Molecular aspects of Ewing’s sarcomas

Introduction

Since the mid-1990s when Ewing's sarcoma started to be studied at the molecular level, a very large number of investigations aimed at deciphering the biology of this disease, particularly by exploring the role of the Ewing's sarcoma–specific oncogene. Even though the conducted research increased our knowledge, a large number of questions remain, such as the cell of origin of Ewing's sarcoma, precise functional roles of the fusion proteins, and modulated genes and signaling pathways. Likewise questions on the relationship of Ewing's sarcoma and its tumor (micro)environment are still poorly understood, and mechanisms of disease progression remains largely unexplored. Thus, this chapter will describe our current understanding of some of these key aspects on the biology of Ewing's sarcoma.

Ewing's sarcoma oncogenes

In 85% of Ewing's sarcoma and primitive neuroectodermal tumors, analysis of cellular DNA revealed translocations between chromosome 22 and chromosome 11 [ ]. On chromosome 22, this translocation always fuses at the N-terminal part of EWSR1 (Ewing's sarcoma breakpoint region 1) in frame to the C-terminal part of FLI1 (Friend leukemia virus integration site 1), leading to the pathognomonic EWSR1-FLI1 oncogenic protein [ ]. Chromosomal breakpoints can differ between tumors, leading to different fusion proteins, linking exons 7–10 of EWSR1 to exons 5–8 of FLI1 [ ], the most prevalent one (EWSR1-FLI1 type I) joining exon 7 of EWSR1 to exon 6 of FLI1 . EWSR1-FLI1 fusion proteins can also differ depending on splice variations [ ]. On the chromosome 11 derivative, on which FLI1 is fused to EWSR1 , the resulting fusion is rarely expressed. Most often the reciprocal fusion is either not expressed due to the very low activity of the FLI1 promoter in Ewing's sarcoma, not in frame or lost due to a wild-type chromosome 11 isodisomy. The involvement of EWSR1 in human oncogenesis is not limited to its fusion with FLI1. Indeed, in Ewing's sarcoma, this gene can also be fused to other genes encoding members of the ETS family (the most common being ERG in roughly 10% of the cases, and ETV1 , ETV4 , or FEV in less than 1% of cases each [ ]). Tumors harboring non-ETS-containing fusions with EWSR1 , such as EWSR1-ATF1 or EWSR1-CREB1 found in angiomatoid fibrous histiocytomas (AFH), EWSR1-PATZ1 in round cell sarcomas, EWSR1-TFPC2 a subtype of rhabdomyosarcomas [ ], or EWSR1-WT1 found in desmoplastic small round cell tumors (DSRCTs), constitute distinct entities that should not be confused with Ewing's sarcoma [ ].

EWSR1 is a protein whose functions are yet to be clarified. It contains an RNA-binding domain in its carboxy-terminal portion and a transcriptional regulatory domain in its amino-terminal portion. EWSR1 is part of the small “FET” family of proteins comprising FUS (fused in liposarcoma) [ , ] and TAF15 (TATA-binding protein–associated factor 15) [ ], three highly homologous proteins. Sarcomas with FUS-FEV fusions are exceedingly rare, but cluster on the transcriptomic level with EWSR1-ETS -positive Ewing's sarcomas and as such should be classified as Ewing's sarcomas as well [ ].

Interestingly, using whole-genome sequencing data, a recent study demonstrated that the pathognomonic EWSR1-ETS fusions found in Ewing's sarcoma can arise through two distinct mechanisms [ ]: The first comprises a balanced chromosomal translocation, the second a more complex reshuffling of chromosome parts involving multiple chromosomes, a process termed chromoplexy. While only around 40% of all EWSR1-FLI1 fusions arise via chromoplexy, all EWSR1-ERG fusions derive from this process, which is related to the strand orientation of ERG [ ].

