doi:10.1016/j.jmb.2004.04.057
J. Mol. Biol. (2004) 340, 599–613
Promoter-targeted Phage Display Selections with
Preassembled Synthetic Zinc Finger Libraries for
Endogenous Gene Regulation
Caren V. Lund, Pilar Blancafort, Mikhail Popkov and
Carlos F. Barbas III*
The Skaggs Institute for
Chemical Biology and
Department of Molecular
Biology, The Scripps Research
Institute, 10550 North Torrey
Pines Road, La Jolla, CA 92037
USA
Regulation of endogenous gene expression has been achieved using
synthetic zinc finger proteins fused to activation or repression domains,
zinc finger transcription factors (TFZFs). Two key aspects of selective gene
regulation using TFZFs are the accessibility of a zinc finger protein to its
target DNA sequence and the interaction of the fused activation or repression domain with endogenous proteins. Previous work has shown that
predicting a biologically active binding site at which a TFZF can control
gene expression is not always straightforward. Here, we used a library of
preassembled three-finger zinc finger proteins (ZFPs) displayed on filamentous phage, and selected for ZFPs that bound along a 1.4 kb promoter
fragment of the human ErbB-2 gene. Following affinity selection by phage
display, 13 ZFPs were isolated and sequenced. Transcription factors were
prepared by fusion of the zinc finger proteins with a VP64 activation
domain or a KRAB repression domain and the transcriptional control
imposed by these TFZFs was evaluated using luciferase reporter assays.
Endogenous gene regulation activity was studied following retroviral
delivery into A431 cells. Additional ZFP characterization included
DNase I footprinting to evaluate the integrity of each predicted protein:DNA interaction. The most promising TFZFs able to both up-regulate
and down-regulate ErbB-2 expression were extended to six-finger proteins. The increased affinity and refined specificity demonstrated by the
six-finger proteins provided reliable transcriptional control. As a result of
studies with the six-finger proteins, the specific region of the promoter
most accessible to transcriptional control by VP64-ZFP and KRAB-ZFP
fusion proteins was elucidated and confirmed by DNase I footprinting,
flow cytometric analysis and immunofluorescence. The ZFP phage display
library strategy disclosed here, coupled with the growing availability of
genome sequencing information, provides a route to identifying gene-regulating TFZFs without the prerequisite of well-defined promoter elements.
q 2004 Elsevier Ltd. All rights reserved.
*Corresponding author
Keywords: phage display selection; zinc finger; gene regulation; ErbB-2;
immunofluorescence
Introduction
Transcription factors (TFs), mediator proteins,
and chromatin remodeling proteins associate with
the promoter regions of genes.1,2 The presence or
absence of these proteins determines the timing
Supplementary data associated with this article can be found at doi: 10.1016/j.jmb.2004.04.057
Abbreviations used: TF, transcription factor; ZFP, zinc finger protein; TFZF, zinc finger transcription factor;
UTR, untranslated region; ELISA, enzyme-linked immunosorbent assay; FACS, fluorescence-activated cell sorting;
KRAB, Kruppel-associated box; HA, hema-glutinin; DAPI, 40 ,6-diamidino-2-phenylindole dihydro-chloride; P/PMR,
polypurine/polypyrimidine mirror repeat; GFP, green fluorescent protein; VP64, tetrameric repeat of herpes simplex
VP16’s minimal activation domain.
E-mail address of the corresponding author: carlos@scripps.edu
0022-2836/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.
600
and duration of gene transcription. One of the
most common classes of TFs contains the DNAbinding zinc finger transcription factors.3,4 In
order to impart artificial control on gene transcription, synthetic zinc finger proteins (ZFPs) have
been engineered to bind specific DNA sequences.
A subset of these engineered proteins (TFZFs) has
been shown to control endogenous gene
expression selectively when fused to activation or
repression domains.5 – 17 The prediction of successful target sites for TFZFs in endogenous promoters
has been based on ZFP affinity, position within the
50 untranslated region (UTR), and their location
within DNase I-hypersensitive regions.6 – 8,10 Application of these criteria, however, has not resulted
consistently in specific gene regulation. With the
roles of endogenous TFs, mediator proteins, and
chromatin remodeling proteins largely uncharacterized for any given promoter, a more empirical
method of sampling binding sites across a promoter is required to achieve robust, imposed gene
regulation.
Synthetic ZFPs have been derived from the
framework of naturally occurring zinc finger TFs,
including Sp1 and the murine Zif268, which have
three zinc finger domains.5 The C2H2 zinc finger
domain consists of an a helix packed against two
antiparallel b sheets and stabilized via coordination
of a zinc ion. The side-chains of amino acid residues at the terminus of the b-sheet and continuing
into the a helix, constitute the DNA reading-head
of the domain that is responsible for sequencespecific interaction with the DNA. A single zinc
finger domain typically recognizes 3 bp of DNA,
primarily through major groove interactions.
Synthetic ZFPs typically consist of a repeat of
three to six zinc finger domains, allowing the
resulting polydactyl proteins to recognize DNA
sequences of 9 – 18 bp. Transcription factors have
been constructed by fusion of an activation or
repression domain to the ZFP, creating TFZFs.
Naturally occurring activators and repressors, or
variants thereof, are commonly used with ZFPs,
particularly the VP16 activation domain and the
Kruppel associated box (KRAB).5,18 – 20
The target of imposed gene regulation is not
simply DNA, but rather the three-dimensional
landscape of chromatin structure constructed from
DNA:protein interactions and protein:protein
interactions along a promoter. The linear sequence
of DNA provides little indication of a promoter’s
dynamic topology. Eukaryotic gene promoter
sequences often stretch over many thousands of
bases, and include both positive and negative
regulatory elements. Genome-wide characterization of promoters employing deletion constructs,
linker scanning, computational TF binding site
analyses, and TF isolation, is ongoing.21 – 25 In
addition, experiments have begun to address
the chromatin state of promoters, the location
of nucleosomes, and the nature of histone
modifications.26,27 In spite of this information, the
understanding of the precise mechanism of tran-
Phage Display Selection of Zinc Finger Transcription Factors
scriptional control is far from complete for any
promoter.
Here, we have studied the application of phage
display to provide ZFP-based TFs that bind
throughout a given promoter sequence with the
aim of applying these proteins to endogenous
gene regulation. This strategy resulted in the
identification of a site at which a three-finger or a
six-finger TFZF could provide regulation of ErbB-2
expression in human epidermoid carcinoma A431
cells. The site identified would not have been predicted from the linear sequence of the promoter,
known TF binding sites, or chromatin characterization, suggesting that this approach can be used
to synthesize gene regulators without prior characterization of a target gene’s promoter.
Results
Phage display optimization and selection
against an ErbB-2 promoter fragment
Previous studies of zinc finger proteins using
phage display have focused on the selection of
zinc finger variants of altered specificity. These
studies have utilized short pieces of DNA as the
target for selection, typically oligonucleotides that
present the 9 bp binding site for a three-finger
protein.9,28 – 30 Our goal was to select zinc fingers
that bound optimally within a 1.4 kb promoter
fragment. The phage display library used in this
study contained 9177 three-finger ZFPs, each
capable of recognizing 9 bp of DNA. The construction of the library was based on a combinatorial
assembly of a subset of defined zinc-finger DNA
sequences recognizing GNN, ANN and TNN
triplets.9,14,30,31 There were only a few TNN
domains available at the time of the library’s
construction, so the library is biased towards
recognition of RNN triplets, where R ¼ G or
A. Theoretically, the double-stranded human
genome contains 750 million (RNN)3 sites.14 In
addition, many promoters contain GNN-rich
regions of DNA, which makes the library particularly applicable to the isolation of promoter binding proteins.32
As the length of the DNA target was far in excess
of any used previously, we first evaluated the
ability of the previously established phage display
protocol to recover ZFPs. A six-finger protein,
E2C, with a dissociation constant of 0.75 nM for
an 18 bp site in the ErbB-2 promoter (see Figure
2), was used as a positive control.5 Using phage
expressing the six-finger protein E2C on its surface,
test selections were done to compare the recovery
of E2C phage by a biotinylated 1.4 kb promoter or
a biotinylated 18 bp hairpin oligonucleotide. Both
selections had the E2C phage diluted into control
phage bearing a different selective marker at a
ratio of 1:10,000. The 1.4 kb promoter DNA yielded
90% fewer phage compared to the number of
phage recovered using the 18 bp oligonucleotide.
