Recent Advances in the Molecular Pathogenesis of Ewing's Sarcoma

Abstract

Tumor development is a complex process resulting from interplay between mutations in oncogenes and tumor suppressors, host susceptibility factors, and cellular context. Great advances have been made by studying rare tumors with unique clinical, genetic, or molecular features. Ewing's sarcoma serves as an excellent paradigm for understanding tumorigenesis because it exhibits some very useful and important characteristics. For example, nearly all cases of Ewing's sarcoma contain the (11;22)(q24;q12) chromosomal translocation that encodes the EWS/FLI oncoprotein. Besides the t(11;22), however, many cases have otherwise simple karyotypes with no other demonstrable abnormalities. Furthermore, it appears that an underlying genetic susceptibility to Ewing's sarcoma, if it exists, must be rare. These two features suggest that EWS/FLI is the primary mutation that drives the development of this tumor. Finally, Ewing's sarcoma is an aggressive tumor that requires aggressive treatment. Thus, improved understanding of the pathogenesis of this tumor will not only be of academic interest, but may also lead to new therapeutic approaches for individuals afflicted with this disease. The purpose of this review is to highlight recent advances in understanding the molecular pathogenesis of Ewing's sarcoma, while considering the questions surrounding this disease that still remain and how this knowledge may be applied to developing new treatments for patients with this highly-aggressive disease.

Introduction

Ewing's sarcoma is a rare but important solid tumor in children and young adults (Arndt and Crist, 1999). Ewing's sarcoma most frequently arises in bone, but <10% of tumors originate in the soft tissues (Horowitz et al., 1997). Approximately 15-25% of patients present with overt metastasis (Terrier et al., 1996); those without clinically-detectable metastases likely have micrometastases, because in the absence of systemic chemotherapy most will relapse with distant metastatic disease following surgical resection (Dahlin et al., 1961; Wang and Schulz, 1953). This propensity to spread contributes to the poor prognosis for Ewing's sarcoma patients and long term cure rate of approximately 60% (Linabery and Ross, 2008). It is hoped that elucidation of the molecular mechanisms at play in these tumors will translate to improved patient outcomes.

Approximately 85% of Ewing's sarcoma tumor specimens harbor the t(11;22)(q24;q12) chromosomal abnormality (Turc-Carel et al., 1988). This translocation fuses the EWSR1 gene on chromosome 22 to the FLI1 gene on chromosome 11 and encodes the EWS/FLI fusion protein (Delattre et al., 1992). EWS/FLI contains the amino-terminus of EWS fused, in frame, to the carboxyl-terminus of FLI. EWS is a protein of uncertain function, while FLI is a member of the ETS family of transcription factors. In the context of this fusion, a strong transcriptional activation domain is contributed by EWS and an ETS-type DNA binding domain is contributed by FLI (Delattre et al., 1992; Lessnick et al., 1995; May et al., 1993a; May et al., 1993b). Both of these domains are required for the oncogenic function of EWS/FLI, supporting the notion that the fusion acts as an aberrant transcription factor (May et al., 1993a; May et al., 1993b). Interestingly, in Ewing's sarcoma cases lacking EWS/FLI, alternate translocations are present that fuse EWS (or a highly-similar protein, TLS/FUS) to other ETS family transcription factors, including, ERG, ETV1, ETV4, and FEV (Jeon et al., 1995; Kaneko et al., 1996; Ng et al., 2007; Peter et al., 1997; Shing et al., 2003; Sorensen et al., 1994), which all likely mimic EWS/FLI (Braunreiter et al., 2006; Sorensen et al., 1994; Teitell et al., 1999; Thompson et al., 1999) (Figure 1). Askin's tumors and peripheral primitive neuroectodermal tumors (pPNET) also harbor EWS/ETS fusions, and are now considered to be manifestations of Ewing's sarcoma. Some investigators have used the term “Ewing's sarcoma family of tumors (ESFT)” to highlight this relationship. However, we prefer the simpler term “Ewing's sarcoma” to refer to all of these genetically-identical tumors.

Figure 1.

Ewing's sarcoma fusion protein organization. The Ewing's sarcoma translocation product is the result of a chromosomal rearrangement involving the N-terminal transcriptional activation domain (TAD) of a TET family member (either TLS/FUS or more commonly EWS) and the C-terminal portion of an ETS family member (FLI, ERG, ETV1, ETV4, or FEV), including the ETS DNA binding domain (DBD).

EWS/FLI and the alternate fusions have been classified as oncogenes based on their ability to transform immortalized murine NIH3T3 cells (Braunreiter et al., 2006; May et al., 1993a). In contrast, EWS/FLI is not sufficient to transform any normal human cell type tested so far (Lessnick et al., 2002; Miyagawa et al., 2008; Riggi et al., 2008), suggesting that critical differences might exist between human and mouse cells in their abilities to respond to EWS/FLI. Nevertheless, ongoing EWS/FLI expression is required for the oncogenic phenotype of Ewing's sarcoma cells, further supporting the assertion that EWS/FLI is the critical oncoprotein in this disease (Chansky et al., 2004; Kinsey et al., 2006; Owen and Lessnick, 2006; Prieur et al., 2004; Smith et al., 2006; Stegmaier et al., 2007).

