Table 2: Binding sites and identity of ZFPs used in VEGF activation
We then generated artificial transcription factors by fusing the three-finger domains to either the p65 or VP16 activation domains and inserting this construct into an expression plasmid. Fig. 3A shows the locations of binding sites for these ZFPs in the VEGF promoter. The human VEGF promoter contains two DNase I–hypersensitive regions that are accessible to transcriptional factors; binding of engineered ZFP transcription factors to these sites activates VEGF gene expression27. We found that regardless of the location of the ZFP binding sites, the four ZFPs we tested activated not only the luciferase reporter gene under the control of the VEGF promoter (Fig. 3B), but also the endogenous VEGF gene itself (Fig. 3C). Also, an enzyme-linked immunosorbent assay (ELISA) performed with medium from the transiently transfected cells indicated that these ZFPs upregulated production of the VEGF protein 13- to 21-fold (Fig. 3D). Control cells that had been transfected with the control plasmid (which contained no ZFP-coding sequences) did not show any change in VEGF mRNA or protein levels. One ZFP, termed F83, did not affect the levels of VEGF mRNA or protein in these assays (Figs. 3C,D). This may be due to the binding of some other protein at the ZFP target site or to the local chromatin structure, which might have rendered the target site inaccessible to the ZFP. There was no strong correlation between the levels of VEFG expression by these ZFPs and their DNA-binding affinities or their expression levels in cells (data not shown).
Figure 3: Regulation of expression of the endogenous VEGF gene by ZFPs assembled by domain shuffling.
(A) Binding sites in the VEGF promoter for the ZFPs used in this study. '+1' indicates the transcription initiation site; 'P' and 'V' represent the p65 and VP16 activation domains, respectively, which had been fused individually to the sequence of the C terminus of the ZFPs. (B) Activation of the luciferase reporter gene under the control of the VEGF promoter (nucleotides -950 to +450) by ZFPs. Plasmids encoding ZFP activators were co-transfected into HEK 293 cells along with a reporter construct encoding the firefly luciferase gene under the control of the VEGF promoter. Luciferase activity was measured 2 d after transfection. (C) Activation of the endogenous VEGF gene by the indicated ZFP activators. We measured, by quantitative RT-PCR, the relative levels of VEGF mRNA in HEK 293 cells that had been transiently transfected with the corresponding ZFPs. (D) Amounts of VEGF protein secreted from HEK 293 cells that had been transiently transfected with the corresponding ZFPs; measured with an ELISA assay kit (Chemicon, Temecula, CA). 'Vector' indicates a control vector that encode no ZFPs. An 'irrelevant ZFP' was also used as a negative control.
To investigate the specificity of ZFP transcription factors on a genome-wide scale, we performed DNA microarray experiments with 293 cell lines that had been stably transfected with DNA constructs encoding each of the following three zinc-finger transcription factors: F121-p65, F435-p65, and F475-VP16. Fifty-one out of 7,458 genes were co-regulated (49 were upregulated and 2 were downregulated more than twofold) by all three ZFP transcriptional activators (Supplementary Table 3 online). Most of these co-regulated genes seem to be closely associated with VEGF function. Many are regulated by VEGF, involved in angiogenesis or hypoxia, or expressed in vascular endothelial cells. Therefore, it is likely that these genes are downstream targets of VEGF. We note, however, that dozens of other genes were regulated by one or two of our ZFP activators but not by all three (data not shown). Considering that these ZFPs that recognize nine bp sites, it is possible that they regulate genes other than VEGF. The use of four-, five-, or six-finger proteins should help to improve the specificity. Taken together, these data indicate that our ZFPs, assembled by shuffling naturally occurring zinc-finger domains, function in cells as transcriptional regulators of specific genes.