Phage Display Selection of Zinc Finger Transcription Factors
Panning conditions were subsequently optimized,
and phage selection using E2C phage was
improved to 85% relative to the phage recovered
with the 18 bp oligonucleotide.
The library of preassembled three-finger proteins
was then selected over four rounds of panning
using the 1.4 kb ErbB-2 promoter fragment. The
first and second rounds of panning contained
125 nM promoter DNA per binding reaction. The
third and fourth rounds contained 62 nM and
31 nM promoter DNA per binding reaction
respectively, to increase the stringency of selection
and to favor recovery of higher-affinity ZFPs. The
effectiveness of each round of panning to select
ZFPs specific for the target was initially evaluated
using a phage enzyme-linked immunosorbent
assay (ELISA) (Figure 1B). As a functional evaluation of the ZFPs recovered, DNA from each round
601
of panning was subcloned into a retroviral vector
and expressed as fusions with the activation
domain, VP64, which is a tetrameric repeat of the
minimal activation domain of herpes simplex
virus VP16 activator protein.5,33 The ability of each
pool of proteins to activate ErbB-2 expression in
A431 cells was evaluated using fluorescenceactivated cell sorting (FACS) analysis (Figure 1C).
The unselected library did show some activation
of ErbB-2 expression. However, proteins selected
in subsequent rounds of phage display, particularly round 4, were more efficient activators of
endogenous gene expression.
Characterization of selected zinc finger proteins
ZFP clones from round 4 of panning were
selected for characterization. Five clones were
Figure 1. Phage display strategy with in vitro and in vivo characterization of the rounds. A, Using a 50 -biotinylated
primer, the promoter of interest was PCR-amplified and conjugated to streptavidin-coated magnetic beads. Incubation
of the phage displaying a preassembled synthetic zinc finger protein library with target-coated beads provided
recovery of zinc finger proteins with binding sites in the promoter. Multiple rounds of selection were conducted with
decreasing amounts of promoter target to select for a narrowed pool of proteins with high binding affinities. B, Phage
ELISA analysis was used to determine the relative binding affinities of the unselected library (US) phage compared
to phage selected from four subsequent rounds of panning. Phage were amplified from each round of panning and
incubated with biotinylated ErbB-2 promoter fragment immobilized on ELISA plates. Bound phage were detected
with an anti-M13/horseradish peroxidase antibody conjugate. Fluorescence of activated ABTS (2,20 -azino-bis(3-ethylbenzthiazoline-6-sulfonic acid) was detected at 405 nm and quantified by a microplate reader. The maximal signal of
round 2 was normalized to 1. C, Each round of panning was subcloned into a retroviral vector for expression as fusion
proteins with VP64 in A431 cells. The ability of the zinc finger fusion proteins from each round to activate endogenous
ErbB-2 expression in A431 cells was monitored by FACS analysis. The tallest, dotted peak is A431 cells infected with
stuffer DNA fused to VP64. The unselected library peak is the second tallest peak and is in bold. The third tallest
peak is round 1 represented with a thin line. The peaks for rounds 2, 3, and 4 overlap, and are shown with a single
bold line that is shifted furthest to the right.
602
Phage Display Selection of Zinc Finger Transcription Factors
Figure 2. A representation of the ErbB-2 promoter fragment, 2 1440 to þ1. The ErbB-2 promoter fragment is separated into three 480 bp regions. Two sections of region 3 (sub-regions 3A and 3B) are magnified to show the binding
sites of the proximal ZFPs. Binding sites of the 3Fn-ZFPs listed in Table 1 are indicated along with the associated promoter sequence (50 -30 ). Open arrowheads designate binding sites of transcription factors, AP-2, p300, Sp1, adenovirus
5 E1A and Ets binding site (EBS), that have been characterized in ErbB-2 gene regulation.35,45,46,55,56 The CCAAT and
TATAA sites are marked with hatched rectangles to represent the boundaries of the DNase I hypersensitive region of
the ErbB-2 promoter.34
initially sequenced and labeled 1A, 2A, 3A, 8A,
and 9A. Subsequently, 25 more clones were
screened for binding to the biotinylated 1.4 kb
ErbB-2 promoter using an ELISA assay (data not
shown). The clones with the best binding signals
were designated 2, 4, 5, 6, 8, 11, 13, and 16, and
were sequenced. On the basis of previous studies,
the DNA sequence that each ZFP is predicted to
bind is based on the amino acid sequence of each
domain’s reading head.9,30 Table 1 lists the predicted DNA-binding sequence for each ZFP and
the corresponding sites within the ErbB-2 promoter. Six of the ZFP proteins sequenced encoded
pairs of identical sequence: 3A and 8, 4 and 11,
and 5 and 13. Most of the proteins were predicted
to bind with eight out of nine matches to sites in
the ErbB-2 promoter, while 8A, 9A, 6, and 8 have
perfect 9 bp matching sites in the promoter
fragment.
Footprinting analysis of the zinc finger
proteins along the promoter
DNase I footprinting analysis was used to confirm the binding of the selected ZFPs at their predicted sites within the promoter DNA. ZFP1A and
2A could not be purified in sufficient quantity for
these studies. The other zinc finger proteins were
footprinted on the fragment of DNA that contained
the predicted sites (only data for ZFP16 are shown
in Figure 3). Each protein was titrated over a 100-
fold concentration range to confirm binding to the
predicted sites. The proteins bound their predicted
sites within the promoter DNA with affinities that
ranged from 1.0 nM to 88 nM (Table 1). Seven out
of nine clones showed additional, lower-affinity
binding sites. None of the footprints observed for
the predicted sites extended outside the expected
9 bp region by more than 3 bp, indicating the
maltose-binding protein fused to each ZFP for protein purification did not interfere with the binding
site analyses. For each footprint, the predicted
binding site of each ZFP is marked with a lowercase letter. Proximal to site 2 285 is the polypurine/polypyridmidine repeat region (P/PMR)
of the ErbB-2 promoter. The P/PMR region is a
GNN-rich region of DNA that is located within a
DNase I-hypersensitive region of the ErbB-2 promoter between the CCAAT and TATAA boxes.34,35
Transfection and retroviral infection of zinc
finger transcription factors
In order to study the potential of each of
the selected ZFPs for gene regulation, the genes
encoding them were transferred into mammalian
expression vectors and expressed as fusions with
the VP64 activation domain. The ability of the
resulting transcription factors to activate a luciferase reporter gene driven by the 1.4 kb ErbB-2 promoter was studied in A431 cells (Table 2). TFZFs
1A and 11 showed the highest level of activation,
Phage Display Selection of Zinc Finger Transcription Factors
603
Figure 3. DNase I footprinting analysis of ZFP16 and three six-finger proteins – 642, – 369 and – 285. Storage phosphor autoradiograms of DNase I footprint titrations analyzed on 8 M urea/6% (w/v) polyacrylamide gels. For each
gel: lane 1 (– ) is the DNA fragment without digestion; lane 2 is the G þ A ladder (L); lane 3, DNase I digestion of the
labeled fragment without protein (D); lanes 4 – 10 contain 100 nM, 50 nM, 20 nM, 10 nM, 5 nM, 2 nM, 1 nM purified
ZFP incubated with the DNA fragment. The location of the predicted binding sites are indicated to the right of each
gel with lower-case letters. All reactions contained 15,000 cpm of 32P end-labeled PCR fragment. Kd values are calculated as described.54
with 52.6-fold and 47.4-fold increases in luciferase
expression over expression of luciferase in the
absence of a transiently expressed transcription
factor. TFZFs 2, 6, and 8 displayed less than twofold
activation, while the level of activation derived
from TFZFs 2A, 8A, 9A, 13, and 16 ranged from
3.2-fold to 15.2-fold. Initially, the various affinities
derived from footprinting analysis (Table 1) were
