HIV protease inhibitors

Molecular dynamics simulations of the HIV protease complexed with fullerene-based inhibitors



Ever since the complementary spatial relationship between fullerene C60 and the cavity of the HIV protease (HIVP) was first reported, there has been a growing interest in the possibility of designing HIVP inhibitors based on C60 derivatives.
Using the PINYMD package developed in my group in collaboration with Prof. Glenn J. Martyna at Indiana University, we have carried out classical molecular dynamics (MD) simulations to investigate the ability of such inhibitors to localize in the active site region of the HIV protease. These simulations utilized multiple time scale integration methodology (r-RESPA) and the CHARMM22 force field. Using a hierarchy of intramolecular, short range and long range force separations, the reversible multiple time scale methodology allowed all simulations to be carried out with a time step of 8 fs. Simulations were carried out in the gas phase, using the recently introduced cluster Ewald method, to reflect the type of calculation often performed in conjunction with drug design. Simulations including explicit solvent are currently under way.

The following depicts the structure of the HIV protease (coordinates obtained from the Brookhaven protein data bank).

It is a dimer composed of two noncovalently bonded 99-residue strands. The cavity region is characterized by two flexible flaps at the top and the active site region at the bottom.

Simulations of the HIV protease complexed with C60 alone, two C60-based inhibitors, and a commercially available inhibitor, Saquinavir, were carried out, each of 2 ns in length at 300 K using canonical MD methods.

When complexed with C60 alone, a flap-opening event was observed to occur after about 1 ns. The figure below shows the progression of the flap distance (defined as the distance between alpha carbons of two isoleucine residues in the flap region.

and the snapshot shows the C60 moving toward the open flaps, away from the active site.

This suggests that the average flap distance could be an important geometrical parameter for characterizing the efficacy of a given inhibitor.

When HIVP is complexed with the two fullerene-based inhibitors, a correlation is found between the average flap separation, the average distance between the catalytic aspartic acid residues (Rasp-asp), and the activity of the inhibitor. In particular, the Rasp-asp values for HIVP complexed with C60 alone, with the less active and with the more active fullerene-based inhibitors are, 9.03, 7.26 and 7.19 angstroms, respectively, while the flap separations are 9.95, 8.95 and 5.80 angstroms, respectively. Thus, we see that both of these parameters decrease with increasing activity. For comparison, the values of the flap separation and Rasp-asp for HIVP complexed with Saquinavir are 6.2 and 5.9 angstroms, respectively. Thus, the more active fullerene-based inhibitor leads to average cavity structural features that are comparable with an active inhibitor, Saquinavir. The figures below shows snapshots of HIVP complexed with the more active fullerene-based inhibitor and HIVP complexed with Saquinavir, respectively:



It should be noted that both Saquinavir and the more active fullerene-based inhibitor share an important design feature in common, which is an OH group that is capable of hydrogen bonding to oxygen atoms in the catalytic aspartic acid residues.

Recently, we have performed multiple time scale molecular dynamics and free energy calculations of the flap-opening event in solution of the HIV protease complexed with the aforementioned fullerene-based inhibitor. In solution, it is not known what the protonation state of the active site residues is. In order to gain some insight into this question and to investigate the binding mechanism, the flap-opening free energy profile as a function of the Asp-Asp distance defined above was computed and the water content of the active site determined. The free energy profiles, computed using both the bluemoon ensemble approach and the recently introduced adiabatic free energy dynamics method, are shown below for different protonation states of the HIV protease alone and the HIV protease complexed with the fullerene-based inhibitor in solution:

The figure shows a flat free energy profile for the protease alone in solution. However, in complexes with the fullerene-based inhibitor, a significant barrier to flap-opening develops which increases as the protonation state of the active site increases. Below, we show typical snapshots from the MD simulation of water content of the active site in the absence and presence of the inhibitor:


These show that water content is significantly reduced when the protease is complexed with the inhibitor. What is observed is that the cavity changes its shape to fit the inhibitor, thereby ``squeezing out'' the water and creating a strong hydrophobic interaction between the cavity and the C60 moiety. The following tables give average cavity structure and the computed water density in the cavity region with and without the inhibitor in different protonation states:

Note that the effective volume occupied by the drug is excluded (Vexcl) These water densities are to be compared with the value 0.033 for bulk water. These show a significant reduction water density in the cavity region upon complexation with the inhibitor and help explain the flap-opening free energy profiles above. The results suggest that binding is most favorable in a diprotonated state, in which an oxygen on each Asp residue is protonated.

References
Molecular dynamics study of the connection between flap closing and binding of fullerene based inhibitors of the HIV-1 protease.
Z. Zhu, D. I. Schuster and M.E. Tuckerman, Biochem. (in press).

A molecular dynamics study of HIV-1 protease complexes with C60 and fullerene-based anti-viral agents .
H. Mi, M.E. Tuckerman, D.I. Schuster and S.R. Wilson, Proc. Electrochem. Soc. (1999).