ABSTRACT
EWS-FLI-1 is a chimeric protein produced in most Ewing's sarcomas. It
results from the fusion of the N-terminal-encoding region of the EWS gene to the C-terminal DNA-binding domain (the ETS domain) encoded by the FLI-1 ets family gene. Both EWS-FLI-1 and FLI-1 proteins function as
transcription factors that bind specifically to ets sequences (the ets boxes)
present in promoter elements. EWS-FLI-1 is a powerful transforming protein, whereas FLI-1 is not. In a search for potential DNA binding sites for
these two proteins, we have tested their ability to recognize the serum
responsive element (SRE) in the
c-fos
promoter. This
cis
element contains an ets box which can be occupied by members of the ETS protein
family which do not bind DNA autonomously but form a ternary complex with a
second protein, p67
SRF
(serum responsive factor). We demonstrate here that EWS-FLI-1, but not FLI-1, is able to form a ternary complex on the
c-fos
SRE. Using a GST pull-down assay, we show that both FLI-1 and EWS-FLI-1 interact
in vitro
with SRF in the absence of DNA. In electromobility shift assays, EWS-FLI-1 binding to the SRE is detectable in the absence of SRF whereas
the binding of FLI-1 is not, suggesting that the interaction with DNA is the step which
limits ternary complex formation by FLI-1. Deletion of the N-terminal portion of FLI-1 resulted in a protein which behaved as EWS-FLI-1, suggesting the existence of an N-terminal inhibitory domain in the normal
protein. Taken together, our data indicate that there are intrinsic differences
in the binding of EWS-FLI-1 and FLI-1 proteins to distinct ets sequences.
A t(11;22) or t(21;22) chromosomal translocation is observed in the vast
majority of Ewing's sarcoma and peripheral neuro-ectodermal tumors of childhood, suggesting a direct role for the corresponding fusion products in the formation of these tumors (
1
,
2
). The translocations juxtapose the 5' end of a gene encoding an RNA binding protein of undetermined function,
EWS, and the 3' end of a proto-oncogene encoding a transcription factor: FLI-1 in the t(11;22)(q24;q12) translocation (
3
) or ERG-1 in the t(21;22) (q22;q12) translocation (
4
). Both gene products belong to the ETS protein family which includes numerous
members involved in developmental processes, cellular responses to various
stimuli and cellular transformation (
5
). All proteins of the ETS family share an 85 amino-acid region, the ETS domain (
6
), usually located at their C-terminus, through which they specifically bind to, and transactivate through, promoter elements displaying a consensus GGAA core sequence
called the ets box. The nucleotides flanking the core sequence define
subclasses of ets boxes which are recognized by distinct ETS family members (
7
).
ETS family members have been shown to cooperate with other nuclear proteins in
transcriptional activation, and often seem to function as parts of larger
protein complexes. Two members of the ETS family, SAP-1 (
8
) and ELK-1 (
9
), are involved in the activation of promoters containing a serum responsive
element or SRE (
10
). They barely bind the SRE autonomously, whereas they strongly bind this
sequence in the presence of the serum response factor (SRF), a dimeric glyco-protein recognizing a CArG box adjacent to the ets element (
11
,
12
). This results in the formation of a ternary complex consisting of SAP-1 (or ELK-1), SRF and DNA (
13
,
14
). The SRE is a target for at least two signal transduction pathways, one which
includes p21
RAS
and MAP kinases and results in the phosphorylation of SAP-1 and ELK-1 (
12
), and one which directly affects SRF through an unknown mechanism (
10
,
15
).
