MOLECULAR AND CELLULAR BIOLOGY, Dec. 1993, p. 7393-7398
Copyright © 1993, American Society for Microbiology
Vol. 13, No. 12
The Ewing's Sarcoma EWS/FLI-1 Fusion Gene Encodes a
More Potent Transcriptional Activator and Is a More
Powerful Transforming Gene than FLI-i
WILLIAM A. MAY,1 STEPHEN L. LESSNICK,2 BENJAMIN S. BRAUN,2 MICHAEL KLEMSZ,324
BRIAN C. LEWIS,' LYNN B. LUNSFORD,1 ROBERT HROMAS,4'5
AND CHRISTOPHER T. DENNY .2*
Department ofPediatrics, Division ofHematology/Oncology, Gwynne Hazen Cherry Memorial Laboratories,
and Jonsson Comprehensive Cancer Center, School ofMedicine,1and Molecular Biology Institute,2
University of California-Los Angeles, Los Angeles, California 90024, and Department of
Microbiology and Immunology,3 Division ofHematology/Oncology,5 and Walther
Oncology Center,4 Indiana University Medical Center,
Indianapolis, Indiana 46202
Received 12 July 1993/Returned for modification 18 August 1993/Accepted 8 September 1993
EWS/FLI-1 is a chimeric protein formed by a tumor-specific 11;22 translocation found in both Ewing's
sarcoma and primitive neuroectodermal tumor of childhood. EWS/FLI-I has been shown to be a potent
transforming gene, suggesting that it plays an important role in the genesis of these human tumors. We now
demonstrate that EWS/FLI-1 has the characteristics of an aberrant transcription factor. Subcellular
fractionation experiments localized the EWS/FLI-1 protein to the nucleus of primitive neuroectodermal tumor
cells. EWS/FLI-1 specifically bound in vitro an ets-2 consensus sequence similarly to normal FLI-1. When
coupled to a GALA DNA-binding domain, the amino-terminal EWS/FLI-1 region was a much more potent
transcriptional activator than the corresponding amino-terminal domain of FLI-1. Finally, EWS/FLI-1
efficiently transformed NIH 3T3 cells, but FLI-1 did not. These data suggest that EWS/FLI-1, functioning as
a transcription factor, leads to a phenotype dramatically different from that of cells expressing FLI-1.
EWS/FLI-1 could disrupt normal growth and differentiation either by more efficiently activating FLI-1 target
genes or by inappropriately modulating genes normally not responsive to FLI-).
Aberrant expression and structural alteration of transcrip-
tion factors are frequent, primary molecular mechanisms in
oncogenesis (11). In lower mammals and avian species,
these alterations are often mediated by retroviral insertion.
In humans, deregulation or structural alteration of transcrip-
tion factors is frequently the result of somatic genomic
The 11;22 chromosomal translocation found in Ewing's
sarcoma and primitive neuroectodermal tumor of childhood
(PNET) juxtaposes the 5' sequences from a newly described
gene, termedEWS, with the 3' sequences from FLI-1, which
encodes a member of the Ets transcription factor family (6).
Like most Ets family members (for a review, see reference
27), the carboxyl domain of FLI-1 mediates sequence-
specific DNA binding. The FLI-1 amino terminus contains a
putative transcription activation domain that may interact on
a protein-protein level with other transcription factors. As
with other Ets proteins, it is probably the combination of
protein-DNA and protein-protein binding specificities that
determines which genes are transcriptionally modulated by
FLI-1. As a result of the 11;22 rearrangement, the amino-
terminal domain of FLI-1 is replaced by a portion of EWS
containing a series of degenerate, glutamine-rich repeats.
The carboxyl terminus of EWS has amino acid similarity to
proteins involved in RNA synthesis and processing (6).
However, the function in tumor cells of the EWS amino-
terminal domain that is fused to FLI-1 is unknown.
We have recently demonstrated that the EWS/FLI-1 chi-
mera is a potent transforming gene in NIH 3T3 cells (14). We
have also shown that both EWS and FLI sequences are
necessary for the transforming activity of this chimeric
oncogene. These data suggest that EWS/FLI-1 may function
as an aberrant transcription factor contributing to transfor-
mation in Ewing's sarcoma and PNET cells. The work
presented here supports this hypothesis by demonstrating
that (i) EWS/FLI-1 localizes to the cell nucleus, (ii) EWS/
FLI-1 is able to bind DNA in a sequence-specific manner,
and (iii) the EWS region can serve as a transcription activa-
tion domain that is more potent than the corresponding
region of normal FLI-1. We also show that the EWS/FLI-1
fusion, coupling the activation domain of EWS to the DNA-
binding domain of FLI-1, transforms rodent fibroblast cells
much more effectively than wild-type FLI-1 does.
