A µ-opioid receptor modulator that works cooperatively with naloxone

Abstract

The µ-opioid receptor (µOR) is a well-established target for analgesia¹, yet conventional opioid receptor agonists cause serious adverse effects, notably addiction and respiratory depression. These factors have contributed to the current opioid overdose epidemic driven by fentanyl², a highly potent synthetic opioid. µOR negative allosteric modulators (NAMs) may serve as useful tools in preventing opioid overdose deaths, but promising chemical scaffolds remain elusive. Here we screened a large DNA-encoded chemical library against inactive µOR, counter-screening with active, G-protein and agonist-bound receptor to 'steer' hits towards conformationally selective modulators. We discovered a NAM compound with high and selective enrichment to inactive µOR that enhances the affinity of the key opioid overdose reversal molecule, naloxone. The NAM works cooperatively with naloxone to potently block opioid agonist signalling. Using cryogenic electron microscopy, we demonstrate that the NAM accomplishes this effect by binding a site on the extracellular vestibule in direct contact with naloxone while stabilizing a distinct inactive conformation of the extracellular portions of the second and seventh transmembrane helices. The NAM alters orthosteric ligand kinetics in therapeutically desirable ways and works cooperatively with low doses of naloxone to effectively inhibit various morphine-induced and fentanyl-induced behavioural effects in vivo while minimizing withdrawal behaviours. Our results provide detailed structural insights into the mechanism of negative allosteric modulation of the µOR and demonstrate how this can be exploited in vivo.

Extended Data Fig. 1 |. Initial biochemical characterization of μOR allosteric modulators from DEL screen.

(a) Chemical properties of 368. MW; molecular weight (Da). logP; predicted octanol/water partition coefficient. PSA; polar surface area (Ų). ROTN; rotatable bonds. HBD; hydrogen bond donors. HBA; hydrogen bond acceptors. RCount; number of rings. ARCount; number of aromatics. CNSMPO; central nervous system multiparameter optimization. LogP was predicted using QikProp in Schrödinger, and other properties were calculated using ChemDraw. (b) Excess concentrations of 368 have no impact (P = 0.084) on G_i1 intrinsic turnover in the absence of receptor. Data are displayed as the average ± s.d. with n = 4 individual experiments. (c) We show using a direct ³H-naloxone binding experiment that increasing concentrations of 368 result in an increased antagonist affinity for μOR-containing membranes. Fitted affinity values are shown along with 95% confidence intervals in parentheses. Data are displayed as average values with error bars corresponding to the standard deviations of n = 4 measurements (d) The GTP turnover assay was used to show that 20 μM 368 inhibits turnover for a wide variety of orthosteric site conditions, ranging from slight inhibition of basal signaling (P = 0.0115) to no detectable effect on naloxone turnover (P = 0.314) to substantial inhibition of moderate partial agonist turnover (mitragynine pseudoindoxyl, MP, P = 0.005) to peptide (DAMGO, P < 0.0001) or small molecule (BU72, P < 0.0001) full agonists (all orthosteric molecules also present at 20 μM, data are displayed as the average ± s.d. with n = 4 individual experiments). P values for all of the above were calculated using an unpaired t-test (two tailed) and are denoted as follows: ns (P > 0.05), * (P ≤ 0.05), ** (P ≤ 0.01), *** (P ≤ 0.001), and **** (P ≤ 0.0001). Titrations of (e) morphine, (f) fentanyl, and (g) met-enkephalin result in activation of an assortment of G-proteins (G_i1, pink; G_i2, orange; G_i3, pale green; G_oA, green; G_oB, blue; and G_z, purple) as observed in the TRUPATH assay by a change in BRET signal as the Gα and Gβγ subunits separate. Activation by all 3 agonists was also calculated in the presence of 90 μM 368 (bottom panels). The data are displayed as the average the average ± s.d. with n = 6 individual experiments. The average log(EC₅₀) for all 6 G-protein activation curves within each panel is shown as a black line (with dashed grey lines representing the standard deviation among different G protein subtypes). The average fold change in EC₅₀ upon addition of 368 for morphine (14.5), fentanyl (16.2) and met-enkephalin (15.8) is shown. (h) The presence of excess (90 μM) 368 results in decreased agonist potencies for a variety of orthosteric agonists (fentanyl, circles; met-enkephalin, squares; morphine, triangles) across a series of G_i/o family G-protein effectors as observed by the TRUPATH assay for G-protein activation. The calculated log(EC₅₀) for all agonist/G-protein combinations are right-shifted by ~1–1.5 units in the presence of 368. Data are displayed as average changes in log(EC50 values) for each condition with error bars corresponding to the additive fitted error 95% confidence intervals for EC₅₀ values with and without 368. (i) This G_i/o family inhibition results in dampened cAMP inhibition in cells with morphine (dark blue vs. pink) and met-enkephalin (green vs. purple). Titration of 368 at EC₈₀ concentrations of orthosteric agonists results in the reversal of cAMP inhibition, though with weaker potencies than those observed biochemically. Data are displayed as the average the average ± s.d. with n = 5 individual experiments.