The Ewing's sarcoma associated fusion protein EWS-FLI-1 retains both the ETS domain and the transcriptional activity of
FLI-1 (
16
,
17
). EWS-FLI-1, but not FLI-1, transforms NIH-3T3 cells (
18
) and induces in these cells the expression of transformation-associated genes including Stromelysin 1, cytokeratin 15 and CYP4F1 (
17
). Both the EWS domain and the FLI-1 portion of the fusion protein are required for transformation (
16
). The molecular basis for the difference between FLI-1 and EWS-FLI-1 is not fully understood. Both FLI-1 and the fusion protein localize to the nucleus (
19
). Both proteins display similar patterns of ets-box recognition (
16
). In co-transfection assays, both FLI-1 and EWS-FLI-1 display a transcriptional transactivating activity on
promoters containing ets binding sites, such as the HTLV-1 and the glyco-protein IIb promoters (
6
,
20
). In these assays, EWS-FLI-1 is a more potent transactivator than FLI-1 (
16
,
19
), which might account, at least in part, for the difference between EWS-FLI-1 and FLI-1 with regard to cell transformation. However, another
possibility is that EWS-FLI-1 and FLI-1 may recognize distinct panels of ets boxes. We show here that EWS-FLI-1, but not FLI-1, is able to form a ternary complex with
the
c-fos
SRE, and thus displays a ternary complex factor (TCF) activity, as previously
observed for ELK-1 and SAP-1. Both FLI-1 and EWS-FLI-1 are able to associate with the SRF protein
in vitro
, in the absence of DNA. However, EWS-FLI-1 binds to the ets-box of c-
fos
SRE in an autonomous manner in the absence of SRF, whereas binding of FLI-1 under the same conditions is barely detectable. We have analysed a
deletion mutant of FLI-1, FLI-1(C) which is reduced to the C-terminal portion of the protein present in EWS-FLI-1. FLI-1(C) forms a ternary complex with SRF on the
SRE and binds to the ets-box in an autonomous manner. This result suggests the existence of an
inhibitory domain in the N-terminal part of the normal FLI-1 protein. Taken together, these data suggest that the loss of the N-terminal part of FLI-1 in EWS-FLI-1 modifies DNA binding-site selection by the fusion product.
A 2993 bp EWS-FLI-1 1 cDNA (EF11), lacking the initiation codon but including the stop
codon and the 3' untranslated sequence, was isolated from a Ewing's sarcoma cDNA library
(
21
). A
Sal
I site and a Kozak consensus translation initiation sequence were inserted at
the 5'-end of the EF11 cDNA by PCR amplification, using as forward primer:
5'-TACAAAGTCGACCACCATGGCGTCCACGGATTACAGTACC-3'; the reverse primer, 5'- ATCTTAGAGCTCTAGTAGTAGCTGCCTA-3', included a
Sac
I site and annealed to the 3'-end of the translated sequence. The amplification product (~1.6 kb) was subcloned between the
Sal
I and
Sac
I sites in the pSP64 poly(A) expression vector (Promega), giving the pSP64 EWS-FLI-1 plasmid.
The plasmid pCR3 FLI-1, which allows the
in vitro
translation of FLI-1 protein to high levels, was a gift of M. Duterque (Lille, France).
The [Delta]EB-78 FLI-1(C) vector, encoding for a truncated FLI-1 protein limited to C-terminal amino acid residues 225-452 of FLI-1 and conserving the ets binding
domain, was described elsewhere and was a gift of J. Ghysdael (Orsay, France) (
16
). The FLI-1(C) cDNA was subcloned into the pSP64 vector.
pT7 ELK-1 (
12
) was a gift of R. Treisman (London, UK). pCDNA3 SRF was constructed by
subcloning the SRF coding sequence from pG3.5 (also from R. Treisman) into
pCDNA3 (Invitrogen). The SRFcore cDNA, encoding amino acid residues 131-266 of pG3.5 SRF and conserving the DNA binding domain and the ets
protein interaction domain, was obtained by PCR amplification, using as forward
primer 5'-CGGAAGCTTACCATGGTGAGCGGGGCCAA-3', and as reverse primer 5'-TCACAGGTTGGTGACTGTGAACGCCGGC-3'. The amplification product
was subcloned between the
Hin
dIII and
Hin
cIII sites in the HIV vector.
The ectodomain of the TM envelope glyco-protein of the HTLV-I retrovirus, subcloned in the pCDNA3 vector, and used as a negative
control in this study, was a gift from A. Rosenberg (Villejuif, France).
Proteins were translated
in vitro
using a TnT kit, as recommended by the manufacturer (Promega). Experiments were
run in parallel, in the presence or absence of
35
S-methionine. Radioactive products were analysed by SDS-PAGE using standard procedures, and non-radioactive products were used in EMSA.
The GpIIb, wtETS SRE and mETS SRE oligonucleotides, described in Figure
1
, were purified on denaturing acrylamide gels, annealed and end-labeled using T4 polynucleotide kinase (Biolabs) and [[gamma]-
32
P]ATP.
