EWS-CREB1: ARecurrentVariant Fusionin Clear Cell Sarcoma?
Association with Gastrointestinal Location and Absence of
Cristina R. Antonescu,1Khedoudja Nafa,1Neil H. Segal,2Paola Dal Cin,3and Marc Ladanyi1
Purpose: Clear cell sarcoma (CCS) usually arises in the lower extremities of young adults and
is typicallyassociated with a t(12;22) translocation resultingin the fusionof EWS (EWSR1) with
ATF1, agene encodingamemberof the cyclic AMP^responsive element bindingprotein (CREB)
family of transcription factors. CCS arising in the gastrointestinal tract is rare and its pathologic
and molecular features are not well defined.
Experimental Design: We report a novel variant fusion of EWS to CREB1, a gene at 2q32
encoding another CREB family member highly related to ATF1, detected in three women with
gastrointestinal CCS. All three cases contained an identical EWS-CREB1 fusion transcript that
was shown by reverse transcription-PCR. In two of the cases tested, EWS gene rearrangement
was also confirmed by fluorescence in situ hybridization and the EWS-CREB1genomic junction
fragments were isolatedby long-range DNAPCR.
Results: Morphologically, all three tumors lacked melanin pigmentation. By immunohisto-
were negative. Ultrastructurally, two of the cases lacked melanosomes.The melanocyte-specific
transcript of MITF was absent in two cases, and only weakly expressed in the third case.
The Affymetrix gene expression data available in one case showed lower expression of the
melanocytic genes MITF,TYR, and TYRP1, compared with four EWS-ATF1-positive CCSs of
Conclusions: EWS-CREB1 may define a novel subset of CCS that occurs preferentially in
the gastrointestinal tract and shows little or no melanocytic differentiation. Thus, evidence of
melanocytic lineage or differentiation is not a necessary feature of sarcomas with gene fusions
involving CREB family members.
Clear cell sarcoma (CCS), also known as melanoma of soft
parts, typically presents in the deep soft tissues of the lower
extremity, in close proximity to tendons, fascia, or aponeuroses.
Young adults are preferentially affected and the clinical course is
often marked by regional and distant metastases. Most CCS
show immunoreactivity for melanoma markers, such as
HMB45, and contain melanosomes. Indeed, most CCS share
a melanocytic gene expression signature with melanomas (1).
However, CCS are also genetically distinct from melanomas, as
they lack BRAF mutations (2) and show, in most cases, a
recurrent chromosomal translocation t(12;22)(q13;q12),
resulting in the fusion of the EWS gene (also known as
EWSR1) on 22q12 with the activating transcription factor-1 gene
(ATF1) on 12q13 (3–6).
Primary CCSs of the gastrointestinal tract are rare (7, 8).
Gastrointestinal CCS includes a histologic variant rich in
osteoclast-type giant cells which uniformly express S100
protein, but lack melanocytic differentiation by immunohisto-
chemistry, being negative for HMB45 and Mart-1 (9). As a
result of its rarity in the gastrointestinal tract, the differential
diagnosis of CCS in this site includes more common
mesenchymal or neuroectodermal neoplasms, such as gastro-
intestinal stromal tumors, schwannoma, carcinoid, or meta-
In this study, we report three cases of CCS of the
gastrointestinal tract showing distinctive pathologic and mo-
lecular characteristics compared with their soft tissue counter-
parts, including lack of melanocytic differentiation and the
presence of a novel EWS-CREB1 fusion transcript.
Materials and Methods
Case history patient 1.
hemicolectomy for a moderately differentiated adenocarcinoma of the
transverse colon. The hemicolectomy specimen showed a synchronous
CCS of the gastrointestinal tract in the ascending colon. Surgical
This 81-year-old woman underwent a
Human Cancer Biology
Kettering Cancer Center, NewYork, NewYork, and3Department of Pathology,
BrighamandWomen’s Hospital, Boston, Massachusetts
Received12/27/05; revised 2/7/06; accepted 3/7/06.
The costs of publicationof this article were defrayedinpartby thepaymentof page
charges.This article must therefore be hereby marked advertisement in accordance
with18 U.S.C. Section1734 solely toindicate this fact.
Note: Supplementary data for this article are available at Clinical Cancer Research
Requestsfor reprints:MarcLadanyi, Departmentof Pathology, Memorial Sloan-
Kettering CancerCenter,1275 York Avenue, NewYork, NY10021. Phone: 212-639-
6369; Fax: 212-717-3515; E-mail: email@example.com.
