Introduction

Opioid Receptors

The opioid receptors are the primary sites of action of opiates, and transduce such information by activating G proteins that in turn alter the membrane conductance for K⁺ and Ca²⁺ and the levels of second messengers such as cAMP and inositol 1,4,5-trisphosphate. Pharmacological studies suggest that multiple opioid receptor subtypes, namely m-, d- and k- receptors, might be responsible for the actions of the opioid drugs including the regulation of pain [1, 2]. The development of opioid drugs that provide analgesics free from abuse potential and the adverse side effects of morphine remains a goal of medicinal chemistry in spite of the range of such agents as pentazocine and buprenorphine already in clinical use.

Recently, substantial progress has been made in elucidating the molecular features underlying the multiple opioid receptor subtypes through cloning their cDNA and genes, and the distinct genes for these multiple opioid receptor subtypes have been cloned [3]. Opioid-receptor genes encode proteins with seven hydrophobic membrane spanning domains that display a close homology with the G protein coupled receptors (GPCRs).

Analysis of Ligand-Receptor Interactions in the Opioid Receptors

It is clear that three dimensional understanding of the molecular interactions between the ligands and opioid receptors can give useful information for the rational design of opioid drugs. Medicinal chemists are able to propose new compounds with enhanced activity or selectivity profiles based on the analysis of the binding mode at atomic level. However, for this strategy, the 3D structure of target protein is required with atomic resolution, generally determined by either high resolution crystallography or multidimensional and multi-NMR spectroscopy. GPCRs of current pharma-ceutical interest are membrane-bound proteins and their 3D structures are unknown at present.

Due to the lack of 3D structures of these membrane protein receptors, indirect under-standing of the ligand-receptor interactions by synthetic modifications and the analysis of structure activity relationships have been performed [4]. Especially, a growing number of three-dimensional quantitative structure-activity relationship (3D QSAR) studies have been described on various GPCR ligands using the Comparative Molecular Field Analysis (CoMFA) methodology [5]. CoMFA studies allow the identification of the pharmacophoric arrangement of molecular fragments in space, and provide guidelines for the design of next-generation compounds with enhanced biological performance. However, in practice, attempts in converting the information from 3D QSAR into new synthetic candidates often failed to give satisfactory results because the non-atomistic models obtained by CoMFA studies do not always give clear directions of chemical modification.

An alternative approach to 3D QSAR is modeling of the ligand-receptor complexes based on three-dimensional computer models of target GPCRs, and molecular modeling of various GPCRs have been widely reported [6-8]. In contrast to the non-atomistic models, atomistic receptor models might give detailed insights into the key interactions between ligand and receptor in more straightforward fashion. Also, identifications of the ligand binding sites in these models have been made from the results of site-directed mutagenesis and studies of chimeric receptors, and the mode of the receptor-ligand interaction has also been analyzed by docking ligands into the receptor models. In the rat opioid receptors, the amino acid residue (m: Asp147; d: Asp128; k: Asp138) in the transmembrane domain III (TM III) was predicted to be involved in ligand binding pocket by the studies of the mutant opioid receptors [9,10]. Site-directed mutagenesis studies of the m- opioid receptor inidicate that Asp147 is probably the primary binding site as the counter ion for the ammonium head group of the opioid ligands in analogy with many other GPCRs. Some groups reported expression of chimeric receptors among their subtypes and analyzed their ligand binding properties to elucidate which portions of the receptor molecules are involved in the ligand binding and selectivity. They demonstrated the existence of the major binding determinants for the selective ligands in the opioid receptors [11-17].

The studies on the analysis of ligand-opioid receptor interaction based on the ligand-receptor complex model can be found in the literatures published recently [18-20]. For example, the ligand binding mode of a series of fentanyl derivatives was examined using a combination of conformational analysis and ligand docking to the m-opioid receptor. Also, the interaction of the k-opioid receptor with arylacetamide and benzo-morphane derivatives were modeled through pharmacophore-based and docking calculations, and potentially bioactive conformations of representative ligands were identified by systematic conformational analysis and ligand docking to the k-opioid rceptor. The reported results are consistent with ligand binding data derived from chimeric and mutant receptor studies as well as reported structure-activity relationships, and several topics such as agonist-antagonist binding and subtype selectivity were discussed based on the complex models.

The potential issues in these approaches are uncertainty of the models of GPCRs and predicted bioactive conformers of ligands. The main problem in modeling GPCRs is the low sequence homology of the receptors to that of bacteriorhodopsin or rhodopsin, which is currently used as the template for 3D modeling of GPCRs. It makes the sequence alignment difficult using as a template even though the resolution of the rhodopsin structure is improved, and this makes it difficult to define the starting and ending points of a-helices through the membrane. Selection of potentially bioactive conformers of the ligand molecule is another issue, and conformer selection significantly affects the results of the ligand-receptor docking. Therefore, tremendous efforts have been expended for the prediction of probable bioactive conformers of flexible ligand molecules. Considering these limitations of the receptor-ligand complex models, it is obvious that the models should be validated by additional experiments before they are subjected to drug design.

Validation of the Ligand-Receptor Complex Models

As one of the most effective methods to validate the complex models, site-directed mutagenesis has been widely employed and found to give significant information on ligand-receptor interactions [6, 21, 22]. Site-directed mutagenesis of single sequence position of interest can be used to determine their importance for ligand binding. In the 3D models of the target receptor, the binding site that is suggested by the data obtained by mutagenesis can be hypothesized and used to forecast a next set of mutated receptors, and these will lead to a further optimization of the receptor-ligand complex model. The problem of site-directed mutagenesis is that frequently mutants with loss of function or binding are generated and it becomes impossible to distinguish between indirect effect caused by change of conformation of the receptor or direct effect caused by the loss of original interactions in the wild receptors [8].

