Naturally occurring enzymes are widely used for metabolic engineering, but they frequently do not exhibit the exact functional properties desired for a given application (e.g., they lack the desired stability, expression level, catalytic eciency, or substrate-specicity). Within the colossal space of possible polypeptide sequences, proteins are thought to exist with functions that meet the specications for almost any engineering goal that you can imagine. In some cases, knowledge-based design can nd proteins with structures and functions distinct from those observed in nature [1-6]. However, our understanding of protein sequence-structure-function relationships is not yet sophisticated enough to consistently alter enzyme functions in a desired manner, especially when the design goal is to optimize a preexisting property. Directed evolution, in contrast, has repeatedly proven quite eective at optimization when applied alone or when used in tandem with knowledge-based mutagenesis [7]. By coupling mutagenic and selection (or screening) strategies as occurs in nature, this approach sieves through libraries of variants to lter out the desired ones with improved tness.