Ewing's Sarcoma Cell of Origin

The cell of origin for Ewing's sarcoma is unknown. This lack of knowledge has hindered progress in the field by precluding the identification of an appropriate model system in which to study the development of the disease. Identification of a permissive cellular environment for the expression of EWS/FLI has also been problematic: expression of EWS/FLI in many primary cell lines results in cell death or growth arrest, while expression in more primitive cells or tumor cell lines causes differentiation defects (reviewed in Kovar, 2005). As stated, NIH3T3 cells provide a useful cellular context for EWS/FLI expression as they are transformed by the fusion (May et al., 1997; May et al., 1993a). However, concerns have recently been raised about how well the NIH3T3 model recapitulates the human disease. For example, recent comparative analyses have demonstrated poor correlation in gene expression patterns between the NIH3T3 model and other Ewing's sarcoma model and primary tumor datasets (Braunreiter et al., 2006; Hancock and Lessnick, 2008). Furthermore, some EWS/FLI gene targets that have been shown to be critical for oncogenic transformation in patient-derived Ewing's sarcoma cell lines, such as NKX2.2 and NR0B1, are not induced by the fusion protein in NIH3T3 cells (Gangwal et al., 2008; Kinsey et al., 2006; Owen and Lessnick, 2006; Smith et al., 2006). These data suggest that the molecular pathways used in NIH3T3 mouse fibroblasts may be different from those used in bona fide Ewing's sarcoma.

Unlike other sarcomas, such as osteosarcoma and liposarcoma that show some lineage specific differentiation (reviewed in Charytonowicz et al., 2009), Ewing's sarcoma presents as an undifferentiated “small round blue cell tumor” that reveals little insight into its cell of origin (Triche et al., 1987). When first described in 1921, James Ewing proposed an endothelial origin (Ewing, 1921). Since that time, numerous hypotheses have been put forth regarding the histogenesis of this tumor including hematopoietic (Kadin and Bensch, 1971), fibroblastic (Dickman et al., 1982), neural crest (Cavazzana et al., 1988), and mesenchymal progenitor/stem cells (Riggi et al., 2005; Tirode et al., 2007). The latter two cell types have been the focus of much investigation.

Several lines of evidence provide support for the neural crest cell of origin hypothesis. Early studies found that cell surface antigens associated with the neuroectodermal lineage were expressed on Ewing's sarcomas (Lipinski et al., 1986; Lipinski et al., 1987a; Lipinski et al., 1987b). Consistent with this observation, both Ewing's sarcoma and pPNET harbor the same t(11;22)(q24;q12) rearrangement. This suggests that these are the same tumor demonstrating differences in extent of neural differentiation (Kovar, 1998; Turc-Carel et al., 1984; Whang-Peng et al., 1984). A gene expression profiling study recently found that genes expressed in neural tissues or during neuronal differentiation are highly expressed in Ewing's sarcomas, and that Ewing's sarcomas clustered with fetal and adult brain tissue (Staege et al., 2004). This study also reported that EWS/FLI expression in bone marrow cells upregulated neural genes. This latter observation suggested an alternate hypothesis: that the Ewing's sarcoma neural phenotype is a result of EWS/FLI expression, rather than a reflection of the cell of origin. This hypothesis is further supported by work demonstrating that genes critical for neural crest development were upregulated when EWS/FLI was ectopically expressed in rhabdomyosarcoma, neuroblastoma, or human foreskin fibroblast cell lines (Hu-Lieskovan et al., 2005; Lessnick et al., 2002; Rorie et al., 2004).

There is a growing body of evidence that suggests Ewing's sarcoma is derived from a mesenchymal stem or progenitor cell. Early studies demonstrated that EWS/FLI and EWS/ERG blocked the differentiation of pluripotent murine bone marrow-derived mesenchymal progenitor cells (mMPCs; Torchia et al., 2003). Subsequent studies established that introduction of EWS/FLI into unselected primary murine bone marrow cells or into mMPCs allowed the transduced cells to form tumors in immunocompromised mice with a small round cell morphology (Castillero-Trejo et al., 2005; Riggi et al., 2005). Although it was suggested that these tumors also expressed CD99, a classical Ewing's sarcoma marker, there is some controversy as to whether the murine genome harbors a CD99 allele that is paralogous to the human version (Kovar and Bernard, 2006; Riggi et al., 2005).

A reciprocal approach complemented the gain-of-function data. Ewing's sarcoma cell lines in which EWS/FLI had been silenced with RNAi displayed a mesenchymal stem cell (MSC) gene expression profile, and these cells demonstrated the capacity to differentiate along both osteogenic and adipogenic lineages, consistent with a MSC phenotype (Tirode et al., 2007). Importantly, human MSCs also provide an appropriate cellular context for EWS/FLI expression. In contrast to other normal human cell types with forced EWS/FLI expression, human MSCs retain the ability to propagate in the presence of the fusion protein (Riggi et al., 2008). Gene expression profiles from these cells are similar to Ewing's sarcoma, but not to other bone and soft tissue tumors (Miyagawa et al., 2008; Riggi et al., 2008). Because these cells were unable to form tumors when injected into immunocompromised mice (Riggi et al., 2008), it appears that EWS/FLI is necessary, but not sufficient, for oncogenic transformation of human MSCs.

Interestingly, it has been recently suggested that the proposed neural crest and MSC origins may not be mutually exclusive (Riggi et al., 2009). On the one hand, it was demonstrated that neural derived MSCs are present in the bone marrow of developing mice (Takashima et al., 2007). Conversely, it has also been shown that neural crest stem cells contain some mesenchymal lineage plasticity (Lee et al., 2007). Taken together, it is possible that Ewing's sarcoma might arise from a neural-derived MSC or from a neural crest stem cell that harbors mesenchymal potential.