F2006 American Association for Cancer Research.
www.aacrjournals.org Clin Cancer Res 2006;12(18) September15, 20065356
margins were negative but there were regional lymph node metastases
of CCS. The patient was then treated with 5-fluorouracil/leucovorin for
7 months for colon carcinoma. Intra-abdominal and liver metastases
from the CCS were diagnosed after a 5-year interval and she underwent
partial hepatectomy and removal of the peritoneal implants which
provided the specimen submitted for molecular analysis. Electron
microscopy was also done.
Case history patient 2. This 42-year-old woman underwent a
segmental resection of the ileum for a primary CCS of the
gastrointestinal tract. The surgery achieved negative margins and there
was no lymph node involvement. This being a very recent case, no
significant follow-up is available.
Case history patient 3.This 42-year-old woman underwent a liver
biopsy in her work-up for the clinically presumed diagnosis of
metastatic carcinoid tumor. A segmental resection of the ileum was
done revealing a 3.7 circumferential partially obstructing lesion. In
addition, a 5.0 cm mesenteric metastatic focus was identified. No
further follow-up is available, this also being a recent case.
Immunohistochemistry and electron microscopy.
histochemical studies were done. Prediluted antibodies were used for
HMB45, A103, vimentin, CD34, Cam5.2, AE:AE3, NSE, CD56,
synaptophysin, chromogranin, SMA, and HHF35 (all from Ventana
Medical Systems, Tucson, AZ). For the other antibodies tested, we used
the following conditions: CD117 (1:500; DAKO, Carpinteria, CA),
desmin (1:50; DAKO), S100 protein (1:500; DAKO), MITF (1:200;
DAKO), and tyrosinase (1:200; Novocastra, Newcastle upon Tyne,
United Kingdom). Tissue for electron microscopy was available in case
no. 1. Representative fresh tumor was fixed in 3% formaldehyde/3%
glutaraldehyde, postfixed in 1% osmium tetroxide, embedded in epoxy
resin, and stained with uranyl acetate-lead citrate by using standard
procedures. Submitted electron microscopy prints were also available
for review in case no. 3.
Fluorescence in situ hybridization analysis for the presence of EWS
rearrangement.This was done in case nos. 1 and 2 using 50-Am
paraffin sections. Whole nuclei preparations were obtained using
standard protocols. The fluorescence in situ hybridization probes were
the LSI EWSR1 translocation probe pair (Abbott/Vysis, Downers Grove,
IL), based on a break-apart fluorescence in situ hybridization assay
Reverse transcription-PCR for EWS-ATF1 and EWS-CREB1. Frozen
tissue, collected under an Institutional Review Board–approved
protocol, was available in case no. 1. Total RNA was extracted using
1 mL of Trizol reagent (Life Technologies Inc., Gaithersburg, MD). In
case nos. 2 and 3, RNA was extracted from a representative paraffin
block of formalin-fixed tumor tissue, using the Paraffin Block RNA
Isolation Kit (Ambion, Austin, TX). In all three samples, reverse
transcription-PCR (RT-PCR) for the EWS-ATF1 fusion was attempted
first, using the forward primer EWSEx7-F1 and the reverse primer
ATF1-R1 as reported previously (5). For RT-PCR detection of the EWS-
CREB1 transcript, we used the EWSEx7-F1 forward primer with either
the CREB1ex7-REVC primer (specific for CREB1; sequence: GTACCC-
CATCGGTACCATTGT) or the consensus CREB1ex7-REVA primer
(binds both CREB1 and ATF1; sequence: TCCATCAGTGGTCTGTG-
Long-range DNA PCR. DNA was extracted from fresh-frozen
tissue using the Puregene DNA isolation Kit (Gentra Systems, Inc.,
Minneapolis, MN). PCR amplification of genomic DNA was done with
HotStart master mix (Qiagen, Valencia, CA), first with EWSEx7-F1/
CREB1-ex7REVB (5¶-GTAGTACCCGGCTGAGTGGCTG-3¶) and then
with internal primers EWS-IVS7F (5¶-GAGGCAGCTATTGCAGGC-3¶)
and CREB1-IVS6R (5¶-CAACTGAAATATCCTATGGAACA-3¶). The PCR
amplification consisted of 15 minutes at 95jC, followed by 35 cycles of
30 seconds at 95jC, 30 seconds at 60jC, and 3 minutes at 72jC; a final
extension of 10 minutes ended the reaction.
Sequencing.Direct sequencing of PCR products was done using the
Big-Dye Terminator kit (Applied Biosystems, Foster City, CA) and run
on an Applied Biosystems model 3100-Avant DNA sequencing system.