Discussion
Two approaches exist for making DNA-binding proteins that bind to predetermined DNA sequences. In one, site-directed mutagenesis is used to replace amino acid residues at a few key positions with residues that interact with the desired bases in the DNA target sequence27-29. The zinc finger–DNA 'recognition code', compiled from extensive biochemical and structural studies by many groups11-13, has made this approach practical. The other approach is based on zinc-finger phage display2-7. ZFPs with altered DNA-binding specificities can be selected from a library of zinc-finger variants in which the amino acid residues at key positions are randomly mutated.
Our approach, termed GeneGrip, differs from these approaches in several ways. (i) Our selection and screening procedures8-10 are done in vivo. Since genomic DNA is packaged as chromatin in eukaryotic cells, we reasoned that ZFPs selected in vivo may be more useful for regulating gene expression in higher eukaryotic cells than those selected in vitro (using phage display) or in prokaryotic cells30. Indeed, using our system we were able to construct ZFPs that functioned efficiently in human cells. (ii) Unlike previous approaches, in which mutations were incorporated at key positions in a given zinc-finger framework27, 31-33, our approach uses intact, wild-type zinc fingers derived from DNA sequences in the human genome. A naturally occurring linker sequence was used to connect individual zinc fingers. ZFPs composed of human zinc fingers may be preferable in therapeutic applications because they would be less likely to induce a host immune response than would proteins composed of mutated zinc fingers (although an immune response involving the linkers between the ZFP moieties and the effector domains is still possible.) (iii) Unlike previous protocols such as phage display selection, our shuffling approach is easily scalable. Thousands of highly active ZFPs can be constructed simultaneously. For example, shuffling 20 domains to make three-finger proteins would yield 8,000 (= 20 
20 20) ZFPs in a single step.
Designer zinc-finger transcription factors generated using the GeneGrip technique could be used to regulate the expression of as yet uncharacterized genes in vivo, an approach that could help in identifying new genes and determining their functions. In addition, a ZFP library could be screened to isolate improved phenotypes induced by specific ZFPs. Our technology should thus find wide application in basic research, medicine, and biotechnology.
Experimental protocol
Construction of a yeast expression plasmid for a zinc-finger library.
We constructed an expression plasmid (termed pPCFM-Zif) encoding a zinc-finger transcription factor by modification of pPC86 (ref. 34). To construct plasmids encoding human zinc-finger libraries, DNA segments from human genomic DNA encoding zinc fingers were amplified by PCR and the 100 bp PCR products were inserted into pPCFM-Zif. The plasmid library was prepared from a total of 1.2 106 E. coli transformants. Gap-repair cloning35 was also used to construct plasmids that encode individual zinc-finger domains. The construction of these plasmids are described in detail in the Supplementary Experimental Protocol online.
Reporter plasmids were prepared by inserting one of 64 pairs of complementary oligonucleotides (primer 4 in Supplementary Table 4 online) that contained three copies of a 9-bp target sequence into pRS315His (a gift from R. Reed at Johns Hopkins University's School of Medicine, Baltimore, MD) and pLacZi (Clontech, Palo Alto, CA).
In vivo selection of zinc-finger domains.
Yeast mating was used to facilitate identification of zinc fingers that bind to each 3-bp target site. The binding affinity and specificity of each zinc finger fused to fingers 1 and 2 of Zif268 were determined both in yeast and by EMSA. These methods are described in detail in Supplementary experimental protocol online.
Construction of three-finger proteins using selected zinc fingers as modular building blocks.
A mammalian expression plasmid, pcDNA3 (Invitrogen, Carlsbad, CA), was used as a parental vector for expressing ZFPs in mammalian cells. DNA segments that encode individual zinc-finger domains were subcloned into the plasmid, and the resulting plasmids were used as starting material for ZFP construction. The scheme for constructing plasmids that encode new three-finger proteins is outlined in Supplementary Figure 2 online. The constructed ZFPs were tested for their DNA binding ability and affinity in mammalian cells as described previously24-26. The reporter plasmid for the assay was constructed using pGL3-TATA/Inr24-26 which harbors the firefly luciferase gene as the reporter. The sequence of and method for making the reporter are described in Supplementary Figure 3 online. In addition, SELEX was performed to test whether these proteins recognize the appropriate target DNA sequences. These methods are described in detail in Supplementary Experimental Protocol online.