An alternative to the approach using mutant receptors is the affinity labeling experiment, which cross-link the ligand and the receptor at the binding site and will give topological information of the ligand binding site. A ligand with photoactive substructure that generates highly reactive species, for example, a carbene, nitrene or biradical is typically used in this approach. Affinity cross-linking experiments of NPY Y₂ receptors by photo- affinity labeling have been reported, and photo-cross linking using a ligand bound photo reactive amino acid has been used successfully to characterize hormone receptors. Furthermore, direct determination of the receptor sequence positions involved in ligand binding is made possible by photo-affinity labeling experiments followed by the purification of the ligand-receptor complex. Issues in this approach are the difficulties to develop amino acid analog that meets all requirements for a perfect photo affinity labelling (c.f., chemical stability prior to photo activation, smooth photolysis at wavelengths long enough to cause no photochemical damage to the proteins, high cross-linking yields, uniformity and stability of the cross-linking products, and low tendency of the intermediate to react with the solvent water). However, if such ligands can be designed, the affinity labeling experiment will be a powerful method to obtain direct evidence of ligand-receptor interactions.

An alternative to the photo-affinity labeling reagents is the use of ligands with electrophilic substituents that covalently bind to nucleophiles in the binding pocket. There have been many reports on the affinity labeling of the opioid receptors, and some opioid ligands that were attached to electrophilic groups were found to label the opioid receptors. Determination of the location of covalently bound residue was also reported, and b-funaltrexamine [27, 28] was found to bind to Lys233 of the m-receptor. Recently, o-phthalaldehyde derivative of b-naltrexamine was found to cross-link neighboring lysine and cysteine residues in m-, d- and k-opioid receptors through a highly selective labeling process .

The involvement of thiol (SH) groups in maintenance of structure and function has been observed for some GPCRs, including acetylcholine and adrenergic receptors [30-32]. The interaction of the opioid ligands with their receptors has also been proposed to be regulated by SH groups. The binding of the opioid ligands with opioid receptors in rat brain is effectively inhibited by reagents that react with SH groups such as N-ethylmaleimide (NEM) [33-35]. Larsen et al. suggested that in opioid receptors, there are at least two different types of SH groups sensitive to NEM. Both of the SH groups are originated from the cysteine residues: one is the cysteine b-SH in the GTP-binding regulatory protein Gi, which exists inside the plasma membrane and couples with the opioid receptor to let the receptor react with agonists, and the other is in the ligand binding site of the receptor protein.

From these observations, it was assumed that topological information rgearding cysteine residues in the ligand binding pocket can be obtained from direct labeling of the cysteine residues by the affinity ligand. Affinity probes with functional groups that can specifically label the binding site cysteine residues must be developed for this purpose. These experiments might add useful information regarding relative location of the cysteine residue and the ligand in the binding sites, and consequently make it possible to predict the binding site involves the cysteine residue and validate the ligand-receptor complex models combined with other experimental data.

These considerations prompted us to investigate the features of the opioid receptor-ligand interaction based on the experimentally validated receptor models by the following process. As the first step, the affinity-labeling experiments are performed to obtain topological information of the cysteine residue in the binding pocket. Then, based on the results of affinity labeling experiments, the ligand binding sites are predicted in the opioid receptor models which are constructed by homology modeling. Finally, possible interactions between the ligand and the receptor is analyzed after the refinement of the complex model with energy minimization.

Affinity Labeling of the Opioid Receptors

Affinty Labeling of m-Opioid Receptors in Guinea-Pig Ileum

In our approach, morphine molecule was selected as the scaffold for the affinity-labeling ligand. There are some advantages for using morphine as the molecular probe. The morphine molecule possess a non-peptidic nature and good solubility as a biological tool and possess a unique T-shaped, highly rigid structure. In particular, the highly rigid framework of morphine makes it significantly easier to understand the relationships between the structure of the ligand and the results in affinity labeling and ligand docking experiments compared to flexible molecules. Furthermore, candidates of bioactive conformation and alignments of multiple ligands might be easily obtained for morphine analogs. Therefore, morphine is a classical drug but still very attractive as small molecular probe.

In the next step, the electrophilic center was introduced into morphine molecule. Selection of the electophilic substructure and the position to be attached is the critical process in designing affinity labeling ligands. Portoghese et al. suggested that two consecutive recognition processes are involved in electrophilic affinity labelling that lead to the covalent binding receptors. The first recognition step is reflected by receptor affinity, and the second one involves the proper alignment of the electrophilic center (attached to the reversibly bound ligand) with a compatible, proximal, receptor-based nucleophile. Because two recognition steps rather than one lead to the covalent binding of the affinity ligand~~abel~~, enhanced receptor selectivity (recognition amplification) is attainable. Thus, the specific covalent binding is dependent upon the nature and orientation of the electrophilic center in the affinity ligand.

Scheme 1. Synthesis of affinity labeling ligand 1.

In vitro functional activities of ligand 1 were tested by the inhibition of the electrically stimulated contraction of guinea pig ileum (GPI) and mouse vas deferens (MVD); the IC₅₀ of ligand 1 in GPI and MVD was 9.3 and 76 nM, respectively [43]. Ligand 1 showed about three times higher potency than morphine in mice. The SH-disulfide exchange reaction with opioid receptor in GPI was examined by the wash-out method [44] as shown in (Fig. 1).

Fig. (1). Recording of the electrically stimulated contraction of Guinea Pig Ileum. A : Irreversibility of the activity of 1 mM compound 1 by washing; B : Reversibility of the activity of 1 mM compound 1 by treatment with 1 mM naloxone and recovery of the original inhibitory activity by washing.

