Molecular modeling studies suggest that zinc ions inhibit HIV-1 protease by binding at catalytic aspartates.

Human immunodeficiency virus type 1 protease is inhibited in vitro by zinc ions at neutral pH. The binding site of these ions is not known; however, experimental data suggest that binding may occur in the active site. To examine the possibility of zinc binding in the active site, molecular dynamics simulations in the presence and absence of zinc have been carried out to 200 psec. The results are compared with the 2.8-A crystallographic structures of a synthetic HIV-1 protease, and a zinc binding site at the catalytic aspartate residues (Asp-25, Asp-25') is proposed. Molecular dynamics simulations show that the zinc ion remains stably bound in this region, coordinating the carboxylate side chains of both aspartate residues. Interaction with zinc does not disrupt the dimeric structure of the protein or significantly alter the structure of the active site. These data are consistent with experimental studies of HIV-1 protease inhibition by zinc and give strong evidence that this is the binding site that leads to inactivation. Imagesp246-aFigure 1.Figure 2.Figure 3.

Human immunodeficiency virus type 1 protease (HIV-1 PR) is a 99-amino acid, virally encoded protein responsible for proteolytic cleavage of viral gag and gag-pol fusion polyproteins into functional products (1)(2)(3). The activity of the protease is required for viral infectivity in vitro (4). Consequently, HIV-1 PR is an attractive therapeutic target for rational drug design (5,6), and the focus of a tremendous amount of research.
HIV-1 PR has been characterized as an aspartic acid protease based on sequence homology to related enzymes (7, catalytic pH studies (8), and inhibition by wellknown aspartyl protease inhibitors (9,10).
One feature that distinguishes the HIV-1 protease from cellular aspartyl pro-teases is the pH dependence of its activity. The optimum pH range for activity of the HIV-1 protease (4.5-6.0) (8) is considerably higher than for cellular enzymes in vitro (typically 2.0-4.0) (21). A notable exception is human renin, which has a pH optimum in the range of 5.5-7.5 (22). The pH optima of these enzymes presumably reflect the catalytic mechanism generally accepted for aspartyl proteases which requires one of the catalytic aspartate residues to be protonated. Recently,Zhang et al. (23) reported that zinc ions are inhibitors of both renin and HIV-1 PR at neutral pH. Although the zinc binding site(s) that lead to inactivation have not been determined, evidence suggests that one site may occur at or near the catalytic aspartate residues. In this study, molecular dynamics methodology was used to explore the possibility of zinc binding in the active site of HIV-1 PR and predict the effects of this binding on the structure. Our results are consistent with the requirements of zinc binding reported by Zhang et al. (23) and give supporting evidence that this is the site that leads to inhibition of the protease.

Methods
We performed molecular mechanics and dynamics calculations using a modified version ofAMBER3.0 (Revision A) (24,25). The all-atom force field (26) was used for all standard residues, and solvent was treated explicitly using the TIP3P model (27). Parameters for chloride ions and zinc ions were taken from Lybrand et al. (28) and Bartolotti et al. (29), respectively. Electrostatic and van der Waals interactions were treated using a "twin range" (9/18 A) residue-based cutoff, updated every 20 steps. We performed simulations under constant temperature (300 K) and pressure (1 bar) conditions using a 1-fsec integration time step, carried out to 200 psec.
We obtained the starting positions of the heavy atoms for the unbound and Zn2-bound dimers from the crystallographic structure of the synthetic (Aba6795) HIV-1 protease at pH 7.0 (14).
The net charge of the unbound dimer was assumed to be +4, consistent with the normal protonation states of the component amino acids at neutral pH. It is possible that the active site aspartates share a proton near neutral pH, as maximum protease activity occurs in the range 4.5-6.0 (8). However, inflections in the log V/K versus pH profile observed at pH 3.1 and 5.2 suggest one of the aspartates is protonated at pH 5.2. Since we were simulating the protease at neutral pH, we chose to treat both aspartates as being fully charged in the simulations. In the case of the zincbound protease, it is reasonable to assume that Zn2+ ion binding (pKa 9.5) (30) would involve both aspartates in the unprotonated form.
There was no particular bias in placing 2+.. a Zn ion in the active site a priori. Instead, the goal was to find the position of a Zn + ion resulting in the most favorable ion-protein potential energy based on the crystallographic coordinates for the protein heavy atoms (with hydrogens energy minimized by AMBER). Using this position for initial placement of the Zn2+ ion, molecular dynamics may then be used to refine the binding interactions and determine the resulting structural changes in the protein. Initial placement of a Zn2+ ion was determined by a grid search procedure, whereby the AMBER nonbond interaction energy (no cutoff) of the ion was tested on a 0.5 A grid over a 5.0 A box around the crystallographic structure. We used the grid points corresponding to the 1000 lowest energies as starting points for . . . . r * 2 + . minimization of a single Zn ion to a root mean square gradient tolerance of 0.000001 kcal/mol. The position corresponding to the global potential energy minimum was chosen for the placement of the ion. Interestingly, this procedure 99-.I 1 -J -resulted in the placement of the Zn2+ ion in the active site cleft directly between the carboxylate side chains of the catalytic aspartate residues. The net charge of the protease and protease/ion complexes were neutralized with negatively charged chloride ions using the same procedure as for the zinc ion. The solute was solvated in a 13 A water bath (8013 water molecules) and equilibrated using the same procedures as for previous simulations (31 ). Unconstrained dynamics was then performed on each system at constant pressure and temperature to 200 psec. Figure 1 shows the time evolution of the root mean square deviation of a carbon atoms with respect to the crystallographic structure for the molecular dynamics structures of the unbound and Zn2+-bound protease. Both systems are well equilibrated after 150 psec. The unbound protease equilibrates with a carbon root mean square deviation of about 1.5 A, whereas the Zn2-bound structure equilibrates at a lower value of approximately 1.2 A. Secondary structural analysis of the crystallographic structure and the simulation (150-200 psec) average structures was performed using the Kabsch and Sander program DSSP (32). Of the 63 secondary structural assignments made for the crystallographic monomer, 46 (73%) were conserved in each monomer of the unbound structure, and 54 (86%) were conserved in each monomer of the Zn2+-bound structure. 2 Hydrogen Bonding at the Dimer Interface An analysis of H-bond contacts at the dimer interface over the last 50 psec of the simulations has been performed ( Table 1). The primary hydrogen bond interactions at the dimer interface (33)  These include interactions at the amino-and carboxy-terminal g strands (residues 1-A, 95-99), which form a four-stranded antiparallel B sheet, and in the region of the active site triads (residues [25][26][27], which interlock in the "fireman's grip" characteristic of aspartyl proteases (33). In particular, inter-subunit hydrogen bonds involving residues at or near the active site (Leu-24, Thr-26, and Arg-87) were all maintained in the presence of bound zinc. Figure 2 shows the pair distribution function (34), gab(r), for the carboxylate oxygens of the catalytic aspartate residues (Fig.  2a) and water oxygens (Fig. 2b) 16. vation sphere at a distance of around 2.25 A. No exchange of oxygens in the first coordination sphere of zinc was observed during the last 50 psec of the simulation. Zinc has been observed to be coordinated to multiple carboxylate ligands in other enzymes such as DNA polymerase I from Escherichia coli (35). Figure 3 is a stereo picture of the active-site triads at the dimer interface with the Zn ion bound to the catalytic aspartates. The crystallographic structure of Wlodawer et al. (14) is also shown superimposed. Inter-subunit H bonds between Thr-26-*Leu-24' and between Thr-26 <-Thr-26' (primes indicate residues of the symmetry-related monomer), termed the "fireman's grip," are clearly maintained. Additionally, the ODI atoms of the aspartate side chains maintain H-bond contact with the backbone amino group of Gly-27 on the same strand. These H bonds are also pronounced in the crystallographic structure.

