Differential Transactivation by Alternative EWS-FLI1 Fusion Proteins Correlates with Clinical Heterogeneity in Ewing’s Sarcoma

Introduction

Ewing’s sarcoma is a primitive neuroectodermal tumor arising in bone or soft tissue, which strikes over 300 children and young adults in the United States annually. The t(11;22)(q24;q12) translocation is present in up to 95% of cases of Ewing’s sarcoma and results in the formation of the EWS-FLI1 fusion gene that encodes an oncogenic chimeric transcription factor (reviewed in Ref. 1 ). EWS is a functional RNA-binding protein that, along with TLS and a novel TATA-binding protein-associated factor, TAF_II68 (also known as RBP56), forms a new family of related proteins. Both TLS and EWS, but not EWS-FLI1, can function as TATA-binding protein-associated factors (2) . Thus, EWS may mediate direct positive interactions of nascent transcripts with the basal transcription apparatus. However, the EWS portion of EWS-FLI1 may be redirected to other protein targets because of possible conformational changes induced by the gene fusion, as suggested by the recent demonstration of specific binding in vitro and in vivo of the EWS portion of EWS-FLI1—but not of native EWS—to hsRPB7, a subunit of RNA polymerase II (2 , 3) . FLI1 encodes a member of the ETS family of transcription factors and seems to be involved in early hematopoietic, vascular, and neuroectodermal development in model organisms (4) . The highly restricted tissue expression of FLI1 contrasts with that of native EWS, which is high and ubiquitous. Because of the genomic structure of the fusion, the EWS promoter drives the expression of EWS-FLI1.

In standard transactivation assays, the EWS-FLI1 chimeric transcription factor functions as a 5- to 10-fold stronger transactivator than native FLI1 at promoters containing binding sites for FLI1 (5) . The sequences recognized by its DNA binding domain are indistinguishable from those recognized by native FLI1 (5) . The oncogenicity of EWS-FLI1 may depend on a number of features, including constitutive, tissue-inappropriate, differentiation stage-inappropriate expression driven by the EWS promoter, the stronger transactivating potential of the EWS NH₂-terminal domain, and its possible interaction with a different subset of coactivators. It is thus likely that the proximate effect of EWS-FLI1 is the deregulated activation of both FLI1 target genes as well as some genes whose expression is not normally regulated by FLI1.

Many alternative forms of EWS-FLI1 exist because of variations in the locations of the EWS and FLI1 genomic breakpoints (6) . At least 12 types of in-frame EWS-FLI1 chimeric transcripts have been observed clinically. They contain different combinations of exons from EWS and FLI1, reflecting different combinations of EWS and FLI1 genomic breakpoints (6) (Fig. 1) ⇓ . The two main types, fusion of EWS exon 7 to FLI1 exon 6 (type 1) and fusion of EWS exon 7 to FLI1 exon 5 (type 2), account for about 60 and 25%, respectively, of EWS-FLI1 fusions (6, 7, 8) . Biologically significant alternative splicing of the EWS-FLI1 transcript is rarely detected (8) ; therefore, each tumor usually expresses only a single type of fusion transcript. The minimal components of each EWS-FLI1 fusion protein are the NH₂-terminal domain of EWS (encoded by exons 1–7), which functions in vitro as a strong transactivation domain, and the intact DNA-binding domain of FLI1 (encoded by exon 9). The portion of the chimeric transcript between these two domains is variable in size and composition, and this heterogeneity is clinically significant.

Specifically, we and others have recently shown (7 , 8) that the survival of patients whose Ewing’s sarcomas bear the type 1 EWS-FLI1 fusion is markedly better than those with tumors containing other fusion types, regardless of tumor site, stage, or size. The clinical impact may be substantial, and it is possible that patients with non-type 1 fusions may be candidates for more aggressive therapy such as bone marrow ablation and total body irradiation. Before such measures are actually implemented, it will be important to better understand the molecular mechanism by which such differences in survival may occur. However, functional studies of EWS-FLI1 have thus far exclusively examined transactivation by the type 1 fusion product. To determine whether the observed differences in clinical behavior are correlated with functional differences, we compared transactivation by different EWS-FLI1 fusion proteins by complementary approaches in homologous and heterologous cells. We found that the different EWS-FLI1 fusion proteins differ in their transactivation potential in a manner that seems to correlate with the clinical behavior of the associated tumors.

