Structure-based design of bitopic ligands for the µ-opioid receptor

Abstract

Mu-opioid receptor (µOR) agonists such as fentanyl have long been used for pain management, but are considered a major public health concern owing to their adverse side effects, including lethal overdose¹. Here, in an effort to design safer therapeutic agents, we report an approach targeting a conserved sodium ion-binding site² found in µOR³ and many other class A G-protein-coupled receptors with bitopic fentanyl derivatives that are functionalized via a linker with a positively charged guanidino group. Cryo-electron microscopy structures of the most potent bitopic ligands in complex with µOR highlight the key interactions between the guanidine of the ligands and the key Asp^2.50 residue in the Na⁺ site. Two bitopics (C5 and C6 guano) maintain nanomolar potency and high efficacy at G_i subtypes and show strongly reduced arrestin recruitment-one (C6 guano) also shows the lowest G_z efficacy among the panel of µOR agonists, including partial and biased morphinan and fentanyl analogues. In mice, C6 guano displayed µOR-dependent antinociception with attenuated adverse effects, supporting the µOR sodium ion-binding site as a potential target for the design of safer analgesics. In general, our study suggests that bitopic ligands that engage the sodium ion-binding pocket in class A G-protein-coupled receptors can be designed to control their efficacy and functional selectivity profiles for G_i, G_o and G_z subtypes and arrestins, thus modulating their in vivo pharmacology.

Extended Data Fig. 1 |. Docking of a fentanyl based bitopic targeting the Na⁺ binding site.

Molecular docking of a fentanyl based bitopic ligand shows that the functional head group can target the Na⁺ pocket.

Extended Data Fig. 2 |. Cryo-EM data processing work-flows.

Representative micrographs, 2D classes, 3D classes and data processing procedures for (A) C5-guano and (B) C6-guano bound μOR-G_i-scFv16 complex.

Extended Data Fig. 3 |. Global and local resolutions for cryo-EM maps.

(A) Gold-standard FSC curves for C5-guano and C6-guano bound μOR-G_i structures. Overall resolution is 3.2 Å for C5-guano bound μOR–G_i-scFv16 and 3.3 Å for C6-guano bound μOR–G_i using the gold Standard FSC = 0.143 criterion. (B) Local resolution map of C5 guano and C6 guano bound μOR–G_i structures. (C) Data collection, refinement, and model statistic of two structures. Extended Data Table 2. Cryo-EM data collection, refinement and validation statistics.

Extended Data Fig. 4 |. Comparison of bitopic structures to BU72 structure.

A, C, Side chains of μOR orthosteric pocket residues are shown for the C5-guano (A) and C6-guano (C) bound μOR–G_i complex (green) in comparison with the BU72 bound μOR (PDB code 5C1M; pink). The orthosteric pocket residues of μOR in complex with bitopic ligands and BU72 show nearly identical conformations. B, D, Side chains of μOR site-2 and Na⁺ site residues are shown for the C5 guano (B) and C6 guano (D) bound μOR–G_i complex (green) in comparison with the BU72 bound μOR (PDB code 5C1M; pink). The site-2 and Na⁺ site residues of μOR in complex with bitopic ligands and BU72 show nearly identical conformations.

Extended Data Fig. 5 |. Analysis of dynamics of direct and water mediated interactions of bitopic ligands.

A) Overlay of three examples of C5 guano conformations bound to active state MOR (pink cartoon/sticks) during MD simulations (B) Detailed view of the interactions between guano moiety of C5 guano (orange sticks) and D114^2.50 mediated by two water molecule (C) Direct salt bridge interactions between C5 guano (light green sticks) and D116^2.50 supplemented by an additional water-mediated hydrogen bond. (D) direct salt bridge interactions between C5 guano (cyan sticks) and D114^2.50 (E) Probability densities of distances between guano nitrogen atoms and D114^2.50 carboxylate oxygens. Each chart plots probability density for frames with two bridging waters (orange), one bridging water (green), and no bridging waters (cyan). (F) Categorization and relative proportion of D2.50 and D3.32 mediated interactions in 10 independent C5 guano-μOR MD trajectories for 1 μs each. Among the cumulative frames from the 10 μs MD runs, close to 1/3^rd of the frames-maintained guano-D2.50 interactions exclusively through water-mediated hydrogen bonds, while ~57% frames formed direct salt- bridges with or without supplementary water mediated interactions. Therefore, close to 90% of the frames maintained D2.50-guano interactions. The piperidine-D3.32 interactions were observed to be even more stable, with over 96% of the frames indicating direct salt bridge or water-mediated hydrogen bonds. (G) Categorization and relative proportion of D2.50 and D3.32 mediated interactions in 5 independent C6-μOR trajectories for 1 μs each. Overall, the number of direct interactions with D2.50 increased from 57% to 85% (compared to C5), perhaps resulting from the increase in linker length by a carbon atom that decreases the overall distances to D2.50 residue.

