ELECTRON DENSITY AVERAGING USING MULTIPLE CRYSTAL FORMS OR DIFFRACTION DATASETS IN STRUCTURE DETERMINATION OF PROTEIN AT MODERATE RESOLUTION. Kalyan Das, Jianping Ding, Yu Hsiou, Karen Lentz, Wanyi Zhang, and Edward Arnold Center for Advanced Biotechnology and Medicine (CABM) and Rutgers University Chemistry Department, 679 Hoes Lane, Piscataway, NJ 08854

Obtaining accurate phases and hence reliable structures using moderate resolution data remains challenging. In the presence of non-crystallographic symmetry (NCS), electron density map averaging techniques have been successfully used to improve the quality of phases. However, many proteins form crystals without useful noncrystallographic redundancy in the asymmetric unit. A multiple electron density maps averaging technique has been developed to improve the phase quality and the interpretability of electron density maps. We averaged the electron density maps computed from structures of HIV-1 reverse transcriptase (RT) crystallized in different crystal forms (i.e., with dramatically different unit cells or space groups). We also averaged the electron density maps calculated from multiple diffraction datasets collected from the same crystal form at different temperatures. For example, the unit cell parameters were slightly different for the crystal frozen at -165deg.C and the same crystal cooled at -10deg.C. In this averaging technique, the protein structures are subdivided into segments that can be superimposed well and the averaging is carried out over the masks covering the segments. The averaged electron density maps for individual subdomains are then combined to cover the whole protein. Also, to minimize the model bias, the conventional omit maps were calculated and used as the input maps for averaging.

This technique was applied to the structure determination of HIV-1 RT in complexes with various nonnucleoside inhibitors (NNRTIs) at 3.0 Å resolution, unliganded HIV-1 RT in two crystal forms at 2.7 Å and 3.5 Å resolution, respectively, and in the structure refinement of the HIV-1 RT/DNA/Fab complex at 2.8 Å resolution. The omit maps and the averaging of the multiple electron density maps have significantly reduced the model bias and improved the interpretability of electron density. The improvement of the phases after map averaging are evident from comparison of electron densities for the bound inhibitors calculated at different stages of HIV-1 RT/NNRTI structure refinements. A detailed description of the algorithm and the results will be discussed. This approach will be useful in solving various other protein structures without NCS symmetry.