Extended Data Fig. 2 |. CryoEM structure determination of μORκ-Nb6 bound to naloxone and 368.

(a) Schematic of μORκ purification and complex formation with Nb6. (b) Cryo-EM data collection and processing pipeline, showing representative micrographs of the μORκ-Nb6 complex, reference-free 2D cryo-EM class averages, and processing flow chart. This includes motion and CTF correction in Relion, followed by particle selection, 2D and 3D classifications, density map reconstructions, “gold standard” FSC curves in Cryosparc, and the final density map colored by local resolution. Also included is the cryo-EM density map and model for the seven transmembrane helices of the μORκ.

Extended Data Fig. 3 |. Stereochemistry & MD analysis of 368.

(a) Comparison of the chemical structures and raw sequencing counts in the DEL selections for a series of molecules in the enriched 368 family. S-stereoisomers are substantially more enriched than their R-counterparts. (b) Chiral chromatography trace demonstrating separation of racemic 368, conditions which were then used to calculate enantiomeric excess in (R)-368 (c) and (S)-368 (d). Calculated peak parameters for (b-d) are shown below the respective traces. (e) Comparison of the individually synthesized stereoisomers of the racemic 368 “hit”, demonstrating that the R-368 is >100 fold weaker than the S-368 isomer, consistent with the DEL enrichment data. Data are displayed as the average value ± s.d. with n = 4 individual experiments. (f) GTP turnover assay again demonstrating that S-368 retains the ability to potently inhibit fentanyl (5 μM)-induced GTP turnover like the racemic 368 “hit”, while the R-368 isomer does not display full inhibition even at 20 μM. Data are displayed as the average ± s.d. with n = 6 individual experiments. (g) Accordingly, the S-isomer of 368 was modeled and placed into the cryo-EM density map (blue), along with an alternate, sub-optimal pose (green) that fits into overlapping but distinct areas in cryoEM density. (h) Both poses were then subjected to MD simulation for three independent 200 ns simulations (without Nb6) (green). An alternate pose from the “opposite” orientation of the NAM in the binding site that (sub-optimally) fits into the cryoEM density was also subjected to three independent 200 ns simulations (blue). The overall root mean square deviations (r.m.s.d.) throughout the trajectory were calculated for protein Cα and all atoms in 368 and naloxone for comparison. The time-dependent r.m.s.d. of 368 throughout all trajectories is displayed in (h) and the average of each of the runs with error bars representing the s.d. of three independent simulation averages is displayed in (i). (j) Example simulation snapshots were overlaid by Cα alignment for both poses at 4 ns increments. For our chosen pose, the conformation of 368 remains nearly constant throughout all three simulations, with r.m.s.d. values near that for the protein Cα (h-j). While Cα and naloxone remain stable in both poses (i), the alternate 368 pose is unstable (h-j), resulting in very high r.m.s.d. values and extensive conformational sampling in the orthosteric vestibule region (j).

Extended Data Fig. 4 |. Mutational analysis of 368 binding site.

