This document summarizes a study that examined the protein-binding potential of C2H2 zinc finger domains. The researchers used the human OLF-1/EBF associated zinc finger (OAZ) protein, which contains 30 zinc fingers arranged in 6 clusters, as a model. They assessed DNA binding using a target site selection assay and protein binding using a yeast two-hybrid assay. They found that clusters known to bind DNA could also facilitate specific protein interactions, but clusters known to bind protein did not facilitate specific DNA interactions. They concluded that DNA binding is a more restricted function of zinc fingers, and their potential to mediate protein interactions is likely greater than previously estimated.
1. ORIGINAL PAPER
The Protein-Binding Potential of C2H2 Zinc Finger Domains
Kathryn J. Brayer Ć Sanjeev Kulshreshtha Ć
David J. Segal
Published online: 20 February 2008
Ć Humana Press Inc. 2008
Abstract There are over 10,000 C2H2-type zinc ļ¬nger
(ZF) domains distributed among more than 1,000 ZF pro-
teins in the human genome. These domains are frequently
observed to be involved in sequence-speciļ¬c DNA binding,
and uncharacterized domains are typically assumed to
facilitate DNA interactions. However, some ZFs also
facilitate binding to proteins or RNA. Over 100 Cys2-His2
(C2H2) ZF-protein interactions have been described. We
initially attempted a bioinformatics analysis to identify
sequence features that would predict a DNA- or protein-
binding function. These efforts were complicated by sev-
eral issues, including uncertainties about the full functional
capabilities of the ZFs. We therefore applied an unbiased
approach to directly examine the potential for ZFs to
facilitate DNA or protein interactions. The human OLF-1/
EBF associated zinc ļ¬nger (OAZ) protein was used as a
model. The human O/E-1-associated zinc ļ¬nger protein
(hOAZ) contains 30 ZFs in 6 clusters, some of which have
been previously indicated in DNA or protein interactions.
DNA binding was assessed using a target site selection
(CAST) assay, and protein binding was assessed using a
yeast two-hybrid assay. We observed that clusters known
to bind DNA could facilitate speciļ¬c protein interactions,
but clusters known to bind protein did not facilitate speciļ¬c
DNA interactions. Our primary conclusion is that DNA
binding is a more restricted function of ZFs, and that their
potential for mediating protein interactions is likely
greater. These results suggest that the role of C2H2 ZF
domains in protein interactions has probably been under-
estimated. The implication of these ļ¬ndings for the
prediction of ZF function is discussed.
Keywords Transcription factors Ć Protein-DNA
interactions Ć Protein chemistry Ć Structural biology Ć
Functional annotations
Introduction
Zinc ļ¬nger (ZF) domains are small, self-folding protein
structures that coordinate the binding of a zinc ion between
conserved amino acids, generally cysteines and/or histi-
dines. Although there are 20 different types of ZF domains,
each categorized by the structure of their zinc stabilizing
residues, the most common type is the Cys2-His2 (C2H2)
type [1, 2]. C2H2 ļ¬ngers contain the consensus sequence
(F/Y)-X-C-X2-5-C-X3-(F/Y)-X5-W-X2-H-X3ā4-H, where X
is any amino acid and W is any hydrophobic residue [3].
This motif folds to form a bba structure, and is so named
for the coordinated binding of a zinc ion by two conserved
cysteine and histidine residues that provide additional sta-
bility to the fold [4ā7]. C2H2 domains were initially
identiļ¬ed as the sequence-speciļ¬c DNA binding domain of
transcription factor TFIIIA in Xenopus laevis [8]. The
domain is frequently found in DNA binding proteins where
amino acids in the C-terminal a-helix of the fold recognize
and bind speciļ¬c DNA sequences, typically 2ā4 nucleo-
tides [3]. Since their discovery over 20 years ago,
recognition of the importance of the C2H2 ZF domain has
K. J. Brayer
Department of Pharmacology and Toxicology,
College of Pharmacy, University of Arizona, Tucson, AZ 85721,
USA
S. Kulshreshtha Ć D. J. Segal (&)
Genome Center and Department of Pharmacology,
University of California, 4513 GBSF, 451 Health Sciences
Dr., Davis, CA 95616, USA
e-mail: djsegal@ucdavis.edu
Cell Biochem Biophys (2008) 51:9ā19
DOI 10.1007/s12013-008-9007-6
2. grown, in part because of their broad distribution
throughout all kingdoms, representing as much as 2ā3% of
the eukaryotic proteome [2, 9, 10]. They have proven to be
as diverse as they are abundant. Although they are still
primarily considered the DNA binding domain of many
transcription factors, C2H2 ZFs also mediate interactions
with proteins and RNA, perform structural roles, or have
undetermined functions [11ā14]. For example, the review
adjoining this article describes over 100 C2H2 ZF-protein
interactions that have been documented in the literature
[15].
Zinc ļ¬nger proteins may contain between 1 and 40 ZF
domains, which are frequently arranged in groups or
clusters of tandem arrays. Typically, only 3ā4 ZFs in a
multi-ļ¬nger protein are involved in DNA binding. This
suggests that the remaining ļ¬ngers could be involved in
other types of interactions. Some illustrative examples
include the 6-ļ¬nger IKAROS protein, which contains four
DNA-binding ZFs and two protein-binding ZFs, the 9-
ļ¬nger FOG protein, which contains four DNA-binding ZFs
and four protein-binding ZFs, and the 9-ļ¬nger TFIIIA
protein, which contains two clusters of three DNA-binding
ZFs, three or four RNA-binding ZFs, and some ļ¬ngers that
appear to be involved in both DNA- and protein-binding
[14, 16ā26]. Even the 3-ļ¬nger proteins Zif268/EGR1 and
SP1, in which all three ZFs are known to bind DNA, have
been shown to mediate protein interactions with their ZFs
[27ā29]. Therefore, the common functional annotation of a
ZF-containing protein as DNA-binding, while not inaccu-
rate, is not sufļ¬cient to describe the full functionality
provided by the ZF domains. To fully understand the
diverse functions mediated by the enormous family of
C2H2 ZF proteins, a more accurate annotation of their
component ZF functions is required.
We initially attempted to identify sequence features that
would predict a DNA-binding or protein-binding function
for a ZF. It had been proposed previously that certain
sequences of the linker that joins tandem ZF domains were
indicative of DNA binding, particularly linkers with a
sequence similar to TGEKP [30, 31]. However, as discussed
in the adjoining review [15], many ZFs involved in proteinā
protein interactions (PPIs) also possess a TGEKP-like linker.
We therefore performed a bioinformatics analysis to search
for features that could better distinguish protein binding from
DNA-binding ZF. However, it became apparent that several
issues complicated our analysis of the existing ZF-DNA and
ZF-protein interaction data. Among these was an uncertainty
about the extent to which DNA-binding ZFs were involved in
protein interactions and vice versa. Would dual-function
ZFs, such as those found in Zif268/EGR1 and SP1, constitute
rare exceptions or frequent occurrences? Without such basic
information, it would be difļ¬cult to segregate ZFs into
mutually exclusive lists to search for sequence differences.
Previous studies aimed at determining the biological
roles of speciļ¬c ZF proteins have introduced a bias in that
only a subset of all possible interactions are investigated. In
this study, we investigated the hypothesis that a less biased
approach would produce a more complete description of
ZF function. As a model, we analyzed the human 30-ļ¬nger
protein OLF-1/EBF associated zinc ļ¬nger (OAZ), which
contains clusters of ZFs that had been previously impli-
cated in DNA or protein binding [32, 33]. To test our
hypothesis, six clusters of OAZ ZFs were independently
analyzed for DNA or protein-binding capacity by target
site selection or yeast two-hybrid analysis, respectively.
