The principal objective of this project is a quantitative description of the physical chemistry that drives the amino acid sequence of staphylococcal nuclease into its three dimensional structure. Experimental studies of structure that persists when nuclease is not folded will employ NMR spectroscopy to measure long range structural features reflected in residual dipolar couplings (RDCs). Previous RDC data have demonstrated that a native-like topology"""""""" persists in denatured nuclease, even in the presence of 8 M urea after mutation of 10 large hydrophobic residues. While the argument is compelling for this conclusion, a much more quantitative understanding of the information in these couplings is needed to complete the picture of this poorly understood ensemble of many inter-converting conformations. Two strategies of data interpretation will be pursued that do not rely on single structures for representation of the ensemble average structure. To achieve the most detailed structure possible, many sets of RDCs will be collected with different alignment tensors, using electric fields or chemical modification to alter the alignment tensor. Attempts will be made in staphylococcal nuclease and three other proteins (ubiquitin, eglin C, and fyn-SH3 domain) to demonstrate a native-like topology in compact denatured states by direct correlation of dipolar couplings from the native and the denatured states. A novel strategy for predicting the structure of new protein folds, based on modeling side-chain/backbone interactions with phi/psi/chil propensities, will be pursued. Initial successes at CASP5 suggest that better sampling of the conformations of turns between helices and strands could lead to significant advances in predicting new folds at low resolution. Recently developed statistical potentials for phi/psi/chil angles and for local side-chain/side-chain interactions will be combined with torsion angle dynamics and applied to the prediction of protein structures at higher resolution, in refinement of both de novo models and homology models.
Showing the most recent 10 out of 29 publications
