http://biotech.nature.com • FEBRUARY 2002 • VOLUME 20 • nature biotechnology

Engineering polydactyl zinc-finger transcription factors

Roger R. Beerli1,2 and Carlos F. Barbas, III1*

The availability of rapid and robust methods for controlling gene function is of prime importance not only forassigning functions to newly discovered genes, but also for therapeutic intervention. Traditionally, gene func-tion has been probed by often-laborious methods that either increase the level of a gene product or decreaseit. Advances now make it possible to rapidly produce zinc-finger proteins capable of recognizing virtually any18 bp stretch of DNA—a sequence long enough to specify a unique address in any genome. The attachmentof functional domains also allows the design of tailor-made transcription factors for specific genes. Recentstudies demonstrate that artificial transcription factors are capable of controlling the expression of endoge-nous genes in their native chromosomal context with a high degree of specificity in both animals and plants.Dominant regulatory control of expression of any endogenous gene can be achieved rapidly and can be alsoplaced under chemical control. A wide range of potential applications is now within reach.

REVIEW

Phenotypic variation results not only from changes in genes them-selves but also from changes in the coordination of gene expression andin expression levels that together tremendously broaden the phenotyp-ic spectrum available to an organism. Transcription is controlled bytranscription factors. Natural transcription factors usually have at leasttwo functional domains, a DNA-binding domain and an effectordomain. The DNA-binding domain acts as a targeting device to localizethe protein to a specific site on chromosomal DNA, whereas the effec-tor domain functions to direct the localization of specific enzymes tothis site, ultimately enabling transcription of the gene to be up- ordownregulated1. Attempts at artificial control of gene expression areoften based on the application of this fundamental principle.

To this end, large numbers of effector domains have been describedthat mediate either gene activation2 or repression3. These domains areusually modular and retain their function when attached to a heterolo-gous DNA-binding domain. However, for the construction of designertranscription factors, DNA-binding domains with customized speci-ficities have become available only recently. This review summarizesrecent progress in the production of zinc-finger DNA-bindingdomains, describes the first successful applications of artificial tran-scription factors, and discusses the potential use of these proteins inresearch and therapy.

Engineering zinc-finger DNA-binding specificityThe Cys2-His2 zinc-finger domain represents the most common DNA-binding motif in eukaryotes and the second most frequently encodedprotein domain in the human genome4–6. Some 700 or more of our30,000–40,000 genes encode zinc-finger domains of this type, ∼ 2% ofour genes. Although these approximately 30 amino acid domains canhave functions beyond DNA binding, they typically function by bind-ing 3 base pairs of DNA sequence. This structurally simple ββα domainis stabilized by hydrophobic interactions and the coordination of a zincion by the eponymous cysteine and histidine residues7. Cys2-His2 zinc-finger domains are particularly well suited for the construction of syn-

thetic transcription factors as they are commonly arranged as covalenttandem repeats, allowing the recognition of extended asymmetricalsequences. This modularity in both structure and function is the keyadvantage of zinc-finger proteins (ZFPs) compared with other types ofDNA-binding domains that typically recognize DNA as dimers and usenonmodular recognition domains8.

Sequence-specific DNA recognition in Cys2-His2 zinc-fingerdomains is achieved by presentation of an α-helical reading head intothe major groove of the double helix. The first x-ray crystal structure ofa ZFP bound to DNA was that of the murine three-finger transcriptionfactor Zif268 and revealed that base-specific interactions are made byamino acids protruding from the N terminus of the α-helix7,9. Contactsare primarily made with one strand of the DNA double helix, withpositions –1, 3, and 6 of each finger (with respect to the start of the α-helix) contacting the 3′-, middle, and 5′-nucleotides of a 3 bp subsite,respectively.

To further elucidate the principles of zinc-finger DNA recognition,rational design10–12, the selection of large combinatorial libraries dis-played on phage13–17, and a combination of these two approaches havebeen applied18–20. Indeed, phage display of ZFPs was one of the firstapplications of protein phage display proposed21. Selection protocolsoften lead to impressive amino acid conservation for recognition of thesame nucleotide in different target sequences13–20. Most notably, anarginine was typically found at positions –1 and 6 for recognition of a 3′and 5′ guanine, respectively.

Other, albeit less impressive, amino acid conservations have alsobeen observed. However, although these studies laid the foundation forfuture applications of designer ZFPs, early hopes that DNA recognitioncould be explained by a simple recognition code of one amino acid toone nucleotide have not materialized. DNA binding is more complex,and the residues flanking the primary contact residues at positions –1,3, and 6 play important roles in specificity, both allowing specific inter-actions to be made and excluding bases from the recognition site18–20.Practically, this means that design alone is not sufficient and that selec-

1The Skaggs Institute for Chemical Biology and the Department of Molecular Biology, The Scripps Research Institute, La Jolla, CA 92037. 2Current address: CytosBiotechnology AG, Wagistrasse 21, 8952 Zurich-Schlieren, Switzerland. *Corresponding author (carlos@scripps.edu).

135

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://b

iote

ch.n

atu

re.c

om

tion strategies must remain an integral part in the modification of zinc-finger DNA-binding specificity. Domains obtained through selectiveapproaches are, however, not fail-safe with respect to specificity andoften benefit from refinement through a rational design cycle18–20.

Polydactyl proteinsJust how many fingers are required to specify a unique address in agenome as complex as ours? Assuming random base distribution, anygiven 16 bp sequence will only occur once every 4.3 billion nucleotidesand would be sufficiently long to specify a unique address in the 3.5 bil-lion bp human genome. An 18 bp address would be specific within 68billion bp—or 20 human genomes—and could be targeted by a poly-dactyl protein containing six zinc-finger domains22. One might alsoargue that chromatin infrastructure serves to protect most of thegenome from DNA-binding proteins and the target sequence lengthrequired for specific regulation in the chromosomal context is muchlower. In any case, the genomic specificity of any designed transcriptionfactor needs to be probed empirically.