Materials and Methods

Cloning of Alternative EWS-FLI1 cDNAs and Construction of Reporter Plasmids.

Seven different naturally occurring forms of EWS-FLI1 (Fig. 1) ⇓ were cloned into the mammalian expression vector pcDNA3.1 (Invitrogen), as follows. A pBluescript clone of the type 1 EWS-FLI1 fusion cDNA (gift of C. Denny, University of California at Los Angeles, Los Angeles, CA) was digested with NotI and HindIII, and the insert consisting of the full-length cDNA was then ligated into the NotI and HindIII sites of the pcDNA3.1(−) vector to produce the pEF7-6 clone. The clones pEF7-5 (type 2 fusion) and pEF10-6 were made by performing RT-PCR 3 to obtain the junction site from clinical tumor samples, using primers flanking the junctional site and encompassing naturally occurring BamHI and EcoRI sites in EWS and FLI1, respectively. The EWS22.3Bam EWS exon 7 forward primer (CTGGATCCTACAGCCAAGCTCCAAG) and the FLI1C FLI1 exon 6 reverse primer (GTTGAGGCCAGAATTCATGTTA). The RT-PCR product was digested with BamHI and EcoRI, and this fragment was used to substitute for the corresponding junction fragment of pEF7-6. The pEF7-4, pEF7-7, pEF7-8, and pEF7-9 clones were generated artificially, because these fusion types are rare, and no suitable tumor samples were readily available (details of cloning strategy available upon request). The reporter plasmid pS2 was constructed by ligating a custom oligonucleotide that included one copy of a FLI1 consensus binding sequence ACCGGAAGT (5 , 9) into the SacI and KpnI sites of pGL3-Promoter (Promega). All of the clones were verified by direct sequencing. The sequence ACCGGAAGT is known to be bound by both FLI1 and EWS-FLI1 in EMSAs and by FLI1 (EWS-FLI1 not previously tested) in transactivation assays (5 , 9) . Two other sites, ACCGGAACT (pS1) (10) and ACCGGAAACGGA (pS3) (4) , both bound by FLI1 in EMSAs but not previously tested with EWS-FLI1 or in transactivation assays, were also tested in preliminary cotransfection experiments with pEF7-6 or pEF7-5 and were found to be less active in reporter constructs than the sequence ACCGGAAGT in pS2 (results not shown).

Transfection and Cell Culture.

HeLa and NIH3T3 cells were seeded (separately) in 12-well plates at a density of 0.5 × 10⁵ cells and allowed to grow for 18 h before transfection. Cells were washed once with serum-free medium, and a mixture containing 5 μl of Lipofectamine (Life Technologies, Inc.) and 1.5 μg of DNA in 0.5 ml of Opti-MEM (Life Technologies, Inc.) was overlaid on the cells. The DNA mixture included 1.0 μg of a EWS-FLI1-containing pEF series plasmid or pcDNA3.1 (as control empty expression plasmid), 0.25 μg of reporter plasmid pS2 or pGL3-Promoter (as reporter control), and 0.25 μg of the Dual Luciferase system (Promega) transfection control plasmid pTKRL. The cells remained serum-free for 5 h, and then serum was added. Cells were harvested at 48 h. For Ewing’s sarcoma cell lines, the procedure was modified because these cells were more sensitive to the transfection procedure. The cell lines with a type 1 EWS-FLI1 fusion included SK-PN-DW, TC32, A673, and SK-ES-17. The cell lines carrying a type 2 fusion included SK-ES-1, RD-ES, and 6647. The fusion types were reverified by RT-PCR, using the EWS22.3Bam and FLI1C primers. Cells were seeded in 12-well plates at a density of 1.5 × 10⁵ cells and allowed to grow for 24 h. They were washed once with serum-free media and overlaid with a mixture containing 4 μl of Lipofectamine and 1.0 μg of DNA in 0.5 ml of Opti-MEM. The DNA mix included 0.5 μg of reporter plasmid (pS2 or pGL3-Promoter) and 0.5 μg of pTKRL. Serum was added at 2 h, and cells were harvested at 48 h. For the NIH3T3, HeLa, and Ewing’s sarcoma cell lines, transfection cultures were done in triplicate for each condition, and experiments were repeated four times.

