The new era of opioid studies began with the isolation of endogenous ligands in the brain for analgesic receptors, which led in a short time to structure determination of enkephalins, dynorphins, and endomorphins, as well as the opioid receptors µ (MOR), δ (DOR), and k (KOR). The putative physiological roles of opioid peptides and their associated receptors have been under intensive investigation for many years and, in the beginning, the cloning of opioid receptors has provided direct structural evidence of “multiple opioid receptors” [1
]. Although the cloned opiate receptors represent powerful tools for the physiological and pharmacological evaluation of their roles in normal and painful states, very few new insights have been made to discover ligands capable of maximizing efficacy, especially in vivo, against neuropathic pain. Opiate therapy is one of the most commonly prescribed treatments for chronic neuropathic pain; however, opioids do not address the mechanisms of neuropathic pain and often have limited efficacy against this type of pain [3
]. Opioids are plagued by analgesic tolerance, addiction, medication overuse, hypersensitivity, and other physical side effects. In addition to social and legal issues associated with their use for non-medical and recreational purposes, several adverse effects (e.g., dysphoria, constipation, respiratory depression, nausea, vomiting, etc.), hinder their clinical usefulness and justify the discovery of safer opioid therapeutics and/or non-addictive medications [4
G protein-coupled receptors (GPCRs) are characterized by a basal signaling activity in the absence of agonists in over-expressed receptor cell lines and at physiological receptor levels [5
]. Compounds acting as antagonists can block agonist-mediated receptor activation but may exert different effects at basally active receptors, thus suppressing basal signaling activity as inverse agonists (antagonists with negative intrinsic activity) or neutral antagonists. However, there are also antagonists that don’t affect basal signaling. Dopamine D1
, and D3
-adrenergic receptors, adenosine receptors, serotonin (5HT2A
) receptors, and δ-opioid receptors (DOR) are able to bind inverse agonists and neutral antagonists [6
]. Intrinsic efficacy depends on the context of the cells, receptors, and experimental conditions. Information about the physiological role of basal signaling activity of receptors and the therapeutic importance of inverse agonists or neutral antagonists are still scarce and uncertain. For instance, diverse intrinsic activities of β-adrenergic receptor blockers influence cardiac contractility and possibly the therapeutic outcome of patients with cardiac failure [7
]; moreover, the high constitutive activity of native H3
receptors regulates histamine neurons in the brain [8
]. Considering the therapeutic perspective, MOR agonists such as morphine, hydromorphone, oxycodone, and fentanyl discovered through traditional approaches, such as natural product isolation (e.g., morphine), semi-synthetic natural product derivatives (e.g., oxycodone, hydromorphone), and synthetic manipulation of natural product scaffolds (e.g., fentanyl), have been used for decades in antinociceptive therapies (Figure 1
These are potent μ-opioid agonists that exert addiction liability and numerous side effects, such as euphoria, addiction, respiratory depression, and gastrointestinal adverse reactions; therefore; circumventing these drawbacks is of extensive importance. A large amount of literature documented novel approaches to disconnecting the analgesic efficacy of μ-opioid agonists from morphine-like side effects, for example biasing the GPCRs over β-arrestin2 recruitment (TRV130, PZM21, HS665) or designing positive allosteric modulators of the MOR (BMS-986122), MOR inverse agonists/KOR antagonists [9
], and multiple agonists of opioid receptors subtypes (SNC80, DPI-125) [10
]. Different peptides and peptidomimetics have been reported in the literature to exert a strong antinociceptive effect on opioid receptors without involving the β-arrestin protein, which is thought to be responsible for respiratory depression and constipation associated with the use of naturally occurring and synthetic opioids. The natural peptides rubiscolin-5 and -6 are excellent examples of molecules that activate G protein signaling pathways at the δ-opioid receptors, with minimal β-arrestin recruitment [11
]. Thus, nowadays the design of small molecules with a peptidic structure could be a valuable strategy to achieve selective activation of a desired opioid signaling pathway, reducing the incidence of unwanted side effects.
Receptor-based drug design is a precious tool to furnish a detailed molecular-level understanding of the interactions between opioids and their receptors. High-resolution crystal structures of all four opioid receptor subtypes, i.e., the MOR, DOR, KOR, and nociceptin/orphanin FQ receptors, allowed for the development of virtual screening (VS) campaigns that led to the discovery of novel chemotypes targeting these receptors. Structure-based drug design, now an indispensable component of drug discovery, principally employs methods of receptor-based virtual screening and molecular docking for binding pose prediction [12
], hit identification, and lead optimization [13
]. As part of our ongoing effort to discover new MOR modulators with novel structures, the study herein described focused on the crystal structure of the MOR inactive-state for the discovery of novel MOR ligands using VS. In particular, we performed a VS study employing an in-house built library of tetrapeptide compounds with the aim to identify novel peptides acting as MOR modulators (Figure 2
). To the best of our knowledge, the study herein reported represents one of the first examples of a VS campaign focused on the identification of new peptide ligands as opioid modulators.