Specificity can also be obtained by combinatorial interactions ofZFPs with shorter recognition sequences on the target gene providedthat the individual constituents alone are inactive. Binding can eitherbe independent, if the two 9 bp sites are far apart, or mediated by spe-cific dimerization domains, if the two sites are close23,24. Novel synthet-ic peptides suitable for the heterodimerization of ZFPs have alsorecently been reported25. The use of a single six-finger protein, however,has the important advantage that it requires only one gene to be deliv-ered18,20,22,23,26–28.

A variety of zinc-finger-based recognition devices have now beenreported (see Fig. 1). Three phage display strategies for the constructionof such polydactyl proteins have been described and involve either par-allel, sequential, or bipartite selection (Fig. 2).All methods have distinctadvantages and disadvantages when compared with one another, andthe protein products from any given methodology should be rigorous-ly characterized8. The recent development of rapid DNA array assays asapplied to DNA–protein interactions should assist in this analysis.Microarrays containing all possible 3 bp binding sites for a given

domain within a zinc-finger protein can permit a rapid assessmentof any given domain’s specificity29,30. This approach is conceptional-ly similar to the ELISA-based assays that are typically employed inthese studies but allows a more rapid and quantitative assay of bind-ing against a much larger array of oligonucleotides. For the assess-

ment of the binding specificity of more than one binding domain,selection of the preferred binding site from random sequence poolsremains the method of choice as many billions of sequences can beprobed this way31.

Parallel selectionWork in our laboratory has focused on the development and use of theparallel selection strategy. The basic assumption of this approach is thatzinc-finger domains are functionally independent and can therefore berecombined with one another in any desired sequence (Fig. 2A). Thus,a complete zinc-finger domain complement of the genetic code wouldconsist of 64 domains specific for each DNA triplet. Stitching togetherpredefined zinc-finger domains in the appropriate sequence thenallows any DNA sequence to be targeted. Zinc fingers can either belinked together using the canonical linker22, or using one or more non-canonical linkers27,28,32. The use of some noncanonical linker peptideshas been reported to lead to improvements in DNA affinity and speci-ficity for certain proteins. While a consensus in the field has not yetbeen formed as to the most optimal noncanonical linker and (perhapsmore importantly) the most general strategy for domain linking, a vari-ety of approaches all seem to work sufficiently well. Ideally, no furtherselection of the assembled protein is needed in this approach.

To date, we have described zinc-finger domains specific for half thegenetic code. These domains bind the 5′-GNN-3′ and 5′-ANN-3′(where N is any one of the four nucleotides) families of DNA tripletsand were produced by phage display selection protocols followed byoptimization through systematic site-directed mutagenesis studies18–20.Although these domains constitute half of the possible domains, it isimportant to consider how many polydactyl proteins these domainscan access. The 5′-GNN-3′ set of domains alone is sufficient for theconstruction of 1.7 × 107 new proteins binding the 5′-(GNN)6-3′ fami-ly of DNA sequences. Assuming random base distribution, 5′-(GNN)6-3′ sequences should occur approximately every 2,048 bp; however, aseukaryotic promoter regions are relatively G+C rich, 5′-(GNN)6-3′sequences are more abundant and multiple sites are found in mosthuman promoters. For instance, a 1,400 bp fragment of the human

REVIEW

nature biotechnology • VOLUME 20 • FEBRUARY 2002 • http://biotech.nature.com136

Figure 1. Zinc-finger–based DNA recognition and modificationdevices. (A) Transcription of transgenes and endogenous target genescan be regulated by fusing designer ZFPs recognizing asymmetricalDNA sequences of varying length to transcriptional effector domains.Transcription factors containing three zinc-finger domains andrecognizing 9 base pairs (bp) of DNA sequence have been describedto mediate gene activation or repression22,35,51,52. (B) For genomicspecificity, transcription factors containing six zinc-finger domains thatrecognize 18 bp of sequence have been reported20,22,26. (C) Genes canbe placed under the control of a chemical inducer by fusion of a ZFPwith the ligand-binding domain of a nuclear hormone receptor, aprocess involving the formation of homodimers. The dimeric formrecognizes 18 bp of DNA through the interaction of paired three-fingerproteins23,57. (D) Monomeric gene switches that directly bind 18 bp ofDNA sequence are produced by fusion of six zinc-finger domains andan effector domain to a designed single-chain nuclear hormonereceptor ligand-binding domain23. (E) Peptides can be selected fromlibraries for the dimerization of proteins containing two zinc-fingerdomains25. (F) Various types of other interaction domains, such as theleucine zippers of Jun and Fos, have been used to direct theheterodimerization of two zinc-finger proteins24. (G) Fusion with anendonuclease domain from Fok1 mediates dimerization that restoresnuclease activity capable of directing the introduction of sequence-specific double-strand breaks, for example, to stimulate homologousrecombination60. 2F, two-finger protein; 3F, three-finger protein; 6F, six-finger protein; ED, effector domain; LBD, nuclear hormone receptorligand-binding domain.

EDEDED

6F

ED

LBDLB

D

6F

ED

3F

3F3F 3F

3F3F

2F 2F 2F2F

Fos Jun

3F

ED

LBD

Fok1

A B

C D

E F

G

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://b

iote

ch.n

atu

re.c

om

REVIEW

erbB-2 promoter contains 11sequences of 5′-(GNN)6-3′ (seeFig. 3). Practically, this meansthat every gene might be regulat-ed by targeting 5′-(GNN)6-3′sequences.

