Fig 1.
fABAMBER benchmarking and application to ligand binding.
a, fABAMBER performance compared to AMBER’s native GPU enabled simulation engine and NFE toolkit for systems comprising 30k, 64k, and 120k atoms. b, Example initial system configuration for ligand binding simulations where the μOR is represented in white and the morphine derived agonist BU72 is shown in orange. c, Averaged free energy landscape for BU72 mABP ligand binding simulations. Contours are drawn at 1, 3, and 5 kcal/mol. d, Structural overlap of BU72’s predicted low-energy conformation (white cartoon and orange sticks) and crystallographic position (PDB Code: 5C1M; blue).
Fig 2.
A common pose for fentanyl binding.
a, Chemical structure of fentanyl, carfentanil, and lofentanil. b-c, Structural superposition and sidechain interactions of fentanyl and its derivatives within the μOR orthosteric site and their corresponding averaged free energy landscapes. Contours are drawn at 1, 3, and 5 kcal/mol. d, Side-by-side comparison of side-chain interactions mediated through lofentanil’s 3-cis methyl group and a common 4-carbomethoxy shared by carfentanil.
Fig 3.
Piperidine rigidity influences both affinity and selectivity.
a, N-aniline-piperidine connecting dihedral free energy landscapes. S2 dihedral occupancy for fentanyl, carfentanil, and lofentanil was calculated to be 38.5%, 43.4%, and 75.9% respectfully. b, Superposition of lofentanil’s optimal S2 dihedral conformation with low-energy conformations identified in mABP ligand binding simulations. c, Comparison of μOR, 𝛿OR, κOR orthosteric site amino acid compositions and selectivity structural analysis. d, Correlation between S2 dihedral propensity with in-vitro potencies taken from ref [30].
Fig 4.
Docking and modelling of various fentanyl derivatives.
a-b, Docking and sidechain interactions of fentanyl, sufentanil, and alfentanil (a), and carfentanil, and remifentanil (b). Relative potencies (in-vivo ED50) are taken from ref [29] and are displayed as a fold-difference from fentanyl in the upper left corner of each panel. c-e, Modelling of various 3-cis alkyl groups. Relative potencies (in-vivo ED50) are taken from ref [31] and are displayed as a fold-difference from fentanyl. f, Docking and sidechain interactions of fentanyl derivatives bearing amide (Compound 4D [33]) or n-alkyl phenyl (Fen-Acry-PEO7 [32]) extensions. g, Modeling of fused-ring fentanyl derivatives.
Fig 5.
Effects of M153 mutations on ligand induced β-arrestin and Gi complex coupling.
a, Side-by-side structural comparison of orthosteric site interactions between DAMGO (PDB code: 6DDE) and fentanyl bound μOR. b-c, Dose-response curve for DAMGO and fentanyl induced Gi complex and β-arrestin coupling for wildtype and M1533.36 mutant receptors measured by nanoBiT direct interaction assays. EC50 and maximal coupling efficacy compared to DAMGO are shown. Values and error bars reflect mean ± s.e.m. normalized to DAMGO of three technical replicates.
Fig 6.
Fentanyl derivatives with no β-arrestin signaling.
a-b, Structure based rational and chemical structure of newly synthesized fentanyl derivatives. c, Gi complex and β-arrestin coupling dose-response curve for DAMGO, fentanyl, and compounds FD1-3 measured by nanoBiT direction interaction assays. EC50 and maximal coupling efficacy compared to DAMGO are shown. Values and error bars reflect mean ± s.e.m. normalized to DAMGO of three technical replicates.