The FLI1 protein is part of the ETS family of transcription factors defined by the presence of a highly conserved domain responsible for binding onto specific DNA target sequences [ ]. The EWSR1-FLI1 fusion protein, in which the RNA-binding domain of EWSR1 is replaced by the ETS DNA-binding domain of FLI1, then possesses the ability to bind DNA and is therefore considered as an aberrant transcription factor. Experiments on model promoters showed that EWSR1-FLI1 has a transcriptional activation potential stronger than FLI1, leading to the supposition that the fusion protein exerts its oncogenic potential through the transcriptional activation of target genes [ ]. However, experiments conducted in vivo on the TGF-β receptor type II showed that EWSR1-FLI1 was also a potent transcriptional repressor [ ]. These studies demonstrated that EWSR1-FLI1 is able to either activate or repress transcription. Most interestingly, in a heterologous NIH3T3 cells overexpression system, EWSR1-FLI1 ETS domain mutants lacking DNA-binding ability were shown to retain some transforming properties, suggesting that the oncogenic properties of EWSR1-FLI1 were not solely due to its capacity to bind DNA [ ]. Protein–protein interactions may thus also be of prime importance for EWSR1-FLI1 activities [ ].

EWSR1 and the two other members of the FET family are expressed in all tissues, localized in the nucleus [ ], and involved in many fusion genes associated with cancer. In each case, the N-terminus of EWSR1, FUS, or TAF15 is fused to the DNA-binding domain of a transcription factor, leading to the expression of an aberrant transcription factor. Many publications described different potential roles of these proteins and their implications in various processes such as transcription, splicing, metabolism, and transport of RNA (mRNA and microRNA), cell response to stress, as well as in the maintenance of genomic integrity. Indeed, EWSR1 can interact with proteins involved in splicing [ ] and regulates alternative splicing of genes involved in DNA repair and genotoxic stress signaling [ ]. EWSR1 is also involved in DNA repair mechanisms during meiosis and B lymphocyte development [ ] and may play a role in miRNAs maturation according to its presence in the microprocessor complex [ ].

Does EWSR1-FLI1 need to be activated for being oncogenic? Probably, as several posttranslational modification altering EWSR1-FLI1 function have been identified. Indeed, the EWSR1 part that is conserved in the fusion protein contains several potential phosphorylated tyrosine residues [ ] and a threonine at position 79 that is phosphorylated depending on DNA damage signaling [ ]. A protein kinase C phosphorylation site at Ser266 [ ] that may be important for EWSR1-FLI1 oncogenic properties [ ] has been described, but is not conserved in the prevalent EWSR1-FLI1 type I fusion. Moreover, glycosylated (O-linked N-acetylglucosamine) residues have been identified in the EWSR1 moiety, which were dependent on the phosphorylation status of EWSR1-FLI1 and were linked to transcriptional activities [ ]. Finally, it has also been proposed that acetylation on several residues of the C-terminus FLI1 moiety may influence the binding properties of EWSR1-FLI [ ].

How EWSR1-FLI1 is regulated is still unknown. However, recent studies demonstrated differences in EWSR1-FLI1 expression levels among tumor cells in the same tumor. Single-cell RNA-sequencing of Ewing's sarcoma cell lines revealed specific expression signatures correlated to EWSR1-FLI1 expression level, such as an increased hypoxia pathway in cells expressing low levels of the fusion protein [ ]. Heterogeneity of EWSR1-FLI1 expression may have a dramatic role in tumor evolution. Indeed, Franzetti et al. demonstrated that cells expressing low levels of EWSR1-FLI1 were more prone to migrate and evade the boundaries of 3D culture spheroids than EWSR1-FLI1 ^high cells [ ]. These results should be taken into account when pursuing new therapeutic approaches attempting to directly inhibit EWSR1-FLI1 activity.

While our knowledge of Ewing's sarcoma oncogenes is still fragmentary, our understanding of its biology is challenged with the descriptions of small round cell tumors with a “Ewing-like” appearance but with atypical fusion proteins whose functions may differ from “classic” translocations [ ] (see Chapter 42 ).

