Progress in the Molecular Biology of Ewing Tumors

Purpose/results/discussion. Rearrangement of the EWS gene with an ETS oncogene by chromosomal translocation is a hallmark of the Ewing family of tumors (EFT). Detectability, incidence, tumor specificity and variability of this aberration have been matters of intense investigation in recent years. A number of related alterations have also been found in other malignancies. The common consequence of these gene rearrangements is the generation of an aberrant transcription factor. In EFT, the ETS partner is responsible for target recognition. However, synergistic and possibly tissue-restricted transcription factors interacting with either the EWS or the ETS portion may influence target selection. Minimal domains of both fusion partners were defined that have proved necessary for the in vitro transformation of murine fibroblasts. These functional studies suggest a role for aberrant transcriptional regulation of transforming target genes by the chimeric transcription factors. Also, fusion of the two unrelated protein domains may affect overall protein conformation and consequently DNA binding specificity. Recent evidence suggests that EWS, when fused to a transcription factor, interacts with different partners than germ-line EWS. Variability in EWS–ETS gene fusions has recently been demonstrated to correlate with clinical outcome. This finding may reflect functional differences of the individual chimeric transcription factors. Alternatively, type and availability of specific recombinases at different time-points of stem cell development or in different stem cell lineages may determine fusion type. Studies on EFT cell lines using EWS–ETS antagonists do suggest a rate-limiting essential role for the gene rearrangement in the self-renewal capacity of EFT cells. The presence of additional aberrations varying in number and type that may account for immortalization and full transformation is postulated. Knowledge about such secondary alterations may provide valuable prognostic markers that could be used for treatment stratification.


Introduction
Ew ing' s sarcom a (ES), being a rare m alignant disease affe cting bone and soft tissue in children and young adults, was hardly known to people other than pediatric oncologists until the characterization of a chim eric gene product presum ed to be causally involved in the generation of this neoplasm . It dram atically gained attention w hen, from investigating other m alignancies, it becam e apparent that the ES-derived oncoprotein constitutes the prototype of a whole class of aberrant proteins speci® cally asso ciated with certain tum or types. C onsequently, ES m ay be considered a m odel system to study m alignant conversion on a subclinical level. T he discovery of the ES-asso ciated gene rearrangement transiently halted a controversy among pathologists about the existence of distinct categories of ES (i.e. osseous ES, extra-sk eletal ES, Askin tum or, peripheral prim itive neuroectoderm al tum or) because it was found to be expressed in all of them . Clinically, however, there is a need for diversi® cation.
Although m ore than half of the patients can be cured by m ultimodal therapy, one third of cases with localized disease and about 80% of patients presenting with m etastases succum b to the disease (for a recent review, see K ovar et al. 1 ). Current treatment protocols have largely com pensated for classical prognostic m arkers such as tumor volum e and localization of the prim ary except for the rather unfavo rable presence of metastases at diagnosis. It is likely, therefore, that biological differences exist between so far incurable aggressive disease and clinically m anageable localized disease inexplicable by the m ere presence of the ES-assoc iated gene rearrangement. W hile Ewing' s tum or research has focused on the clinical exploitability and the function of the ES-spe ci® c gene rearrangem ent since its discovery in 1992, this review will also consider extensively the role of additional m olecular aberrations in the search for useful prognostic m arkers. N eoplastic transform ation and m etastatic spread is com m only believed to result from a m ulti-step pro-H. Kovar cess. In this context, the ES-spe ci® c gene rearrangement obviously constitutes a rate-lim iting event. According to Knudson' s legendary tw o-hit hypothesis, at least one additional aberration should be present in a Ewing tum or. It is possible that this second hit is less speci® c and affects different genes at different times during developm ent of the enigmatic Ewing tumor stem cell, thus de® ning distinct subcategories of the disease. C onsequently, Ew ing tum or research is slowly m oving towards m olecular subclassi® cation and staging.