(a) Titration of (S)-368 against either wild type μOR (red), A323L μOR (purple) or I71W μOR (pink) expressing membranes demonstrates that both mutations substantially inhibit the ability of the NAM to enhance ³H-naloxone affinity. Data are displayed as the average ± s.d. with n = 4 individual experiments. (b) Expression levels of wild type human μOR (normalized to 100%) compared to mutations used in TRUPATH studies. Data are displayed as the average ± s.e.m. with n = 9 individual experiments. (c) Morphine dose-response curves observed by G_i1 recruitment to human μOR using the TRUPATH assay, comparing wild type receptor with a series of point mutants (A323L, I144E, H319L). The residue numbers correspond to those in the mouse μOR sequence used for the structural studies. Data are displayed as the average ± s.e.m. with n = 9 individual experiments. (d) Fentanyl dose-response curves observed by G_i1 recruitment to human μOR using the TRUPATH assay, again comparing wild type with point mutants. Data are displayed as the average ± s.e.m. with n = 9 individual experiments. Dose-response curves in c and d were fit using a three-parameter model for bottom, top, and logEC50 values. Log EC50 values for each curve, along with the 95% confidence interval range, are displayed for all curves.

Extended Data Fig. 5 |. Probe dependence and opioid receptor selectivity of 368.

(a) Comparison of the extracellular vestibule regions of the previous β-FNA-bound μOR and the current naloxone-NAM-bound μOR. The presence of 368 sterically restricts the ability of small molecule antagonists to enter/exit the orthosteric site. (b) Alignment of the receptor regions of various ligand-bound μOR structures demonstrate that small molecule orthosteric compounds (top; naloxone [present work], lofentanil [PDB: 7T2H], MP [PDB:7T2G]) have little steric clash with 368 (calculated as the number of atoms in the orthosteric ligand within 1.5 Å of 368), while both peptide agonists (bottom; β-endorphin [PDB: 8F7Q], DAMGO [PDB: 6DDE]) have clear and substantial steric clash when overlaid with the 368 binding site. (c-e) TRUPATH assays for G_i1 activation for three different opioid receptor/ligand pairs; (c) μOR/DAMGO, (d) δOR/DPDPE, (e) κOR/U50488. Data are displayed as the average ± s.d. with n = 6 individual experiments. Fitted agonist EC50 values for each curve are shown, along with the error values which correspond to the 95% confidence interval for the fitted values. 368 has the largest impact on DAMGO activation of μOR and shows some activity against δOR activation by DPDPE, but has no effects on κOR activation by U50488. (f) Sequence alignment of human μOR, δOR and κOR at structural elements with interaction with 368 (residues within 6 Å denoted with an *). Numbering refers to positions within the human μOR. Red asterisks denote residues with important interactions with 368 but have side chains that are predicted to clash or no longer form productive interactions with 368.

Extended Data Fig. 6 |. In vivo behavior of 368.

(a) Pharmacokinetics measurements of 368 at 10 mg/kg administered intravenously. 368 enters the brain with a maximum concentration of 1.66 uM, well above the observed affinity of the compound from radioligand binding (133 nM), but significantly lower than the concentration needed to inhibit agonist activity in cell assays. Data are displayed as the averages ± s.e.m. with n = 4 individuals. (b) In the absence of naloxone, 368 does not have significant effects on morphine-induced antinociception in the 55 °C warm-water tail-withdrawal assay (F_(14,84) = 0.80, p = 0.67; Two-way RM ANOVA; n = 8 mice per group). This contrasts to the observed potentiated antagonism when in the presence of a low-dose (0.1 mg/kg) of naloxone (Fig. 4b). (c,d) The CLAMS assay with n = 8 mice demonstrates that in the absence of orthosteric compounds, 368 alone (10 mg/kg) has no significant impact on ambulation (F_(6,108) = 0.36, p = 0.90; Two-way RM ANOVA) (c) and respiration (F_(6,108) = 1.19, p = 0.32; Two-way RM ANOVA) (d). (e,f) CLAMS assay with n = 12 mice/group (n = 8 for morphine alone) demonstrating a dose- and time-dependent ability of 368 to potentiate low-dose (0.1 mg/kg) naloxone in reversing morphine-induced hyperlocomotion (F_(24,306) = 4.27, p < 0.0001; Two-way RM ANOVA; e) and respiratory depression (F_(24,294) = 5.00, p < 0.0001; Two-way RM ANOVA; f), with a maximal effect observed at 100 mg/kg 368 for both. (g) Antinociceptive time course experiment with n = 8 mice/group demonstrating that a 0.3 mg/kg dose of naloxone is able to partially reverse fentanyl (0.1 mg/kg)-induced antinociception on its own (factor: dose x time; Two-way RM ANOVA, F_(7,98) = 3.883, p = 0.001; purple vs. orange curves), addition of 368 results in a significant enhancement of this naloxone-induced reversal (Two-way RM ANOVA,F_(7,98) = 7.069, p = <0.0001; orange vs. dark red curves, n = 12 mice). Data for are displayed as the average percent antinociception ± s.e.m. of n = 8 or 12 mice as detailed above. (h) Antinociceptive time course experiment demonstrating that a 1.0 mg/kg dose of naloxone are not significantly able to reverse fentanyl (0.1 mg/kg)-induced antinociception (F_(7,98) = 1.947, p = 0.070; purple vs. orange curves), addition of 368 results in a significant enhancement of this naloxone-induced reversal (F_(7,98) = 7.077, p = <0.0001; orange vs. dark red curves). Data for are displayed as the average percent antinociception ± s.e.m. of n = 8 mice.

