A series of early folding events of BBA1 protein domain.

James Caldwell, University of California, San Francisco

Research Objectives
(1) To study the early stage of the folding processes of small proteins. (2) To identify important folding intermediates by limited folding simulations using locally enhanced sampling (LES). (3) To refine and evaluate the free energy of structure predictions for small proteins. (4) To elucidate the mechanisms of the enzymatic catalysis for beta-lactamases classes A and C as well as inhibition mechanisms of penicillin-binding proteins by beta-lactam antibiotics.

Computational Approach
We use the AMBER6 molecular mechanics simulation program suite and Gaussian98 quantum mechanical calculation packages. The AMBER suite was developed by our group and enjoys a large user community.

Accomplishments
We have significantly improved both the single-CPU and parallel performance of the molecular dynamics module (Sander) of AMBER. Currently for periodic systems, our parallel performance is of the order of 170/256 on the Cray T3E. We have also made significant progress in speeding up the more complex particle mesh Ewald version of Sander as well.

We have recently completed microsecond scale simulations of the folding motion of the small proteins villin headpiece and BBA1. These simulations marked the beginning of the ability of theory to directly simulate the initial stages of protein folding.

We have successfully reproduced the inherent structure of a heptapeptide which has been implicated as a folding nucleation peptide in folding of SH3 domains. Our calculations not only find it stable on a multinanosecond scale, but the calculated NMR spectra of the peptide over the trajectory is in excellent agreement with available experiments.

The quantum mechanical-free energy method that has been under development here is also yielding exciting insights into the nature of enzyme catalysis for the beta-lactamases.

Significance
Elucidation of the mechanism of protein folding has remained a scientific challenge for decades. Molecular dynamics simulation with full representation of solvent possesses a unique advantage to study protein folding due to its atomic-level resolution and accuracy. This method and associated simulation parameters (i.e., the force field) have been tested and refined thoroughly using smaller systems in comparison with many experimental results, although the results of our research may also offer a further critical test of their accuracy. Most proteins take milliseconds to seconds to fully fold, much longer than typical simulation scales, which are presently on the order of nanoseconds. Many methods have been proposed to circumvent this difficulty, including high temperature unfolding, efficient sampling such as LES, and long time dynamics using massive parallelism. These approaches are mutually complementary and cover different parts of the folding process. A combination of these methods can be extremely powerful.

Publications
Y. Duan and P. A. Kollman, "Toward a realistic simulation of the folding process of small proteins," J. Phys. Chem. (in press).

B. R. Krueger and P. A. Kollman, "Molecular dynamics simulations of a peptide from SH3 domain: A possible sequence-function relationship," J. Mol. Biol. (in press).

B. Kuhn and P. A. Kollman, "QM-FE and molecular dynamics calculations on catechol O-methyltransferase: Free energy of activation in the enzyme and in aqueous solution and regioselectivity of the enzyme-catalyzed reaction," J. Am. Chem. Soc. 122, 2586 (2000).