Rational Design of PZM21: A Review on Novel OR Agonist Achieving Analgesia Without Traditional Opioid Side Effects

Since the 19th century, attempts to create safer and more efficient painkillers have been spurred by the addiction to opioids and their harmful side effects, such as respiratory depression. Although they offer more reliable pain relief than raw opium, natural substances like morphine and codeine, as well as the semi-synthetic opioid heroin, nonetheless have comparable hazards^1,2. The identification of opioid receptor subtypes, including μ, δ, κ, and nociception, raised hopes that focusing on particular receptors could lessen the negative effects of conventional morphinan-based opioids. Although powerful synthetic opioid agonists such as methadone and fentanyl have been introduced, and endogenous opioid peptides³ have been identified, creating pain relievers without the undesirable effects of traditional opioids has proven challenging. Recent research indicates that opioid-induced pain relief is linked to μ-opioid receptor (μOR) signaling through the G protein Gi. In contrast, many side effects, including respiratory depression and constipation, are believed to result from β-arrestin pathway signaling that occurs following μOR activation (Fig. 1a)^4-6. Researchers are actively exploring μ-opioid receptor (μOR) agonists that selectively activate the Gi signaling pathway. These agonists are being investigated both as potential therapeutic agents and as tools to better understand μOR signaling mechanisms. Recent advancements have demonstrated the practicality and possible clinical benefits of developing such biased μOR agonists^7,8. The identification of crystal structures for the μ, δ, κ, and nociceptin opioid receptors^9-12 has opened new possibilities for developing μOR agonists through structure-based methods. Recent research efforts have leveraged the crystal structures of other Family A G protein-coupled receptors (GPCRs) to virtually screen extensive molecular libraries, successfully finding ligands with novel scaffolds and potent activity in the nanomolar range^13-17. With this approach in mind, we focused on the μOR receptor for structure-based docking, aiming to identify ligands with distinct chemical structures. We hypothesized that these new chemotypes could exhibit unique signaling behaviors and biological effects, similar to outcomes seen in other structure-based discovery efforts^18,19.

2.Structure-based docking to the μOR

In a structure-based approach to identify novel μOR agonists, researchers conducted an extensive virtual screening by docking over 3 million commercially available lead-like compounds²⁰ against the orthosteric pocket of the inactive μOR. Their strategy involved prioritizing ligands that showed interactions with key affinity-determining residues and potential specificity residues that vary among the four opioid receptor subtypes (Fig. 1b, d). This targeted approach aimed to enhance receptor selectivity and improve therapeutic outcomes. A recent study employed a structure-based docking approach using DOCK3.6 to evaluate an extensive library of compounds for their potential as μOR agonists^21-23. On average, 1.3 million configurations per compound were assessed using a physics-based energy function to predict receptor complementarity. As is standard in docking and screening protocols, the highest-ranking molecules underwent further manual inspection to assess key features not fully captured by the scoring function. This evaluation emphasized molecular novelty, interactions with critical polar residues such as Asp147^3.32 (Ballesteros-Weinstein numbering²⁴), and minimized conformational strain. Out of over 3 million docked compounds, 23 high-ranking molecules—spanning ranks 237 to 2,095—were selected for experimental validation (Fig. 1e). These candidates demonstrated Extended Connectivity Fingerprint 4 (ECFP4)-based Tanimoto coefficients (Tc) ranging between 0.28 and 0.31, indicating the presence of novel scaffolds. Notably, seven compounds exhibited μOR binding affinities (Ki) in the range of 2.3 μM to 14 μM (Data Table 1, Extended Data Fig. 1), underscoring the potential of this structure-based docking strategy in identifying promising new chemotypes for opioid receptor modulation²⁶. The newly designed ligands are expected to interact with the μ-opioid receptor (μOR) through novel binding mechanisms (Fig. 1f and Extended Data Fig. 1). Typically, most opioid ligands rely on a cationic amine group to form an ionic bond with Asp147^3.32. This well-known interaction has been identified in the structural analysis of μOR, δOR, κOR, and the nociceptin receptor when bound to ligands with various scaffolds^9–12,28. As expected, the docked ligands replicated this interaction. There is limited evidence for the formation of an additional hydrogen bond with the anchor aspartate, which is often facilitated by a urea amide in the docking models. In some of the newly designed ligands, the urea carbonyl is predicted to form a hydrogen bond with Tyr148^3.33, while the remaining ligands frequently occupy regions not previously explored by morphinan compounds (Extended Data Fig. 1). Notably, the double hydrogen bond coordination involving Asp147^3.32, as modeled in the docking poses, appears to be a novel finding for opioid ligands. Out of the 5,215 opioid ligands recorded in ChEMBL16, only 50 are reported to contain a urea group. Although the initial docking hits had unique structures, their binding affinities were relatively weak. To improve both binding strength and selectivity, researchers docked 500 analogues of compounds 4, 5, and 7. These analogues retained essential recognition features but included packing substituents or extended towards the extracellular side of the receptor, where opioid receptors tend to vary more. Out of the 15 highest-scoring analogues that underwent testing, seven showed Ki values ranging from 42 nM to 4.7 μM (Extended Data Table 2). Notably, several analogues (compounds 12–15) demonstrated selectivity for the μOR over the κOR. Further investigation into the more potent analogues revealed that compounds 8 and 12–14 effectively activated Gi/o signaling, despite being designed using the inactive state of the μOR. This result aligns with previous findings in a κOR²⁸ docking study, where a similar trend of agonist enrichment was observed²⁹. Among these, compound 12 stood out as the most potent, significantly activating Gi/o with minimal β-arrestin-2 recruitment (Fig. 2a–d).