MATERIALS AND METHODS
Subcellular localization. Subcellular localization of EWS/
FLI-1 protein was performed by modification of previously
published protocols (29), using TC-32, a PNET-derived cell
line containing the t(11;22) rearrangement (31), and HeLa
cells. Briefly, cells grown in Dulbecco's minimal essential
medium plus 10% calf serum were metabolically labeled with
[35S]methionine, washed with phosphate-buffered saline
(PBS), subjected to hypotonic lysis in 10 mM Tris (pH 8)-i
mM MgCl2-0.1 mM phenylmethylsulfonyl fluoride, and
sheared through a 26-gauge needle. Nuclear and cytoplasmic
fractions were separated as pellet and supernatant by low-
speed centrifugation (2,000 rpm, 700 x g). Nuclei were
washed and then resuspended in hypotonic lysis buffer.
MAY ET AL.
Staining with both hematoxylin-eosin and vimentin demon-
strated less than 10% whole cell contamination in the nuclear
fraction. Triton X-100 and sodium dodecyl sulfate (SDS)
were added to both nuclear and cytoplasmic fractions to final
concentrations of 1 and 0.1%, respectively. Two-cycle im-
munoprecipitations were performed on both cytoplasmic
and nuclear fractions, using rabbit antisera raised to an
EWS/FLI-1 fusion polypeptide, as previously described
(14). Precipitated proteins were fractionated by SDS-poly-
Gel shift assay. In vitro-translated proteins were made by
using a rabbit reticulocyte lysate as recommended by the
manufacturer (Promega). The constructs used for these in
vitro translations have been described elsewhere (14). Pro-
tein production was confirmed by SDS-PAGE and autora-
diography of 35S-labeled proteins. The gel shift probes (see
Fig. 2 for sequence) were excised from pBluescript II KS+
(Stratagene) with XbaI and HindIll and gel purified. The
probe was labeled by Klenow filling in the presence of
[32P]dATP. Cold competitors were isolated in a similar
fashion and left unlabeled.
DNA binding reactions were carried out for 15 min at
room temperature in a final volume of 20 ml containing
10,000 cpm of labeled probe in a final buffer concentration of
acid (HEPES; pH 7.9)-75 mM NaCl-0.5 mM EDTA-1 mM
dithiothreitol-5% glycerol, using 2 jig of poly(dI-dC) (Phar-
macia). Two microliters of in vitro-translated protein was
used, and a 100-fold excess of cold competitor was added in
indicated lanes. Samples were loaded onto a prerun (140 V
for 30 min) 6% polyacrylamide gel (acrylamide/bisacryla-
ratio of 30:1) prepared with 0.25 x
EDTA-5% glycerol and subjected to electrophoresis at 140
V at room temperature. The gels were dried, and autoradiog-
raphy was performed by using two sheets of Kodak XAR
film at -70°C with an intensifying screen.
Transformation assay. The cDNA for human FLI-1 was
modified by the addition of a nine-amino-acid epitope from
influenza virus hemagglutinin to the amino-terminal end via
polymerase chain reaction amplification as previously de-
scribed (14). The modified cDNA was blunted with Klenow
and cloned into the blunted EcoRI site of the retroviral
vector SRaMSV tk neo(AHindIII) (19). The EWS/FLI-1
cDNA constructed with the same epitope as previously
described (14) and was also cloned into the EcoRI site of
SRaMSV tk neo-(AHindIII). Replication-deficient retroviral
stocks were created by transiently transfecting COS cells
with EWSIFLI-1 or FLI-I constructs together with a psi-
minus packaging plasmid (19) by either calcium phosphate
precipitation or electroporation. Conditioned medium con-
taining virus was harvested and used to infect NIH 3T3 cells.
These cells were then selected in G418 for 1 week. Selected
populations were plated in soft agar at either 5,000 or 50,000
cells per 6-cm-diameter plate and at either 10 or 20% fetal
calf serum in Iscove's medium as previously described (14).
Agar plates were photographed approximately 3 weeks after
plating. The levels of EWS/FLI and FLI proteins in these
selected cell lines were determined by using two cycles of
immunoprecipitation with polyclonal antisera as previously
GALA constructs. Constructs fusing amino acids 1 to 147 of
the yeast transcription factor GAILA to domains of interest
were constructed as follows.
(i) GAL4/EWS. The 900-bp PstI-PvuII fragment from
EWS/FLI-1 including nearly all of the EWS domain and a
portion of the FLI-1 domain but excluding nearly all of the
consensus DNA binding sequence was selected for fusion to
GAL4 (see Fig. 4A). It was initially subcloned into pBlue-
script II KS+ (Stratagene). From this intermediate, two
fragments were gel purified: a 5' 450-bp BamHI-AvrII frag-
ment and a 3' 430-bp AvrII-Asp 718 fragment. The EWS
fragment was fused to GAL4 in a trimolecular ligation of
these two fragments to the GAL4 fusion vector pSG424 (32)
cut with BamHI and Asp 718. The structure of the construct
was confirmed with analytic digests.