D iagnostic tools
In 1988, the cytogenetic translocation t(11;22) (q24;q12) was described as speci® cally asso ciated w ith histopathologically diagnosed ES and peripheral prim itive neuroectoderm al tum or (pPN ET). 2 T he presence of this aberration in a largely undifferentiated small round cell tum or of childhood turned out to be a form idable diagnostic m arker. 3,4 H ow ever, cytogenetic analysis was restricted to tum or cells with at least lim ited in vitro proliferation potential. T he generation of an antibody, H BA71, 5 speci® cally reacting with the surface glycoprotein encoded by the M IC2 gene, 6 which w as found to be abundantly present in tum or cells carrying a chrom osom e 22q12 aberration, 7,8 enlarged the spectrum of diagnostic tools. H owever, em bryonal rhabdomyosarcom as, asterocytom as, neuroendocrine tum ors and carcinom as occasionally stained positive with HBA71 9 and, w hen using a m ore sensitive antibody (12E7 10 ), high level expression of this antigen was also noted in early hem atopoietic precursor cells 11 and several lymphom as. 12 The characterization of the ES break-po int regions on chrom osom es 22 and 11 13 and the subsequent cloning of a chim eric cD N A resulting from a gene fusion between a novel gene, designated E W S, and the E TS transcription factor gene FLI1 14 allowed for sensitive detection of tum or cells carrying a 11;22 translocation even in sm all sam ples of fresh, frozen or paraf® nem bedded m aterial by m eans of reverse transcriptase polym erase chain reaction (RT -PC R). 15± (Fig. 1). T LS w as identi® ed as a heterogenous nu clear ribon ucleoprotein (hnRN P) in nonspliceosom al com plexes on m R N A continuously shuttling betw een the nucleus and the cytoplasm 45 and SA RFH w as found to be asso ciated w ith regions of the D rosophila chrom atin transcribed by R N A polym erase II. C onsistent w ith a role of E W S fam ily m em bers in gene transcription, hT A FII68, T L S and E W S have been identi® ed in subpopu lations of the general transcription factor T F IID . 126 H owever, recent evidence suggests that the oncogenic derivatives of T L S an d E W S are not stably associated w ith the RN A polym erase II com plex and T F IID . 12 6 Accum ulation in nuclear inclusions such as the coiled bod y and the nucleolus have been reported for Pigpen, 46 the 69-kD a snRN Passociated protein, 37 and, after transcriptional inhibition, for T L S. 47 Such nuclear subcom partm ents m ight either constitute the site of norm al function of these EW S-related proteins or serve as their reservoir. Interestingly, on coproteins that contain the am ino terminal dom ain of EW S or T LS are also targeted to the sam e structure 47 . So far, the functional relevance of this ® nding is com pletely unknown.
A role for EW S and its partner genes in determining the EFT phenotyp e Figure 2 sum m arizes all known gene fusions involving either EW S or TLS in human malignancies. In an N IH 3T3 transform ation study, the type of transcription factor contributing to the chim eric gene product determined cell m orphology. 35 T his observation m ight in part explain w hy only m em bers of the ET S transcription factor fam ily are found in gene fusions with EW S associated w ith an EFT phenotype. H ow ever, w hile EW S and TLS am ino termini appear to be functionally interchangable w hen fused to the transcription factor CHO P in the in vitro m odel, 35 as well as in m yxoid chondrosarcom a, 33,34,48 T LS has never been found to replace EW S in EF T. In contrast, fusion of T LS to the ET S fam ily m ember ERG, which is involved in 10% of EF T, has been reported for poor prognosis, t(l6;21) positive, acute m yeloid leukem ia. 49± 51 Rearrangem ent of E W S with other transcription factor genes such as ATF1, the W ilm s' tum or gene W T1 and the nuclear receptor C HN /TEC have been shown to be asso ciated w ith m alignant m elanom a of soft parts, desm oplastic sm all round cell tumor and myxoid chondrosarcom a, respectively. 52± 55 T hus, it is the speci® c com bination of EW S w ith a subset of ETS transcription factor genes and/or a particular stem cell in which these genes are sensitive to illegitimate recom bination that determ ine the EFT phenotype. Accessibility to rearrangement by an as yet unde® ned recom binase m ight also determine the incidence of EFT . Z ucm an et al. reported recently that sequence analysis of the entire EW S intron 6 region close to the m ajor break-point region in EFT from Caucasian origin revealed a very high density of Alu elements resulting from repeated retroposition during evolution. 56 T he Alu fam ily of short interspersed repetitive D N A elements has previously been dem onstrated to be frequently involved in hum an gene rearrangements. 57 T his region was found to be reduced by 50% due to deletion in the African population. This inter-ethnic polymorphism in the EW S gene is accom panied by a striking difference in the incidence of EF T between populations of European and African origin. 58,59 It should be noted that the m ajority of EW S genomic breakpoints occur in intron 7 and that intron 6 is, in fact, never directly rearranged in EF T. So far, only three E W S genom ic break-po ints have been sequenced, two in EF T and one in a desm oplastic sm all round cell tum or, 60± 62 , none of which contained Alu elem ents in the im mediate vicinity of the rearrangem ent sites. T hus, direct proof for the involvem ent of Alu elements in E W S translocation is not availab le.
