Mol. Cells, Vol. 21, No. 3, pp. 376-380

One-Step Selection of Artificial Transcription Factors Using an In Vivo Screening System

Kwang-Hee Bae1,†

and Jin-Soo Kim1,2,

* 1 ToolGen, Inc., Daejeon 305-390, Korea; 2 Department of Chemistry, Seoul National University, Seoul 151-747, Korea.

(Received January 16, 2006; Accepted April 4, 2006)

Gene expression is regulated in large part at the level of

transcription under the control of sequence-specific

transcriptional regulatory proteins. Therefore, the abil-

ity to affect gene expression at will using sequence-

specific artificial transcription factors would provide

researchers with a powerful tool for biotechnology re-

search and drug discovery. Previously, we isolated 56

novel sequence-specific DNA-binding domains from the

human genome by in vivo selection. We hypothesized

that these domains might be more useful for regulating

gene expression in higher eukaryotic cells than those

selected in vitro using phage display. However, an un-

predictable factor, termed the “context effect”, is asso-

ciated with the construction of novel zinc finger tran-

scription factors--- DNA-binding proteins that bind

specifically to 9-base pair target sequences. In this study,

we directly selected active artificial zinc finger proteins

from a zinc finger protein library. Direct in vivo selec-

tion of constituents of a zinc finger protein library may

be an efficient method for isolating multi-finger DNA

binding proteins while avoiding the context effect.

Keywords: Artificial Transcription Factor; Context Ef-

fect; In vivo Selection; Zinc Finger.

Introduction

Tissue-specific and developmentally regulated gene ex-

pression is controlled by transcription factors that interact

† Present address: Systemic Proteomics Research Center, Korea

Research Institute of Bioscience and Biotechnology (KRIBB),

Daejeon 305-333, Korea.

* To whom correspondence should be addressed.

Tel: 82-42-863-8166; Fax: 82-42-863-3840

E-mail: jskim01@snu.ac.kr

with specific DNA regulatory elements to activate or in-

hibit transcription initiation. Typical transcriptional regu-

latory proteins have modular structures that consist of a

DNA-binding domain (DBD) and an effector (transcrip-

tional activation or repression) domain. The effector do-

mains of many transcription factors remain active when

transferred to heterologous DBDs (Brent and Ptashne,

1985; Lee et al., 2003). Therefore, novel transcription

factors can be constructed by first preparing DBDs with

the desired DNA-binding specificities, then linking them

to appropriate effector domains.

A wide range of organisms possess transcription factors

that contain zinc-finger motifs, which are ~30 amino acid

peptides that bind to DNA in a sequence-specific manner.

Although there are several types of zinc-finger, members

of the Cys2-His2 class are especially useful for generating

artificial transcription factors owing to their diversity and

modular structure. Several groups have used the phage

display method to identify zinc finger domains (ZFDs)

that bind specifically to a diverse set of DNA triplet sub-

sites (Beerli and Barbas, 2002; Choo and Klug, 1994;

Dreier et al., 2001; Jamieson et al., 1994; Joung et al.,

2000; Rebar and Pabo, 1994; Segal et al., 1999; Tan et al.,

2003). Although phage display is an effective method, it

is labor intensive and time-consuming. Furthermore, be-

cause it uses an in vitro selection method, the zinc finger

proteins (ZFPs) identified may not function efficiently in

vivo. In addition, with phage display, ZFPs are selected

for binding to “naked” DNA rather than to DNA packaged

as chromatin, the form assumed by DNA in eukaryotic

cells. To overcome these problems, we developed an in

vivo selection method that makes use of a yeast one-

hybrid system (Bae et al., 2003). Using this technique, we

isolated 56 novel ZFDs from the human genome and used

Abbreviations: EMSA, electrophoretic mobility shift assay;

VEGF, vascular endothelial growth factor; ZFD, zinc finger

domain; ZFP, zinc finger protein.

Molecules

and

Cells©KSMCB 2006

Kwang-Hee Bae & Jin-Soo Kim 377

them as modular building blocks in the construction of

novel, sequence-specific, multi-finger ZFPs. In addition,

by fusing these complex ZFPs with appropriate effector

domains, we constructed artificial transcription factors

that specifically activate or repress expression of the vas-

cular endothelial growth factor gene (VEGF).

