| Abstract | The ever-increasing desire to study challenging biological systems by X-ray crystallography often leads to weakly diffracting crystals with minimum Bragg spacings (dmin) significantly below 3 Å. Such diffraction data pose challenges to existing structure solution and refinement methods since the data are often noisy and the observable to parameter ratio is very low. We employed significantly modified and enhanced methods in order to solve and refine a particular set of low resolution structures for the ATPase p97/VCP, consisting of a tandem pair of ATPase domains (D1, D2), and an N-terminal domain. Based on the known structure of the N-D1 fragment, the combination of molecular replacement phasing and SeMet phasing of the entire molecule allowed approximate building of the D2 domain. Modification and enhancement of the bulk solvent model and B-factor sharpening of electron density maps were essential for successful refinement. Reasonable Rfree values (below 32%) were achieved upon refinement that included experimental phase information sing the MLHL target along with judicious use of NCS and secondary structure restraints at diffraction limits as low as 4.7 Å. At this resolution, the topology and the backbone chain trace of the D2 domain was visible, some side chain positions could be assigned, and the type of bound nucleotide determined. Furthermore, large conformational changes could be discerned by comparison of structures solved in three nucleotide states ATP (at 3.5 Å resolution), ATP-AlFx (at 4.5 Å resolution), and ADP (at 4.25 Å resolution. However, the secondary structural elements of the D2 domain showed significant deviations from canonical ATPase structures. Recently, we were able to solve the D2 domain alone at 3 Å resolution with greatly improved geometry over existing models of active AAA domains. The revised model of D2 in conjunction with the high-resolution structure of the N-D1 fragment were used as starting models for re-refinement against the low-resolution structure factors for full-length p97 in the three nucleotide states. Although only a few additional residues were added to the models, the revised full-length models show significant improvement in both model geometry and R values with final Rfree values below 30% compared to the original models. The free R value dropped as much as 5%, indicating that there is information in the diffraction data even at ~ 4 Å resolution that objectively assesses the quality of the model even though the low resolution electron density maps do not give sufficient clues how to improve the model. Thus, one approach to address this vexing problem is to determine the structures of (sub)-domains or fragments at high resolution of a macromolecule that only gives raise to low-resolution diffraction. An alternative and more general approach would be the appropriate use of homologous structures where available. We are currently developing and testing a method that allows one to incorporate structural knowledge of homology models into the refinement process at low resolution. We have developed a general geometry-based algorithm that efficiently samples conformational space under constraints imposed by low-resolution density maps obtained from electron microscopy or X-ray crystallography experiments. A deformable elastic network (DEN) is used to restrain the sampling to prior knowledge of an approximate or homologous structure. The DEN restraints dramatically reduce over-fitting, especially at low resolution. Cross-validation is used to optimally weigh the structural information and experimental data and adjust the deformability factor. Our algorithm is robust even for noisy density maps and it has a large radius of convergence. The DEN refined structures are much superior in terms of free R value, phase difference and coordinate rmsds to the known structure, and Ramachandran statistics compared to refinement without DEN. |