The cell of origin of Ewing's sarcoma

From the histological point of view, Ewing's sarcoma tumors comprise a phenotypic spectrum, ranging from completely undifferentiated small round cells to a pronounced neuronal character of peripheral primitive neuroectodermal tumors (pPNETs). Since the proposition by James Ewing in 1921 of an endothelial source [ ], the cellular origin of Ewing's sarcoma has remained controversial. The identification of the Ewing's sarcoma–specific oncogenic events in pPNET [ , ] led to hypothesis that cells derived from the neural crest could be the origin of these tumors. This theory was reinforced by the fact that Ewing's sarcoma cell lines are able to implement a program of neural differentiation under certain culture conditions [ , ] and that Ewing's sarcoma cells express a variety of genes and proteins found in neural tissues [ ]. Nevertheless, numerous neural genes were shown to be transcriptionally activated by EWSR1-FLI1 [ , ], indicating that the neural phenotype observed in Ewing's cell may not be indicative of the tissue of origin but rather a consequence of EWSR1-FLI1 expression [ ]. In support of this notion, human neural crest–derived mesenchymal stem cells (NCSCs) with ectopic EWSR1-FLI1 expression only partially recapitulated the typical transcriptomic signature of Ewing's sarcoma [ ].

Interestingly, at the exception of few stem cells, no untransformed cell type tested so far survives ectopic EWSR1-FLI1 expression and rather undergoes apoptosis. This strongly indicates that few cell types are permissive to EWSR1-FLI1. Accordingly, ectopic expression of EWSR1-FLI1 inhibits osteoclast and adipocyte differentiation of mice bone marrow stromal cells [ ] as well as the differentiation of C2C12 myoblast cells [ ]. In addition, upon ectopic expression of the specific Ewing's sarcoma oncogene, murine bone marrow cells are able to give rise to tumors resembling Ewing's sarcoma when being reinjected into mice [ , ]. A number of experiments have indicated that MSCs are permissive to EWSR1-FLI1, probably revealing the cellular origin of Ewing's sarcoma cells. With the advent of small interfering RNA allowing the inhibition of EWSR1-FLI1 in tumor cells, demonstration has been made that inhibited Ewing's sarcoma cells could reexpress markers that are specific to their tissue of origin and furthermore retrieve, at least in part, their original phenotype. Indeed, comparisons of expression profiles from a variety of different cell types and tissues with those of siRNA-treated Ewing's sarcoma cells demonstrated that EWSR1-FLI1 inhibited Ewing's sarcoma cells closely resemble to MSCs [ ]. These in silico analyses were confirmed by in vitro differentiation experiments, in which inhibited Ewing's sarcoma cells were able both to induce the expression of specific MSC surface markers and to differentiate into adipocytes and osteocytes, two lineages specific of MSCs [ ]. Interestingly, expression profiles of “pediatric” MSCs (derived from 6- to 14-year-old individuals), in which the Ewing's sarcoma oncogene is expressed, resemble more Ewing's sarcoma cell lines than adult transformed MSCs do [ ]. This result may be indicative of the importance of MSCs' stemness/genetic background/state of differentiation for EWSR1-FLI1 transformation but also of the lack of a strict definition of MSCs, as each study used different criteria for the definition of MSCs and therefore different sources of cells.

Finally, what is really the cell of origin of Ewing's sarcoma between MSCs and NCSC? Actually, both hypotheses could coexist. Indeed, there is a growing body of evidence that MSC could arise from multiple developmental origins, including neural crest. As a matter of fact it has been shown that the neuroepithelium could give rise to MSCs in part via a neural crest intermediate phase [ , ] and that NCSC could be found in bone marrow together with MSCs [ ]. Thus, a reconciling hypothesis is that the cell of origin of Ewing's sarcoma may be an MSC-derived from the neural crest that remains in the bone marrow. Further investigations are needed to definitively clarify this point.

In conclusion, the origin of Ewing's sarcoma is not as uncertain as it was a few years ago, but still remains debated. The only certitude is that the cell of origin, either MSCs or NCSC, has to possess peculiar features to be transformed by the Ewing's sarcoma oncogene.