There are no known precursor lesions for Ewing's sarcoma, preventing the definitive identification of the cell of origin through analysis of early stage precancerous cells. An alternate approach towards the identification of the cell of origin could be the development of a mouse model for the disease through conditional EWS/FLI expression in the correct progenitor cell. Such an approach has been effective for the identification of cells of origin of other sarcomas, including alveolar rhabdomyosarcoma (Keller et al., 2004a; Keller et al., 2004b), synovial sarcoma (Haldar et al., 2007), and osteosarcoma (Walkley et al., 2008). The development of genetically engineered mouse models of Ewing's sarcoma remains a challenge. Early attempts to express EWS/FLI, or equivalent translocations, in mice induced leukemia, and not sarcoma (Codrington et al., 2005; Torchia et al., 2007). Forced expression of EWS/FLI in mesenchymal cells of the developing limb bud in mice resulted in abnormalities of limb development without tumor formation (Lin et al., 2008). In this latter study, however, EWS/FLI accelerated the tumor development typically seen with p53 deletion, and changed the predominant tumor type from osteosarcoma to an “undifferentiated” sarcoma. Whether this undifferentiated sarcoma represents a mouse equivalent of Ewing's sarcoma is unclear, and further study will be required to better define these tumors. Interestingly, it has been recently suggested that genomic differences in microsatellite content near critical EWS/FLI target genes may prohibit the development of a mouse model (see below; Gangwal and Lessnick, 2008; Gangwal et al., 2008).

EWS/FLI Transcriptional Regulation

It is well appreciated that EWS/FLI acts as an aberrant transcription factor and that several of its downstream targets contribute to tumorigenesis. While early studies classified EWS/FLI as a strong transcriptional activator (Bailly et al., 1994; Lessnick et al., 1995; May et al., 1993b; Ohno et al., 1993), later studies suggested that it also functions as a transcriptional repressor at some gene targets (Hahm et al., 1999; Nakatani et al., 2003). The repressive capacity of EWS/FLI was further supported by expression profiling studies using patient derived Ewing's sarcoma cell lines (Prieur et al., 2004; Smith et al., 2006). These microarray analyses identified over one-thousand EWS/FLI regulated genes (which are comprised of direct and indirect target genes) in A673 cells; surprisingly, downregulated targets comprised 80% or more of the dysregulated genes. Subsequent comprehensive transcriptional analyses comparing multiple Ewing's sarcoma cell lines and tumor samples revealed a more equal representation of EWS/FLI induced and repressed genes (Hancock and Lessnick, 2008; Kauer et al., 2009; Kinsey et al., 2006). While the exact magnitude of transcriptional repression induced by EWS/FLI remains to be clarified, studies demonstrating the importance of target gene repression in tumorigenic phenotype suggest that this may be a critical EWS/FLI-mediated activity (Hahm et al., 1999; Prieur et al., 2004).

ETS family members (including EWS/FLI) bind to sequences containing a GGAA “core” motif surrounded by bases that provide affinity and specificity to the interaction (Seth and Watson, 2005; Sharrocks, 2001; Szymczyna and Arrowsmith, 2000). In vitro binding site selection approaches identified ACCGGAAGTG as a site that binds wild-type FLI and EWS/FLI (and other ETS family members) with high affinity (Mao et al., 1994; Seth and Watson, 2005; Sharrocks, 2001). Whole genome localization analysis (ChIP-chip) demonstrated that EWS/FLI binds to this same sequence element in vivo (Gangwal et al., 2008). This latter study also identified GGAA-containing microsatellite sequences as EWS/FLI binding sites associated with a number of genes involved in oncogenic transformation in Ewing's sarcoma. This result was confirmed by a ChIP-sequencing approach as well (Guillon et al., 2009). These findings were surprising because microsatellite sequences have been described as “junk DNA” without biologic function (Gangwal and Lessnick, 2008). The identification of microsatellites as cancer-relevant EWS/FLI binding sites suggests that such elements have important roles in the development of Ewing's sarcoma. Follow-up studies have begun to unravel the mechanistic basis for EWS/FLI binding to microsatellite sequences and have suggested that similar mechanisms might also be utilized by other cancer-relevant ETS family members (Gangwal et al., 2010).

EWS/FLI binding to GGAA microsatellites was associated solely with gene activation. Other mechanisms of transcriptional activation and repression by EWS/FLI are slowly becoming defined. Protein-protein interactions and post-translational modifications appear to take part in both activating and repressing functions of EWS/FLI. For example, EWS/FLI has been shown to be phosphorylated in Ewing's sarcoma cell lines (Klevernic et al., 2009), and phosphorylation modulates DNA binding and transcriptional activity in a heterologous system (Olsen and Hinrichs, 2001), suggesting that this modification may be relevant to the oncogenic function of EWS/FLI. It was also recently reported that EWS/FLI undergoes O-GlcNAcylation (Bachmaier et al., 2009). This may be involved directly in EWS/FLI-mediated transcriptional activation, or alternately, may indirectly modulate EWS/FLI transcriptional function by altering the protein's intracellular half-life.