Extended Data Fig. 6 |. Profiling of chemically and pharmacologically distinct μOR agonists using TRUPATH, arrestin signaling.

A) Peptides: Endomorphin-1, Leu-enkephalin, Met-enkephalin, Beta- endorphin and Dynorphin A (1-17). Dynorphin A (1-17) showed reduced arrestin recruitment while other peptides retained robust arrestin recruitment among peptides tested. B) Opioid biased agonists and partials: PZM21, TRV130, Gα- subtype selectivity and arrestin recruitment on μOR. PZM21, 7-OH and TRV130 showed <50% efficacy for arrestin1/2. Highest efficacy for all three biased agonists was seen at the G_z-subtype. μOR partial agonist pentazocine was a full agonist at the G_z subtype. C) Oxycodone and Carfentanil, Gα- subtype selectivity and arrestin recruitment on μOR. Carfentanil showed near maximal efficacy at all Gα- subtypes and arrestin1/2. Oxycodone was a full agonist at G_z and showed >50% efficacy at β-arrestin2. D) Fentanyl guano bitopics show differential G-protein and arrestin efficacy with increased chain length.

Fig. 1 |. Targeting the Na⁺ site with fentanyl-based bitopic ligands and characterization of lead compounds in binding, G-protein and arrestin signalling assays.

a, Structure of Bu72 bound to μOR, showing the orthosteric site and the unoccupied Na⁺ site. The μOR structure is shown in green ribbon as well as transparent grey surface. b, Docking of fentanyl in μOR, showing the orthosteric site and the unoccupied Na⁺ site. c, Chemical structures of fentanyl and designed bitopic ligands. d, Chemical structures of the lead fentanyl guano bitopic ligands. e, Binding affinities at the μOR. Lead bitopic ligands (guano fentanyls) were characterized in binding assays in CHO cells expressing μOR using [¹²⁵I]IBNtxA as radioligand. C5 guano and C6 guano had similar affinity to fentanyl. Data are mean ± s.e.m. (n = 3 experiments each done in triplicate). f,g, cAMP inhibition (f) and Tango assay for β-arrestin-2 recruitment on μOR (g) with the bitopic ligands. The guano bitopic derivatives show high G-protein agonism with poor recruitment of β-arrestin-2; C5 guano was the most active relative to DAMGO. Data are mean ± s.e.m. (n = 3 experiments each done in duplicate). Extended Data Table 1 shows values for all panels and data for all bitopic ligands.

Fig. 2 |. Structures of bitopic ligands bound to μOR.

a,e, Cryo-EM structure of C5 guano (a) and C6 guano (e) bound to μOR–G_i–scFv16 complex coloured by subunit. b,f, View of the C5 guano (b) and C6 guano (f) cryo-EM density. The cryo-EM maps are contoured at 5.0σ. The functional groups of the bitopic ligands target site 2 above the Na⁺-binding site, as expected. c,g, Salt-bridge interactions between C5 guano (c) or C6 guano (g) and residues in the μOR pocket. d,h, Cation–π interactions between C5 guano (d) or C6 guano (h) and residues in the μOR pocket.

Fig. 3 |. Profiling of C5 guano, C6 guano and μOR using TRUPATH Gαβγ biosensors and β-arrestin-1 and β-arrestin-2 efficacy.

a,b, BRET assays for β-arrestin-2 (a) and β-arrestin-1 (b) recruitment in the presence of 10 μM C5 guano, C6 guano, morphine, fentanyl, buprenorphine, 7-OH and DAMGO (all relative to DAMGO). a, C5 guano and C6 guano showed significantly reduced β-arrestin-2 recruitment compared with DAMGO. Minimum, median and maximum values–C5 guano: 23.9%, 29.0%, 42,5%; C6 guano: 2.7%, 14.6%, 38.4%; morphine: −6.0%, 25.8%, 62.6%; 7-OH:−13.1%, −2.4%, 25.2%; DAMGO: 100%, 100%, 100%; fentanyl: 93.5%, 99.7%, 113.8%. b, Similarly, C5 guano and C6 guano showed significantly reduced β-arrestin-1 recruitment compared with DAMGO. Minimum, median and median values–C5 guano: 17.0%, 26.6%, 29.4%; C6 guano: −7.9%, 4.2%, 4.8%; morphine: 7.2%, 25.6%, 50.2%; 7-OH: −8.0%, −1.4%, 0.3%; DAMGO: 100%, 100%, 100%; fentanyl: 77.4%, 77.4%, 98.4. Data are mean ± s.e.m. (n = three experiments each done in duplicate). Primary statistics for a,b are provided in Supplementary Table 2. c–i, Gα-subtype selectivity using TRUPATH on μOR for DAMGO (c), C5 guano (d), C6 guano (e), buprenorphine (f), fentanyl (g), morphine (h) and 7-OH (i). C6 guano showed distinct potencies and efficacies for all six Gα-protein subtypes (G_i1, G_i2, G_i3, G_oA, G_oB and G_z) with the lowest efficacy for the G_z subtype. j, Efficacy heat map for opioid peptides, biased agonists, partial agonists, morphine or fentanyl template agonists and bitopic ligands using TRUPATH and β-arrestin-1 and β-arrestin-2 activity. Raw curves and values for all compounds are presented in Extended Data Fig. 6 and Extended Data Tables 2 and 3. Data are mean ± s.e.m. (n = three experiments each done in duplicate).

