Addressing the numbers problem in directed evolution

Our previous contribution to increasing the efficiency of directed evolution is iterative saturation mutagenesis (ISM) as a systematic means of generating focused libraries for the control of substrate acceptance, enantioselectivity, or thermostability of enzymes. We hove now introduced a crucial element to knowledge-guided targeted mutagenesis in general that helps to solve the numbers problem in directed evolution. We show that the choice of the amino acid (aa) alphabet, as specified by the utilized codon degeneracy, provides the experimenter with a powerful tool in designing "smarter" randomized libraries that require considerably less screening effort. A systematic comparison of two different codon degeneracies was made by examining the relative quality of the identically sized enzyme libraries in relation to the degree of oversampling required in the screening process. The specific example in our case study concerns the conventional NNK codon degeneracy (32 codons/20 aa) versus NDT (12 codons/12 aa). The model reaction is the hydrolytic kinetic resolution of a chiral trans-disubstituted epoxide, catalyzed by the epoxide hydrolase from Aspergillus niger. The NDT library proves to be of much higher quality, as measured by the dramatically higher frequency of positive variants and by the magnitude of catalyst improvement (enhanced rate and enantioselectivity). We provide a statistical analysis that constitutes a useful guide for the optimal design and generation of "smarter" focused libraries. This type of approach accelerates the process of laboratory evolution considerably and can be expected to be broadly applicable when engineering functional proteins in general.