3.Synthetic optimization guided by structure

To enhance the properties of compound 12, researchers synthesized stereo chemically pure isomers and added a phenolic hydroxyl group (Fig. 3a). The (S,S) isomer of compound 12 demonstrated improved binding affinity (Ki) of 4.8 nM and a signalling EC50 of 65 nM, making it the most potent and effective Gi/o signalling agonist among the four tested isomers (Fig. 3e). The phenolic hydroxyl group, incorporated to create compound (S,S)-21, was intended to form a water-mediated hydrogen bond with His297^6.52 — a key interaction seen in the μOR-β-FNA complex (Fig. 3b) and other δOR²⁸ and κOR11 structures. This hydroxyl group was well-integrated in the docked μOR-12 complex, enhancing its predicted docking energy (Fig. 3c). Compound (S,S)-21 achieved an EC50 of 4.6 nM in a Gi/o activation assay with 76% efficacy (Fig. 3f) and displayed a Ki of 1.1 nM in radioligand binding assays (Extended Data Table 3), showing a 40-fold improvement compared to compound 12. The remaining three stereoisomers of (S,S)-21 were significantly less potent or effective (Extended Data Fig. 2a, b), indicating a precise stereochemical requirement for optimal potency and efficacy. This aligns with the predicted docked poses of (S,S)-21 in both inactive and active μOR structures²⁹ (Fig. 3c, Extended Data Fig. 2c, d). Compound (S,S)-21 is referred to as PZM21 in subsequent discussions. Since PZM21 was initially identified based on the inactive structure of the μ-opioid receptor (μOR), its binding pose in the active μOR structure remained uncertain. To better understand this interaction, researchers performed detailed docking studies and molecular dynamics simulations. This model was further validated by synthesizing molecules designed to alter or leverage specific modeled interactions. For instance, compound PZM28 was created by amidating the tertiary amine, neutralizing its charge. This modification resulted in a 1,000-fold reduction in potency, reinforcing the significance of the ionic interaction between PZM21’s tertiary amine and Asp147³^.³². Similarly, compound PZM27 was synthesized with additional steric bulk on the tertiary amine to disrupt hydrophobic interactions involving the N-methyl group, Met151³^.³?, and Trp293?^.??. This alteration led to a 30-fold decrease in potency and reduced efficacy. Compounds PZM23, PZM24, and PZM25 were designed to interfere with hydrogen bonds between the urea group and residues such as Asp147³^.³², Tyr326?^.?³, and Gln124²^.??. These modifications resulted in potency losses ranging from 30- to 230-fold, even though their solvation penalties were reduced. Notably, key ionic and hydrogen bonding interactions remained stable throughout 3 μs of molecular dynamics simulations of PZM21 bound to the active μOR. Additionally, the phenolic hydroxyl maintained bridging water interactions with His297?^.?², further supporting the modeled pose. The thiophene group in PZM21, designed to fit within the open specificity region of the μOR, could be replaced with a larger benzothiophene without any reduction in potency. This thiophene’s interactions with residues unique to opioid receptor subtypes may contribute to PZM21’s selectivity. Molecular simulations also predicted that the PZM21 thiophene approaches Asn127²^.?³ within 6 Å in the active μOR. To confirm this, researchers synthesized PZM29 — an irreversible PZM21 variant — which successfully formed a covalent bond with a mutant μOR containing an N127C substitution. PZM29 bound irreversibly to the mutant receptor but not the wild-type, while retaining its agonist efficacy. These findings strongly support the modeled orientation of PZM21 in the μOR’s orthosteric site.