compared with activation of a luciferase reporter
construct (Table 2). The TFZFs with affinities stronger than 52 nM generated at least fivefold acti-
vation of luciferase expression. Considering TFZFs
11 and 13, whose affinities were determined to be
similar for the same region of DNA, their activation of luciferase expression was 47-fold versus
5.8-fold, respectively. Therefore, there was no strict
correlation between stronger affinity and greater
activation. This may be due, in part, to the multiplicity of binding sites available to a three-finger
protein in the 1.4 kb target region and elsewhere
in the cell. However, the inability of transient transfection reporter assays to correlate activation with
604
Phage Display Selection of Zinc Finger Transcription Factors
Table 1. Selected sequences and their position in the ErbB-2 promoter used
Zinc finger protein
Predicted DNA recognition
Times selected
Binding sites in promoter
3Fn 1A
3Fn-2A
3Fn-8A
TAG GAG GTG
GAG GCG GTG
GAG AAG GAG
1
1
1
3Fn-9A
3Fn-2
3Fn-6
3Fn-8
TAG GAG GGA
GGA GCG GAT
GGG GCA GAG
GAG GCG GGA
1
1
1
2
3Fn-11
GAG GAG AAG
2
3Fn-13
GCG GAG GAG
2
3Fn-16
TAG GAG GGG
1
237
237
341
243
237a
228
369
–
454
1014
878
416
246
237a
228
231
228
642
369
284
Affinity (nM)
ND
ND
ND
8.9 ^ 3.8
4.6 ^ 1.4
3.6 ^ 1.2
ND
ND
88.0 ^ 58
ND
ND
ND
3.9 ^ 1.6
1.0 ^ 0.3
1.2 ^ 0.3
5.1 ^ 1.8
4.3 ^ 2.0
2.9 ^ 0.6
51.6 ^ 29
5.9 ^ 1.8
The three finger (3Fn) ZFP proteins used in this work are listed with their predicted binding sites, the recognition sequences
predicted for each ZFP (50 -30 ), the number of times each binding motif was identified, the location of each predicted binding site in
the promoter target, and the corresponding affinity of the ZFPs at the predicted sites.
a
Sites that were observed in DNase I footprinting experiments and for which associated affinities are included.
affinity or, more importantly, predict endogenous
regulation, has been observed with other threefinger TFZFs.7 Therefore, the activity of the
transcription factors on the endogenous gene was
studied.
The ability of the TFZFs to activate or repress
endogenous ErbB-2 expression was studied using
retroviral delivery of the transcription factors into
cells. The retroviral vector pMX, which contains a
transcription factor expression cassette linked to
an internal ribosome entry sequence (IRES) fused
to the coding sequence of green fluorescent protein
(GFP), was used for these studies. 53GFP expression
from this construct is then correlated with transcription factor activity. Immunofluorescence
studies were used to confirm the nuclear
expression of zinc finger proteins and the cytoplasmic expression of GFP in retrovirally infected
A431 cells.
Table 2 lists the values for activation and repression levels obtained with each TFZF. Transcriptional
activation was studied using VP64, and repression
was studied using the Kox-1 Kruppel-associated
box (KRAB) domain.5,6,19 Six TFZFs had greater
than twofold activation of endogenous ErbB-2
expression in A431 cells. TFZFs 1A and 16 enhanced
expression 5.2-fold and 4.8-fold, respectively,
relative to cells transduced with control virus. In
contrast, only two of the ten TFZFs were able to
repress ErbB-2 expression greater than twofold:
2A with 2.2-fold repression, and 16 with 7.1-fold
repression. Figure 4(a – d) presents the FACS profiles of the two best activating (1A and 16) and
repressing (2A and 16) TFZFs. Only 16 was able to
both up and down-regulate ErbB-2 expression.
The six proteins that showed activation or repression of ErbB-2 expression had predicted binding
sites within a GNN-rich region of DNA characterized previously as a P/PMR mirror repeat.35
The P/PMR region can form a triple helix and
prevent binding of endogenous TFs adjacent to
the neighboring TATAA box.35,36 The P/PMR
region has been characterized to have an increased
level of chromatin-relaxing proteins associated
with it in cell-lines expressing high levels of
ErbB-2, compared to cells with low-level ErbB-2
expression.37 Considering the substantial level of
ErbB-2 expression in A431 cells, the P/PMR region
would be expected to be an accessible region of
binding. The inability of the same ZFP to provide
both up and down-regulation, depending on the
fused effector domain, in spite of being in a presumably accessible region of DNA, supports
the hypothesis that the topology of a promoter
varies from position to position, depending on
endogenous factors.
With the predicted binding sites and functional
data collected, we were surprised to find that
TFZF9A and TFZF16 did not share the same transcriptional control profiles. Both ZFP9A and
ZFP16 are predicted to bind at site 2 369, with 9A
being a perfect match. While TFZF16 is able to both
up-regulate and down-regulate ErbB-2 expression
robustly, TFZF9A up-regulates ErbB-2 expression
only moderately. Studies are ongoing to determine
if subtle differences in binding between these two
proteins explains their differences in regulation.
Affinity of ZFP16 for its three potential
binding sites
Since TFZF16 was differentiated by its ability to
both up-regulate and down-regulate, we sought to
determine which of its three potential binding
Phage Display Selection of Zinc Finger Transcription Factors
605
Figure 4. FACS analysis and immunofluorescence of A431 cells transduced with zinc finger fusion proteins.
a – d, FACS profiles are shown for cells transduced with the fusion proteins that displayed activation and repression of
endogenous ErbB-2 expression. Black indicates the staining of A431 cells with IgG1 isotype control antibody, blue is
the level of ErbB-2 expression in cells transduced with a stuffer fragment, and red is the level of ErbB-2 expression as
a result of transduction with the fusion protein expression construct indicated in each panel. e –l, Maximal projections
of 3D datasets show the total cellular distribution of TFZF (g and k). A431 cells were transduced with TFZF 16K (i – l) and
control StufferK (e– h). The top panels show A431 nuclei in blue; the upper middle panels show GFP in green; the
lower middle panels show TFZF in red; and the bottom panels show merge signals of TFZF (red), GFP (green), and
nuclei (blue). TFZF localization in nuclei was observed in TFZF 16K-transduced cells (l). No TFZF was detected in
control, StufferK-tranduced A431 (h).
sites within the ErbB-2 promoter sequences (Table
1) was responsible for the biological phenotype.
The prediction of the biologically relevant site was
first addressed by determining the affinity of
ZFP16 for each of the three sites. Figure 3 shows
the DNase I footprinting for each DNA:ZFP interaction site. Three independent binding reactions
were used to calculate Kd values. DNase I footprint
analysis of ZFP16 bound on the fragment containing site 2 642 (Figure 3 panel a) shows one
606
Phage Display Selection of Zinc Finger Transcription Factors
Figure 5. FACS analysis of six finger proteins 2642, 2 369 and 2285. Flow cytometric analyses of ErbB-2 expression
in cells retrovirally transduced with the six different fusion proteins are indicated. Two days after transduction, cells
were stained with the ErbB-2 specific antibody TA-1 in combination with Cy5-labeled secondary antibody. The broken
line represents transduced cells stained with IgG1 isotype control antibody, the thin line represents ErbB-2 expression
of cells transduced with stuffer DNA, and the bold line is ErbB-2 expression as a result of TFZF transduction.
binding site that extends over the sequence
50 -GGGGAGGAA-30 . The affinity of ZFP16 for this
site was determined to be 2.9 (^ 0.6) nM. ZFP16
was also predicted to bind a site at 2 369. Footprint
analysis of ZFP16 on a 189 bp fragment containing
the 2 369 binding site showed binding of ZFP16 at
two regions. The predicted binding site on this
fragment at 2 369 was 50 -TAGGAGGGA-30 and
the associated affinity was determined to be 51.6
(^ 29) nM. ZFP16 also bound a second site with
seven out of nine matches to the promoter
sequence: 50 -GGGGAGGGG-30 . The affinity of
ZFP16 for this site was not determined. The third
predicted binding site for ZFP16 was site 2 285.
