Solving the structures of macromolecular complexes.

The examples cited show that three-dimensional reconstructions of non-crystalline objects embedded in vitreous ice can yield reliable electron density maps. The resolution of such maps (20– 30 a) is presently limited by the relatively high degrees of underfocus (1–2 mm) required to achieve adequate contrast in the image. This resolution is, however, generally adequate for positioning subunits of known structure. Nevertheless, a serious problem arises in defining details of the interface since all the side chains find themselves in a new environment. Moreover, macromolecules have a strong tendency to indulge in ‘induced fit’ on mating. This trend has become abundantly clear for protein–DNA complexes [19xPlus ca change, plus c'est la meme chose. Burley, S.K. Nature Struct. Biol. 1994; 1: 207–208Crossref | PubMed | Scopus (6)See all References][19]. Furthermore, in the case of actin there are two forms, F and G, which differ from each other at atomic resolution, the structure switch being controlled by the polymerization. Moreover, there are ‘loops’ in actin which only acquire structure in the complex. The structure of these must be computed. This is also true for the actin–myosin interface. The question is how accurately we may be able to predict the structure of a protein– protein interface by calculation if we can define the general way that the proteins fit together using the methods discussed above. A well defined example, an antigen–antibody complex, has recently been tackled ab initio [20xDetailed ab initio prediction of lysozyme-antibody complex with 1.6 a accuracy. Totrov, R. and Abagyan, M. Nature Struct. Biol. 1994; 1: 259–263Crossref | PubMed | Scopus (95)See all References][20], with some success. If ab initio methods can fold protein loops in more or less well defined environments, we may get what we need.Another possibility is to improve cryo-EM. Here there is indeed hope. In particular the introduction of energy-filter cryo-EM leads to dramatic improvement in the signal to noise ratio in the image [[21]xZero-loss energy filtering as improved imaging mode in cryoelectron microscopy of frozen hydrated specimens. Schroder, R.A., Hofmann, W., and Menetret, J.-F. J. Struct. Biol. 1990; 105: 28–34Crossref | Scopus (34)See all References, [22]xQuantitation of molecular densities by cryo-electron microscopy. Determination of the radial density distribution of tobacco mosaic virus. Smith, M.F. and Langmore, J.P. J. Mol. Biol. 1992; 226: 763–774Crossref | PubMed | Scopus (22)See all References]. This improvement in turn allows images to be obtained closer to focus and therefore with higher resolution. If one can attain 8–10 a resolution from macro-molecular complexes in vitreous ice, a difficult but not impossible task, then one will have enough data to define some elements of secondary structure. This should allow an experimental verification of any ‘induced fit’ and provide the basis for more constrained, and therefore more precise, attempts to calculate the protein–protein interface.