Identifying Improved Variants

Another major challenge in directed evolution is to measure changes in the property of interest. The problem is that typically thousands of assays have to be performed to identify variants with improved properties. However, for traditional biochemical assays this can be prohibitively time consuming and reagent intensive. One appealing solution to this problem is to identify a selection system for the improved enzyme. In this scenario, the host organism is unable to survive unless a variant of the expressed enzyme attains a particular property that allows the host to survive under defined growth conditions. Such a system was reported for plant fatty acid desaturase genes. An E. coli strain MH13 is an unsaturated fatty acid auxotroph that has to be supplemented with unsaturated fatty acids in the growth medium for survival (Cahoon and Shanklin, 2000; Clark et al., 1983). The enzyme encoded by the plant desaturase gene was specific for 18-carbon substrate, but E. coli contains insufficient 18-carbon substrate for the desaturase to convert to the unsaturated fatty acid necessary for survival (Cahoon and Shanklin, 2000). However, E. coli does contain sufficient levels of 16-carbon substrate for the enzyme to desaturate, but the enzyme was far more active on 18- versus 16-carbon substrates. So, a library of variants was constructed from the 18-carbon preferring desaturase and E. coli containing these variants was challenged to grow on media lacking unsaturated fatty acids. This method allowed the identification of many variants specific for 16-carbon substrates (Whittle et al., 2001). When reintroduced into plants, these enzymes efficiently desaturated 16-carbon fatty acid resulting in the accumulation of unusual fatty acids in seed oils.

The benefits of such selection systems are immediately apparent, that is, that all growing colonies are ‘‘winners,’’ and that millions of variants can be assessed in a short period of time. However, it should be noted that there are also problems using this approach. It can be very difficult or impossible to design such selection systems because the product of a desired reaction may not be essential for survival. It can also be difficult to manipulate the threshold necessary for survival. This means that one might have too tight or too loose a criterion for survival, in which cases one might get no colonies, or get too many to perform follow-up analysis. Even with the extremely powerful fatty acid auxotrophy selection described above, it proved difficult to alter the survival constraints, and so it was relatively easy to identify the first round of improved variants, but the system was of little use in identifying further improved variants after subsequent recombination experiments of the type described below.

The alternative to selection systems is to employ screening techniques. Because precise assays are relatively time consuming, the use of tiered screens has become routine for high-throughput applications. The idea of a tiered screen is that improved variants are first assessed for improvement with a fast but low precision methodology and that variants that meet some minimal criterion are subjected to a secondary screen that is more precise but more time- and reagent consuming. A final very precise biochemical assay is added in some cases to distinguish between the improved variants. Examples of such screens include fluorescentactivated cell sorting (FACS) or phage display, technologies that can quickly screen >10⁷ variants (Crameri et al., 1998; Gao et al., 1997; Naki et al., 1998); solid state colorimetric or fluorescence assays for between 104 and 107 variants (Moore and Arnold, 1996; Zhang et al., 1997); microtiter format assays for 102–104 (Joo et al., 1999; Zhang et al., 1997); and individual high-precision assays involving gas chromatography, high performance liquid chromatography, or mass spectrometry for 10¹–10² samples (Altamirano et al., 2000; Reetz et al., 1997).