4.PZM21 is a selective Gi-biased μOR agonist

PZM21 has shown no measurable agonist activity at the κ opioid receptor (κOR) or the nociceptin receptor. Instead, it acts as an 18 nM κOR antagonist and is 500 times weaker as a δ opioid receptor (δOR) agonist, establishing it as a selective μ opioid receptor (μOR) agonist. To assess its specificity, PZM21 underwent screening against 316 additional G-protein-coupled receptors (GPCRs)³⁰. While some activity was noted at 10 μM in certain peptide and protein receptors, no potent effects were confirmed through full dose–response experiments. These findings highlight PZM21’s strong selectivity as a μOR agonist. Further tests examined PZM21's effects on the hERG ion channel and key neurotransmitter transporters, including dopamine, norepinephrine, and serotonin systems. PZM21’s IC50 value at the hERG channel ranged between 2 and 4 μM, which is 500- to 1,000-fold weaker than its μOR agonist potency. Its inhibition of neurotransmitter transporters was even weaker, with IC50 values between 7.8 and 34 μM. Collectively, these results demonstrate PZM21’s potency, selectivity, and effectiveness as a μ opioid agonist. One primary objective of this study was to identify novel chemotypes capable of biased signaling, potentially improving in vivo safety profiles compared to traditional opioids. PZM21’s signaling through the μOR was predominantly mediated by the Gi/o protein pathway, as shown by the elimination of its effect on cAMP levels following pertussis toxin treatment. Additionally, no activity was detected in calcium release assays. At maximal concentrations, PZM21 failed to stimulate β-arrestin-2 recruitment in the PathHunter assay and showed minimal μOR internalization when compared to DAMGO and morphine. The β-arrestin-2 response was too weak to allow a formal calculation of signaling bias. Even with GRK2 co-transfection, PZM21 displayed weak arrestin recruitment compared to DAMGO and morphine³¹. Interestingly, its signaling bias closely resembled that of TRV130, a Gi-biased agonist undergoing Phase III clinical trials, and surpassed the G-protein bias of herkinorin^32,33. Notably, PZM21’s distinct signalling profile includes its lack of κOR agonist activity, in contrast to TRV130, which activates κOR with potency similar to morphine. Despite their comparable signalling bias, molecular modelling suggests that PZM21 and TRV130 interact with the μOR in different ways.