Hybridization of Affymetrix oligonucleotide chips.
tissue for RNA extraction was available in case no. 1 (CCS no. 1). RNA
was isolated using RNAwiz RNA isolation reagent (Ambion) and run
through a column with RNase-free DNase (Qiagen). Twenty-five
nanograms of total RNA were tested for quality by RNA 6000
NanoAssay on a Bioanalyzer 2100 (Agilent, Palo Alto, CA). Two
micrograms of high-quality total RNA (A260/280ratio > 1.8) was then
labeled according to the manufacturer’s instructions. Ten micrograms of
labeled and fragmented cRNA were then hybridized onto a Human
Genome U133A expression array (Affymetrix, Santa Clara, CA).
Posthybridization staining, washing, and scanning were done according
to instructions from the manufacturer (Affymetrix). The raw expression
data were derived using the Affymetrix Microarray Analysis 5.0 (MAS
5.0) software. The data were normalized using a scaling target intensity
of 500 to account for differences in the global chip intensity. The
expression values were transformed using the logarithm base two.
Affymetrix U133A gene expression data were also available in four cases
of non-gastrointestinal CCS, all four carrying an EWS-ATF1 fusion, as
previously reported in a separate study (1).
Morphology, immunohistochemical findings, and ultrastructure.
Grossly, the lesions had a white-tan cut surface with infiltrative
borders. Areas of hemorrhage and necrosis were also seen in
some tumors. The mean tumor size was 5.7 cm and ranged
from 3.7 to 7.5 cm. Microscopically, the tumors had a
multinodular and infiltrative growth pattern, being centered
in the submucosa and muscularis propria, but focally extending
into the mucosa, as well as into subserosal fat. At low power,
the tumors showed a variable growth pattern, with areas of
solid, nested, single-file, and pseudo-papillary growth noted
even within the same tumor (Fig. 1A and B). Morphologically,
the tumors displayed predominantly uniform small epithelioid
cells, with limited variation in the nuclear size and shape.
However, ovoid cell morphology and even spindling was also
noted. The nuclei showed a smooth nuclear contour, with open
and fine chromatin, and prominent nucleoli. Although the
main pattern included epithelioid cells with a minimal amount
of cytoplasm arranged in solid nests, focal areas of more
abundant cytoplasm, either clear or granular-eosinophilic, were
noted in two cases (Fig. 1C). Particularly striking was the
pseudo-papillary pattern with preservation of the tumor cells
around blood vessels, mimicking an epithelial neoplasm with
papillary architecture (Fig. 1B). No melanin pigment was
identified. Scattered multinucleated, osteoclast-type giant cells
were identified in one case (case no. 2, Fig. 1D). Mitotic activity
was high in all three cases, with a mean of eight mitotic figures
per 10 high-power fields.
The immunohistochemistry findings showed a consistent
pattern, with diffuse and strong reactivity for S100 protein in
the overwhelming majority of the tumor cells (Fig. 2A). All
three cases showed staining for neural markers, such as NSE,
CD56 and synaptophysin, either in a diffuse (case nos. 2 and 3)
or focal (case no. 1) manner (Fig. 2B). Vimentin was only
focally positive. The tumors were completely negative for all
melanoma markers, such as HMB45, A103, MITF, and
tyrosinase. All the other markers tested, including CD117,
CD34, cytokeratins, muscle markers, and chromogranin were
Ultrastructurally, the tumors examined (case nos. 1 and 3)
showed moderate amounts of cytoplasm, rich in organelles
Novel EWS-CREB1Fusion in Clear Cell Sarcoma
www.aacrjournals.org Clin Cancer Res 2006;12(18) September15, 20065357
such as mitochondria, polyribosomes, and lysosomes. A
significant number of electron-dense granules of variable sizes
and shapes were identified, but no diagnostic melanosomes
or dense-core neurosecretory-type granules were identified
(Fig. 2C). A number of primitive or simple cell junctions were
seen. Pertinent negative findings included lack of basement
membrane material, glycogen, myofilaments, and tonofilaments.
Detection of novel EWS-CREB1 fusion. Fluorescence in situ
hybridization for EWS rearrangement was done in case nos. 1
and 2 and showed multiple nuclei with split signals indicative
of EWS rearrangement in both (Fig. 3A). RT-PCR for the EWS-
ATF1 fusion was attempted in case no. 1, using the forward
primer EWSEx7-F1 and the reverse primer ATF1-R1. This
showed a weak but distinct product that differed in size from
the product obtained from the EWS-ATF1-positive control cell
line SU-CCS-1 (data not shown). Direct sequencing of the
product in case no. 1 showed a chimeric transcript consisting of
a junction between EWS exon 7 and a nucleotide sequence
similar to ATF1 exon 7, which BLAST analysis revealed to be a
perfect match with exon 7 of CREB1 (Fig. 3B). The novel EWS-
CREB1 fusion transcript was confirmed by RT-PCR using
EWSEx7-F1 with a specific reverse primer CREB1-Ex7bR
(Fig. 3C). Using the same respective primer pairs as above,
case nos. 2 and 3 were negative by RT-PCR for the EWS-ATF1
fusion, but were positive for EWS-CREB1 showing the same
fusion structure as case no. 1 (Fig. 3C and D). Thus, these three
patients provide evidence for the existence of a recurrent variant
chromosomal translocation of EWS (22q12) and CREB1
(2q32.3), presumably a t(2;22)(q32.3;q12), in CCS. Alignment
of the EWS-CREB1 fusion product with native ATF1 highlights
the extensive similarity in CREB1 and ATF1 (Fig. 4).
