ABSTRACT
We have previously mapped a repression domain from the active transcriptional
repressor E4BP4 to a 65 amino acid segment near the C-terminus of the polypeptide. Here we show that the E4BP4 repression domain
interacts specifically with the TBP binding repressor protein Dr1. Mutants that
affect the ability of E4BP4 to bring about transcriptional repression are also
deficient in their binding of Dr1.
The results are discussed in the light of evidence for squelching of a
`global' repressor by a DNA binding defective E4BP4 mutant.
Transcriptional repression is an important component of gene regulation in eukaryotes and can be mediated through nucleosomal and higher-order chromatin structures, or more selectively through the action of
specific transcriptional repressors. Such repressors fall into two functional
classes that either down regulate the activity of one or more positively acting
transcription factors by, for example, competing for their DNA binding sites or
that possess intrinsic repressing activity. The latter group can be considered
as active transcriptional repressors and include The
Drosophila melanogaster
proteins Krüppel (
1
), Even-skipped (
2
), Engrailed (
3
) and Snail (
4
), the mammalian protein WT1 (
5
) derived from the Wilm's tumour gene, the virally transduced oncogene vErbA,
the thyroid hormone receptor in its unliganded state (
6
) and the bZIP factors ATF3 (
7
) and E4BP4 (
8
,
9
). In addition, the Retinoblastoma gene product Rb has recently been shown to
repress transcription when targeted to promoter DNA (
10
).
Distinct mechanisms of transcriptional repression are suggested by variation in
the ability of different repressors to reduce activated versus basal
transcription. While evidence suggests that repressors such as snail and Rb
interfere with or quench the communication between upstream activators and the
basal transcription machinery (
4
,
10
), other repressors inhibit basal as well as activated transcription as
evidenced by their ability to repress basal promoters lacking activator binding
sites (
2
). These repressors presumably suppress transcription initiation (or some later
stage) directly. The mammalian bZIP factor E4BP4 falls into the latter category
of factors (
11
). Discrete transferable repression domains have been mapped for some repressors
including E4BP4 (
9
,
11
) and as is the case for classical transcriptional activators there is little
consensus in their amino acid sequence, composition or (predicted) secondary
structure (
9
). By analogy with what is known about activators, active repressors such as
E4BP4 might be predicted to destabilise or discourage some stage of initiation
complex formation resulting in a decreased probability of initiation at a given
promoter and hence a reduced initiation rate. This idea is supported by the
work of Fondell
et al.
(
12
,
13
) who have shown
in vitro
that the thyroid hormone receptor (T
3
R), which is an active repressor in the absence of its ligand, interacts
directly with TBP and inhibits transcription at an early step during
preinitiation complex formation. Um
et al.
(
14
) have shown a similar interaction between the
Drosophila
Even-skipped protein and TBP. Similarly, transcriptional repression by Krüppel appears to involve interaction with the [beta] subunit of TFIIE (
15
).
Other active repressors could be expected to act through the same or similar
targets which might include the family of TBP-associated factors (TAFs) that make up
holo
-TFIID or one or more of a number of recently purified and/or cloned
TFIID/TBP-interacting factors, some of which, including NC1 (
16
), NC2 (
17
), Dr
1
(
18
), Dr
2
(topoisomerase I) (
19
,
20
), have inhibitory effects on transcription
in vitro
. Dr
1
is a general repressor and represses both basal and activated transcription in
a reconstituted
in vitro
transcription system (
18
) and represses a range of promoters
in vivo
(
21
). Factors such as Dr
1
presumably add a level of transcriptional `fine-tuning' and it is also possible that the activity of factors like this may
be increased in the presence of appropriate DNA binding active repressors. This
mode of repression has been suggested for the bZIP factor ATF3 (
7
) and for the Wilm's tumour gene product WT1 (
22
). In both cases, expression of non-DNA binding forms of the repressor resulted in transcriptional activation,
potentially due to sequestration of a transcriptional inhibitory component.
This is analogous to repression or squelching by the sequestration of co-activators through protein-protein interactions with over-abundant activators (
23
).
