We recently reported the development of nonhomologous random recombination (NRR) as a method for nucleic acid diversification and applied NRR to the evolution of DNA aptamers. Here, we describe a modified method, protein NRR, that enables proteins to access diversity previously difficult or impossible to generate. We investigated the structural plasticity of protein folds and the ability of helical motifs to function in different contexts by applying protein NRR and in vivo selection to the evolution of chorismate mutase (CM) enzymes. Functional CM mutants evolved using protein NRR contained many insertions, deletions, and rearrangements. The distribution of these changes was not random but clustered in certain regions of the protein. Topologically rearranged but functional enzymes also emerged from these studies, indicating that multiple connectivities can accommodate a functional CM active site and demonstrating the ability to generate new domain connectivities through protein NRR. Protein NRR was also used to randomly recombine CM and fumarase, an unrelated but also ␣-helical protein. Whereas the resulting library contained fumarase fragments in many contexts before functional selection, library members surviving selection for CM activity invariably contained a CM core with fumarase sequences found only at the termini or in one loop. These results imply that internal helical fragments cannot be swapped between these proteins without the loss of nearly all CM activity. Our findings suggest that protein NRR will be useful in probing the functional requirements of enzymes and in the creation of new protein topologies. D irected evolution has been used to alter (1-4) enzyme activity as well as to investigate sequence constraints on protein folding (5-7) and catalysis (8). The sequences and therefore the properties of proteins that emerge from directed evolution depend on the diversification method used to generate the protein library before screening or selection. Theoretical studies (9) suggest that diversification methods, including point mutagenesis, homologous recombination, secondary structure swapping, and nonhomologous recombination differ in their ability to access protein sequences representing new folds and improved functions. Of the methods listed above, nonhomologous recombination has been theorized to be the most effective at enabling new structures and functions to emerge during protein evolution. Additional studies on enzyme superfamilies suggest that the reorganization of structurally similar components can result in altered substrate specificity or function (10). Because these components often lack DNA sequence homology, their combinatorial diversification can in general only be accomplished through nonhomologous recombination.Methods that enable recombination to take place at defined sites without sequence homology have been recently described (11). For example, it is possible to recombine unrelated proteinencoding genes by using synthetic oligonucleotides to encode each desired crossover (12, 1...