Structural, Analytical and Medicinal Chemistry, Pharmacia & Upjohn, Kalamazoo, MI 49001, USA
Structure-based drug design relies on the integration of many experimental methods, structural and sequence databases, and bioinformatics. As an example of a highly successful structure-based drug design effort, the development of potent non-peptidic HIV protease inhibitors will be used. In the development of these inhibitors with good pharmacokinetic properties about 80 crystal structures of protein/inhibitor complexes were determined. The initial protein crystallographic model was obtained from data deposited in the Brookhaven Protein Data Bank. Based on an SAR database dissimilarity study, an Upjohn compound library of approximately 5,000 organic molecules had been setup. An initial search for competitive inhibitors of HIV protease was carried out leading to the identification of warfarin as a weak competitive inhibitor. Then through a similarity search of The Upjohn Company database for related compounds, a better lead (phenprocoumon) with good pharmacokinetic characteristics was next identified. A model for the structure-based drug design paradigm was carried-out with synthetic chemistry, competitive binding assays, crystallography, modeling and biological testing. Crystallography proved invaluable since modeling did not always predict the binding details correctly at key points during the process. Modeling was invaluable in testing various hypothesis before going back to synthesizing new compounds. Starting with the 4-hydroxy coumarins, the central template bound into the catalytic pocket of the protease evolved to pyrones and then dihydropyrones. Through substitutions including various aromatic sulfonamides off the central template into additional binding pockets, the binding of the inhibitors was greatly improved. Later design efforts were directed towards maintaining the tight binding while improving the pharmacokinetic properties and drug efficacy. This entire process was facilitated by an internal database using data entry from the individual scientists of the compound structures, chemical data and biological data allowing the results to be available immediately to the other members of the research team. As a result, orally bioavailable, non-peptidic HIV protease inhibitors were discoverd with in vitro activity comparable to the peptide-derived inhibitors. Various methods have been used to produce a drug candidate that would be less effected by mutations of the protease including information derived from the various sequence databases. In addition, many crystal structures were determined using HIV-2 protease which has only a 50% sequence identity to HIV-1 protease. The crystal structures lend insight as to why these inhibitors are less effected by the usual mutations in the substrate binding region. Through comparison with other aspartyl proteases we can ascertain why these inhibitors are highly specific for HIV proteases. As a consequence, HIV-1 isolates highly resistant to the peptide-derived inhibitors remain sensitive to these third generation inhibitors. Finally, from the unusual binding of some of the inhibitors, binding of transition-state analogues, binding of accidental binding of auto-catalytic peptide fragments and binding of small synthetic peptides, the mechanism of action of aspartyl proteases is further clarified.