When GPI was incubated with 1 mM of compound 1 for 10 min, the activity was not reversed at all even by washing 50 times (Fig. 1A), indicating a continuous stimulation of the receptor by tight binding of 1. With 1 mM of naloxone, a m-selective antagonist, the retained activity was completely reversed (Fig. 1B). Then washing 10 times elicited 95% of the activity of compound 1 again, suggesting that compound 1 can be replaced by naloxone at the binding site, and still be retained near the binding site. In contrast, incubation of GPI, of which contraction had been fully suppressed by compound 1, with 1 mM of dithiothreitol (DTT) for 30 min eliminated almost all of the activity after washing as shown in (Fig. 2).

These results indicate that compound 1 reacted with an SH group in opioid receptors near the position, where compound 1 interacts, through a disulfide linkage and that this linkage was cleaved reductively by DTT as illustrated in (Fig. 3). This is the evidence of compound 1 linking to the m-opioid receptor through the SH-disulfide exchange reaction.

Fig. (2). Reversibility of the activity of 1 _M compound 1 by treatment with 1 mM DTT and washing

Affinity Labeling of the Opioid Receptors in Guinea Pig Brain

Since the covalent linkage of compound 1 to the peripheral opioid receptors was proved, we have prepared the S-activated (-)-8b-sulfhydryl-dihyd-romorphine analogue 2, expecting the differential regiochemical effect between 1 and 2 on predo-minant affinity labeling of opioid receptors. To evaluate their binding abilities for the opioid receptors in detail, we tested compounds 1 and 2 by radio labeled ligand binding assays using membrane preparations from guinea pig brain . The synthesis of compound 2 from morphine acetate is shown in Scheme 2. Specific binding affinities of compounds 1 and 2 for m- opioid receptor in guinea pig brain were determined by evaluating their abilities to displace [³H]-[naloxone] [46]. The radioligand receptor binding assays were carried out as described previously [47]. Scheme 2

Fig. (3). Schematic illustration of the interaction of compound 1 with an opioid receptor.

Both 1 and 2 potently interact with the m-receptor. The IC₅₀ values, the half-maximal concentration of unlabeled ligand for inhibition of binding of labeled ligand, were 14.2 and 13.4 nM for 1 and 2, respectively. Binding affinities of 1 and 2 with d- opioid receptor were evaluated using [³H]-[D-Ser², Leu⁵]enkephalyl-Thr⁶ ([³H]DSLET) as a tracer [48], and 1 and 2 exhibited IC₅₀ values of 5.7 nM and 26.8 nM, respectively. Their binding affinities for k-opioid receptor were evaluated using [³H]U-69593 as a tracer [49], and 1 and 2 exhibited IC₅₀ values of 56.7 nM and 136.9 nM, respectively. Ligand 1 showed higher affinity for d-receptor sites than m- or k-receptor sites.

Scheme 2. Synthesis of affinity labeling ligand 2.

The ability of compounds 1 and 2 to label irreversibly opioid receptors was examined. The affinity labeling of receptors would substantially reduce the number of receptors available for binding of the ligands added afterwards. Thus, after incubation of membranes with S-activated ligands, the subsequent receptor binding assay would estimate the amount of receptors unlabeled and consequently the amount of labeled receptors. To estimate the total amount of the unlabeled m-receptors, the amount of naloxone that displaces radiolabeled [³H]naloxone was measured. For the d- or k- opioid receptors, to estimate the total amount of the free receptors, the amount of DSLET that displaces [³H]DSLET or U-69593 that displaces [³H]U-69593 was measured, respectively. The affinity labeling experiment was performed essentially as described previously [50].

When 1 was incubated with the guinea pig brain membranes, it was found that the amount of free receptors diminished sharply, depending upon the concentrations of 1 as shown in (Fig. 4).

Fig. (4). The concentration-dependent affinity labeling of m(A), d(B) and k(C) opioid receptors by ligands 1 (m) and 2 (l).

(Fig. 4A) indicates, for example, that when the guinea pig brain membrane is incubated with 1 mM of the 6b-S-activated ligand 1, it occupies approxi-mately 60 % of m- receptor, while the 8b-S-activa-ted isomer 2 cross-links only about 25% of the m- receptor. The concentration-dependent curves of affinity-labeling of d- and k- receptors are depicted as in (Fig. 4B and Fig. 4C), respectively. The effective concentrations (EC₅₀) of compound 1 to occupy the half-maximal amounts of receptors were estimated from (Fig. 4) and shown in (Table 1).

It should be noted that 2 was not potent enough to affinity-label the opioid receptors completely in our experiments. Since the IC₅₀ and EC₅₀ values of compound 1 for m- receptor are almost comparable to each other (14.2 and 420.2 nM, respectively), it appears that the compound 1 bound to the binding site almost inevitably forms a cross-link with the receptor molecule by a disulfide bonding. In sharp contrast, compound 2 lacks the potential ability to affinity-label the opioid receptor. In particular, in spite of strong binding affinity with the opioid receptor, its ability to form a cross-link with the receptor molecule was found to be extremely weak. This is certainly the reflection of regiochemistry of SH groups activated by the Npys group, which is attached to position C₆ or C₈. The receptor’s SH group seemed to be present near the portion where the morphine 6b-substituent is located, and the spatial proximity may cause a SH-disulfide exchange reaction between compound 1 and the SH group in the m-receptor. On the other hand, the 8b-SH group of 2 is in a much less favored position to interact with the SH group in the receptor.