(ii) GAL4/FLI-1. The 800-bp 5' MscI-PvuII fragment from
FLI-1 (see Fig. 4A) was cloned into the SmaI site of pUC19
in the proper orientation. The integrity of the 5' junction was
confirmed by sequencing. From this intermediate, the insert
was excised with BamHI and Asp 718 and cloned into
pSG424 cut with the same enzymes.
Reporter gene assays. Approximately 106 HeLa cells were
transfected by overlaying them with a mixture of 50 jil of
Lipofectin (Gibco-BRL) plus 20 jig of total DNA (5 jig of
GNE4tkCAT [courtesy of Mike Carey] with 2 jig of cyto-
megalovirus luciferase [courtesy of Tina Rhodes] and 8 jig of
pBluescript II KS+) in serum-free Dulbecco's minimal es-
sential medium. The cells were fed after 5 h to bring the
serum concentration to 10% and harvested 48 h later. Cells
were harvested with trypsin and washed in 1x PBS. Fifteen
percent of each transfection was set aside and assayed for
luciferase activity by using a commercially available kit and
The remaining cells were resuspended in 0.25 M Tris (pH
7.8) and lysed by three consecutive freeze-thaw cycles. The
amount of lysate used was standardized by reference to the
luciferase values. Chloramphenicol acetyltransferase (CAT)
activity was assayed by incubation for 1 h at 37°C with 0.5 ml
of [14C]chloramphenicol (CAT assay grade; NEN) and 10 ,ul
of acetyl coenzyme A in 0.25 M Tris (3.5 mg of lithium salt
per ml; Sigma) in a final reaction volume of 70 ml of 0.25 M
Tris (pH 7.8). Control reactions using 1 U of purified CAT
enzyme (Pharmacia) were performed under the same condi-
tions. Reactions were extracted with ethyl acetate, dried,
and fractionated by thin-layer chromatography, using a 95:5
mixture of chloroform-methanol. Autoradiography was per-
formed overnight, and bands were cut out and quantitated in
5 ml of scintillation fluid (Beckman).
of reporter plasmid
EWS/FLI-1 localizes to the nucleus. Although FLI-1 is a
known DNA-binding protein, the biologic function of EWS
is unknown. The EWS/FLI-1 fusion protein could therefore
be involved in a variety of cellular functions. To begin to
determine the activity of EWS/FLI-1, subcellular localiza-
tion analyses were performed. TC-32 cells, a PNET cell line
containing the 11;22 translocation (31), and HeLa cells,
which do not contain the 11;22 translocation and thus do not
express the EWS/FLI-I fusion gene, were both metabolically
labeled with [35S]methionine. Nuclei were separated from
cytoplasm and membranes by hypotonic lysis and low-speed
centrifugation. Two-cycle immunoprecipitation experiments
were then performed on both nuclear and cytoplasmic/
membrane fractions, using an antiserum raised to a polypep-
tide from the TC-32 EWS/FLI-1 junction point (14). This
polypeptide fragment included EWS amino acid residues 244
to 264 (6) fused to residues 241 to 284 of FLI-1 (30).
A 68-kDa band corresponding to the EWS/FLI-1 fusion
protein was consistently present in the nuclear fraction of
MOL. CELL. BIOL.
EWING'S SARCOMA EWSIFLI-1 FUSION GENE
FIG. 1. Subcellular localization of EWS/FLI-1. Two-cycle im-
munoprecipitation of fractionated protein extracts with a polyclonal
antiserum raised to an EWS/FLI-1 polypeptide fragment demon-
strated that nearly all of the EWS/FLI-1 protein was present in the
nuclear (N) fraction of TC-32 cells, a PNET cell line containing the
11;22 translocation. The additional 90- and 100 kDa bands present in
TC-32 and also in t(11;22)-negative HeLa cells may represent
unrearranged EWS proteins (see text). C, cytoplasmic fraction.
TC-32 cells but absent in the cytoplasmic/membrane fraction
(Fig. 1). This band comigrated with in vitro-translated EWS!
FLI-1 cDNA as well as immunoprecipitated protein from
NIH 3T3 cells transformed by EWS/FLI-1 (14). Additional
bands of 90 and 100 kDa were also detected in TC-32 and
HeLa cells. It is likely that at least in part, these represent
germ line EWS proteins. In vitro-translated EWS cDNA
generated a protein that comigrated with the 90 kDa band
and that immunoprecipitated with our EWS/FLI-1 antiserum
(data not shown). Together, these data indicate that the
EWS/FLI-1 fusion protein localizes to the cell nucleus.
EWS/IFU-1 binds specific DNA sequence. Many transcrip-
tion factors bind DNA in
Though there is considerable variation in EWS/FLI-1 fusion
products from different translocation-positive tumors, in all
instances the DNA-binding domain of FLI-1 remains intact
(6). We previously showed that if the DNA-binding domain
is disrupted, EWS/FLI no longer can transform NIH 3T3
cells (14). This result suggests that EWS/FLI-1 retains the
ability to bind DNA and that its binding specificity might be
similar to that of FLI-1.
