(361d) Flap-Opening Dynamics and Ligand Unbinding of HIV-1 Protease Studied Using Accelerated MD Simulations

HIV-1 protease (PR) is a functional protein within the HIV virus. Due to it's necessity for viral maturation, PR is a commonly targeted protein for drug development. PR has been extensively studied using MD simulations. A complete mutagenesis study indicated that there exists 3 regions of residues which must be highly conserved for proper function (Loeb et al, 1989). These regions have been identified as the binding pocket at residues 22 to 33, the flap-tip region at residues 47 to 52, and a hydrophobic region spanning residues 74 to 87. In MD simulations, a fourth region of importance was discovered as a mobile elbow region which acts as a hinge for wide flap opening (Harte et al, 1992). Two competing theories of ligand binding emerged from these regions of importance. The first involves wide flap opening which engages the elbow. When the flaps are wide open, the ligand has uninhibited access to the binding pocket. The second theory involves a slightly open flap conformation common in the unbound protein. The ligand threads into the binding pocket through the open side of the dimer. In both theories, the flaps tightly close following ligand binding and prior to proteolysis. Our current work aims to uncover the most efficient and probable protein movement which allows for ligand binding and eventual proteolysis.

We generated an all-atom, explicit water simulation system using the PR from the PDB entry 1F7A, a closed-form PR with the 9-residue C-peptide representing the capsid/p2 cleavage site of HIV-1 Gag bound. The system was prepared for simulations through solvation and a long time scale MD equilibration simulation. Long MD exhibits a strong preference towards the closed, ligand-bound state. Ligand unbinding was studied using Temperature Accelerated MD (TAMD) simulations. TAMD acts as an equalizer to increase the accessibility of high energy states by tethering a collective variable (CV) to an unrestrained fictitious particle. CV's must be selected with care to accelerate the desired outcome without biasing inherent behavior. In the current study, the desired outcome includes opening of the flap tips and movement of the ligand out of the binding pocket. We anticipate that both of these movements will have significantly slower inherent time scales than normal protein fluctuations and will require acceleration to be accessible on normal MD time scales. We have chosen to accelerate the cartesian coordinates of the N- and C-termini of the ligand to observe ligand movement out of the pocket. To observe flap opening, we have tested 2 independent CV sets. The first biases the distance of each flap tip from the binding pocket and will be referred to as the distance CV. This CV forces flap opening and biases a wide open state which is believed to be required for ligand unbinding. The second CV set forces a highly localized curling of the flap tips and will be referred to as the flap curling CV. Localized flap curling was simulated by accelerating the distance between the functional isoleucine in the center of the flap tips (ILE-50) and the center of mass of two isoleucines at the edges of the flap tips (ILE-47 and ILE-54). This CV forces the unbinding of the closed flap conformation without biasing towards the open state.

TAMD simulations were performed for each of these CV sets in 10 replicate simulations. Each TAMD simulation runs independently starting from an equilibrated, closed, and bound form of PR. The distance CV produced ligand unbinding in each replicate. The CV highly biases the wide open conformation with a large percentage of the simulation occurring with flaps further than 3 nm from the binding site. The flap curling CV also produced ligand unbinding in each of the 10 simulations. This CV causes an unbinding of the flaps by disturbing the interactions between the flap tips. Once the flaps are in the semi-open state, the ligand typically moves laterally, normal to the binding pocket, forcing the flaps to open more widely and allow for ligand unbinding and exit. Wide flap opening was rarely seen outside of the process of ligand unbinding. Simulations maintain a tighter flap opening even as the ligand exited the binding pocket. In each of the 20 independent simulations, the flaps must open in order for ligand unbinding to occur, no instances of ligand threading with closed flaps were observed. Ligand threading out of the binding pocket did occur when flaps were slightly open rather than fully open. Flap opening was not significantly skewed towards symmetric or asymmetric flap opening with similar appearances of both pathways.

Similar TAMD simulations were performed with a ligand-free solvated protein initially in the fully closed state. The closed state is highly stable in MD simulations even in the unbound state and the closed, unbound protein exhibited no change in state for hundreds of nanoseconds. TAMD simulations performed with the distance-based and flap-curling CV sets show fast unbinding of the protein flaps and no reemergence of the closed conformation throughout the simulations. Disproportionately high incidents of wide flap opening were observed in the distance-based CV. The flap curling CV forced flap opening and unbinding, but flaps maintained semi-closed or slightly open conformations for the majority of the simulation.

The TAMD results relating to ligand unbinding indicate wide-flap opening is not necessary for the ligand to exit the binding pocket and slight opening of flaps is sufficient. In addition, the TAMD results indicate that specific flap curling and orientation are necessary for maintaining a closed conformation and strong flap binding. Flaps spontaneously open to release the ligand as a result of removing inter-flap bonding through localized flap curling. Flaps bias towards a slightly open state and some appearance of a wide open state without bias indicates that the wide open state is not preferred but energetically available.