Vol. 47 No. 4/2000 1061–1066 QUARTERLY Communication

Physiological and pathological roles of cysteine proteases make them important targets for inhibitor development. Although highly potent inhibitors of this group of enzymes are known, their major drawback is a lack of sufficient specificity. Two cysteine protease covalent inhibitors, viz. (i) Z-RL-deoxo-V-peptide-epoxysuccinyl hybrid, and (ii) Z-RLVG-methyl-, have been developed and modeled in the catalytic pocket of papain, an archetypal thiol protease. A number of configurations have been generated and relaxed for each system using the AMBER force field. The catalytic pockets S3 and S4 appear rather elusive in view of the observed inhibitors' flexibility. This suggest rather limited chances for the development of selective structure-based inhibitors of thiol proteases, designed to exploit differences in the structure of catalytic pockets of various members of this family.

Although highly potent inhibitors are known, their major drawback is a lack of sufficient specificity to allow the targeting of a particular thiol protease.The inspection of the reported enzyme-inhibitor structures indicates a wide dispersion of the ligand binding modes (Drenth et al., 1976;Varughese et al., 1989;Stubbs et al., 1990;Yamamoto et al., 1991;1992;Kim et al., 1992;Schroder et al., 1993).This, in turn, strongly suggests the absence of a general principle governing binding of the inhibitor.In order to analyze in detail the binding of a specific inhibitor, we have modeled two ligands, developed in our laboratory (Hall et al., 1992), in the active site of papain.Thus, the aim of the study was to evaluate the mobility of two new inhibitors of thiol proteases in the catalytic pocket of papain.
Both subjects of the present study: (i) Z-RL-deoxo-V-peptide-epoxysuccinyl hybrid (further referred to as InhA), and (ii) Z-RLVG-methyl-(called InhB), were developed on the assumption that a reactive group responsible for the covalent linkage (i.e., epoxysuccinyl or diazomethyl) should be combined with a peptide fragment, modeled after the cystatin C N-terminal sequence (residues R 8 LVG 11 ) which determines structural features responsible for tight non-covalent binding of cystatins to thiol proteases (Hall et al., 1992;Czaplewski et al., 1999).

COMPUTATIONAL METHODS
All simulations were carried out using the AMBER 5.0 program (Case et al., 1997).The all-atom AMBER force field was used (Cornell et al., 1995).
Starting coordinates for all heavy atoms of papain were obtained from the crystal structures from aqueous solution, Protein Data Bank, PDB (Bernstein et al., 1977), files 1pe6 and 1pad for InhA and InhB, respectively.The topology and definitions of particular residues of both protein-ligand complexes are given in Fig. 1.The design of the starting model and development of specific constraints for the papain-InhA complex (not having the Cbz residue) are described in detail elsewhere (Czaplewski et al., 1999).The initial model for the papain-InhB complex was designed in Sybyl 6.1 program (1994, Tripos, Inc.) on the basis of point-substitutive mutations of Acaapack (Ac-AAPA-CH 2 ) complex (1pad), and a set of specific constraints for this starting model was set (Table 1) to maintain the papain-inhibitor interactions (Drenth et al., 1976).New residues, absent in the original AMBER force field, have been parameterized according to standard procedures (Bayly et al., 1993).
The details of modeling procedure have been described in our previous article (Czaplewski et al., 1999).Briefly, a sequence of C a -positional and "softly" (as indicated in Table 1) constrained minimization and simulated annealing in vacuo, was followed by a series of C a -positional constrained minimizations in vacuo and in water, and subsequent unconstrained minimization in water.A consecutive unconstrained thermalization preceded a 230 ps molecular dynamics (MD) runs of the complexes in water, under periodic boundary conditions in a closed, isothermal, isobaric (NTP) ensemble.Throughout the simulation the sol- ute and solvent were coupled to a constant-temperature (T = 300 K) heat bath and a constant-pressure (P = 1 bar) bath (Berendsen et al., 1984).Counterions (10 Cl -) to neutralize the system were added using AMBER 5.0 program LEAP.All bonds were constrained using the SHAKE algorithm (Ryckaert et al., 1977) allowing a time step of 1 fs.A double residue-based cutoff distance of 10/14 Å was used for nonbonded interactions.The TIP3P model was used for water molecules (Jorgensen et al., 1983).A typical box size was 78 66 ´60 Å.Approximately 8200 TIP3P water molecules were in the box, i.e. the whole system consisted of a total of 28 000 atoms.