The natural target genes of FLI-1 are unknown. FLI-1 is
able to specifically bind to an ets-2 consensus sequence but
not a PU-llspi-i sequence (10). To determine whether EWS/
FLI-1 and FLI-1 share the same specificity for these DNA
targets, gel shift analyses were performed (Fig. 2). In vitro-
translated full-length FLI-1 (50 kDa) and EWS/FLI-1 (68
kDa) proteins were assayed for the ability to specifically bind
a labeled ets-2 consensus oligonucleotide. Approximately
equal amounts of eachprotein,as measured by SDS-PAGE,
were incubated with 3 P-labeled ets-2 oligonucleotide and
then fractionated on nondenaturing gels. Protein-DNA com-
plexes were detected for both FLI-1 and EWS/FLI-1 pro-
teins (Fig. 2, lanes 1 and 4). Binding was greatly attenuated
by a molar excess of unlabeled ets-2 oligonucleotide (lanes 2
a sequence-specific manner.
ETS-2 - GAGACCGGAAGTGGGG
FIG. 2. Sequence-specific DNA binding by EWS/FLI-1 and
FLI-1 proteins. In vitro-translated EWS/FLI-1 and FLI-1 proteins
form stable complexes with 32P-labeled ets-2 oligonucleotide. La-
beled DNA-protein complexes are attenuated by molar excess of
unlabeled ets-2 probe but not by PU-1/spi-1 oligonucleotide. EWS/
FLI-1 A22, a nontransforming mutant that lacks most of the EWS
domain, also binds to the ets-2 probe in a similar fashion. However,
EWS/FLI-1 AETS, which lacks the FLI-1 DNA-binding domain,
fails to bind labeled ets-2 oligonucleotide. The background band
visible in lanes 1, 4, 7, and 10 is probably due to endogenous Ets
proteins present in the rabbit reticulocyte lysate since it is also seen
in a mock in vitro translation mix lacking input RNA (lane 13).
and 5) but not by PU-llspi-J oligonucleotide (lanes 3 and 6),
demonstrating that for both EWS/FLI-1 and FLI-1 proteins,
these protein-DNA interactions are sequence specific.
Two mutant EWS/FLI-1 constructs that were unable to
induce transformation in 3T3 cells were also tested for the
ability to bind ets-2 sites. Mutant EWS/FLI-1 AETS contains
a 54-amino-acid deletion within the FLI-1 DNA-binding
domain (14) and was unable to bind ets-2 oligonucleotide
(lanes 10 to 12). Only a background band also present in the
in vitro translation mix lacking input RNA (lane 13) was
visible. This result directly confirms that sequence-specific
DNA binding is essential to the transforming activity of
EWS/FLI-1. In contrast, the mutant EWS/FLI-1 A22, which
lacks nearly all of the EWS sequences, was able to specifi-
cally bind the ets-2 probe (lanes 7 to 9). These data suggest
that the inability of the A22 mutant to transform cells is due
to the loss of biochemical function other than sequence-
specific ets binding.
EWS/FLI-1 transforms NIH 3T3 cells, but FLI-1 does not.
The EWS/FLI-1 fusion gene is a potent single-step trans-
forming gene in NIH 3T3 cells (14). Since FLI-1 and EWS/
FLI-1 have similar DNA-binding characteristics, we wished
to determine whether FLI-1 could also transform NIH 3T3
Low-passage NIH 3T3 cells were infected with recombi-
nant retrovirus containing either EWSIFLI-1 or FLI-I under
the control of the retroviral long terminal repeat promoter.
Pure polyclonal populations selected by growth in G418-
containing medium were plated into soft agar at a variety of
seeding densities and serum concentrations. In seven inde-
pendent infections and under all plating conditions, EWS/
FLI-I-infected 3T3 cells consistently formed numerous mac-
roscopic colonies in agar after 8 to 10 days (Fig. 3). Three
independent infections with FLI-I retroviruses produced no
VOL. 13, 1993
MAY ET AL.
FIG. 3. Agar assays demonstrating transformation of NIH 3T3
cells by EWS/FLI-1 but not FLI-1. A G418-selected, polyclonal
population of 3T3 cells infected with FLI-1-containing retrovirus
failed to form colonies in soft agar. Clonal or polyclonal populations
of NIH 3T3 cells expressing the EWS/FLI-1 fusion both formed
macroscopic colonies in agar at equivalent efficiencies. For these
plates, a polyclonal population 3T3 cells expressing FLI-1 was
compared with a clonal EWS/FLI-J-expressing 3T3 cell line. Cells
were seeded into agar at a density of 50,000 cells per plate and at a
serum concentration of 10%.
visible colonies in agar under any conditions, even after
incubation for more than 4 weeks.