5.Reduced adverse effects while using analgesia

PZM21, a μOR agonist, demonstrated dose-dependent pain relief in a mouse hotplate test, achieving 87% of the maximum possible effect (% MPE) within 15 minutes at the highest dose administered. In comparison, morphine reached 92% after 30 minutes. Interestingly, PZM21 showed no analgesic effect in the tail-flick test, which is highly unusual among opioid analgesics. The hotplate test evaluates pain relief involving both central nervous system (CNS) pathways and spinal nociceptive circuits, whereas the tail-flick test primarily assesses spinal reflexive responses³⁴. Further analysis of the hotplate data revealed that PZM21 targets only the affective (CNS-mediated) aspect of pain, unlike morphine, which influences both affective and reflexive pain responses. This unique effect of PZM21 has only been seen previously in selective chemo genetic activation³⁵ or toxin-induced inactivation of CNS neurons in rodents³⁶. Additionally, PZM21 showed activity in the formalin injection nociception test, likely due to its action on supraspinal descending inhibitory circuits³⁷ (Fig. 4d). The analgesic effect of PZM21 is confirmed to result from μOR activation, as genetic knockout of the μOR eliminated the observed pain relief in the hotplate test. PZM21 is also metabolized slowly by mouse liver microsomes, with only 8% undergoing metabolism within one hour. Analysis of the metabolite pool showed no evidence of a metabolite with stronger μOR activation, confirming that the initial dose of PZM21 drives its analgesic effect. Based on prior research involving arrestin knockout mice and biased opioid compounds^4-7, PZM21 was expected to offer prolonged pain relief with reduced respiratory depression and constipation — both common side effects of traditional opioids. PZM21’s analgesic effects last up to 180 minutes, significantly longer than morphine’s maximal dose and the biased agonist TRV130 (Fig. 4a). While PZM21 reduces defecation, its constipating effect is much lower than that of morphine (Fig. 4f). Respiratory depression studies revealed that PZM21, at an equi-analgesic dose of 40 mg/kg, produced minimal respiratory depression similar to the vehicle control (Fig. 4g). In contrast, TRV130 (1.2 mg/kg) and morphine (10 mg/kg) both significantly reduced respiration within 15 minutes. The rapid respiratory depression caused by morphine and TRV130 may involve G-protein-coupled inwardly rectifying potassium channels (GIRKs). However, opioids can still suppress respiration in GIRK-deficient mice³⁸, supporting a G-protein-independent mechanism linked to β-arrestin-2 signalling. PZM21’s minimal respiratory depression at later time points, despite its effective analgesia, contrasts with morphine’s prolonged respiratory suppression even after its pain-relieving effects diminish. These findings indicate that PZM21 likely triggers minimal β-arrestin-2 signalling in vivo, consistent with its biased signalling profile (biased signalling, Fig. 1a). One significant concern with current opioid painkillers is their potential to cause reinforcement and addiction, which is believed to be linked, at least partially, to the activation of dopaminergic reward pathways³⁹. In mice, a key indicator of this activation is an acute increase in movement, signifying mesolimbic dopamine system stimulation⁴⁰. In an open-field test, morphine was shown to trigger hyperactive behavior in mice (Fig. 4h). However, PZM21, administered at a nearly equivalent analgesic dose, did not produce any noticeable increase in movement compared to the control vehicle. The reduced movement observed with PZM21 was not due to catalepsy, Supporting the idea of reduced reward circuit activation, PZM21 administration also failed to produce a conditioned place preference response (Fig. 4i), a behavior commonly induced by morphine and other opioids⁴¹. While TRV130 showed a slight tendency to induce place preference, its effect was not statistically significant compared to the control vehicle. This absence of conditioned place preference in both biased agonists may suggest that G-protein bias contributes to the lack of opioid-induced reinforcing behaviour. The variation in conditioned place preference between morphine and PZM21 is not merely a result of differences in their ability to penetrate the central nervous system (CNS) indicate that a substantial amount of PZM21 successfully crosses the blood-brain barrier. There are several important considerations to note. While structure-based drug discovery successfully identified novel scaffolds and facilitated efficient optimization, some characteristics of PZM21 may have been coincidental. The compound’s biased signaling via G protein and arrestin pathways results from stabilizing conformations located over 30 Å away from the orthosteric binding site of PZM21. Instead of intentionally selecting molecules that stabilize these conformations, the approach focused on chemical novelty to introduce unique biological effects. Achieving receptor subtype selectivity was accomplished by choosing molecules that extended into variable regions of the receptor, although this strategy may not consistently yield success. Certain aspects of the pharmacology discussed here are still preliminary, including studies on metabolic stability and pharmacokinetics. It remains uncertain whether PZM21’s remarkable in vivo activity is due to its biased and selective agonism or some other property arising from its novel chemotype. Lastly, while identifying agonists through docking to an inactive receptor structure is not always reliable, there are established precedents for successfully applying this method to opioid receptors^{13, 16, 42, 43}, with some previous successes specifically in this area ²³.