In the published cases, the lack of uniform ity of sequences affe cted by the gene rearrangem ent does not allow the identi® cation of a speci® c recom binase responsible for the translocation. Chromosom e 22 alteration m ay occur as the only cytogenetically visible aberration in EF T suggesting that the EW S± E TS gene rearrangem ent is not the consequence of a general genomic destabilization. However, the frequent involvem ent of m ore than two chrom osom es in complex chrom osom e 22 aberrations and evidence for deletion of considerable am ounts of sequences from the directly involved genes on the untranscribed counterpart of the derivative chrom osom e 22 60 im ply a complex m echanism for gene rearrangement in EFT. In addition, w hile EW S and FLI1 are equally oriented on the long arm s of chrom osom es 22 and 11 from the centrom ere to the telom ere allow ing for sim ple reciprocal transloca-tion, the E RG gene is oriented in the opposite direction. Consequently, EW S± ERG gene fusions m ay result from either interstitial deletion/insertion m echanism s 63 or from com plex genom ic rearrangem ents involving additional chrom osom es. 24 Although a high variability in EW S fusion partners and genom ic break-point locations has been noted, rearrangement of EW S intron 7 w ith intron 5 or 4 of the ETS fam ily gene FLI1 predom inates (about 80% of EF T cases). 16,18,24 Interestingly, the only known three cases of fusion between EW S and the ETS transcription factor gene FE V involve E W S intron 10 which is otherw ise affe cted in only 9% of EFT . Since, as outlined later, the m inim al portions of EW S and its fusion partners contained in all EFT -derived oncoproteins and required for full in vitro transfo rm ation and transcription activation function are signi® cantly sm aller than the portions present in the m ost frequently observed EW S fusions, it is unlikely that in EFT rearrangement sites are determined by functional constraints only. Rather, genom ic structure and accessibility m ight direct illegitimate recom bination to speci® c regions in the involved genes. G enom ic accessibility and availability of recom binases m ight vary during the developm ent of speci® c stem cell lineages. Therefore, it cannot be excluded that the different E W S rearrangements de® ne different histogenetic starting points of EF T developm ent. This m odel would provide an intriguing explanation for our recent observation of prognostic differences in EFT correlating with different EW S fusion types. 18 Alternatively, the various chim eric oncoproteins m ight display functional differences. Since the com plete EW S and FLI1 genes have been cloned and sequence inform ation is readily availab le, 64± 65 the description of m ore genom ic rearrangement points in EFT will hopefully throw m ore light on the m echanism of gene rearrangem ent in this disease.
The ETS partner in the E W S fusion gene T he ET S transcription factor fam ily currently counts m ore than 30 mem bers. It is characterized by the presence of a unique D N A binding domain which is highly conserved from¯ies to hum ans and has been ® rst described for the viral oncogene v-ets of avian erythroblastosis virus E 26 (E twentysix speci® c). Several E TS subfam ilies can be de® ned on the basis of evolutionary sequence conservation. In ET , 95% of cases show EW S fusion to the FLI1 (Friend leukemia virus integration site 1)/ERG (ET S related gene) subfam ily of transcriptional activators. 16,18 These two gene products share, in addition to alm ost identical D N A binding dom ains, a 16-am ino acid stretch im m ediately upstream of the ETS dom ain which is 100% conserved betw een X enopus and humans, suggesting an im portant but as yet unidenti® ed functional role. 66 T his portion is retained in alm ost all EW S± FLI1 and EW S± ERG ET S1, G ABPa , and F LI1 have recently been demonstrated to bind to the transcription factor PAX 5 in vitro. 78 A¯anking PAX5 binding site allowed for ET S1 binding to an im perfect ET S recognition m otif indicating that cooperativity m ight change ET S binding speci® city to som e extent. Som e ET S proteins, including ETS1, ETS2, F LI1, ERG , G ABPa and TEL, share a further conserved region¯anking the am ino terminal transactivation dom ain, refered to as the`pointed or B-dom ain' . T his structure has been show n to de® ne a speci® c oligom erization interface. For T EL, it governs hom otypic aggregation. 79 N o interaction partner has been de® ned for FLI1 and ERG so far. The conserved fold of the am ino terminal dom ain of ET S proteins is, however, likely to underlie a conserved function. T hus, replacem ent of this portion by the EW S am ino term inus may alter not only quantitatively but also qualitatively the transcriptional activation properties of these two ETS fam ily m em bers by placing them into a different protein context. In addition, several ET S fam ily m embers have been demonstrated to be a target for the RAS/RAF M AP kinase signaling pathway. For ET S1, ras regulation involves phospho rylation of residues within the am ino terminal`pointed' dom ain. So far, no link between this signalling pathw ay and FLI1 or ERG has been reported. Consequently, it rem ains unclear if fusion to EW S w ill uncouple FLI1/ERG target gene regulation from extracellular signaling.