Expression of EWS/FLI-1 and FLI-1 proteins was docu-
mented by immunoprecipitation from [35S]methionine meta-
bolically labeled, infected 3T3 cell populations by using an
EWS/FLI-1 antiserum (see above). EWS/FLI-1 and FLI-1
were present in 3T3 cells at approximately equal levels (data
not shown). The difference in the transforming activity
between EWS/FLI-1 and FLI-1 in NIH 3T3 cells is not due
to disparate levels of intracellular protein but reflects differ-
ences in biologic function of these two proteins.
The EWS domain functions as a potent transcriptional
activator. The amino-terminal domains of most Ets family
members are thought to function as transcriptional activa-
tors. The principle difference between EWS/FLI-1 and
FLI-1 is the substitution of the amino-terminal domain of
EWS for the putative transcriptional activation domain of
FLI-1. This substitution, which converts a nontransforming
gene (FLI-1) into a transforming gene (EWS/FLI-1), might
also alter the transcriptional activation properties of the
amino-terminal domains of these two proteins.
To test this hypothesis, reporter gene experiments were
performed in HeLa cells. FLI-1 has been shown to activate,
at low levels, constructs containing concatemerized ets-
binding sites upstream of an attenuated promoter and re-
porter gene (10, 33). We encountered high levels of nonspe-
cific background with use of similar reporter constructs
containing concatemerized ets-2-binding sites. To circum-
vent these problems, we used a reporter gene system based
on the yeast GAL4 gene. This system has been successfully
applied to quantitating the transactivation properties ofother
mammalian transcription factors, including members of the
Ets family (3, 22).
The DNA-binding domain of GAL4 was fused to the
amino termini of either EWS/FLI-1 or FLI-1, truncated at
the same point within the common FLI-1 DNA-binding
domain (Fig. 4A). These constructs were transfected into
HeLa cells, together with a reporter construct containing
either one or five GAL4 binding sites upstream of a minimal
promoter driving a CAT gene. Also included was a control
plasmid consisting of a cytomegalovirus-driven luciferase
gene. Protein lysates from these transfections were normal-
ized by luciferase activity for transfection efficiency and
assayed for CAT activity. These experiments were per-
formed in triplicate at various DNA concentrations, all with
similar results (Fig. 4B). GAL4-EWS/FLI-1 constructs effi-
ciently activated reporter constructs containing either one or
five GALA binding sites. At identical conditions, low-level
CAT activity was detected with use of the FLI-1 amino-
terminal region. GAL4 fusion constructs containing the
EWS amino terminus
greater CAT activity than those containing the correspond-
ing FLI-1 amino-terminal domains (Fig. 4C). Immunoprecip-
itation from extracts ofHeLa cells transfected with the same
FLI-1 expression plasmid demonstrated stable expression of
GALA/FLI-1 protein (data not shown). These data demon-
strate that the EWS amino-terminal domain, present in the
EWS/FLI-1 fusion protein, is a much more potent transcrip-
tion activator than the amino terminus of FLI-1.
Our data suggest that EWS/FLI-1, a transforming gene
caused by a human tumor-specific chromosomal rearrange-
ment, encodes a protein that acts as an aberrant transcrip-
tion factor. We show that EWS/FLI-1 localizes to the
nucleus of t(11;22)-positive PNET cells and that it can bind
DNA in a sequence-specific manner. The fact that EWS/
FLI-1 and FLI-1 can specifically bind the same nucleotide
sequence is consistent with a modular nature for transcrip-
tion factors, since both molecules have the same DNA-
binding domain (for a review see reference 17). Finally, we
demonstrate that the EWS sequences that displace the
normal FLI-1 transactivation domain in the t(11;22) rear-
rangement can act as a potent transcriptional activation
EWS/FLI-1 appears to act as a transcription factor that
causes a phenotype that is distinct from that conferred by
FLI-1. Although FLI-1 does not transform NIH 3T3 cells,
EWS/FLI-1 does when expressed at comparable levels.
These results suggest that the oncogenic effect of the 11;22
translocation is not simply deregulating an analog of FLI-1
by placing it under the transcriptional control of the ubiqui-
tously expressed EWS gene. Rather, the structural fusion of
EWS to FLI-1 and the creation of a chimeric protein is
important for transformation activity.
There are several possible explanations for why EWS!
FLI-1 is a transforming gene but FLI-i is not. Both mole-
cules may interact with the same target genes, but EWS/
FLI-1 might be able to activate these genes much more
effectively. We have shown that EWS/FLI-1 and FLI-1 can
specifically bind the same ets-2 consensus sequence. Part of
the transforming activity of EWS/FLI-1 could be due to
enhanced binding of FLI-1 target sites. EWS/FLI-1 differs
from FLI-1 at its amino-terminal domain. The amino-termi-
nal sequences of some Ets proteins can modulate the avidity
of DNA binding (12, 28). Alternatively, EWS/FLI-1 and
FLI-1 may bind target sites with similar affinities, but
EWS/FLI-1 may be a more potent transcriptional activator.