Zinc-Protein Interaction
The flap regions of the protease (residues 42-58), which are observed in crystallographic structures to significantly rearrange upon inhibitor binding (16)(17)(18)(19)(20), were not observed to interact with the zinc ion in the simulation. Examination  average structure of the zinc-bound protease indicates the nearest backbone atoms of the flaps reside more than 9 A away from the zinc ion.

Discussion
Kinetic studies have shown that zinc is an effective inhibitor of renin and HIV-1 PR at pH 7, two aspartyl proteases with pH optima near neutrality (23). Other studies of metal ion inhibition of HIV-1 PR performed at pH 5 did not reveal significant inhibition of the enzyme by Zn2+ ions (36). If inhibition of HIV-1 PR is a result of Zn2+ binding at the active-site aspartate residues, it is reasonable to expect that conditions that favor both aspartates in the unprotonated form would enhance Zn2+ binding, and hence inhibition. Because both catalytic aspartates presumably are unprotonated around neutrality, the observed pH dependence of Zn+2 inhibition is consistent with binding at or near the catalytic aspartates. Although the bacterially expressed HIV-1 PR has three residues that commonly bind zinc (Cys-67, His-69, and Cys-95), Zn2+ also inhibits the related HIV-2 protease, which does not contain cysteines or histidines [for a discussion of HIV-2 protease, see Gustchina and Weber (37) and Tomasselli et al. (38)], in an identical manner as the HIV-1 protease (23). This suggests that these residues are not required binding sites for inactivation by zinc. Considered with the fact that inhibition was observed to be first order (23) in Zn2+, we may infer that there is a single important binding site that causes inactivation. Furthermore, inactivation of the protease by zinc was observed to be reversible, as protease activity could be restored upon addition of EDTA (23), which has a higher affinity than carboxylate moieties for Zn2+ ions. These data are, therefore, consistent with the requirement of noncovalent binding of Zn 2 in the active site.
The present study of Zn2+ binding at the catalytic aspartate residues provides evidence that this is the site responsible for inactivation of the protease. Evaluation of the potential energy of a Zn2+ ion in the field of the crystallographic structure suggests a probable binding site for zinc occurs in the active site at the catalytic aspartates (see Methods). Subsequent molecular dynamics simulations predict that the zinc ion remains stably bound in this region; i.e., no exchange of carboxylate oxygens was observed in the first coordination sphere of the zinc ion after initial equilibration. Because the hydrolysis mechanism mediated by HIV-1 PR requires one of the aspartates to be protonated, forming a hydrogen bond with the carbonyl oxygen of the scissile bond and the other in the unprotonated form to activate a water molecule for hydrolysis (6), the coordination of both of these residues by a zinc ion would conceivably inactivate the enzyme. Gel filtration and equilibrium sedimentation analyses suggest that zinc binding does not involve disruption of the dimeric structure of the protease (23). Our simulation with bound Zn2+ shows no weakening of the dimer. The overall simulation results indicate that Zn2+ can bind stably with the catalytic aspartate residues of the HIV-1 protease without significantly disrupting the overall structure. These results are consistent with experimental studies and taken together suggest that the site of Zn + binding that leads to catalytic inhibition is the active site.