Although the number of suit-able target sites in a given gene ofinterest may be sufficient, notevery target may be available forbinding by an artificial transcrip-tion factor because of local chro-matin structure, covalent modifi-cation, or occupation of the siteby an endogenous DNA-bindingprotein. If we now consider thataddition of the 5′-(ANN)-3′binding domains20 provide tar-geting of 5′-(RNN)6-3′ (where R = G or A) sequences, more than onebillion transcription factors become accessible. Binding sites for theseproteins should occur every 32 bp, and an analysis of the 1,400 bphuman erbB-2 promoter fragments reveals the presence of 77 suchsites. Thus, with just half the set, ∼ 27,000 times more transcription fac-tors can be constructed than there are genes in the human genome. Todate, the parallel selection approach is the only one validated at the levelof endogenous gene regulation.

The parallel selection approach for the production of new specifici-ties has the important advantage that it uses predefined domains andshould not require additional design or selection, making it extremelyrapid and accessible to any laboratory. However, it should be noted thatcomplete functional independence of zinc-finger domains oversimpli-fies their binding mechanism in some cases. The four 5′-(GNG)-3′zinc-finger domains appear to specify a 4 bp site rather than the typical3 bp site. This was first noted on inspection of the Zif268 structure, inwhich an aspartate in position 2 of finger 2 contacts the binding site offinger 1, forcing that site to be either GNN or TNN (refs 7,9,33,34).

Fortunately, several strategies can be used to overcome or avoidproblems imposed by target site overlap resulting from the occurrenceof an aspartate at position 2 of the 5′-GNG-3′ domains. First, poly-dactyl proteins can be produced within the structural constraintsimposed by target site overlap. Specifically, the exclusive use of 5′-GNN-3′ or 5′-TNN-3′ domains avoids this problem, as does the place-ment of a 5′-GNN-3′- or 5′-TNN-3′-type finger located N-terminal ofany 5′-GNG-3′ finger in otherwise mosaic proteins constructed with5′-ANN-3′ or 5′-CNN-3′ domains20,26,35. Alternatively, if the necessityarises to combine two fingers in an otherwise incompatible manner,residues at the interface of the two α-helices including position 2 can berandomized, allowing selection of optimal amino acid residues on acase-by-case basis36. The practical implications of target site overlap tothe parallel approach exist primarily on paper. In reality, the large num-ber of predefined zinc-finger domains currently available makes thisproblem a practical nullity for gene targeting.

Sequential selectionThe sequential selection protocol pioneered by the Pabo laboratoryaddresses the problems imposed by target site overlap and cooperativi-ty in zinc-finger DNA recognition (Fig. 2B). As the name implies,DNA-binding specificities of individual zinc-finger domains are alteredsequentially in the context of the other zinc fingers, rather than in par-allel17. Thus, finger three of the three-finger protein Zif268 is replacedby a finger one in which the critical amino acid residues have been ran-domized. This library is then selected in the context of the two originalfingers, which serve as “anchors”.After selection, the N-terminal anchorfinger is removed and a finger two library is attached to the C terminus.

Selection of this library ensures that the new finger two works well inthe context of the finger one selected in the previous round. In the finalstep, the last remaining anchor finger is discarded and a randomizedfinger three is attached to the C terminus, again followed by selection.In this manner, each finger of the new three-finger protein is selected inthe context of its neighboring finger, preventing any potential problemscaused by target site overlap.

The target specificity of these three-finger proteins is similar to thatof stitched proteins8,35,37. Although sequential selection undoubtedly isa useful method, it does require the generation and selection of six zinc-finger libraries for each protein, making this approach inaccessible tomost laboratories. Sequential selection may be essential for targetingsome sequences wherein no flexibility exists in target choice, for exam-ple, in using the zinc finger to protect a specific site in the genome toprobe the function of an endogenous factor binding at the same site.

Recently the crystal structure of a sequentially selected protein incomplex with its TATA box target sequence has been reported38. Thisstructure demonstrates how interwoven the interactions of zinc fingerscan be when this type of sequential selection strategy is applied, imply-ing that proteins developed using the three methods described here willlikely use different mechanisms for the molecular recognition of DNA.

Bipartite selectionThis recently developed method aims at combining the distinct advan-tages of the two approaches described above39. It makes use of a pair ofprefabricated zinc-finger libraries, in each of which one-and-a-half fin-gers of the three-finger ZFP Zif268 are randomized (Fig. 2C). Selectionof these two libraries is carried out in parallel against DNA sequences inwhich either the first or the last 5 bp of the 9 bp Zif268 target site areexchanged against a sequence of interest. After phage display selection,the two libraries are combined and further rounds of selection are per-formed on the 9 bp sequence of interest. The bipartite selectionapproach has yielded several three-finger proteins binding predefinedsequences in various regions of the HIV-1 promoter, with Kd values inthe nanomolar range. Though not extensively analyzed, these proteinsappear to have specificities similar to those produced using the paralleland sequential selection strategies.

The bipartite selection strategy represents a potential advancein the field of zinc-finger engineering, as not only concerns of tar-get site overlap and zinc-finger cooperation are addressed, but alsothe time required to evolve new specificities is dramatically short-ened compared with sequential selection. Although this approachmay allow the targeting of any DNA sequence, it is important tokeep in mind that it does involve many rounds of phage displayselection. For the regulation of a specific gene of interest, typicallythere is a lot of flexibility in the choice of the precise target sitelocation. As a result, the use of predefined domains that do not

http://biotech.nature.com • FEBRUARY 2002 • VOLUME 20 • nature biotechnology 137

Figure 2. Strategies for the production of ZFPs with desired DNA-binding specificity. (A) Parallel selection. (B)Sequential selection. (C) Bipartite selection. Asterisks indicate preselected libraries. See text and references fordetails.

