Figure 4.

Evolution of peptides 1 and 5 by sequential block reoptimization. The sequences selected from each reoptimization step are shown in bold, with the number of selection and amplification cycles given in parentheses. Sequences roughly matching the consensus that were used in later steps have been boxed. In some cases, such as in reoptimization step 3 for peptide 1 and reoptimization step 1 for peptide 5, we isolated clones that carried spurious mutations at a nondegenerate position of the peptide extension. In addition, the E21D mutation in the zinc finger region (which also was seen in the original peptide 2 sequence) arose several times; this mutation may stabilize complex formation by improving contacts at the protein–DNA interface. [Note: some confusion was caused by this E21D mutation, which occurred in the first reoptimization step for peptide 1, but was discovered only after reoptimization step 2. Thus, the “consensus” sequence from reoptimization step 1 (VPKQR), chosen after selection–amplification cycle 9, had a glutamine that did not occur in sequences isolated after cycle 6. To double-check this position of peptide 1, it was randomized again during reoptimization step 3. The corresponding position was allowed to vary as Q, M, I, or L, along with the complete randomization of the third block. The reselections showed that methionine or leucine is preferred at this position.]