Modular polyketide synthases (PKSs), such as the 6-deoxyerythronolide B synthase (DEBS), catalyze the biosynthesis of structurally complex and medicinally important natural products. These large multifunctional enzymes are organized into "modules", where each module contains active sites homologous to those of higher eucaryotic fatty acid synthases (FASs). Like FASs, modular PKSs are known to be dimers. Here we provide functional evidence for the existence of two catalytically independent clusters of active sites within a modular PKS. In three bimodular derivatives of DEBS, the ketosynthase domain of module 1 (KS-1) or module 2 (KS-2) or the acyl carrier protein domain of module 2 (ACP-2) was inactivated via site-directed mutagenesis. As expected, the purified proteins were unable to catalyze polyketide synthesis (although the KS-1 mutant could convert a diketide thioester into the predicted triketide lactone). Remarkably however, the KS-1/KS-2 and the KS-2/ACP-2 mutant pairs could efficiently complement each other and catalyze polyketide formation. In contrast, the KS-1 and ACP-2 mutants did not complement each other. On the basis of these and other results, a model is proposed in which the individual modules of a PKS dimer form head-to-tail homodimers, thereby generating two equivalent and independent clusters of active sites for polyketide biosynthesis. Specifically, each subunit contributes half of the KS and ACP domains in each cluster. A similar complementation approach should also be useful in dissecting the organization of the remaining types of active sites within this family of multienzyme assemblies. Finally, blocked systems, such as the KS-1 mutant described here, present a new strategy for the noncompetitive conversion of unnatural substrates into polyketides by modular PKSs.