The binding site extended over the predicted
binding sequence, 50 -TTGGAGGGG-30 , with an
associated affinity of 5.9 (^ 1.8) nM. Additional
binding sites were found between positions 2 243
and 2 218 that included three adjacent sites, each
with seven out of nine matches with ZFP16’s
predicted DNA recognition:
50 -GAGAAGGAGGAGGTGGAGGAGGAGGGCT-30 .
In vitro and in vivo characterization of three
six-finger proteins for specific regulation of
ErbB-2
In order to determine the biologically relevant
site for TFZF16, three six-finger proteins were constructed; 6Fn-642, 6Fn-369 and 6Fn-285. Six-finger
proteins recognize 18 bp of DNA and were
designed to bind each of ZFP16’s three predicted
sites independently. The six-finger proteins were
characterized by sequencing, by evaluation of
binding to oligonucleotides displaying the target
sequence, and by DNase I footprinting. Figure 3d
shows the DNase I footprinting analysis of
6Fn-642 on a 180 bp fragment containing the predicted site. The site bound is 50 -CACATCCCCCT
CCTTGA-30 . Figure 3e is the DNase I footprint
of 6Fn-369 on a 189 bp fragment at the site 50 GGCGCTAGGAGGGACG-30 . Figure 3f is the
DNase I footprint of 6Fn-285 on a 165 bp fragment
at the site 50 -CGTCAACCTCCCCCGCTC-30 . The
affinities of 6Fn-642, 6Fn-369 and 6Fn-285 for their
respective sites were 4.8 (^ 1.1) nM, 4.8 (^ 1.3) nM,
and 9.5 (^ 4.8) nM. It was interesting to find
that two of the three six-finger proteins had lower
affinities for the same target sites as their threefinger counterpart, ZFP16. Typically, six-finger
proteins are associated with higher affinity.5,20,38
To evaluate the biological effect of retroviral
expression of each of the six-finger proteins fused
to activation or repression domains, cell-surface
expression of ErbB-2 was monitored by FACS
analysis. Each of the constructs was infected into
A431 cells in two independent experiments and
the resulting up-regulation and down-regulation
profiles of each are illustrated in Figure 5. 6Fn-285
showed no ability to up-regulate or down-regulate
ErbB-2 expression in A431 cells, 6Fn-642 could
slightly down-regulate ErbB-2 expression, and
6Fn-369 was able to up-regulate and down-regulate
ErbB-2 expression approximately tenfold from
basal expression levels. DNase I footprint analysis
showed that 6Fn-369, with an affinity of 4.8 nM
for site 2 369, does not bind at sites 2 285 or 2 642
with equal or better affinity (data not shown).
Immunofluorescence studies included A431 cells
transduced to express the three-finger protein
ZFP16 fused to KRAB (ZFP16K), the six-finger protein 6Fn-369 fused to KRAB (6F3K), and cells
Phage Display Selection of Zinc Finger Transcription Factors
607
Figure 6. TFZF and ErbB-2 visualization on A431. Maximal projections of 3D datasets show the total cellular distribution of ErbB-2 (d and f) and TFZF (g– i). A431 cells were transduced with the following fusion proteins: ZFP16K
(a, d, g, and j), 6Fn-369K (b, e, h, and k), and control StufferK (c, f, i, and l). The top panels show nuclei in blue; the
upper middle panels show ErbB-2 receptor in green; the lower middle panels show ZFP in red; and the bottom
panels show merge signals of TFZF (red), ErbB-2 (green), and nuclei (blue). TFZF localization in nuclei was observed
in ZFP16K-transduced and 6Fn3K-transduced cells, and correlated with ErbB-2 down-regulation (j and k). No ZFP
was detected in control Stuffer-transduced A431 (l).
608
transduced with a negative control DNA fragment
fused to KRAB (StufferK) (Figure 6). Immunofluorescence studies confirmed the down-regulation of
cell surface ErbB-2 expression shown by FACS
analysis of ZFP16-KRAB and 6Fn369-KRAB
(Figure 6d and e versus f). Immunofluorescence
studies also showed a correlation of TFZF
expression of ErbB-2 (Figure 6j and k versus l). The
bottom three panels of Figure 6 have corresponding movies in Supplementary Material.
Discussion
Previous guidelines for the selection of TFZFs for
control of endogenous gene expression have
included affinity, proximity to transcription start
sites, and proximity to DNase I hypersensitive
regions in cellular chromatin.5 – 8,10 None of these
parameters have proven to be absolute rules for
successful, specific gene regulation. With the roles
of endogenous TFs, mediator proteins, and chromatin remodeling proteins largely uncharacterized
for any given promoter, an empirical method of
sampling different binding sites is needed. The
availability of the phage display library used in
this work, coupled with the growing availability
of genome sequencing information, allowed us to
identify gene-regulating TFZFs without the need
for a well-characterized promoter, or a detailed
zinc finger assembly strategy. Here, we describe
isolation of ZFPs that bind along the ErbB-2 promoter, and characterize the ability of the selected
zinc finger fusion proteins to up-regulate or
down-regulate endogenous ErbB-2 expression in
human epidermoid carcinoma A431 cells.
Phage display provided means for selecting
ZFPs with high affinity for the binding sites provided by a 1.4 kb ErbB-2 promoter fragment.
In previous work, protein:DNA affinity was considered a key requirement for achieving gene
regulation with a synthetic transcription factor
such as a TFZF. With more examples of endogenous
gene regulation by TFZFs available, the data indicate that the position of targeting is as important
as affinity. For example, six-finger proteins pE2X,
pE3Y and pE3Z were designed to regulate ErbB-2
and ErbB-3 expression.9 Only pE3Y both up-regulated and down-regulated ErbB-3 expression, and
yet all three proteins had DNA-binding affinities
of , 10 nM. Similarly, only one of three TFZFs
targeted to the erythropoietin (EPO) promoter upregulated EPO transcription, and only three of ten
TFZFs were able to up-regulate VEGF-A secretion,
despite all these ZFPs having dissociation
constants of , 10 nM.7,8 The minimum affinity
required for a ZFP to function as an effective TFZF
amongst endogenous factors is not known. Theoretically, this value would vary depending on the
accessibility of a binding site(s), the number of
binding sites, and the position of the binding
site(s) amongst promoter-associated proteins. The
advantage of using the phage display method is
Phage Display Selection of Zinc Finger Transcription Factors
that ZFPs can be isolated that bind various regions
of the promoter and with a range of affinities. The
resulting pool of proteins is screened to identify
the combination of position and affinity that provides transcriptional control of the target gene.