The transcriptional repressor E4BP4 is widely expressed in cell lines of human
origin, is a member of the bZIP family of transcription factors and recognises
an overlapping but distinct pattern of DNA binding sites to the CREB/ATF family
of factors (
8
). The optimum binding sequences for E4BP4 (TTATGTAA or TTACGTAA) are also
highly related to binding sites for the hlf/PAR family of bZIP factors (
24
,
25
). We have shown previously that E4BP4 is an active transcriptional repressor
when transiently expressed in human cell lines and that the repressing activity
resides fully in a 65 amino acid segment near the C-terminal of the protein (
11
). Here we present evidence that the E4BP4 repression domain interacts with the
TBP-binding repressor protein Dr
1
and discuss the significance of this in relation to evidence for titration of a
negatively acting factor by E4BP4
in vivo
.
Eukaryotic expression plasmids
. pGAL[delta]BstB has been described previously (
11
). pCMVP4 was constructed by ligating the
Bam
HI-
Xho
I fragment, containing the full-length E4BP4 coding sequence, from pSVKP4 (
8
) into plasmid pCDNA3 (Invitrogen). Plasmid pP4[delta]ZIP, which encodes the E4BP4 leucine zipper deletion mutation was
generated by digesting pSVKP4[delta]BstB (
11
) with
Nde
I and
Sal
I, filling in the recessed ends with Klenow fragment of
Escherichia coli
DNA polymerase I and dNTPs and then recircularizing. Plasmids pGAL-pm1 and pGAL-pm4 were constructed as follows. A fragment was amplified by PCR
from pP4RS2 (
8
) using the primers: 5'-TCCGGATCGAAGCCG- AAGCCATGCAGATC-3' or 5'-TCCGGATCGCAGCCGCAGCCATGCAGATC-3' respectively and 5'-TCTAGAAATTGTCTTTTAGATGTC. The resulting fragments were ligated into the
Eco
RV site of pBluescript (Stratagene) to generate pBS-pm1 and pBS-pm4. pBS-pm1 and pBS-pm4 were digested with
Bsp
EI and
Xba
I and the fragments were ligated into pGAL-CT3 (
11
) cleaved with the same enzymes. Mutations were verified (as was the fidelity of the PCR reaction) by DNA sequencing. Plasmids dz-pm1 and dz-pm4 were created as follows: BS-pm1 and BS-pm4 were digested with
Bsp
EI and
Xho
I and the resulting fragments were ligated into pPSVKP4 digested with the same
enzymes.
Reporter plasmids.
Plasmids
p[pi]S12(34)CAT, p[pi]GALCAT (
11
) and G5E1BCAT (
26
) have been described previously.
Bacterial expression plasmids.
Plasmid pGST[delta]BstB was generated by digesting pGST-CT4 (
11
) with
Bst
BI and
Sma
I and recircularizing the plasmid after filling the recessed ends.
Plasmids for in vitro expression.
Plasmids pET11aDr
1
(
18
) and pBShTOP1 (
27
) [containing the human topoisomerase I (Dr
2
) cDNA in pBluescript] have been described elsewhere. pET11aDr
1
was a gift from D. Reinberg and pBSTopI was a gift from R. Hania and J. C.
Wang. The E1a expression plasmids pSPNC, pSPCS, pSP5/3x pSP13S and pSP12S
containing various E1a deletion mutations and cDNA encoding the 13S and 12S
forms of E1a are all based on the vector pSP65a and have been described
previously (
28
).
cDNAs to be transcribed and translated were added as plasmid DNAs to a coupled
transcription translation rabbit reticulocyte lysate (Promega) according to the
manufacturer's instructions. Transcription required variously T3, T7 or SP6 RNA
polymerase and was carried out in the presence of [
35
S]l-methionine. Transcription/translation reactions were generally performed
for 60 min at 30oC after which time an equal volume of 50 mM HEPES-KOH, pH 7.9, 150 mM NaCl, 20% (v/v) glycerol was added.