Analysis of genomic structure of EWS-CREB1 fusion. Detailed
genomic analyses of EWS breakpoints, available only for the
EWS-FLI1 fusions seen in Ewing’s sarcomas, have shown that
all fusions involving EWS exon 7 as the 5¶ partner arise from
genomic rearrangements within 4.2 kb downstream of EWS
Fig.1. Microscopic appearance of gastrointestinal CCS.
A, well-defined nested growth pattern (case no. 2;
no.1 ; magnification,?100). C, areas of clear cell appearance
(case no. 2; magnification,?200). D, osteoclast-type giant
cell (arrow; case no. 2; magnification,?400).
Fig. 2. Immunoprofile including (A) strong
S100 protein reactivity (case no.1, lymph
node metastasis; magnification,?200), and
(B) focal synaptophysin staining (case no.1 ;
magnification,?200). C, electron
micrograph showing moderate amount of
cytoplasm with a number of variably sized
electron-dense granules (arrows).
Diagnostic melanosomes or dense core
neurosecretory granules were not identified
(case no.1, magnification,?60,840).
Human Cancer Biology
www.aacrjournals.orgClin Cancer Res 2006;12(18) September15, 20065358
exon 7 (10). Likewise, it is likely that fusions involving CREB1
exon 7 as the 3¶ partner arise from genomic rearrangements of
CREB1 intron 6, which measures 4.95 kb. To characterize the
genomic fusion point of the EWS-CREB1 chimeric gene, we did
PCR amplification of genomic DNA from case no. 1 with the
EWSEx7-F1 and CREB1-ex7REVB primers. This revealed a
strong discrete product of about 2 kb (data not shown).
Stepwise direct sequencing of this fragment was done. PCR
amplification and DNA sequencing with intronic forward
primer EWS/IVS7F and the intronic reverse primer CREB1/
IVS6R localized the genomic fusion point to nucleotide
position g20493 of EWS intron 7 and nucleotide position
g44998 of CREB1 intron 6 (Supplementary Fig. S1A). The EWS-
CREB1 fusion gene in case no. 1 retained 1,052 bp from the 5¶
end of EWS intron 7 and 913 bp from the 3¶ end of CREB1
intron 6 resulting in the formation a chimeric intron of 1,965
bp, which corresponded with the size of the initial long-range
DNA PCR product.
In case nos. 2 and 3, only paraffin-embedded tissue was
available for DNA extraction. PCR amplification of genomic
DNA extracted from paraffin-embedded tissue was attempted in
case no. 2 only. Long-range PCR using primers EWSEx7-F1 and
Fig. 3. A, fluorescence in situ hybridization for EWS rearrangement.The probe centromeric to EWS (red); the probe on the telomeric side of EWS (green). Four nuclei
with split signals indicative of EWS rearrangement (case no.1). B, electrophoretogram of direct sequencing of junction point of EWS-CREB1fusion transcript within-frame
fusion(caseno.1).C andD,agarosegels ofEWS-CREB1RT-PCRproductsincasenos.1and2usingprimers EWSex7F1andCREB1ex7REVcandincaseno. 3usingthesame
forward primer but with reverse primer CREB1ex7REVa (consensus primer for CREB1and ATF1-see Fig. 4). Lane C is the SU-CCS-1positive control cellline containing
EWS-ATF1. No R.T. designates thenegative controls lacking reverse transcriptase.
Fig. 4. Partial alignment of EWS-CREB1
fusion withnativeATF1. Exons are indicated
by alternating upper and lower case letters.
The positions of reverse PCR primers in
CREB1and ATF1are showninitalics.
CREB1ex7REVC is a CREB1-specific primer,
whereas CREB1ex7REVA is a consensus
primer for CREB1and ATF1.
Novel EWS-CREB1Fusion in Clear Cell Sarcoma
www.aacrjournals.org Clin Cancer Res 2006;12(18) September15, 20065359
CREB1-ex7REVB primers did not yield a discrete PCR product.
