Investigation of Ligand-HIV Protease Dissociation: Pathway, Interactions, Mechanism, and Free Energy Landscape

Drugs with desired kinetic properties have better efficacy. Non-covalent small molecule drugs can bind/unbind from their targets, which normally are macro biomolecules. Experimental methods can measure kinetics parameters at ensemble level, but they are unable to track the exact drug binding/unbinding pathways or explain whether drugs show different kinetic behaviors from different pathways. Investigation of ligand binding/unbinding pathways can deepen our understanding of ligand-protein molecular recognition. In this work, possible pathways of ligands-protein were sampled and post-analysis were conducted to investigate ligand binding kinetics (residence time), free energy landscape of ligand unbinding, and key factors that affect binding kinetics. We are the first utilizing unbiased molecular dynamics to sample a pair of ligands, ritonavir and xk263, unbinding from HIV protease comprehensively, and classify the unbinding pathways based on the contacts between ligand-protein regions during dissociation. I identify key residues that form hydrogen bond with ligands that results in a meta-stable state for ligand-protein complex. Four distal mutation sites were observed forming interactions with ligands during unbinding, which explains why distal mutations in HIV protease affect drug binding affinity. Molecular mechanics Poisson–Boltzmann surface area (MM/PBSA) was performed to calculate free energy landscape during ligands dissociation. HIV protease backbone conformations during ligands binding/unbinding were compared based on root-mean-square deviation (RMSD). High similarity of protein conformation during ritonavir binding/unbinding suggests that ritonavir follows conformation-selection mechanism during dissociation. In the contrast, low similarity of protein conformation during xk263 binding/unbinding suggests that xk263 follows induced-fit mechanism during dissociation.Reconstruction of real free energy landscape of ritonavir-HIV protease dissociation was performed using Binding Kinetic Toolkit (BKiT) package. First two principal components (PC) of alpha carbon of protein along with heavy atoms in ritonavir were used as reaction coordinates to guide dissociation path on PC space. Energy barriers on free energy landscape under different unbinding pathways were explained by the molecular recognition of ritonavir-HIV protease. Furthermore, free energy landscape using ligand RMSD were computed, giving a much more accurate binding free energy and residence time approximation but in the lack of detailed ligand-protein interactions