Our understanding of protein function, and our ability to modify that function, depends on the X-ray structures being complemented by information about dynamics. Current X-ray refinement models for dynamics are inaccurate, and information is lost. We are developing methods to locate disorder and to model anisotropic motion. A variation of X-ray restrained molecular dynamics is being used as a means of estimating the errors in atomic positions, of searching for alternative conformations, and for evaluating the need to include anisotropic temperature factors. Two structures are used simultaneously in the dynamics procedure, thereby allowing the X-ray refinement to search for multiple peaks in the electron density. The method will be tested using X-ray data for ribonuclease-A (S.K. Burley & G.A. Petsko) and crambin (M.M. Teeter and W.A. Hendrickson). As an application of the method in a functionally relevant situation, we shall study disorder at the active site of myoglobin. It has been shown that pH can be used to tune the predominant structure: at neutral pH there is a strained CO in a closed binding pocket, and spectroscopic results suggest that at low pH is a relaxed CO in an open binding pocket. The lack of X-ray data at extreme pH's has obscured the interpretation of the spectroscopic experiments in terms of the protein structure. We shall measure X-ray data at low, intermediate, and high pH, and use our newly developed refinement methods to model the disorder and correlate it with the spectroscopic data. MD simulations of crambin and myoglobin will be performed in the crystal environment. Crambin is of interest because of the availability of extremely high resolution diffraction data enabling stringent comparison with, and tests of, the simulation results. The myoglobin simulations will be correlated with the pH dependent X-ray studies mentioned above. In order to compare the simulations with experiments on an equivalent basis, the trajectories will be used to generate simulated electron density and structure factors. We will use the simulated data to evaluate models for conformational disorder, as well as the effectiveness of various constraints on the anisotropic temperature factors. In order to clarify the connection between crystallographic data and picosecond dynamic, we will investigate the importance of rigid-body librations in determining the X-ray temperature factors. These motions are expected to have very long time scales, and are likely to be poorly sampled by computer simulations. We have shown that treating the entire protein molecule as a rigid-body is an extremely effective means of refining temperature factors for very large proteins, such as influenza virus hemagglutinin. We will develop a refinement protocol based on this approximation, and we shall estimate an upper limit for the amplitude of the rigid-body motions.