6.METHODS

6.1 Chemicals, reagents, and cell lines

The chemicals and reagents utilized in this investigation were either synthesized in accordance with the Supplementary Information or acquired from commercial sources (Sigma, Tocris, Fisher Scientific, and ZINC database vendors). The ATCC provided the HEK293 cells (ATCC CRL-1573; 60113019; certified mycoplasma free and authentic by ATCC) and HEK293-T cells (HEK293T; ATCC CRL-11268; 59587035; certified mycoplasma free and authentic by ATCC), which have been thoroughly validated for signaling investigations. Additionally, cells were verified through the examination of short tandem repeat (STR) DNA profiles, which displayed a 100% match with the ATCC's STR database. The human μOR-expressing U2OS cells were acquired from DiscoverX as cryopreserved supplies and were not subjected to additional authentication.

6.2 Molecular docking and analogue selection

The inactive-state μ-opioid receptor structure (PDB: 4DKL) served as the starting point for receptor preparation using DOCK Blaster. A total of 45 matching spheres were employed, based on a modified form of the crystallized ligand. For sphere generation, the covalent bond and linker region of the antagonist β-funaltrexamine were removed. Ligand sampling parameters were configured with a bin size of 0.4 Å, bin size overlap of 0.1 Å, and distance tolerances of 1.5 Å for both matching spheres and docked molecules. Ligand poses were evaluated by combining receptor-ligand electrostatics and van der Waals interaction energy, with adjustments for ligand desolvation. Receptor atom partial charges followed the united atom AMBER force field, except for Lys233 and Tyr326, where the dipole moment was increased as previously reported. Over 3 million commercially available compounds from the ZINC20 lead-like set were docked into the receptor using DOCK3.621. Among the top-ranking 0.08% of molecules, 23 were chosen for experimental evaluation in the primary screen. A publicly accessible resource is available for conducting these docking studies via DOCK Blaster. For the secondary screen, analogues of the top three hits from the primary screen (compounds 4, 5, and 7) with a similarity score above 0.7, as defined by the ZINC search tool, were identified in the ZINC database. Substructure searches were also carried out using the scaffolds of these three compounds. This process identified 500 purchasable compounds, which were then docked following the same procedure as in the primary screen. Selected analogues were manually reviewed for interactions before advancing to further experimental testing.