HeLa cells were maintained in Dulbecco's Modified Eagle's Medium (DMEM)
supplemented with 10% FCS. Transfections and CAT assays were performed as
described in reference (
11
) with details given in figure legends.
Escherichia coli
BL21(DE3) were transformed with plasmids pGEX2TK (Pharmacia) or pGEX[delta]BstB. Glutathione transferase (GST) or GST fusion proteins were prepared
from 500 ml cultures of each strain by standard procedures (
11
). Purified proteins were stored bound to GSH agarose beads in PBS containing
0.5 mM DTT and 0.2 mM PMSF at 4oC.
For
in vitro
protein interaction assays (GST pull down assays) a standard binding reaction
contained up to 10 [mu]l of diluted rabbit reticulocyte lysate and 10 [mu]l protein-loaded GSH-Sepharose beads equilibrated in 25 mM Tris-HCl, pH 7.9, 1 mM dithiothreitol, 150 mM NaCl, 0.01% NP-40. Each reaction contained an equivalent quantity of
recombinant fusion protein as judged by SDS-PAGE and protein staining with Coomassie blue. The final volume of the
binding reaction was made up to 50 or 100 [mu]l with the same buffer. Unless stated otherwise binding was carried out at
20-22oC for 20 min. The bound complexes were briefly sedimented and washed
twice with 1 ml each time of 25 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM dithiothreitol, 0.05% NP-40. Beads were transferred to fresh tubes and were washed twice
with the above buffer lacking NP-40. The pelleted GSH-Sepharose beads were finally boiled in 20-50 [mu]l SDS-PAGE sample buffer and bound proteins were resolved by SDS-PAGE. After electrophoresis, gels were fixed in
25% isopropanol, 10% acetic acid and treated with Amplify (Amersham) according
to the manufacturer's instructions. Gels thus treated were dried and exposed to
film at -70oC.
Most models for the mechanism of action of a eukaryotic repressor involve direct
or indirect interaction with components of the general transcription machinery.
We therefore set out to identify nuclear proteins capable of interacting with
E4BP4. The minimal repression domain from E4BP4 (residues 299-363) fused to
S.japonicum
glutathione transferase (GST) and thus bound to glutathione (GSH) agarose beads
was used as a substrate to test for the binding of candidate proteins. We have
previously used this assay to test for interaction of E4BP4 with human TBP,
TFIIB (
11
) and TFIIE[beta] (unpublished results) with negative results. We therefore extended the
analysis of proteins that might interact with the repression domain of E4BP4 to
the TBP-associated regulatory factors Dr
1
and Dr
2
(DNA topoisomerase I).
Dr
1
and Dr
2
polypeptides were translated
in vitro
in the presence of [
35
S]methionine (Fig.
1
, lanes 1 and 4). GSH-agarose beads loaded with either GST (Fig.
1
, lanes 2 and 5) or a GST E4BP4 repression domain fusion protein (GST[delta]BstB beads, Fig.
1
, lanes 3 and 6) were incubated with rabbit reticulocyte lysates containing
either
in vitro
translated Dr
1
or Dr
2
. The beads were washed and any bound proteins eluted into SDS-PAGE sample buffer. Eluted proteins were visualized by fluorography after
SDS-PAGE. Dr
2
(topoisomerase I) was not retained on the GST[delta]BstB beads (Fig.
1
, lane 3), but Dr
1
was specifically retained on the GST[delta]BstB, but not control GST beads (Fig.
1
, lanes 6 and 5). Binding was not affected by inclusion of ethidium bromide in
the binding and washing buffers, nor was it affected appreciably by the
inclusion or omission of the non-ionic detergent NP-40 up to 0.05% (data not shown).
Two mutations in the E4BP4 repression domain were created that either partially
or fully abolish transcriptional repression activity in transient expression
assays (see below, Fig.
3
). Mutant pm1 in which two lysine residues (Lys-330 and Lys-332) were changed to glutamate residues has been described
previously (
11
) and renders the repression domain inactive when transiently expressed in HeLa cells as a chimera with the GAL4 DNA binding domain (GAL-pm1, Fig.
