How drugs bind to orthosteric sites of family A GPCRs: MD simulations of ligand binding with a focus on alprenolol binding to β2-adrenoceptor
RO Dror. a, AC Pan. a, DH Arlowa, DW Borhani. a, P Maragakis. a, Y Shan. a, H Xu. a, & DE Shaw a.b.
a. D. E. Shaw Research, New York, NY 10036; and b. Center for Computational Biology and Bioinformatics, Columbia University, New York, NY 10032
Key recognitions by the study
1. The major ligand binding pathway ascribing to how the ligands approach orthosteric ligand binding sites of β-adrenoceptors.
2. Two prominent energetic barriers in the binding and unbinding kinetics: a barrier at the receptor surface in drug-receptor dehydration; and a barrier brought by the receptor conformations at the binding site.
MD simulations
They performed simulations with the CHARMM force field with Anton accelerator. The system was equilibrated with Desmond on a commodity cluster.
Their Supporting Information contains protocols in detail, relevant numerical values and additional figures.
Initial conformation of receptors
The crystal structure of β1-adrenoceptor (PDB: 2VT4, chain B) and β2-adrenoceptor (PDB: 2RH1) without co-crystalised antagonists, cyanopindolol and carazolol, respectively. T4 lysozyme fused to ICL3 of β2-adrenoceptor was omitted. Each receptor was embedded in a hydrated lipid bilayer with all atoms. In β1-adrenoceptor, the 6 thermostabilising mutations introduced for the crystalisation were reverse-mutated to the wild-type with Maestro, which was also employed to add hydrogens to the receptors. Residues were left in the dominant protonation state at pH7, except for E3.41 and D2.50, both of which were protonated.
Ligands subjected in the study
The binding of three antagonists (propranolol, alprenolol & dihydroalprenolol) and an agonist (isoproterenol) to β2-adrenoceptor, also dihydroalprenolol binding to β1-adrenoceptor, were investigated. Unbiased simulations, which were free from any prior knowledge of the bindings, led the ligands to find their target sites spontaneously.
Ligands data in CHEMBL
Frequency of the bindings in simulations
In 82 simulations, 21 binding events were observed. The duration of each simulation was in a range of 1 to 19 μs.
Initial positions of ligands
10 identical ligands were introduced in the bulk solvent, at least 30 Å away from the binding pocket and 12 Å away from the receptor surface.
Ligand behaviours
The ligands diffused around the receptor before entering the binding pocket. The bound ligands then remained there for the rest of simulation period.
Binding rate
On-rate of alprenolol or dihydroalprenolol binding to β2-adrenoceptor:
3.1 x 10 ^(−7) L/mol/s at 37 ˚C (close to the lab-determined: 1 x 10 ^(−7) L/mol/s, Limbird & Lefkowitz 1976).
Binding energy of dihydroalprenolol to β2-adrenoceptor determined by free energy perturbation:
−13.4 ± 1.6 kcal/mol ≈ −560924 ± 6698 J/mol (similar to the lab-determined: −12.2 kcal/mol ≈ −51069 J/mol, Caron & Lefkowitz 1976).
Comparison with the crystal structure of alprenolol-bound β2-adrenoceptor
In 6 out of 12 simulations of alprenolol binding to β2-adrenoceptor, the bound conformations closely resembled the corresponding crystal structure (PDB: 3NYA) with RMSD less than 1 Å between the crystal and the average simulation poses; Figure 1B in the paper compares the two poses in an overlay.
The pathway of alprenolol binding to β2-adrenoceptor
The hydrophobic ring of alprenolol tended to associate with the lipid bilayer. However, the ligand was reported never to have entered the binding cavity from the lipid bilayer.
Alprenolol-binding simulated was reported to share a mutual pathway, which was observed in 11 out of 12 simulations. The pathway comprising two steps was briefly described as follows.
First, alprenolol passed between ECL2 and ECL3 then through the crevice between ECL2, TM5, TM6 and TM7 to reach and settle in the binding cavity.
