ = 60°), in agreement with the high affinities of [^zPhe³] and [(2R, 3S)MePhe³] analogs of JOM-13 (Mosberg et al., 1994b, 1996), in which the ¹ angles of residue 3 are fixed in this orientation. The reduced  binding affinity observed for [^EPhe³]JOM-13, in which ¹ of Phe³ is fixed at 180° (Mosberg et al., 1996