The effect of DTT on the affinity labeling was also examined in guinea pig brain assay. After incubation of membranes with S-activated ligand 1, the membranes were treated by DTT (5 mM or 7 mM) and the amount of labeled receptors was estimated by subsequent receptor binding assay [51]. As shown in Fig. 5), m-receptor was recovered in time-dependent manner, and almost 100% of total receptor was recovered by 8 hours treatment of 7 mM of DTT. These results are direct evidence that the observed affinity labeling of 1 in the guinea pig brain is caused by the formation of covalent disulfide bonding between the ligand and the m-receptor.

Fig. (5) Receptor recovery from affinity labeling of ligand 1 iμ opioid receptor.

Although the role of SH group in the molecular mechanism of receptor responses has not been clarified yet, the present results indicate that the opioid receptor protein contains a distinct free SH group, from cysteine a residue, in the ligand binding site in a specific position relative to the receptor-bound ligand molecule.

As an additional study for the further confirmation of this hypothesis, ligand 3 (Fig. 6) was examined in a similar affinity labeling test in rat brain membrane using [³H]-[D-Ala², MePhe⁴, Gly-ol⁵]enkephalyl-Thr⁶ ([³H]DAGO) as a tracer (unpublished data), and the results showed significant level of recovery of the labeling ability was observed for the ligand 3 with longer side chain compared to 2 as shown in (Table 2). This observation also supports the hypothesis; the SH group exists in a specific position relative to the receptor-bound affinity ligand, and the feasibility of the cross-linking was significantly affected by the regiochemistry of the electrophilic center of the labeling agents.

Fig. (6). The structure of affinity ligand 3.

Table 2. Effects of Ligands 1, 2 and 3 in Binding (IC₅₀) and Affinity Labeling (EC₅₀) Experiments in Rat Brain

Molecular Modeling of Multiple Opioid Receptor Subtypes

As useful information of the ligand-receptor interaction was obtained by the affinity labeling experiments, the 3D models for opioid receptor subtypes were constructed as the next step. The amino acid sequences of the rat m-, d- and k-opioid receptors were aligned based upon representative studies by Fukuda et al. [52], Yasuda et al. [531], and Minami et al. [542], respectively. For the identification of the hydrophobic helical regions, the parameter sets of Kyte-Doolittle were used [553]. Primary amino acid sequence alignment clearly defined seven highly conserved hydrophobic sequences corresponding to the TM regions as shown in (Fig. 7). Because it was not possible to localize precisely the starting and ending amino acids of the TM regions from hydropathy analyses, we predicted the putative TM regions of the three receptor subtypes according to the report described by Yasuda et al. [531].

A detailed comparison of the TM regions showed the high percentage (69-72 %) of sequence identity within the opioid receptor subtypes. The extremely high homologies observed in the TM regions among the opioid receptor subtypes correlate very well with the traditional structure-activity relationships of opioid ligands. On the other hand, the extracellular domains of the opioid receptors showed a low percentage (25-33%) of sequence identity. The studies of the chimeric opioid receptors indicated that nonpeptidic opioid ligands such as morphine derivatives possess the distinctive binding sites from those for the opioid peptide ligands [11-13, 15-17]. The opioid peptide ligands were assumed to bind the receptor spanning TM and the extracellular loop regions. On the other hand, nonpeptidic opioid ligands bind to a hydrophobic pocket buried in the TM regions. Based on this assumption, only the TM regions were modeled and subjected to further analysis as the binding sites for nonpeptidic opioid ligands.

Fig. (7). Alignment of the amino acid sequences of m-,d- and k-opioid receptors. The predicted transmembrane segments (TM 1-7) are indicated; the termini of each segment are tentatively assigned according to the report.

The structures of the membrane spanning helices in the bacteriorhodopsin protein (PDB code 1BRD) [56, 57] were assumed to construct three-dimensional structures of the TM domains of the aforementioned three opioid receptors. The receptor models were constructed from the amino acid sequences according to the following procedures: (1) The hydrophobic moment [58] of each helix of bacteriorhodopsin was calculated, and the amplitude and direction of the moment were indicated by a vector. (2) The average hydrophobic moment of the three opioid receptors was obtained from the aligned amino acid sequences. (3) The seven TM helices in each opioid receptor were constructed by taking the backbone (f, y) angles as (-59°, -44°) [59] and the preferable side-chain rotamer structures, as determined by Ponder and Richards [60]. (4) Each helix was rotated around the helical axis so that the direction of the hydrophobic moment agreed with that of bacteriorhodopsin. Here, the second, sixth and seventh helices were further rotated so that the side-chains of aspartate, cysteine and the aromatic residues would be oriented to the interior of the helices. (5) Each helix was moved in the direction normal to the membrane on the graphics screen. (6) The whole structural energy of each TM domain in the three opioid receptors was minimized to get rid of bad contacts in each final model structure.

The construction of the model for opioid receptors was done using Insight II/Discover (Molecular Simulation Inc.). The energy minimizations were carried out by molecular mechanics calculations with CVFF force field in Discover. The high degree of similarity within these hydrophobic stretches leads to the assumption that the homologous TM regions in all opioid receptors possess the same secondary structures and folds in the same way. The distribution of the conserved and charged amino acids on the same face of the helices imply that the assumptions made above are correct. The receptor models posses a number of features that we believe to be essential for this class of membrane-embedded receptors : the seven a-helices are tightly placed and define a central narrow, dihedral cleft.

Prediction of the Ligand Binding Sites in the Opioid Receptor Subtypes

We attempted to estimate the ligand recognition sites in the opioid receptors from the results of our affinity labeling experiments and the homology between the opioid receptors and other GPCRs. To suggest amino acid residues participating in the ligand binding, our basic idea was that the conserved amino acid residues within all subtypes, which are involved in the ligand binding sites, play an important role in the overall folding and function of the receptors. On the other hand, the amino acids specific to the receptor subtype might be responsible for the binding of the corresponding ligands and for some of the specific triggering mechanisms.

