News and Publications

Computer Modeling of Protein and Peptide Structure and Intermolecular Interactions

V.A. Roberts, J.L. Pellequer, M.E. Pique, M.M. Thayer, S.J. Benkovic,* A.E. Karu,** R.J. Nachman,*** M.B. Rittenberg,**** L.F. Ten Eyck*****

* The Pennsylvania State University, University Park, PA
** University of California, Berkeley, CA
*** U.S. Department of Agriculture, College Station, TX
**** Oregon Health Sciences University, Portland, OR
***** San Diego Supercomputer Center, San Diego, CA

The rapid increase in the number of known protein sequences and structures is fueling the need for methods to predict protein structure and intermolecular interactions. We use computational and computer graphics techniques in conjunction with site-directed mutagenesis, protein crystallography, and peptide synthesis to develop testable hypotheses and direct protein engineering.

HIGH-RESOLUTION ANTIBODY MODELS

Our database of superimposed crystallographic antibody structures reveals the structural conservation of both the antibody backbone fold and the side chains that shape the antigen-binding pocket. Using the database, we constructed 3-dimensional models of antibodies to investigate antigen-antibody interactions, the efficiency of somatic mutation, and the mechanism of catalysis. These models provide a structural basis for directing mutagenesis experiments for enhancing binding, selectivity, and catalysis.

Our model of the catalytic antibody 43C9, which efficiently hydrolyzes specific amides and esters, led to identification of key catalytic amino acid side chains and the design of 3 metal-binding sites with potential for interacting with bound antigen. A long-term goal of this project is to build a catalytic metal site in 43C9. Because this goal requires precise structural information, we are collaborating with E. Getzoff, Department of Molecular Biology, in the determination of crystallographic structures of 43C9 and designed metal-binding mutants (Fig. 1).

As part of a research project to develop antibodies for detecting environmental contaminants, we built models of 2 antibodies to benzo[a]pyrene, a carcinogenic polyaromatic hydrocarbon present in cigarette smoke and industrial waste. Both antibodies have exceptionally deep binding pockets created in part by an unusual substitution of a highly conserved tryptophan side chain by a small, hydrophobic side chain. Docking studies indicated that benzo[a]pyrene has 4 close energy orientations in the antibody-binding pocket and has weaker, nonspecific binding outside the binding pocket (Fig. 2). These structural results provide explanations for apparent contradictions in experimental binding studies, suggest why obtaining good polyaromatic hydrocarbon binders from synthetic antibody libraries may be particularly difficult, and target 2 antibody segments for combinatorial mutagenesis to develop antibodies with enhanced binding and selectivity.

We also focus on structures of antiphosphocholine antibodies in which random mutagenesis is used to simulate somatic mutation. Combination of the structural and mutational results revealed that mutations in a loop distant from the binding site can, in conjunction with other mutations, enhance or decrease binding. This finding may explain why this loop shows considerable sequence variation among antibodies despite its distance from the binding pocket. Mutations in this loop may shift the relative geometry between the 2 domains that form the binding pocket, influencing the shape of the pocket.

PREDICTING MACROMOLECULAR COMPLEXES

The computer program DOT was developed to predict intermolecular interactions. DOT performs a complete, 6-degree-of-freedom search of all configurations between 2 molecules. The search algorithm is extremely fast and does not directly depend on the size of the molecules being investigated, factors that allow it to be applied to proteins. DOT is being tested on 2 types of systems: electron-transport proteins, which represent transient protein-protein interactions, and the DNA-repair enzyme uracil-DNA glycosylase, which forms an irreversible complex with the protein uracil glycosylase inhibitor. Results from DOT are being verified by comparing them with the crystallographic structures of the complexes.

ACTIVE CONFORMATIONS OF INSECT NEUROPEPTIDES

Neuropeptides control diverse functions in living organisms. Surprisingly, neuropeptides found in insects often have sequence similarity to mammalian neuropeptides, a finding that suggests the existence of neuropeptide superfamilies with shared conformational determinants. Conformational preferences determined by molecular dynamics simulations combined with structure-activity studies led to the design of constrained cyclic analogs, highly active simplified linear analogs, and the first nonpeptidal ligand for an insect neuropeptide-receptor system.

PUBLICATIONS

Wiens, G.D., Roberts, V.A., Whitcomb, E.A., O'Hare, T., Stenzel-Poore, M.P., Rittenberg, M.B. Harmful somatic mutations: Lessons from the dark side. Immunol. Rev. 162:197, 1998.