Here, the most successful TFZF had a dissociation
constant of 52 nM for the biologically relevant target site, an affinity much lower than would be
expected. However, the affinity of a ZFP for this
site was improved to 4.8 nM with assembly of a
six-finger protein. Therefore, even if the affinity
that the phage display screen provides is not
optimal, if function is demonstrated, an extended
ZFP can typically improve the affinity.5,6,20,39
The lack of a strict correlation between affinity
and transcriptional regulation using TFZFs has
been acknowledged in the field. Investigation of
the role of chromatin structure as it effects TFZF targeting is ongoing.40,41 One method of determining
the accessible regions of chromatin-associated
DNA, is DNase I-hypersensitive site mapping. The
identification of a hypersensitive region was a
valuable approach in the targeting of VEGF-A.42
However, the same approach in the design of
TFZFs selective for PPARg2 regulation provided
differing levels and extents of regulation of both
of the PPARg isoforms, g1 and g2.10 In the targeting
of the EPO gene, there was no hypersensitive site
associated with site 2 862, although additional
characterization indicated proximity of the site to
a positioned nucleosome.7 In this study, 6Fn-369
was able to up-regulate and down-regulate ErbB-2
expression, and did not map within the DNase Ihypersensitive region of the ErbB-2 promoter. Of
the six TFZFs that showed binding in the hypersensitive region by DNase I footprinting analysis,
three showed activation of ErbB-2 expression, one
showed repression of ErbB-2 expression, and one
was unable to modulate the transcription of ErbB2 in either direction. The inconsistency of targeting
based on hypersensitive regions and the incomplete characterization of chromatin structure for
many promoters, supports the use of an empirical
method, without biases, to target TFZFs as provided
by phage display and a given linear DNA sequence.
Retroviral expression of each selected ZFP as a
fusion protein with an activation or repression
domain was assessed by FACS analysis of ErbB-2.
The most significant finding from the FACS data
was that activation and repression of ErbB-2
expression were not usually accessible from the
same DNA-binding domain. For example, TFZF11
was able to up-regulate, but not down-regulate
ErbB-2 expression. Considering that both fusion
proteins had the same DNA-binding domain, the
up-regulation data suggests that the DNA-binding
domain was able to bind the DNA of the promoter
in the context of cellular chromatin. Therefore, the
inability of TFZFs 1A, 2A, 8A, 9A or 11 to effect
transcriptional control of ErbB-2 is hypothesized to
be associated with the proteins recruited by KRAB
or VP64 that were unable to form a functional unit
or, if the functional unit was assembled, was
609
Phage Display Selection of Zinc Finger Transcription Factors
unable to modify the promoter’s transcriptional
program.18,43 By sampling different sites along a
promoter, sites amenable to VP64 and KRAB
assembly and function can be elucidated to provide gene regulation. Site 2 369 was a site from
which activation and repression could be coordinated. In the context of the ErbB-2 promoter, site
2 369 lies 20 bp from an upstream AP-2 binding
site and 20 bp from a downstream p300 binding
site. The AP-2 site acts in coordination with an
upstream AP-2 site to activate ErbB-2 expression
in breast cancer cells.44 AP-2 activation can be
countered by an estrogen-suppressible enhancer in
the first intron of the ErbB-2 gene.45 Binding of
p300 in the ErbB-2 promoter provides a target for
the adenovirus 5 E1A mediated repression of
ErbB-2, which is in clinical trials for breast cancer
treatment.46,47 Considering the characterization of
endogenous factors for the ErbB-2 promoter, site
2 369 is in proximity to regions accessible for the
recruitment of activation and repression mediator
proteins.
Once a successful three-finger TFZF was identified, it was unclear which of ZFP16’s three potential binding sites was biologically relevant, or if
binding at all three sites contributed to the transcriptional regulation observed. Six-finger fusion
proteins, which have extended recognition of
18 bp of DNA, were designed to bind independently at each of TFZF16’s best match sites to
evaluate transcriptional regulation derived from
each site. Even before ZFPs were described, vonHippel’s study of transcription factors and nucleotide sequence recognized that binding of extended
sequences would increase the information content
of a DNA-binding protein and reduce binding to
other DNA sites.48 This theory is supported by the
DNase I footprint analysis of the six-finger proteins
assembled for this study. Figure 3d– f shows that
only one 18 bp site was bound by each of the
six-finger proteins on the fragments footprinted.
This is in contrast to the upper panels of Figure 3
that show cross-reactive binding of the three-finger
protein, ZFP16, outside of the region labeled as the
best match site on each DNA fragment. Cross-reactivity of the three-finger ZFPs was most extensive
in the P/PMR region that contains nine consecutive GNN triplets. Although the footprinting data
were used to confirm the ability of the phage display selection method to isolate ZFPs that could
be associated with sites in the promoter fragment,
the footprinting analysis provided valuable insight
into the type of recognition provided by threefinger proteins versus six-finger proteins. The
subtlety of amino acid interactions within the ZFP
protein structure as well as in proximity to DNA,
continues to be an area of study for the zinc finger
field.49 DNase I footprinting presents a ZFP with a
biologically relevant diversity of binding sites not
represented by specificity ELISAs, gel mobilityshift assays, CAST assays, or available microarray
chips.38,50,51 While statistics predict a 9 bp perfect
match site for a given ZFP should be once every
Table 2. Luciferase reporter activation by TFZFs and retroviral transduction mediated activation and repression of
endogenous ErbB-2 expression by ZFPs
Zinc finger
protein
(ZFP)
1A
2A
8A
9A
2
6
8
11
13
16
Luciferase
reporter assay
Retroviral transduction
Fold
Fold
activation
repression
52.6 ^ 4.4
5.6 ^ 3.1
15.2 ^ 10.2
3.2 ^ 0.7
0.55 ^ 0.2
1.85 ^ 0.13
1.63 ^ 0.21
47.4 ^ 15.2
5.8 ^ 1.0
10.9 ^ 2.1
5.2 ^ 2.5
–
2.9 ^ 2.1
3.2 ^ 1.2
–
–
–
3.1 ^ 2.0
–
4.8 ^ 0.84
–
2.2 ^ 0.23
–
–
–
–
–
–
–
10.7 ^ 3.6
A431 cells were cotransfected with each TFZF listed and an
ErbB-2 promoter-luciferase reporter construct.5 Luciferase
activity in total cell extracts was measured 48 hours after transfection. Fold activation and standard deviation are derived
from triplicate measurements. Fold activation and repression of
the endogenous ErbB-2 promoter were evaluated in A431 cells
using FACS analysis following retroviral expression of each
TFZF listed. Two independent transductions were used to determine the fold activation and repression from basal ErbB-2
expression levels and the corresponding standard deviation
values.
2.6 £ 105 bp,50 this is an underestimate of the potential binding sites given the binding of eight out of
9 bp, or seven out of 9 bp match sites with affinities
, 100 nM as observed here. Taking into consideration this degeneracy, by increasing the number of
contacts needed to provide a minimum threshold
of binding affinity, the importance of using sixfinger proteins for specific gene regulation is
emphasized.
Phage display for the selection of TFZFs has
many advantages. First, the investigation is not
limited to sites that fit the guidelines previously
used for the selection of TFZF binding sites. Second,
the method provides TFZFs more rapidly than by
individual construction and testing. Finally, this
empirical approach requires only the preassembled
library of ZFPs and a linear sequence of DNA to
allow selection. Our application of this approach
has resulted in the successful regulation of the
human ErbB-2 promoter. An advantage of using
phage display is that each round of panning can
be tested in various cell lines to determine which
pool of TFZFs is amenable to transcriptional regulation in a particular cell type. We suggest that
this strategy holds potential for the rapid preparation of transcription factors for the characterization of genes and for therapeutic application via
gene therapy.
Materials and Methods
Selection by phage display
Construction of zinc finger libraries has been
610
described.14 PCR-generated libraries were subcloned in
pComb3H and amplified. Growth and precipitation of
phage were as described.30 Streptavidin-coated magnetic
beads (Dynal) were optimized to bind the 1.4 kb target
(0.05 mg target/ml beads). The beads were washed four
times with zinc buffer A (ZBA) (10 mM Tris, 90 mM
KCl, 1 mM MgCl2, 90 mM ZnCl2), 5 mM DTT and once
with ZBA, 5 mM DTT, 5% (w/v) non-fat dry milk. Target
was prepared in ZBA, 5 mM DTT, 5% non-fat dry milk,
1 M NaCl, added to the washed beads, and incubated
for one hour at 37 8C on a rotating wheel. After incubation, beads were washed once with ZBA, 5 mM DTT,
5% non-fat dry milk. Then 100 ml of filtered phage (1013
colony-forming units), 100 ml of ZBA, 5 mM DTT, 5%
Blotto, 4 mg of sheared herring sperm DNA (Sigma),
294 ml of ZBA and 2.5 ml of 1 M DTT were added to the
beads and the samples were incubated for three hours
at room temperature on a rotating wheel.
