Nature, 409, 215–219 (2001).

In light of the recent X-ray structure determination of the SarR-maltose binding fusion protein by Zhang et al.1, we re-examined our crystal structure determinations of the Staphylococcus aureus transcription regulator, SarA, which were reported in the 11 January 2001 issue of Nature. Although suggested homologues (27% identity), the SarA and SarR structures are significantly dissimilar. We now believe that our structures of the apo and DNA-bound forms of SarA may harbour some anomalies.

We are performing a comprehensive examination of the structure and function relationships of a global virulence gene regulator in S. aureus, SarA. The full-length, recombinant protein was expressed in E. coli and purified to apparent homogeneity. This protein, whether generated in the laboratory of B.K.H. or that of R.G.B., behaved as expected. It bound with high affinity to cis regulatory sequences upstream of virulence genes previously reported to be controlled by SarA. These data were reported in refs 2 and 3.

To investigate why the SarA and SarR structures are different, we first reassessed our structure-determination processes. The SIR and MAD (as well as averaged)-derived phases that were used to calculate our SarA–DNA complex structure resulted in electron density maps that showed consistent secondary structure features different from those of SarR. In addition, MAD data for one of the ‘apo’ SarA crystals revealed similar features to the DNA-bound form of SarA. These results suggested to us that we were working with a highly flexible molecule that could undergo remarkable structural changes. One point is the non-standard, although we feel correct, manner by which we calculated the value of Rfree, that is, both structures were not refined with CNS and therefore we may have erred in the use of this quality-control standard. It remains a possibility that our intensity data were simply not of high enough quality to discern the structure of Zhang et al.1. Although attempted multiple times, however, molecular replacement experiments to solve the apo SarA and the SarA–DNA protein structures, using our intensity data and SarR as the search model, have failed. Moreover, our heavy-atom sites are inconsistent with their SarR location. We are left with the possibility that in the absence of a carrier protein, such as MBP, or cognate DNA of sufficient length, we have crystallized a protein with some anomalously folded regions, which in our structures display a great deal of conformational flexibility. In support of this are the facts that there is a remarkable change in space group (from P212121 to P21212) and c cell edge (from 141 Å to 27 Å) upon freezing; all cryocooled crystals were non-isomorphous; our protein becomes inactive over time and degrades; our SarA–DNA complex reveals only nonspecific contacts; and there is an unprecedented change in protein conformation upon ligand binding.

To test the validity of our structures and to assess the importance of residues shown in the structures as important for function, we generated mutants of SarA that should result in aberrant activity. Using both an in vivo assay for virulence gene regulation and an in vitro DNA-binding assay for SarA function, most of the mutants have significantly altered activity (K. Sterba, M. S. Smeltzer and B.K.H., unpublished results). At this point, we do not know the significance and origin of the differences between the reported structures of SarA and SarR.