Cundari Th.R. -- Computational Organometallic Chemistry-0824704789

might be involved in metal ion chelation at the zinc finger region in the amino terminus of IN (18).

Subsequent work with flavones and CAPE analogs identified other active inhibitors (19,20). The ability of these molecules to inhibit the disintegration reaction catalyzed by integrase lacking the amino-terminal domain clarified that inhibitors bind in the catalytic core domain. Speculation on the role of metal ions in inhibition moved to the organometallic active site. The role of the metal ion in inhibition was questioned by the discovery that Cu2 -phenanthroline complexes (Fig. 2, 4) were also able to inhibit IN (21). The inhibition observed was selective for copper over magnesium, indicating that the inhibition by the phenanthroline complex did not involve the divalent cation in the organometallic active site. Independent observation that most polynucleotidyltransferases contain two

˚

organometallic active sites separated by 4.4 A, as well as the commonly found diaryl functionality in known inhibitors, led to explorations of linker type and length between two metal chelating functionalities (Fig. 2, 5) (22–24). Computations on the strength of the cation–π interaction of benzene and catechol with the divalent cations of manganese and magnesium were undertaken in 1997 (25). These calculations showed that the interaction energy is much larger than for previously studied cation–π interactions with monovalent cations and provided no reason to rule out such interactions in the organometallic active site of IN. Experimental work probing metal ion selectivity showed that different metal cations can be used to separate the stages involved in the activity of IN (26). Integrase activity can be divided into an assembly stage, during which the enzyme complexes with DNA to form a preintegration complex, and a catalytic stage, during which the 3′ processing and strand transfer reactions occur. It was shown that although both magnesium and manganese can serve as cofactors for both stages, calcium promotes only assembly and cobalt promotes only catalysis. Further studies used sequential addition of calcium and cobalt to allow addition of inhibitors between the stages as well as prior to both. These studies showed that selected inhibitors, many containing polyhydroxylated aromatic systems, were effective only if added prior to the assembly stage and showed no effect on catalysis in a preassembled complex. These results neither support nor contradict the hypothesis that the inhibitors interact at the active-site metal ion. Additional research identified tetracyclines (Fig. 2, 6) as a new class of IN inhibitors (27). Not only was a new class of inhibitors identified, their inhibition was also evaluated in the presence of both manganese and magnesium ions. No difference in potency was observed, although the earlier computational work had identified a significant difference in gas-phase interaction energies between aromatic systems and these two metal ions (25). These data seem to add support to a lack of a direct interaction between the organometallic active site and enzyme inhibitors. Our understanding of the metal ion role in the inhibition of IN was further muddied in 1998. Neamati and coworkers published the discovery of another class of

IN inhibitors, the salicylhydrazines (Fig. 2, 7) (28). These inhibitors, however, showed selective inhibition in the presence of manganese, but not magnesium. This result indicates that the metal ion can certainly influence the binding of some inhibitors, although not specifically proving that the metal ion must interact directly with the inhibitors. No comparative structural information is currently available for the IN protein containing different divalent metal ions in the active site. Thus the metal ion may influence the shape of a distant inhibitor binding site (an allosteric site). A final result to add to this history is the recent publication of another structural class of inhibitors that show no preference between manganese and magnesium, the thiazolothiazepines (Fig. 2, 8) (29).

The prior history of evidence both supporting and contradicting a direct role for metal ion interaction with IN inhibitors demonstrates the need to perform additional research on this issue. This chapter presents our multipronged computational approach to elucidate the exact involvement of the organometallic active site in the binding of inhibitors to the catalytic core of the IN enzyme. Quantum mechanical calculations are applied to examine differences in metal ion interactions with inhibitors and differences in metal ion interactions with the IN active site. Molecular mechanics calculations are used to explore the best geometric (steric) and electrostatic fit of the IN inhibitors to the IN catalytic core. Current results from each of these three avenues of investigation are described in the following sections.

2. METAL ION COMPLEXES WITH INHIBITORS

This section describes our studies on the interactions of Mg2 and Mn2 with two classes of inhibitors, the salicylhydrazines (Fig. 2, 7) (28) and the thiazolothiazepines (Fig. 2, 8) (29). Inhibitors used in these studies as well as in the docking studies described later are shown in Table 1 along with their biological activities. The salicylhydrazines are of interest because they show an inhibitory effect only when assayed with Mn2 as the active-site metal ion. The thiazolothiazepines, on the other hand, are equally active with either Mg2 or Mn2 present in the active site. Clearly the metal ion has an impact on inhibition by some inhibitors. The modeling studies described in this section seek to evaluate two different mechanisms through which this impact could arise from the direct metal ion interaction with the inhibitors that has been postulated in the literature since 1993. First, the impact could be due to geometric differences that arise from alternate sites of chelation to the inhibitor for different metals. Second, a direct interaction with the metal ion might be reflected in a correlation between the strength of the interaction and the biological activity of inhibitors in a structural class. It is entirely possible that the metal ion impact is a complex mix of these two factors, although that possibility is not tested in this work.