We observed that all six clusters were capable of speciļ¬c
protein interactions, but only Cluster 1 was capable of a
speciļ¬c DNA interaction. Our data therefore support our
hypothesis that ZFs can potentially mediate more functions
than are typically observed when studied in a particular
biological context. We further conclude that DNA binding
is a more restricted function of ZFs, while their potential
for mediating protein interactions is much larger. These
results suggest that the role of ZFs in mediating protein
interactions has probably been underestimated and under-
annotated. Useful functional assignments will likely
require a further exploration of ZF DNA- and/or protein-
binding capacity.
Material and Methods
Cyclical Ampliļ¬cation and Selection of Targets
(CAST) Assays
Human full-length OAZ cDNA was provided as a kind gift
from Dr. Joan MassagueĀ“ at the Memorial Sloan-Kettering
Cancer Center. The cDNA was subcloned into clusters
corresponding to ZFs 2ā5 (amino acids 78ā189), 6ā8 (aa
203ā289), 9ā13 (aa 349ā533), 14ā20 (aa 572ā775), 21ā25
(aa 826ā987), and 26ā30 (aa 1060ā1227) by PCR ampli-
ļ¬cation using primers containing appropriate restriction
sites. CAST assays were conducted as described previously
with minor changes detailed below [34]. Brieļ¬y, OAZ ZF
clusters were inserted into the BamHI and HindIII sites of
the prokaryotic expression vector pMAL-c2X (New Eng-
land Biolabs), to create fusion proteins consisting of an N-
terminal maltose-binding protein (MBP) and C-terminal
ZF domain. Proteins were over-expressed in BL21 (DE3)
E. coli (Invitrogen) after IPTG induction (0.3 mM). Pro-
teins were puriļ¬ed on an amylose resin column (New
England Biolabs), washed with Zinc Buffer A (ZBA;
10 mM Tris pH 7.5, 90 mM KCl, 1 mM MgCl2, 90 lM
ZnCl2, 5 mM DTT), and eluted with ZBA + 10 mM
maltose. Puriļ¬ed protein was aliquoted and stored at
-80Ā°C until use.
10 Cell Biochem Biophys (2008) 51:9ā19
3. Randomized libraries of double-stranded DNA were
created by PCR ampliļ¬cation of 150 pmole of a library
oligonucleotide, 50
-GAGCTCATGGAAGTACCATAG-
(N21)-GAACGTCGATCACTCGAG-30
, with the primers
50
-GAGCTCATGGAAGTACCATAG-30
and 50
-CTCGAG
TGATCGACGTTC-30
(10 cycles; 15 s at 94Ā°C, 15 s at
70Ā°C, and 60 s at 72Ā°C). Library oligos were trace labeled
by a single round of PCR with 10 lCi [a32
P]-dATP.
Labeled DNA library was incubated with protein in 1x
Binding buffer (ZBA, 1% BSA, 5 mM DTT, 0.17 lg/ll
herring sperm DNA, 10% glycerol) for 30 min at room
temperature, and then separated on a 5% TBE polyacryl-
amide gel in 0.5xTBE buffer. A well-characterized DNA
binding ZF (Aart, [35]) was also screened against the
random N-21 library as a positive control. Dried gels were
imaged on a Storm 860 (Molecular Dynamics). Shifted
bands were excised from the gel and the DNA was eluted
overnight at room temperature in elution buffer (0.1%
SDS, 0.5 M NH3OAc, 10 mM MgOAc).
Eluted DNA was precipitated, washed with ethanol, and
re-suspended in 20 lL DEPC water (Ambion). A non-
radioactive PCR reaction was performed to determine the
optimal number of ampliļ¬cation rounds, prior to a radio-
active reaction to generate the next labeled library for
CAST selection. Selection rounds were repeated until 50%
of the input was shifted or DNA could no longer be
ampliļ¬ed for the subsequent round (4ā7 rounds). Target
sequence was determined by TOPO-TA cloning ampliļ¬ed
DNA into pCR2.1-TOPO (Invitrogen), and submitting
samples to the UC Davis DNA Sequencing Facility.
Yeast Two-Hybrid
Bait constructs were created by cloning each OAZ ZF cluster
into the NcoI and XmaI sites of the bait vector, pGBTK7
DNA BD (Clontech). This vector appends an N-terminal
GAL4 DNA-binding domain (GAL4-BD) and a C-MYC
epitope tag. Cluster 3 contained an internal NcoI site, so NdeI
was used instead. Following sequence veriļ¬cation of the
insert, each bait vector was transformed into the AH109
yeast strain and grown on SD/-Trp/X-a-gal agar plates to
verify transformation and reporter gene activation. Expres-
sion of GAL4-BD-ZF fusion proteins was veriļ¬ed by
western blot using antibodies to the C-MYC epitope tag
(Sigma). OAZ ZF clusters were individually screened
against a commercial human fetal brain cDNA library
(Clontech #638804) that had been pre-cloned into pGADT7-
Rec and pre-transformed into yeast strain Y187. Screening
was performed using Clontechās Matchmaker Two-Hybrid
System 3 according to the manufacturerās protocol for high
stringency screening. Individual colonies displaying a
positive phenotype were re-struck twice on SD/-Ade/-His/-
Leu/-Trp/+X-a-gal for phenotype veriļ¬cation, prior to DNA
isolation for sequencing and co-immunoprecipitation. Yeast
plasmid DNA was isolated using the Zymoprep yeast plas-
mid miniprep kit (Zymo Research) per the manufacturerās
instruction.
Co-immunoprecipitation
Each OAZ ZF cluster was cloned into the BamHI and
HindIII sites of pcDNA3.1(-) (Invitrogen), in frame with an
engineered N-terminal FLAG epitope tag. Putative binding
partners selected from the yeast two-hybrid library were
PCR ampliļ¬ed using the Matchmaker 50
and 30
AD
LD-insert screening amplimers (Clontech), and cloned into
a pcDNA3.1(-) vector containing engineered Sļ¬I sites and
an N-terminal HA epitope tag. Full-length human Early B-
cell factor (EBF) cDNA (ATCC #9891743) was also sub-
cloned by PCR ampliļ¬ed (aa 221ā570) and cloned into the
BamHI and HindIII sites of a pcDNA3.1(-) vector con-
taining an engineered N-terminal HA epitope tag.
HEK293T cells were cultured in Dulbeccoās modiļ¬ca-
tion of eagleās medium (DMEM; Cellgro) with 4.5 g/L
glucose and L-glutamine and without sodium pyruvate, plus
5% new born calf serum (JR Scientiļ¬c) and 1X penicillin/
streptomycin/neomycin (Sigma). Prior to transfection, cells
were plated at a density of 4.4 9 106
cells/mL on 10 cm
plates. The next day, FLAG-ZF constructs were transfected
into HEK293T cells with either empty pcDNA3.1(-) vector,
or HA-partner pcDNA3.1(-) vector using Lipofectamine
2,000 per the manufacturerās protocol. A total of 12 lg of
DNA was mixed with Lipofectamine (1:3) in 3 mL Opti-
MEM 1 reduced serum media (Gibco) and incubated for 4 h
at 37Ā°C, before addition of 7 mL DMEM. After 16 h the
transfection media was removed and replaced with fresh
DMEM, and the cells were incubated an additional 24 h
before harvesting.
