PROTEIN FLEXIBILITY AND ADAPTABILITY SEEN IN 25 CRYSTAL FORMS OF T4 LYSOZYME

The structures of various mutants of T4 lysozyme have been determined in 25 non-isomorphous crystal forms. This provides an unusually diverse data base to compare the structures and dynamics of a closely related set of proteins in different crystal packing environments. In general, the more tightly packed crystals diffract better than those that are highly hydrated although the wild-type crystal form is an exception. The ability of the protein to form a relatively open but stable lattice may help explain why many of the mutants crystallize in this form. In different crystalline environments, the lysozyme molecules associate with 2-fold, 3-fold, 4-fold, and 5-fold symmetry, as well as with various types of screw associations. A ''back-to-back'' dimeric association, and a ''head-to-tail'' 2(1) screw association, are especially common, each occurring in more than half a dozen crystal forms. The 4-fold and 5-fold modes of association are closely related and provide an example of quasi-equivalent association as envisaged by Caspar and Klug. In different crystal environments the lysozyme molecules display a range of over 50 degrees in the hinge-bending angle between the amino and carboxyterminal domains. Large variations in the hinge-bending angle are observed not only for lysozymes with mutations in the hinge region, but for molecules with mutations far from this site. This suggests that hinge-bending is an intrinsic property of the lysozyme molecule and is not an artifact due to mutation. As the hinge-bending angle increases about 15 degrees beyond that seen in wild-type there is a distinct conformations change in the side-chains of five residues in the hinge-bending region. Changes in the backbone are localized near residues 13, 59 and 80, but do not include significant changes in phi psi. Comparison of the different structures indicates that crystal contacts perturb the backbone structure of the protein by 0.2 to 0.5 Angstrom. These perturbations are of the same magnitude for helices and beta-sheet strands, suggesting that protein structures can be defined and maintained equally well by hydrogen-bonding (i.e. strand-strand) or by non-hydrogen-bonding (i.e. helix-helix) interactions. The discrepancies between the lysozyme structures in different crystallographic environments are in line with other comparisons of independently determined protein crystal structures. They suggest that protein structures in general are subject to low energy changes in conformation of 0.2 to 0.5 Angstrom. The thermal factors of the side-chains in the individual crystal structures agree well with the average values (overall correlation 0.74), suggesting that side-chain mobility in the crystals is representative of side-chain mobility in solution. The thermal factors of the main-chain atoms, however, do not agree so well (correlations with the average ranging from 0.16 to 0.76). This suggests that in extreme cases the main-chain motion in some crystal forms may be constrained by intermolecular crystal contacts to such a degree that it is no longer representative of dynamic behavior in solution.