In about 1% of ET cases, EW S is rearranged with FEV (® fth Ew ing' s tum or varian t) on chrom osom e 2. 21 T his ETS fam ily m ember displays 90% identity to FLI1 in the D N A binding dom ain but the F LI1and E RG -sp eci® c¯anking 16 am ino acid sequence is m issing in this protein. Interestingly, F EV also lacks an am ino terminal transactivation dom ain present in m ost other ETS fam ily m embers. Instead, it carries a long C-term inus w hich, because of the presence of abundant alanine residues, m ight serve as a putative repressor dom ain. H owever, experim ental proof for such an activity is not yet available. Since this portion is retained in the EW S fusion, it rem ains to be established if the gene rearrangement w ould result in a functional conversion to an activator. Interestingly, germ -line FE V cD N A has been cloned from an E W S± FLI1 expressing Ew ing tum or cell line, indicating coexpression of the two genes w ithin the sam e cell. If E W S± FLI1 and FEV target the sam e genes and FEV operates as a repressor, it is possible that the EFT gene rearrangement results in the release of these genes from transcriptional inhibition by com petitive binding.
So far, three cases of suspected EF T have been reported in w hich EW S w as fused to ETS fam ily genes of a different subclass on chromosom es 7p22 and 17q21, ETV1 (ETS translocation varian t 1) and E1AF (Adeno virus E1A enhancer binding factor), the putative hum an hom ologues of m ouse ER81 and fusions, w hile the genuine F LI1 and ERG transactivation dom ain is alw ays replaced by the EW S am ino term inus resulting in a potentiation of transcriptional activation properties. 67± 69 The 85-am ino acid D N A binding domain folds into three helices and a four-stranded û sheet (w inged helix-turn-helix m otif), 70± 72 m ost frequently refered to as`the ET S dom ain' . In som e ETS fam ily m embers, this dom ain is¯anked by auto-inhibitory a helical structures that fold back and interact with the ET S dom ain. 73 Structural studies on m urine ET S1 suggest that upon speci® c binding to D N A a conform ational change takes place that m ight expose distinct portions of the ET S dom ain and its¯anking regions for interactions with other proteins. 74 D N A bindingdependent complex formation with other transcription factors m ediated by the ET S dom ain and additional residues has been reported for several ET S fam ily m embers including GABP~, ELK1, SAP1, Pu1, ETS1 and ETS2. 75 As dem onstrated for ET S1, w hen binding to the speci® c recognition sequence, intercalation of a tryptophan into the m inor groove induces a sharp kink and a w idening of D N A that m ight facilitate synergistic binding of other regulatory proteins. 76 Since alm ost all ET S proteins bind to a (G /C)(A/C )GG A(A/T)T consensus m otif, 75 NIH3T 3), several potential EW S± FLI1-speci® c target genes were identi® ed. 80 Am ong them w ere the m urine hom ologue of cytochrom e P-450 F 1, cytokeratine 15, a novel SH 2 dom ain containing protein, EAT 2 (EW S± FLI1 activated transcript 2) 81 and the strom elysin gene. Strom elysin is a m atrix m etalloproteinase involved in m etastatic invasio n.
Previously, E1AF has been demonstrated to regulate the stromelysin gene. In transfection experim ents, E1AF was suf® cient to confer an invasive phenotype to non-m etastatic hum an breast cancer cells (M CF7) 82 and antisense RN A to E1AF was able to revert it in a squam ous cell carcinom a cell line. 83 Strom elysin has also been demonstrated to be activated by ETS2, a m ember of another ET S subfam ily. 84 However, coexpression of ERG, w hich also strongly binds to the same prom oter but is by itself unable to activate it, resulted in inhibition of ET S2m ediated strom elysin gene activation. 85 T hus, different ET S fam ily m embers, irrespective of their subclass, app ear to com pete for binding to speci® c target genes. It is likely, therefore, that target selectivity and the speci® c m ode of gene regulation by the EW S chim eric ETS transcription factors strongly depends on the cellular background. A spectrum of ET S-related gene products coexpressed w ith EW S± FLI1 in EFT cell lines by am pli® cation of ETS D N A binding dom ain encoding cD N As has recently been de® ned using degenerate prim ers. 86 Am ong them, E TS2, E4TF1-60, ELK , ELF1, the putative hum an hom ologue of ER71, and a novel gene product ELF R w ere identi® ed. 127 N one of the ET S fam ily m embers involved in EFT -spe ci® c gene rearrangements w ere found to be expressed in their germ -line con® guration. T his assay, however, m ay have m issed low level expression of som e ET S family m embers (i.e. FEV). As show n in Fig. 3 com paring the D N A binding dom ains of the ET S gene products alternatively fused to EW S in EFT with those found to be coexpressed with the chim eric transcription factors, ELK is the m ost likely ET Srelated gene product that m ight interfere with the EF T fusion proteins in target site selection because the D N A-contacting third a -h elix of the ET S dom ain is identical to that of FLI1, ERG and FEV. EL K is one of several alternative ternary com plex factors regulating a num ber of growth factor inducible genes. 87 In fact, EW S± F LI1 can replace EL K within the ternary com plex form ed on the serum response elem ent of the cfos and the EGR1 promoters. 88,89 In contrast to germ -line FLI1, binding of EW S± FLI1 to the serum response element did not require interaction with the serum response factor SRF. T ernary com plex form ation by FLI1 and EW S± FLI1 was m ediated by a dom ain preceding the D N A binding dom ain and present in the m ajority of EFT -derived EW S± FLI1 fusions that show lim ited sim ilarity to the ELK1± SRF interaction dom ain. H ow ever, no hom ologous structure can be identi® ed in ERG and evidence for EW S± ERG involvem ent in ternary com plex form ation is not available.