After removing DMEM and washing cells twice with
PBS, 1-mL ice-cold lysis buffer (10 mM TrisāHCl, pH 7.5,
10 mM NaCl, 0.5% Triton X-100, 1x Complete EDTA-
free protease inhibitor cocktail (Roche)) was added to each
plate and cells were harvested by scraping with a rubber
spatula. Lysed cells were transferred to a 1.5 mL micro-
fuge tube and incubated on ice for 10 min prior to the
addition of 3 M NaCl to a ļ¬nal concentration of 150 mM,
followed by an additional incubation on ice for 5 min.
Whole cell lysate was cleared by centrifugation at
14,000 rpm for 15 min at 4Ā°C, and supernatant was
transferred to a clean microfuge tube. Total protein con-
centration was determined by Bradford analysis. After
equilibrating protein concentrations, 1 mL of sample was
mixed with 40 lL of M2 anti-FLAG agarose beads
(Sigma) pre-washed with TNT buffer (50 mM TrisāHCl,
pH7.5, 150 mM NaCl, 0.05% Triton X-100). About
100 lM ZnCl2, 5 mM EDTA, or 20U TURBO DNase
Cell Biochem Biophys (2008) 51:9ā19 11
4. (Ambion) were added, where indicated, and rotated over-
night at 4Ā°C.
Beads were washed 8x with 1 mL of either 150 mM
wash (50 mM Tris 7.5, 150 mM NaCl, 0.5% triton X-100,
20 lM ZnCl2), 300 mM wash (50 mM Tris 7.5, 300 mM
NaCl, 0.5% triton X-100, 20 lM ZnCl2), or EDTA- wash
(50 mM Tris 7.5, 150/300 mM NaCl, 0.5% triton X-100).
Between each wash beads were collected by centrifugation
at 1,000 rpm for 1 min at 4Ā°C and supernatant was
removed. After ļ¬nal wash, beads were re-suspended in
SDS-loading buffer (50 mM Tris pH 6.8, 100 nM DTT,
2% SDS, 0.1% bromophenol blue, 10% glycerol), and
resolved on a 12% SDS-PAGE gel at 25 mA. Proteins were
transferred to a PVDF membrane (75 mA for 1.5 h). After
transfer, membrane was blocked with 5% MPT (100 mM
NaH2PO4, 27 mM KCl, 1.37 mM NaCl, 0.05% Tween, 5%
Nonfat dry milk), then incubated with anti-HA-HRP anti-
body (1:500; Sigma) for *4 h. Bands were visualized
using the ECL chemiluminescence detection kit (GE
Healthcare). Membranes were stripped (100 mM 2-mer-
captoethanol, 2% SDS, 62.5 mM TrisāHCl ph 6.7) and
re-probed with anti-Flag-HRP antibody (1:10,000; Sigma).
Results
A Bioinformatics-Based Approach to ZF Functional
Assignments
We initially tried to use hidden Markov modeling (HMM)
to identify protein sequence differences between
DNA-binding and protein-binding ZFs. A total of 56 DNA-
binding ZFs were compiled from the public transcription
factor database TRANSFAC (release 7.0, [36]) and addi-
tional sources. Also, 72 protein-binding ZFs were obtained
from the protein interaction databases BOND [37], HPRD
[38], and MINT [39], as well as additional sources. These
lists were used as input for the HMM modeler HMMER
[40]. Known cases of functional overlap, such as Zif268/
EGR1 and SP1, were treated as exceptions and excluded.
No obvious pattern differences emerged from this analysis
(data not shown).
There are several possible explanations for our inability
to detect sequences differences between ZFs that bind
protein and those that bind DNA. One concern is that the
interface surfaces for protein interaction are considerably
more heterogeneous than those for DNA binding. Virtually
all contacts to DNA bases extend from the N-terminus of
the a-helix. However, as detailed in the adjoining review
[15], C2H2 ZF-protein interaction surfaces can involve the
āāDNA faceāā of the helix, the side of the helix, the
b-strands, or regions involving both the a-helix and the
b-strands. Therefore, a simple alignment of known
protein-binding ZF sequences might not be expected to
reveal a protein-binding signature. A priori, we considered
that there might be a limited number of protein-interaction
modes; for example, the four classes mentioned above. If
that were true, it might be possible to sort the protein-
binding ZF sequences into bins representing the major
interaction modes. We performed cluster analysis on the
ZF lists using the PHYLIP algorithm in the ClustalW
program to generate phylograms [41]. However, even the
best clusters found were difļ¬cult to rationalize with known
interaction interfaces, and alignments of the cluster mem-
bers were not obviously different from each other, DNA-
binding ZF, or a mixture of protein and DNA-binding ZF
(data not shown). Therefore, these experiments were also
unrevealing in our hands.
In addition to the complications of detecting a ZF-pro-
tein-binding signature, our inability to distinguish protein-
binding ZFs from DNA-binding ZFs raised concerns about
the functional annotations of the C2H2 domains in our data
sets. Most of the database entries came from studies that
did not attempt a rigorous identiļ¬cation of the full potential
functionality of the ZFs. A typical ZF-protein interaction
study would identify a ZF protein in a yeast two-hybrid
assay, and then implicate the role of particular ZFs by
deletion analysis. However, such studies rarely investi-
gated, for example, if the protein-binding ļ¬ngers might
support DNA-binding in a different context. Based on these
concerns, we postulated that many ZFs might have
sequence features that allowed them to perform multiple
functional roles. We therefore decided to investigate the
hypothesis that the functional roles of ZFs could be more
accurately revealed by using an approach that was less
biased by a known set of interacting partners or biological
pathways.
hOAZ-DNA Interactions Detected by CAST Assay
Our experimental design for the unbiased elucidation of
potential ZF functions was to expose ZFs to all possible
DNA or protein sequences and select for interacting pairs.
In the case of DNA, this objective can be easily accom-
plished using a target site selection assay such as cyclical
ampliļ¬cation and selection of targets (CAST, [34]). In this
assay, puriļ¬ed ZF protein is exposed to a library of short
double-stranded DNA molecules. Bound ZF-DNA com-
plexes are separated in an electromobility shift assay
(EMSA). The bound DNA is then recovered and
re-ampliļ¬ed for additional rounds of selection or
sequenced to determine the preferred ZF binding sites.
Here we randomized a 21 bp region of DNA, producing a
library of 421
or 4.4 9 1012
DNA sequences. Since each ZF
recognizes a 3 bp site (or a 4 bp overlapping site), a DNA
library of this size would present every possible DNA
12 Cell Biochem Biophys (2008) 51:9ā19
5. binding site to a protein containing up to seven ZFs.
Although some proteins have been shown to require only
two ZFs for DNA binding [42], typically three or more
ļ¬ngers are required for high afļ¬nity binding. Cluster sizes
therefore needed to contain no less than three ļ¬ngers and
no more than seven.
For these experiments we wanted to analyze a protein
that contained known examples of DNA- and protein-
binding ZFs. The human O/E-1-associated zinc ļ¬nger
protein (hOAZ) was one of the ļ¬rst proteins recognized for
its ability to mediate PPIs with its C2H2 domains. A large
ZF protein containing 30 C2H2 domains arranged into six
clusters (Fig. 1), hOAZ uses different sets of ZFs to
interact with DNA, homodimerize, and to interact with
other proteins [32, 33, 43]. The six clusters of hOAZ,
corresponding to ZFs 2ā5, 6ā8, 9ā13, 14ā20, 21ā25, and
26ā30, were subcloned, puriļ¬ed from bacteria as fusions
with MBP, and applied to CAST analysis. As a positive
control for these experiments, a well-characterized 6-ZF
protein, Aart, was also puriļ¬ed as an MBP fusion and used
in the CAST assay [35]. This protein not only controls for
the technical aspects of the experiment, but also produces a
strong enrichment of its preferred target in the early rounds
of selection due it is unusually high afļ¬nity (70 pM). The
visible band of shifted DNA serves to indicate the position
of the nearly invisible bands of shifted DNA in samples
with proteins of lower afļ¬nity.