Fig. 4 |. C6 guano exhibits μOR-mediated antinociception without CPP, CPA or hyperlocomotor effects.

a, Antinociceptive time course: C57BL/6J mice were administered C6 guano by ICV injection and antinociception was measured using the 55 °C tail-withdrawal assay. Data are shown as mean percentage antinociception (MPE) ± s.e.m. An ED₅₀ (with 95% confidence interval) of 18.77 nmol (5.49–55.54 nmol) was calculated for C6 guano. b, Antinociception by 100 nmol ICV C6 guano was attenuated in μOR-knockout (KO) mice compared with wild-type (C57BL/6J) or vehicle-treated (veh) mice. Data are MPE ± s.e.m. c,d, Locomotor effects and respiratory depression. C57BL/6J mice were administered either saline, vehicle, morphine (10, 30 or 100 nmol ICV), C6 guano (100 or 300 nmol ICV) and ambulation (c) or number of breaths (d) was measured every minute and averaged in 20-min bins. Data are presented as percentage of vehicle response ± s.e.m. c, Morphine treatment resulted in hyperlocomotion, whereas the C6 guano (300 nmol ICV) effect was not significantly different from ICV vehicle at any time point. d, Morphine did not result in any significant decrease in respiration rate at 10 nmol, but caused respiratory depression at 30 nmol and 100 nmol doses. C6 guano caused increased respiration rate at 100 nmol and 300 nmol doses compared with vehicle. There were no significant differences between wild-type and μOR-knockout mice treated with 300 nmol ICV C6 guano. e, C6 guano (100 nmol ICV) treatment did not result in CPP or CPA changes relative to vehicle, whereas morphine (30 nmol ICV) resulted in CPP and U50,488H (100 nmol ICV) resulted in CPA. Minimum, median and maximum values–U50,488H pre-conditioning (pre-CPP): −207 s, 52 s, 344 s; U50,488H post-conditioning (post-CPP): −1,436 s, −109 s, 178 s; morphine pre-CPP: −542 s, −23 s, 413 s; morphine post-CPP: −299 s, 154 s, 819 s; C6 guano pre-CPP: −364 s, 43 s, 499 s; C6 guano post-CPP: −878 s, −42 s, 1,164 s; saline pre-CPP: −416 s, −93 s, 595 s; saline post-CPP:−411 s, 129 s, 438 s; vehicle pre-CPP: −377 s, 43 s, 573 s; vehicle post-CPP: −865 s, −75 s, 533 s. f, Comparison of C6 guano and U50,488H in an operant model of antinociception using CCI–CPP. A schematic representation of the CCI–CPP model protocol is presented in the Supplementary Methods. Points represent the difference in time spent on the drug-paired side. Data are mean ± s.e.m. U50,488H treatment in this model did not result in CPP or CPA, whereas C6 guano treatment resulted in CPP. Minimum, median and maximum values: U50,488H pre-CPP: −771 s, −52 s, 699 s; U50,488H post-CPP: −1,170 s, −67 s, 1,293 s; C6 guano pre-CPP: −767 s, 101 s, 438 s; C6 guano post-CPP: −566 s, 388 s, 1,397 s. g, Dose- and time-dependent anti-allodynic activity of C6 guano in a CCI model of neuropathic pain. Mechanical allodynia produced by sciatic nerve ligation was reduced between 20 and 100 min after treatment with 30 nmol or 100 nmol ICV C6 guano but not by 10 nmol ICV guano or vehicle. Data are mean ± s.e.m. h, Antinociception by C6 guano in the mouse acetic acid writhing test. Treatment with 100 nmol ICV C6 guano or 30 nmol ICV morphine showed significant antinociception compared with vehicle or saline. Data plotted are number of writhes counted over 15 min after administration of compound. Minimum, median and maximum values: vehicle: 7, 19.5, 37; morphine: 0, 0, 32; C6 guano: 0, 0.5, 17. i, Evaluation of C6 guano for antinociceptive effects in the formalin-induced inflammation assay in mouse. Treatment with 100 nmol ICV C6 guano or 30 nmol ICV morphine showed significant antinociception compared with vehicle or saline. Minimum, median and maximum values: vehicle: 34 s, 93.5 s, 151 s; morphine: 1 s, 7 s, 95 s; C6 guano: 0 s, 0 s, 78 s; saline: 29 s, 74 s, 150 s. n values and primary statistics for all panels are provided in Supplementary Table 2.