Proteins that share common ancestry may differ in structure and function because of divergent evolution of their amino acid sequences. For a typical diverse protein superfamily, the properties of a few scattered members are known from experiment. A satisfying picture of functional and structural evolution in relation to sequence changes, however, may require characterization of a larger, well chosen subset. Here, we employ a ''stepping-stone'' method, based on transitive homology, to target sequences intermediate between two related proteins with known divergent properties. We apply the approach to the question of how new protein folds can evolve from preexisting folds and, in particular, to an evolutionary change in secondary structure and oligomeric state in the Cro family of bacteriophage transcription factors, initially identified by sequence-structure comparison of distant homologs from phages P22 and . We report crystal structures of two Cro proteins, Xfaso 1 and Pfl 6, with sequences intermediate between those of P22 and . The domains show 40% sequence identity but differ by switching of ␣-helix to -sheet in a C-terminal region spanning Ϸ25 residues. Sedimentation analysis also suggests a correlation between helix-to-sheet conversion and strengthened dimerization.conformational switching ͉ structural evolution ͉ transitive homology ͉ x-ray crystallography T he amino acid sequences of proteins evolve faster than the structures and functions encoded by these sequences. This neutral sequence drift allows annotation of an uncharacterized protein-coding gene based on common ancestry (homology) with a characterized gene, even if the protein sequences are quite different. Conservation of structure and function may hold even for homology so distant that no clear sequence similarity has survived evolutionary divergence. Yet there are limits: the structural and functional evolution of proteins is not completely static, and the likelihood of two proteins evolving divergent properties increases with the extent of sequence change separating them. Remote homology detection methods [for example, PSI-BLAST (1), COMPASS (2), and HHpred (3)] thus yield diminishing returns for gene annotation by grouping distantly related proteins into superfamilies that encompass diverse properties and biological roles. Simultaneously, however, the excavation of distant relationships opens a rich field for experimental studies of protein evolution, with the promise of recovered annotation power as one elucidates how structure and function vary across the ''sequence space'' of a superfamily.Transitive sequence comparison is one method for detecting distant homology between highly diverged sequences (4-8). In this approach, two dissimilar sequences, A and C, are indirectly linked if a third ''intermediate'' sequence B exists with sufficient similarity to both A and C to imply homology with both proteins. The relationships between A and B and between B and C combine to support distant common ancestry between A and C. Transitivity can extend ...