MAD PHASING STRATEGIES EXPLORED WITH A BROMINATED OLIGONUCLEOTIDE CRYSTAL AT 1.65 ANGSTROM RESOLUTION

The crystal structure of a brominated oligonucleotide d(CGCG(Br)CG), c hemical formula C114N48O68P10Br2, has been analysed by multiwavelength anomalous dispersion (MAD) methods. The oligonucleotide crystallizes in space group P2(1)2(1)2(1) with a = 17.97, b = 30.98, c = 44.85 Angs trom, alpha = beta = gamma = 90 degrees. Data to a resolution of 1.65 Angstrom were collected at four wavelengths about the K-absorption edg e of the bromine atom (lambda(1) = 0.9323 Angstrom, a reference wavele ngth at the long-wavelength side of the edge; lambda(2) = 0.9192 Angst rom, at the absorption-edge inflection point; lambda(3) = 0.9185 Angst rom, at the 'white line' absorption maximum; lambda(4) = 0.8983 Angstr om, a reference wavelength at the short-wavelength side) using synchro tron radiation at Station PX9.5, SRS, Daresbury. Multiwavelength data could be collected on a single crystal as the sample was radiation sta ble. Anomalous and dispersive Patterson maps were readily interpretabl e to give the bromine anomalous scatterer positions. Phase calculation s to 1.65 Angstrom resolution, using all four wavelengths, gave a figu re of merit of 0.825 for 2454 reflections. The electron-density map wa s readily interpretable showing excellent connectivity for the sugar/p hosphate backbone and each base was easily characterized. The two nucl eotide strands paired up as expected in an antiparallel Watson-Crick-t ype manner. The structure was refined to 1.65 Angstrom using all the d ata (R factor = 17.0% based on 3151 reflections, with a data-to-parame ter ratio of 2.6). In addition to the four-wavelength analysis, a vari ety of other phasing strategies, and the associated quality of the res ulting electron-density maps, were compared. These included use of eit her of the reference wavelength data sets in the two possible three-wa velength phasing combinations to assess their relative effectiveness. Moreover, the time dependence upon measuring the Bijvoet differences a nd its effect upon phasing was also investigated. Finally, the use of only two wavelengths, including Friedel pairs, is demonstrated (the th eoretical minimum case); this is of particular interest when consideri ng overall beam time needs and is clearly a feasible experimental stra tegy, as shown here.