A B C

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://b

iote

ch.n

atu

re.c

om

require additional design or selection likely remains the fastestroute to gene regulation.

Alternative selection strategiesTypically, phage display strategies require multipleenrichment/amplification cycles and are rather time-consuming,especially if DNA-binding domains with new specificities have tobe evolved de novo. As an alternative to phage display, severalgenetic screens have been described that are suitable for the selec-tion of new DNA-binding specificities.

A genetic screen in bacteria has been used to produce mutantsof the basic region-leucine zipper protein C/EBP that recognizedsequences differing in two of its five half-site nucleotides40. Inaddition, zinc-finger domains with altered DNA-binding specifici-ty can be identified in a genetic screen, as has recently been shownby the Pabo laboratory41. In this work, a bacterial variant of theyeast two-hybrid system has been developed and employed for theselection of zinc-finger domains with novel specificities. Thenewly selected zinc finger variants have affinities and specificitiescomparable to those obtained by phage display17. Finally, newDNA-binding specificities could also be produced in eukaryoticcells, for example, using a yeast one-hybrid screen, a method com-monly used for the identification of DNA-binding proteins42,43.

Trans dominant gene regulationFor the practical regulation of a given gene, one simply requires aZFP to regulate its target gene and not other genes with sequencesrelated to the target gene. As a multitude of protein engineeringstudies have shown in other systems, affinity, specificity, and stability of most proteins can be further improved by in vitroevolution methods. The designer transcription factor need notbind DNA with the highest possible affinity or specificity as longas it can act to specifically regulate the target gene. There does,however, appear to be a threshold time for DNA occupancy neces-sary for transcriptional signaling, but exceeding this occupancytime does not appear to provide additional gains when effectordomains are used. This threshold time appears to be met with pro-teins with DNA-binding affinities of ∼ 10 nM or better, and to date,endogenous gene regulation has been accomplished only withsuch proteins.

While a six-finger protein provides the obvious mathematicaladvantage in terms of specificity, a three-finger protein may have suf-ficient specificity when accounting for such factors as chromatin

structure and promoterproximity. The distinctadvantage of a transcriptionfactor–based approach isthat both gene repressionand activation can beorchestrated in trans.

Gene repressionWith respect to gene repres-sion, it is possible to make afunctional knockout of agene by simply adding atransgene rather than resort-ing to breeding homozygous knockout organisms usingmore traditional approaches.Theoretically, gene repres-sion by designer DNA-bind-ing proteins can be achievedin a number of ways, includ-

ing the following: (1) fusing a DNA-binding domain to a transcription-al repression domain; (2) targeting a site in the transcribed region of agene and interfering with transcription elongation; (3) targeting asequence in the immediate vicinity of the transcription initiation siteand interfering with the assembly of the transcription machinery; or(4) targeting a site upstream of a gene of interest that is naturally boundby a coactivator, thus preventing coactivator binding. The first three ofthese strategies have been investigated on transiently transfectedreporter plasmids, on integrated reporter constructs, or on endogenousgenes. These approaches should provide access to the spectrum of geneknockdown to knockout phenotypes.

Our laboratory has concentrated on the first strategy using zinc-fin-ger–effector domain fusion proteins. One important advantage of thisstrategy is that it provides great flexibility in the choice of the targetsequence, as most effector domains act in a distance- and orientation-independent manner. Furthermore, the use of effector domains allowsnot only the downregulation of gene activity, but also upregulation oractivation (see below).

Several modular transcriptional repressor domains have beendescribed and can be incorporated into designed transcription factors.These include the Krüppel-associated box (KRAB), a domain com-monly found at the N terminus of naturally occurring zinc-finger pro-teins44, and the mSin3 interaction domain (SID), a short domain foundat the N terminus of Mad that mediates repression via recruitment of ahistone deacetylase45.When fused to the six-finger protein E2C that wasdesigned to bind an 18 bp sequence in the 5′-untranslated region(UTR) of the human proto-oncogene erbB-2, both domains mediatedeffective repression of an erbB-2 promoter luciferase reporter constructin transiently transfected HeLa cells35. Regulation of endogenous geneswas explored using the six-finger proteins E2C and E3 fused to a KRABdomain26. E3 binds to an 18 bp sequence in the 5′-UTR of the humanerbB-3 gene that is identical to the E2C sequence in 15 out of 18 bases.These zinc-finger proteins bind their respective DNA targets with ∼ 0.5nM affinity. Homologous targeting sites were chosen in the two genesto ascertain whether specific gene regulation was possible using ZFPs(i.e., independent regulation through closely related binding sites).Significantly, studies demonstrated that the transcription factors effi-ciently repressed their respective target genes in a highly specific man-ner26. The lack of cross-regulation in this system demonstrates the veryhigh degree of specificity attainable using polydactyl zinc-finger tran-scription factors. Moreover, the erbB-2 transcription factor was shownto regulate its endogenous gene target in mouse, monkey, and humancells that conserve its binding site. Additional regulators of the endoge-

REVIEW

nature biotechnology • VOLUME 20 • FEBRUARY 2002 • http://biotech.nature.com138

Figure 3. Structure of a 1,400 bp human erbB-2 promoter fragment. The major site of transcription initiation (light graytriangle) and the position of the ATG translation initiation codon (dark gray triangle) are indicated. The fragment wassearched for the presence of suitable 18 bp target sequences. White circles, 5´-(GNN)6-3´ sites; black circles, 5′-(RNN)6-3′ sites (R = guanine or adenine).