3
a). In the same assay the wild-type repression domain fused to the GAL4 DNA binding region resulted in an
~10-fold repression of transcription (GAL[delta]BstB, Fig.
3
A). Computer analysis of the primary structure of the E4BP4 repression domain
predicts an [alpha]-helical structure for the central part of the domain (
11
). We therefore constructed the mutant pm4, in which Lys-330 and Lys-332 where exchanged for alanine residues. This represents a less
drastic charge change than pm1 and would not be expected to disrupt potential [alpha]-helical structure. In transient transfection assays, a GAL4
1-147
-pm4 fusion protein (GAL-pm4) was approximately four times less effective as a repressor
than the wild-type repression domain (Fig.
3
A). Minimal 65 amino acid repression domains containing these mutations were
synthesised as GST fusion proteins and compared with the wild type for the ability to bind Dr
1
. As shown in Figure
3
B, mutant pm1, which was inactive in the
in vivo
repression assay, was negative for Dr
1
binding (lane 4) while mutant pm4 that was compromised for repression, but
which retained some activity also retained the ability to bind Dr
1
, albeit with reduced efficiency compared to the wild-type repression domain (lanes 5 and 7). Thus, on the basis of these
mutants at least, the ability of E4BP4 to bind Dr
1
correlates well with its transcriptional repression activity.
Dr
1
is a TBP binding protein and in the light of the above findings it is
interesting to note that careful comparison of the amino acid sequences of the
E4BP4 repression domain with yeast and human TBP revealed a region of
similarity (Fig.
4
) overlapping the basic repeat of TBP, suggesting that E4BP4 and TBP might
contain a similar Dr
1
binding motif (see below). A further similarity between the TBP basic repeat
and the E4BP4 repression domain concerns their affinity for the adenovirus E1a
product. In Figure
5
we present data showing that GST[delta]BstB will bind to E1a proteins containing the constant region 3 (CR3)
peptide; a binding preference also shown by TBP (
29
). It should be pointed out that E1a also has also been shown to bind DR
1
in vitro
(
30
), but this binding depends on the N-terminal region of E1a whereas interaction between E1a and both TBP and
E4BP4 involves constant region 3. Together this evidence suggests that the
basic repeat region of TBP and the repression domain of E4BP4 present a similar
binding surface, both capable of binding Dr
1
and E1a.
Since E4BP4 apparently represses both activated and basal transcription
in vivo
(
11
) we surmised that the mechanism of repression was likely to involve
interference with the basic machinery of transcription initiation at the
promoter. This could involve interaction with a component of the general
transcription machinery such as one of the general RNA polymerase II
transcription factors or accessory factors such as one of the TAFs. In order to
test this we constructed a non-DNA binding E4BP4 mutant, reasoning that overexpression of such a protein
in vivo
would titrate any interacting proteins, potentially resulting in a detectable
change in the level of transcription of a reporter gene. This would be
analogous to the squelching effect first described by Gill and Ptashne (
23
). To this end a variant of E4BP4 was constructed (P4[delta]ZIP; Fig.
6
) that lacks the leucine zipper dimerization domain (residues 102-128) but retains a basic DNA binding region that overlaps a predicted
nuclear localization signal sequence (
31
,
32
). P4[delta]ZIP was not expected to exhibit specific DNA binding activity as
dimerization is necessary for DNA binding of bZIP factors. When P4[delta]ZIP was synthesised by
in vitro
translation no DNA binding activity was detected towards an adenovirus E4
promoter ATF site to which full-length E4BP4 binds avidly (data not shown).
Expression in cells of either wt E4BP4 or non-DNA binding mutant P4[delta]ZIP produced apparently opposite transcriptional effects, with wt
protein repressing promoters containing it's cognate binding site while the
mutant stimulates both basal and activated transcription (Fig.
7
). These observations are reminiscent of the `squelching' phenomenon documented for non-binding versions of a number of transcriptional activators (
23
). Consequently, although it is formally possible that overexpression of P4[delta]ZIP may disrupt the normal mechanism of E4BP4 repression, the simplest
explanation of these results is that P4[delta]ZIP sequesters an inhibitory factor or `global' repressor that normally
mediates the repression function of E4BP4.
In the light of the
in vitro
protein interaction data we present here, a possible candidate for this E4BP4
repression cofactor is Dr
1
. Not only does Dr
1
bind specifically and tightly to the repression domain of E4BP4 (Figs
1
and
2
), but mutants within this domain that are either negative or defective for Dr
1
binding
in vitro
are similarly compromised for repression activity
in vivo
(Fig.
3
). In addition, we have also detected a weak but reproducible
in vivo
interaction between
E4BP4 and Dr
1
using the yeast two hybrid system (
35
) (not shown). Furthermore, the same mutations that cripple the E4BP4 repression
domain and disrupt its interaction with Dr
1
also interfere with the ability of the non-DNA binding form of E4BP4 to stimulate transcription (Fig.
8
). The ability of Dr
1
to down regulate both basal and activated transcription at a variety of
promoters has been well documented (
18
,
21
,
30
) and this would correlate with our findings in Figure
7
discussed above. Unfortunately, our attempts to determine the effects of
expressing exogenous Dr
1
in cells co-transfected by the P4[delta]ZIP expression construct have yielded variable and unreproducible
results probably due to modulation of the P4[delta]ZIP expression construct by exogenous Dr
1
(unpublished results). We therefore cannot unequivocally say that the
sequestered factor is Dr
1
.
Dr
1
also binds to the basic repeat region of the TATA binding factor TBP. Indeed,
the E4BP4 repression domain and this domain of TBP share certain sequence
similarities (Fig.
4
) and the ability to bind the adenovirus E1a product constant region 3 (Fig.
5
). This suggests that Dr
1
binds to either TBP or E4BP4 and we have certainly failed to find evidence for
a ternary complex containing all three components
in vitro
(unpublished data). Therefore, if Dr
1
is the physiological target for E4BP4 the question arises as to the role of
this interaction. Simply tethering Dr
1
to a promoter does not result in transcriptional repression (
21
) suggesting that it is not sufficient for E4BP4 to simply recruit Dr
1
to a promoter to cause repression. It also seems unlikely that E4BP4 `passes'
Dr
1
to TBP as Dr
1
appears to dissociate from E4BP4 very slowly under
in vitro
conditions (Fig.
2
B). However, this situation may be different
in vivo
where other factors may be involved. Notably, Merlmelstein
et al.
(
36
) have recently described a DR
1
-associated protein DRAP1 that increases the stability of the DR1-TBP-TATA complex and it may be that this protein also modifies
the interaction of Dr
1
with E4BP4.
It is also possible that E4BP4 represses transcription through both corepressor
(Dr
1
) dependent and independent mechanisms.
In vitro
transcription experiments using a transcription system composed largely of
recombinant factors have shown that E4BP4 (in the form of GAL[delta]BstB, Fig.
6
) is still capable of repressing a GAL-MLP chimeric promoter in reactions apparently lacking Dr
1
(K. Leung and D. Reinberg, personal communication). This would be analogous to
the suggestion made by Fondell
et al.
(
13
) to explain the fact that the thyroid hormone receptor T
3
R has been shown to interact directly with TBP
in vitro
and to interfere with preinitiation complex assembly (
13
) while persuasive evidence also exists that a corepressor, N-CoR, mediates ligand independent repression by T
3
R (
37
) thus indicating that two pathways leading to transcriptional repression exist.
We have recently identified a novel protein by yeast two-hybrid screening that is distinct from Dr
1
but which also binds E4BP4. We are currently examining whether this protein is
involved in a distinct mechanism of repression by E4BP4 or whether it is a
further component of a Dr
1
-mediated repression mechanism indicated by the protein interaction and
transfection data presented here.
I.G.C. is in receipt of a Postdoctoral Research Fellowship from The Wellcome
Trust, Grant No: 036652/z/92/z/PMG/RB and H.C.H. is supported by the Imperial
Cancer Research Fund.
REFERENCES
Return