The properties of (2F_o − F_c) and (F_o − F_c) electron-density maps at medium-to-high resolutions

This paper reports on the efficacy of (F_o − F_c) versus (2F_o − F_c) electron-density maps at 3.2 Å resolution. Firstly, a study is reported of a simple truncation at 2.3 and 3.2 Å of the 1.6 Å resolution crystal structure of concanavalin A at room temperature [Emmerich et al. (1994), Acta Cryst. D50, 749–756] with 149 known bound water molecules. Secondly, the concanavalin A 1.6 Å resolution model was re-refined but with the data truncated to 3.2 Å. In a similar evaluation, these procedures were repeated for the apocrustacyanin A1 cryotemperature 1.4 Å resolution model [Cianci et al. (2001), Acta Cryst. D57, 1219–1229]. Maps at 1.4, 2.3 and 3.2 Å resolutions were first generated and the structure was then re-refined at 3.2 Å and additionally at 2.3 Å resolution. The results on concanavalin A show that the number of bound water molecules that are resolved decreases by two thirds from 1.6 to 3.2 Å, but that key structural waters, for example at the transition metal and the calcium ion, are still resolved in the (F_o − F_c) map but not in the (2F_o − F_c) map. For apocrustacyanin A1, the results with these two difference maps were less clear-cut. Two key structural bound waters (w93 and w105) were selected that had been previously identified in β-­crustacyanin [Cianci et al. (2002), Proc. Natl Acad. Sci. USA, 99, 9795–9800] in protein–carotenoid interactions. The behaviour of w93 is similar to that of concanavalin A key waters, but that of w105 is not. These behaviours were therefore explored in finer resolution increments, namely 2.9, 2.7 and 2.5 Å. Finally, further tests on `real' data sets for peanut lectin and concanavalin A at medium resolution confirm these map properties, namely that an (F_o − F_c) difference electron-density map is more effective than a (2F_o − F_c) map in showing bound water structure at lower resolutions (∼3.2 Å). This result is important since a growing number of protein crystal structure studies are concerned with multi-macromolecular complexes and are at such resolutions. Details of the bound solvent can still be revealed at 3.2 Å via the (F_o − F_c) map calculation. The physical basis of the limitation of the (2F_o − F_c) map presumably lies in the series-termination error effect on such a map involving the first negative ripple from the protein atom to which a bound water oxygen is hydrogen bonded, sufficiently cancelling its peak. In addition, re-refinements at 3.2 Å show distances that can agree with known values but B values that do not agree with known values.