INDICATION OF MODEL-BIAS AT ATOMIC RESOLUTION: EXPERIMENTAL PHASING OF AN EF-HAND PROTEIN AT 1.05 A
D. E. Brodersen1, E. de La Fortelle2, G. Bricogne2, J. Nyborg1, and M. Kjeldgaard1
1Macromolecular
Crystallography, Inst. of Mol. and Struct. Biol., Aarhus
University, Gustav Wieds Vej 10c, DK-8000 Aarhus C, Denmark.
Email: ding@imsb.au.dk
2MRC-LMB, Hills Road, Cambridge,
CB2 2QH, England
Keywords: Calcium-binding protein, SAD, phasing,
model-bias
A set of experimental phases not biased by any atomic model has been obtained to a resolution of 1.05 A by single wavelength anomalous dispersion (SAD) using the statistical phasing program SHARP [1, 2]. The resulting electron density map is of outstanding quality and offers a unique possibility to gain insight into high-resolution features of proteins such as local disorder, conformational variations, thermal motion, and preferred geometry. Moreover, the data indicate that significant bias towards model phases is likely to occur during refinement of macromolecular structures even at high- to atomic resolution [3]. Traditional conjugate gradient LS refinement including anisotropy and riding hydrogens was done in SHELX and carried out against the same SAD data with separate Friedel mates.
Psoriasin is a small (2*11.7 kDa) dimeric protein belonging to the S100 class of calcium-binding EF-hand proteins [4-6]. In order to obtain very accurate phase information, the two intrinsic calcium ions of the dimer were substituted before crystallisation for the lanthanide element holmium (67Ho) which is similar to calcium in ion radius and preferred coordination geometry. A single wavelength anomalous dispersion (SAD) data set was collected at 0.9091A (far from any of the L absorption edges) to a resolution of 1.05A spacings from a crystal of the metal ion substituted protein [7]. These data alone were used to refine the heavy atom parameters and calculate phase probability distributions with SHARP. In a second step, 130 cycles of solvent flattening were applied in SOLOMON [8] using a special procedure implemented in the SHARP interface that slowly decreases the radius of the 'solvent sphere' from an initial value of 3A until it reaches 1.05A. The effect of this procedure is to create a mask that starts as a molecular envelope and slowly becomes as a 'glove' around the density of each residue and therefore applies a very strong density modification constraint by the end of the procedure. The resulting map is of such quality that most atoms in the structure are clearly resolved from their neighbours, a fact that has allowed real-space refinement of a model directly against the density. It should be noted that the initial map obtained before solvent flattening, although it contains more noise, still is of high quality. A non-iterative refinement in real-space will quantitatively assess preferred rotamer conformation of side chains, local static disorder, and heterogeneity in general.