Coupling LGM and TM Domain Motions to Capture Binding Site Conformational Changes Necessary for Agonist Recognition

The therapeutic action of many drugs and drug candidates is based upon GPCR activation. Agonist binding at the receptor extracellular domain initiates (or modulates) a complex signal transduction process, wherein dynamic changes in the TM helices and receptor loops are coupled to intracellular binding and activation of G proteins [42, 43]. Deciphering the details of agonist/ receptor interaction is essential to the discovery of novel agonist chemotypes, but the GPCR activation mechanism remains poorly understood at the atomic level. Though a crystal structure for activated bovine opsin was recently published - 44] . it is unknown whether the full complement of conformational changes necessary for GPCR activation is represented. All other available GPCR structures are complexed with an antagonist or inverse agonist, and are consistent with an inactive (antagonist bound) conformation. Importantly, useful models for the structure- based design of agonists do not necessitate modeling of the full- l ength activated receptor but only require an accurate representation of conformational changes in the ligand binding pocket. As discussed above, LGM was successful in generating antagonist-bound pocket models for the MCH and M2 GPCRs. However, due to the nature of activation-associated conformational changes, sampling of the ligand binding pocket side chains is inadequate for fully capturing GPCR/agonist interactions. While side-chain minimization can improve the prediction of agonist/receptor contacts, it was recently found that repositioning key residues of p2AR ligand binding pocket is alone insufficient to provide agonist-selective enrichment in VLS; selectivity was obtained only when the model was combined with a molecular interaction fingerprint scoring function [45]. More sophisticated methods that allow flexibility within the TM backbone or include sampling of helix orientation are thus needed. In the current section, we discuss the modeling of conformational changes restricted to the p2AR ligand binding pocket and the sufficiency of these for allowing accurate agonist recognition. Computational approaches for investigating global conformational changes and downstream effects resulting from receptor activation are discussed later in the chapter.

Recently, we constructed an agonist-bound p2AR model from the p2AR/ carazolol crystal structure by conducting LGM with the full agonist isoproter-enol [22]. Flexibility was permitted in the TM5 proline kink, extracellular loop 2 (ECL2), and binding pocket side chains during optimization. Following refinement, an approximately 2-A tilt of the TM5 extracellular end toward the ligand binding pocket was obtained (Fig. 15.2b). This movement enabled optimal engagement of the ethanolamine "tail" and catechol "head" of isoproterenol with corresponding hydrogen bonding partners in TM3 and TM5, consistent with prior mutagenesis data [46-49] . In comparison, a rigid helix model placed TM3 and TM5 too far apart to allow simultaneous interaction with the agonist "head" and "tail" functional groups. Relative binding energies for a diverse set of compounds ranging from full agonists to inverse agonists were accurately calculated with the TM5 flexible model, while a TM5 rigid model underestimated affinities for several ligands (Fig. 15.2c). As discussed below (see "VLS with Agonist-Selective Models"), incorporating shifts in the TM5 position provides a means to generate agonist-selective models of the receptor binding pocket in VLS. Interestingly, smaller shifts in the TM5 position of p2AR were predicted for optimal binding of partial agonists like dopa-mine or salbutamol, while inverse agonists and antagonists (such as carazolol or propranolol) blocked an inward move of TM5. The degree of tilt in the extracellular portion of TM5 then appears to serve as a ligand .dependent "rheostat," regulating the propensity of the receptor for activation. Mobility in TM5 may facilitate differentiation among the spectrum of full and partial agonists, antagonists, and inverse agonists. This model provides a satisfying description of ligand/receptor interactions and suggests a hypothesis for the coupling of ligand structure and receptor activation. However, polar interactions between full agonists, and serine or threonine side chains in positions 5.42, 5.43, and 5.46 of TM5 have been demonstrated for only a select group of GPCRs in the adrenergic, dopamine, and serotonin families. For other GPCR types, the preference for polar side chains in these pocket-exposed TM5 positions is not observed. As such, ligand-induced conformational changes within the binding pocket may vary between receptors and therefore require alternate procedures for modeling activation-associated conformational change.

0 0

Post a comment