Full length EWS directly associates with the general transcriptional machinery, including RNA polymerase II and TFIID (reviewed in Tan and Manley, 2009). In contrast, the EWS/FLI fusion does not appear to be in stable association with the RNA polymerase II complex (Bertolotti et al., 1998), but does interact with the RNA polymerase II subunit RPB7 (Petermann et al., 1998). In addition to the basal machinery, EWS/FLI interacts with other transcriptionally-relevant proteins, such as RNA helicase A (Toretsky et al., 2006). Inhibition of this association with small molecules reduced tumor growth in xenograft mouse models, suggesting that this interaction is biologically relevant (Erkizan et al., 2009). While the use of reporter assays indicate that the RNA helicase A-EWS/FLI interaction increases EWS/FLI transcriptional activity, the presence of RNA helicase A at the promoters of EWS/FLI-repressed targets complicates the interpretation of these findings (Toretsky et al., 2006).

Progress has also been made recently in understanding the mechanistic basis for transcriptional repression by EWS/FLI. For example, interaction between EWS/FLI and p300 serves to inhibit the latter protein's histone acetyl-transferase activity, blocking its transcriptional activation and resulting in transcriptional repression (Nakatani et al., 2003). A similar effect has been reported for wild-type EWS and its highly-related family member TLS/FUS (Wang et al., 2008). In addition to direct transcriptional repression, it should be noted that EWS/FLI also upregulates the expression of transcriptional repressors, such as NKX2.2, co-repressors, such as NR0B1, and polycomb-family repressors, such as EZH2, that appear to account for some of the downregulated signature (Kinsey et al., 2009; Owen et al., 2008; Riggi et al., 2008).

EWS/FLI Targets and Transformation Pathways

Genomic approaches linking RNAi technologies with microarray analysis have identified thousands of genes that are dysregulated by EWS/FLI in Ewing's sarcoma cell lines (Kauer et al., 2009; Kinsey et al., 2006; Prieur et al., 2004; Smith et al., 2006). While not all of these genes are actively involved in the transformation process, a small but growing subset of these have been shown to be critical for oncogenesis (e.g., NKX2.2, NR0B1, and GLI1; Beauchamp et al., 2009; Joo et al., 2009; Kinsey et al., 2009; Kinsey et al., 2006; Owen et al., 2008; Smith et al., 2006; Zwerner et al., 2008). Interestingly, these proteins all have a role in transcriptional regulation. However, the mechanisms by which these transcriptional regulators, or their downstream targets, mediate oncogenic transformation are not fully understood. In contrast, other EWS/FLI target genes are more readily connected to hallmarks necessary for tumor formation and progression (Hanahan and Weinberg, 2000). For example, many Ewing's sarcoma tumors and cell lines express human telomerase reverse transcriptase (hTERT) (Amiel et al., 2003; Ohali et al., 2003; Schuck et al., 2002) and vascular endothelial growth factor (VEGF; Fuchs et al., 2004b). Although it has not been demonstrated in Ewing's sarcoma cell lines per se, these genes are induced by EWS/FLI in heterologous cellular backgrounds (Fuchs et al., 2004a; Fuchs et al., 2004b). The expression of these two genes allows for tumor cell immortalization and vasculogenesis. Other EWS/FLI target genes contribute to additional cancer-specific phenotypes such as preventing differentiation to “pirate” some qualities of stem cells, metastatic spread, and drug resistance as discussed further below.

Cancer stem cells (CSC) were recently identified in Ewing's sarcoma (Suva et al., 2009). These are to be distinguished from the mesenchymal stem cell tumor cell of origin hypothesis discussed above. CSCs have the ability to form tumors from a single cell, and may be more resistant to chemotherapy, suggesting that current approaches that reduce tumor bulk may leave CSCs behind, thereby allowing tumors to reform. The identification of CSCs in Ewing's sarcoma will now allow key aspects of the CSC model to be tested in this disease. As an interesting corollary to the identification of Ewing's sarcoma CSCs, other work has identified a role for “stem cell genes” in Ewing's sarcoma. For example, polycomb repressor complexes play a central role in maintaining stemness and pluripotency (Bracken et al., 2006; Lee et al., 2006; Tolhuis et al., 2006). Members of these complexes, including EZH2 and BMI1, are expressed in Ewing's sarcoma (and in the case of EZH2, regulated by EWS/FLI), are required for Ewing's sarcoma oncogenic transformation, and appear to block normal differentiation pathways in this disease (Burdach et al., 2009; Douglas et al., 2008; Riggi et al., 2008). Thus, the polycomb group family appears critical to Ewing's sarcoma tumorigenesis.

The PI3K-AKT pathway can mediate both cellular proliferation and survival (Vivanco and Sawyers, 2002). The caveolin gene (CAV1) connects EWS/FLI to this pathway. The CAV1 promoter contains a GGAA microsatellite that is occupied by EWS/FLI in vivo (Gangwal et al., 2008), is highly expressed in Ewing's sarcoma tumors, and promotes tumor growth in mouse xenograft models of Ewing's sarcoma (Tirado et al., 2006). The requirement of CAV1 for tumorigenicity was attributed to its ability to indirectly activate E-cadherin expression (Tirado et al., 2006). E-cadherin is also upregulated when Ewing's sarcoma cell lines are grown under anchorage independent conditions, and its expression correlated with phosphorylation of the receptor tyrosine kinase ERBB4 and phosphorylation of AKT (Kang et al., 2007). Cell-cell contacts and signaling mediated by E-cadherin and ERBB4 may promote survival and suppress anoikis to facilitate metastatic spread in Ewing's sarcoma (Kang et al., 2007).