Fig. 1 |. DEL screen for new μOR allosteric modulators.

a, Schematic detailing the DEL selection scheme including target (condition 1, inactive μOR–naloxone), anti-target (condition 2, active μOR–G_i) and no target control (condition 3). b, Enrichment scores for 368 (solid) and related family members (semi-transparent) across the three sample conditions. The chemical structure for 368 is displayed below. c, GTP turnover assay used to initially assess the activity of the NAM. G_i alone has intrinsic GTPase activity (dark grey) that is not significantly affected by the presence of apo μOR (light grey) (P = 0.114), but is enhanced when the receptor is bound to the full agonist peptide met-enkephalin (green) (P = 0.007). The NAM (red) significantly inhibits agonist-induced turnover (P = 0.005) back to levels that are indistinguishable from G_i alone (P = 0.071). Final concentrations of μOR and G_i were 0.5 μM, and met-enkephalin and 368 were present at 20 μM. Data are represented as the mean ± s.d. of n = 4 individual experiments. P values were calculated using unpaired t-test (two-tailed) and are denoted as follows: NS, not significant (P > 0.05), *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001 and ****P ≤ 0.0001. Schematic in a was created using BioRender (https://biorender.com).

Fig. 2 |. 368 inhibits turnover together with naloxone.

a, Introduction of increasing concentrations of 368 to the GTP turnover assay results in nearly full inhibition of excess (20 μM) met-enkephalin-bound receptor-mediated turnover with a potency in the single-digit micromolar range. Data are represented as the mean ± s.d. with n = 6 individual experiments. b, Increasing concentrations of 368 result in an increase in ³H-naloxone binding to the receptor with an observed EC₅₀ of 133 nM (95% confidence interval (CI) of 112–159 nM) (dashed red line). Data are represented as the mean ± s.d. with n = 4 individual experiments. c, GTP turnover assay with excess concentrations of 368, antagonist naloxone or full agonist DAMGO. 368 significantly inhibits basal signalling (grey versus red, P = 0.0029), the weak partial activity of naloxone (yellow versus orange, P = 0.0040) and the full activity of DAMGO (pink versus purple, P < 0.0001). Final concentrations of all ligands were 20 μM, with G_i at 0.5 μM and μOR at 1 μM. Data are represented as the mean ± s.d. with n = 6 individual experiments. d, Reversal of DAMGO-induced G_i1 activation by the μOR through the titration of naloxone in the absence (yellow) or presence (red) of 368. Data are represented as the mean ± s.d. with n = 12 individual experiments. e,f, cAMP levels decrease after increasing concentrations of morphine (e) or fentanyl (f) (light blue dashed curves). Naloxone concentration-response curves (CRCs) (solid curves) at EC₈₀ concentrations of agonist result in reversal of cAMP depletion. Increasing concentrations of 368 result in a left-shift in naloxone potencies, with clear effects observed at single-digit micromolar concentrations (light and dark green curves). Data are represented as the mean ± s.e.m. with n = 6 individual experiments. P values were calculated using unpaired t-test (two-tailed) and are denoted as follows: NS (P > 0.05), *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001 and ****P ≤ 0.0001.