Phage ELISA
Phage ELISA was performed as described.52
Luciferase assays
Luciferase assays were performed as described,5
except that A431 cells were used.
Antibodies
ErbB-2 expression was detected using TA-1 antibody
(Calbiochem). Control staining was done using mouse
F(ab0 )2 IgG1-UNLB (SouthernBiotech). The secondary
antibody for both was Cy5-labeled, affinity-purified donkey F(ab0 )2 anti-mouse IgG (Jackson ImmunoResearch).
Retroviral gene targeting
For retroviral expression of the three-finger and sixfinger proteins, the zinc finger-KRAB and zinc fingerVP64 coding regions were cloned into a modified
pMX-IRES-GFP53 using SfiI restriction sites (IRES,
internal ribosome-entry site; GFP, green fluorescent protein). As a control for the retroviral infection, a construct
containing a stuffer fragment of DNA inserted at the Sfi
sites was prepared. The stuffer sequence coded for a
single-chain Fab modified with stop codons that prevent
expression. The retroviral pMX-IRES-GFP/zinc finger
constructs were transiently transfected into the packaging cell line 293 gag/pols6 by using Lipofectamine
Plus (GIBCO/BRL). Three hours after transfection the
medium was changed to 6 ml of fresh DMEM with 10%
(v/v) fetal calf serum (FCS) and penicillin/streptomycin/anti-mycotic antibiotics. Approximately 42 hours
later, culture supernatants were used for infection of
target cells in the presence of 8 mg/ml of Polybrene.
Four infections were performed, the first after 42 hours
of transfection, then 50, 66 and 74 hours following transfection. At 90 hours post-transfection, the mediumwas
changed to 10 ml of fresh DMEM, 10% FCS, antibiotics.
One week from the start of the transfection the cells
were harvested for analysis.
Flow cytometric analysis
Cells were trypsinized and washed in FACS buffer
(PBS, 1 mM EDTA, 25 mM Hepes (pH 7.0), 1% (w/v)
Phage Display Selection of Zinc Finger Transcription Factors
BSA) prior to staining. Two wells for each sample were
prepared with 105 cells; one received 5 mg/ml of TA-1
antibody and the other 5 mg/ml of IgG1 control antibody
in 100 ml of FACS buffer. After incubation on ice for one
hour, cells were washed twice in FACS buffer. Bound
antibodies were stained with Cy5-labeled donkey antimouse secondary antibody in 100 ml of a 1:100 (v/v)
dilution in FACS buffer. Finally, the cells were washed
twice in FACS buffer, resuspended in 500 ml of FACS
buffer, and analyzed for their fluorescence with a Becton
Dickinson FACSort.
Fold activation and repression were calculated based on
gating the GFP-positive population (examples shown in
Figure 4a–d). From the GFP-gated profiles, the geometric
mean value for the center of the peak derived from the protein expression as a result of TFZF infection (in Figure 4a–d
this peak is red) and the geometric mean value for
the center of the peak derived from cells transduced
with a stuffer DNA fragment fused to VP64 or KRAB
(Figure 4a–d, the blue peak) were calculated using CellQuest software. To calculate fold activation, the TFZF
derived peak value (red) was divided by the basal
expression peak (blue). For fold repression, the basal
expression peak (blue) was divided by the TFZF-derived
peak (red).
Construction and characterization of the
six-finger proteins
For the construction of the six-finger proteins, two
three-finger proteins binding each of the 9 bp half-sites
of each of the three 18 bp target sequences were constructed by grafting the appropriate DNA recognition
helices into the framework of the three-finger protein
Sp1C. DNA fragments encoding the two three-finger
proteins were assembled from six overlapping oligonucleotides as described.5 The oligonucleotides used were
based on DNA recognition helices characterized from
finger 2 variants of Zif268.9,30 6Fn-654 was designed to bind
50 -GTCAAGGAGGGGGATGTG-30 and was assembled
from domains pGTC, pAAG, pmGAG, pmGGG, pGAT
and GTG. 6Fn-369 was designed to bind 50 -GGCGCTAGG
AGGGACGAC-30 and was assembled from domains
pmGGC, pmGCT, pAGG (t), pAGG (j), pmGAC and
pmGAC. 6Fn-275 was designed to bind 50 -GCAGTTGGA
GGGGGCGAG-30 and was assembled from domains
pGCA, pmGTT, GGA, pmGGG, pmGGC, pmGAG.
Purification and footprinting
Zinc finger-coding DNA was subcloned into a modified pMAL-c2 (New England Biolabs) bacterial
expression vector and transformed into XL-1 Blue
(Stratagene). Protein was purified using the Protein
Fusion and Purification System (New England Biolabs)
following the manufacturer’s protocol, except that
ZBA/5 mM DTT was used as the column buffer. Protein
purity was determined by staining 4%– 12% Novex gels
with Coomassie brilliant blue. The protein concentration
was determined by Bradford assay with bovine serum
albumin (BSA) standards. The radiolabeled DNA fragments were generated by PCR using the human ErbB-2
promoter cloned into pGL35 with High Fidelity PCR
Master (Roche) with 60 pmol of the following (50 -32P)labeled primers (designated with an F) and 30 primers
(designated with a B):
342F (GCATTTTGAAGAATTGAGATAGAAGTCTTT
TTGGG)
611
Phage Display Selection of Zinc Finger Transcription Factors
(GGATTACAGGCATGTGCCACCATGACC)
450F (GGAGTTCAAGACCAGCCTCACCAACGTGG
AGAACC)
(GCTCTTGTTGCCCAGGCTGGAGAGC)
(CCAAATTTGTAGACCCTCTTAAGATCATGC)
(CCAAGCCTATTTGTTTTAATATCAAATAATGG)
(CCTGAGACTTAAAAGGGTGTTAAGAGTGGCAGC)
1110B (GCTTCACTTTCTCCCTCTCTTCGC)
(CCTAGGGAATTTATCCCGGACTCC)
1150B (CGAAGTCTGGGAGTCGGCAACTCC)
1090F (GCGAAGAGAGGGAGAAAGTGAAGC)
1270B (GCTTCACAACTTCATTCTTATACTTCC)
Buffer for PCR with all primers except 1270 and 1150
included 10% (v/v) DMSO The PCR program was two
minutes at 94 8C, (30 seconds at 94 8C, 30 s at 55 8C, 30
seconds at 72 8C) for 30 cycles, seven minutes at 72 8C,
then 4 8C. The footprinting assay was carried out in
triplicate as described.54 The 5 £ TKMC buffer (50 mM
Tris – HCl (pH 7.0), 50 mM KCl, 50 mM MgCl2, 25 mM
CaCl2,) was supplemented with 50 mM ZnCl2 and
5 mM DTT was added fresh to the binding reactions
Equilibrations were incubated for 12–18 hours at 4 8C
or for one hour at 37 8C and were subsequently
digested with 13 ml of a 0.00012 unit/ml solution of
DNase I, 1 mM DTT (Roche Diagnostics) Samples were
electrophoresed through an 8 M urea/6% poly-acrylamide
gel The gels were exposed on phosphorimager
plates, recorded by a PhosphorImager SI (Molecular
Dynamics), and subsequently analyzed using ImageQuant (Molecular Dynamics) and KaleidaGraph
software (Synergy, Reading, PA) to give Kd values as
described.54
Immunocytochemical analysis of TFZF localization
For analysis of TFZF expression in transduced A431
cells (TFZFs ZFP16K and 6Fn3K), cells were seeded on
poly(L-lysine)-coated Lab-Tek coverglasses and incubated for 30 minutes. Cells were washed with copious
amounts of PBS, incubated with 4% (v/v) paraformaldehyde for 20 minutes, washed again, and incubated in a
humidifying chamber at room temperature for one hour
with a mixture of biotinylated rat anti-HA mAb (2 mg/
ml, Roche Diagnostics, Indianapolis, IN) and rabbit antihuman ErbB-2 polyclonal antibody (5 mg/ml, Calbiochem, La Jolla, CA) in FACS buffer, 0.1% (w/v) saponin.