Despite improved understanding of transformation pathways operative in Ewing's sarcoma, adequate therapy remains a challenge. Patients often show promising responses to initial courses of chemotherapy, but later relapse with chemotherapy-resistant tumors and metastases (Meyers and Levy, 2000). Glutathione S-transferases (GST) are detoxification enzymes that mediate solubility and excretion of both endogenous and exogenous reactive compounds (Comstock et al., 1994). Recently, GSTM4 was identified as a GGAA-microsatellite containing gene that was required for the transformed phenotype and modulates resistance to chemotherapeutic agents in Ewing's sarcoma cells (Luo et al., 2009). Consistent with this in vitro observation, GSTM4 protein levels in primary tumors inversely correlate with patient survival. Similar relationships between expression levels and patient outcome were recently reported for another GST enzyme, MGST1 (Scotlandi et al., 2009). This latter study demonstrated efficacy of a small-molecule inhibitor of GST enzymes, 6-(7-nitro-2,1,3-benzoxadiazol-4-ylthio)hexanol (NBDHEX), against a number of Ewing's sarcoma cell lines in vitro, suggesting a new therapeutic approach to this disease.

Parallel Pathways and Cooperating Mutations

EWS/FLI is indisputably the central player in Ewing's sarcoma pathogenesis. However, the fact that EWS/FLI has not been sufficient to transform any human cell type in vitro implies that there are parallel pathways and/or cooperating mutations working in conjunction with EWS/FLI. Work in this area has provided promising leads to what such EWS/FLI collaborators may be, yet we remain far from a complete understanding of exactly what molecular combination provides the “Ewing's sarcoma oncogenic cocktail” (Figure 2).

Figure 2.

Ewing's sarcoma transformation requires several distinct events. The EWS/FLI (or other TET/ETS) translocation is the central mediator of this process, dysregulating a number of genes that contribute to oncogenesis and tumor progression. A permissive cell type for EWS/FLI expression is also required. Cooperating mutations such as those in the RB and p53 pathways, as well as growth factor signaling (including IGF) likely contribute to the fully transformed phenotype.

Two decades ago it was appreciated that Ewing's cell lines expressed both insulin-like growth factor 1 (IGF-1) and its receptor (IGF-1R), and that blocking antibodies targeting the receptor could slow cell growth in culture (Yee et al., 1990). Since that time, a variety of studies have supported the importance of that pathway in Ewing's sarcoma (e.g., Hamilton et al., 1991; Scotlandi et al., 1998; Scotlandi et al., 1996; Scotlandi et al., 2002b; Toretsky et al., 1997; van Valen et al., 1992). This work has important clinical implications, as both small molecule and antibody-mediated approaches to IGF pathway blockade have shown efficacy in preclinical models of the disease, and in patients in early-phase clinical trials (Benini et al., 2001; Benjamin et al., 2007; Kolb et al., 2008; Kurmasheva et al., 2009; Manara et al., 2010; Olmos et al., 2009; Scotlandi et al., 2002a; Scotlandi et al., 2005; Tolcher et al., 2007). It should be noted, that although the IGF pathway is often considered a “parallel” pathway to EWS/FLI, it may in fact be modulated by the fusion protein. For example, the IGF binding protein 3 gene (IGFBP3) is downregulated by EWS/FLI (Prieur et al., 2004), which might allow for increased IGF signaling. Additionally, recent work suggests that EWS/FLI may regulate IGF1 itself (Cironi et al., 2008; Herrero-Martin et al., 2009).

CD99 (also called MIC2) is an integral transmembrane glycoprotein that is the most commonly used diagnostic marker in Ewing's sarcoma to distinguish Ewing's sarcoma from other small round blue cell tumors (Ambros et al., 1991; Fellinger et al., 1991; Kovar et al., 1990). Engagement of CD99 with monoclonal antibodies can induce apoptosis in Ewing's sarcoma cell lines (Sohn et al., 1998). This finding was extended to show that CD99 engagement slowed tumor formation in athymic mice and enhanced the growth inhibitory effects of doxorubicin and vincristine in cell culture experiments (Scotlandi et al., 2000). Recent data suggests that CD99 is not only a marker and therapeutic target for Ewing's sarcoma, but may also contribute to the disease phenotype. Knock-down of CD99 in Ewing's sarcoma cell lines resulted in decreased growth in tissue culture, diminished colony formation in soft agar assays, reduced cell motility, and smaller tumors with less metastasis in xenograft models (Rocchi et al., 2010). This study also suggested that CD99 inhibits full neuronal differentiation by decreasing the activity of the MAP kinase pathway.

Alterations in the RB and p53 pathways are likely cooperating mutations in Ewing's sarcoma. For example, expression of the cell cycle progression protein cyclin D1 (which phosphorylates and inactivates RB) is generally upregulated (Fuchs et al., 2004b; Sanchez et al., 2008; Zhang et al., 2004), while the cell cycle inhibitory proteins RB and p16INK4A are often mutated or deleted in Ewing's sarcoma (Huang et al., 2005; Kovar et al., 1997; Wei et al., 2000). Mutations in this pathway also have prognostic value, with patients harboring p53 or p16INK4a/ARF deletions having a worse prognosis than those without (de Alava et al., 2000; Huang et al., 2005; Wei et al., 2000). The biologic basis for these observations may be related to oncogenic stress mediated by the EWS/FLI fusion, as expression of EWS/FLI in primary fibroblasts leads to growth arrest or cell death (Deneen and Denny, 2001; Lessnick et al., 2002). Inhibition of the p53 and/or RB pathways in these models bypasses these effects, suggesting that there may be selective pressure in vivo for cooperating mutations in these pathways.

