The Cys2His2 zinc finger domain (hereafter simply “zinc finger”) provides a useful scaffold for creating customized DNA-binding proteins, a technology with potential applications in biological research, molecular medicine, and gene therapy. Using a combination of targeted randomization and selection methodologies (e.g., phage display), many research groups have successfully altered the DNA-binding specificities of single zinc fingers, which typically recognize three to four base pairs of DNA. In these experiments, potential DNA-binding residues in a finger’s a-helix (or “recognition helix”) were randomized to generate a library of variants and then selection methods were used to identify fingers with desired DNA-binding specificities. To create synthetic multi-finger proteins capable of recognizing longer DNA sequences, various investigators have linked together three or more pre-selected or pre-characterized finger domains (Figure 1A) (see, for example, Choo et al., 1994; Beerli et al., 1998; Liu et al., 2002; and Bae et al., 2003). This type of strategy, which assumes that individual fingers behave in a modular fashion, permits the rapid assembly of multi-finger proteins directed to bind a wide variety of different DNA sequences.
A number of observations suggest, however, that zinc fingers do not always behave as completely modular units. For example, neighboring zinc fingers can in some cases work together to bind their respective DNA sites with residues from one finger “crossing over” into the DNA site bound by a neighboring finger. In addition, structural data has revealed the occurrence of inter-finger interactions between adjacent finger domains. Lastly, the results of selection experiments showed context-dependent effects (i.e., fingers that work well within one multi-finger protein may not necessarily be optimized for functioning within another). Taken together, these “context-dependent” effects on DNA-binding strongly suggest that multi-finger proteins constructed using a “modular assembly” approach may not always be fully optimized for DNA-binding to their intended target DNA sequences.
To address the issue of context-dependent effects on DNA-binding, Choo and colleagues described a “bi-partite” method for constructing synthetic three-finger proteins. In this strategy, two “halves” of a three-finger domain (i.e., one and a half fingers) are first independently randomized and selected and then assembled to create the final desired protein (Figure 1B; Isalan et al., 2001). This method accounts for and permits the optimization of inter-finger co-operativity. However, this approach also significantly limits the spectrum of amino acid residues permitted within each one-and-a half finger library (a restriction required to keep combinatorial parameters manageable). Given our, as yet incomplete, understanding of zinc finger protein-DNA interactions, these limitations to the combinatorial diversity of libraries may lead to the unintentional exclusion of optimal fingers for certain DNA-binding sites.
We recently developed a new selection strategy that identifies groups of fingers that function well together without sacrificing combinatorial diversity (Hurt et al., 2003). Our approach consists of three steps (Figure 1C): initial low-stringency selection of finger pools (one pool for each position in the final protein) from pre-constructed master randomized finger libraries, random shuffling of finger pools to create a library of multi-finger proteins, and a final high-stringency selection for binding to the desired target DNA sequence. As with another previously described strategy (Greisman and Pabo, 1997) our approach ensures the selection of finger combinations that optimally bind the DNA target without limiting library diversity. However, our method requires the construction of fewer randomized libraries (and is therefore less labor intensive) than the method of Greisman and Pabo. In addition, our method uses a bacterial two-hybrid selection system, instead of multi-round phage display, to directly identify candidates from both the randomized and the shuffled finger libraries.
We used our new strategy to select multi-finger proteins capable of binding to specific DNA sites and compared the affinities and specificities of these with comparable proteins (designed to bind to the same sites) constructed using modular or bi-partite assembly methods. We reasoned that if either context-dependent effects and/or full combinatorial diversity are important for creating multi-finger proteins then our strategy should yield proteins with an improved DNA-binding function relative to those produced by either the modular or bi-partite methods. To perform our test, we chose three different target DNA sequences. For all three target sequences, our selection yielded proteins with better affinities and/or specificities than proteins made by modular or bi-partite assembly (Hurt et al., 2003) as judged by in vitro electrophoretic mobility shift assays. In addition, we demonstrated that for one of the target sites, the proteins we selected exhibit better function in the context of a mammalian cell nucleus than a protein created by modular assembly. In this experiment we fused zinc finger proteins to a transcriptional activation domain and compared their abilities to stimulate transcription of an endogenous mammalian cell gene harboring a target DNA site (Hurt et al., 2003).
