Morphine is widely used to manage pain in patients, although the risk of side effects is significant. The use of biased agonists to the G protein of μ-opioid receptors has been suggested as a potential solution, although Oliceridine and PZM21 have previously failed to demonstrate benefits in clinical studies. An amplification-induced confusion in the process of comparing G protein and beta-arrestin pathways may account for previous biased agonist mis-identification. Here, we have devised a strategy to discover biased agonists with intrinsic efficacy. We computationally simulated 430,000 molecular dockings to the μ-opioid receptor to construct a compound library. Hits were then verified by experiment. Using the verified compounds, we performed simulations to build a second library with a common scaffold, and selected compounds which show biased features to μ and δ-opioid receptors through a cell-based assay. Three compounds (ID110460001, ID110460002, and ID110460003) with a dual biased agonistic effect for μ and δ-opioid receptors were identified. These candidates are full agonists for the μ-opioid receptor, and they show specific binding modes. Based on our findings, we expect our novel compound to act as a biased agonist than conventional drugs such as Oliceridine.
Morphine, the archetypal analgesic opioid, is broadly used to manage pain due to external injury or underlying disease. Morphine and opioid-derived drugs target the opioid receptor family, including μ-(OPRM), δ-(OPRD), and κ-(OPRK) receptors. Opioid receptors are G protein-coupled receptors (GPCR), and are activated through G protein subunits and β-arrestin. Although it demonstrates superior efficacy to other analgesics, morphine can also induce several adverse effects.To overcome the side effects of morphine, agonists that target a subgroup of opioid receptors, or only peripheral opioid receptors, have been developed. In another approach, agonists showing functional selectivity have been exploited.
Although opioid receptors are activated by the G protein and β-arrestin2 pathways, these pathways do not contribute equally to the pain relief effect. As demonstrated by in vivo experiments using β-arrestin2 knockout rodents, the analgesic effect of morphine is predominantly determined by G protein, while the side effects were determined by β-arrestin2. Thus, research has focused on G protein functional selective agonists. Oliceridine, an apparent “biased agonist” targeting OPRM, has already been evaluated in preclinical studies and clinical trials. However, Oliceridine was only approved by the US FDA as a full agonist of OPRM, essentially because it did not show significantly reduced respiratory depression in clinical trials. PZM21, another biased agonist against OPRM, was discovered using a structure-based docking simulation. In cell-based assays, PZM21 demonstrated better potency and efficacy than Oliceridine. Subsequent animal experiments confirmed the analgesic effects of PZM21, and provided evidence for reduced side effects, including respiratory depression. However, subsequent studies have reported that PZM21 is a partial agonist for OPRM, and clinical benefits are limited.
Two different approaches have been used to calculate biased agonist features such as functional selectivity: “Equiactive comparison”, used in the discovery of Oliceridine; and the “Black-Leff operational model with transduction coefficient”, used in PZM21 profiling. In the equiactive comparison method, the respective concentrations of agonists in the two pathways are assumed from each concentration-response curve. Using this approach, a bias factor (including the signaling state of relative activity) is obtained. In the Black-Leff operational model, because agonist binding initiates downstream signaling pathways, the efficacy of the receptor pathway can be defined by receptor occupancy populations. The operational model is described by the τ parameter index, which defines the intrinsic coupling efficiency of the agonist, and the logK A parameter, which denotes the functional equilibrium dissociation constant of the agonist. Using these two parameters, the logarithm of the “transduction coefficient” (τ/K A) parameter can be calculated. The difference in transduction coefficient parameter of each pathway is calculated to determine the relative activities of the agonist in the different signaling pathways. For this procedure, test compound data is compared to full agonist data, a reference criteria of maximal receptor signaling efficacy. By using a reference agonist, data variance from environmental factors is reduced. Hence, the transduction coefficient method was the standard for bias calculation. However, PZM21 did not show promising preclinical study despite using the transduction coefficient methods. Thus, there is a disconnect between cell-based assay results and in vivo systems.