Copy Number Alterations in Ewing's Sarcoma

Genome wide approaches offer the opportunity to identify cooperative pathways more efficiently. Several studies in the past decade have used early generation array comparative genomic hybridization (aCGH) technologies on Ewing's clinical samples (Armengol et al., 1997; Brisset et al., 2001; Ferreira et al., 2008; Hattinger et al., 2002; Maurici et al., 1998; Ozaki et al., 2001; Savola et al., 2009; Selvarajah et al., 2007; Shing et al., 2002; Tarkkanen et al., 1999). Despite the small numbers of tumors and relatively low-resolution offered by early generation aCGH, many of these studies have described overlapping regions of copy number gains or losses (collectively referred to as copy number alterations, or CNAs; Table 1). Deletion of 9p21.3, containing the CDKN2A gene that encodes p16INK4a/ARF, was deleted in 14-67% of Ewing's tumors and cell lines (Brownhill et al., 2007; Neale et al., 2008). Conversely, gain of the specific region of 1q22 has been reported to occur in 18-63% of Ewing's samples or cell lines (Armengol et al., 1997; Savola et al., 2009; Shing et al., 2002; Tarkkanen et al., 1999), suggesting the presence of a possible oncogene in this region.

Table 1.

Copy Number Alterations (CNAs) reported in Ewing's sarcoma.