Prior to embarking on a more systematic set of long-range DNA
PCR assays with regularly spaced primers in EWS intron 7 and
CREB1 intron 6, we wished to exclude the possibility of a
similar genomic breakpoint in case no. 2 as in case no. 1.
Therefore, we did a nested PCR using first the EWSEx7-F1 and
CREB1-Ex7bR primers and then the internal primers EWS/
IVS7F and CREB1/IVS6R. This amplified a strong 1 kb fragment
(data not shown). Direct sequencing of this fragment revealed
that the breakpoint was located at nucleotide position g20645
in EWS intron 7 and at nucleotide position g44921 of CREB1
intron 6 (Supplementary Fig. S1B). Thus, the EWS-CREB1
fusion gene in case no. 2 retained 1,204 bp from the 5¶ end of
EWS intron 7 and 992 bp from the 3¶ end of CREB1 intron 6,
forming a chimeric intron of 2,196 bp.
RT-PCR and microarray analysis of melanocytic differentiation
transcripts. Affymetrix U133A gene expression data were
available in case no. 1 and were compared with data from
four cases of non-gastrointestinal CCS, all four carrying the
EWS-ATF1 fusion, focusing on the level of expression of critical
genes involved in melanogenesis. As shown in Fig. 5A, case no.
1 expressed lower levels of the key melanocytic genes MITF,
TYR, and TYRP1, but SOX10 transcript levels were not
substantially different. This was consistent with its poor
expression of melanocytic proteins by immunohistochemistry.
We also examined the transcript levels of CREB1 and ATF1
(probe set 222103_at). The CREB1 transcript level was higher
in case no. 1 than in the four EWS-ATF1 cases, possibly
reflecting the detection of the EWS-CREB1 transcript by this
probe set. However, ATF1 transcript levels did not differ
between case nos. 1 and the 4 CCS with the EWS-ATF1 fusion
(data not shown).
There are several alternative forms of the MITF transcript,
reflecting different promoters and transcription start sites, only
one of which, MITF-M, is highly specific for the melanocytic
lineage (5). We have previously shown that classic soft tissue
CCS with EWS-ATF1 express the melanocyte-specific MITF-M
transcript, further supporting their genuine melanocytic differ-
entiation (5). In contrast, RT-PCR analysis for the MITF-M
transcript in our three EWS-CREB1 cases was negative in two
cases and only weakly positive in one case (Fig. 5B), the latter
being case no. 1, which also expressed the highest level of the
consensus MITF transcript among these three cases (Fig. 5B).
Together with the above immunohistochemical and ultrastruc-
tural studies, these gene expression data confirm that evidence
of melanocytic differentiation in CCS with EWS-CREB1 is
either absent or much less than in classic CCS with EWS-ATF1.
This, in turn, strengthens the notion that melanocytic features
are not a necessary finding in this type of sarcoma.
The classic cytogenetic hallmark of CCS, first described in
cases arising in somatic soft tissues, is a recurrent
t(12;22)(q13;q12) translocation, resulting in the EWS-ATF1
fusion (3, 5, 6), an alteration which was, until recently (see
below), not observed in any other tumor type. Despite this
distinctive gene fusion in CCS (absent in melanoma; refs. 4, 11)
and the absence of the BRAF mutations that are so common in
melanoma (2), immunohistochemistry and ultrastructural
studies of CCS suggest that it is, like melanoma, a neuro-
ectodermal tumor with melanocytic differentiation. Further
evidence of the genuine melanocytic differentiation in CCS has
been provided by a microarray-based gene expression profiling
study which showed that unsupervised clustering grouped CCS
with melanomas rather than with other high-grade sarcomas
(1). This study also confirmed the up-regulation in CCS of
a number of genes involved in melanocytic differentiation
(such as MITF and SOX10), relative to other soft tissue sarcoma
We have identified a novel recurrent variant fusion in CCS,
EWS-CREB1. CREB1 maps to 2q34, and like EWS, is oriented
with its 3¶ end telomeric, suggesting that both of our cases may
have contained a simple t(2:22)(q34;q12) at the cytogenetic
Fig. 5. Low or absent expression of melanocytic differentiation
genes in CCS cases with the EWS-CREB1fusion. A, Affymetrix
expressionlevels of melanocytic differentiation genes comparing
EWS-CREB1case nos.1to 4 EWS-ATF1-positive CCS cases.The
probe sets are 207233_s_at for MITF (microphthalmia-associated
transcription factor), 205694_at forTYRP1 (tyrosinase-related
protein1), 206630_at for tyrosinase (TYR), and 209842_at for
SOX10 (SRY-box10).The data for CREB1 (probe set204314_s_at)
are also shown.The values represent raw expression values
calculated byAffymetrix MAS 5.0 software (no units). B, RT-PCR
for consensus and melanocyte-specific MITF transcripts
(MITF-C and MITF-M, respectively). In case nos. 2 and 3, which
expressed little or no MITF transcripts, robust expression of a
housekeeping gene (phosphoglycerate kinase, PGK) is shown,
confirming RNA quality.