In sum m ary, current know ledge about norm al and aberrant ET S proteins suggest a num ber of interesting candidate target genes for the EF T-speci® c chim eric transcription factors, potentially involved in the regulation of cell growth, signaling and m etastasis, as revealed by the study of heterologous cellular system s such as m urine ® broblasts. H owever, am ple evidence exists that ET S transcription factor action is largely context speci® c. F or EFT , the cell of origin rem ains a m atter of speculation. Because of lim ited neural differentiation potential, EFT is considered as derived from the neuroectoderm . 90 Interestingly, X enopus FLI1 has been shown to be expressed in a restrictive pattern during embryogenesis evocative of neural crest cell invasio n. 66 It is therefore speculated that FLI1 m ight be involved in neural differentiation in the context of these early stem cells. If so, unscheduled activation of F LI-responsive genes by the EFT -sp eci® c EW S fusion proteins in an undifferentiated cell possibly unrelated to the neural crest m ight result in lim ited neural differentiation of the EFT stem cell depending on the degree of determ ination achieved at the time of gene rearrangement. 1 T his m odel could explain the variable degree of neuroectoderm al m arker expression in ES and pP N ET as well as the occurrence of biphenotypic tum ors. 26 The rate-lim iting (® rst) hit in EFT pathogenesis EW S± ETS gene rearrangements are the only genetic aberrations that have so far been identi® ed as highly associated w ith histologically diagnosed EFT. This association is the only available compelling argum ent that EW S± ETS gene rearrangements can be rate-limiting for tum origenesis. Although com plem entary experim ental evidence supports this assu m ption, the m echanism of m alignant transfo rmation by these chimeric oncoproteins rem ains elusive. The best studied biological m odel for the pathogenic role of inappropriately activated FLI1 is Friend m urine leukemia virus(F -M uL V)-induced erythroleukem ia. Insertional activation of the FLI1 gene appears to be the ® rst detectable genetic change asso ciated w ith this disease. T he association between the detection of FLI1 rearrangement and clonal outgrowth of erythroleukem ia cells suggests that the activation of this transcription factor m ay be affe cting the self-renewal potential of the infected erythroid progenitors. However, leukem ogenesis proceeds in m ultiple steps and additional aberrations affe cting viab ility of cells (e.g. inactivation of the tum or suppressor gene p53) can be observed in F -M uLV induced erythroleukem ia. 91,92 In contrast to ERG , 93 norm al FLI1 was reported to be unable to transfo rm m urine ® broblasts (NIH3T 3) while expression of an EW S fusion protein resulted in pronounced anchorage-independent clonogenicity of N IH 3T3 cells. 94,95 H owever, rat embryo ® broblasts and som e m urine ® broblast subclones w ere resistant to EW S± FLI1-mediated transformation. These ® ndings again suggest that the oncogenic potential of norm al and aberrant FLI/ERG ET S subfam ily m embers m ay depend on a cell type-sp eci® c availability of relevant synergistic factors and possib ly on the presence of additional aberrations. N IH3T 3 transfection studies with various recom binant EW S± F LI1 deletion m utants revealed a dependence of transfo rm ation on both the EW S portion and the ETS dom ain. 95 However, optim al transactivation potential m ediated by the 30 EW S am ino term inal degenerate repeats included in almost all EFT -derived fusion proteins was dispensable for m axim al focus form ation of transfected N IH3T 3 in soft agar. T he m inim al EW S dom ain required to transform m urine ® broblasts was delineated to the ® rst 82 am ino acids. 94 Recently, evidence obtained shows that w ithin the EW S± FLI1 fusion protein, but not within germ -line EW S, this peptide directly contacts a com ponent of the RN A polym erase II complex, RP B7, and that this interaction is suf® cient to drive EW S± FLI1-m ediated reporter gene transactivation. 125 Interactions of fulllength EW S with the general transcription factor T FIID , an essential com ponent of the transcriptional preinitiation com plex, w ere absent from EW S fusion proteins. 126 It is therefore possible that fusion of the EW S amino terminus to the FLI1 D N A binding dom ain alters the protein conform ation and directly recruits RN A polym erase II to F LI1 target genes. Since RPB7 displays sim ilarities to prokaryotic sigm a factors, it m ight be involved in EFT-speci® c target site selection. Further protein± protein interactions presum ably occurring downstream of the 82 am ino acids m ight be required for ef® cient gene regulation w ithin the EFT context. In addition, using the yeast two-hybrid protein interaction trap, further candidate proteins not directly related to transcription regulation were identi® ed that interacted with the 82 am ino acids long m inim al transfo rm ation dom ain. T hese interactions aw ait detailed characterization.
