Seeing but not believing: the structure of glycerol dehydrogenase initially assumed to be the structure of a survival protein from Salmonella typhimurium

The determination of the crystal structure of a mutant protein using phases based on a previously determined crystal structure of the wild-type protein is often a straightforward molecular-replacement protocol. Such a structure determination may be difficult if there are large-scale structural differences between the wild-type and mutant proteins. In this manuscript, an interesting case is presented of the unintentional crystallization of a contaminant protein which shared some structural features with the presumed target protein, leading to difficulties in obtaining a completely satisfactory molecular-replacement structure solution. It was not immediately evident that the initial structure solution was incorrect owing to the poor quality of the X-ray diffraction data and low resolution. The structure was subsequently determined by improving the quality of the data and following a sequence-independent MarathonMR protocol. The structure corresponded to that of glycerol dehydrogenase, which crystallized as a contaminant, instead of the presumed mutant of a survival protein encoded by Salmonella typhimurium. The reasons why a solution that appeared to be reasonable was obtained with an incorrect protein model are discussed. The results presented here show that a degree of caution is warranted when handling large-scale structure-determination projects.

Structure determination of contaminant proteins using the MarathonMR procedure

In the recent decades, essential steps of protein structure determination such as phasing by multiple isomorphous replacement and multi wave length anomalous dispersion, molecular replacement, refinement of the structure determined and its validation have been fully automated. Several computer program suites that execute all these steps as a pipeline operation have been made available. In spite of these great advances, determination of a protein structure may turn out to be a challenging task for a variety of reasons. It might be difficult to obtain multiple isomorphous replacement or multi wave length anomalous dispersion data or the crystal may have defects such as twinning or pseudo translation. Apart from these usual difficulties, more frequent difficulties have been encountered in recent years because of the large number of projects handled by structural biologists. These new difficulties usually result from contamination of the protein of interest by other proteins or presence of proteins from pathogenic organisms that could withstand the antibiotics used to prevent bacterial contamination. It could also be a result of poor book keeping. Recently, we have developed a procedure called MarathonMR that has the power to resolve some of these problems automatically. In this communication, we describe how the MarathonMR was used to determine four different protein structures that had remained elusive for several years. We describe the plausible reasons for the difficulties encountered in determining these structures and point out that the method presented here could be a validation tool for protein structures deposited in the protein data bank.