In vitro
translated proteins (2 [mu]l for SRF, 5 [mu]l for the others) were preincubated in 20 [mu]l of EMSA buffer (188 mM NaCl, 50 mM HEPES pH 7.9, 2.5 mM EDTA pH 8, 2.5 mM DTT, 12% glycerol) with 1-2 [mu]g salmon sperm DNA, together with an excess of unlabeled
competitor oligonucleotide (where indicated). After 15 min on ice,
32
P-labeled oligonucleotide probes (2 ng, corresponding to ~9 * 10
3
c.p.m.) were added and the incubation was allowed to continue for an additional 15 min at room temperature. Samples were electrophoresed at room temperature on a 4%
polyacrylamide gel in 0.25* TBE buffer.
GST or GST-SRF coated beads were prepared according to Groisman
et al
. (
22
). GST or SRF-GST beads were incubated with equivalent amounts of
in vitro
translated
35
S-methionine-labeled EWS-FLI-1, FLI-1, FLI-1(C), ELK-1 or HTLV-1 TM proteins, in a final volume
of 200 [mu]l of either NETN buffer (20 mM Tris pH 8, 100 mM NaCl, 1 mM EDTA, 0.5% NP-40) or ZBU buffer (25 mM HEPES pH 7.5, 12.5 mM MgCl
2
, 20% glycerol, 0.1% NP-40, 150 mM KCl, 0.5 M urea). After 1 h incubation at room temperature with
gentle agitation, beads were washed five times in NETN buffer.
Physical interaction between EWS-FLI-1 or FLI-1 and SRF was tested in the absence of DNA, using a GST pull-down assay. FLI-1 or EWS-FLI-1, as well as ELK-1 as a positive control, or the
unrelated protein TM as a negative control, were incubated with beads coated
with GST-SRF or GST alone. Results shown in Figure
3
, indicate that significant amounts of EWS-FLI-1 and FLI-1 were specifically retained on the SRF-coated beads (lanes 3 and 6), suggesting that both FLI-1 and EWS-FLI-1 interact with SRF. In control
experiments, similar amounts of the TM protein did not show any detectable
specific absorption on the SRF beads (lane 12). Furthermore, ethidium bromide
did not inhibit the retention of EWS-FLI-1, ruling out the possibility of an artefactual result due to non-specific binding of the proteins to contaminating DNA (data
not shown). Although this technique has some limitations (in particular a large
excess of SRF on the beads allows the detection of low-affinity binding proteins), these results suggest that both EWS-FLI-1 and FLI-1 bind to SRF, similarly to what is observed with ELK-1 (lane 9). The interaction between SRF and other
members of the ETS family requires a protein domain referred to as the B-box (
8
,
25
,
26
). Comparison of the amino-acid sequences of FLI-1 and the B-boxes of several known TCFs showed only a short stretch of
homology (not shown). The involvement of these amino-acids is currently under investigation.
Thus EWS-FLI-1 and FLI-1 can both interact with SRF, whereas only the chimeric
protein can form a ternary complex on the SRE ets box. This suggests that
interaction with SRF is not sufficient for ternary complex formation.
We next investigated
c-fos
ets box recognition by FLI-1 and EWS-FLI-1.
In vitro
translated proteins were analysed by EMSA in the absence of SRF, using
oligonucleotide probes containing the ets box from either the GpIIb promoter, as a reference, or the c-
fos
SRE. Significant amounts of wtETS SRE probe were retarded by EWS-FLI-1, comparable to that observed with ELK-1 (Fig.
4
A, compare lanes 4 and 6), whereas binding of FLI-1 was hardly detectable (Fig.
4
A, lane 2). The binding of EWS-FLI-1 was specific since it was inhibited by increasing amounts of the autologous
oligonucleotide (Fig.
4
B) but not by an oligonucleotide in which the ets box is mutated (mETS SRE; data
not shown). Furthermore, the complexes were not observed when the mutant
oligonucleotide was used as a probe (Fig.
4
C). Such an autonomous binding could also be detected in the presence of SRF
(see arrow `EWS-FLI-1' in Fig.
2
A), although most of the fusion protein was revealed in ternary complex under
these conditions.