5. Results and Discussion
The µ-opioid receptor (MOR) is the main mediator of narcotic analgesia and addiction, exhibiting a basal signaling activity in SH-SY5Y cells and transfected HEK293 cells via Gαi/Gαo proteins [42
]. With the aim to identify novel tetrapeptides able to bind MOR, we performed a VS study using a virtual combinatorial library of tetrapeptides. By using an in-house python program, we generated the FASTA sequences of all the possible tetrapeptides obtained by the combination of the 20 standard amino acids, which were then converted to the corresponding 3D structures (see Material and Methods for details). In this way, we obtained a virtual library of about 200,000 tetrapeptides to be used for VS studies. A receptor-based pharmacophore model was then developed on the basis of the X-ray structure of MOR bound to the morphinan antagonist β-funaltrexamine (β-FNA) [16
]. The ligand co-crystallized with MOR is an irreversible antagonist that covalently binds to a lysine residue of the receptor (K233); however, since we aimed at identifying novel peptides with MOR inhibitory activity (i.e., acting as antagonists/inverse agonists), the choice of this X-ray structure, which is the only crystal structure of MOR in inactive conformation, was necessary for our VS protocol. Nevertheless, the pharmacophore model was built taking into account the ligand-protein interactions established by the core scaffold of the antagonist. As shown in Figure 3
A, the morphinan core of the ligand forms a salt bridge interaction with D147 through its tertiary amine group, while an H-bond between Y148 and the endocyclic oxygen of the ligand is observed. Moreover, the inhibitor forms a water-bridged interaction with H297 allowed by two structural water molecules that participate in an H-bond network between the side chain of H297 and the phenolic OH group of the ligand. Finally, multiple hydrophobic interactions can be observed between the morphinan core of the inhibitor and the surrounding protein residues. In particular, the ligand cyclopropyl group is placed in a lipophilic pocket delimited by W293, I296, I322, and Y326, thus forming hydrophobic interactions with these residues, while the aromatic ring is sandwiched between M151 and V300, taking additional lipophilic contacts with I296 and H297. Based on these considerations, a receptor-based pharmacophore model comprising five different features was generated. The model included: a) a positively charged feature representing the salt bridge interaction with D147, b) two hydrophobic features representing the lipophilic interactions formed by the cyclopropyl group and the aromatic ring of the ligand, c) two H-bond acceptor features representing the direct and the water-mediated H-bond interactions formed with Y148 and H297, respectively (Figure 3
B). It is worth specifying that this last interaction was modeled through a feature pointing at the region of space among the ligand OH group and the two water molecules involved in the H-bond network, in order to find ligands that could replace the structural waters and directly interact with H297. In addition, the model was refined by adding excluded volume spheres mimicking the steric hindrance represented by the MOR binding site (see Materials and Methods for details).
The receptor-based pharmacophore model was used to screen the virtual library of about 200,000 tetrapeptides in order to identify all peptides able to form the key interactions with MOR. For this purpose, only the tetrapeptides respecting the volume constrains, matching the positively charged feature of the model and at least two additional features were retained by the filter. By using this strategy, only 28,070 peptide ligands were selected and subjected to further analyses based on docking evaluations. Interestingly, we found that 5170 out of the 28,070 peptides retrieved by the pharmacophore filter showed a tyrosine residue in the first position. Therefore, as expected, the filter enriched in peptides containing an amino-terminal tyrosine (as in the natural opioid motif Tyr–Gly–Gly–Phe), the subset of compounds to be considered in the docking step.
To the best of our knowledge, no examples of VS studies focused on peptide libraries have been reported to date; as a consequence, no hints about the reliability of the commonly used docking software in predicting the binding mode of small peptides were available in the literature. For this reason, we performed a docking reliability analysis aimed at assessing the performance of 10 different docking procedures in reproducing the experimental binding mode of 12 small peptides for which ligand-protein co-crystal structures were available in the Protein Data Bank. For each peptide docked into its corresponding receptor using the 10 different docking procedures, the root-mean square deviation (RMSD) of the peptide backbone calculated between the binding modes predicted by docking and the corresponding experimental pose were used to evaluate the reliability of the docking methods. As reported in Figure 4
, Glide with the standard precision (SP) method and Autodock Vina showed the best results in terms of average RMSD (aRMSD) obtained for the whole dataset of peptides, with values below or equal to 3.5 Å. Although it would have been desirable to obtain RMSD values closer to 2.0 Å, these results were expected, since the docking software was developed and calibrated for the docking of small-molecules with non-peptide structures. Nevertheless, since Glide SP and Autodock Vina performed better, on average, than all other software and thus represented the best docking methods among those tested for the docking of small peptides, these two procedures were selected to be applied in our VS study.