Note: Supplementary information is available on the Nature Biotechnology website.
Received 6 August 2002; Accepted 3 January 2003; Published online 18 February 2003.
REFERENCES
Brent, R. & Ptashne, M.A. Eukaryotic transcriptional activator bearing the DNA specificity of a prokaryotic repressor. Cell 43, 729-736 (1985). | ChemPort | Rebar, E.J. & Pabo, C.O. Zinc-finger phage: affinity selection of fingers with new DNA-binding specificities. Science 263, 671-673 (1994). | PubMed | ISI | ChemPort | Jamieson, A.C., Kim, S.-H. & Wells, J.A. In vitro selection of zinc fingers with altered DNA-binding specificity. Biochemistry 33, 5689-5695 (1994). | PubMed | ISI | ChemPort | Choo, Y. & Klug, A. Toward a code for the interactions of zinc fingers with DNA: selection of randomized fingers displayed on phage. Proc. Natl. Acad. Sci. 91, 11163-11167 (1994). | PubMed | ChemPort | Wu, H., Yang, W.P. & Barbas, C.F. III. Building zinc fingers by selection: toward a therapeutic application. Proc. Natl. Acad. Sci. USA 92, 344-348 (1995). | PubMed | ChemPort | Greisman, H.A. & Pabo, C.O. A general strategy for selecting high-affinity zinc-finger proteins for diverse DNA target sites. Science 275, 657-661 (1997). | Article | PubMed | ISI | ChemPort | Isalan, M., Klug, A. & Choo, Y. A rapid, generally applicable method to engineer zinc fingers illustrated by targeting the HIV-1 promoter. Nat. Biotechnol. 19, 656-660 (2001). | Article | PubMed | ISI | ChemPort | Wang, M.M. & Reed, R.R. Molecular cloning of the olfactory neuronal transcription factor Olf-1 by genetic selection in yeast. Nature 364, 121-126 (1993). | Article | PubMed | ISI | ChemPort | Cheng, X., Boyer, J.L. & Juliano, R.L. Selection of peptides that functionally replace a zinc finger in the Sp1 transcription factor by using a yeast combinatorial library. Proc. Natl. Acad. Sci. USA94, 14120-14125 (1997). | Article | PubMed | ChemPort | Bartsevich, V.V. & Juliano, R.L. Regulation of the MDR1 gene by transcriptional repressors selected using peptide combinatorial libraries. Mol. Pharmacol. 58, 1-10 (2000). | PubMed | ISI | ChemPort | Desjarlais, J.R. & Berg, J.M. Redesigning the DNA-binding specificity of a zinc-finger protein: a data base-guided approach. Proteins Struct. Funct. Genet. 12, 101-104 (1992). | PubMed | ISI | ChemPort | Choo, Y. & Klug, A. Physical basis of a protein-DNA recognition code. Curr. Opin. Struct. Biol. 7, 117-125 (1997). | Article | PubMed | ISI | ChemPort | Wolfe, S.A., Greisman, H.A., Ramm, E.I. & Pabo, C.O. Analysis of zinc fingers optimized via phage display: evaluating the utility of a recognition code. J. Mol. Biol. 285, 1917-1934 (1998). | Article | ISI | Pavletich, N.P. & Pabo, C.O. Zinc finger-DNA recognition: crystal structure of a Zif268-DNA complex at 2. 1Å. Science 252, 809-817 (1991). | PubMed | ISI | ChemPort | Elrod-Erickson, M., Benson, T.E. & Pabo, C.O. High-resolution structures of variant Zif268-DNA complexes: implications for understanding zinc finger-DNA recognition. Structure 6, 451-464 (1998). |