After receptor-ligand binding, physiological signaling pathways are initiated, and this signaling becomes more amplified the further downstream. The amplification is mediated by intracellular molecules such as Ca 2+ and cyclic adenosine monophosphate (cAMP). Comparative activation of most existing biased agonists for opioid receptors was conducted using intracellular cAMP measurements for G protein pathway activation and β-arrestin2 recruitment to the intracellular receptor region for β-arrestin2 pathway activation. However, results from this comparison may overestimate the G protein efficacy due to unbalanced signaling amplification. Furthermore, the cell lines used in the in vitro assay systems are genetically engineered to overexpress target receptors. Therefore, these cell lines have a higher population of receptors compared with physiological systems. Thus, in vitro experiments have a high receptor reserve, and most agonists tend to maximal efficacy (like full agonists) under these conditions. Under conditions where signaling amplification and receptor reserve exist simultaneously, partial agonists may behave as biased agonists that reach maximal effects in the G protein pathway. By utilizing a lower receptor population and comparing the G protein pathway and β-arrestin pathway at the same signaling distance, Alexander Gillis et al revealed that Oliceridine, PZM21, and SR-17018 are partial agonists. To discover true biased agonists, it is necessary to confirm that ligands in different signaling cascades within the in vitro system are unaffected by the abovementioned limitations.
To discover a new scaffold for opioid receptor biased agonists, we applied a virtual screening (VS) methodology. This is a widely used technique for accelerating the early drug discovery process and identifying novel scaffolds. In many studies, VS is significantly faster and equally or more successful than high-throughput screening (HTS). Moreover, VS costs around 10-fold less than HTS and takes approximately half the time.
Here, we carried out a sequential ligand-based and structure-based virtual screening approach. The Korea Chemical Bank (KCB) was preprocessed using Pipeline Pilot. The virtual hits were then screened to discover OPRM, OPRD dual biased agonists for the G protein pathway. To minimize uncertainty due to the limitations of signal amplification and receptor reserve, we used cell lines in which receptor population can be tuned by irreversible antagonists. Thus, we obtained compounds with a bias effect emanating from high efficacy. The binding sites of our compounds were subsequently evaluated to improve the efficacy profiles.
An efficient strategy for screening chemical candidates for OPRM, OPRD dual biased agonists was devised. We adopted a virtual screening method using the KCB database and validated the results using the cell-based assay. Using chemistry aided drug discovery (CADD) and the cell-based assay we were able to define a common chemical structure. This was then applied to the docking simulation. We identified three candidates with different mechanisms of action to Oliceridine and PZM21. The bases of these differences were then explained through simulations of molecular models.
We applied pharmacophore-based virtual screening using the reported OPRM biased agonists. The lowest energy conformations of Oliceridine, PZM21, and SR-1708 were geometrically refined using Discovery studio. These structures were used to build a three-dimensional common feature model within Phase. The resulting four-feature pharmacophore comprised one hydrophobic group, two aromatic rings, and one positive ionic feature. Figure 2 shows the three reported biased agonists superimposed on these features and especially, their protonated amine group was mapped the positive ionic feature. The pharmacophore model was then applied as a filter to virtually screen the KCB library. The top 10,000 candidates (by score) were selected for the docking study. To narrow the focus, molecular docking was performed upon active state OPRM using Glide SP. The reported biased agonist binding modes demonstrated docking scores less than −5.9. This value was set as the cutoff to be used for filtering in our structure-based virtual screening (SBVS). In total, 5,544 unique compounds remained following the pharmacophore-based and docking-based filtering stages. To exclude false-positive compounds, we performed clustering with ECFP4 and an additional visual inspection within the binding site of OPRM. Finally, we identified 1,986 potential compounds for biological testing.
The flow chart for the discovery of OPRM, OPRD dual biased agonists is described in Figure 3A. The previously identified set of small molecules was screened through a cAMP assay to evaluate their activation properties. The first screening was accomplished at a single concentration (10±1 μM). Because virtual screening was focused on OPRM, the primary target in this screening was OPRM. As described in Figure 3B, 55 compounds showed at least 50% activation when compared with the E max of DAMGO. These compounds were screened again in triplicate at the same concentration, and the results were reproducible for 32 compounds.
