(634b) Computational Design of High-Resolution Protein Crystals | AIChE

(634b) Computational Design of High-Resolution Protein Crystals


Gray, J. J. - Presenter, Johns Hopkins University
Jeliazkov, J. R., Johns Hopkins University
Robinson, A., Johns Hopkins University
Garcia-Moreno, B. E., Johns Hopkins University
Berger, J. M., Johns Hopkins Medical Institute
X-ray crystallography can reveal intricate details that underlie protein function. However, the information contained in a crystal structure is dependent on its resolution: high-resolution structures can reveal individual atoms whereas low-resolution structures may not show side chains or small-molecule ligands in their entirety. The inability of crystal structures to resolve these crucial elements limits our understanding of catalytic mechanisms, protein-protein interactions, and drug-protein interactions. Using Rosetta, we have developed a computational design method to stabilize low-energy interfaces from low-resolution protein crystals and subsequently enhance resolution. By studying existing crystal structures and past attempts at rational protein-crystal design, we found that protein–protein crystal contact interaction strength was one key determinant of resolution. We tested six protein design strategies on published rational design results to identify the optimal approach. We found that design strategies with large backbone or rigid-body motions yielded a lower success rates (5-12%), while design strategies with only side-chain repacking followed by minimization yielded moderate success rates (36-50%). We applied our design strategy to a model protein (Staphylococcal nuclease) and experimentally characterized our designed variants. We attempted to crystallize our variants with the standard hanging-drop vapor diffusion technique, using wildtype-like conditions for the reservoir solution. We found that five of the fifteen designed variants crystallized and the rest yielded amorphous aggregate, in agreement with our predicted success rate. However, upon collecting diffraction data and solving the crystal structures, we found that only one variant had improved resolution and the improvement arose from an unpredictable change of space group. Based on these preliminary results, we will update our design strategy. We seek to develop a universal method that can be applied to any protein target with a pre-existing low-resolution crystal structure.