Fig. 3 |. Structural mechanism of 368 NAM activity.

a, Cryo-EM density map of naloxone, 368 and Nb6-bound μOR–κOR (μORκ) coloured by subunit. Map contour level 0.93 in Chimera. b, Naloxone (right) forms a series of inactive-state interactions with μOR, although it also directly interacts with 368. 368 forms a series of interactions across the extracellular vestibule of the receptor (bottom), ranging from hydrophobic interactions with I144^3.29 and V143^3.28 in TM3 to hydrogen bonding with H319^7.36 in TM7 and π–π interactions with Y75^1.39 in TM1. Key interactions are shown in dashed boxes. c, Cryo-EM density maps of naloxone (yellow) and 368 (red). Map contour level 0.93 in Chimera. d, The off-rate of ³H-naloxone from μOR was measured in the absence (yellow) or presence (red) of 368. Data are displayed as the mean per cent binding ± s.d. of n = 4 individual experiments, normalized to nonspecific binding (0%) and values in the absence of introduction of cold naloxone (100%). e, Comparing the 368–naloxone-bound inactive state of the μOR with the active, DAMGO–G_i-bound state of the receptor structure (orange, Protein Data Bank (PDB) identifier: 6DDE) and previously published inactive structures (green, PDB identifier: 4DKL) reveals several features about the inactivation mechanism of 368. The pose of 368 observed here is inconsistent with either active or inactive inwards states of TM1 and TM2, resulting in an outward movement in both (clashes in black dashed circles). Most significantly, the active state of TM7 clashes extensively with 368, particularly W318^7.35 and H319^7.36. Binding of 368 stabilizes an extremely outward, inactive state of the extracellular half of TM7. f, ³H-naloxone on-rate experiment to agonist (MP) saturated μOR, demonstrating that 368 (100 μM) can increase its observed off-rate. Data are represented as the mean ± s.d. with n = 4 individual experiments and normalized as in d.

Fig. 4 |. 368 potentiates naloxone activity in vivo.

a, Antinociceptive time-course experiment (tail-flick analgesia with morphine) demonstrating that low doses of naloxone have no impact on morphine-induced antinociception, whereas increasing doses of 368 before morphine treatment in the presence of this low-dose naloxone results in substantial inhibition of morphine-induced antinociception. b,c, Although treatment with either low doses of naloxone or 368 on their own has no or only modest effects, we observed a similar enhancement of low-dose naloxone effects on hyperlocomotion (b) and respiratory depression (c) after co-administration with 368. Data are displayed in 20-min bins as percentages relative to vehicle (XAMB, progressive ambulations; BPM, breaths per minute). d, Intermediate doses of naloxone are not able to significantly reverse fentanyl-induced antinociception (tail-flick analgesia), although addition of 368 resulted in a significant enhancement of this naloxone-induced reversal. e, Morphine produces significant CPP (left blue bars) that is not blocked with low doses of naloxone (−15 min, cyan central bars), whereas pretreatment with 368 and low-dose naloxone eliminated morphine CPP (right purple bars). f–h, Naloxone (NAL) precipitation of withdrawal symptoms were measured, comparing conventional naloxone-induced effects (2) with the low doses used above without (3) and with (4) pretreatment with 368. f, 368 with low-dose naloxone increases teeth chattering frequency relative to low-dose naloxone alone, although not to the same extent as full-dose naloxone. g, Low-dose naloxone alone causes substantial increases in jumping frequency that is not enhanced by 368. h, Low-dose and high-dose naloxone causes significant diarrhoea in mice, although this response is inhibited by 368 such that there is no difference from saline control. See Methods for details on the statistical tests used and their results. All data are displayed as the mean ± s.e.m of cohort sizes described in the Methods. P values are denoted as follows: NS (P > 0.05), *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001 and ****P ≤ 0.0001.