The cells were then stained for one hour at room temperature with the mixture of Cy5-conjugated donkey
anti-rabbit IgG polyclonal antibodies and Streptavidin/
rhodamine Red-X (both from Jackson Immunoresearch,
West Grove, PA) diluted 1:100 (v/v) in FACS buffer,
0.1% (w/v) saponin. Finally, the cells were incubated
with 40 ,6-diamidino-2-phenylindole dihydrochloride
(DAPI) solution for five minutes, washed with PBS, and
covered with SlowFade Antifade reagent. Four-color
(DAPI, GFP, rhodamine Red-X, and Cy5) three-dimensional datasets were collected with a DeltaVision system
(Applied Precision, Issaquah, WA). Exposure times
were 0.2– 0.5 s (2-binning), and images were obtained
with either 60 £ or 100 £ magnification oil objectives.
Three-dimensional reconstruction was generated by
capturing 150 nm serial sections along the z-axis.
Images were deconvolved (based on the Agard-Sadat
inverse matrix algorithm) and analyzed with softWorX
Version 2.5.
Acknowledgements
The authors thank Joel Gottesfeld, Christian
Melander and David Segal for their invaluable
assistance. This work was supported by the
US National Institutes of Health CA86258 and
DK61803. C.V.L. is a Skaggs Predoctoral Fellow.
References
1. Levine, M. & Tjian, R. (2003). Transcription regulation and animal diversity. Nature, 424, 147– 151.
2. Wray, G. A., Hahn, M. W., Abouheif, E., Balhoff, J. P.,
Pizer, M., Rockman, M. V. & Romano, L. A. (2003).
The evolution of transcriptional regulation in
eukaryotes. Mol. Biol. Evol. 20, 1377– 1419.
3. Tupler, R., Perini, G. & Green, M. R. (2001). Expressing the human genome. Nature, 409, 832– 833.
4. Iuchi, S. (2001). Three classes of C2H2 zinc finger
proteins. Cell. Mol. Life Sci. 58, 625–635.
5. Beerli, R. R., Segal, D. J., Dreier, B. & Barbas, C. F.,
3rd (1998). 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. Proc. Natl
Acad. Sci. USA, 95, 14628– 14633.
6. Beerli, R. R., Dreier, B. & Barbas, C. F., 3rd (2000).
Positive and negative regulation of endogenous
genes by designed transcription factors. Proc. Natl
Acad. Sci. USA, 97, 1495– 1500.
7. Zhang, L., Spratt, S. K., Liu, Q., Johnstone, B., Qi, H.,
Raschke, E. E. et al. (2000). Synthetic zinc finger transcription factor action at an endogenous chromosomal site. Activation of the human erythropoietin
gene. J. Biol. Chem. 275, 33850– 33860.
8. Liu, P. Q., Rebar, E. J., Zhang, L., Liu, Q., Jamieson,
A. C., Liang, Y. et al. (2001). Regulation of an
endogenous locus using a panel of designed zinc finger proteins targeted to accessible chromatin regions.
Activation of vascular endothelial growth factor A.
J. Biol. Chem. 276, 11323– 11334.
9. Dreier, B., Beerli, R. R., Segal, D. J., Flippin, J. D. &
Barbas, C. F., 3rd (2001). Development of zinc finger
domains for recognition of the 50 -ANN-30 family of
DNA sequences and their use in the construction
of artificial transcription factors. J. Biol. Chem. 276,
29466 – 29478.
10. Ren, D., Collingwood, T. N., Rebar, E. J., Wolffe, A. P.
& Camp, H. S. (2002). PPARgamma knockdown by
engineered transcription factors: exogenous PPARgamma2 but not PPARgamma1 reactivates adipogenesis. Genes Dev. 16, 27 – 32.
11. Xu, D., Ye, D., Fisher, M. & Juliano, R. L. (2002).
Selective inhibition of P-glycoprotein expression in
multidrug-resistant tumor cells by a designed transcriptional regulator. J. Pharmacol. Expt. Ther. 302,
963– 971.
12. Guan, X., Stege, J., Kim, M., Dahmani, Z., Fan, N.,
Heifetz, P. et al. (2002). Heritable endogenous gene
regulation in plants with designed polydactyl zinc
finger transcription factors. Proc. Natl Acad. Sci.
USA, 99, 13296– 13301.
13. Jouvenot, Y., Ginjala, V., Zhang, L., Liu, P. Q.,
Oshimura, M., Feinberg, A. P. et al. (2003). Targeted
regulation of imprinted genes by synthetic zincfinger transcription factors. Gene Ther. 10, 513– 522.
14. Blancafort, P., Magnenat, L. & Barbas, C. F. (2003).
612
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
Scanning the human genome with combinatorial
transcription factor libraries. Nature Biotechnol. 21,
269– 274.
Falke, D., Fisher, M., Ye, D. & Juliano, R. L. (2003).
Design of artificial transcription factors to selectively
regulate the pro-apoptotic bax gene. Nucl. Acids Res.
31, e10.
Reynolds, L., Ullman, C., Moore, M., Isalan, M.,
West, M. J., Clapham, P. et al. (2003). Repression of
the HIV-1 50 LTR promoter and inhibition of HIV-1
replication by using engineered zinc-finger transcription factors. Proc. Natl Acad. Sci. USA, 100,
1615– 1620.
Segal, D. J., Gonclaves, J., Eberhardy, S., Swan, C. H.,
Torbett, B. E. & Barbas, C. F., 3rd (2004). Attenuation
of HIV-1 replication in primary human cells with a
designed zinc finger transcription factor. J. Biol.
Chem. 279, 14509– 14519.
Wysocka, J. & Herr, W. (2003). The herpes simplex
virus VP16-induced complex: the makings of a regulatory switch. Trends Biochem. Sci. 28, 294–304.
Margolin, J. F., Friedman, J. R., Meyer, W. K., Vissing,
H., Thiesen, H. J. & Rauscher, F. J., 3rd (1994).
Kruppel-associated boxes are potent transcriptional
repression domains. Proc. Natl Acad. Sci. USA, 91,
4509– 4513.
Liu, Q., Segal, D. J., Ghiara, J. B. & Barbas, C. F., 3rd
(1997). Design of polydactyl zinc-finger proteins for
unique addressing within complex genomes. Proc.
Natl Acad. Sci. USA, 94, 5525– 5530.
Ronicke, V., Risau, W. & Breier, G. (1996). Characterization of the endothelium-specific murine vascular
endothelial growth factor receptor-2 (Flk-1) promoter. Circ. Res. 79, 277– 285.
Kappel, A., Schlaeger, T. M., Flamme, I., Orkin, S. H.,
Risau, W. & Breier, G. (2000). Role of SCL/Tal-1,
GATA, and ets transcription factor binding sites
for the regulation of flk-1 expression during murine
vascular development. Blood, 96, 3078– 3085.
Perier, R. C., Praz, V., Junier, T., Bonnard, C. &
Bucher, P. (2000). The eukaryotic promoter database
(EPD). Nucl. Acids Res. 28, 302–333.
Heinemeyer, T., Wingender, E., Reuter, I., Hermjakob,
H., Kel, A. E., Kel, O. V. et al. (1998). Databases on
transcriptional regulation: TRANSFAC, TRRD and
COMPEL. Nucl. Acids Res. 26, 362– 367.
Vityaev, E. E., Orlov, Y. L., Vishnevsky, O. V.,
Pozdnyakov, M. A. & Kolchanov, N. A. (2002). Computer system “gene discovery” for promoter structure analysis. In Silico Biol. 2, 257– 262.
