“Itching” for new strategies in protein engineering

Recent in vitro recombination or "DNA shuffling" experiments have demonstrated that exchanging fragments of closely homologous genes provides an efficient way to generate proteins with new traits. The resulting molecules are very different, at least in sequence, from those that might be obtained by more local searches of protein space, for example by random mutagenesis. The DNA shuffling method, which relies on homologous recombination during the PCR reassembly of gene fragments from multiple parents, generates crossovers at points of high sequence identity. Would crossovers between regions of low or no homology also be useful for molecular breeding of highly similar genes, or to shuffle more distantly related and even unrelated sequences? Would this approach generate a more efficient search of sequence space and, practically, more rapid adoption of novel protein properties? In this issue, Ostermeier et al. report a strategy to perform nonhomologous recombination that should eventually allow these questions to be answered empirically