The fact that EWS acts as a more powerful transcription
activation domain than FLI-1 in GALA reporter gene assays
supports this concept. All of these models suggest quantita-
tive differences in the ability of EWS/FLI-1 and FLI-1 to
activate a common set of target genes.
Another possibility is that EWS/FLI-1 is able to transcrip-
tionally activate target genes that FLI-1 cannot activate.
MOL. CELL. BiOL.
EWING'S SARCOMA EWSIFLI-I FUSION GENE
- - -- -- ---
FIG. 4. GAL4 reporter gene assays demonstrate that the EWS
portion of EWS/FLI-1 is a potent transcriptional activation domain.
(A) Schematic of EWS/GAL4 and FLI-1/GAL4 expression con-
structs. Expression constructs are flanked by simian virus 40 early
promoters and poly(A) signal sequence. Restriction endonuclease
sites shown: P, PstI; A, AvrII; B, BamHI; Pv, PvuII; M, MscI.
Domains depicted are as labeled and as follows: EWS domain,
white; FLI-1 domain, light cross-hatching; ets DNA-binding do-
main, grey. (B) Thin-layer chromatography autoradiogram demon-
strating efficient activation of the CAT reporter gene by an EWS/
construct. An EWS/GAL4
plasmid was transiently transfected into HeLa cells together with a
CAT reporter gene construct containing one (G1E4CAT) or five
(G5E4CAT) GAL4 DNA-binding sites. EWS/GAIA was a potent
activator of both CAT reporter constructs. Under the same condi-
tions, FLI-1/GALA induced a lower level of activation that was
enhanced in the reporter construct containing five GAL4 binding
sites. The control lane reflects the activity of 1 U of purified CAT
enzyme at the same reaction conditions. (Ac)2 Chl, diacetylated
chloramphenicol; (Ac) Chl, acetylated chloramphenicol; Chl, chlor-
amphenicol; ori, origin. (C) Histogram depicting quantitation of
CAT activity. To compare EWS/GAL4 activation with FLI-1/GAL4
activation, lysates were diluted to place peak CAT activity within
ETS proteins frequently work in conjunction with other
proteins to transcriptionally activate genes. These protein-
protein interactions often play a crucial role in defining the
specificity of Ets proteins in transcriptional regulation. Such
auxiliary proteins can be cofactors that do not bind DNA.
For example,GABPPis required together with GABPa, an
Ets protein, for activation of herpes simplex virus early
genes (25). Like GABPa, FLI-1 may require cofactors that
are not expressed in NIH 3T3 cells, while EWS/FLI-1 can
act independently of these factors. In other instances, Ets
proteins work with other DNA-binding proteins to activate
genes. The Ets protein SAP-1 forms a heteromeric complex
with serum response factor for transcriptional induction of
c-Fos (5). In this scenario, target genes would be limited to
those that had binding sites for both a particular Ets protein
and other requisite DNA-binding factors. EWS/FLI-1 may
be able to function independently of DNA-binding factors
needed by FLI-1 for transcriptional activation. In this way,
EWS/FLI-1 could activate a broader repertoire ofgenes than
FLI-1 can. The transcriptional activation domains of VP-16
and SP1 have been shown to interact directly with basal
transcription complex proteins (8, 13, 21, 24). EWS/FLI-1
may be able to function similarly.
EWS/FLI-1 is the first example of an Ets protein that is
directly involved in the genesis of a human malignancy.
However, overexpression and structural alteration of Ets
transcription factors have been implicated in cancers of
lower vertebrates. Overexpression of either FLI-1 or PU-1/
spi-J by Friend virus promotes erythroleukemia in mice (1,
18). Fusion of Ets-1 to c-Myb in E26 virus causes leukemia
in chickens (15, 16). In contrast to FLI-1, both Ets-1 and
Ets-2 transform NIH 3T3 cells (23, 26).
The formation of chimeric transcription factors is a com-
mon theme in human oncogenesis and occurs through tumor-
specific rearrangements in a variety of different cancers.
Such rearrangements were first identified in hematopoietic
malignancies. For example, the t(1;19) translocation associ-
ated with acute lymphoblastic leukemia fuses the transcrip-
tional activation domain of the E2A gene to the DNA-
binding region of the homeobox gene PBX (9, 20). The
t(15;17) translocation found in acute promeylocytic leukemia
juxtaposes the retinoic acid alpha receptor to the previously
undescribed promyelocytic leukemia gene (2, 7).
