|
|
|
||
|
||||
|
|
|||
Traditional drug discovery techniques seek compounds that maximally inhibit a single target enzyme. In most cases, this is an effective approach. Most enzymes are highly specific for their substrate, so an inhibitor of similar shape and chemical nature will bind tightly to the active site. This approach has been widely used to design inhibitors for diverse enzymatic targets, including HIV-1 protease (Fig. 1). HIV-1 protease, however, is somewhat different than most typical enzymes: it binds to and cleaves a collection of different peptides. This broader specificity makes resistance mutation possible, as the active site is remodeled to reject inhibitors but still bind to the appropriate peptide sequences needed for viral maturation. In contrast, the coevolutionary method evaluates each inhibitor for effectiveness against all possible mutant proteases, and searches for the one inhibitor that performs the "best" against the entire range (the term "best" was quantified using a minimax evaluation). A holistic inhibitor was found by this approach which is predicted to bind to a broad range of complex mutants [1].
 |
Figure 1: Complex of HIV-1 protease with a substrate. HIV-1 protease is represented by molecular surfaces and ribbon diagrams with subunit A in red and B in blue. The eleven protease residues that were allowed to mutate in this study (D25, G27, A28, D29, D30, V32, I47, G48, G49, I50, and I84) are colored orange and cyan, respectively. A ball-and-stick representation of a natural polyprotein substrate (SQNYPIVQ) is bound in the active site, with the backbone in yellow and side chains in white. (a) View of the entire complex. (b) Close-up of the active site, with the eight subsites (denoted S4 to S1 and S1' to S4') labeled. Peptides bind to HIV-1 protease in extended form with eight contiguous residues on the peptide, labeled P4 to P4', making contact with the eight enzyme subsites S4 to S4'. I created the images with the Python Molecular Viewer. |
I then developed computational and visualization tools based on the coevolution algorithm to analyze the evolutionary tree of mutants that develop when HIV-1 is challenged with protease inhibitors (Fig. 2). These new tools also adressed the restrictions of the previous approach to search for peptide side chains only. I extended the evolutionary search to process any kind of inhibitors, not just peptide-based ones.
Figure 2: Evolutionary tree diagrams. (a) The set of all viable HIV-1 protease mutants that cleave the polyprotein processing sites in the absence of inhibitors. For ease in comparison, HIV-1 protease mutants are ordered by location within the active site, counterclockwise from S1 to S4 starting at the 3 o'clock position. For each site of mutation, the possible mutants are sorted by amino-acid size, starting with the smallest, glycine, moving counterclockwise to the largest, tryptophan. In this diagram, each mutated protease is assigned a unique position which is used in all subsequent diagrams allowing easy comparison with the diagrams in other figures. Mutated proteases with a viability greater than 5% are represented by a dot which is connected by lines to viable parents and children. Protease mutants with a viability less than 5% are not drawn, leaving holes in the diagram. The colors represent the viabilities of the proteases: dark-blue for low viabilities of 5-10%, light-blue for 11-25%; green for 26-50%, yellow for 51-75%, orange for 76-99% and red for 100% viability and higher. The single letter amino acid codes, such as I50 or A28, refer to the one-site mutants at the first step of each branch. (b) HIV-1 protease resistance when inhibited against the theoretical best inhibitor. Two resistant strains with low viabilities remain if challenged with this highly effective theoretical inhibitor. Yellow numbers indicate one-site, two-site and three-site mutations, "wt" indicates the wild-type protease at the center of the diagram. The individual data points are in close proximity, making it difficult to distinguish them (left panel). In the right panel, the outermost data points (i.e., the three-site mutants) are equidistantly spaced around the circle, allowing access to individual mutants in the large dataset.
The evolutionary diagrams allowed us to analyze the susceptibility of a trial inhibitor to resistance as the virus mutates, and to investigate the relative importance and roles of individual protease mutants in this evolution [2]. Comparing diagrams of individual drugs also provided a powerful method to predict which drug combinations might be the most efficient when applied to a given patient. This allowed us to analyze the susceptibility of existing drugs (Fig. 3) used for treatment of AIDS, and making suggestions for modifying these drugs to improve their robustness in the face of viral resistance mutation.
Figure 3: Drug resistance patterns for saquinavir and ritonavir. (a) Ball-and-stick representations of saquinavir (filled bonds) superimposed on HIV-1 protease natural substrate (SQNYPVIQ, wireframe bonds). The saquinavir backbone is cyan and side chains are green, and side chain atoms which reach into a neighboring protease subsite are red. Saquinavir occupies subsites S3 through S3', leaving S4 and S4' empty. (b) Drug resistance pattern of saquinavir depicted as evolutionary tree diagram. D29NE, G48ASC, and I84N mutants are the most resistant when inhibited by saquinavir. (c) Drug resistance pattern for an optimized version of saquinavir. This theoretical compound lowers viabilities of almost all saquinavir-resistant mutants and completely eliminates the V32ILM resistance mutants. However, it triggers a weakly viable I50V mutant and increases viabilities of G27A and I84F mutants. (d) Ball-and-stick representation of ritonavir, which occupies subsites S3 through S4', leaving S4 empty. (e) Drug resistance pattern of ritonavir. Ritonavir is the most susceptible of the five inhibitors used for this study, allowing development of many resistant strains of the virus, including G27A, D30EN, V32ILM, G48C, and I84NF. (f) Drug resistance pattern of an optimized version of ritonavir. The optimized ritonavir is more efficient than the optimized saquinavir because it offers an additional side chain (P4) to optimize.
The side chain optimization studies highlight a few sites for easy resistance development in HIV-1 protease, which provide quick ways to improve the robustness of a given inhibitor. The P3 side chain is particularly easy to see. Ritonavir was designed to fill the S3 site, with a very large P3 side chain. This allows the virus to develop resistance easily with G48 mutations, by simple constriction of the S3 subsite. Knowing that this mutation is probable, we can reduce the ritonavir P3 side chain, lowering G48 resistance from 100% to 5%, and evading this resistant mutant before it develops. Looking at all six of the inhibitors tested here, we also see the advantages of P4 and P4' side chains. These positions add binding strength, and as long as they are kept small, they are free of susceptibility to resistance mutation. Of course, these advantages must be weighed against any pharmacological disadvantages of larger compounds.
For further information, contact me or read [2].
Discalimer: to keep this article short, I show only the first two of six drugs tested in this study.
References:
1. Rosin, D.C., Belew, R.K., Morris, G.M., Olson, A.J., and Goodsell, D.S. (1999). Coevolutionary analysis of resistance-evading peptidomimetic inhibitors of HIV-1 protease. Proc Natl Acad Sci USA 96, 1369-1374.
2. Stoffler, D., Sanner, M.F., Morris, G.M., Olson, A.J., and Goodsell, D.S. (2002). Evolutionary analysis of HIV-1 protease inhibitors: Methods for design of inhibitors that evade resistance. Proteins 48:63-74.
Homepage Daniel Stoffler: Home | Personal
| Projects | Publications
| Links