Human Cancer Biology
www.aacrjournals.org Clin Cancer Res 2006;12(18) September15, 20065360
level, but no karyotypes were available. Our finding that all
three CCS cases with this novel fusion arose in the gastroin-
testinal tract suggests that this fusion may be preferentially
associated with a gastrointestinal location, whereas non-
gastrointestinal CCS show the EWS-ATF1 fusion in at least
90% of the cases (5, 6).
ATF1, cyclic AMP (cAMP)–responsive element binding
protein (CREB1), and cAMP response element modulatory
protein constitute a subfamily of the basic leucine zipper
superfamily of transcription factors and have been implicated
in cAMP and Ca2+-induced transcriptional activation. CREB1 is
a nuclear protein that binds cAMP response elements as a
homodimer or heterodimer (with members of the ATF and AP1
transcription factor families). Overexpression of CREB tran-
scription factors contributes to the acquisition of the metastatic
potential in human melanoma cells and is also oncogenic in
the myeloid lineage (12, 13). Genome-wide screens for
promoters bound by CREB reveal a very large number of
potential target genes (f4,000) and a critical role for tissue-
specific coactivators in determining target gene activation (14).
Like other members of the basic leucine zipper superfamily,
CREB1 has a modular structure consisting of a carboxyl
terminal basic leucine zipper domain mediating DNA binding
and dimerization, and an amino terminal transactivation
domain that contains a kinase-inducible domain mediating
interactions with CBP and p300 (Fig. 6). The predicted protein
structure of EWS-CREB1 thus parallels that of EWS-ATF1
(Fig. 6). Specifically, all CCS cases with EWS-ATF1 contain
fusion transcripts in which the basic leucine zipper domain is
retained, and this is also the case for EWS-CREB1. The kinase-
inducible domain, which is either excluded or truncated in
different forms of EWS-ATF1, is not part of EWS-CREB1 in the
three current cases (Fig. 6). Thus, the structure of EWS-ATF1
and the predicted structure of EWS-CREB1 indicates that
mediating cAMP-inducible transcription through PKA-mediat-
ed phosphorylation is not a necessary feature of either fusion
protein. Interestingly, whereas the MITF-M promoter is cAMP-
inducible in the presence of SOX10 in melanoma cells through
binding of its cAMP response elements by CREB (15), EWS-
ATF1 does not seem to transactivate the MITF-M promoter in
CCS cells, at least in exogenous constructs (16). Thus, the
expression of the MITF-M transcript in CCS may represent a
feature of the precursor cell in which the translocation occurs.
Indeed, a chromatin immunoprecipitation–based screen for
EWS-ATF1 target genes in a CCS cell line did not detect MITF
(17). In fact, none of the nine putative target genes isolated in
that study were related to the melanocytic lineage. Another line
of evidence that EWS-ATF1 and MITF-M expression are not
tightly linked comes from data in angiomatoid fibrous
histiocytoma. Some angiomatoid fibrous histiocytomas seem
to contain a novel FUS-ATF1 fusion (18, 19) but there is a
recent report of a case with an EWS-ATF1 fusion, and that case
lacked expression of the melanocytic splice form of MITF (20).
The occurrence of CCS in the gastrointestinal tract is
exceptionally rare. Data from eight genetically confirmed cases
along with the three current EWS-CREB1 cases and two
additional unpublished cases from our center are summarized
in Supplementary Table S1 (7–9, 21–23). Based on these 13
cases, CCS of the gastrointestinal tract seems slightly more
prevalent in females, with a wide age distribution, 15 to 85
years, but with eight cases (62%) occurring between 30 and
51. The most common location is the small bowel (69%)
followed by colon and stomach. A common characteristic of
these tumors is the transmural involvement of the bowel wall,
often with mucosal ulceration and spreading to regional
lymph nodes. Microscopically, the predominant pattern is of
nested or solid growth of small epithelioid cells with
amphophilic or clear cytoplasm and uniform nuclei with
conspicuous nucleoli. However, certain cases contain some-
what distinctive morphologic features, due to a component of
admixed reactive, KP1-positive, and osteoclast-type giant cells
(9, 23). One of our cases showed a similar, although very
minor, component of osteoclast-type giant cells. These cells
are most likely reactive histiocytic cells, and are different
morphologically and immunohistochemically from the mul-
tinucleated tumor cells seen in conventional CCS. This latter
type of giant tumor cell was not seen in most cases of primary
gastrointestinal CCS. Due to its unusual visceral presentation,
coupled with the inconsistent expression of melanocytic
markers and variant EWS fusion partners, CCS of the
gastrointestinal tract can often be misdiagnosed. This is
evident in our experience, in which the original or second-
opinion consultation diagnoses included: poorly differentiated
carcinoma, most likely metastatic from the other synchronous
colonic primary adenocarcinoma (case no. 1); carcinoid
tumor, gastrointestinal stromal tumor, and undifferentiated
neoplasm, possibly neurogenic in origin (case no. 2); and
metastatic melanoma and carcinoid tumor (case no. 3).