6.3 Gi/o induced cAMP inhibition

To assess μOR Gi/o-mediated cAMP inhibition, HEK-293T cells were co-transfected using calcium phosphate in equal parts with human μOR and a split-luciferase-based cAMP biosensor (pGloSensor™-22F; Promega). In experiments involving GRK2 co-expression, cells were transfected with 1 μg of GRK2 per 15-cm dish. After a minimum of 24 hours, the transfected cells were washed with phosphate-buffered saline (PBS) and dissociated using trypsin. Following centrifugation, the cells were resuspended in plating media containing 1% dialyzed FBS in DMEM, then plated at a density of 15,000–20,000 cells per 40 μl per well in poly-lysine coated 384-well white clear bottom cell culture plates. The plates were incubated overnight at 37 °C with 5% CO2. For experiments involving pertussis-toxin (PTX) Gαi/o inactivation, cells were plated with PTX at a final concentration of 100 ng/ml. The next day, drug dilutions were prepared in a fresh assay buffer containing 20 mM HEPES, 1× HBSS, 0.1% bovine serum albumin (BSA), and 0.01% ascorbic acid at pH 7.4, with drugs diluted to three times their intended concentration. Plates were emptied, and 20 μl of drug buffer (20 mM HEPES, 1× HBSS, pH 7.4) was added to each well. FLIPR was used to add 10 μl of the drug solution per well, bringing the total volume to 30 μl. Plates were incubated in the dark at room temperature for exactly 15 minutes. To activate endogenous cAMP via β adrenergic-Gs stimulation, 10 μl of 4× isoproterenol (200 nM final concentration) in drug buffer with GloSensor assay substrate was added per well. The cells were incubated in the dark at room temperature for another 15 minutes before measuring luminescence intensity using a Wallac TriLux microbeta luminescence counter (Perkin Elmer). Data were normalized to DAMGO-induced cAMP inhibition and analyzed via nonlinear regression in GraphPad Prism 6.0. The functional activity of PZM21-29 for SAR studies was assessed through a BRET-based cAMP accumulation assay. HEK-293T cells were transiently co-transfected with pcDNA3L-His-CAMYEL42 (purchased from ATCC via LCG Standards, Wesel, Germany) and human μOR at a cDNA ratio of 2:2 using Mirus TransIT-293 transfection reagent. After 24 hours, the transfected cells were seeded into white half-area 96-well plates at a density of 20 × 10? cells per well and incubated overnight. On the following day, the phenol-red-free medium was replaced with PBS, and the cells were serum-starved for one hour before treatment. The assay began with the addition of 10 μl coelenterazine h (Progmega, Mannheim, Germany) to each well at a final concentration of 5 μM. After a five-minute incubation, compounds were added in PBS containing forskolin (final concentration 10 μM). BRET readings started 15 minutes after the addition of the agonist. Emission signals from Renilla Luciferase and YFP were simultaneously measured using a CLARIOstar plate reader (BMG LabTech, Ortenberg, Germany) with a BRET1 filter set (475-30 nm/535-30 nm). BRET ratios (emission at 535-30 nm/emission at 475-30 nm) were calculated, and dose–response curves were analyzed using nonlinear regression in GraphPad Prism 6.0. Curves were normalized to the basal BRET ratio obtained from dPBS and the maximum response induced by morphine and DAMGO. Each curve was based on three to five independent experiments, each conducted in duplicate.

6.4 Calcium release

Calcium release was assessed using a FLIPRTETRA fluorescence imaging plate reader from Molecular Devices. These experiments were conducted alongside Gi/o Glosensor assays using the same HEK-293T cells transfected with μOR. For the FLIPR assay, cells were plated in 384-well black clear bottom cell culture plates coated with poly-lysine. The cells were incubated overnight at 37 °C with 5% CO2. The following day, the culture media was removed and replaced with Fluo-4 direct calcium dye from Life Technologies, prepared in HBSS containing 20 mM HEPES at pH 7.4. After a one-hour incubation at 37 °C, the cells were brought to room temperature. Fluorescence readings were taken for an initial 10 seconds to establish a baseline. Subsequently, 10 μl of drug solution at three times the concentration was added to each well, and the maximum increase in fluorescence was measured relative to the baseline. The drug solutions for the FLIPR assay were identical to those used in the Gi/o Glosensor experiments. As a positive control for calcium release through endogenous Gq-coupled receptors, TFLLR-NH2 at a concentration of 10 μM, a PAR-1 selective agonist, was employed.

6.5 Receptor internalization

Internalization was assessed using the eXpress DiscoveRx PathHunter GPCR internalization assay, which utilizes split β-galactosidase complementation. In short, cryopreserved U2OS cells expressing the human μOR were quickly thawed and plated in the provided medium within 96-well culture plates. The following day, the cells were treated with drugs at a concentration of 10× and incubated for 90 minutes at 37°C with 5% CO2. After incubation, a substrate was added to the cells, and chemiluminescence was recorded using a TriLux (Perkin Elmer) plate counter. The data were normalized to DAMGO and analyzed with Graphpad Prism 6.0.

6.6 Pharmacokinetics of PZM21

Studies were performed by the Preclinical Therapeutics Core and the Drug Studies Unit at the University of California San Francisco. Ten mice were injected subcutaneously with 20 mg/kg of PZM21. At each time point, 1 ml of blood was collected from three mice and the serum concentration of PZM21 determined by liquid chromatography–mass spectrometry (LC/MS). Mice were subsequently sacrificed and entire brains were homogenized for determination of PZM21 concentrations by LC/MS. All studies were performed with approved mouse protocols from the institutional animal care and use committees.

