Statistical Analysis of the Benefits of Focused Saturation Mutagenesis in Directed Evolution Based on Reduced Amino Acid Alphabets

Directed evolution of stereo-, regio-, and chemoselective enzymes has enriched the toolbox of synthetic organic chemistry. Among the different gene mutagenesis techniques, saturation mutagenesis (SM) at sites lining the enzyme's binding pocket has emerged as a particularly viable approach to control selectivity and activity. Traditionally, NNK codon degeneracy encoding all 20 canonical amino acids is used, but as the size of the randomization site increases beyond a single residue, oversampling of transformants needed to ensure ≥95% library coverage rapidly reaches astronomical dimensions, impossible to screen in a practical manner. Therefore, many groups have been content with screening only a small segment of the designed protein sequence space, but this means that the best mutants will be missed. Alternatively, it has been shown that the use of highly reduced amino acid alphabets allows the generation of small and smart libraries requiring less screening. Here we address the question of which approach is more efficient. Two different enzyme types serve as model systems in stereoselective reactions, limonene epoxide hydrolase and P450-BM3. Equal numbers of transformants were screened for differently sized SM libraries that were constructed using NNK codon degeneracy and other codons corresponding to only one or just a few amino acids. Conversion as a rough indication of activity was used as the parameter in primary screening followed by GC-based ee-determination. Statistical analyses clearly show that it is more efficient to opt for rationally designed reduced amino acid alphabets because this approach results in a distinctly higher frequency of active mutants, stereoselectivity also being notably higher.