Interestingly, whereas non-gastrointestinal CCS shows the
expression of melanocytic markers (HMB45, A103, MITF, etc.)
in the overwhelming majority of cases (5), the reverse seems to
be the case in the gastrointestinal tract, where most tumors
(69%) are negative for these markers (Supplementary Table
S1). In spite of the negativity for melanocytic markers, the
morphologic appearance coupled with the t(12;22) transloca-
tion supports a diagnosis of CCS in such cases. Although all
three of our EWS-CREB1-positive CCS lacked melanocytic
expression by immunohistochemistry, this was also noted in
Fig. 6. Schematic scale diagram of normal EWS, CREB1,
and ATF1proteins and corresponding fusion products.
Zn Ran BD, zinc finger RAN binding domainin EWS; pKID,
kinase-inducible domain, part of the transactivation domain
of CREB family transcription factors; bZIP, basic-leucine
zipper domain, representing the DNA-binding domain.
Scale is10 amino acid residues/segment. In the predicted
EWS-CREB1fusion protein, codon 265 of EWS is joined
to codon182 of CREB1.The diagram shows only the most
common type of EWS-ATF1fusion protein, in which
EWS codon 325 is joined toATF1codon 65. Data for
native proteins is from InterPro database at
Novel EWS-CREB1Fusion in Clear Cell Sarcoma
www.aacrjournals.orgClin Cancer Res 2006;12(18) September15, 20065361
four of eight previously reported EWS-ATF1-positive CCS of the Download full-text
gastrointestinal tract (9, 21, 23, 24), suggesting that gastrointes-
tinal location rather than the fusion transcript type might
determine the lack of melanocytic differentiation. Furthermore,
two additional gastrointestinal CCS from our files (not
previously reported, see Supplementary Table S1), showing
either EWS-ATF1 by RT-PCR or EWS rearrangement by FISH
lacked evidence of melanocytic differentiation by immunohisto-
chemistry. The consistent expression of neuroectodermal
markers, such as S100, NSE, CD56, and synaptophysin, in our
cases, raises the possibility of a novel gastrointestinal neuro-
ectodermal tumor carrying an EWS-CREB1 fusion and lacking
melanocytic differentiation. However, the detection of EWS-
ATF1 fusion in a number of gastrointestinal tract CCS also
lacking expression of melanocytic markers, argue in favor of a
common histogenesis, possibly from a gastrointestinal neuro-
ectodermal precursor cell which has lost or does not have
potential to differentiate along the melanocytic lineage, in
contrast to the putative neuroectodermal precursor cell of non-
Finally, we note that the clinical behavior of gastrointestinal
CCS seems to be aggressive, regardless of the fusion type, with a
high incidence of both distant and regional metastases. Most
patients developed liver metastases, but intraperitoneal spread
was noted in a few cases as well. The time to distant recurrence
was variable, ranging from 9 months to 5 years.
The authors thankTao Zheng for expert technical assistance with RT-PCR and
Dr. Agnes Viale and the staff of the Memorial Sloan-Kettering Cancer Center
Genomics Core Laboratory forassistancewithmicroarray studies.
Human Cancer Biology
www.aacrjournals.orgClin Cancer Res 2006;12(18) September15, 2006 5362
1. Segal NH, Pavlidis P, Noble WS, et al. Classifica-
tion of clear-cell sarcoma as a subtype of melano-
ma by genomic profiling. J Clin Oncol 2003;21:
2. Panagopoulos I, Mertens F, Isaksson M, Mandahl N.
Absence of mutations of the BRAF gene in malignant
melanoma of soft parts (clear cell sarcoma of tendons
and aponeuroses). Cancer Genet Cytogenet 2005;
3. Zucman J, Delattre O, Desmaze C, et al. EWS and
ATF-1gene fusion induced by t(12;22) translocation
in malignant melanoma of soft parts. Nature Gene
4. Langezaal SM, Graadt van Roggen JF, Cleton-
Jansen AM, Baelde JJ, Hogendoorn PC. Malignant
melanoma is genetically distinct from clear cell sar-
coma of tendons and aponeurosis (malignant mel-
anoma of soft parts). Br J Cancer 2001 ;84:535^8.