Conserved Aspartic Acid Residues

All subtypes of opioid receptor possess three invariant acidic residues. One aspartic acid (m : Asp114; d : Asp95; k : Asp105) is highly conserved across the whole GPCR family, indicating that this residue probably plays an essential role in the folding and/or the function of the receptor. The second aspartic acid (m : Asp164; d : Asp145; k : Asp155) might be a member of the sequence coupling with G proteins. On the other hand, the conserved aspartic acid (m : Asp147; d : Asp128; k : Asp138), which is located in the middle of TM III near the extracellular domain, is conserved in opioid receptors as well as cationic neurotransmitter receptors. Mutagenesis experiments of m- opioid receptor inidicate that the Asp residue in TM II is of importance for the binding of agonists, and the Asp residue in TM III is of importance for both agonists and antagonists . The Asp residue in TM III is probably the primary binding site for the opioid ligands with a protonated nitrogen. The Asp residue in TM II may have an allosteric effect on agonist affinity in accordance with opioid receptors where the corresponding residue has mutated. This resulted in an impaired sodium effect when binding agonist. Considering these results, it was suggested that the binding site of opioid receptor might be in the cleft centered in the Asp residue in TM III.

Conserved Cysteine Residues

As described above, our results of affinity labeling experiments using S-activated dihydromorphine derivatives suggested that the opioid receptor protein contains at least one Cys residue in the ligand binding site and the SH group of the Cys residue seemed to be present near the portion where the morphine 6b-substituent was located rather than 8b-substituent, and the conserved Cys residues might participate in the ligand binding in the opioid receptors.

The primary sequence analysis defined several conserved Cys residues within the three types of opioid receptors, but did not suggest which Cys residues could be located in the ligand binding sites and labeled by our S-activated ligands. However, our 3D receptor models indicated that conserved Cys residues (m : Cys79; d : Cys60 in TM I or m : Cys321; d : Cys303 ; k : Cys315 in TM VII) could be involved in the pocket containing the Asp residue in the center, and interact with the activated disulfide bond of S-activated ligands. These Cys residues are conserved within the opioid receptor subtypes and are absent in other cationic neurotransmitter receptors. When S-activated ligand 1 was manually docked into the receptor model so that a salt bridge can be made between the cationic nitrogen of the ligand and aspartate residue in TM III, the SH group of the residue in the TM VII ( m : Cys321; d : Cys303 ; k : Cys315) or the residue in the TM I (m : Cys79; d : Cys60) are close enough to the activated disulfide bond of 1 to undergo the SH-disulfide exchange reaction. Therefore, we presumed that the clefts containing these Cys residues might be the binding sites for the morphine derivatives, and defined the pocket involving the Asp residue in TM III (m : Asp147; d : Asp128; k : Asp138) and the Cys residue in the TM I (m : Cys79; d : Cys60) as the binding site I, and the pocket involving the Asp residue in TM III (m : Asp147; d : Asp128; k : Asp138) and the Cys residue in TM VII ( m : Cys321; d : Cys303 ; k : Cys315) as the binding site II as shown in (Fig. 8).

Conserved Aromatic Residues in the Predicted Binding Sites

As mentioned in the primary sequence alignment, all opioid receptors contain many conserved aromatic residues. It is generally known that the aromatic residues can be involved in important internal cross-linking hydrogen bonds and conformational changes [61]. In particular, the Tyr residues in TM III, the Phe residues in TM VI (Tyr in m-receptor) and Tyr residues in TM VI that are located in the predicted ligand binding sites I and II are conserved within all opioid receptor subtypes and are absent in the other cationic neurotransmitter receptors. The conserved Tyr residue in TM III (m : Tyr148; d : Tyr129; k : Tyr139) is located at the position adjacent to the aspartate, which may stabilize the ligand-receptor complex by an electrostatic interaction between its side chain aromatic ring and an ammonium cation head group of the opioid ligands. When a morphine molecule was manually docked into the predicted binding sites I and II, where the quaternary ammonium cationic head group was placed to form electrostatic interactions with the aspartate in TM III, the phenyl ring of the morphine in both of the binding sites I and II were located close to the phenyl ring of the Tyr residue (m : Tyr325; d : Tyr308; k : Tyr320) in TM VII and the Phe or Tyr residue (m : Tyr299; d : Phe280; k : Phe293) in TM VI, respectively. These aromatic-aromatic interactions might stabilize the ligand-receptor binding.

Fig. (8). Putative binding sites I and II in the m-opioid receptor model.

Variant Amino Acid Residues in the Predicted Binding Sites

As described above, the primary sequence analysis of the three subtypes indicates high homology in the predicted TM regions, and a number of conserved amino acid residues that may play an important role in the ligand-binding were defined. On the other hand, a variant amino acid residue involved in the ligand-binding sites could be characteristic for the ligand-binding of each subtype of opioid receptor. A comparison of the 3D models of opioid receptors defined the variant amino acid residues, which are different in charge and hydrophobic properties, within the three subtypes. In the upper region of TM VI in the putative ligand-binding sites, the m-, d- and k- receptors possess Lys303, Trp284 and Glu297, respectively. The variant amino acid residues are able to participate in the mechanism that controls the selectivity of the ligand binding. For example, Glu297 was identified as the key residue in the binding and selectivity of the k-ligand such as norBNI and GNTI [623] and Trp284 was a key recognition element for d-selective piperazine derivative SNC80 [634].