TABLE 2 PM3(tm) Heats of Formation for Metal Ion Complexes of Integrase Inhibitors

a ∆∆Hf ∆Hf (π) ∆Hf (σ).

ences are seen for Mn2 . There is a 24-kcal/mol preference for π over σ complex formation.

These results fail to provide support for a mechanism of direct metal ion interaction in which geometric differences for complexes of different metals lead to differences in inhibition of the enzyme.

Further studies were performed using the single preferred chelation site for each metal interacting with the salicylhydrazine series of inhibitors. These studies evaluated the metal ion binding energy using the PM3(tm) semiempirical method as well as the 3-21G* ab initio basis set in the Spartan program. The metal ion binding energy was determined by subtracting the energies of the isolated ion and the isolated inhibitor from the energy of the ion:inhibitor complex. The semiempirical calculations provided energies in the form of heats of formation, whereas the ab initio calculations provided electronic energies. However, the trends in the differences are expected to be similar by both methods. The results of the calculations on 10 salicylhydrazines are shown graphically in Figure 4. The correlation coefficients (R2) for Mn2 binding versus biological activity determined in the presence of Mn2 are 0.211 and 0.416 by the semi-empirical and ab initio methods, respectively. The correlation coefficients for Mg2 binding versus biological activity determined in the presence of Mn2 are essentially the same, 0.284 and 0.405 by the semiempirical and ab initio methods, respectively. These data clearly show that the correlations for both metal ions are the same at a given level of theory even though biological activity is only observed experimentally in the presence of Mn2 . Thus a direct interaction with the metal ion is not reflected in a correlation between the strength of the interaction and the biological activity of these inhibitors.

The current results provide no support for either mechanism by which the metal ion impact on inhibition could arise from direct interactions between the metal ion and the inhibitors. The results reported here used medium-sized basis

FIGURE 4 Metal ion binding energies shown as a function of biological activity measured in the presence of Mn2 . Top: Semiempirical results. Bottom: Results of 3-21G* calculations.

sets and did not include a full solvation sphere for each metal ion. Additional studies are under way using larger basis sets as well as explicit water molecules completing the inner coordination sphere. The next section describes studies in progress designed to assess the impact of metal ion differences on the integrase active site.

3.INFLUENCE OF METAL IONS ON THE INTEGRASE ACTIVE SITE

The interactions of Mg2 , Mn2 , Ca2 , and Co2 with the active site of IN were investigated. These metal ions clearly have different impacts on the ability of the enzyme to assemble and catalyze the strand transfer reaction as well as showing selectivity toward some structural classes of inhibitors (26,28,29). These effects may be due to minor geometric differences in the inner coordination sphere that propagate into nearby regions of the enzyme structure, thus affecting allosteric sites at which other viral proteins involved in the preintegration complex need to interact. Because Ca2 is only able to promote assembly, and Co2 is only supportive of catalysis, these two ions are expected to induce the structural extremes, with Mg2 and Mn2 having more similar structural impact. The crystallographic complex of the IN catalytic core with Mg2 shows that two carboxylates from aspartate residues 64 and 116 chelate in an η1 fashion with the metal ion (31). Four water molecules occupy the remaining sites of the octahedral inner coordination sphere. The crystallographic geometry with hydrogen atoms added by the MOE program (32) was used to initiate quantum mechanical optimization of the active site with each metal ion. Optimizations were performed at the Hartree–Fock level of theory using the SBK basis set (33–35) and effective core potentials (36) as implemented in the GAMESS program (37). Figure 5 shows the truncated model of the active site that was included in these computations.

FIGURE 5 Truncated active-site region modeled, with M Mn2 , Mg2 , Ca2 , and Co2 .