6xGNN

100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400

transcription start

translation start

6xRNN

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://b

iote

ch.n

atu

re.c

om

REVIEW

nous erbB-2 and erbB-3 genes have also been reported20.The use of DNA-binding domains lacking effector domains for gene

regulation has also been studied extensively. This strategy attempts toplace a steric blockade in the path of RNA polymerase. Klug and col-leagues46 have transiently expressed a three-finger ZFP in BaF3 cellspreviously made interleukin 3 (IL-3)-independent by integration of aplasmid encoding a p190BCR-ABL transgene. The ZFP-transfected cellsreverted to IL-3 dependence through repression of p190BCR-ABL expres-sion. These results are remarkable as the ZFP employed was designedto bind deep within the transcribed region of the gene and had a ratherlow affinity (Kd = 0.6 µM). Quantitative studies from the Pabo labora-tory using ZFPs that bind 1,000 times more strongly have revealed atmost twofold repression of basal transcription and no repression ofactivated transcription by binding within a transcribed gene47. Kangand Kim48 have also reported no repression upon targeting a site >50bp downstream of the transcriptional start site of a reporter gene, againwith a protein with far greater affinity for DNA. Consistent with thesereports, our own studies with the six-finger protein E2C that bindswithin the transcribed region of the erbB-2 gene with subnanomolaraffinity, yielded a modest 30% repression when this ZFP was expressedwithout a fused repression domain35. In contrast, fusion of E2C to aKRAB domain could effect complete repression of the endogenousgene. Taken together, these studies suggest that a polymerase blockademechanism is typically not very effective.

Pabo and coworkers47–49 have also evaluated the potential of repress-ing transcription by interfering with the assembly of the transcriptionmachinery (i.e., mechanism 3 described above). In initial transienttransfection studies, Zif268 binding sites were placed in an artificialpromoter, at various distances from the TATA box or the transcriptioninitiation site47. Substantial repression of both basal and activated tran-scription was found when Zif268 bound either close to the TATA boxor to the initiator element. Efficient repression by blocking access to theTATA box of a synthetic promoter was also shown with previouslyunknown fusion proteins consisting of Zif268 and the TATA-bindingprotein connected by a flexible peptide linker49. Finally, inhibition oftranscription by targeting sites close to the initiator element of synthet-ic promoter constructs has been recently shown using six-finger pro-teins, either upon transient transfection27 or after integration of thereporter construct into chromosomal DNA48.

Although these data convincingly demonstrate the feasibility of tar-geting the transcription initiation complex, downregulation of anendogenous gene based on this mechanism has yet to be shown.Importantly, the targeting of endogenous genes by this strategy has thecomplications that many promoters lack a TATA box and that manygenes use multiple transcription initiation sites.

Gene activationThe zinc-finger strategy is not limited to gene repression, but is readilyadapted to gene activation. Upregulation of fetal hemoglobin, forexample, could impact sickle-cell anemia, while upregulation ofgrowth hormone could impact dwarfism and controlled regulation oferythropoietin and vascular endothelial growth factor (VEGF) couldbe important in cancer therapy as well as diabetes.

Two major mechanisms can be employed to upregulate expressionof a specific gene: first, a DNA-binding domain can be fused to a tran-scriptional activation domain, upregulating transcription by activelyrecruiting the transcriptional machinery50; second, a DNA-bindingprotein can be designed to compete with a transcriptional repressorbound upstream of a gene of interest, leading to gene activation with-out the need for an effector domain.

The first strategy has the significant advantage that it allows flexibil-ity in choosing the precise location of the target sequence. Using zinc-finger-activation domain fusion proteins, our group has recentlyshown efficient upregulation of endogenous genes26. The six-finger

proteins E2C and E3 fused to the synthetic transcriptional activationdomain VP64, a derivative of the herpes simplex virus protein VP16,efficiently upregulated the endogenous human erbB-2 and erbB-3genes, respectively. Flexibility in target site location was demonstratedby the finding that both target sites are located within the 5′-UTR of therespective gene. Thus, efficient activation can be achieved by targetingnot only promoter sequences but also sites within the transcribedregion of a gene. Several additional six-finger activators of the endoge-nous human erbB-2 and erbB-3 genes have also been reported20.

Other studies have further demonstrated the potential of zinc-fin-ger-based transcriptional activators. Panels of three-finger proteinshave been designed and constructed to bind in various locations in thepromoter regions of the human erythropoietin and VEGF-A genes.Upon fusion with VP16 or p65 activation domains, the zinc-fingertranscription factors were capable of upregulating expression from theendogenous chromosomal loci51,52. These reports demonstrate the effi-cacy of three-finger proteins in activation and the manner in whichchromatin structure can effectively enhance their specificity. The acti-vation of endogenous VEGF-A illustrates an important advantage ofthis approach over simple cDNA delivery to enhance gene expression.The VEGF-A transcript is naturally processed into three major splicevariants, and proper functional responses require the appropriate rela-tive levels of these variants to be expressed. To date, this has only beenachieved with zinc-finger activation of the endogenous gene.

Regulating the regulatorsDesigner transcription factors with tailored DNA-binding specificityfacilitate useful strategies for manipulating gene expression, both incultured cells and in whole organisms. However, constitutive regulationof a given target gene may not always be desirable and reversible regula-tion of target gene expression, preferentially by a small chemical induc-er, would broaden application. Inducible gene regulation is particularlydesirable in a gene therapy setting, as it could limit side effects andenhance safety by rendering the modulation of gene activity reversible.

One way of regulating an endogenous gene in an inducible manneris to place the expression of the designed transcription factor under thecontrol of an inducible promoter. Several inducible expression systemshave been described23,53–55; one of the most prominent involves the useof an expression cassette regulated by a tetracycline-controlled transac-tivator53. We recently have demonstrated that even endogenous genescan be placed under chemically regulated control. The tetracycline-reg-ulated expression system was adapted for the control of the endoge-nous erbB-2 gene in human cervical carcinoma cells26. Stably integratedgenes encoding the repressor E2C–KRAB or the activator E2C–VP64were induced by removal of the tetracycline analog doxycycline, leadingto repression or activation of the endogenous erbB-2 gene. Level of thechemical inducer correlated well with the level of gene product.

One complication of inducible promoter systems is that they requirethe delivery of two genes: one encoding the ZFP under the control of aninducible promoter, the other encoding the regulatory protein.Delivery of multiple vectors is a laborious process and a hurdle in agene therapy setting. For the inducible control of target gene expres-sion, it would be much more simple and efficient to regulate the activi-ty of the transcription factor itself, rather than its expression.

Members of the nuclear hormone receptor protein superfamily areprototypical ligand-regulated transcription factors and offer featuresthat can be exploited in protein engineering56. Thus, to create chemical-ly regulated zinc-finger transcription factors, our group23 has preparedfusion proteins containing ligand-binding domains (LBDs) derivedfrom progesterone, estrogen, and ecdysone receptors. These types ofzinc-finger fusion proteins could prove useful not only for targetingendogenous genes, but also for the inducible regulation of transgenes.To achieve inducible regulation of gene expression, we have fuseddesigned ZFPs with novel DNA binding specificities to a transcription-

http://biotech.nature.com • FEBRUARY 2002 • VOLUME 20 • nature biotechnology 139

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://b

iote

ch.n

atu

re.c

om

al activation domain, as well as to the LBDs derived from either theestrogen or progesterone receptor (Fig. 1C). Together with optimizedminimal promoters, the progesterone- and estrogen-based transcrip-tion factors provide 4-hydroxytamoxifen- or RU486-inducible expres-sion systems with induction ratios of up to three orders of magnitude23.As distinct DNA-binding specificities were used, these inducible sys-tems are functionally independent and can be selectively switched onwithin the same cell, allowing the concomitant regulation of multipletransgenes.

The therapeutic potential of an estrogen-based zinc-finger systemhas now been demonstrated in vivo with the regulation of an aden-ovirus-delivered endostatin transgene in a mouse model57. In this sys-tem, endostatin expression could be controlled by the administrationof low levels of 4-hydroxytamoxifen to the mice.As the LBDs of nuclearhormone receptors mediate dimerization of their fused three-fingerproteins, these transcription factors act on symmetrical 18 bp sites, siteslong enough also for the specific targeting of endogenous genes.

To further enhance the potential of ligand-dependent zinc-fingertranscription factors for the regulation of endogenous genes, our grouphas used two LBDs connected by a flexible peptide linker. In this sys-tem, the response to the chemical inducer results in an intramolecularrearrangement, rather than dimerization, leading to transcriptionallyactive proteins (Fig. 1D). The significant advantage of this type offusion protein is that it allows targeting of extended asymmetricalsequences, which enables ligand-dependent regulation of any promot-er. By obviating the need to deliver multiple genes, these monomericsingle-chain regulators hold great promise for the inducible control ofgene expression.

PerspectivesWith the relative ease of preparation of designer transcription factors, avery broad range of applications is now possible. Diverse applicationsin gene therapy offer the possibility of selectively regulating endoge-

nous as well as foreign genes. The production of viral gene productsand transcripts, as well as the expression of oncogenes or bovinespongiform encephalopathy–perpetuating prions, might be silenced,whereas disease-fighting genes might be activated to counteract orcompensate for damaged genes of homologous function. With properdesign, entire gene families or biosynthetic pathways might be regulat-ed by a single zinc-finger transcription factor.

Aside from these potential therapeutic applications, designed tran-scription factors should also be of great use in basic research and DNA-based diagnostic applications.As these proteins act in a trans-dominantmanner, functional gene knockouts can be quickly realized, as recentlyreported in studies with Arabidopsis plants58,59. This strategy greatlyreduces the time required to produce a gene knockout in whole organ-isms and is readily adapted to the tissue-specific or chemicallyinducible knockout of gene expression. Disease resistance and otherenhanced agronomic phenotypes might readily be achieved by modu-lating endogenous gene expression. In other studies, fundamentalquestions regarding the mechanisms underlying the regulation of geneexpression and chromatin structure can now be addressed more effec-tively, as proteins can be placed at defined genomic sites.

While all of these applications result from the ability to reprogramthe software of the genome, an analogous approach may soon allowprecise genomic alterations to be orchestrated. Pieces of the genomemight be cut and pasted with surgical precision. Gene therapy mightbecome a more precise science if gene integration can be specificallytargeted, and one can imagine that integrated viruses might be forev-er removed by their controlled excision from the germline.Preliminary studies using zinc-finger–endonuclease fusions toenhance homologous recombination60 provide a glimpse at futureapplications of zinc-finger proteins in modifying the hardware of thegenome itself.

Received 27 June 2001; accepted 22 December 2001

REVIEW

nature biotechnology • VOLUME 20 • FEBRUARY 2002 • http://biotech.nature.com140

1. Ptashne, M. Control of gene transcription: an outline. Nat. Med. 3, 1069–1072(1997).

2. Seipel, K., Georgiev, O. & Schaffner, W. Different activation domains stimulatetranscription from remote (“enhancer”) and proximal (“promoter”) positions.EMBO J. 11, 4961–4968 (1992).

3. Hanna-Rose, W. & Hansen, U. Active repression mechanisms of eukaryotic tran-scription repressors. Trends Genet. 229–234 (1996).

4. Tupler, R., Perini, G. & Green, M.R. Expressing the human genome. Nature 409,832–833 (2001).

5. Venter, J.C. et al. The sequencing of the human genome. Science 291,1304–1351 (2001).

6. Consortium, I.H.G.S. Initial sequencing and analysis of the human genome.Nature 409, 860–921 (2001).

7. Pavletich, N.P. & Pabo, C.O. Zinc finger–DNA recognition: crystal structure of aZif268–DNA complex at 2.1 Å. Science 252, 809–817 (1991).