Turner, B. M. (2002). Cellular memory and the
histone code. Cell, 111, 285– 291.
Georgel, P. T. (2002). Chromatin structure of eukaryotic promoters: a changing perspective. Biochem.
Cell. Biol. 80, 295– 300.
Choo, Y. & Klug, A. (1994). Selection of DNA binding
sites for zinc fingers using rationally randomized
DNA reveals coded interactions. Proc. Natl Acad. Sci.
USA, 91, 11168– 11172.
Wu, H., Yang, W. P. H. & Barbas, C. F., 3rd (1995).
Building zinc fingers by selection: toward a therapeutic application. Proc. Natl Acad. Sci. USA, 92,
344– 348.
Segal, D. J., Dreier, B., Beerli, R. R. & Barbas, C. F.,
3rd (1999). Toward controlling gene expression at
will: selection and design of zinc finger domains
recognizing each of the 50 -GNN-30 DNA target
sequences. Proc. Natl Acad. Sci. USA, 96, 2758–2763.
Dreier, B., Segal, D. J. & Barbas, C. F., 3rd (2000).
Phage Display Selection of Zinc Finger Transcription Factors
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
Insights into the molecular recognition of the
50 -GNN-30 ( family of DNA sequences by zinc finger
domains. J. Mol. Biol. 303, 489– 502.
Hapgood, J. P., Riedemann, J. & Scherer, S. D. (2001).
Regulation of gene expression by GC-rich DNA ciselements. Cell. Biol. Intl, 25, 17 – 31.
Seipel, K., Georgiev, O. & Schaffner, W. (1992). Different activation domains stimulate transcription
from remote (“enhancer”) and proximal (“promoter”) positions. EMBO J. 11, 4961– 4968.
Scott, G. K., Daniel, J. C., Xiong, X., Maki, R. A.,
Kabat, D. & Benz, C. C. (1994). Binding of an ETSrelated protein within the DNase I hypersensitive
site of the HER2/neu promoter in human breast
cancer cells. J. Biol. Chem. 269, 19848– 19858.
Scott, G. K., Chang, C. H., Erny, K. M., Xu, F.,
Fredericks, W. J., Rauscher, F. J., 3rd et al. (2000). Ets
regulation of the erbB2 promoter. Oncogene, 19,
6490– 6502.
Noonberg, S. B., Scott, G. K., Hunt, C. A., Hogan,
M. E. & Benz, C. C. (1994). Inhibition of transcription
factor binding to the HER2 promoter by triplexforming oligodeoxyribonucleotides. Gene, 149,
123– 126.
Mishra, S. K., Mandal, M., Mazumdar, A. & Kumar,
R. (2001). Dynamic chromatin remodeling on the
HER2 promoter in human breast cancer cells. FEBS
Letters, 507, 88 – 94.
Tan, S., Guschin, D., Davalos, A., Lee, Y. L.,
Snowden, A. W., Jouvenot, Y. et al. (2003). Zinc-finger
protein-targeted gene regulation: genomewide
single-gene specificity. Proc. Natl Acad. Sci. USA, 100,
11997– 12002.
Kim, J. S. & Pabo, C. O. (1998). Getting a handhold
on DNA: design of poly-zinc finger proteins with
femtomolar dissociation constants. Proc. Natl Acad.
Sci. USA, 95, 2812– 2817.
Urnov, F. D. (2002). A feel for the template: zinc
finger protein transcription factors and chromatin.
Biochem. Cell. Biol. 80, 321– 333.
Urnov, F. D., Rebar, E. J., Reik, A. & Pandolfi, P. P.
(2002). Designed transcription factors as structural,
functional and therapeutic probes of chromatin in
vivo. Fourth in review series on chromatin dynamics.
EMBO Rep. 3, 610– 615.
Liang, Y., Li, X. Y., Rebar, E. J., Li, P., Zhou, Y., Chen,
B. et al. (2002). Activation of vascular endothelial
growth factor A transcription in tumorigenic glioblastoma cell lines by an enhancer with cell typespecific DNase I accessibility. J. Biol. Chem. 277,
20087– 20094.
Peng, H., Begg, G. E., Schultz, D. C., Friedman, J. R.,
Jensen, D. E., Speicher, D. W. & Rauscher, F. J., 3rd
(2000). Reconstitution of the KRAB-KAP-1 repressor
complex: a model system for defining the molecular
anatomy of RING-B box-coiled-coil domainmediated protein – protein interactions. J. Mol. Biol.
295, 1139– 1162.
Vernimmen, D., Begon, D., Salvador, C., Gofflot, S.,
Grooteclaes, M. & Winkler, R. (2003). Identification
of HTF (HER2 transcription factor) as an AP-2
(activator protein-2) transcription factor and contribution of the HTF binding site to ERBB2 gene overexpression. Biochem. J. 370, 323– 329.
Perissi, V., Menini, N., Cottone, E., Capello, D., Sacco,
M., Montaldo, F. & De Bortoli, M. (2000). AP-2 transcription factors in the regulation of ERBB2 gene
transcription by oestrogen. Oncogene, 19, 280– 288.
Chen, H., Yu, D., Chinnadurai, G., Karunagaran, D.
Phage Display Selection of Zinc Finger Transcription Factors
47.
48.
49.
50.
51.
52.
53.
& Hung, M. C. (1997). Mapping of adenovirus 5 E1A
domains responsible for suppression of neumediated transformation via transcriptional repression of neu. Oncogene, 14, 1965– 1971.
Wang, S. C., Zhang, L., Hortobagyi, G. N. & Hung,
M. C. (2001). Targeting HER2: recent developments
and future directions for breast cancer patients.
Semin. Oncol. 28, 21– 29.
von Hippel, P. (1979). On the molecular bases of the
specificity of interaction of transcriptional proteins
with
genomic
DNA.
Biological
Regulation
and Development (Goldberger, R., ed.), vol. 1, pp.
279–347, Plenum Press, New York.
Miller, J. C. & Pabo, C. O. (2001). Rearrangement of
side-chains in a Zif268 mutant highlights the complexities of zinc finger-DNA recognition. J. Mol. Biol.
313, 309– 315.
Segal, D. J., Beerli, R. R., Blancafort, P., Dreier, B.,
Effertz, K., Huber, A. et al. (2003). Evaluation of
a modular strategy for the construction of novel
polydactyl zinc finger DNA-binding proteins. Biochemistry, 42, 2137– 2148.
Bulyk, M. L., Huang, X., Choo, Y. & Church, G. M.
(2001). Exploring the DNA-binding specificities of
zinc fingers with DNA microarrays. Proc. Natl Acad.
Sci. USA, 98, 7158– 7163.
Barbas, C. I., Burton, D. R., Scott, J. K. & Silverman,
G. J. (2001). Phage Display: A Laboratory Manual, Cold
Spring Harbor Laboratory Press, Cold Spring
Harbor, NY.
Liu, X., Sun, Y., Constantinescu, S. N., Karam, E.,
Weinberg, R. A. & Lodish, H. F. (1997). Transforming
613
growth factor beta-induced phosphorylation of
Smad3 is required for growth inhibition and transcriptional induction in epithelial cells. Proc. Natl
Acad. Sci. USA, 94, 10669– 10674.
54. Trauger, J. W. & Dervan, P. B. (2001). Footprinting
methods for analysis of pyrrole– imidazole polyamide/DNA complexes. Methods Enzymol. 340,
450 –466.
55. Chen, H. & Hung, M. C. (1997). Involvement of
co-activator p300 in the transcriptional regulation
of the HER-2/neu gene. J. Biol. Chem. 272, 6101– 6104.
56. Chen, Y. & Gill, G. N. (1994). Positive and negative
regulatory elements in the human erbB-2 gene
promoter. Oncogene, 9, 2269– 2276.
Edited by F. Schmid
(Received 14 January 2004; received in revised form
21 April 2004; accepted 22 April 2004)
Supplementary Material for this paper comprising one movie is available on Science Direct