Using phage display, binding site code-based design,

and in vivo selection, several groups have identified indi-

vidual ZFDs that appear to bind specifically to a diverse

set of triplet DNA subsites. However, a number of other

studies have suggested that ZFDs in multi-finger proteins

could interact with bases outside their predicted triplet

subsites (Cheng et al., 1997; Jamieson et al, 2003; Pail-

lard et al., 2004; Wolfe et al., 2001). Typical “target site

overlap” occurs when certain amino acid residues at posi-

tion 2 in the recognition helix of the ZFD interact with the

first base of the neighboring DNA subsite (Pabo et al.,

2001). Recently, Pabo and colleagues reported that there

are other target site overlaps in ZFPs assembled from

ZFDs selected using phage display (Wolfe et al., 2001).

These target site overlaps are typical examples of the

“context effect”. This effect may be the main stumbling

block to generating highly functional, multi-finger ZFPs,

because many of the assembled proteins have low target

site binding activity in vivo. This problem can arise when

ZFDs selected using phage display or an in vivo selection

method are used to produce the multi-finger proteins. Al-

though human ZFDs selected in vivo may be more useful

than ZFDs selected by phage display for constructing ac-

tive ZFPs (Bae et al., 2003), nearly half the assembled

ZFPs did not function efficiently. Recently, Barbas and

collaborators developed an in vivo selection method for

active three-finger proteins using a mammalian cell line

(Blancafort et al., 2003; Magnenat et al., 2004). However,

mammalian cells are more difficult to transfect or trans-

duce with plasmids or retrovirus than yeast cells. In this

study, we used yeast as host to isolate novel DNA-binding

proteins that bind specifically to 9-base pair target DNA

sequence elements of choice.

Materials and Methods

Yeast strains, chemicals, and enzymes The host yeast strains

used for in vivo screening were S. cerevisiae yW1 (MATα Δgal4

Δgal180 ura3 his3-Δ200 trp1-Δ63 leu2 ade2-101ochre CYH2) (a

derivative of yWAM2 (Wang and Reed, 1993)) and yW1a

(MATa Δgal4 Δgal180 ura3 his3-Δ200 trp1-Δ63 leu2 ade2-

101ochre CYH2). yW1a was constructed as described previously

(Herskowitz and Jensen, 1991). yPH499 (MATa ade2-101 ura3-

52 lys2-801 trp1-Δ63 his3-Δ200 leu2-Δ1 GAL+) was also tested

in order to identify a strain that functioned efficiently in our

screen. Oligonucleotides used to construct reporter plasmids

were synthesized commercially (Genotech, Korea) and gel-

purified prior to use. All chemicals and enzymes were from

Promega or Sigma.

Construction of a three-finger library using ZFDs selected in

vivo The yeast expression plasmid, pPCFM, which was derived

from pPC86 (Chevray and Nathans, 1992), was used as a paren-

tal vector for expressing three-finger ZFPs. DNA segments that

encode individual ZFDs were subcloned into pPCFM. The re-

sulting plasmids were used as starting material for ZFP con-

struction and, subsequently, for production of a three-finger ZFP

library, as described previously (Bae et al., 2003).

In vivo selection of active three-finger ZFPs from a ZFP li-

brary For in vivo selection of active three-finger ZFPs from a

ZFP library, we first constructed a reporter yeast strain that car-

ried a reporter plasmid containing three tandem copies of a tar-

get DNA binding site for a specific ZFP upstream of the reporter

gene. Next, mating was carried out with a strain of the opposite

mating type that carried the ZFP library. In order to isolate the

most active ZFPs, ~2.5 × 105 mated cells were screened by

growth on X-gal-containing plates. Clones that were deep blue

were picked for further analysis.

Electrophoretic mobility shift assay (EMSA) DNA segments

that encode ZFPs were isolated by digestion with SalI and NotI,

and inserted into pGEX-4T2 (GE Health Care, Sweden). The

ZFPs were expressed in E. coli as fusion proteins linked to glu-

tathione-S-transferase (GST). The fusion proteins were purified

using glutathione affinity chromatography and digested with

thrombin. The probe DNAs used in EMSAs were synthesized

and annealed, and labeled with 32P using T4 polynucleotide

kinase (Kim et al., 2005). EMSAs were performed as previously

described (Bae et al., 2003; Kim and Pabo, 1997).

In vivo repression assay In vivo repression assays were carried

out as described previously (Kim and Pabo, 1997). ZFPs effi-

ciently repress VP16-activated transcription of a reporter gene

when they bind to DNA sequences near the transcription start

site. We used this assay to determine whether the selected ZFPs

were functional in human cells. The plasmids encoding selected

ZFPs were cotransfected into HEK293 cells with appropriate

reporter plasmids. The latter carried the firefly luciferase gene

under the control of synthetic promoters in which appropriate

binding sites were incorporated at positions near the initiator

element (Bae et al., 2003). Repression levels (fold repression)

were obtained by dividing the luciferase activities of cells trans-

fected with empty plasmid by those of cells transfected with

plasmids encoding ZFPs. Active ZFPs were defined as those

that showed greater than 2.0-fold repression (Bae et al., 2003).