5. Antonescu CR,TschernyavskySJ,WoodruffJM, et al.
Moleculardiagnosis ofclearcellsarcoma: detectionof
EWS-ATF1and MITF-M transcripts and histopatho-
logical and ultrastructural analysis of 12 cases. J Mol
6. Panagopoulos I, Mertens F, Debiec-Rychter M, et al.
Molecular genetic characterization of the EWS/ATF1
fusion gene in clear cell sarcoma of tendons and apo-
neuroses. IntJCancer 2002;99:560^7.
7. Covinsky M, Gong S, RajaramV, PerryA, Pfeifer J.
previously diagnosed as malignant melanoma. Hum
8. Taminelli L, Zaman K, Gengler C, et al. Primary clear
cell sarcomaof theileum: anuncommon andmislead-
ing site.Virchows Arch 2005;447:772^7.
9. Zambrano E, Reyes-Mugica M, Franchi A, Rosai J.
An osteoclast-rich tumor of the gastrointestinal tract
with features resembling clear cell sarcoma of soft
parts: reportsof6 cases ofa GISTsimulator. IntJSurg
10. Zucman-Rossi J, Legoix P,Victor JM, Lopez B,
Thomas G. Chromosome translocation based on ille-
gitimate recombination in human tumors. Proc Natl
Acad Sci US A1998;95:11786^91.
11. Patel RM, Downs-Kelly E,Weiss SW, et al. Dual-
color, break-apart fluorescence in situ hybridization
for EWS gene rearrangement distinguishes clear cell
sarcoma of soft tissue from malignant melanoma.
Mod Pathol 2005;18:1585^90.
12. Jean D, Bar-Eli M. Regulation of tumor growth and
metastasis of human melanoma by the CREB tran-
scription factor family. Mol Cell Biochem 2000;212:
13. Shankar DB, Cheng JC, Kinjo K, et al. The role
of CREB as a proto-oncogene in hematopoiesis
and in acute myeloid leukemia. Cancer Cell 2005;
14. Zhang X, Odom DT, Koo SH, et al. Genome-wide
analysis of cAMP-response element binding protein
occupancy, phosphorylation, and target gene activa-
tion in human tissues. Proc Natl Acad Sci U S A
15. Huber WE, Price ER,Widlund HR, et al. A tissue-
restricted cAMP transcriptional response: SOX10
modulates a-melanocyte-stimulating hormone-
triggered expression of microphthalmia-associated
transcription factor in melanocytes. J Biol Chem
16. Li KK, GoodallJ, Goding CR, et al.The melanocyte
inducing factor MITF is stably expressed in cell lines
from human clear cell sarcoma. BrJCancer 2003;89:
17. Jishage M, FujinoT,YamazakiY, Kuroda H, Nakamura
T. Identificationof targetgenes for EWS/ATF-1chime-
ric transcription factor. Oncogene 2003;22:41^9.
18.Waters BL, Panagopoulos I, Allen EF. Genetic char-
tifies fusion of the FUS and ATF-1genes induced by a
16p11 [In Process Citation]. Cancer Genet Cytogenet
19. Raddaoui E, Donner LR, Panagopoulos I. Fusion of
the FUS and ATF1 genes in a large, deep-seated
angiomatoid fibrous histiocytoma. Diagn Mol Pathol
20. Hallor KH, Mertens F, Jin Y, et al. Fusion of
the EWSR1 and ATF1 genes without expression
of the MITF-M transcript in angiomatoid fibrous
histiocytoma. Genes Chromosomes Cancer 2005;
21. Donner LR,Trompler RA, Dobin S. Clear cell sarco-
maof theileum:thecrucialrole ofcytogenetics for the
diagnosis. AmJSurg Pathol1998;22:121^4.
22. FukudaT, KakiharaT, Baba K, et al. Clear cell sarco-
maarisinginthe transverse colon. Pathol Int 2000;50:
23. Friedrichs N,Testi MA, Moiraghi L, et al. Clear cell
sarcoma-like tumor with osteoclast-like giant cells in
the small bowel: further evidence for a new tumor en-
tity. IntJSurg Pathol 2005;13:313^8.
24. Pauwels P, Debiec-Rychter M, Sciot R, et al. Clear
cellsarcomaof the stomach.Histopathology2002;41: