Genetic selection was exploited in combination with structure-based design to transform an intimately entwined, dimeric chorismate mutase into a monomeric, four-helix-bundle protein with near native activity. Successful reengineering depended on choosing a thermostable starting protein, introducing point mutations that preferentially destabilize the wild-type dimer, and using directed evolution to optimize an inserted interhelical turn. Contrary to expectations based on studies of other four-helix-bundle proteins, only a small fraction of possible turn sequences (fewer than 0.05 percent) yielded well-behaved, monomeric, and highly active enzymes. Selection for catalytic function thus provides an efficient yet stringent method for rapidly assessing correctly folded polypeptides and may prove generally useful for protein design.