Computational and mutagenesis studies of the streptavidin native dimer interface.

Wt streptavidin forms a domain swapped tetramer consisting of two native dimers. The role of tetramerization has been studied previously and is known to contribute to biotin binding by allowing the exchange of W120 between adjacent subunits. However, the role of dimer formation in streptavidin folding and function has been largely overlooked to date, although native dimers are necessary for tetramer formation and thus for high affinity biotin binding. To understand how the side chain interactions at the dimer interface stabilize the subunit association, we studied the structural and functional consequences of introducing interfacial mutations by a combination of molecular dynamics (MD) simulation and biochemical characterization. In particular, we introduced rational mutations at the dimer interface to engineer new side chain interactions and measured the stability and function of the resulting mutants. We focused on two residues that form a "knob" and a "hole" pair, G74 and T76, since steric complementarity plays an important role at these positions. We introduced mutations that would change the polarity and side chain packing to test if the interface can be rationally redesigned. Both energy calculation and geometric parameterization were used to interpret the simulated structures and predict how the mutations affect the dimer stability. In this regard, obtaining precise energy estimates was difficult because the simulated structures have large stochastic variations and some mutants did not reach an equilibrium by the end of the simulation. In contrast, comparing the wt and mutants to one another and parameterizing the simulation using a geometric parameter, i.e. the degree of solvation of the buried interface, resulted in a testable prediction regarding which mutations would result in a stable dimer. We present experimental data (denaturation and binding measurements) to show that an intuitive parameter based on physical reasoning can be useful for characterizing simulations that are difficult to analyze quantitatively.