Molecular characterization of chimeric transcription fac-
tors in nonhematopoietic human tumors has only recently
been achieved. The t(12;16) translocation found in myxoid
liposarcoma is of particular interest, because it generates a
chimeric product similar to EWS/FLI-1 (4). This rearrange-
ment fuses the amino terminus of a newly described gene,
termed TLS, with the carboxyl portion of the CHOP gene, a
member of the C/EBP family of transcription factors. Like
EWS, the TLS gene is composed of a glutamine-rich amino-
terminal domain coupled to a carboxyl region that is very
similar to the EWS carboxyl terminus. This suggests that
EWS and TLS belong to the same gene family. The TLS
carboxyl-terminal domain will bind poly(A)+ RNA in vitro,
suggesting that this gene might normally be involved in RNA
The Ewing's sarcoma 11;22 and myxoid liposarcoma 12;16
the linear range of the assay. As indicated, EWS/GALA constructs
resulted in approximately 30-fold-greater CAT activity than did
FLI-1/GAL4 constructs whether a reporter with one (Gi) or five
(G5) binding sites was used.
VOL. 13, 1993
MAY ET AL.
translocations are two of the first nonhematopoietic tumor-
specific rearrangements to be molecularly characterized. It
is striking that these two different tumors both form chimeric
transcription factors, utilizing members of what appears to
be the same gene family. No hematopoietic tumor-associ-
ated rearrangements that involve genes resembling EWS or
TLS have yet been described. Whether this
rearrangement preference in the genome of the respective
cells of origin or because EWSITLS chimeric genes are most
active in sarcomas is a point of future investigation.
is due to
This work was supported by NCI grants CA12800 and CA50443.
Cancer Society grant PRTA-5. S.L.L. and B.S.B. are supported by
NIH grant GM08042. C.T.D. is supported by a grant from the
Leukemia Society of America.
We gratefully acknowledge the following individuals for their
invaluable assistance: Michael Carey for help with GAL4 reporter
gene experiments, Owen Witte and Mikhail Gishizky for assistance
in retroviral infection procedures and transformation experiments,
Poul Sorensen and Timothy Triche for help with TC-32 histochem-
ical staining, Kathleen Sakamoto and Neil Shah for assistance with
reporter gene assays, and Harvey Herschman and Steven Smale for
critical review of the manuscript.
is supported by NIAID grant AI07126 and American
1. Ben-David, Y., E. B. Giddens, K. Letwin, and A. Bernstein.
1991. Erythroleukemia induction by Friend murine leukemia
virus: insertional activation of a new member of the ets gene
family Fli-i, closely linked to c-ets-1. Genes Dev. 5:908-918.
2. Borrow, J., A. D. Goddard, D. Sheer, and E. Solomon. 1990.
Molecular analysis of acute promyelocytic leukemia breakpoint
cluster region on chromosome 17. Science 249:1577-1580.
3. Chumakov, A. M., D. L. Chen, E. A. Chumakova, and H. P.
Koeffler. 1993. Localization of the c-ets-2 transactivation do-
main. J. Virol. 67:2421-2425.
4. Crozat, A., P. Amant, N. Mandahl, and D. Ron. 1993. Fusion of
CHOP to a novel RNA-binding protein in human myxoid
liposarcoma. Nature (London) 363:640-644.
5. Dalton, S., and R. Treisman. 1992. Characterization of SAP-1, a
protein recruited by serum response factor to the c-fos serum
response element. Cell 68:597-612.
6. Delattre, O., J. Zucman, B. Plougastel, C. Desmaze, T. Melot,
M. Peter, H. Kovar, I. Joubert, P. de Jong, G. Rouleau, A.
Aurias, and G. Thomas. 1992. Gene fusion with an ETS DNA-
binding domain caused by chromosome translocation in human
tumours. Nature (London) 358:62-165.
7. de The, H., C. Chomienne, M. Lanotte, L. Degos, and A. Dejean.
1990. The t(15;17) translocation of acute promyelocytic leukae-
mia fuses the retinoic acid receptor alpha gene to a novel
transcribed locus. Nature (London) 347:558-561.
8. Hoey, T., R. 0. Weinzierl, G. Gill, J. L. Chen, B. D. Dynlacht,
and R. Tjian. 1993. Molecular cloning and functional analysis of
Drosophila TAF110 reveal properties expected of coactivators.
9. Kamps, M. P., C. Murre, X. H. Sun, and D. Baltimore. 1990. A
new homeobox gene contributes the DNA binding domain of the
t(1;19) translocation protein in pre-B ALL. Cell 60:547-555.
10. Klemsz, M. J., R. A. Maki, T. Papayannopoulou, J. Moore, and
R. Hromas. 1993. Characterization of the ets oncogene family
member, fli-1. J. Biol. Chem. 268:5769-5773.
11. Lewin, B. 1991. Oncogenic conversion by regulatory changes in
transcription factors. Cell 64:303-312.
12. Lim, F., N. Kraut, J. Frampton, and T. Graf. 1992. DNA
binding by c-Ets-1, but not v-Ets-1, is repressed by an intramo-
lecular mechanism. EMBO J. 11:643-652.
13. Lin, Y. S., and M. R. Green. 1991. Mechanism of action of an
acidic transcriptional activator in vitro. Cell 64:971-981.