Our results clearly demonstrate that multi-finger proteins assembled using either the modular or bi-partite assembly methods will not always be fully optimized for DNA-binding affinity and/or specificity. Although the modular and bi-partite assembly strategies are conceptually appealing as they permit the rapid construction of multi-finger proteins from pre-selected or pre-characterized finger collections, our data suggest that the DNA-binding characteristics of the final proteins assembled using these methods should be carefully examined. Our results also suggest that failure to consider either context-dependent effects on DNA-binding when constructing proteins, or limitations on combinatorial diversity when preparing randomized finger libraries, may produce proteins with sub-optimal affinity and/or specificity. Our new selection strategy provides a simple and highly effective alternative for taking both these factors into account and for producing multi-finger proteins with high DNA-binding affinity and specificity, characteristics critical to the successful use of these proteins for either research or therapeutic applications.
In vivo repression by a site-specific DNA-binding protein designed against an oncogenic sequence.
Choo Y, Sanchez-Garcia I, Klug A.
Medical Research Council, Cambridge, UK.
Nature 372:642-645, Dec. 15, 1994.
Summary: The authors describe the assembly of a three-finger protein (targeted to a sequence unique to the BCR-ABL translocation) from pre-selected finger units. Evidence of specific DNA-binding by this protein in vitro and in cultured cells is presented.
Toward controlling gene expression at will: specific regulation of the erbB-2/HER-2 promoter by using polydactyl zinc finger proteins constructed from modular building blocks.
Beerli RR, Segal DJ, Dreier B, Barbas CF 3rd.
The Scripps Research Institute, La Jolla, CA, USA.
PNAS 95:14628-14633, Dec. 8, 1998.
Summary: This report describes the assembly of multi-finger proteins from pre-selected finger modules. Synthetic three-finger and six-finger proteins are constructed and tested for their ability to bind specific sequences in vitro and in mammalian cells.
Validated zinc finger protein designs for all 16 GNN DNA triplet targets.
Liu Q, Xia Z, Zhong X, Case CC.
Sangamo BioSciences Inc., Richmond, CA, USA.
J Biol Chem 277:3850-3856, Feb. 8, 2002.
Summary: The authors describe the in vitro characterization of three sets (one for each position in a three-finger protein) of sixteen individual fingers designed to bind three base pair sites of the form 5′GNN3′.
Human zinc fingers as building blocks in the construction of artificial transcription factors.
Bae KH et al.
Toolgen, Inc., Daejeon, South Korea.
Nature Biotechnol 21:275-280, Mar. 2003.
Summary: Description of the cloning and characterization of a collection of naturally occurring zinc fingers from the human genome. The authors use these pre-characterized fingers to construct multi-finger proteins and demonstrate their activities in mammalian cells.
A rapid, generally applicable method to engineer zinc fingers illustrated by targeting the HIV-1 promoter.
Isalan M, Klug A, Choo Y.
Medical Research Council Laboratory of Molecular Biology, Cambridge, UK.
Nat Biotechnol 19:656-660, Jul. 2001.
Summary: First description of the “bi-partite” assembly method for creating three-finger proteins. The authors construct libraries in which residues in one and a half recognition helices are randomized and perform phage display selections to identify one-and-a-half finger modules with altered DNA binding specificities. These “halves” are then recombined to construct three-finger proteins, which the authors test for binding in vitro.
Highly specific zinc finger proteins obtained by directed domain shuffling and cell-based selection.
Hurt JA*#, Thibodeau SA*, Hirsh AS*, Pabo CO#, Joung JK*#.
Massachusetts General Hospital (*), Boston, MA, USA and Howard Hughes Medical Institute and Massachusetts Institute of Technology (#), Cambridge, MA, USA.
PNAS 100:12271-12276, Oct. 14, 2003.
Summary: This report describes validation of a method for selecting multi-finger proteins that accounts for context-dependent effects on DNA-binding without limiting combinatorial diversity of randomized finger libraries. Proteins selected by this new strategy exhibit improved DNA-binding affinities and specificities compared with counterparts constructed by modular or bi-partite assembly approaches.
A general strategy for selecting high-affinity zinc finger proteins for diverse DNA target sites.
Greisman HA, Pabo CO.
Howard Hughes Medical Institute and Massachusetts Institute of Technology, Cambridge, MA, USA.
Science 275:657-661, Jan. 31, 1997.
Summary: The authors describe a strategy for creating multi-finger proteins in which each finger in the final protein is sequentially selected from large randomized libraries. This approach yields combinations of fingers that work well together to bind the intended target DNA site with high affinity and specificity (as demonstrated in vitro).
[Discovery Medicine, 3(19):32-35, 2003]