8. Segal, D.J. & Barbas, C.F., III. Design of novel sequence-specific DNA-bindingproteins. Curr. Opin. Chem. Biol. 4, 34–39 (2000).

9. Elrod-Erickson, M., Rould, M.A., Nekludova, L., & Pabo, C.O. Zif268 protein–DNAcomplex refined at 1.6 A: a model system for understanding zinc finger–DNAinteractions. Structure 4, 1171–1180 (1996).

10. Desjarlais, J.R. & Berg, J.M. Toward rules relating zinc finger protein sequencesand DNA binding site preferences. Proc. Natl. Acad. Sci. USA 89, 7345–7349(1992).

11. Nardelli, J., Gibson, T. & Charnay, P. Zinc finger–DNA recognition: analysis of basespecificity by site-directed mutagenesis. Nucleic Acids Res. 20, 4137–4144(1992).

12. Taylor, W.E. et al. Designing zinc–finger ADR1 mutants with altered specificity ofDNA binding to T in UAS1 sequences. Biochemistry 34, 3222–3230 (1995).

13. Rebar, E.J. & Pabo, C.O. Zinc finger phage: affinity selection of fingers with newDNA-binding specificities. Science 263, 671–673 (1994).

14. 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. USA91, 11163–11167 (1994).

15. Jamieson, A.C., Kim, S.-H. & Wells, J.A. In vitro selection of zinc fingers withaltered DNA-binding specificity. Biochemistry 33, 5689–5695 (1994).

16. Wu, H., Yang, W.-P. & Barbas, C.F., III. Building zinc fingers by selection: toward atherapeutic application. Proc. Natl. Acad. Sci. USA 92, 344–348 (1995).

17. Greisman, H.A. & Pabo, C.O. A general strategy for selecting high-affinity zinc fin-ger proteins for diverse DNA target sites. Science 275, 657–661 (1997).

18. Segal, D.J., Dreier, B., Beerli, R.R. & Barbas, C.F., III. Toward controlling gene

expression at will: selection and design of zinc finger domains recognizing each ofthe 5′-GNN-3′ DNA target sequences. Proc. Natl. Acad. Sci. USA 96, 2758–2763(1999).

19. Dreier, B., Segal, D.J. & Barbas, C.F., III. Insights into the molecular recognition ofthe 3′-GNN-3′ family of DNA sequences by zinc-finger domains. J. Mol. Biol. 303,489–502 (2000).

20. Dreier, B., Beerli, R.R., Segal, D.J., Flippin, J.D. & Barbas, C.F., III. Development ofzinc finger domains for recognition of the 5′-ANN-3′ family of DNA sequences andtheir use in the construction of artificial transcription factors. J. Biol. Chem. 276,29466–29478 (2001).

21. Barbas, C.F., III & Lerner, R.A. Combinatorial immunoglobulin libraries on the sur-face of phage (Phabs): rapid selection of antigen-specific Fabs. Methods:Companion Methods Enzymol. 2, 119–124 (1991).

22. Liu, Q., Segal, D.J., Ghiara, J.B. & Barbas, C.F., III. Design of polydactyl zinc-fingerproteins for unique addressing within complex genomes. Proc. Natl. Acad. Sci.USA 94, 5525–5530 (1997).

23. Beerli, R.R., Schopfer, U., Dreier, B. & Barbas, C.F., III. Chemically regulated zincfinger transcription factors. J. Biol. Chem. 275, 32617–32627 (2000).

24. Wolfe, S.A., Ramm, E.I. & Pabo, C.O. Combining structure-based design withphage display to create new Cys(2)His(2) zinc finger dimers. Structure Fold. Des.8, 739–750 (2000).

25. Wang, B.S. & Pabo, C.O. Dimerization of zinc fingers mediated by peptidesevolved in vitro from random sequences. Proc. Natl. Acad. Sci. USA 96,9568–9573 (1999).

26. Beerli, R.R., Dreier, B., & Barbas, C.F., III. Positive and negative regulation ofendogenous genes by designed transcription factors. Proc. Natl. Acad. Sci. USA97, 1495–1500 (2000).

27. Kim, J.S. & Pabo, C.O. Getting a handhold on DNA: design of poly-zinc finger pro-teins with femtomolar dissociation constants. Proc. Natl. Acad. Sci. USA 95,2812–2817 (1998).

28. Moore, M., Klug, A. & Choo, Y. Improved DNA-binding specificity from polyzinc fin-ger peptides by using strings of two-finger units Proc. Natl. Acad. Sci. USA 981437–1441 (2001).

29. Bulyk, M., Gentalen, E., Lockhart, D.J. & Church, G.M. Quantifying DNA–proteininteractions by double-stranded DNA arrays Nat. Biotechnol. 17, 573–577 (1999).

30. Bulyk, M.L., Huang, X., Choo, Y. & Church, G.M. Exploring the DNA-binding speci-ficities of zinc fingers with DNA microarrays Proc. Natl. Acad. Sci. USA 98,7158–7163 (2001).

31. Thiesen, H.J. & Bach, C. Target Detection Assay (TDA): a versatile procedure todetermine DNA binding sites as demonstrated on SP1 protein Nucleic Acids Res.