Since no tissue of EFT origin has been identi® ed so far, transfo rm ation studies of authentic EFT stem cells cannot be perform ed. Alternatively, several investigators have used EW S± F LI1 antagonists (antisense RN A expression vectors, antisense oligonucleotides, dom inant negative proteins) to m odulate expression of the chim eric oncoprotein in EFT cell lines. 86,96± 98 T hese studies revealed a growth inhibitory and anti-tum origenic effect of these agents. Reduction in cell growth appeared to result from cell cycle arrest and not from reduced tum or cell viability. Recently, EW S± FLI1-m ediated transform ation of m urine ® broblasts was demonstrated to require the presence of a functional insulin-like growth factor-1 (IGF 1) receptor. 99 Interestingly, consistent expression of IG F1 and its receptor was previously reported for EFT and IGF1 was demonstrated to act as a potent growth factor for EFT cells in the absence of serum . l00± l03 This cytokine appears to regulate negatively several m echanism s of program m ed cell death at a far dow nstream step. 104 It has been show n that inhibition of the IG F1 autoregulatory circuit by anti-IGF1 receptor antibodies resulted in increased apoptosis and reduced tumorigenicity of EFT cells. l03 T aken together, the EW S± ETS gene rearrangement appears to be involved in the aberrant self-renewal capacity of E FT cells but m ight not be suf® cient to guarantee survival of initiated tum or cells. However, as demonstrated recently, there m ight still be som e role for F LI1, ERG and their E W S fusions to play in the protection from stress (i.e. calcium ionophore and serum deprivation)-induced cell death. 105

The second hit
Assum ing that the EW S± ETS gene rearrangement is able to initiate EFT pathogenesis but is not suf® cient to generate m alignant transfo rm ation, the presence of additional m utations m ust be postulated. These aberrations m ight not necessarily be tum or speci® c but m ay display inter-individual variation that could account for variations in EF T phenotype as w ell as in clinical behavio r. Also, they m ay determ ine differentiation capacity, invasive potential and treatm ent resistance. Consequently, while from a clinical point of view the EW S± ETS gene rearrangement provides a valuable diagnostic tumor m arker, knowledge about the nature of additional aberrations in EFT m ay assist subclassi® cation and provide prognostic tools. Since studies on facultative genetic anomalies in EF T have been largely neglected since the discovery of the E W S± ETS gene rearrangements, enhanced efforts to de® ne the m ultitude of additional aberrations are warranted for the bene® t of patients.

C lues from cytogenetics
In three independent reviews of cytogenetically inform ative EF T cases, non-random num erical and structural chrom osom al aberrations were reported to occur with variable frequencies in addition to the tum or speci® c t(11;22)(q24;q12) 106± 107 (H attinger et al., unpublish ed). T hese include trisom y 8 in about 50% of cases frequently coupled with trisom y 12 occurring in roughly 20% , and a derivative chrom osom e 16 as a result of an unbalanced t(1;16) in 18% of EF T. In rarer cases, other aneuploidies have been identi® ed. Structural chrom osom e 1 aberrations that either result in gains of chrom osom e 1q21± 22 or relative losses of the short arm of chrom osom e 1, a frequent alteration in neuroblastom a and other neuroectoderm al tum ors, have also been observed in EFT (Hattinger et al., unpublished). Interestingly, this chrom osomal region harbours a gene encoding a protein (p73) that is structurally and functionally related to the tum or suppressor p53, a transcription factor involved in the regulation of cell growth and apoptosis, and frequently inactivated during the progression of m any tum ors. 108,109 W hile research currently focuses on the role of p73 for neuroblastom a pathogenesis, its relevance for a subset of EFT is som ething that has to be explored.
In general, excluding chrom osom e 22q12 translocations, num erical chrom osome changes are the m ost frequent cytogenetic ® ndings in EF T. T hese are likely to affect gene dosage. However, no candidate genes that could prom ote EW S± ETS-initiated EF T pathogenesis when expressed at aberrant levels have been identi® ed so far. Also, inform ation on genes affe cted by the recurrent chrom osom e 16 and 1 structural alterations is not available yet. Frequently, these cytogenetic alterations occur in only a subpopulation of neoplastic cells within the tum or suggesting that they m ay be associated with late stages of tumor progression.