The 28,070 peptide ligands previously selected through the pharmacophore screening, were thus initially docked into MOR binding site, by using Glide SP and ranked according to the docking score associated to their predicted binding mode. Since in our previous studies Glide SP scoring function showed to be particularly reliable [43
], and all peptides accurately docked by Glide SP in the reliability analysis showed docking scores lower than −8.0 kcal/mol, all screening compounds for which a docking score higher than –8.0 kcal/mol was obtained were discarded, while the remaining top-scored 913 peptides were subjected to a qualitative post-docking filter. Precisely, the selected compounds were superimposed with the receptor-based pharmacophore model and only those still matching the positively charged feature and at least two additional features of the model in their predicted binding mode were further considered. Based on this qualitative filter, only 146 tertrapeptides were retained and subjected to additional docking studies using Autodock Vina. The docked compounds were directly subjected to the pharmacophore-based post-docking filter and only 15 tetrapetides showed to maintain the ligand-protein interactions represented by the model. These compounds were thus further analyzed through molecular dynamics (MD) simulations studies, in order to evaluate the stability of their binding disposition into MOR. A total of 12.5 ns of MD simulations with explicit solvent were performed for each ligand and the persistence of the key pharmacophoric interactions during the simulations was analyzed. The peptide ligands that couldn’t maintain the salt bridge interaction with D147 and at least one of the H-bonds with H297 and Y148 for at least 80% of the whole MD simulation were discarded. As a result, only 3 tetrapeptides were considered as potential MOR ligands and were then synthesized to be tested for their biological activity: peptide 1 (Tyr–Lys–Arg–Cys), peptide 2 (Tyr–Trp–Trp–Trp) and peptide 3 (Tyr–Trp–Tyr-Trp). Interestingly, the three selected compounds presented a tyrosine residue in the first position as observed in the natural opioid motif (Tyr–Gly–Gly–Phe). For initial screening purposes, the compounds were tested in two high concentration points, 3 and 10 µM. However, the peptides did not show a strong affinity for MOR nor for DOR, since none of the test compounds inhibited radioligand total specific binding lower than 50% even at the highest (10 µM) concentrations (Figure 5
A and 5B). For comparison, DAMGO and IleDelt II, MOR and DOR selective agonist compounds, respectively, inhibited the specific binding of their radiolabeled homologs to non-specific binding level (0%) (Figure 5
The G-protein activity of the test compounds was firstly investigated at 10 µM concentrations. The results were in agreement with the affinity data, with the only exception of peptide 1, since peptide 2 and 3 did not alter significantly the basal activity (100%) of the monitored G-protein in 10 µM concentrations (Figure 5
C). For comparison, DAMGO and IleDelt II selective ligands for MOR and DOR respectively, significantly increased the specific binding of the [35
S]GTPγS compared to the basal level, which demonstrates their well-documented agonist activity. However, peptide 1 displayed an appreciable inverse agonist effect, indicated by a reduced (~20%) G-protein basal activity (Figure 5
C). Since in the binding affinity assay peptide 1 displayed inhibition of [3
H]DAMGO specific binding and not against [3
H]IleDelt II, we investigated the MOR specificity of the inverse agonist effect. When DAMGO was added to peptide 1 at 10 µM concentration, the inverse agonist effect was reversed back to basal activity level, indicating a MOR mediated effect. Interestingly the MOR selective antagonist cyprodime did not cause any significant change (Figure 5
D), indicating that our receptor-based VS protocol yielded the discovery of a novel peptide acting as an inverse agonist at the MOR receptor.
shows the predicted binding mode of peptide 1 within MOR binding site refined through MD simulations. As observed in many endogenous opioid ligands, the tyrosine residue of the peptide forms key interactions with the protein and efficiently mimics the morphinan scaffold of the crystallographic antagonist β-FNA. Precisely, the protonated amino group of the ligand’s tyrosine forms a stable salt bridge with the side chain of D147, which is maintained for the whole MD simulation; the aromatic ring of the residue is placed between M151 and V300, forming hydrophobic interactions with these residues as well as with W293, I296, and H297; finally, the phenolic group of the ligand’s tyrosine succeeds in establishing a direct H-bond with H297, thus replacing the two structural water molecules observed in the reference X-ray structure of MOR. Interestingly, at the beginning of the MD simulation, the imidazole ring of H297 undergoes a 180 degrees rotation, which allows the residue to form an H-bond with the backbone oxygen of W293. Nevertheless, the H-bond between H297 and the ligand is maintained for almost the whole simulation. The cysteine residue of the ligand forms additional interactions with the protein. In particular, the terminal carboxylic group of the peptide showed a stable ionic interaction with the side chain of K223 (the alkylated residue in the reference X-ray complex) and an H-bond with the phenol group of Y148. Finally, the lysine residue of the peptide did not show particularly relevant interactions with the receptor, while the guanidine moiety of the ligand’s arginine residue formed strong a π–π stacking with H319 and Y128. Interestingly, peptides 2 and 3, which resulted to be inactive against MOR, were able to form stable H-bond interactions with both H297 and Y148, as well as the salt bridge with D147, in their predicted binding mode. Moreover, both peptides formed an additional H-bond with W318 not shown by peptide 1 (Figure S1
). However, the inactive peptides lacked the stable salt bridge with K233 and the strong π–π stacking with H319 and Y128 shown by peptide 1. Moreover, peptides 2 and 3 were less buried within the protein binding site and more protruded toward the extracellular side, with a widely solvent-exposed tryptophan residue. These binding mode differences with respect to peptide 1 might explain the inactivity of peptides 2 and 3.