The bacterial cell division protein FtsZ assembles into straight protofilaments, one subunit thick, in which subunits appear to be connected by identical bonds or interfaces. These bonds involve the top surface of one subunit making extensive contact with the bottom surface of the subunit above it. We have investigated this interface by site-directed mutagenesis. We found nine bottom and eight top mutants that were unable to function for cell division. We had expected that some of the mutants might poison cell division substoichiometrically, but this was not found for any mutant. Eight of the bottom mutants exhibited dominant negative effects (reduced colony size) and four completely blocked colony formation, but this required expression of the mutant protein at four to five times the wild-type FtsZ level. Remarkably, the top mutants were even weaker, most showing no effect at the highest expression level. This suggests a directional assembly or treadmilling, where subunit addition is primarily to the bottom end of the protofilament. Selected pairs of top and bottom mutants showed no GTPase activity up to 10 to 20 microM, in contrast to the high GTPase activity of wild-type FtsZ above 1 muM. Overall, these results suggest that in order for a subunit to bind a protofilament at the 1 microM K(d) for elongation, it must have functional interfaces at both the top and bottom. This is inconsistent with the present model of the protofilament, as a simple stack of subunits one on top of the other, and may require a new structural model.