The role of non-speci® c cancer genes
In the absence of recurrent candidate progressionasso ciated genetic alterations identi® able in EFT by the means of cytogenetics, work has focused on the analysis of m utations generally associated w ith a broad range of hum an m alignancies. Genes investigated during the last years include the oncogenes ras, cmyc and M DM 2, the tumor suppressors p53, p16, and Rb, the metatasis-asso ciated splice varian ts of the CD 44 adhesion m olecule, and the tum or-speci® c m etastasis suppressor gene nm 23H1. 110± 113,127 N one of the studied oncogenes w as found to be altered by m utation or in expression although occasional low level am pli® cation of M DM 2 has been reported in an independent study on a sim ilar sized cohort of EFT patients. 111 N either m utation nor differences in expression levels of the nucleotide diphosph ate-kinase nm 23H1 were observed irrespective of the disease extension. 112 O nly standard C D44 expression was detectable in EF T. 127 In contrast, hom ozygous deletions of the p16 tum or suppressor was identi® ed in about onethird of prim ary EFT sam ples. This ® nding was surprising since chrom osom al aberrations of band 9q21 containing the p16 gene were not reported before suggesting a high-freq uency of m icrodeletions. Expression studies on EFT cell lines suggested that the frequency of p16 inactivation m ight be even higher since post-transcriptional gene silencing w as observed in several cases. 113 p16 acts as an inhibitor of the cyclin D 1/cyclin-dependent kinase 4 (CD K 4) com plex that inactivates the cell cycle inhibitor pRb by phosph orylation. Inactivation of p16 should com prom ise the G1 cell cycle checkpoint. Over-expression of either cyclin D 1 or C D K4, or loss of pRb function, is believed to m ediate a sim ilar effect. 114 In fact, we observed frequent cyclin D 1 over-expression as well as variable CD K4 abundancy in EFT cell lines and loss of Rb in one case. How ever, expression data from prim ary tumor m aterial are not available yet. In addition, low level C DK4 am pli® cation w as previously reported for two of 30 EFT sam ples. 111 In virally induced m alignancy, G1 check-point control is frequently com promised by concom ittant inactivation of the pRb and p53 pathways. In addition, F-M uLV-induced erythroleukem ia involves not only the activation of the FLI1 oncogene as a rate-lim iting step but also m utation of p53. T he author and others have, therefore, investigated the status and expression of p53 and related genes in EFT . 110,115 T he frequency of p53 m utations in prim ary tumors w as found to be low er than 10% as opposed to a m ean frequency of 40± 60% in m ost hum an m alignancies. In contrast, in about half of EFT cell lines, the p53 gene was mutated and show ed loss of heterozygosity. Com pariso n of p53 gene status between cell lines and the respective prim ary tum ors of origin, w hen available, demonstrated that the observed increase in m utation frequency was due to selection and was not acquired during in vitro expansion of tumor cells. T his result suggested that p53 m utation m ight release EFT cells from som e in vivo grow th or survival factor dependency. Transient transfection and over-expression of wildtype p53 in cell lines with endogenous m utant or wildtype gene status demonstrated frequent but variable reduction in apoptotic responsiveness, 116 suggesting the presence of som e as yet unidenti® ed cell-protective m echanism in EF T cell lines. Preliminary expression analysis of m embers of the cell death regulatory Bcl2 gene fam ily did not reveal any signi® cant variations between individual EFT cell lines (our unpublished observations). Previously, high levels and activity of poly(AD P-rib ose) polym erase, a nuclear enzym e that participates in DN A replication, repair and the triggering of apoptosis induced by D N A strand breaks, have been reported for som e EFT cell lines. 117 Sensitivity of EFT cell lines to D N A dam aging agents (etoposide, actinom ycin D , X-rays) varied considerably in a m anner independent from p53 responsiveness and endogenous p53 gene status suggesting that com plete m utational or partial inhibition of the p53 apoptosis pathway in EFT cell lines plays a role different from radio-and chemosensitivity. However, the physiolo gical signals that stim ulate p53-dependent cell death have not been de® ned so far.

