The mature, active HIV-1 protease is a homodimer that is made up of monomer subunits containing 99 amino acid residues. At least eight FDA-approved anti-protease drugs have been developed. Each of these drugs binds tightly to the dimer active site, thereby inhibiting catalytic activity and preventing propagation of the AIDs virus; however, one unfortunate consequence of targeting a single site on the protease has been the emergence of viral strains which carry multi-drug resistant mutations within their protease molecules. For this reason, there is considerable interest in identifying alternative protease sites that are suitable drug targets. Of particular interest are compounds that bind to the interface of the dimer. Such compounds will block formation of the active protease dimer, but be insensitive to current multidrug-resistant protease variants. In spite of these attractive features, dimerizatioin inhibitors face the formidable challenge of dissociating the tightly bound mature protease homodimer. We have shown that, in contrast with the mature protease, protease constructs, among them those that contain N-terminal extensions (similar to those of the protease precursor, which is embedded within the Gag-Pol polyprotein), have a million-fold larger dissociation constants and are predominantly monomeric at concentrations required for NMR studies. These discoveries suggest that the precursor monomer rather than the mature protease monomer is the target of choice of dimerization inhibitors. We have therefore initiated NMR studies to screen for molecules that interact with models of protease precursor monomer constructs, in order to identify inhibitors of precursor dimerization. Structural information has guided the design of the available anti-protease drugs. In a similar way knowledge of the monomer structure, as well as understanding of inter- and intra molecular interactions that affect dimer stability, should aid in the design of dimerization inhibitors. Therefore, we have determined the structure of the protease monomer. It contains a folded core domain that is common to all monomer constructs that we have examined. The core domain presents preorganized target sites to which small compounds can bind. In addition we have found that mutations in core amino acid residues, which are not in the dimer interface, can destabilize the dimer. These sites are therefore also potential targets for dimer inhibitors. We have continued studies of the dynamics and interactions of the N- and C-terminal strands that form the primary interface of the protease homodimer. This work is aimed at further understanding the mechanism of protease precursor processing. In the static structural model of the protease, derived from X-ray and NMR work, terminal residues 1-4 and 96-99 from both monomers form a 4-stranded b-sheet (the primary dimer interface) to which the exposed autolysis susceptible loop, containing residues 5-9, is connected. However, NMR transverse spin relaxation (R2) dispersion data, hydrogen-deuterium exchange rates and two-dimensional lineshapes provide strong evidence that residues 1-9 of the dimer interface sample two conformations, in dynamic equilibrium. The labile nature of the N-terminus, suggested by the data, is consistent with conclusions described in previous reports on a variety of protease mutants, which revealed that interactions involving the solvent exposed N-terminal strands of the protease are much less important in stabilizing the homodimer than interactions involving the interior C-terminal strands. Furthermore, the proposed dynamic structure rationalizes kinetics data which show that the N-terminal strand folds into the active site of the protease precursor. This conformational switch permits intermolecular cleavage of the immature N-terminus (an early step in Gag-Pol processing). Finally, the flexible N-terminus reveals how the loop containing residues 5-9 becomes fully accessible for autoproteolysis. The model discussed above was derived from data obtained for the protease bound to the inhibitor DMP323. The free protease exhibits R2 dispersion that is qualitatively similar to that obtained for the protease/DMP323 complex. However, the amplitude of the R2 dispersion is significantly smaller in the case of the free protease. Work is currently underway to find the reason for the quantitative difference in relaxation dispersion observed between free and ligand-bound proteases. Analysis of R2 relaxation dispersion profiles can, in principle, provide quantitative information about the chemical shifts and populations of conformations in dynamic equilibrium as well as the rate of the conformational interconversion. Accurate values of these physical parameters provide a basis for characterizing the structural and energetic changes associated with the dynamic process. However , extracting the best values of these parameters and their uncertainties requires careful assessment of experimental errors (both random and systematic) as well as correct application of statistical criteria to ensure that the data are fit with the appropriate statistical model. We have developed a systematic procedure for obtaining best-fitted values of the desired physical parameters and their uncertainties. We show that a judicious choice of initial fitting parameters is required in order that the final fitting parameters correspond the global, rather than local, chi square minimum in the parameter space. In addition, we show that when exchange is outside the fast limit for some sites, robust parameter estimates are obtained by simultaneously fitting 1H and 15N R2 dispersion data of multiple residues in a protein. In contrast, when exchange is in the fast limit for all sites, uncertainties in populations and chemical shifts are large and correlated. This problem can be often be over come by recording data at lower temperature and/or at larger external field, so that exchange is not in the fast limit at all sites.
Showing the most recent 10 out of 14 publications
