AUTOMATED SCATTERING CURVE MODELLING OF PLASMA MULTIDOMAIN PROTEINS. S. J. Perkins, Dept. Biochemistry and Molecular Biology, Royal Free Hospital Sch. Med., Rowland Hill St., London NW3 2PF, UK

Solution scattering analyses of multidomain proteins have been much enhanced by a recently-developed method of constrained scattering curve fits. The domains are represented by known atomic coordinates from homologous proteins. Scattering curves are calculated using a previously calibrated procedure. This process can be automated in order to generate a range of sterically allowed structures which are then evaluated systematically. Error limits on the ensuing models can be determined. Three types of applications have been developed to date for single proteins, glycosylated proteins and protein-protein complexes: (1) Bovine immunoglobulin G was analysed in terms of two Fab subunits and one Fc subunit to show that an extended planar structure gave good fits. The six-domain structure of the Fc subunit of human immunoglobulin E was studied to show that only bent non-linear Fc structures satisfied the experimental scattering curves. (2) The heavily glycosylated seven-domain cell surface protein, carcinoembryonic antigen (>50% carbohydrate) was represented by an adaptation of the crystal structure of the two-domain protein CD2. This yielded an extended model that satisfied the scattering data. (3) Homodimers and homotrimers of the bacterial cytosol amide sensor protein AmiC, which is structurally related to two-domain monomeric periplasmic binding proteins, were modelled to show that compact associations of the monomers accounted for their scattering curves. The heterodimeric complex of tissue factor and factor VIIa, which are two- and four-domain plasma proteins implicated in blood coagulation, were modelled to result in a compact complex. While unique structures cannot be determined by scattering analyses, application of the constraints inherent in the use of known crystal structures for scattering curve fits usually permitted a limited family of multidomain structures to be determined in each case. In combination with other information in relation to known ligand binding sites, these studies can provide useful biological insights on the steric accessibility or exposure of domains in multidomain proteins and their complexes.