ABSTRACT
ZF5 is a ubiquitously expressed protein originally identified by its ability to
bind and repress the murine
c-myc
promoter. It contains five C-terminal zinc fingers and a conserved N-terminal ZiN/POZ domain. This motif, found in a growing number of
zinc finger proteins, can inhibit DNA binding and mediate dimerization
[Bardwell,V.J. and Treisman,R. (1994)
Genes Dev.,
8
,
1664-1677]. In the current study, a cyclic amplification and selection of
targets (CAST) protocol detected preferred ZF5 binding sites which are highly
GC-rich. Binding to these sites by ZF5 depended upon the zinc fingers and was
enhanced when the ZiN/POZ domain was removed. Using transient cotransfection
assays, ZF5 was shown to activate the HIV-1 LTR and repress the
[beta]
-actin promoter. The ZiN/POZ domain was shown to mediate ZF5-dependent transcriptional activation and repression. From these
data, we conclude that ZF5 can both activate and repress in the context of
different natural promoters and that its ZiN/POZ domain can affect two
functions; DNA binding and transcriptional modulation.
ZF5 cDNA was cloned based on the ability of the ZF5 protein to bind a regulatory
region of the murine
c-myc
promoter (
2
). In addition to five zinc finger domains at the C-terminus, ZF5 contains an N-terminal ZiN (
ZF5 binds to two elements within the -290 to -240 bp region in the
c-myc
promoter (
2
). These sites flank a YY1 activator site (
4
,
5
) and overlap a Blimp-1 repressor site (
6
,
7
, Lin
et al.
submitted). ZF5 also binds to the -50 `Sp1' site of the herpes simplex virus 1 (HSV-1) thymidine kinase (tk) promoter (
2
). In these contexts and in a Gal4 fusion assay ZF5 is a transcriptional
repressor. Additional target genes for ZF5 have not been identified.
ZF5 is ubiquitously expressed with highest levels found in brain and ovary
tissues and fibroblast cell lines (
2
). Recently, ZF5 was independently cloned using a differential display technique
as a gene specifically expressed in slow-growth phenotype female preimplantation embryos (James Crane, personal
communication). This is consistent with a recent report demonstrating impaired
growth of cell lines which overexpress ectopic ZF5 (
8
). Thus, ZF5 is expressed early in development and may be important for
retarding cellular proliferation.
The number of known zinc finger proteins containing ZiN/POZ domains is growing
rapidly; they have been found in species as diverse as
Drosophila melanogaster
, mice and humans. ZiN/POZ proteins usually either activate or repress
transcription. Although the biological roles of mammalian ZiN/POZ proteins are
poorly understood, two human ZiN/POZ proteins, BCL6 and PLZF, appear to play a
role in leukemogenesis (
9
-
14
). In drosophila, approximately 40 ZiN/POZ family members have been detected (
3
) and several have been shown to regulate important developmental decisions (
15
-
19
).
The ZiN/POZ domain of the human ZID protein inhibits DNA binding by the ZID zinc
fingers and is a dimerization domain with a high degree of specificity for
dimerization partners (
1
). Consistent with its function as a dimerization domain, the ZiN/POZ domain of
drosophila bric a brac has an alpha-helical structure with a highly hydrophobic face rich in leucine residues
(
18
).
The studies reported here were undertaken to characterize the functional domains
of ZF5 and to identify natural promoters which are subject to ZF5 regulation.
Using a truncated form of ZF5 with enhanced DNA binding ability, a binding site
selection protocol was used to define the preferred ZF5 binding sequence. This
sequence is GC rich and shows similarity to Sp1 binding sites. Therefore, ZF5
was tested for its ability to regulate transcription of the HIV-1 LTR which depends upon three Sp1 sites. Interestingly, ZF5 activates
this promoter. Activation of the HIV-1 LTR by ZF5 depends on ZF5 binding to DNA, requires intact Sp1 sites, is
synergistic with the viral TAT protein and requires the ZF5 ZiN/POZ domain. The
ZiN/POZ domain was also shown to be required for repression by ZF5 on a natural
promoter and in a Gal4 fusion assay.