Results and Discussion

One of the unpredictable factors associated with the con-

struction of novel ZFPs is the so called “neighboring

finger effect” or “context effect”. Although many ZFDs

378 One-Step Selection of Artificial Transcription Factors

Fig. 1. Control experiment for direct in vivo selection of active

ZFPs. Zif26 refers to a two finger protein composed of the two

N-terminal ZFDs of Zif268. The sequence of the target site was

5′-GCG TGG GCG-3′.

can be used as modular building blocks to construct novel

sequence-specific DNA-binding proteins, it is difficult to

predict which ZFDs will function well in combination,

and which ones won’t. For example, a certain ZFD,

termed ‘A’, might function well at the N terminus of ZFD

‘B’, but not at the N terminus of ZFD ‘C’. We previously

reported the results of shuffling ZFDs to construct new

DNA-binding proteins (Bae et al., 2003). To our surprise,

none of the 24 randomly-chosen ZFPs composed of mu-

tated or engineered ZFDs were strong transcriptional rep-

ressors in co-transfection assays. In contrast, 28% of the

ZFPs composed of 3 ZFDs derived from the human ge-

nome had high repression activity. Apparently the ZFDs

with sequences derived from the human genome make

better building blocks for multi-finger ZFPs than do the

engineered ZFDs. Still, it is difficult to predict whether a

given combination of human ZFDs will generate a func-

tional in vivo transcription factor.

To overcome the context effect, Greisman and Pabo

(1997) developed the sequential selection strategy and

Isalan et al. (2001) reported their bipartite selection

method. Although these methods can yield high-quality

ZFPs, they each require several rounds of selection and

amplification and are based on in vitro selection using

phage display. We attempted to use the yeast one-hybrid

screening method to directly isolate, from a ZFP library,

highly active ZFPs that bind in vivo to 9-bp DNA se-

quence elements in vivo.

To construct the ZFP library, we chose 36 ZFDs with

distinct DNA-binding specificities and randomly shuffled

the nucleic acid segments encoding each of these ZFDs to

assemble three-finger chimeric ZFPs (Bae et al., 2003;

Park et al., 2003). The resulting plasmid library of hybrid

transcription factors was introduced into the yW1 ‘α’

strain, and the transformed α cells were than mated with

the yW1 ‘a’ yeast strain, which harbors a reporter plasmid

with three copies of a variable 9-bp target DNA sequence

element upstream of lacZ. Figure 1 shows the result of a

control experiment in which a typical ZFP, Zif268, bound

the target sequence and activated transcription of lacZ. In

our procedure active ZFPs binding the 9 bp target DNA

were identified by isolating the expression plasmids from

Fig. 2. Target sequences, blue density test, and IDs of active

three-finger proteins isolated from the ZFP library.

Fig. 3. Determination of dissociation constant (Kd) by EMSA.

The amounts of substrate added were 0, 0.04, 0.08, 0.16, 0.33,

0.63, 1.3, 2.5, 5, and 10 nM, beginning with the left-most lane.

The sequence of probes is shown below the gel image. Underlin-

ing represents mutation. Under our experimental conditions, the

Kd of Zif268 was 0.056 nM.

blue colonies and sequencing them. Several pale blue

colonies were also detected but only those that were a

deep blue like the Zif268 control were analyzed. Figure 2

gives the ZFPs that were selected with this yeast one-

hybrid assay.

The structures of the selected ZFPs agree with those

predicted by the DNA binding specificities of the individ-

ual fingers (Bae et al., 2003). For example, the selected

ZFP termed DSNR-RSHR-HSSR was isolated with the

target DNA sequence 5′-GCT GGG GAC-3′. According to

previous reports (Bae et al., 2003; Rebar and Pabo, 1994),

the DSNR finger at the N terminus of the selected ZFP is

expected to bind to the GAC site, and the RSHR finger to

the GGG site. The HSSR finger is expected to bind to a

GTT site, but not to a GCT site. However, the serine resi-

due at position 3 of the HSSR finger might be able to

form a hydrogen bond with thymine. Thus the DSNR-

RSHR-HSSR ZFP is one of the ZFPs expected to be se-

lected from the library with the 5′-GCT GGG GAC-3′

Kwang-Hee Bae & Jin-Soo Kim 379

A B

C

Fig. 4. Transcriptional repression by ZFPs selected with the yeast

one-hybrid assay. Human 293 cells were transiently transfected

with an activator plasmid encoding GAL4-VP16, an internal con-

trol plasmid encoding Renilla luciferase, a reporter plasmid en-

coding firefly luciferase, and an effector plasmid encoding a se-

lected hZFP. Renilla and firefly luciferase activities were meas-

ured 48 h after transfection. The firefly luciferase activities were

normalized with respect to the Renilla luciferase activities to cor-

rect for transfection efficiency. Repression levels (fold repression)