Luciferase Reporter Assay.

The Dual-Luciferase assay with firefly luciferase as reporter (based on pGL3-Promoter plasmid) followed by renilla luciferase as transfection control (pTKRL plasmid) was performed according to the manufacturer’s instructions (Promega). Cells were lysed in 200 μl of passive lysis buffer, and 10 μl of the sample was used for luciferase assays. Light readings were made in a luminometer (Turner). The fold transactivation was the ratio of luciferase activity in experimental cultures compared with control cultures. For HeLa and NIH3T3 cells, control cultures were cotransfected with empty pcDNA3.1 expression plasmid, pS2, and the pTKRL transfection control plasmid. For Ewing’s sarcoma cell lines, the control cultures were transfected with unmodified pGL3-Promoter plasmid and pTKRL. The results of the transfection experiments were combined and entered into the SPSS 7.5 statistical software package. Student’s t test was used to compare means of luciferase assays and fold-transactivation.

EMSA.

The TNT coupled reticulocyte lysate system (Promega) was used to produce EWS-FLI1 protein. Briefly, starting with 1 μg of plasmid DNA, in vitro transcription with the T7 RNA polymerase was followed by in vitro translation with a rabbit reticulocyte lysate. The EWS-FLI1 protein was preincubated in the following buffer for 10 min at room temperature: 4% glycerol, 1 mm MgCl₂, 0.5 mm EDTA, 0.5 mm DTT, 50 mm NaCl, 10 mm Tris-HCl (pH 7.5), 50 μg/ml poly dI-dC. Labeled olignocleotide was added to the mix and incubated for 20 min at room temperature. The oligonucleotide GTATGGGTACCGGAAGTGAGCTCGTTAAG was end-labeled with [γ³²P]ATP and T4 kinase. The protein-DNA mix was applied to a 4% nondenaturing acrylamide TBE (Tris-boric acid-EDTA) gel. Autoradiographs were exposed for 24–48 h at −70°C. Controls included the addition of a 10-fold excess of unlabeled identical or unrelated oligonucleotide during the preincubation step.

Results

Comparison of Transactivation by Alternative forms of EWS-FLI1.

Seven different expression constructs containing full-length cDNAs of alternative forms of EWS-FLI1 (Fig. 1) ⇓ were cotransfected with a FLI1-responsive luciferase reporter plasmid into two heterologous cell lines, HeLa (human cervical carcinoma) and NIH3T3 (mouse fibroblast). The expression constructs, all of which were based on the same vector (pcDNA3.1), included pEF7-6 (type 1 fusion), pEF7-5 (type 2 fusion), pEF10-6, pEF7-4, pEF7-7, pEF7-8, and pEF7-9 (Fig. 1) ⇓ . All of these correspond to naturally occurring in-frame EWS-FLI1 fusion transcripts clinically observed in Ewing’s sarcomas (6, 7, 8) . The luciferase reporter plasmid pS2 contained the FLI1 binding sequence ACCGGAAGT, known to be bound by both FLI1 and EWS-FLI1 in EMSAs and by FLI1 in transactivation assays (5 , 9) . Our present results confirm its responsiveness to EWS-FLI1 in transactivation assays. Specifically, in HeLa cells, the type 1 EWS-FLI1 was a significantly weaker transactivator than EWS-FLI1 type 2 (P = 0.003) and the other less common EWS-FLI1 types tested (Fig. 2) ⇓ . Transactivation by the type 2 product was approximately 50% higher than with type 1 EWS-FLI1. Progressive absence of the portion encoded by FLI1 exons 4 through 8 seemed to have a bimodal effect on transactivation, with the lack of exons 4 and 5 resulting in decreased activity, reaching an apparent nadir with the type 1 (construct 7-6) fusion, and with transactivation increasing with further losses corresponding to FLI1 exons 7 and 8. An essentially identical pattern of transactivation was obtained in NIH3T3 cells, with the type 1 (construct 7-6) fusion showing the lowest activity, significantly lower than type 2 (P = 0.033; data not shown).