14. May, W. A., M. L. Gishizsky, S. L. Lessnick, L. B. Lunsford,
B. C. Lewis, 0. Delattre, J. Zucman, G. Thomas, and C. T.
Denny. 1993. Ewing sarcoma 11;22 translocation produces a
chimeric transcription factor that requires the DNA-binding
domain encoded by FLI1 for transformation. Proc. Natl. Acad.
Sci. USA 90:5752-5756.
15. Metz, T., and T. Graf. 1991. Fusion of the nuclear oncoproteins
v-Myb and v-Ets is required for the leukemogenicity of E26
virus. Cell 66:95-105.
16. Metz, T., and T. Graf. 1991. v-myb and v-ets transform chicken
erythroid cells and cooperate both in trans and in cis to induce
distinct differentiation phenotypes. Genes Dev. 5:369-380.
17. Mitchell, P., and R. Tjian. 1989. Transcriptional regulation in
mammalian cell by sequence-specific DNA binding proteins.
18. Moreau-Gachelin, F., A. Tavitian, and P. Tambourin. 1988.
Spi-1 is a putative oncogene in virally induced murine erythro-
leukemias. Nature (London) 331:277-280.
19. Muller, A. J., J. C. Young, A.-M. Pendergast, M. Pondel, N. R.
Landau, D. R. Littman, and 0. N. Witte. 1991. BCR first exon
sequences specifically activate the BCR/ABL tyrosine kinase
oncogene of Philadelphia chromosome-positive human leuke-
mias. Mol. Cell. Biol. 11:1785-1792.
20. Nourse, J., J. D. Mellentin, N. Galili, J. Wilkinson, E. Stan-
bridge, S. D. Smith, and M. L. Cleary. 1990. Chromosomal
translocation t(1;19) results in synthesis of a homeobox fusion
mRNA that codes for a potential chimeric transcription factor.
21. Roberts, S. G. E., I. Ha, E. Maldonado, D. Reinberg, and M. R.
Green. 1993. Interaction between an acidic activator and tran-
scription factor TFIIB is required for transcriptional activation.
Nature (London) 363:741-744.
22. Seneca, S., B. Punyammalee, M. Bailly, J. Ghysdael, and M.
Crabeel. 1991. Etsl, when fused to the GAL4 DNA binding
domain, efficiently enhances galactose promoter dependent
gene expression in yeast. Oncogene 6:357-360.
23. Seth, A., D. K. Watson, D. G. Blair, and T. S. Papas. 1989.
c-ets-2 protooncogene has mitogenic activity. Proc. Natl. Acad.
Sci. USA 86:7833-7837.
24. Stringer, K. F., C. J. Ingles, and J. Greenblatt. 1990. Direct and
selective binding of an acidic transcriptional activation domain
to the TATA-box factor TFIID. Nature (London) 345:783-786.
25. Thompson, C. C., T. A. Brown, and S. L. McKnight. 1991.
Convergence of Ets- and notch-related structural motifs in a
heteromeric DNA binding complex. Science 253:762-768.
26. Topol, L. Z., A. G. Tatosyan, R. Ascione, D. M. Thompson,
D. G. Blair, I. Kola, and A. Seth. 1992. c-ets-1 protooncogene
expression alters the growth properties of immortalized rat
fibroblasts. Cancer Lett. 67:71-78.
27. Wasylyk, B., S. L. Hahn, and A. Giovane. 1993. The Ets family
of transcription factors. Eur. J. Biochem. 211:7-18.
28. Wasylyk, C., J.-P. Kerckaert, and B. Wasylyk. 1992. A novel
modulator of Ets transcription factors. Gene Dev. 6:965-974.
29. Watanabe, S. M., N. E. Rosenberg, and 0. N. Witte. 1984. A
membrane-associated, carbohydrate-modified form of the v-abl
protein that cannot be phosphorylated in vivo or in vitro. J.
30. Watson, D. K., F. E. Smyth, D. M. Thompson, J. Q. Cheng,
J. R. Testa, T. S. Papas, and A. Seth. 1992. The ERGB/Fli-1
gene: isolation and characterization of a new member of the
family of ETS transcription factors. Cell Growth Differ. 3:705-
31. Whang-Peng, J., T. J. Triche, T. Knutsen, J. Miser, S. Kao-
Shan, S. Tsai, and M. A. Israel. 1986. Cytogenetic characteriza-
tion of selected small round cell tumors of childhood. Cancer
Genet. Cytogenet. 21:185-208.
32. Wu, Y. L., S. P. Tam, and P. L. Davies. 1990. A modified CAT
expression vector with convenient cloning sites. Nucleic Acids
33. Zhang, L., V. Lemarchandel, P.-H. Romeo, Y. Ben-David, P.
Greer, and A. Bernstein. 1993. The FLI-1 proto-oncogene,
involved in erythroleukemia and Ewing's sarcoma, encodes a
transcriptional activator with DNA-binding specificities distinct
from other Ets family members. Oncogene 8:1621-1630.
MOL. CELL. BiOL.