were obtained by dividing the normalized luciferase activities of

cells transfected with empty plasmid by those of cells transfected

with the effector plasmid encoding the selected hZFPs. The data

are averages of three independent experiments. The reporter se-

quences used are shown at the top of the figure.

DNA sequence. Considering the high complexity of the

ZFP library (36 ZFDs × 36 ZFDs × 36 ZFDs = 46,656

ZFPs), it is remarkable that the expected ZFP was isolated

in a single screening step.

To further characterize the activity of the selected ZFPs,

we carried out EMSAs. The recombinant ZFPs were puri-

fied from Escherichia coli and showed tight binding to

their cognate target DNA sequences, but not to a mutated

DNA sequence (Fig. 3). Indeed, the two ZFPs tested

showed a roughly 10-fold preference for their target DNA

sequences over the mutated DNA sequence. In addition,

the selected ZFPs caused strong repression of luciferase

expression in the in vivo repression assay (Fig. 4). These

results demonstrate that the selected ZFPs are functional

both in vitro and in vivo.

In the course of our experiments, we noted one limita-

tion of the one-step ZFP screening system: when we used

5′-GAG GCG GAA-3′ as the target DNA sequence, all the

mated yeast colonies turned blue and we were unable to

screen for ZFPs binding to this particular DNA site. The

blue color of the yeast cells harboring this reporter may

result from the presence of endogenous transcription fac-

tors that bind to the target sequence and activate lacZ.

Thus far, we have tested dozens of different target DNA

sequences in the yeast mating assay and observed this

pervasive blue phenotype only in this one case. Thus this

problem is expected to be very limited.

We hypothesized that the problem of selecting ZFPs

that are inactive in vivo could be solved by screening the

assembled ZFPs. Using our direct 9-bp selection system,

we were able to screen the three-finger ZFP library for

ZFPs that displayed reasonably tight binding to their

DNA target sequences in vivo. This method is also useful

for detecting more than one ZFD that can bind to a single

3-bp subsite. For example, we isolated six fingers, desig-

nated ‘RDER1 to 6’ that recognize the GCG subsite (Bae

et al., 2003). We do not yet know which of these fingers

has the tightest DNA binding activity or the highest

transcriptional activity; however, further routine experiments

should allow us to determine which context produces the

most active ZFP.

In conclusion, our one-hybrid selection method prom-

ises to be useful for obtaining novel ZFDs and for screen-

ing for active ZFPs while avoiding the “context effect”.

Acknowledgments We thank Dr. Randall Reed for providing

yeast strain yWAM2. We also thank Mrs. M.-S Hwang and E. H.

Ryu for excellent technical assistance and helpful advice, and Dr.

K. LaMarco for carefully reading our manuscript.

References

Bae, K.-H., Kwon, Y. D., Shin, H.-C., Hwang, M.-S., Ryu, E.-H.,

et al. (2003) Human zinc fingers as building blocks in the

construction of artificial transcription factors. Nat. Biotech.

21, 275−280.

Beerli, R. R. and Barbas, C. F. (2002) Engineering polydactyl

zinc-finger transcription factors. Nat. Biotechnol. 20, 135−

141.

Blancafort, P., Magnenat, L., and Barbas, C. F. (2003) Scanning

the human genome with combinatorial transcription factor li-

braries. Nat. Biotechnol. 21, 269−274.

Brent, R. and Ptashne, M. A. (1985) Eukaryotic transcriptional

activator bearing the DNA specificity of a prokaryotic rep-

ressor. Cell 43, 729−736.

Cheng, X., Boyer, J. L., and Juliano, R. L. (1997) Selection of

peptides that functionally replace a zinc finger in the Sp1

transcription factor by using a yeast combinatorial library.

Proc. Natl. Acad. Sci. USA 94, 14120−14125.

Chevray, P. M. and Nathans, D. (1992) Protein interaction clon-

ing in yeast: identification of mammalian proteins that inter-

act with the leucine zipper of Jun. Proc. Natl. Acad. Sci. USA

89, 5789−5793.