Comparison of Transactivation of an Exogenous FLI1 Reporter in Ewing’s Sarcoma Cell Lines with Type 1 or Type 2 EWS-FLI1.

As a complementary approach, we also studied transactivation in Ewing’s sarcoma cell lines carrying either the type 1 or type 2 fusion. Because individual tumor cell lines may display idiosyncratic differences in the expression of other closely related ETS-family transcription factors or of as yet uncharacterized cofactors involved in EWS-FLI1-mediated transactivation, we analyzed statistically the results from multiple cell lines with either the type 1 or type 2 EWS-FLI1 fusion. The cell lines were transfected with the same reporter plasmid (pS2) used in the cotransfections in heterologous cells described above. Transactivation of the transiently transfected reporter was found to be approximately 50% greater in Ewing’s sarcoma cell lines carrying the type 2 fusion (P = 0.003; Fig. 3 ⇓ ). Thus, the findings in Ewing’s sarcoma cell lines were entirely consistent with the data from the cotransfections in HeLa and NIH3T3.

EMSAs.

To test the possibility that the lower transactivation mediated by the type 1 fusion was due to lower DNA binding affinity, we performed electromobility shift assays. In vitro translated type 1 and 2 EWS-FLI1 were bound to an end-labeled oligonucleotide containing the same ACCGGAAGT FLI1 binding sequence used in reporter plasmid pS2 described above. The assay demonstrated equivalent binding of the oligonucleotide by the type 1 and the type 2 fusion EWS-FLI1 proteins (Fig. 4) ⇓ . Thus, the differences in transactivation between the two major forms of EWS-FLI1 observed in HeLa, NIH3T3, and Ewing’s sarcoma cell lines do not seem to be due to gross differences in the ability to recognize and bind the FLI1 binding site sequence used in the reporter constructs.

Discussion

Our results suggest that there are demonstrable functional differences between alternative forms of EWS-FLI1. The clinically observed EWS-FLI1 types studied in our experiments constitute a series of six progressive “internal deletions” corresponding to individual exons in the portion contributed by FLI1, namely fusions 7-4 through 7-9 (Fig. 1) ⇓ . We found in cotransfection assays in both HeLa and NIH3T3 cells that loss of the portion correponding to FLI1 exons 4 and 5 reduced transactivation. This is in keeping with studies of native FLI1 that have shown that these exons encode most of the NH₂-terminal transactivation domain of FLI1 (a pointed domain) (11) . Our results indicate that the inclusion of this portion of the FLI1 transactivation domain is functionally significant even when juxtaposed to the stronger EWS transactivation domain.

We also found that further loss of the portion encoded by FLI1 exon 6 seemed to restore transactivation to a level equivalent to EWS-FLI1 type 2, which suggested the presence of an inhibitory domain encoded by this exon. This inhibitory effect does not seem to operate by reducing DNA binding affinity inasmuch as the binding of type 1 and type 2 EWS-FLI1 to the FLI1 binding site of the pS2 reporter was equivalent in EMSAs (Fig. 4) ⇓ . Thus, the inhibitory effect of the portion encoded by FLI1 exon 6 may be mediated by protein-protein interactions, either intramolecular or involving other transcriptional proteins. A similarly positioned region in several other ETS proteins has been implicated in intramolecular inhibition (12) , but the corresponding portion of FLI1 (exons 5–8) shows no homology (11) . More recently, it has been demonstrated that the closely related ERG transcription factor, which fuses with EWS in 5% of cases of Ewing’s sarcoma (6 , 13) , forms homo- and heterodimers through the pointed domain and the ETS domain, and that a central domain, corresponding to exons 6 to 8, may inhibit both dimerization and transactivation (14) . A similar analysis of FLI1 is not available, although its pointed domain, ETS domain, and the portion of the central domain encoded by exon 8 are all identical or highly homologous to ERG, and FLI1 is one of the ETS proteins capable of heterodimerizing with ERG (14) . Hypothetically, the absence of FLI1 exon 6-encoded sequences in EWS-FLI1 could be sufficient to inactivate a similar putative inhibitory domain in FLI1. Other protein-protein interactions have been mapped to this region of FLI1. Specifically, a domain encoded by exons 6 and 7 interacts with serum response factor at serum response elements flanked by ETS-binding sites (15, 16, 17) .