From these observations, we presumed the ligand recognition mechanisms in the binding sites I and II as: (1) the carboxylate of the aspartate residue in TM III (m : Asp147; d : Asp128; k : Asp138) electrostatically interacts with the quaternary ammonium cation in the opioid ligand. (2) The tyrosine residue in TM VII (in binding site I, m : Tyr325; d : Tyr308; k : Tyr320) or the Phe or Tyr residue in TM VI (in binding site II, m : Tyr299; d : Phe280; k : Phe293) interacts with the phenol part of opioid ligands with an aromatic interaction. (3) The cysteine residue in TM I (in binding site I, m : Cys79; d : Cys60) or TM VII (in binding site II, m : Cys321; d : Cys303 ; k : Cys315) is the second recognition site that undergoes the SH-disulfide exchange reaction with the S-activated dihydromorphine derivatives. (4) The variant amino acid residue (m : Lys303; d : Trp284 ; k : Glu297) which can be found only in the binding site II may interact with the ligand and participate in the mechanism that controls the selectivity of the ligand binding.

Analyses of the Predicted Ligand Binding Sites by Ligand Docking

We tested the predicted binding sites I and II by ligand docking analyses. The structures of ligands, whose crystal coordinates have been known, were derived from CSD (Cambridge Structural Database System). The ligands were manually docked into the 3D models of the opioid receptors, and refinement of the initial receptor-ligand complex was achieved by in vacuo energy minimizations using Insight II/Discover with CVFF force field in Discover (Molecular Simulation Inc.). The parameters were the distance-dependent dielectric constant of 2 and the non-bonded cut off of 15 Å, and no solvent molecules were included in the calculation. The models were energy minimized for 500 steps with steepest descent minimizer and subsequently until rms energy gradient was less than 0.1 kcal/molÅ with the conjugate gradient minimizer. An initial model for the opioid ligand-receptor complex was obtained by docking ligand into receptor model by manual adjustment.

Fig. (9). Schematic representation of the observed interactions between morphine and the predicted binding site II of the m-opioid receptor.

Morphine was docked into the binding sites I and II of the m-opioid receptor model as described above. In the binding site I, the sulfur atom of Cys 79 in TM I and the phenyl ring of Tyr325 in TM VII were considerably apart from morphine (7.1 Å and 5.8 Å respectively). Moreover, no other amino acid residue that was participating in ligand binding was found. On the other hand, the cationic nitrogen of the morphine in the binding site II was positioned to possess an electrostatic interaction with the carboxylate side chain of Asp 147 (2.7 Å) and the phenyl ring of the morphine may have aromatic interactions with Tyr148 and Tyr 299. Moreover, the morphine molecule was stabilized by hydrogen bonding between the phenolic hydroxyl group of morphine and the phenolic hydroxyl group of Tyr148 and the amino group of side chain of Lys303, respectively. The schematic illustration of morphine molecule in the binding site II is shown in (Fig. 9).

Fig. (10). Schematic representation of the observed interactions between the predicted binding site II of the m-opioid receptor and ligand 1 (A) or ligand 2 (B).

The docking of the morphine molecule into the binding sites of d- and k-opioid receptors was also analyzed, and the similar interactions were hypothesized only in the binding site II of the d- and k-opioid receptors. Only conserved aromatic residues in TM III (d : Tyr129; k : Tyr139) and TM VI (d : Phe280; k : Phe293) in the predicted binding site II were able to participate in the hypothesized ligand-receptor interactions. From these results, we concluded that the binding site II in the opioid receptor subtypes might be the ligand binding site for the morphine derivatives. It should be noted that Fukuda et al. expressed chimeric receptors between the rat m- and d-opioid receptors and analyzed their ligand binding properties, and major determinant for binding of the m-selective morphine is demonstrated to exist in the region spanning the TM V-VII. These results are consistent with our conclusion. For the further characterization of the binding site II, we have analyzed docking of the affinity labeling ligands 1 and 2 in the m-opioid receptor (Fig. 10).

After optimization of the structure of the complex models of the affinity ligands 1 and 2, the distances were measured between the amino acid residues in the predicted binding sites and each ligand. Although the distance between the carboxylate side chain of Asp147 and the cationic nitrogen of each ligand (2.7-2.8 Å) and the distance between the phenyl ring of Tyr299 in TM VI and the phenyl ring of each ligand (3.3-3.5 Å) were almost comparable, only the distance between the SH group of Cys321 and the activated SH group of compounds 1 and 2 were considerably different (3.4 Å and 6.5 Å, respectively). This observation may explain the difference in the irreversible binding abilities between the compounds 1 and 2, so the present complex model may account for our results of affinity labeling experiments.

Next, our interest was focused on the elucidation of the features of the antagonistic binding in m-opioid receptors. According to the studies of structure-activity relationships of m-selective ligands [64], it was found that replacement of the methyl group on the nitrogen atom with an allyl or cyclopropylmethyl group dramatically affects the pharmacological reactivity to transform opioid agonists into antagonists. The explanation for this observation was made on the basis of the difference of the location of the cationic nitrogen atom in the binding site between the agonists and antagonists. It was suggested that allyl or cyclopropylmethyl substituents of m-antagonists have Van der Waals interactions with the receptor that results in moving the nitrogen away from the position required for agonistic binding. Although several groups have suggested that the agonist/antagonist activity of opioid ligands is determined by stereoelectronic interactions of the amino moiety with the receptor, there is no agreement in the literature. To obtain some insight into agonist/antagonist binding, the N-cyclopropylmetyl derivative of morphine (naltrexone, (Fig. 11) was selected as a representative m-antagonist and was docked into the predicted binding site in the m-opioid receptor model.

Fig. (11). The structure of naltrexone.