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://b

iote

ch.n

atu

re.c

om

REVIEW

http://biotech.nature.com • FEBRUARY 2002 • VOLUME 20 • nature biotechnology 141

18 3203–3209 (1990).32. Moore, M., Choo, Y. & Klug, A. Design of polyzinc finger peptides with structured

linkers Proc. Natl. Acad. Sci. USA 98, 1432–1436 (2001).33. Kim, C.A. & Berg, J.M. A 2.2 Å resolution crystal structure of a designed zinc finger

protein bound to DNA. Nat. Struct. Biol. 3, 940–945 (1996).34. Isalan, M., Choo, Y. & Klug, A. Synergy between adjacent zinc fingers in sequence-

specific DNA recognition. Proc. Natl. Acad. Sci. USA 94, 5617–5621 (1997).35. Beerli, R.R., Segal, D.J., Dreier, B. & Barbas, C.F., III. Toward controlling gene

expression at will: specific regulation of the erbB-2/HER-2 promoter by using poly-dactyl zinc finger proteins constructed from modular building blocks. Proc. Natl.Acad. Sci. USA 95, 14628–14633 (1998).

36. Isalan, M., Klug, A. & Choo, Y. Comprehensive DNA recognition through concertedinteractions from adjacent zinc fingers. Biochemistry 37, 12026–12033 (1998).

37. Wolfe, S.A., Greisman, H.A., Ramm, E.I. & Pabo, C.O. Analysis of zinc fingers opti-mized via phage display: evaluating the utility of a recognition code. J. Mol. Biol.285, 1917–1934 (1999).

38. Wolfe, S.A., Grant, R.A., Elrod-Erickson, M. & Pabo, C.O. Beyond the “recognitioncode”: structures of two Cys2His2 zinc finger/TATA box complexesStructure 9,717–723 (2001).

39. Isalan, M., Klug, A. & Choo, Y. A rapid, generally applicable method to engineer zincfingers illustrated by targeting the HIV-1 promoter Nat. Biotechnol. 19, 656–660(2001).

40. Sera, T. & Schultz, P.G. In vivo selection of basic region-leucine zipper proteins withaltered DNA-binding specificities. Proc. Natl. Acad. Sci. USA 93, 2920–2925 (1996).

41. Joung, J.K., Ramm, E.R. & Pabo, C.O. A bacterial two-hybrid selection system forstudying protein–DNA and protein–protein interactions. Proc. Natl. Acad. Sci. USA97, 7382–7387 (2000).

42. Cathomen, T., Stracker, T., Gilbert, L. & Weitzman, M. A genetic screen identifies acellular regulator of adeno-associated virus. Proc. Natl. Acad. Sci. USA 93,14991–14996 (2001).

43. Calvo, S. et al. Molecular dissection of DNA sequences and factors involved in slowmuscle-specific transcription Mol. Cell. Biol. 21, 8490–8503 (2001).

44. Margolin, J.F. et al. Krüppel-associated boxes are potent transcriptional repressiondomains. Proc. Natl. Acad. Sci. USA 91, 4509–4513 (1994).

45. Heinzel, T. et al. A complex containing N-CoR, mSin3 and histone deacetylasemediates transcriptional repression. Nature 387, 43–48 (1997).

46. Choo, Y., Sanchez-Garcia, I. & Klug, A. In vivo repression by a site-specific DNA-binding protein designed against an oncogenic sequence. Nature 372, 642–645(1994).

47. Kim, J.S. & Pabo, C.O. Transcriptional repression by zinc finger peptides. Exploringthe potential for applications in gene therapy. J. Biol. Chem. 272, 29795–29800(1997).

48. Kang, J.S. & Kim, J.-S. Zinc finger proteins as designer transcription factors. J. Biol.Chem. 275, 8742–8748 (2000).

49. Kim, J.S., Kim, J., Cepek, K.L., Sharp, P.A. & Pabo, C.O. Design of TATA box-bindingprotein/zinc finger fusions for targeted regulation of gene expression. Proc. Natl.Acad. Sci. USA 94, 3616–3620 (1997).

50. Ptashne, M. & Gann, A. Transcriptional activation by recruitment. Nature 386,569–577 (1997).

51. Liu, P.-Q. et al. Regulation of an endogenous locus using a panel of designed zincfinger proteins targeted to accessible chromatin regions. J. Biol. Chem. 276,11323–11334 (2001).

52. Zhang, L. et al. Synthetic zinc finger transcription factor action at an endogenouschromosomal site. J. Biol. Chem. 275, 33850–33860 (2000).

53. Gossen, M. & Bujard, H. Tight control of gene expression in mammalian cells bytetracycline-responsive promoters. Proc. Natl. Acad. Sci. USA 89, 5547–5551(1992).

54. Wang, Y., O’Malley, B.W., Jr., Tsai, S. & O’Malley, B.W. A regulatory system for usein gene transfer. Proc. Natl. Acad. Sci. USA 91, 8180–8184 (1994).

55. No, D., Yao, T.-P. & Evans, R.M. Ecdysone-inducible gene expression in mammaliancells and transgenic mice. Proc. Natl. Acad. Sci. USA 93, 3346–3351 (1996).

56. Allgood, V.E. & Eastman, E.M. Chimeric receptors as gene switches. Curr. Opin.Biotechnol. 8, 474–479 (1997).

57. Xu, L. et al. A versatile framework for the design of ligand-dependent, transgene-specific transcription factors. Mol. Ther. 3, 262–273 (2001).

58. Stege, J., Guan, X., Briggs, S. & Barbas, C.F., III. p. 12 in Keystone SymposiaSystems Approaches to Plant Biology (Big Sky, MT, 2001).

59. Eckardt, N.A. Meeting report: The new biology: genomics fosters a systemsapproach and collaborations between academic, government, and industry scien-tists. Plant Cell 13, 725–732 (2001).

60. Bibikova, M. et al. Stimulation of homologous recombination through targetedcleavage by a chimeric nuclease. Mol. Cell. Biol. 21, 289–297 (2000).

©20

02 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://b

iote

ch.n

atu

re.c

om