Structural uncertainty of proteins in solution by NMR. A re-evaluation of the structure of theLac repressor headpiece

Abstract

A number of recent studies have shown that the determination of protein structure in solution by NMR data is subject to bias introduced by the method of data analysis. Specifically, it has been shown that certain commonly used algorithms for the interpretation of distance data may not fully explore the space of conformational possibilities (Levy R.M., Bassalino D.A., Kitchen D.B., Pardi A.: Biochemistry28, 9361–9372 (1989); Metzler W.J., Hare D.R., Pardi A.: Biochemistry28, 7045–7052 (1989); Pachter R., Altman R.B., Jardetzky O.: J. Magn. Res.89, 578–584 (1990); Pachter R., Altman R.B., Czaplicki J., Jardetzky O.: J. Magn. Res.92, 468–479 (1991); Jardetzky O. in: Computational Aspects of the Study of Biological Macromolecules by NMR Spectroscopy (Hoch J.C., Poulsen F.M., Redfield C., eds.). NY: Plenum 1991; Liu Y, Altman R., Jardetzky O.: (1992), manuscript in preparation.) We report here a new analysis of previously existing data on thelac repressor headpiece protein and a comparison of structures obtained by different methods. Our own method of analysis (Altman R.B., Jardetzky O.: Methods in Enzymology177, 218–246 (1989) and references therein) has been specifically designed to calculate not only a structure, but also an explicit estimate of the uncertainty in that structure. It uses a sequence of steps, each of which attempts to maintain a conservative upper bound on the uncertainty of each atomic position; a sampled system calculates the position of the backbone atoms, then a parametric representation refines these positions. These refined backbone positions, with their associated uncertainly, are then used as a starting point for a full atomic refinement. We show that previous structures have captured the essential topology of thelac repressor headpiece molecule accurately. However, there are important differences in the relative certainty of atomic positions within the molecule which imply numerous possible variations in the detailed structure. We show that our representation is a reasonable way to summarize these uncertainties for purposes of cautious interpretation of the data.