Other genetic events

EWSR1-FLI1 is a proven oncogene being able to transform NIH3T3 [ ]. Nevertheless, EWSR1-FLI1 is not able to transform MSCs [ ] or NCSC [ ] unless acquirement of another genetic event, like TP53 mutations [ ]. This strongly suggests that EWSR1-FLI1 is not the only genetic event to be causing Ewing's sarcoma; in other words, EWSR1-FLI1 needs other genetic events or a permissive background for being able to transform a cell. Genome-wide analysis of copy number alteration and/or expression profiles as well as sequencing of key cancer genes (oncogenes or tumor suppressor) have identified genetic abnormalities of prognostic significance. Indeed, the losses or mutations of CDKN2A [ ] and TP53 [ , ] genes, the gain of chromosomes 1q [ ], the burden of copy number alterations [ , ], as well as expression level of some genes were associated with an unfavorable prognosis. Nevertheless, it is not known if these events are required for EWSR1-FLI1-mediated transformation, if they are acquired during tumor development or if they are a necessary evolution for metastatic spread. Nevertheless, as described in recent literature, Ewing's sarcoma demonstrated a strikingly low number of genetic alterations [ ]. Nevertheless, whole-genome sequencing of large series of tumors has led to the discovery of STAG2, a member of the cohesin complex, as an important player in the metastatic evolution of Ewing's sarcomas [ ]. Indeed, mutation in STAG2 gene is found in 15%–20% of tumors and is linked to unfavorable prognostic, especially in conjunction with TP53 mutations [ ]. In a recent publication, Surdez et al. demonstrated that STAG2 loss-of-function decreases CTCF-mediated promoter-enhancer interactions leading to increased migratory and invasive properties that are associated with the aggressive behavior in Ewing sarcoma .

The identification of several susceptibility loci shed new lights on Ewing's sarcoma development. The epidemiological distribution of Ewing's sarcoma is remarkable: the vast majority of cases have been reported in populations of European descent/origin while very few cases have been described in the African or Asian populations, independently of their geographical localization [ ]. Moreover, despite a low incidence rate (0.15 case/100,000 individuals) [ ] and the lack of familial cases of Ewing's sarcoma, rare cases of sibling pairs with Ewing's sarcoma have been described. These observations suggested the presence of susceptibility factors. Through an initial genome-wide association study (GWAS), three loci associated with the risk of developing Ewing's sarcoma were identified on chromosomes 1, 10, and 15 [ ]. Although the true causative genetic variations still remain to be identified, it appears that modulations in the expression of some genes present in close vicinity of these regions may be involved in Ewing's sarcoma development. Interestingly, one of the genes that is close to the chromosome 1 locus is TAR DNA-binding protein (TARDBP or TDP-43), a gene involved in amyotrophic lateral sclerosis [ ] and which is closely related to FUS, a member of the FET family. It has been postulated that FUS (and the other FET proteins) and TARDBP could play similar roles in many different processes [ ]. In fact, interaction and/or competition of EWSR1-FLI1 with the FET family members has become a growing area of interest since EWSR1-FLI1 activity may also happen through dominant effect over the native FET protein functions. Thus a deregulation of TARDBP expression (even modest), as found in patients with the surrogate at-risk SNPs allele, may directly influence EWSR1-FLI1 activity. A second interesting gene is early growth response 2 (EGR2 or KROX20) which is located within the chromosome 10 susceptibility locus. EGR2 is a transcription factor that could modulate numerous genes and is associated to Charcot–Marie–Tooth disease type 1D [ ]. With respect to Ewing's sarcoma's biology, EGR2 is upregulated by EWSR1-FLI1 and is highly expressed in Ewing's sarcoma tumors. Most interestingly, EGR2 has been shown to regulate osteoprogenitors and bone formation [ , ]. A subsequent study has shown that a conserved A/T SNP in linkage disequilibrium (LD) with the previously identified sentinel SNPs of the GWAS contributed to EGR2 overexpression in Ewing's sarcoma [ ]. The risk-allele (A) of this SNP connects two stretches of an interrupted EGR2 -associated GGAA-microsatellite and allows a more efficient binding of EWSR1-FLI1, which then leads to higher EGR2 expression that in turn contributes to Ewing's sarcoma tumorigenesis in preclinical models [ ]. Interestingly, a second, more comprehensive GWAS firmly validated the susceptibility loci near TARDBP and EGR2 , but could identify several additional risk loci near BMF , NKX2-2 , KIZ , and RREB1 [ ]. Strikingly, although formal experimental proof is still lacking, all new susceptibility loci are enriched in EWSR1-FLI1-bound GGAA-microsatellites that contain potential regulatory SNPs being in LD with the respective lead SNPs [ ]. These findings imply that the interplay of EWSR1-FLI1 with germline SNPs located in GGAA-microsatellites may have a pivotal role in Ewing's sarcomagenesis and perhaps interindividual heterogeneity of clinical outcomes [ ].