Deletion	Gain	Frequency (%)	Sample Type	Technology	Study
	1q	12/41 (29%)	ESFT Samples (Primary, Metastatic)	Karyotyping and CGH	Hattinger et al. (2002)
		14/83 (17%)	ESFT Samples (Primary, Localized)	Karyotyping and CGH	Hattinger et al. (2002)
		6/22 (27%)	ESFT Samples (Metastatic)	CGH	Brisset et al. (2001)
		7/28 (25%)	ESFT Samples (Metastatic)	Karyotyping (G-Band)	Roberts et al. (2008)
		8/45 (18%)	ESFT Samples (Localized)	Karyotyping (G-Band)	Roberts et al. (2008)
		9/52 (17%)	ESFT Samples (Primary, includes 20 Metastatic)	CGH	Ozaki et al. (2001)
	1q 21-q22	5/20 (25%)	ESFT Samples (17 patients, two specimens from same pt. in 3 cases)	CGH	Armengol et al. (1997)
	1q21-22	5/28 (18%)	ESFT Samples	CGH	Tarkkanen et al. (1999)
	1q21-q32	5/8 (63%)	Cell line	CGH	Shing et al. (2002)
	1q22-qter	10/31 (32%)	ESFT Samples	CGH	Savola et al. (2009)
	2	11/100 (11%)	ESFT Samples (Primary, Localized + Metastatic)	Karyotyping and CGH	Hattinger et al. (2002)
		5/21 (24%)	ESFT Samples (Localized)	CGH	Brisset et al. (2001)
		9/31 (29%)	ESFT Samples	CGH	Savola et al. (2009)
	2q	4/10 (40%)	ESFT Samples (Relapse)	CGH	Ozaki et al. (2001)
	4p	4/10 (40%)	ESFT Samples (Relapse)	CGH	Ozaki et al. (2001)
	4q	3/20 (15%)	ESFT Samples (17 patients, two specimens from same pt. in 3 cases)	CGH	Armengol et al. (1997)
	5	12/100 (12%)	ESFT Samples (Primary, Localized + Metastatic)	Karyotyping and CGH	Hattinger et al. (2002)
	5p	5/25 (20%)	ESFT Samples (23 Primary, 2 Relapse)	CGH	Ferreira et al. (2008)
	7	10/100 (10%)	ESFT Samples (Primary, Localized + Metastatic)	Karyotyping and CGH	Hattinger et al. (2002)
	7p21.1-p11.2	2/9 (22%)	ESFT Samples vs. cell line xenografts	SNP Microarray (Affy 100K)	Neale et al. (2008)
	7q	5/28 (18%)	ESFT Samples	CGH	Tarkkanen et al. (1999)
	8	10/28 (36%)	ESFT Samples	CGH	Tarkkanen et al. (1999)
		10/28 (36%)	ESFT Samples (Metastatic)	Karyotyping (G-Band)	Roberts et al. (2008)
		14/25 (56%)	ESFT Samples (23 Primary, 2 Relapse)	CGH	Ferreira et al. (2008)
		20/45 (44%)	ESFT Samples (Localized)	Karyotyping (G-Band)	Roberts et al. (2008)
		21/31 (67%)	ESFT Samples	CGH	Savola et al. (2009)
		24/52 (46%)	ESFT Samples (1°, Mets, Recurrence)	FISH (24) or Karyotyping (24)	Maurici et al. (1998)
		25/42 (60%)	ESFT Samples (Primary, Metastatic)	Karyotyping and CGH	Hattinger et al. (2002)
		5/22 (23%)	ESFT Samples (Metastatic)	CGH	Brisset et al. (2001)
		43/89 (48%)	ESFT Samples (Primary, Localized)	Karyotyping and CGH	Hattinger et al. (2002)
		7/20 (35%)	ESFT Samples (17 patients, two specimens from same pt. in 3 cases)	CGH	Armengol et al. (1997)
		8/10 (80%)	ESFT Samples (Relapse)	CGH	Ozaki et al. (2001)
		8/21 (38%)	ESFT Samples (Localized)	CGH	Brisset et al. (2001)
	8p	22/52 (42%)	ESFT Samples (Primary, includes 20 Metastatic)	CGH	Ozaki et al. (2001)
	8q	24/52 (46%)	ESFT Samples (Primary, includes 20 Metastatic)	CGH	Ozaki et al. (2001)
	8q11.21-q22.3	6/9 (67%)	ESFT Samples vs. cell line xenografts	SNP Microarray (Affy 100K)	Neale et al. (2008)
	8q11.2-q22	7/8 (88%)	6 cell line, 1 primary culture	CGH	Shing et al. (2002)
	8q23-q24.1	5/8 (63%)	4 cell line, 1 primary culture	CGH	Shing et al. (2002)
	8q24.11-q24.21	7/9 (78%)	ESFT Samples vs. cell line xenografts	SNP Microarray (Affy 100K)	Neale et al. (2008)
	12	13/43 (30%)	ESFT Samples (Primary, Metastatic)	Karyotyping and CGH	Hattinger et al. (2002)
		17/52 (33%)	ESFT Samples (1°, Mets, Recurrence)	FISH (24) or Karyotyping (24)	Maurici et al. (1998)
		23/88 (26%)	ESFT Samples (Primary, Localized)	Karyotyping and CGH	Hattinger et al. (2002)
		3/28 (11%)	ESFT Samples	CGH	Tarkkanen et al. (1999)
		3/28 (11%)	ESFT Samples (Metastatic)	Karyotyping (G-Band)	Roberts et al. (2008)
		4/22 (18%)	ESFT Samples (Metastatic)	CGH	Brisset et al. (2001)
		5/20 (25%)	ESFT Samples (17 patients, two specimens from same pt. in 3 cases)	CGH	Armengol et al. (1997)
		5/25 (20%)	ESFT Samples (23 Primary, 2 Relapse)	CGH	Ferreira et al. (2008)
		6/21 (29%)	ESFT Samples (Localized)	CGH	Brisset et al. (2001)
		7/45 (16%)	ESFT Samples (Localized)	Karyotyping (G-Band)	Roberts et al. (2008)
		9/31 (29%)	ESFT Samples	CGH	Savola et al. (2009)
	12p	9/52 (17%)	ESFT Samples (Primary, includes 20 Metastatic)	CGH	Ozaki et al. (2001)
	12q	11/52 (21%)	ESFT Samples (Primary, includes 20 Metastatic)	CGH	Ozaki et al. (2001)
	12q12-q15	6/8 (75%)	Cell line	CGH	Shing et al. (2002)
	12q13.2-q14.1	9/31 (29%)	ESFT Samples	CGH	Savola et al. (2009)
	12q14.1-q15	2/9 (22%)	ESFT Samples vs. cell line xenografts	SNP Microarray (Affy 100K)	Neale et al. (2008)
	14	9/100 (9%)	ESFT Samples (Primary, Localized + Metastatic)	Karyotyping and CGH	Hattinger et al. (2002)
	14q	3/20 (15%)	ESFT Samples (17 patients, two specimens from same pt. in 3 cases)	CGH	Armengol et al. (1997)
	14q11.2	2/9 (22%)	ESFT Samples vs. cell line xenografts	SNP Microarray (Affy 100K)	Neale et al. (2008)
	17q21.31-q25.3	6/9 (67%)	ESFT Samples vs. cell line xenografts	SNP Microarray (Affy 100K)	Neale et al. (2008)
	18	3/25 (12%)	ESFT Samples (23 Primary, 2 Relapse)	CGH	Ferreira et al. (2008)
	20	12/92 (13%)	ESFT Samples (Primary, Localized + Metastatic)	Karyotyping and CGH	Hattinger et al. (2002)
		3/25 (12%)	ESFT Samples (23 Primary, 2 Relapse)	CGH	Ferreira et al. (2008)
		5/28 (18%)	ESFT Samples (Metastatic)	Karyotyping (G-Band)	Roberts et al. (2008)
		5/45 (11%)	ESFT Samples (Localized)	Karyotyping (G-Band)	Roberts et al. (2008)
		6/21 (29%)	ESFT Samples (Localized)	CGH	Brisset et al. (2001)
		8/52 (15%)	ESFT Samples (Primary, includes 20 Metastatic)	CGH	Ozaki et al. (2001)
	20p	10/52 (19%)	ESFT Samples (Primary, includes 20 Metastatic)	CGH	Ozaki et al. (2001)
	20q	10/52 (19%)	ESFT Samples (Primary, includes 20 Metastatic)	CGH	Ozaki et al. (2001)
	20q11.23-q13.33	2/9 (22%)	ESFT Samples vs. cell line xenografts	SNP Microarray (Affy 100K)	Neale et al. (2008)
	21q22.3	2/9 (22%)	ESFT Samples vs. cell line xenografts	SNP Microarray (Affy 100K)	Neale et al. (2008)
	22q11.21	2/9 (22%)	ESFT Samples vs. cell line xenografts	SNP Microarray (Affy 100K)	Neale et al. (2008)
1p		3/10 (30%)	ESFT Samples (Relapse)	CGH	Ozaki et al. (2001)
		4/42 (10%)	ESFT Samples (Primary, Metastatic)	Karyotyping and CGH	Hattinger et al. (2002)
		6/80 (8%)	ESFT Samples (Primary, Localized)	Karyotyping and CGH	Hattinger et al. (2002)
1p13-1p36.3		10/125 (8%)	ESFT Samples (Primary, Localized + Metastatic)	Karyotyping and CGH	Hattinger et al. (2002)
1p36.32-p36.11		2/9 (22%)	ESFT Samples vs. cell line xenografts	SNP Microarray (Affy 100K)	Neale et al. (2008)
3p		3/8 (38%)	Cell line	CGH	Shing et al. (2002)
7q11.2		5/25 (20%)	ESFT Samples (23 Primary, 2 Relapse)	CGH	Ferreira et al. (2008)
9p		7/31 (23%)	ESFT Samples	CGH	Savola et al. (2009)
9p21		6/42 (14%)	ESFT Samples (Primary)	MLPA	Brownhill et al. (2007)
		6/9 (67%)	ESFT Cell lines	MLPA	Brownhill et al. (2007)
		5/9 (56%)	ESFT Samples vs. cell line xenografts	SNP Microarray (Affy 100K)	Neale et al. (2008)
10		4/25 (16%)	ESFT Samples (23 Primary, 2 Relapse)	CGH	Ferreira et al. (2008)
		7/52 (13%)	ESFT Samples (Primary, includes 20 Metastatic)	CGH	Ozaki et al. (2001)
16q		10/31 (32%)	ESFT Samples	CGH	Savola et al. (2009)
		11/52 (21%)	ESFT Samples (Primary, includes 20 Metastatic)	CGH	Ozaki et al. (2001)
		12/78 (15%)	ESFT Samples (Primary, Localized)	Karyotyping and CGH	Hattinger et al. (2002)
		13/41 (32%)	ESFT Samples (Primary, Metastatic)	Karyotyping and CGH	Hattinger et al. (2002)
		3/28 (11%)	ESFT Samples	CGH	Tarkkanen et al. (1999)
		4/25 (16%)	ESFT Samples (23 Primary, 2 Relapse)	CGH	Ferreira et al. (2008)
		5/8 (63%)	4 cell line, 1 primary culture	CGH	Shing et al. (2002)
		8/45 (18%)	ESFT Samples (Localized)	Karyotyping (G-Band)	Roberts et al. (2008)
16q22.3		5/9 (56%)	ESFT Samples vs. cell line xenografts	SNP Microarray (Affy 100K)	Neale et al. (2008)
17p		8/52 (15%)	ESFT Samples (Primary, includes 20 Metastatic)	CGH	Ozaki et al. (2001)
17p13-p11.2		4/8 (50%)	3 cell line, 1 primary culture	CGH	Shing et al. (2002)
19		4/25 (16%)	ESFT Samples (23 Primary, 2 Relapse)	CGH	Ferreira et al. (2008)
19p		6/52 (12%)	ESFT Samples (Primary, includes 20 Metastatic)	CGH	Ozaki et al. (2001)
19q		10/52 (19%)	ESFT Samples (Primary, includes 20 Metastatic)	CGH	Ozaki et al. (2001)
Y		3/5, Males (60%)	Cell line	CGH	Shing et al. (2002)