When naltrexone was located in the same position as morphine in the m-receptor model, the cyclopropylmethyl group was hindered by repulsive interactions with the main chain of TM III, and naltrexone bound the receptor in a manner slightly different from morphine. In the morphine-m-opioid receptor complex, the distance from the nitrogen atom of morphine to Asp147, Cys 321 and Tyr 299 was 2.7 Å, 10.1 Å and 6.6 Å respectively. On the other hand, in the complex of naltrexone with the m- opioid receptor, the distance was 2.8 Å, 8.6 Å and 4.8 Å, respectively. The differences between the nitrogen atom of each ligand to each amino acid residue except Asp147 were more than 2.0 Å.

Hypothesis regarding the interaction of receptor and agonist/antagonist was reported [65], and GPCRs are assumed to exist in at least two conformations. The active conformation interacts with G proteins, but the inactive (resting) conformation can not bind G protein, and the hypothesis can explain the existence of agonists, antagonists and partial or inverse agonists . Although this docking analysis could not suggest the actual mode of agonist/antagonist interactions, the present results might lend support for the existence of active and inactive conformation of the opioid receptor, and significant conformational difference might be expected between these two states.

As a result of the foregoing work, possible interactions of morphine derivatives and opioid receptors were predicted based on the combination of the results from affinity labeling experiments and ligand docking experiments. To study the predicted binding mode in detail and to test the usefulness of the current models for drug design of morphine derivatives, attempts at model-based ligand design were made.

Design of a k-Agonist Based on the Complex Model

Design and Synthesis of KT-95

Among the three opioid receptor subtypes, the k-opioid receptor has been of most interest because its activation produces analgesia with minimum physical dependence and respiratory depression. It was also assumed that the k-opioid receptor system is significantly activated under conditions of severe pain [66]. Unilateral inflammation in the hindpaw of the rat results in an increase in the expression of preprodynorphin and preproenkephalin mRNA in the spinal cord. Thus, k-opioid agonists may provide useful analgesics free from the abuse potential and the adverse side effects of m-agonists like morphine. Recently, we have described the synthesis of (-)-3-acetyl-6b-acetylthio-N-cyclopropylmethyl-normorphine (KT-90), (Fig. 12) and its pharmacological activities [67, 68]. Analgesic actions of KT-90 were mediated through an activation of k-opioid receptors and about 10 times as potent as morphine. Also, KT-90 might be free from undesirable physical and psychic dependence liabilities, as KT-90 acted as k-agonist, but at the same time, behaved as strong antagonist for m- and d-opioid receptors.

Fig. (12). The structure of KT-90.

Chronic high-dose morphine therapy has been widely used for severe cancer pain in palliative care [69], resulting in the discovery of extra-analgesic actions of opioids, such as immuno-suppression [70] or unexpectedly prolonged life span of cancer patients. Recently, anti-cancer activities of morphine derivative were reported [71-73], and KT-90 was found to significantly inhibit the growth of human cancer cell lines up to 80 times more potently than morphine [74]. Although this anti-cancer activity was not mediated by the activation of opioid receptors, these data suggested that KT-90 may be a useful prototype for the clinical analgesics for the cancer treatment.

Thus, we decided to apply rational drug design approach based on the obtained complex models for the development of potent KT-90 analogs with improved k-agonistic activity. Several potent k-ligands like ketocyclazocine (KCZ) or ethylketocyclazocine (EKC) possess an oxo group in the benzylic position [64, 75]. It appears that the oxo group of these ligands might contribute to the affinity to the k-receptor. Based on this consideration, new compound KT-95 which possess an additional benzylic oxo group to KT-90 was designed [76] as shown in (Fig. 13).

Fig. (13). The structures of KT-95.

As mentioned before, a comparison of our 3D models of the three opioid receptor subtypes defined variant amino acid residues (m: Lys 303, d: Trp284, k: Glu297) in TM VI. It is likely that the variant amino acid residues are able to participate in the mechanism that controls the type- selectivity of the ligand binding. When KT-95 was tested by ligand-docking in k-opioid receptor by the method previously mentioned, the positions of the carboxylate side chain of Glu297 and the benzylic oxo group in the ligand molecule were close enough to allow interaction with each other. The observed interactions in the complex model are shown in (Fig. 14).

The importance of Glu297 in ligand recognition in k-opioid receptors has also been reported by Metzger et. al. NorBNI, a selective k-antagonist was predicted to form an ion pair between one of the cationic amine moieties in norBNI and Glu297 in their model, and this was consistent with the results of the site-directed mutagenesis experiments. On the other hand, point mutation of Glu297 did not affect the binding of highly selective arylacetoamide k-agonists, and the different binding mode of aryl acetoamide from norBNI was predicted by ligand docking study . The 3D structures of their k-opioid receptor model [18, 77] and our model were compared. Although the orientation of the a-helices and the juxtapositions of the amino acid residues involved in the predicted binding site of morphine derivatives were well-conserved between these two models, a slight difference was found in the portion involving Glu297. When the distances between the a-carbon of three amino acid residues (Asp138 in TM III, Glu297 in TMVI and Cys315 in TMVII) were measured, the distances between Asp138 and Cys315, or Glu297 and Cys315 were almost comparable. However, the distance between Asp138 and Glu297 was slightly different (15.8 Å in our model and 19.6 Å in their model). This may give slightly different binding mode of the ligands in the docking studies between these two models.

Fig. (14). The interactions between KT-95 and the predicted binding site of k-opioid receptor model.

To validate our complex model and the importance of the benzylic oxo group in KT-95, we decided to examine the biological properties of KT-95. KT-95 was prepared [76] as shown in Scheme 3.

Biological Evaluations of KT-95

The affinity of KT-95 was examined in the crude synaptosomal fractions of rat brain [2, 76,78-80], and pIC₅₀ values, being the negative logarithm of 50% inhibitory concentrations, are summarized in (Table 3). The affinity of KT-95 to k-receptors was found to be about 18 times higher than that of morphine and about 5 times than that of KT-90.