6.7 Metabolism of PZM21

Metabolism experiments followed previously established methods. Briefly, pooled microsomes from male mouse liver (CD-1) were obtained from Sigma Aldrich and stored at a temperature of −75 °C until needed. NADPH, sourced from Carl Roth, was kept at −8 °C. The incubation reactions took place in polyethylene caps (Eppendorf 1.5 ml) at a controlled temperature of 37 °C. Each incubation mixture contained PZM21 at a concentration of 80 μM or positive controls such as imipramine and rotigotine, along with pooled liver microsomes at a concentration of 0.5 mg of microsomal protein per ml of incubation mixture, and Tris-MgCl2 buffer (48 mM Tris, 4.8 mM MgCl2, pH 7.4). The total incubation volume was 0.5 ml. The microsomal reactions were started by adding 50 μl of enzyme cofactor solution NADPH, with a final concentration of 1 mM. At time intervals of 0, 15, 30, and 60 minutes, the enzymatic reactions were stopped by adding 500 μl of ice-cold acetonitrile containing 8 μM internal standard. The precipitated protein was then separated by centrifugation at 15,000 rcf for 3 minutes. The resulting supernatant was analyzed using HPLC/MS with a binary solvent system. The elution involved acetonitrile in 0.1% aqueous formic acid, progressing from 10% to 40% acetonitrile in 8 minutes, then increasing to 95% in 1 minute, followed by 95% acetonitrile for another minute, with a flow rate of 0.3 ml/min. The experiments were repeated three times independently. Control incubations were also performed without the cofactor solution to assess nonspecific binding to the matrix. The remaining substrate and metabolite formation were calculated as mean values ± s.e.m. based on three independent experiments. These values were determined by comparing the AUC of metabolites and substrate at each incubation time to the AUC of the substrate at the starting point, assuming similar ionization rates and adjusting using a correction factor derived from the AUC of the internal standard at each time point.

6.8 Chemical synthesis

The stereochemically pure isomers of compounds 12 and PZM21 were synthesized using their respective (R)- and (S)-amino acid amides, which were either commercially sourced or easily prepared from the corresponding acid or ester (refer to Supplementary Information). The primary amino group underwent dimethylation through the use of an excess of aqueous formaldehyde and sodium triacetoxyborohydride in aqueous acetonitrile. To obtain the primary amines, carboxamides 16a,b were treated with a borane-tetrahydrofuran complex under reflux, resulting in the formation of diamines 17a,b. A Henry reaction involving thiophene-3-carbaldehyde and nitroethane produced the nitropropene derivative 18, which was further transformed into the racemic alkylamine 19. This compound was activated with 4-nitrophenyl chloroformate to form carbamates 20. These carbamates were subsequently combined with the enantiopure primary amines 17a,b to yield diastereomeric mixtures of the corresponding ureas 12 and 21. Purification through HPLC with a semi-preparative Chiralpak AS-H column provided eight pure stereoisomers of compounds 12 and 21, including PZM21. To confirm the absolute configuration of the final products and efficiently prepare PZM21, an enantiomerically enriched carbamate 20 was synthesized and coupled with the corresponding primary amines. The racemic primary amine 19 was subjected to chiral resolution via repeated crystallization with di-p-anisoyl-(S)-tartaric acid. After three crystallizations, 19 was obtained in its dextrorotatory form with a specific rotation of +20.5°. Given that the (R)-acetamide form has a known specific rotation of +49.8°, the enriched 19 sample was treated with acetic anhydride and triethylamine. The resulting acetamide exhibited a specific rotation of −46.6°, confirming the major isomer's absolute configuration as (S). The enriched (S)-20 was then used for synthesizing the final urea derivatives, and the absolute configuration of the diastereomers was determined by matching their retention times in chiral HPLC. Comprehensive details regarding the synthetic routes and analytical data for compounds 12, PZM21, and related analogues PZM22-29 are provided in the Supplementary Information.