After four rounds of selection, a substantial portion of
the probe DNA was shifted by Aart positive control protein
(Fig. 2). Only faint bands of shifted DNA were visible in
the samples containing Clusters 1ā6, which could have
been due to either speciļ¬c or non-speciļ¬c ZF-DNA inter-
actions. However, it was no longer possible to amplify the
Fig. 1 Schematic
representation of the human
OAZ protein. The 30 C2H2 zinc
ļ¬nger domains are indicated by
rectangles. Brackets above
indicate previously described
functions of zinc ļ¬ngers [27,
28]. The ZF clusters used in this
study are indicated. The aligned
amino acid sequence of hOAZ
is shown below the schematic,
with clusters and ZFs labeled.
Positions -1 to 6 of the a-helix,
which are important for DNA
binding, and the linker regions
are indicated. Conserved
cysteines and histidines are
highlighted
Fig. 2 CAST analysis of speciļ¬c DNA-binding activity. TOP: After
4 rounds of CAST, the autoradiogram shows that some portion of the
labeled DNA library is still present in the assay (lower band), but very
little is bound by the six clusters of zinc ļ¬ngers (C1-C6). In contrast, a
large proportion of labeled DNA is clearly bound (upper band) by the
positive control protein Aart (+). BOTTOM: After 7 rounds of CAST,
a large portion of the DNA in the Cluster 1 sample and nearly all the
DNA in the positive control sample is bound to protein, indicating a
strong enrichment for binding sites in these samples. None of the
other clusters contained bound or unbound library DNA (not shown),
indicating the interactions were too weak to allow for selection and
subsequent ampliļ¬cation
Cell Biochem Biophys (2008) 51:9ā19 13
6. excised shifted DNA for Clusters 2ā6 in subsequent rounds
(data not shown), suggesting the observed bands did not
represent speciļ¬c interactions and were not selected sub-
sequently. These results indicated that Clusters 2ā6 were
unable to bind speciļ¬c DNA sequences under these
experimental conditions. In contrast, the shifted DNA in
the sample containing Cluster 1 continued to intensify in
subsequent rounds, indicating positive selection of speciļ¬c
targets. After seven rounds of selection, 15 random clones
of the Cluster 1 output were sequenced and revealed a
consensus sequence of 50
-TGGGTGTCA-30
(Fig. 3).
hOAZ-Protein Interactions Detected by Yeast Two-
Hybrid Assay
In comparison to DNA, it is not possible to expose a set of
ZFs to all possible protein sequences. The combinatorial
permutations using 20 amino acids are vast. Unlike the
largely 2-dimensional major groove of DNA, a PPI surface
is typically a large (&800 AĖ 2
) 3-dimensional structure
composed of any number of secondary structural elements
[44, 45]. As a practical approach, we used the yeast two-
hybrid method to screen a large commercial library from
human fetal brain (Clontech). This tissue was chosen
because it is among the most highly complex in terms of
the number of genes expressed within it [46]. The library
contained approximately 3.5 9 106
independent clones,
which represents the upper limit of the yeast two-hybrid
assay. Each of the six hOAZ clusters were subcloned into
the Matchmaker Two-Hybrid 3 bait vector (Clontech) and
checked for protein expression and autoactivation. The
clusters were then examined in the yeast two-hybrid assay.
We observed that all clusters tested positive for protein
interactions, after multiple rounds of re-streaking to retest
phenotypes, as indicated by blue colony growth on media
that was adenine-
, histidine-
, leucine-
, tryptophan-
, and
X-a-gal+
(Table 1). Approximately 10 random positive
colonies from each cluster were sequenced to sample the
identities of the putative interacting proteins (Table 2).
Several results from the yeast two-hybrid assay seemed
striking. Tsai and Reed [33, 43] showed that Cluster 6 of
the rat OAZ homolog was responsible for protein
interactions with itself as well as OLF/EBF. We were
therefore surprised that neither hOAZ nor EBF were among
the 14 positive clones for Cluster 6. Similarly surprising
was that Cluster 1, which was shown both here and by Tsai
and Reed [33] to interact with DNA, produced more than
80 positive yeast colonies. However, the yeast two-hybrid
assay is associated with a high rate of false-positives. We
therefore chose to verify a sample of the putative protein-
interactions involving Clusters 1 and 6 by the independent
method of mammalian cell co-immunoprecipitation.
Veriļ¬cation of Select Protein Interactions
by Mammalian Cell Co-immunoprecipitation
The ZF clusters were subcloned into the mammalian
expression vector pcDNA 3.1(-) containing an N-terminal
FLAG-epitope tag. The putative binding partners were
subcloned into pcDNA 3.1(-) containing an N-terminal
HA-epitope tag. For experiments with Cluster 1, we found
the putative partner XRCC6 to be particularly interesting
since it is known to be involved in V(D)J recombination
and hOAZ has a demonstrated role in B-cell maturation
[33, 43, 47]. For experiments with Cluster 6, we tested
RNF157, an uncharacterized putative partner isolated from
the yeast two-hybrid assay. As a positive control, we used
amino acids 221ā570 of the protein EBF, which had been
previously shown to interact with Cluster 6 in the rat
homolog rOAZ [33, 43]. Plasmids expressing FLAG-
Fig. 3 The binding site selected
by Cluster 1 in the CAST assay.
Sequencing of 15 random
clones from the CAST output
for Cluster 1 revealed a
consensus sequence of 50
-
TGGGTGTCA-30
Table 1 Yeast two-hybrid results
Zinc ļ¬ngers No. clones screened No. coloniesa
Cluster 1 (ZF2ā5) 7.4 9 105
82
Cluster 2 (ZF6ā8) 1.4 9 105
10
Cluster 3 (ZF9ā13) 4.0 9 104
N/Ab
Cluster 4 (ZF14ā20) 3.0 9 105
11
Cluster 5 (ZF21ā25) 1.2 9 106
288
Cluster 6 (ZF26ā30) 2.8 9 106
14
a
As determined after two rounds of re-streaking (see methods)
b
Not all colonies were re-struck. Original plates had 508 colonies;
*50 were re-struck, only a subset of the original colonies, roughly
38% grew on third plate
14 Cell Biochem Biophys (2008) 51:9ā19
7. tagged ZF clusters were co-transfected into the cell line
HEK263T with an equal amount of either HA-tagged
putative partner or empty vector. After harvesting cells,
protein complexes were immunoprecipitated with anti-
FLAG beads, resolved on a 12% SDS-PAGE gel, and
transferred to a PVDF membrane for visualization using
anti-HA antibodies. The co-immunoprecipitation assay in
HEK293T cells veriļ¬ed the interaction of both Cluster 1
with XRCC6 (Fig. 4a) and Cluster 6 with EBF and
RNF157 (Fig. 4b). No signiļ¬cant interactions were
observed for the reciprocal combinations of Cluster 1 with
RNF157 and Cluster 6 with XRCC6 (data not shown).
The Observed Protein Interactions are Dependent
on Zinc
To further examine if the proper folding of the ZFs was
important to the interaction, we repeated the co-immuno-
precipitations in the presence of 5 mM EDTA. This
method had been shown previously to chelate the zinc ions
and disrupt the ZFsā activity [33]. Addition of EDTA
abolished the interaction between RNF157 and Cluster 6
demonstrating the importance of an intact C2H2 domain to
this interaction (Fig. 4b). Although the Cluster 6-EBF and
Cluster 1-XRCC5 interactions were not completely abol-
ished by the addition of EDTA, the interactions were
reduced to background levels. This result suggested that
Table 2 Potential interaction partners for OAZ Zinc ļ¬nger clusters
Cluster Frequency Protein name Amino acid
start
1 1 GAPDH 97
1 DBNDD2 45
1 FADH1 136
1 XRCC6 311
1 TUBB2A 1
1 Weakly similar to PTPRC 72
1 Copine 6 397
1 FTL 103
1 GPBP1L1 18
1 PIK4CB 620
2 1 VDAC2 108
1 DHX15 86
1 JARID1D 31
1 C2orf16 1,719
2 XPO6 616
2 DGKI 1,060
2 ADNP 532
3 1 AKAP9 3,748
1 OR2Z1 130
1 STY13 mRNA
1 Similar to VDAC2 212
1 Alu subfamily J
1 CCD57 1
1 OLIG1 1
1 ALDOA 1
1 PGAM1 143
1 NDUFB8 1
4 1 PGM1 456
1 PCBP1 21
4 Similar to p40 7
5 2 PSMA6 85, 113
1 PPIA 1
1 TNNI3K 4
1 PRKAG2 453
1 ETFA 197
1 ACOT4 58
1 CLU 187
2 PSAP 420, 252
1 NGFRAP1 1
6 1 RNF157 2
1 DUS1L 353
1 MAP1B 2,350
1 Genomic contig
1 PRR6 STOP
1 SYTL1 11
1 ATP1B1 70
2 Genomic contig
Fig. 4 Proteināprotein interactions mediated by Clusters 1 and 6.
HEK293T cells were co-transfected with (a) FLAG-tagged Cluster 1
and HA-tagged XRCC6, (b) FLAG-tagged cluster 6 and HA-tagged
RNF157 or HA-tagged EBF DNA, followed by immunoprecipitation
using anti-FLAG agarose beads, and probing with anti-HA (top
panels) or anti-FLAG (lower panels) antibodies. About 5 mM EDTA
or 20 U DNAse (+/-) was added to the whole cell lysates prior to
immunoprecipitation, as indicated. Experiments were repeated a
minimum of 3 times and results presented here are representative
Cell Biochem Biophys (2008) 51:9ā19 15
8. proper folding of the ZFs was important to these interac-
tions (Fig. 4a and b).
The Observed Protein Interactions are not Dependent
on DNA
Both Cluster 1 and XRCC6 are known to bind DNA. It was
therefore possible that the observed protein interaction was
actually mediated by the simultaneous binding to an
unexpected DNA. EBF also has a DNA binding domain.
The DNA-binding capability of RNF157 is unknown, and
Cluster 6 was found in this study not to interact with DNA.
The DNA-dependence of all four protein interactions was
tested by adding 20 U of DNase to the co-immunoprecip-
itation reaction, based on similar previous experiments
[48]. Although somewhat reduced overall, XRCC6 clearly
co-precipitated with Cluster 1 compared to background
(Fig. 4a). These results indicate that the interaction
between Cluster 1 and XRCC5 is likely not DNA-depen-
dent. As expected, the addition of 20 U of DNase had no
effect on the interaction between Cluster 6 and either EBF
or RNF157 (Fig. 4b).
Discussion
C2H2 ZF domains are well known as the DNA binding
domains in many transcription factors. A DNA-binding
function is frequently assumed for uncharacterized
domains, especially in high-throughput and bioinformatics-
based assays. However, some ZFs are also known to
mediate interactions with RNA and proteins. This study
focused on the protein-binding potential of C2H2 ZF
domains, due to our long-term interests in evaluating ZF
PPIs as a novel class of speciļ¬c drug targets. Transcription
factors have traditionally been poor drug targets because
the DNA-protein interface is often large, and neither the
DNA nor the interacting surface of the protein presents a
well-deļ¬ned binding pocket for small molecule binding.
On the other hand, many transcription factors also partic-
ipate in PPIs, which are more likely candidates for drug
targeting [49, 50]. To avoid common protein interaction
motifs such as the KRAB domain, which is found on a third
of all ZF proteins [51], we decided to study PPIs mediated
by the ZFs themselves. ZF-RNA interactions, which might
also be useful as speciļ¬c therapeutic targets, were not
examined. In part, this was because even less is know about
these interactions. Only a handful of examples have been
described [11, 12], and structures of the well-characterized
TFIIIA-5S RNA interaction have only recently been
reported [13, 14]. Like C2H2 PPIs, the role of ZF-RNA
interactions has probably been underappreciated and
deserves further investigation.
Our initial attempts to create algorithms that could
predict the protein or DNA binding behavior of unchar-
acterized ZFs were unsuccessful. There are several possible
explanations for our inability to detect sequence differ-
ences between ZFs that bind protein and those that bind
DNA. As discussed in greater detail in the accompanying
review article [15], there are signiļ¬cant differences in the
binding interfaces used by C2H2 domains for interacting
with DNA compared to protein. Virtually all C2H2 ZF-
DNA interactions rely on a small number of amino acids
localized in the N-terminus of the a-helix. ZF-protein
interaction surfaces are larger, and rely on a greater
diversity of amino acids that can be located anywhere on
the surface of the fold [52]. Our inability to detect a pro-
tein-binding signature from aligned protein-binding ZF
sequences, or even clusters of sequences, could be inter-
preted to indicate that the diversity of binding interfaces is
quite large. The heterogeneous nature of protein interaction
surfaces could be the major reason for the failure of the
HMM approach. However, it was still not obvious to us
why an āaveragedā alignment of diverse protein-binding
ZFs would necessarily look similar to an alignment of
DNA-binding ZFs.
Another explanation for our failure to detect a āāprotein-
binding signatureāā was that several of the ZFs might in fact
have more than one function. We further hypothesized that
a less biased functional analysis could reveal these addi-
tional functions. We therefore attempted to directly
investigate the ability of clusters of ZF domains from a
model protein, hOAZ, to interact with DNA or protein. The
human O/E-1-associated zinc ļ¬nger protein is a large ZF
protein containing 30 C2H2 domains arranged into six
clusters (Fig. 1). It uses different sets of ZFs to interact
with DNA, homodimerize, and to interact with other pro-
teins [32, 33, 43]. Working with the rat ortholog, rOAZ,
Tsai and Reed [33, 43] determined that ļ¬ngers 1ā7
(Clusters 1 and 2) bind DNA, and ļ¬ngers 25ā29 (Cluster 6)
mediate homodimerization as well as an interaction with
Olf-1/Early B-cell Factor (OLF1/EBF). More recently,
Hata et al. [32] reported ļ¬ngers 9ā13 (Cluster 3) of hOAZ
bound to a different DNA target than the target of the N-
terminal Cluster 1. They also reported that hOAZ inter-
acted with SMA and MAD related protein (SMAD) 1 and
SMAD 4 using ļ¬ngers 14ā19 (Cluster 4) [32]. Importantly,
they found that hOAZ interaction with O/E-1 inhibited
hOAZ-SMAD1/4 interactions, suggesting that these are
two separate transcriptional pathways [32]. The function of
Cluster 5, comprising ļ¬ngers 21ā25, is unknown.
We used two library-based methods, CAST and yeast
two-hybrid, to expose the domains to a wide variety of
potential DNA and protein-interaction partners. These
experiments were designed to avoid any pre-conceived
expectations of the function of the ZFs. However, we
16 Cell Biochem Biophys (2008) 51:9ā19
9. acknowledge that there are several limitations to this
approach. An interaction partner might be poorly repre-
sented or absent from the fetal brain library. This might be
a reasonable explanation for our failure to detect the
expected Cluster 6-EBF interaction in our yeast two-hybrid
assay. Some interactions may require speciļ¬c post-trans-
lational modiļ¬cations, or might only occur in certain cell
types in response to particular stimuli. This could explain
our failure to detect the expected Cluster 4-SMAD protein
interaction, which had been shown to be dependent on the
activation of bone morphogenetic protein (BMP) [32].
Finally, despite our use of clusters containing 3ā7 ļ¬ngers to
maximize the opportunity for interactions, some interac-
tions may require additional ļ¬ngers of hOAZ. Hata et al.
[32] demonstrated binding of a fragment containing Clus-
ters 2 and 3 to a BMP response element DNA binding site
that was independent of BMP stimulation. Our separation
of these clusters may explain why Cluster 3 was not
observed to bind DNA in our CAST assay.
The CAST assay only detected DNA binding by Cluster
1, which selected a consensus binding site of 50
-
TGGGTGTCA-30
. These results are consistent with the
ļ¬ndings of Tsai and Reed [33], who detected a palindromic
consensus sequence of 50
-GCACCC(A/T)(A/T)GGGTG-30
by CAST assay using the full-length rat homolog of OAZ
(similarities underlined). They also demonstrated homodi-
merization mediated by the equivalent of Cluster 6, and
speculated that selection of a palindromic DNA binding
sequence was dependent on homodimerization. Here we
separated the DNA binding domain (Cluster 1) from the
homodimerization domain (Cluster 6), and thus likely
selected for only half of the palindrome.
In principle, the limitations of our approach could have
diminished our ability to detect either DNA or protein
interactions. However, in contrast to the paucity of DNA-
interactions observed, we were able to select several
potential protein-interaction partners for all clusters in our
yeast two-hybrid assay. The yeast two-hybrid assay itself
has several limitations, such as a high rate of false-posi-
tives [53]. For this reason, hits identiļ¬ed in a yeast two-
hybrid assay are frequently veriļ¬ed using a different
method such as mammalian cell co-immunoprecipitation.
We chose to not verify all 47 hits listed in Table 2 and
instead decided to examine a few examples in some depth
to illustrate key points. Our co-immunoprecipitation results
conļ¬rmed the interaction between Cluster 1 and XRCC6.
Although the biological signiļ¬cance of the interaction was
not examined here, XRCC6 has been observed to interact
with C2H2 ZFs in other transcription factors [54]. We also
conļ¬rmed a novel interaction between Cluster 6 and
RNF157, possibly suggesting an additional function of
hOAZ. However, since RNF157 is functionally uncharac-
terized, the biological signiļ¬cance of the interaction
remains elusive. We conļ¬rmed that, like their rat homo-
logs, human OAZ and human EBF also interact through
Cluster 6. Addition of DNase did not disrupt any of the
interactions, but all were found to be sensitive to disruption
by EDTA. These results indicate that the PPIs were not
facilitated by the presence of DNA, and that a folded C2H2
domain was required for the interaction.
These data support our hypothesis that unanticipated
functional information can be obtained by using a less
biased approach, and that some ZFs that were determined
to have a particular function in one context may in fact be
capable of supporting other types of interactions. If further
studies were to ļ¬nd these conclusions to be generally true
for C2H2 ZFs in other proteins, it would suggest that a
simple algorithm for predicting a DNA- or protein-binding
function might be impossible, and perhaps meaningless.
Finally, the ļ¬nding that putative protein interactions were
observed (albeit largely unveriļ¬ed) for all ZF clusters, but
DNA interactions were only observed for one cluster,
strongly supports the conclusion that DNA binding is likely a
more restricted function of ZFs while their potential for
mediating protein binding is likely greater. In some respects,
this conclusion is obvious. There are a wide variety of ZF
protein-interaction interaction surfaces [15], while ZF DNA-
interactions are restricted to the N-terminus of the a-helix
and the major groove of DNA [3]. This conclusion is also
supported by the observations of others, such as Mackay
[55], who recently pointed out that in some cases āāa single
classical ZF is capable of mediating PPIs, and that an array of
such domains is necessary for high afļ¬nity DNA binding.āā
However, if the potential of C2H2 ZFs for mediating protein
interactions is indeed greater than their potential for DNA
binding, it would suggest a dramatic reevaluation of our
understanding of ZF function. The role of ZFs in protein
binding has probably been underestimated and under-anno-
tated. The possibility of developing ZF PPIs as a new class of
drug targets, or designing custom ZFs protein-binding
domains, has probably been underappreciated. It is therefore
critical that future studies of C2H2 domains consider protein
binding as well as DNA binding, so that we can gain better
insights into the biology of one of the largest protein super-
families of found in nature.
Acknowledgments We thank Dr. Joan MassagueĀ“ for the hOAZ
cDNA. This work was supported in part by Contract 9016 from the
Arizona Disease Control Research Committee. KJB received support
from an NSF IGERT-Genomics award.
References
1. Krishna, S. S., Majumdar, I., & Grishin, N. V. (2003). Structural
classiļ¬cation of zinc ļ¬ngers: Survey and summary. Nucleic Acids
Research, 31, 532ā550.
Cell Biochem Biophys (2008) 51:9ā19 17
10. 2. Matthews, J. M., & Sunde, M. (2002). Zinc ļ¬ngersāfolds for
many occasions. IUBMB Life, 54, 351ā355.
3. Wolfe, S. A., Nekludova, L., & Pabo, C. O. (2000). DNA rec-
ognition by Cys2His2 zinc ļ¬nger proteins. Annual Review of
Biophysics and Biomolecular Structure, 29, 183ā212.
4. Frankel, A. D., Berg, J. M., & Pabo, C. O. (1987). Metal-
dependent folding of a single zinc ļ¬nger from transcription factor
IIIA. Proceedings of the National Academy of Sciences of the
United States of America, 84, 4841ā4845.
5. Lee, M. S., Gippert, G. P., Soman, K. V., Case, D. A., & Wright,
P. E. (1989). Three-dimensional solution structure of a single zinc
ļ¬nger DNA-binding domain. Science, 245, 635ā637.
6. Parraga, G., Horvath, S. J., Eisen, A., Taylor, W. E., Hood, L.,
Young, E. T., & Klevit, R. E. (1988). Zinc-dependent structure of
a single-ļ¬nger domain of yeast ADR1. Science, 241, 1489ā1492.
7. Pavletich, N. P., & Pabo, C. O. (1991). Zinc ļ¬nger-DNA rec-
ognition: Crystal structure of a Zif268-DNA complex at 2.1 A.
Science, 252, 809ā817.
8. Miller, J., McLachlan, A. D., & Klug, A. (1985). Repetitive zinc-
binding domains in the protein transcription factor IIIA from
Xenopus oocytes. The EMBO Journal, 4, 1609ā1614.
9. Kersey, P., Bower, L., Morris, L., Horne, A., Petryszak, R., Kanz,
C., Kanapin, A., Das, U., Michoud, K., Phan, I., Gattiker, A.,
Kulikova, T., Faruque, N., Duggan, K., McLaren, P., Reimholz,
B., Duret, L., Penel, S., Reuter, I., & Apweiler, R. (2005). Integr8
and genome reviews: Integrated views of complete genomes and
proteomes. Nucleic Acids Research, 33, D297āD302.
10. Mackay, J. P., & Crossley, M. (1998). Zinc ļ¬ngers are sticking
together. Trends in Biochemical Sciences, 23, 1ā4.
11. Brown, R. S. (2005). Zinc ļ¬nger proteins: Getting a grip on RNA.
Current Opinion in Structural Biology, 15, 94ā98.
12. Hall, T. M. (2005). Multiple modes of RNA recognition by zinc
ļ¬nger proteins. Current Opinion in Structural Biology, 15,
367ā373.
13. Lee, B. M., Xu, J., Clarkson, B. K., Martinez-Yamout M. A.,
Dyson, H. J., Case, D. A., Gottesfeld, J. M., & Wright, P. E.
(2006). Induced ļ¬t and āālock and keyāā recognition of 5S RNA by
zinc ļ¬ngers of transcription factor IIIA. Journal of Molecular
Biology, 357, 275ā291.
14. Lu, D., Searles, M. A., & Klug, A. (2003). Crystal structure of a
zinc-ļ¬nger-RNA complex reveals two modes of molecular rec-
ognition. Nature, 426, 96ā100.
15. Brayer, K. J., & Segal, D. J. Keep your ļ¬ngers off my DNA: proteinā
protein interactions mediated by C2H2 zinc ļ¬nger domains. Cell
Biochemistry and Biophysics. doi:10.1007/s12013-008-9008-5.
16. Del Rio, S., & Setzer, D. R. (1993). The role of zinc ļ¬ngers in
transcriptional activation by transcription factor IIIA. Proceed-
ings of the National Academy of Sciences of the United States of
America, 90, 168ā172.
17. Fox, A. H., Liew, C., Holmes, M., Kowalski, K., Mackay, J., &
Crossley, M. (1999). Transcriptional cofactors of the FOG family
interact with GATA proteins by means of multiple zinc ļ¬ngers.
The EMBO Journal, 18, 2812ā2822.
18. Friesen, W. J., & Darby, M. K. (1997). Phage display of RNA
binding zinc ļ¬ngers from transcription factor IIIA. The Journal of
Biological Chemistry, 272, 10994ā10997.
19. Honma, Y., Kiyosawa, H., Mori, T., Oguri, A., Nikaido, T.,
Kanazawa, K., Tojo, M., Takeda, J., Tanno, Y., Yokoya, S.,
Kawabata, I., Ikeda, H., & Wanaka, A. (1999). Eos: A novel
member of the Ikaros gene family expressed predominantly in the
developing nervous system. FEBS Letters, 447, 76ā80.
20. Kelley, C. M., Ikeda, T., Koipally, J., Avitahl, N., Wu, L.,
Georgopoulos, K., & Morgan, B. A. (1998). Helios, a novel
dimerization partner of Ikaros expressed in the earliest hemato-
poietic progenitors. Current Biology, 8, 508ā515.
21. Morgan, B., Sun, L., Avitahl, N., Andrikopoulos, K., Ikeda, T.,
Gonzales, E., Wu, P., Neben, S., & Georgopoulos, K. (1997).
Aiolos, a lymphoid restricted transcription factor that interacts
with Ikaros to regulate lymphocyte differentiation. The EMBO
Journal, 16, 2004ā2013.
22. Pelham, H. R., & Brown, D. D. (1980). A speciļ¬c transcription
factor that can bind either the 5S RNA gene or 5S RNA. Pro-
ceedings of the National Academy of Sciences of the United
States of America, 77, 4170ā4174.
23. Perdomo, J., Holmes, M., Chong, B., & Crossley, M. (2000). Eos
and pegasus, two members of the Ikaros family of proteins with
distinct DNA binding activities. The Journal of Biological
Chemistry, 275, 38347ā38354.
24. Simpson, R. J., Yi Lee, S.H., Bartle, N., Sum, E. Y., Visvader, J.
E., Matthews, J. M., Mackay, J. P., & Crossley, M. (2004). A
classic zinc ļ¬nger from friend of GATA mediates an interaction
with the coiled-coil of transforming acidic coiled-coil 3. The
Journal of Biological Chemistry, 279, 39789ā39797.
25. Sun, L., Liu, A., & Georgopoulos, K. (1996). Zinc ļ¬nger-medi-
ated protein interactions modulate Ikaros activity, a molecular
control of lymphocyte development. The EMBO Journal, 15,
5358ā5369.
26. Theunissen, O., Rudt, F., & Pieler, T. (1998). Structural deter-
minants in 5S RNA and TFIIIA for 7S RNP formation. European
Journal of Biochemistry, 258, 758ā767.
27. Chapman, N. R., & Perkins, N. D. (2000). Inhibition of the
RelA(p65) NF-kappaB subunit by Egr-1. The Journal of Bio-
logical Chemistry, 275, 4719ā4725.
28. Lee, J. S., Galvin, K. M., & Shi, Y. (1993). Evidence for physical
interaction between the zinc-ļ¬nger transcription factors YY1 and
Sp1. Proceedings of the National Academy of Sciences of the
United States of America, 90, 6145ā6149.
29. Seto, E., Lewis, B., & Shenk, T. (1993). Interaction between
transcription factors Sp1 and YY1. Nature, 365, 462ā464.
30. Laity, J. H., Dyson, H. J., & Wright, P. E. (2000). DNA-induced
alpha-helix capping in conserved linker sequences is a determi-
nant of binding afļ¬nity in Cys(2)-His(2) zinc ļ¬ngers. Journal of
Molecular Biology, 295, 719ā727.
31. Ryan, R. F., & Darby, M. K. (1998). The role of zinc ļ¬nger
linkers in p43 and TFIIIA binding to 5S rRNA and DNA. Nucleic
Acids Research, 26, 703ā709.
32. Hata, A., Seoane, J., Lagna, G., Montalvo, E., Hemmati-Bri-
vanlou, A., & Massague, J. (2000). OAZ uses distinct DNA- and
protein-binding zinc ļ¬ngers in separate BMP-Smad and Olf sig-
naling pathways. Cell, 100, 229ā240.
33. Tsai, R. Y., & Reed, R. R. (1998). Identiļ¬cation of DNA rec-
ognition sequences and protein interaction domains of the
multiple-Zn-ļ¬nger protein Roaz. Molecular and Cellular Biol-
ogy, 18, 6447ā6456.
34. Segal, D. J., Beerli, R. R., Blancafort, P., Dreier, B., Effertz, K.,
Huber, A., Koksch, B., Lund, C. V., Magnenat, L., Valente, D., &
Barbas, C. F. 3rd (2003). Evaluation of a modular strategy for the
construction of novel polydactyl zinc ļ¬nger DNA-binding pro-
teins. Biochemistry, 42, 2137ā2148.
35. Dreier, B., Beerli, R. R., Segal, D. J., Flippin, J. D., & Barbas, C.
F. 3rd (2001). Development of zinc ļ¬nger domains for recogni-
tion of the 50
-ANN-30
family of DNA sequences and their use in
the construction of artiļ¬cial transcription factors. The Journal of
Biological Chemistry, 276, 29466ā29478.
36. Matys, V., Fricke, E., Geffers, R., Gossling, E., Haubrock, M.,
Hehl, R., Hornischer, K., Karas, D., Kel, A. E., Kel-Margoulis,
O.V., Kloos, D. U., Land, S., Lewicki-Potapov, B., Michael, H.,
Munch, R., Reuter, I., Rotert, S., Saxel, H., Scheer, M., Thiele, S.,
& Wingender, E. (2003). TRANSFAC: Transcriptional regulation,
18 Cell Biochem Biophys (2008) 51:9ā19
11. from patterns to proļ¬les. Nucleic Acids Research, 31,
374ā378.
37. Alfarano, C., Andrade, C. E., Anthony, K., Bahroos, N., Bajec,
M., Bantoft, K., Betel, D., Bobechko, B., Boutilier, K., Burgess,
E., Buzadzija, K., Cavero, R., DāAbreo, C., Donaldson, I., Do-
rairajoo, D., Dumontier, M. J., Dumontier, M. R., Earles, V.,
Farrall, R., Feldman, H., Garderman, E., Gong, Y., Gonzaga, R.,
Grytsan, V., Gryz, E., Gu, V., Haldorsen, E., Halupa, A., Haw,
R., Hrvojic, A., Hurrell, L., Isserlin, R., Jack, F., Juma, F., Khan,
A., Kon, T., Konopinsky, S., Le, V., Lee, E., Ling, S., Magidin,
M., Moniakis, J., Montojo, J., Moore, S., Muskat, B., Ng, I.,
Paraiso, J. P., Parker, B., Pintilie, G., Pirone, R., Salama, J. J.,
Sgro, S., Shan, T., Shu, Y., Siew, J., Skinner, D., Snyder, K.,
Stasiuk, R., Strumpf, D., Tuekam, B., Tao, S., Wang, Z., White,
M., Willis, R., Wolting, C., Wong, S., Wrong, A., Xin, C., Yao,
R., Yates, B., Zhang, S., Zheng, K., Pawson, T., Ouellette, B. F.,
& Hogue, C. W. (2005). The biomolecular interaction network
database and related tools 2005 update. Nucleic Acids Research,
33, D418āD424.
38. Peri, S., Navarro, J. D., Amanchy, R., Kristiansen, T. Z., Jon-
nalagadda, C. K., Surendranath, V., Niranjan, V., Muthusamy, B.,
Gandhi, T. K., Gronborg, M., Ibarrola, N., Deshpande, N.,
Shanker, K., Shivashankar, H. N., Rashmi, B. P., Ramya, M. A.,
Zhao, Z., Chandrika, K. N., Padma, N., Harsha, H. C., Yatish, A.
J., Kavitha, M. P., Menezes, M., Choudhury, D. R., Suresh, S.,
Ghosh, N., Saravana, R., Chandran, S., Krishna, S., Joy, M.,
Anand, S. K., Madavan, V., Joseph, A., Wong, G. W., Schie-
mann, W. P., Constantinescu, S. N., Huang, L., Khosravi-Far, R.,
Steen, H., Tewari, M., Ghaffari, S., Blobe, G. C., Dang, C. V.,
Garcia, J. G., Pevsner, J., Jensen, O. N., Roepstorff, P., Desh-
pande, K. S., Chinnaiyan, A. M., Hamosh, A., Chakravarti, A., &
Pandey, A. (2003). Development of human protein reference
database as an initial platform for approaching systems biology in
humans. Genome Research, 13, 2363ā2371.
39. Chatr-aryamontri, A., Ceol, A., Palazzi, L. M., Nardelli, G.,
Schneider, M. V., Castagnoli, L., & Cesareni, G. (2007). MINT:
The Molecular INTeraction database. Nucleic Acids Research,
35, 72ā74.
40. Eddy, S. R. (1998). Proļ¬le hidden Markov models. Bioinfor-
matics, 14, 755ā763.
41. Chenna, R., Sugawara, H., Koike, T., Lopez, R., Gibson, T. J.,
Higgins, D. G., & Thompson, J. D. (2003). Multiple sequence
alignment with the clustal series of programs. Nucleic Acids
Research, 31, 3497ā3500.
42. Fairall, L., Harrison, S. D., Travers, A. A., & Rhodes, D. (1992).
Sequence-speciļ¬c DNA binding by a two zinc-ļ¬nger peptide
from the drosophila melanogaster tramtrack protein. Journal of
Molecular Biology, 226, 349ā366.
43. Tsai, R. Y., & Reed, R. R. (1997). Cloning and functional
characterization of Roaz, a zinc ļ¬nger protein that interacts with
O/E-1 to regulate gene expression: Implications for olfactory
neuronal development. The Journal of Neuroscience, 17,
4159ā4169.
44. Jones, S., & Thornton, J. M. (1996). Principles of proteināprotein
interactions. Proceedings of the National Academy of Sciences of
the United States of America, 93, 13ā20.
45. Stites, W. E. (1997). Proteināprotein interactions: Interface
structure, binding thermodynamics, and mutational analysis.
Chemical Review, 97, 1233ā1250.
46. Velculescu, V. E., Madden, S. L., Zhang, L., Lash, A. E., Yu, J.,
Rago, C., Lal, A., Wang, C. J., Beaudry, G. A., Ciriello, K. M.,
Cook, B. P., Dufault, M. R., Ferguson, A. T., Gao, Y., He, T. C.,
Hermeking, H., Hiraldo, S. K., Hwang, P. M., Lopez, M. A.,
Luderer, H. F., Mathews, B., Petroziello, J. M., Polyak, K., Za-
wel, L., Kinzler, K. W., et al. (1999). Analysis of human
transcriptomes. Nature Genetics, 23, 387ā388.
47. Gu, Y., Jin, S., Gao, Y., Weaver, D. T., & Alt, F. W. (1997).
Ku70-deļ¬cient embryonic stem cells have increased ionizing
radiosensitivity, defective DNA end-binding activity, and
inability to support V(D)J recombination. Proceedings of the
National Academy of Sciences of the United States of America,
94, 8076ā8081.
48. Nieto, J. M., Madrid, C., Miquelay, E., Parra, J. L., Rodriguez, S.,
& Juarez, A. (2002). Evidence for direct proteināprotein inter-
action between members of the enterobacterial Hha/YmoA and
H-NS families of proteins. Journal of Bacteriology, 184,
629ā635.
49. Jung, D., Choi, Y., & Uesugi, M. (2006). Small organic mole-
cules that modulate gene transcription. Drug Discovery Today,
11, 452ā457.
50. Arndt, H. D. (2006). Small molecule modulators of transcription.
Angewandte ChemieāInternational Edition in English, 45,
4552ā4560.
51. Collins, T., Stone, J. R., & Williams, A. J. (2001). All in the
family: The BTB/POZ, KRAB, and SCAN domains. Molecular
and Cellular Biology, 21, 3609ā3615.
52. Zhou, H. X., & Qin, S. (2007). Interaction-site prediction for
protein complexes: A critical assessment. Bioinformatics, 23,
2203ā2209.
53. W. van Criekinge, Cornelis, S., M. Van De Craen, Vandenabeele,
P., Fiers, W., & Beyaert, R. (1999). GAL4 is a substrate for
caspases: Implications for two-hybrid screening and other GAL4-
based assays. Molecular Cell Biology Research Communications,
1, 158ā161.
54. Ishiguro, A., Ideta, M., Mikoshiba, K., Chen, D. J., & Aruga, J.
(2007). ZIC2-dependent transcriptional regulation is mediated by
DNA-dependent protein kinase, poly(ADP-ribose) polymerase,
and RNA helicase A. The Journal of Biological Chemistry, 282,
9983ā9995.
55. Gamsjaeger, R., Liew, C. K., Loughlin, F. E., Crossley, M., &
Mackay, J. P. (2007). Sticky ļ¬ngers: Zinc-ļ¬ngers as protein-
recognition motifs. Trends in Biochemical Sciences, 32, 63ā70.
Cell Biochem Biophys (2008) 51:9ā19 19