Scheme 3. Synthesis of KT-95.

Table 3. Inhibitory Effects (pIC₅₀values) of KT-95, KT-90 and Morphine on the specific Binding of ³H-Labeled Opioid Ligands to Rat Brain Crude Synaptosomal Fractions

The electrically stimulated ileal preparation of the guinea-pig was used as a model to study the action of narcotic analgesics in the central nervous system. KT-95 inhibited the twitch response of the longitudinal muscle of the GPI, which contains both m- and k-opioid receptors and rabbit vas deferens (RbVD), which contains only k-receptors, to electrical stimulation in a concen-tration-dependent manner, and the inhibitory effect of KT-95 was about 17 times more potent than morphine and about 4 times more potent than KT-90 in GPI (Table 4).

Table 4. PD₂ values of KT-95, KT-90, Morphine and Dynorphine (1-13) on the Electrically Stimulated Longitudinal Muscle of the Guinea Pig Ileum and/or Rabbit vas Deferens

The site of action was examined by the antagonism between KT-95 and norBNI, a k-antagonist [83]. NorBNI caused parallel rightward shifts of the concentration-response curves to KT-95 in a concentration-dependent manner. As shown in (Table 5), the pA₂ values for norBNI, calculated from the Schild plot, is in agreement with the generally accepted value found on k-receptors in several tissues. These results suggested that KT-95 is a potent k-agonist.

Table 5. PA₂ Values and Slopes of the Schild Plot for norBNI Against KT-95 on the Electrically Stimulated Longitudinal Muscle of Guinea Pig Ileum and/or Rabbit vas Deferens

Also, KT-95 was found to antagonize the response of morphine (m-agonist) in the longitudinal muscle of the GPI pretreated with norBNI, and the pA₂ value was 8.51 ± 0.08. Moreover, KT-95 was found to antagonize the response of Leu-enkephalin (d-agonist) in electrically stimulated MVD pretreated with norBNI with pA₂ value of 8.16 ± 0.15. These results indicate that KT-95 behaves as an agonist, mediated through k-receptors, with m- and d-antagonistic activities.

The analgesic activities of KT-95, given s.c., were 20 and 58 times more potent than morphine in the pressure test [843] and in the acetic acid-induced writhing test, respectively. The potency of KT-95 in the pressure test was similar to that in functional studies and in the binding assay. The analgesic effect induced by KT-95 was abolished by simultaneous administration of norBNI, a selective k-antagonist, in the pressure test and acetic acid-induced writhing test as shown in (Fig. 15) and (Fig. 16), respectively. These data suggest that analgesia induced by KT-95 is mainly mediated through k-receptors. The analgesic effects of KT-90 were 5 and 10 times more potent than morphine in the pressure test and in the acetic acid-induced writhing test, respectively [68]. Therefore, KT-95 might be more potent than KT-90, especially in the writhing test.

Fig. (15). Effects of norBNI on the antinociceptive action of KT-95 in the pressure test. NorBNI was administered 30 min before KT-95 was applied. Ordinate: Analgesic index, which is shown as the ratio of the threshold pressure (g) 30 min before and after injection, and is calculated and expressed as percentage of control responses. Each column represents the mean±S.E. of 4 experiments. ***p<0.001: Compared to the control. ^###p<0.001: Compared to KT-95

These results suggested that newly introduced oxo group of KT-95 may play an important role in improving its k-agonistic activity relative to KT-90. The hypothesis discussed in our analysis of the KT-95-receptor complex may be a possible explanation for the remarkably improved k-agonistic activity of KT-95. Thus, further improvement of the ligand-receptor complex models might be made based on the present results.

Fig. (16). Effects of norBNI on the antinociceptive action of KT-95 in the acetic acid-induced abdominal constriction test. NorBNI was administered 30 min before KT-95 was applied. Ordinate: Counts of writhes, expressed as percentage of control responses. Each column represents the mean±S.E. of 5 experiments. ***p<0.001: Compared to the control. ^###p<0.001: Compared to KT-95.

The potential of KT-95 to produce the side effects of opiates such as physical or psychic dependence liabilities and dysphoria was also examined [85]. As shown in (Fig. 17), no significant physical dependence was observed in KT-95-treated mice when morphine-treated mice showed physical dependence such as significant loss of weight, jumping and tremor when treated with antagonist after chronic administration of each drug.

Also, no significant place-preference was observed for KT-95 ithe conditioned place preference paradigm tesshown in (Fig. 1). The result indicated that KT-95 might be free from psychological psychic dependence liabilities which are are mediated by m- and d-opioid receptors. Dysphoria was known as the characterisitic side effect of k-agonist, and the typical k-selective agonist U-50488 [86] showed a significant place aversion in non-inflamed group using the conditioned place preference paradigm as shown in (Fig. 18B). It is known that rewarding effect of morphine is attenuated in the presence of inflammatory nociception. In the present study, we found that U-50488-induced place aversion is also attenuated in the presence of inflammatory nociception. On the other hand, KT-95 showed no significant place aversion as shown in (Fig. 18A). These results indicated that KT-95 is free from dysphoria, although KT-95 acts as potent k-agonist.

Fig. (17). Time course change in naloxone-precipitated body weight loss. Drug-dependent mice were treated with morphine (8-45 mg/kg, s.c.) or KT-90 (1.6-9 mg/kg, s.c.), KT-95 (0.53-3 mg/kg, s.c.) twice a day for 5 days. Withdrawal was precipitated by naloxone (3 mg/kg, s.c.). Each point represents the mean body weight loss rate ±S.E.M. of 10 animals for 15 min.

Several groups have suggested the presence of k-opioid receptor subtypes based on physiological and binding evidence, i.e. k₁-, k₂- and k₃-opioid receptors