Our cotransfection studies in HeLa and NIH3T3 cells also included EWS-FLI1 type 10-6, thereby providing the first data on the contribution of the portion encoded by EWS exons 8–10 to EWS-FLI1 function. In both cell types, pEF10-6 was a significantly stronger transactivator than pEF7-6, suggesting that the inclusion of a domain encoded within EWS exons 8, 9, and 10 may enhance transactivation by the NH₂-terminal domain of EWS (or may abrogate the apparent inhibitory effect of the central portion of FLI1 corresponding to exon 6). Previous studies of transactivation by the NH₂-terminal domain of EWS in the context of EWS-FLI1 have only examined the region encoded by exons 1–7. Within the EWS-ATF1 translocation product of clear cell sarcoma, however, the inclusion of EWS exon 8 contributes significantly to transactivation (18) . EWS exon 8 also contains a domain that mediates interaction with calmodulin, and phosphorylation at this site by protein kinase C inhibits RNA binding by native EWS (19) . Thus, calmodulin binding or phosphorylation could also affect the function of EWS-FLI1 fusions containing EWS exon 8.

Alternately, the possible impact of pairing with different portions of FLI1 on the conformation of the EWS NH₂-terminal domain could alter, qualitatively or quantitatively, the interactions of the latter with relevant proteins, such as the hsRPB7 subunit of RNA polymerase II (2 , 3) or the ZFM1 transcriptional repressor (20) .

It may seem paradoxical that the type 1 EWS-FLI1 fusion product is the least effective transactivator but is the form of EWS-FLI1 observed in the majority of Ewing’s sarcomas. Although the type 1 fusion protein may be less functionally active than other fusion types, it seems clinically sufficient for transformation. A recent detailed analysis of EWS and FLI1 genomic breakpoints in Ewing’s sarcoma cell lines failed to identify putative recombinogenic sequence elements in the rearranged introns, suggesting that the frequency of different EWS-FLI1 fusion types may be partly determined by the sizes of the introns in which the genomic breaks occur after selection for a functional chimeric protein (21) . Within the FLI1 breakpoint cluster region, breaks producing the two most common forms of EWS-FLI1, type 1 and type 2, involve the largest and second largest introns of FLI1, respectively. Intriguingly, however, in relation to intron size, the density of breakpoints in FLI1 intron 4 (associated with the type 2 fusion) was 3-fold higher than in FLI1 intron 5 (associated with the type 1 fusion) (21) , which suggests the possibility of a slight in vivo or in vitro selective advantage for tumors with the type 2 fusion.

In summary, the variably included regions of EWS-FLI1 seem to determine functional differences that correlate closely with significant clinical heterogeneity in Ewing’s sarcoma. Specifically, type 1 EWS-FLI1 is less effective in transactivation assays, and may, therefore, be less effective in activating critical target genes. These are just beginning to be identified (22 , 23) ; the forthcoming characterization of the enhancer elements of these actual in vivo targets of EWS-FLI1 should provide reagents to address experimentally one of the caveats of the present study, namely that the reporter element used was synthetic and had been primarily defined by its interaction with native FLI1 in vitro or in model in vivo assays. The functional comparison of the different alternative EWS-FLI1 types using clinically relevant targets will also be important to begin to link mechanistically the moderate (2-fold) differences in transactivation potential demonstrated in the present study to the clinical biology of Ewing’s sarcoma. Nonetheless, our results may provide the first data toward a molecular explanation of the clinical association of the type 1 fusion with better patient survival, compared with other fusion types.