pGEX-ZF5, pGEX-[Delta]ZF or pALEX-[Delta]ZiN bearing
Escherichia coli
cultures were grown to an OD
600
of ~0.4-0.6 and induced with 1 mM IPTG for 2 h. The bacteria were
centrifuged at 4000
g
for 15 min and resuspended in GST binding buffer [50 mM Tris-HCl pH 8.0, 100 mM NaCl, 100 [mu]M ZnCl
2
, 1 mM DTT, 0.1% Triton X-100, 5% glycerol, 6.3 [mu]g/ml aprotonin, leupeptin, pepstatin and benzamidine, 63 [mu]g/ml
N
[alpha]
-p-
tosyl-l-lysine chloromethyl ketone (TLCK) and
N-
tosyl-l-phenylalanine chloromethyl ketone (TPCK), 725 [mu]M phenyl methane sulfonyl fluoride (PMSF)]. Cells were lysed by
sonication, the bacterial debris centrifuged at 6000
g
for 15 min at 4oC. The supernatant was saved for quantification and analysis. Because of
the significant insolubility of bacterially expressed ZF5 proteins, the pellets
were resolubilized by dialysis in GST binding buffer plus 725 [mu]M PMSF, 500 mM NaCl and 6 M urea and sequentially renatured in 4 M, 2 M, 1
M, 500 mM and 0.0 M urea at 4oC. GST-ZF5 and GST-[Delta]ZiN-ZF5 were purified by incubation with glutathione
(GSH)-agarose beads (Sigma), washed three times with GST binding buffer and
eluted with GST binding buffer plus 10 mM reduced glutathione but lacking
Triton X-100. ZF5-[Delta]ZiN was cleaved from the GSH-agarose beads/GST-[Delta]ZiN-ZF5 complex by activated factor X
(Boehringer Mannheim) in GST binding buffer at 4oC for 8 h. Proteins were analyzed by SDS-PAGE and quantified as described (
20
).
For each round of selection, 1.5 mg crude GST or GST-[Delta]ZiN-ZF5 extract was incubated with a 50% slurry of GSH-agarose in a final volume of 100 [mu]l in GST binding buffer. Binding of protein to
the beads occurred for 2 h at room temperature followed by two washes in GST
binding buffer. The bound beads were then equilibrated to ZF5 binding buffer
(20 mM Tris-HCl pH 7.5, 50 mM NaCl, 5% glycerol, 1 mM DTT, 100 [mu]M ZnCl
2
) by two consecutive washes. Duplex degenerate oligonucleotide (10 [mu]g), gift of Riccardo Dalla Favara (AGACGGATCCATTGCA[N
20
]CTGTAGGAATTCGGA) was added and binding to the immobilized protein occurred in a
50 [mu]l reaction for 20 min at room temperature. The resulting complex was washed
four times in ZF5 binding buffer and directly subjected to PCR amplification as
described in the text. Following six rounds of selection and amplification, the
resulting sequences were cloned into the pGEM-T vector (Promega). Nucleotide analyses were performed by the Consensus
Program offered by the Genetics Computer Group (
32
).
Probes for EMSA were produced by phosphorylation of oligonucleotides with [[gamma]-
32
P]ATP, PCR amplification and polyacrylamide gel purification. Unless otherwise
noted, 20 ng of bacterially expressed, purified ZF5-derived proteins were used; binding reactions occurred in ZF5 binding
buffer at room temperature for 20 min. Some EMSA experiments included 50 ng
poly(dA-dT)(Pharmacia) as competitor for non-specific DNA binding activity. Competitors for specific DNA binding
activity were preincubated with proteins in the binding reactions for 10 min
before addition of 20 000-40 000 c.p.m. of labeled probe per reaction. Bound and free complexes
were separated on a native 5% polyacrylamide gel in 0.25* TBE (22 mM Tris, 22 mM borate, 500 [mu]M EDTA) at 4oC.
End labeled probes were bound to the indicated proteins in 50 [mu]l ZF5 binding buffer. DNase I (Worthington) was added at the indicated
concentrations with a final CaCl
2
concentration of 2.5 mM for 1 min followed by addition of 100 [mu]l stop buffer [1% SDS, 20 mM EDTA, 200 mM NaCl, 200 ng/[mu]l glycogen (Boehringer Mannheim)]. Samples were phenol/chloroform
extracted, ethanol precipitated and resolved on a 7 M urea/TBE, 8%
polyacrylamide sequencing gel followed by autoradiography.
NIH 3T3 cells growing in 10 ml Iscove's Modified Dulbecco's Medium (IMDM) + 10%
newborn calf serum were split at a density of 5 * 10
5
cells/10 cm plate the day before transfection and the cells were again fed 3 h
before transfection. For co-transfection experiments in NIH 3T3 cells, 1 [mu]g of each reporter and the indicated amount of expression vectors were
combined with pBluescript II SK(+) carrier DNA (Stratagene) to a total of 11 [mu]g plasmid. These DNAs were added as a CaPO
4
precipitate (25 mM HEPES; 140 mM NaCl; 750 [mu]M Na
2
HPO
4
; 125 mM CaCl
2
) to the media of the cells to be transfected. The following day, 3T3 monolayers
were shocked for 2 min (15% glycerol; 25 mM HEPES; 140 mM NaCl, 750 [mu]M Na
2
HPO
4
) and incubated another 24 h in 10 ml medium. Cells were harvested and
luciferase activity was assayed as described (
2
).
NIH 3T3 cells were transfected as described with 20 [mu]g of each expression construct. The cells were harvested by scraping in PBS
on ice, counted, centrifuged at 2500
g
at 4oC, resuspended in Western substrate buffer (WSB) (50 mM Tris-HCl pH 6.8, 0.2% SDS, 1 mg/ml bromophenol blue, 0.1 M DTT, 10%
glycerol), and boiled for 5 min. 1.6 * 10
5
cell equivalents (ce) were loaded into each lane of a 10% SDS-PAGE gel, followed by electroblot onto nitrocellulose. The blot was fixed
for 1 min in isopropanol, rehydrated in water and blocked with 5% dry milk in
PBS. For monitoring the expression of the Gal4 fusion proteins, anti-yeast Gal4 DNA binding domain antibody (Upstate Biotechnology
Incorporated) was used at a 1:500 dilution in 2% dry milk in PBS; goat anti-rabbit IgG, peroxidase conjugated (Boehringer Mannheim), was used at a
1:10
4
dilution in 2% dry milk/PBS. For the detection of FZF5LexA, FZF5-[Delta]ZFLexA, FZF5[Delta]ZiNLexA and FZF5 (Flu tagged), monoclonal anti-influenza hemaglutinin antibody 12CA5 (Boehringer Mannheim) was
used at a 1 * 10
4
dilution and detected with rabbit anti-mouse IgG, peroxidase conjugated (Boehringer Mannheim). The bands were
visualized by ECL Western Detection (Amersham) and exposure to X-ray film (Kodak).
To construct pGEX-ZF5, ZF5 cDNA was PCR amplified with a synthetic N-terminal
Bam
HI site engineered into the 5' primer. A
Bam
HI-
Fsp
I fragment was blunt end cloned into the
Bam
HI site of pGEX-2T (Pharmacia). pGEX-[Delta]ZF resulted from an internal deletion which removed the ZF5
zinc fingers from pGEX-ZF5 by partial
Bst
EII digestion, end filling and religation. To construct pALEX-[Delta]ZiN, the same 5' primer was used to amplify ZF5 cDNA and a
Hin
cII-
Sac
I (blunt) fragment was cloned into the
Sma
I-
Not
I (Blunt) sites of pALEX (
33
). To construct Gal4-ZF5, the
Bam
HI-
Fsp
I fragment was cloned into the
Bam
HI-
Ecl
136II sites of Gal4 1-147 (
34
). Gal4-ZF5 and Gal4(.76) were constructed as described (Numoto
et al.
). Gal4-ZiN and Gal4-ZiNAc contain ZF5 amino acids 1-90 and 1-209, respectively. The 5'-ends both use the synthetic
Bam
HI site, the 3'-ends use synthetic
Sal
I sites to clone into the
Bam
HI and
Sal
I sites of Gal4 1-147. Gal4 AcX has synthetic
Sma
I and
Xba
I sites bracketing the sequences encoding amino acids 159-283 and cloned into the same sites in Gal4 1-147. pFLexA was constructed by PCR amplifying the first 85 codons
of LexA from pBTM116 (
35
) with primers which introduced
Sma
I and
Bam
HI sites on the 5'- and 3'-ends, respectively. Also, on the 3'-end is an engineered stop codon after
codon 85, this fragment was cloned into the
Sma
I and
Bam
HI sites of pGCN (
36
). To construct pFZF5LexA, ZF5 was PCR amplified from Gal4-ZF5 with a 5' primer which bound to the Gal4 1-147 multiple cloning sequence and contained an engineered
Xba
I site. The 3' primer bound to the last five codons of ZF5 before the first stop codon
and contained an engineered
Sac
I site. This fragment was cloned into the pFLexA
Xba
I and
Sac
I sites (a partial
Sac
I digestion strategy was employed). For pFZF5, an
Xba
I stop-linker (New England Biolabs #1062) was cloned into the
Sma
I site between the ZF5 and LexA sequences, ensuring the expression of only a ZF5
protein. pFZF5-[Delta]ZFLexA was constructed identically to pFZF5LexA except the 3' primer bound to codons 274-278, just upstream from the zinc fingers.
Construction of pFZF5-[Delta]ZiNLexA was also identical to that of pFZF5LexA except the original
Gal4 fusion construct contained a truncation in the 5' ZF5 sequence at the
Acc
I site. LTR-Luc, -158[Delta]LTR-Luc, -93[Delta]LTR-Luc and p[beta]actinLuc were as
described (
23
,
24
). To construct pGL2wtLTR and pGL2mSp1LTR, the wild-type and mutant HIV-1 LTR sequences from -177 to 84 were PCR amplified from HIV-CAT and NSPALL (
25
) and blunt end cloned into the
Eco
RV site of pBluescript II SK+ (Stratagene).
Hin
cII/
Bam
HI fragments with the HIV sequences were cloned into the
Sma
I and
Bgl
II sites of pGL2Basic (Promega). RSV
tat
is as described (
25
).
In order to identify potential ZF5 binding sites in natural promoters, a CAST
protocol (
19
) was employed to determine the consensus binding site for ZF5. Briefly, a pool
of synthetic oligonucleotides was designed such that 20 bases of degeneracy
were flanked by 15 base constant regions. Based on Bardwell and Treisman's data
(
1
) showing that the ZiN/POZ domain of ZID inhibited its DNA binding ability, we
used a truncated form of ZF5 lacking the ZiN/POZ domain in the CAST protocol.
Bacterially expressed glutathione acetyl transferase (GST) and GST-[Delta]ZiN-ZF5 (a GST fusion protein containing the C-terminal portion of ZF5 but lacking the ZF5 ZiN/POZ
domain, Fig.
1
A) were immobilized on glutathione (GSH)-agarose beads and incubated with the double-stranded oligonucleotide pool. The bound complexes were isolated
and subjected to 10, 14 or 18 cycles of PCR amplification. By 10 cycles, a
specific product from the GST-[Delta]ZiN-ZF5 matrix could be detected by agarose gel electrophoresis.
This product was selected by immobilized GST-[Delta]ZiN-ZF5 in a second round of CAST. Six rounds of selection were
performed followed by cloning and sequencing.
A
Based on the consensus binding sequence for ZF5, we hypothesized that ZF5 might
regulate transcription from the HIV-1 LTR which contains a very GC-rich region and depends on three Sp1 sites. A co-transfection assay in NIH 3T3 fibroblasts was employed to test
this possibility. A ZF5 expression plasmid, pFZF5, was engineered so that the
influenza hemaglutinin (Flu) epitope tag was fused in frame to the N-terminus of ZF5 so that protein expression could be monitored. A similar
construct, pFZF5-[Delta]ZFLexA, has a truncation just upstream of the zinc fingers which
are replaced by the LexA DNA binding domain. Expression of these fusion
proteins in NIH 3T3 cells is shown in Figure
3
A. pFLexA, a similar plasmid expressing the LexA DNA binding domain was used as
a vector control in the co-transfection experiment shown in Figure
3
B. Unexpectedly, co-transfection of pFZF5 resulted in a 12-fold activation of the pGL2wtLTR reporter plasmid. This was
dependent upon the ability of ZF5 to bind DNA since the FZF5-[Delta]ZFLexA protein could not activate the reporter (Fig.
3
B).
A
B
To determine the regions of ZF5 which are necessary for transcriptional
repression, a previously described Gal4 fusion/co-transfection assay (
2
) was employed to execute a deletional analysis. A luciferase reporter driven by
the HSV-1 tk promoter with or without five binding sites for the Gal4 protein was
used in co-transfections with expression plasmids encoding ZF5 fused to the Gal4 DNA
binding domain (Gal4 1-147, here abbreviated to G4 1-147). An expression plasmid, G4(.76), which encodes a fusion
protein containing an N-terminal fragment of ZF5 lacking the zinc fingers, repressed transcription
only in the presence of Gal4 sites, demonstrating that DNA binding was
dependent upon the Gal4 1-147 portion of the fusion protein (
2
). Since this system eliminated ambiguity which might result from ZF5 binding to
the HSV-1 tk promoter, G4(.76) and derivatives of it were used to monitor
transcriptional repression by ZF5 fusion proteins.
Repression activities were monitored by co-transfecting 1 [mu]g G5-TK-Luc with 5 [mu]g of each effector plasmid (Fig.
5
A). A construct expressing only the ZiN/POZ and acidic domains, G4 ZiNAc,
repressed transcription nearly as efficiently as G4(.76). G4 ZiN, containing
only the ZiN/POZ domain, partially repressed transcription. However, G4 AcX,
which is comparable to G4(.76) except for the deletion of the ZiN/POZ domain,
was unable to repress transcription (Fig.
5
A). Using antibodies to the Gal4 DNA binding domain, the expression levels of
all the fusion proteins were monitored by western blot and found to be roughly
equivalent (Fig.
5
B). These data show that the ZF5 ZiN/POZ domain plays an important role in
transcriptional repression. They also suggest that regions C-terminal to the ZiN/POZ domain participate in repression. This is
consistent with the location of two repression domains in the BCL6 protein (
26
).
A
The studies reported here provide information about the mechanism of action of
the ZF5 protein. We have demonstrated that the zinc fingers are required for
ZF5 to bind DNA and that the ZiN/POZ domain reduces the affinity of the protein
for DNA. Using a sensitive selection technique, a consensus high affinity
binding site for ZF5 was determined and shown to be GC rich with similarity to
Sp1 sites. This led us to examine its role in the transcriptional regulation of
the HIV-1 LTR, a Sp1-dependent promoter. We have shown that the HIV-1 LTR is transactivated by ZF5 while the human [beta]-actin promoter is repressed by ZF5. Finally, we
have demonstrated that the ZF5 ZiN/POZ domain participates in transcriptional
activation as well as repression.
Utilizing a binding site selection technique (CAST), high affinity binding
sequences for ZF5 were enriched and from these a consensus sequence was
derived. Consistent with previous findings, this sequence is rich in guanine
and cytosine nucleotides and resembles the Sp1 consensus (
2
). Other zinc finger proteins such as MAZ (
27
,
28
) have also been found to have overlapping binding specificity with Sp1 family
proteins.
Binding studies on the HIV-1 LTR (Fig.
4
B) and thymidine kinase (
2
) promoters show that ZF5 only binds a subset of Sp1 sites, consistent with the
finding that the ZF5 consensus is slightly different from the Sp1 consensus
(Fig.
2
B). However, for the `Sp1' sites which are recognized by ZF5, the two proteins
may compete for binding
in vivo
and the regulation of some promoters previously attributed to Sp1 may in fact
involve ZF5. It will be important in future studies to identify which known
`Sp1' sites are recognized by ZF5, to determine whether ZF5 activates or
represses when bound to these sites and to determine whether it binds the sites
in vivo
. In addition, it may be that ZF5 and Sp1 act together or synergistically in
some contexts such as the HIV-1 LTR. Our data show that ZF5 binds some but not all the Sp1 sites in this
promoter and also binds sites not occupied by Sp1 (Fig.
4
B). Our data are consistent with models in which ZF5 and Sp1 cooperate in
binding or transcriptional activation, although additional studies will be
necessary to test these possibilities.
The ZiN/POZ domain clearly confers on ZF5 a conformation which decreases binding
affinity (Fig.
1
E). Suppression of DNA binding by a region which is required for transcriptional
modulation seems paradoxical since transcriptional regulation requires DNA
binding. However our co-transfection data show that full-length ZF5 is indeed transcriptionally active, either activating or
repressing transcription, depending on the gene context. In addition, it
appears that the ZiN/POZ domains of other family members may have similar
properties since the ZiN/POZ region of BCL6 was recently shown to be necessary
for transcriptional repression (
26
).
To solve this paradox, it seems likely that
in vivo
there is a mechanism to regulate the ability of ZiN/POZ proteins to bind DNA
and to modulate transcription. Indeed, Bardwell and Treisman noted that the ZID
ZiN/POZ domain can direct assembly into subnuclear structures, presumably by
dimerization, suggesting that ZiN/POZ-containing proteins may be unavailable for gene regulation unless they are
modified (
1
). Phosphorylation or other post-translational modifications might alter the ability of the ZiN/POZ domain
to suppress DNA binding. Alternatively, association with transcriptional or
other regulatory proteins might stabilize ZiN/POZ proteins bound to DNA as part
of a multiprotein complex or might induce a conformation with higher affinity
for DNA.
When ZF5 is tethered to an artificial promoter by the GAL4 DNA binding domain,
transcriptional repression is observed. ZF5 also represses natural promoters
including those of the murine
c-myc
,
HSV-1 TK (
2
) and human [beta]-actin genes. However, ZF5 strongly transactivates the HIV-1 LTR promoter. Thus, as with many activator/repressor
proteins, an important question is what determines whether ZF5 will activate or
repress a promoter.
There are several models which can provide an explanation for the ability of ZF5
to both activate and repress transcription. The binding site consensus for ZF5
is not palindromic and its orientation may affect ZF5 activity. Alternatively,
ZF5 may function differently at high and low affinity binding sites, a
possibility suggested by studies on the drosophila protein GAGA (
29
,
30
) and by our isolation of both high and low affinity sites by CAST. Binding of
adjacent proteins may also affect ZF5 activity. It is interesting that there
are multiple ZF5 binding sites interspersed with Sp1 sites on the HIV-1 LTR where ZF5 functions as an activator and, as suggested above, ZF5 and
Sp1 may affect one another's activity. Adjacent proteins may affect the ability
of ZF5 to associate with co-modulators, TAFs or the basal transcription machinery. YY1, another zinc
finger protein which, like ZF5, can either activate or repress transcription,
has recently been shown to require association with a co-repressor to repress transcription (
31
).
The growing list of ZiN/POZ proteins, their involvement in human tumors and
their roles in drosophila development all underscore the importance of this
class of zinc finger proteins. The experiments reported here provide the
groundwork for addressing many intriguing questions regarding ZF5 and ZiN/POZ
proteins in general. It will be important to study the paradox of the ZiN/POZ
domain functions and to determine how DNA binding and transcriptional activity
are regulated
in vivo.
It will also be important to determine the relationship between Sp1 and ZF5
binding in different genes and to identify additional target genes for ZF5.
Finally, studies on the HIV-1 LTR and [beta]-actin promoters should help us understand what determines
whether ZF5 activates or represses transcription.
We would like to thank members of the Calame laboratory, Dr Andrew Henderson, Dr
Christos Panagiotidis and Bob Soliman for their insightful discussions and
generosity. We would also like to thank Dr Christian Schindler and Dr Gerald
Siu for critically reading this manuscript. This work was supported by grants
from the National Institutes of Health, the American Cancer Society and the
Council for Tobacco Research to KC.
REFERENCES
Return