In the first step, it contacted hydrophobic residues: F193 at ECL2, Y7.35, A 5.39, H6.58, and V6.59, spending typically several hundred nanoseconds while adopting its conformations. The ammonium group of alprenolol interacted ionically with D300 of ECL3, then the same ammonium group moved to associate temporarily with the carbonyl oxygen of D192.
In the second step, alprenolol migrated into the binding cavity through narrow passages between ECL2, TM5, TM6 and TM7. In some simulations, it immediately adopted conformations resembling the crystal counterpart, whereas in other simulations the ammonium group of alprenolol temporarily formed a salt bridge with D3.32 before settling as in the crystal. Alternatively, a twisted aliphatic chain of alprenolol by its β-hydroxyl group was seen for 2.6 μs before the β-hydroxyl group form a H-bond with D6.55.
Two prominent energetic barriers in binding and unbinding
The first highest barrier was observed outside the binding cavity at a distance of 15 Å away, before entering the extracellular vestibule. Alprenolol migrated from bulk solvent into the ECL vestibule; once entered, the ligand was less likely to dissociate back into the solvent. The second barrier was in the passage from the ECL vestibule into the binding cavity. Figure 3 in the paper graphically illustrates the energy barriers.
The contributing reasons for energetic barriers
The first barrier was reported neither to involve conformational changes nor to be an electrostatic in nature.
The binding process was not diffusion-controlled, since the experimentally determined binding rate of alprenolol to β2-adrenoceptor was two orders of magnitude slower than that obtained for typically diffusion-controlled associations; moreover, an estimated activation enthalpy of alprenolol association based on experimentally determined equilibrium dissociation constant (Kd) and dissociation kinetics is much larger than that observed for ligand diffusion in the water.
Hence the authors have suggested based on their simulations that the first barrier may be due to dehydration of alprenolol and the ECL vestibule upon the ligand entering into the region. Alprenolol was reported to lose about 80% of associated water upon binding to the receptor; 63% of the ligand dehydration was estimated to occur at this stage. As ligand enter the vestibule, about 50 nm x m of hydrophobic surface area buries itself, while about 15 water molecules escape from the vicinity within 1 ns.
The second barrier was proposed to involve conformational changes accompanied with dehydration. The binding separate Y7.35 and F193 at ECL2, in turn breaking an intramolecular salt bridge between D192 at ECL2 and K7.32. The events alone were suggested not to be rate-limiting, but dehydration again contribute to limit the rate.
Similar pathways of propranolol and isoproterenol binding to β-adrenoceptors
β2-adrenoceptor
Isoproterenol (agonist)
As entering the binding cavity, the agonist isoproterenol formed a salt bridge with D3.32; it however showed relatively higher mobility than the antagonists when bound. The duration of simulations did not extend long enough to observe receptor activation which occurs on millisecond time scale (Vilardaga et al. 2003).
Propranolol & Isoproterenol
Both binding of propranolol and isoproterenol recapitulated that of alprenolol, through similar pathways, energy barriers and dehydration needs.
β1-adrenoceptor
Dihydroalprenolol
Similar binding pathway was observed also for this ligand-receptor pair. In β1-adrenoceptor, the lack of a salt-bridge, equivalent to the one formed between D192 of ECL2 and K7.32 of β2-adrenoceptor, allows the vestibule to extend to TM2. In certain simulations dihydroalprenolol interacted hydrophobicly with TM2 more frequently than β2-adrenoceptor.
Comment
The study has demonstrated orthosteric ligand binding behaviours of β-adrenoceptors in detail. The investigation could be extended further in order to simulate receptor activation upon agonist binding in a longer time scale. In a challenging attempt, simulations could include multiple receptors to observe and dissect cooperativity in its presence.
References
Caron MG, Lefkowitz RJ. 1976. Solubilization and characterization of the β-adrenergic receptor binding sites of frog erythrocytes. J Biol Chem 251:2374–2384.
Limbird LE, Lefkowitz RJ. 1976. Negative cooperativity among β-adrenergic receptors. J Biol Chem 251:5007–5014.
Vilardaga J-P, Bünemann M, Krasel C, Castro M, Lohse MJ. 2003. Measurement of the millisecond activation switch of G-protein-coupled receptors in living cells. Nat Biotechnol 21:807–812.