Copy number studies may also have prognostic value. For example, gains of the entire chromosome 8 have been reported by nearly every CGH study and occur in as many as 23-80% of Ewing's samples (Armengol et al., 1997; Brisset et al., 2001; Ferreira et al., 2008; Hattinger et al., 2002; Maurici et al., 1998; Ozaki et al., 2001; Roberts et al., 2008; Savola et al., 2009), with the highest frequency found in metastatic and relapsed tumors. Additionally, gain of 1q or independent loss of 16q was found to predict worse overall and event-free survival, regardless of localized or metastatic disease at diagnosis (Hattinger et al., 2002). This same study showed chromosome 12 gain to be associated with adverse event-free survival in patients with localized disease at diagnosis. Deletion of 6p has also been associated with worse outcome in diagnostic Ewing's sarcoma samples (Tarkkanen et al., 1999).

Total ploidy and absolute number of CNAs have also been shown to correlate with clinical outcome in Ewing's sarcoma. Patients with primary Ewing's tumors with 3 or less CNAs were reported to do better than patients with more than 3 copy number changes in two different studies (Ferreira et al., 2008; Savola et al., 2009), whereas another group found that 5 CNAs could stratify patients (Ozaki et al., 2001). Taken together, it appears that increasing genomic instability in Ewing's sarcoma is associated with worse outcome.

Summary and Conclusions

Ewing's sarcoma is a complex disease that requires the coordination of many events to arise. The enigmatic cell of origin that provides a permissive environment for the complex transcriptional dysregulation mediated by EWS/FLI, combined with cooperation of parallel pathways that may differ from case to case make research and treatment of this disease a challenge. Better insight into the cell of origin and the cancer stem cell potential in Ewing's sarcoma will provide a model in which to study this malignancy. This should afford researchers with a more comprehensive list of proteins necessary for Ewing's sarcoma tumorigenesis and an opportunity to understand the contribution of each. The mechanism by which EWS/FLI up- or downregulates these critical genes remains an important question. Finally new technology will allow for the identification of mutations that cooperate with EWS/FLI and its targets to bypass the many mechanisms in place to prevent uncontrolled proliferation and other cancerous phenotypes. Research in all of these areas will lead to a more comprehensive model of Ewing's sarcoma genesis and disease progression which may inform the study of other cancers caused by ETS protein dysregulation. This improved molecular insight can also be incorporated into clinical trials that will eventually provide more targeted treatment for this disease.