RESULTS AND DISCUSSION
The initial structures of the papain-ligand complexes were subjected to the constrained simulated annealing so as to perform the simulation at a very high, physically unrealistic temperature.The additional kinetic energy enhances the ability of the system to explore the energy surface and can prevent the molecule from getting stuck in a localized region of conformational space.Afterwards the system was thermalized and subjected to 230 ps of unconstrained molecular dynamics at 300 K.
Time-averaged residue-based deviations as a function of residue number for all molecular dynamics runs are shown in Fig. 2. Changes up to 5 Å were observed for some residues (C-terminus), but the overall C a mobilities figured approximately 1 (for InhA) and 2 (for InhB) Å.The average structures displayed similarly significant changes in some flexible loops of the enzyme.It should be stressed that the protein structures reproduced very well the mobility pattern over the whole molecule, represented experimentally by the atomic temperature factors in the crystal structure of the papain-E-64c complex (Kim et al., 1992).This result validated the use of the AMBER 5.0 force field as a suitable tool for scanning the conformational space of both ligands nested in the catalytic cleft of papain.
The detailed atomic-level analysis of mobilities of the inhibitors' backbones reveals that the dispersion of the relaxed positions of residues increases steeply towards the N-terminus of the inhibitors (Fig. 3).The catalytic pocket S 3 , as defined by the pioneering studies of Schechter & Berger (1967), appears rather elusive in view of the evident inhibi-Vol.47 Cysteine protease inhibitors 1063 Table 1.The two sets of "soft" constraints for the papain-InhB starting complex.
A specific atom pair was assigned to one or the other constraint category, in accordance with the initial distance linking the constrained atoms.The harmonic force-constant in both sets of constraints equaled 100 kcal/mol., 1993).The location and definition of substrate binding site S 4 is even more questionable.These features are also reflected in the two selected binding modes of papain-InhA and papain-InhB, given in Fig. 4. The lack of a single site in space that would be clearly defined for binding of inhibitor's main chain may indicate that the surface of the papain-inhibitor interaction spreads over a relatively large area.Keeping in mind the clear diversity in the architecture, accommodated by covalent inhibitors in experimental structures, and the fact that the exact location of the residues and the subsites more distant to the catalytic cysteine cannot be clearly identified (Turk et al., 1998), it should be emphasized that the localization and definition of the substrate binding sites S 3 and S 4 is precarious.As the substrate binding sites cannot be precisely localized, the term "binding site" may be somewhat misleading.
These findings prompt us to suggest that the considerably weak definition of catalytic pockets S 3 and S 4 of cysteine proteases predicts rather limited chances for the development of selective structure-based inhibitors of thiol proteases, designed to exploit tiny differences between the catalytic pockets of various members of this family.Blue and yellow indicate low and high atomic mobility limits, respectively.The catalytic dyad (C25 in blue and H159 in red) is also shown.The obvious scatter of the inhibitors covering a considerable part of space can be easily seen.

Figure 2 .
Figure 2. The time-averaged C a -based deviations (mobilities) along the papain sequence accompanying MD runs, solid line.Symbols A1 and B1 refer to InhA and InhB MD runs, respectively.The distribution of the C a temperature factors (as given in the Protein Data Bank file 1pe6), being an independent experimental measure of the papain-chain mobility, is shown as a dotted line.A correlation between the two plots is evident.It can be seen that some loops of the papain L lobe undergo high fluctuations during the molecular dynamics runs.

Figure 3 .
Figure 3.The time-averaged C a -based (except the Cbz and Oxi, residues, for which the benzyloxy oxygen and sulfur, respectively, serve as probes) mobility of inhibitors' individual residues during the molecular dynamics simulations.

Figure 4 .
Figure 4.The selected binding modes of relaxed InhA and InhB configurations (black lines), resulting from random sampling of MD productive runs.