The quantitative difference in the amount of probe (wtETS SRE) retarded by EWS-FLI-1 and FLI-1 (Fig.
4
A, lanes 2 and 4) was not due to a difference in the protein amounts used in
these experiments. Indeed (Fig.
4
D), both FLI-1 and EWS-FLI-1 retarded similar amounts of a GpIIb probe (described in Fig.
1
A), which is expected since the two proteins display a similar affinity for this
sequence (
16
). In addition, FLI-1 and EWS-FLI-1 proteins were repeatedly translated with similar
efficiencies, as assessed by quantification of
in vitro
translated
35
S-labeled products (Fig.
4
E) using a Bas 1000 Phosphoimager. Quantification of the amount of SRE probe (Fig.
4
A) retained by EWS-FLI-1 and FLI-1 proteins showed that EWS-FLI-1 retained 10 times more probe than FLI-1 did.
It therefore appears that a region outside the ETS domain (which is identical in
the two proteins) may be critical for binding affinity to the c-
fos
ets sequence. For example, an inhibitory domain could be present in the FLI-1 protein, which could be masked or deleted in the fusion product EWS-FLI-1. In order to test this hypothesis, we analysed next a
deletion mutant of FLI-1, FLI-1(C) which is reduced to the portion of the protein present in the
fusion product, EWS-FLI-1. FLI-1(C) was tested by EMSA in the presence or absence of SRF (in
order to increase the resolution of potential ternary complexes with FLI-1(C), a short polypeptide, a short version of SRF, the SRF `core' (
11
), was used instead of the full-length SRF). Results (Fig.
5
) indicate that FLI-1(C) behaves like the fusion product EWS-FLI-1. Like EWS-FLI-1 (Fig.
5
A, lane 4) and in contrast to FLI-1 (lane 2), FLI-1(C) induces a super shift of the core SRF-DNA complex (lane 3). Furthermore, FLI-1(C) binds to c-
fos
ets box in an autonomous manner, i.e. in the absence of SRF (Fig.
5
, lane 5) as did EWS-FLI-1 (lane 6) but not FLI-1 (lane 4), and this autonomous binding of FLI-1(C) is completely abrogated when the SRE probe is
modified by substitutions in the ets sequence (mETS SRE, Fig.
5
B, lane 8). The difference between FLI-1 and FLI-1(C) was not observed when the GpIIb
ets
sequence was used as a probe (Fig.
5
B, lanes 1-3). The amount of probe retained by FLI-1(C) in the presence of SRF was larger than that retained in the
absence of SRF, suggesting a cooperative effect of SRF on probe recognition by
FLI-1(C) This was also true for EWS-FLI-1. Finally, FLI-1 and FLI-1(C) display identical SRF-binding capacities in a GST-pull down assay (Fig.
5
C), indicating that the region of the FLI-1 molecule responsible for the interaction with SRF is located in the C-terminal part of the molecule.
Taken together, the data show that,
in vitro
, EWS-FLI-1 recognizes target sequences distinct from those specifically bound
by FLI-1 protein and that region ouside the ETS-domain may be responsible for this difference. In conclusion, EWS-FLI-1 and FLI-1 can be distinguished not only on the basis of
quantitative differences in their transactivation capability (
16
), but also qualitatively, by the target sequences which they are able to
recognize. This suggests that the tumor-specific EWS-FLI-1 protein may have the potential to regulate the expression of
various genes whose promoters cannot be bound by FLI-1. The functional relevance of this observation is currently under
investigation.
We thank R. Treisman, M. Duterque, A. Rosenberg and J. Ghysdael for kindly
providing us with plasmid gifts, and J. Ghysdael, M. C. Dokhélar and L. Pritchard for helpful discussions. This work was supported by
grants from the Association pour la Recherche sur le Cancer, from the Ligue
Nationale Contre le Cancer and from the Groupement des Entreprises Francaises
dans la Lutte conntre le Cancer. L.M.-J. was a recipient of a fellowship from the Ligue Nationale Contre le
Cancer and is currently supported by the Association pour la Recherche sur le
Cancer. H.M. was supported by the Centre National de la Recherche Scientifique
and is a recipient of a travel award from the Ryoichi Naito Foundation for
Medical Research.
REFERENCES
Return