Differences in the calculated geometries of the active-site model with different metal ions were modest, in agreement with crystallographic results for the avian sarcoma virus (ASV) integrase with Mn2 , Mg2 , Ca2 , Zn2 , and Cd2 (38,39). Our results indicate that the distances between the metal ion and the

and Ca2 complexes, respectively. Distances between the metal ion and the water

complexes, respectively. Thus distances to the metal ions follow the same trend regardless of the ligand. The distances to the transition metals, cobalt and manganese, are consistent with those determined in an evaluation of the PM3(tm) method, in which the bond lengths between these metals and water ligands in octahedral complexes were predicted within 5% and 3% of the experimental ref-

erence value (40). A more significant difference can be seen in the angle formed at the metal ion by the bonds to the two acetate ligands. These angles are 85.0°, 103.5°, 106.1°, and 107.9° in the Ca2 , Co2 , Mg2 , and Mn2 complexes, respec-

tively. This angle in the calcium complex is clearly quite different from that observed in the three other complexes. The previously mentioned crystal studies of the ASV IN, in fact, identified that the calcium complex was not actually octahedral, but had only three water ligands (38,39). This difference may relate to the inability of the integrase enzyme to perform its catalytic function with calcium in the active site. These results, however, do not clearly delineate a reason why cobalt is unable to promote assembly of the viral complex even though it is able to facilitate catalysis. An overlay of the four optimized model active sites is shown in Figure 6.

FIGURE 6 An overlay of the model active sites containing Mn2 , Mg2 , Ca2 , and Co2 .

Additional calculations are under way with full aspartate amino acid residues rather than acetate ligands. Finally, calculations need to be done with Ca2 as an incomplete octahedron, as was observed in the ASV IN crystal structure (38,39). The quantum mechanically optimized geometries of the active-site models will be used to determine if the small changes induced by the presence of different metal ions have an effect on the larger HIV IN catalytic core structure. The optimized active-site geometries will be fitted back into the catalytic core and will be held fixed during molecular dynamics simulations that allow the remainder of the IN structure to adapt to the different metal ion environments. It will be of interest to see whether or not molecular dynamics calculations will perpetuate minor geometry differences around the metal ion into the nearby catalytic loop region. These studies may provide insight into a possible allosteric role of the metal ion in IN inhibition by certain inhibitor classes.

4.DOCKING INHIBITORS TO THE INTEGRASE CATALYTIC CORE

An independent series of calculations was performed to explore the steric and electrostatic complementarity of the salicylhydrazine inhibitors for different regions of HIV IN. These calculations used empirical docking and binding affinity evaluations to find favorable regions on the crystal structure of the IN catalytic core (31) for inhibitor binding. Two regions of the IN catalytic core were emphasized in these docking studies, the region around the metal ion and the region above the highly flexible catalytic loop. The active-site metal region and the catalytic loop are labeled in Figure 7. The direct and indirect evidence for a role of the metal ion in both catalysis and inhibition provided motivation to closely examine the inhibitor interactions with the active site. The crystal structure of an HIV IN inhibitor complexed with ASV IN (41) that showed inhibitor binding above the catalytic loop provided motivation to closely examine this second region. Figure 7 shows an overlay of the HIV IN catalytic core crystal structure used in our docking studies on the ASV IN complexed with an HIV IN inhibitor. This figure demonstrates that differences between the two structures are most significant in the region occupied by the inhibitor. Thus the ASV IN crystallographic complex does not definitively identify the inhibitor binding site in the HIV IN structure.

Several methods are available to explore the conformations and configurations of a small molecule in the environment of a larger, rigid biomolecule. These are called docking methods. One method used in these studies is the docking module within the MOE program. This docking method precalculates a grid of steric and electrostatic interaction energies for probe atoms within a user-defined volume, the docking box, containing all or only part of the biomolecule. Pre-

FIGURE 7 Overlay of the ASV IN:inhibitor crystallographic complex on the HIV IN crystal structure.

computation of these interaction energies allows for rapid evaluation of ligand orientations in the environment of the biomolecule by summing the steric and electrostatic energies at grid points occupied by ligand atoms. Random initial ligand conformations and orientations are subjected to Monte Carlo optimization of the interaction energy. A second docking method was also used in order to investigate if a consensus on the optimal inhibitor binding site would be achieved. The second docking method uses fast Fourier transforms to optimize the orientations of a rigid ligand in the environment of a rigid target biomolecule using geometric complementarity (42). These two docking methods have complementary strengths and weaknesses. The docking method implemented in the MOE program offers the advantages of ligand flexibility and consideration of electrostatic complementarity. The fast Fourier transform method has the advantage of speed and can be used to scan the entire protein surface for favorable binding pockets.

In order to determine the most likely location for ligand binding in the HIV IN catalytic core, it is necessary not only to fit the ligand into potential binding sites, but also to energetically evaluate the resulting complexes, known as scoring. Both docking methods generate multiple geometries for the complex. Fifty geometries were generated by MOE and 100 by the fast Fourier transform in these