Dissection of ligand-CDK8/CycC unbinding free energy barriers and kinetics by molecular simulations

This study applies a novel computational strategy to investigate molecular recognition and binding kinetics using five pyrazolourea ligands dissociating from cyclin-dependent kinase 8 with cyclin C (CDK8/CycC) as an example. The computed free energy barriers guide designing compounds using the transient conformations unavailable in experiments. The intermediates and their free energy profile during ligand association and discussion processes control ligand-protein binding kinetics and bring a more complete picture of ligand-protein binding. We used metadynamics and a pathway search method to sample pathways and applied combined reduced dimensionality, molecular dynamics (MD) simulations and milestoning theory to construct the free energy profile and estimate the residence time. The binding free energy and the trend of binding kinetics agreed with experiments. We explain the why of the barriers and the kinetics and use the information to assist ligand design. Guided by a barrier of a ligand passing an C helix and activation loop, we introduced one hydroxyl group to parent compounds to design our ligands with increased residence time and validated our prediction by experiments. This work provides a novel and robust approach to investigate dissociation kinetics of large and flexible systems for understanding unbinding mechanisms and designing new small molecule drugs with desired binding kinetics.\n\nSignificance StatementThe transient conformations during ligand binding/unbinding control non-covalent binding kinetics. However, the transient structures and their free energy landscape of flexible ligand-protein systems are unavailable in experiments and challenging to model. Due to lack of understanding in binding kinetics, even scientists know that kinetic properties can be important in drug development, calculations using the intermediate states to design ligands with preferred binding kinetics are absent. We overcome these challenges and compute ligand-protein unbinding free energy profile using a novel method with molecular dynamics simulations, reduced dimensionality, and milestoning theory to deepen our understanding in molecular recognition. We also designed compounds based on the computed free energy barriers and experimentally validated that our designed compound can increase residence time.