Understanding the principles of GPCR selectivity by homology modeling and simulation of ternary complexes

Membrane proteins play a critical role in cellular signaling. G-protein coupled receptors (GPCRs), a membrane receptor family, are of particular interest as the largest family of drug targets. Drug-like molecules (ligands) can modulate GPCRs by either activating (agonists) or blocking their signaling (antagonists). Moreover, some ligands can decrease the basal signaling induced by the unbound GPCRs (inverse-agonists). The molecular criteria that determine the role of drug-like molecules on signaling weren’t revealed. The ability of some ligands to interact with different related members of GPCRs may lead to side effects that can be useful in some clinical indications or result in undesired complications. Recent research studies have proven that some ligands can initiate different signaling at the same GPCR and became highly interesting due to the therapeutic value of these extraordinary ligands. A molecular basis for designing these ligands remained, however missing. The aim of this thesis was to reveal the molecular determinants for the subtype and functional selectivity of ligands on GPCRs using computational structure-based modeling and simulations. The major challenge in addressing these investigations has been the lack of accurate computational schemes that can accurately sample and compare the dynamic binding profiles of versatile set of ligands on different receptors at a feasible computational cost. Arginine-vasopressin (AVP) is a peptide hormone that regulates different physiological functions by activating one of three different subtypes of the Vasopressin receptors (VR). A particular interest has been given to the V2R because of the therapeutic potential of its ligands in treatment of heart failure and chronic nephrogenic diabetes insipidus. I constructed a homology model of the receptor and used microseconds of molecular dynamics simulations to refine the model. Several trials to simulate spontaneous binding from several microseconds of atomistic simulations gave us insights into an initial binding complex, where the AVP peptide intimately interacts with the surface of V2R. Professor Francesco Gervasio and Dr. Giorgio Saladino have assisted me in understanding how to develop metadynamics simulations that I used to simulate and quantify the binding profile of AVP to V2R. I discovered that the AVP has three binding sites: a vestibule site that we captured in the unbiased simulations, an intermediate site and an orthosteric site that satisfy the criteria set by previous experiments for the binding partners. These results have shown that the extracellular vestibule of the receptor plays an important role in GPCR-ligand recognition and filter the possible binding modes and pathways. I have developed, with assistance from our collaborators, a simplified metadynamics scheme for the small non-peptide ligands that can be generalized on different GPCRs with less computational demand to allow application on bigger datasets. Binding metadynamics simulations for a set of ligands with varying subtype selectivity and functional activity toward the V1aR and V2R, has shown a preference of the V2R selective antagonist to bind on the vestibule site and the V1aR selective antagonists to bind to the intermediate site. The non-selective antagonists bound both sites with similar affinity and the agonists had better affinity to the orthosteric binding site. The metadynamics scheme for the non-peptide ligands was applied on five different GPCRs in different states (active/inactive) with different coupling partners (arrestin/G-protein). We have proven that we can generalize the scheme for different GPCRs with remarkable accuracy. The accuracy and correlation with experimental data encouraged us to apply the method to study more complex behaviors related to the signaling and regulation of the receptors. Three different GPCRs were crystalized in the active, high–affinity states by means of stabilizing engineered protein nanobodies, a protein mimic to their natural stabilizing G-protein or arrestin. Only One agonist–GPCR complex was crystalized in both the G-protein– and nanobody–stabilized states that allowed general comparison, whereas some experiments have shown different effect of both intracellular binding partners (IBP) on the agonist affinity. We used unbiased molecular dynamics simulations of microseconds on the agonists poses of the three GPCRs crystalized in the nanobody-stabilized active and inactive states. The agonist poses were carefully transferred to G-protein coupled models of their GPCRs and subsequently subjected to both unbiased simulation and the binding metadynamics that we developed. Our results have shown that the different IBPs can allosterically propagate different changes in the binding pockets. These changes can either affect the binding modes of the ligands, affinity and/or binding kinetics. These results indicate a basis that may explain the functional bias toward arrestin and G-protein signaling. 7.3 Molecular cooperativity between agonist and G-protein on β2–adrenergic receptor The ternary complex model is a widely accepted principle in GPCRs that explains the enhanced affinity of agonist binding to a GPCR upon coupling to an intracellular binding partner. Crystal structures have shown the change in the binding pocket upon activation. The direct comparison between the high– and low–affinity agonist poses is not possible because of the absence of crystal structures for agonists bound the low–affinity state. The static description of the crystal structures also lacks any information about the change in binding kinetics caused by G-protein stabilizations. I used molecular dynamics simulation to allow the agonist poses to adapt to the low affinity state in the inactive GPCRs and used unbinding metadynamics to compare the changes in the binding affinity and kinetics. Our results show that the G-protein stabilization decreases the barrier for the agonist to reach the orthosteric binding site and stabilizes intermediate/vestibule binding sites. In addition, I developed a metadynamics scheme to simulate the coupling and activation of the β2–adrenergic receptor, which allowed me to compare the coupling of G-protein in presence and absence of a stabilizing agonist. Many G-protein coupled receptors can interact with more than one IBP to initiate several independent signaling pathways. Recent studies have characterized some ligands to selectively modulate β2–adrenergic receptor to signal through either arrestin or G-protein pathways independently. I have designed ternary complex for β2–adrenergic receptor with arrestin. I used atomistic simulations to determine the β-arrestin- and Gs-coupled global minima of β2-adrenergic receptor in the apo-form and in presence of four different ligands. I have shown that using ternary complex simulations and metadynamics can identify and predict the efficacy and functional bias at the β2–adrenergic receptor. Metadynamics enhanced sampling allow us to quantify the cooperative behaviour between ligand and intracellular binding partner (G-protein or -arrestin). The direction of this effect (positive or negative) and its magnitude are characteristic for ligands with different functional selectivity.