Toward controlling gene expression at will: selection and design of zinc finger domains recognizing each of the 5'-GNN-3' DNA target sequences

Abstract

We have taken a comprehensive approach to the generation of novel DNA binding zinc finger domains of defined specificity. Herein we describe the generation and characterization of a family of zinc finger domains developed for the recognition of each of the 16 possible 3-bp DNA binding sites having the sequence 5'-GNN-3'. Phage display libraries of zinc finger proteins were created and selected under conditions that favor enrichment of sequence-specific proteins. Zinc finger domains recognizing a number of sequences required refinement by site-directed mutagenesis that was guided by both phage selection data and structural information. In many cases, residues not expected to make base-specific contacts had effects on specificity. A number of these domains demonstrate exquisite specificity and discriminate between sequences that differ by a single base with >100-fold loss in affinity. We conclude that the three helical positions -1, 3, and 6 of a zinc finger domain are insufficient to allow for the fine specificity of the DNA binding domain to be predicted. These domains are functionally modular and may be recombined with one another to create polydactyl proteins capable of binding 18-bp sequences with subnanomolar affinity. The family of zinc finger domains described here is sufficient for the construction of 17 million novel proteins that bind the 5'-(GNN)6-3' family of DNA sequences. These materials and methods should allow for the rapid construction of novel gene switches and provide the basis for a universal system for gene control.

Figure 1

The finger-2 recognition helices of randomly chosen clones from the seventh round of selection. The selection target site is shown to the left of each set, followed by the frequency with which each sequence was observed. The helix position of each amino acid is shown at the top, with positions −1, 3, and 6 shown in bold. Boxed sequences were studied in detail.

Figure 2

Multitarget ELISA titration assay for binding specificity. At the top of each graph is the DNA finger-2 target site for which each protein was selected or designed, and the recognition helix of that protein (positions −2 to 6). Helix positions −1, 3, and 6 are in bold. Proteins modified by site-directed mutagenesis have the prefix “m” before their DNA target. Columns 1–16 (filled bars) represent target oligos with different finger-2 subsites: 1 = GGG; 2 = GGA; 3 = GGT; 4 = GGC; 5 = GAG; 6 = GAA; 7 = GAT; 8 = GAC; 9 = GTG; 10 = GTA; 11 = GTT; 12 = GTC; 13 = GCG; 14 = GCA; 15 = GCT; 16 = GCC. Columns 17–20 (empty bars) represent oligonucleotide pools with a unique 5′ nucleotide in their finger-2 subsite: 17 = GNN; 18 = ANN; 19 = TNN; 20 = CNN. (j) Column 22 = CGC. (i) Column 22 = TAC. (bb) Column 22 = TGG. All data are background subtracted (column 21 = no target oligo). The height of each bar represents the average normalized titer from two independent experiments, with the highest signal normalized to the greatest value in columns 1–16 and 17–20. Error bars represent the deviation from the average. An arrow indicates the position of the cognate target oligonucleotide.