Choo, Y. and Klug, A. (1994) Toward a code for the interactions

of zinc fingers with DNA: selection of randomized fingers

displayed on phage. Proc. Natl. Acad. Sci. USA 91, 11163−

11167.

Dreier, B., Beerli, R. R., Segal, D. J., Flippin, J. D., and Barbas, C.

F. (2001) Development of zinc finger domains for recognition

of the 5′-ANN-3′ family of DNA sequences and their use in the

380 One-Step Selection of Artificial Transcription Factors

construction of artificial transcription factors. J. Biol. Chem.

276, 29466−29478.

Greisman, H. A. and Pabo, C. O. (1997) A general strategy for

selecting high-affinity zinc finger proteins for diverse DNA

target sites. Science 275, 657−661.

Herskowitz, I. and Jensen, R. E. (1991) Putting the HO gene to

work: practical uses for mating-type switching. Methods En-

zymol. 194, 132−146.

Isalan, M., Klug, A., and Choo, Y. (2001) A rapid, generally

applicable method to engineer zinc fingers illustrated by tar-

geting the HIV-1 promoter. Nat. Biotech. 19, 656−660.

Jamieson, A. C., Kim, S.-H., and Wells, J. A. (1994) In vitro

selection of zinc fingers with altered DNA-binding specific-

ity. Biochemistry 33, 5689−5695.

Jamieson, A. C., Miller, J. C., and Pabo, C. O. (2003) Drug dis-

covery with engineered zinc-finger proteins. Nat. Rev. Drug

Discov. 2, 361−368.

Joung, J. K., Ramm, E. I., and Pabo, C. O. (2000) A bacterial

two-hybrid selection system for studying protein-DNA and

protein-protein interactions. Proc. Natl. Acad. Sci. USA 97,

7382−7387.

Kim, J.-S. and Pabo, C. O. (1997) Transcriptional repression by

zinc finger peptides. Exploring the potential for applications

in gene therapy. J. Biol. Chem. 272, 29795−29800.

Kim, S.-J., Kang, S.-Y., Shin, H.-H., and Choi, H.-S. (2005)

Sulforaphane inhibits osteoclastogenesis by inhibiting nu-

clear factor-kB. Mol. Cells 20, 364−370.

Lee, D.-k., Seol, W., and Kim, J. S. (2003) Custom DNA-

binding proteins and artificial transcription factors. Curr. Top.

Med. Chem. 3, 645−657.

Magnenat, L., Blancafort, P., and Barbas, C. F. (2004) In vivo

selection of combinatorial libraries and designed affinity

maturation of polydactyl zinc finger transcription factors for

ICAM-1 provides new insights into gene regulation. J. Mol.

Biol. 341, 635−649.

Pabo, C. O., Peisach, E., and Grant, R. A. (2001) Design and

selection of novel Cys2His2 zinc finger proteins. Annu. Rev.

Biochem. 70, 313−340.

Paillard, G., Deremble, C., and Lavery, R. (2004) Looking into

DNA recognition: zinc finger binding specificity. Nucleic Ac-

ids Res. 32, 6673−6682.

Park, K. S., Lee, D.-K., Lee, H., Lee, Y., Jang, Y. S., et al.

(2003) Phenotype alteration of eukaryotic cells using ran-

domized libraries of artificial transcription factors. Nat. Bio-

tech. 21, 1208−1214.

Rebar, E. J. and Pabo, C. O. (1994) Zinc finger phage: affinity

selection of fingers with new DNA-binding specificities. Sci-

ence 263, 671−673.

Segal, D. J., Dreier, B., Beerli, R. R., and Barbas, C. F. (1999)

Toward controlling gene expression at will: selection and de-

sign of zinc finger domains recognizing each of the 5′-GNN-

3′ DNA target sequences. Proc. Natl. Acad. Sci. USA 96,

2758−2763.

Tan, S., Guschin, D., Davalos, A., Lee, Y. L., Snowden, A. W.,

et al. (2003) Zinc-finger protein-targeted gene regulation:

genomewide single-gene specificity. Proc. Natl. Acad. Sci.

USA 100, 11997−12002.

Wang, M. M. and Reed, R. R. (1993) Molecular cloning of the

olfactory neuronal transcription factor Olf-1 by genetic selec-

tion in yeast. Nature 364, 121−126.

Wolfe, S. A., Grant, R. A., Elrod-Erickson, M., and Pabo, C. O.

(2001) Beyond the “recognition code”: structures of two

Cys2His2 zinc finger/TATA box complexes. Structure 9,

717−723.