6.9 Data analysis and reporting

Other than the in vivo studies, no statistical analysis was applied to in vitro or cell-based signaling assays. Sample size (number of assays for each compound or receptor) was predetermined to be in triplicate or quadruplicate for primary screening assays at a single concentration. For concentration-response assays, the sample size (number of assays for each compound at selected receptors) was also predetermined to be tested for a minimum of three assays, each in triplicate or quadruplicate. None of the functional assays were blinded to investigators.

7.Results

Extended Data

Signaling data are mean ± s.e.m. of normalized results (n = 3 measurements).

       
           
       
Figure 4. Signalling properties of PZM21 at the opioid receptors

Raw luminescence data from a Gi/o Glosensor experiment are shown. Agonists reduce luminescence in agonist mode, but inverse agonists increase it by reducing basal signaling.
A prototypical, well-characterized agonist (black curves) and antagonist (red curves) were employed for each opioid receptor in order to verify the experiment. In antagonist mode, 50 nM agonist and increasing amounts of the tested medication are used to perform a competition reaction. In this case, the observed signal is increased by true antagonists, which is consistent with their capacity to compete with the agonist without triggering Gi signalling. Non-normalized results (n = 3 measurements) are represented by the mean ± standard error. aThe ECFP4 Tanimoto similarity (Tc) to the most similar µOR ligand in ChEMBL16.

aThe ECFP4 Tanimoto similarity (Tc) to the most similar µOR ligand in ChEMBL16. bNo measurable activity. aNot determined. bNo measurable activity.

Table 3: Binding and signalling properties of compounds 12 and PZM21

a, Opiate-induced μOR signalling through Gi activates G-protein-gated inwardly rectifying potassium channels (GIRKs) and inhibits adenylyl cyclase, leading to analgesia. Conversely, recruitment of β-arrestin is implicated in tolerance, respiratory depression, and constipation. b, Cutaway of the μOR orthosteric site to which β-FNA binds. Highlighted regions on the extracellular side diverge between the opioid receptors. c, Conserved features of opioid ligand recognition in the μOR. d, Overlaid docking poses of 23 compounds selected for experimental testing. e, Single-point competition binding assay of 23 candidate molecules against the μOR antagonist 3H-diprenorphine. Each ligand was tested at 20 μM and for those with > 25% inhibition affinity was calculated in full displacement curves; data represent mean ± s.e.m. (n = 3 measurements). One of these hits, compound 7, was subsequently optimized. f, Docking pose of compound 7.

a. Among a series of analogues derived from compound 7, compound 12 stood out due to its µOR selectivity and effectiveness as a µOR agonist. b. The docking pose of compound 12 is illustrated. c. In a Gi/o signaling assay, compound 12 exhibited agonist activity with an EC50 of 180 nM. DAMGO, a well-known unbiased opioid agonist, served as a reference. d. Although compound 12 effectively activated Gi/o signaling, it caused significantly less β-arrestin recruitment compared to DAMGO. Data for both c and d are presented as mean ± s.e.m. from 3 to 6 measurements of normalized results.

a. The development of PZM21 was guided by structural insights. b. A docking study compared compound 12 with β-FNA. The phenolic hydroxyl group in β-FNA interacts with His2976.52 through two water molecules, which provided clues for optimizing compound 12. c. When PZM21 binds to the active µOR, it forms a water-mediated network connecting its phenol group to His2976.52. The reference agonist BU72 is shown in orange for comparison. d. Key interactions between µOR and PZM21 include hydrogen bonds (blue dashes), hydrophobic interactions (green dashes), and an ionic bond (red dash). Additional data from related studies are shown in the extended data section. e. Different stereoisomers of compound 12 were tested in a Gi/o signaling assay. f. PZM21 demonstrated strong µOR agonist activity in the same assay.g. In a β-arrestin-2 recruitment test (PathHunter assay), PZM21 showed no detectable activity, indicating a reduced risk of side effects linked to this pathway. The data shown are average results with error margins from 3–6 measurement