We have engineered a soluble, stable two-chain dimeric streptavidin (TCD) in Escherchia coli. Examination of the three-dimensional structure of streptavidin aided by empirical binding free-energy calculations helped us to select mutations at subunit interfaces that dissociate the native tetramer and stabilize the desired dimer. We chose positions W120, L124, V125 and H127 and mutated them to 120D/124D/125D/127D (TCD-1); 120D/124N/125S/127D (TCD-2); and 120D/124D/125S/127D (TCD-3). The H127D mutation creates electrostatic repulsion that disrupts the dimer-dimer interface, but leaves it very hydrophobic. Therefore, W120, L124 and V125 were mutated to hydrophilic residues to increase dimer solubility. Among the three candidates, TCD-2 gave the best result: a stable, active dimer with K(d) for biotin of approximately 1x10(-7)M after purification by gel-filtration chromatography. The experimental results confirm the possibility of rational engineering of low-pI dimeric streptavidins. Reduced-size streptavidin mutants with a net negative charge may be more suitable than antibodies or wild-type streptavidin for the targeting step in radioimmunotherapy because they should clear faster from the bloodstream and the kidney.