M olecular markers of prognosis
As a result of variable break-po int localization in the involved genes, EW S± ETS gene products vary considerably in size. M ost fusions include EW S exons 1 to 7 (89% ) and FLI1 exons 6 to 9 (54% ). E W S/FLI1 exon 7/6 fusions (type 1) predominate independent of the disease extension (51% ). In about one-third of EFT , FLI1 exon 5 is included into the chim eric gene product, m ost frequently joined to EW S exon 7 (type 2) (27%). In rare cases, the chrom osom al translocation results in the inclusion of EW S exons 9 (1%) or 9 plus 10 (10% ) or FLI1 exon 4 (1% ). In about 3% of cases, FLI1 exon 6 or exon 6 plus 7 are m issing from the gene fusion. T his variability has prom pted us to investigate a possib le prognostic im pact of the gene fusion type. The study, perform ed on 55 patients with localized disease and 30 patients with metastases at diagnosis, treated according to the European Intergroup Coordinated Ewing' s Sarcom a Studies (C ESS 86 and EIC ESS 92), revealed a signi® cantly better outcom e for patients with localized disease carrying a type 1 EW S± FLI1 expressing tum or as com pared to non-type 1 cases. 18 A recent update after a m edian observation time of 3 1 2 years con® rm ed this result (Zoubek et al., unpublished). In addition, an independent Am erican study perform ed on a sim ilar sized cohort of patients after a m edian follow up of 31 m onths, using a sim ilar treatment regimen, not only supported our ® ndings but also identi® ed the EW S± ETS gene fusion type as a prognostic m arker independent from the presence of m etastases at diagnosis in a m ulti-variate analysis. 128 About 55% of the`non-type 1' group in the two studies w ere com prised of type 2 gene fusions. Because of the low incidence of the individual`other-gene' fusion types, no distinction has been m ade betw een various non-type 1 subgroups so far. In the absence of a biological explanation for the observed prognostic differences, large collaborative prospective studies are warranted to highlight the speci® c chim eric m olecules and the protein dom ains associated w ith adverse patients' outcom e.
Still, about 20% of patients with localized tum ors and m ore than half of the patients with m etastases succum b from the disease despite the expression of a type 1 EW S± FLI1 gene fusion, suggesting the existence of additional adverse factors.
Com pariso n of pl6 gene status and clinical course of 23 EFT patients analyzed so far suggested an adverse prognosis asso ciated w ith this aberration independent from the extension of the disease. 113 T hese results, w hich have not been subjected to statistical analysis, m ust be considered as prelim inary since patients' num bers in the study were sm all and the median observation period did not exceed 2 years. Retrospective im m unohistochem ical analysis of biopsy m aterial from a large num ber of patients will help to clarify the prognostic relevance of a disrupted pRb cell cycle regulatory pathw ay in EFT . Also, m utation of p53 m ight be linked to an adverse outcom e, since none of the three EFT patients from our series carrying such an aberration survived. H ow ever, because of the rarity of this alteration it cannot serve as a useful prognostic m arker.
A prognostic relevance for EFT of the observed num erical and structural cytogenetic changes has not been demonstrated with con® dence due to low sample num bers in the studies perform ed so far. M ost recently, deletion at 1p36, occurring in 6/22 localized EFT, was discussed as being asso ciated with unfavo rable outcom e in this group (Hattinger et al., unpublish ed).
In the absence of reliable m olecular m arkers to predict outcom e in EFT , the presence of clinically overt m etastases at diagnosis is com m only considered as the only prognostic criterion that is used for treatment strati® cation. The EW S± ETS gene rearrangement as a tum or cell speci® c m arker detectable by the highly sensitive RT -PCR m ethod provides a powerful m eans for the detection of m inute num bers of circulating tum or cells that m ay be the source of clinically occult m icrom etastases. 118,119 However, except shortly after surgical intervention, 120 m obilization of PCR detectable am ounts of tum or cells ( . 1/10 6 ) into the bloodstream has rarely been observed. In contrast, tumor cells were detected at diagnosis by this m ethod in the bone m arrow of 30% of patients with localized disease, 50% of cases with isolated lung m etastases and all patients w ith bone m etastases. 120,121 In a prelim inary series of 23 patients lacking clinically overt dissem ination, RT -PC R screening for bone m arrow involvem ent did not allow the prediction of early relapse after a m edian observation time of 30 m onths. It is, however, necessary to recall several factors that m ay affect tum or cell detection in the bone m arrow by RT -PC R: (1) tum or cell in® ltration m ay be focal and bone m arrow aspiration m ay m iss these sites, (2) bone m arrow aspirates m ay contain variable am ounts of diluting blood resulting in insuf® cient sensitivity, (3) prim ary EFT cells m ay differ in vitality, although diluted tumor cells from cell lines can be detected in blood sam ples even after 48 h at 4°C, 119 and (4) bone m arrow in® ltrating tum or cells m ay be in a resting state and express lower levels of chim eric EW S RN A than proliferating tumor cells. RT -PCR m easures RN A quantity rather than tum or cell abundance. In a recent study, up to 10-fold variations in the content of chim eric EW S± ETS tran-scripts between individual EFT cell lines have been reported. 122 M oreover, germ -line EW S expression in T -cells has been demonstrated to depend on the proliferative activity. 123 Since the EFT -sp eci® c chrom osom al rearrangement places the chim eric gene under the control of the EW S regulatory sequences, it is possible that EW S± ETS gene expression m ay also vary. Consequently, detectability of EW S± ETS chim eric transcripts does not necessarily re¯ect true tum or cell content. Even if RT-PCR studies fail to dem onstrate signi® cance of positive blood or bone m arrow screening results for relapse in patients with localized disease, the question of prognostic relevance of tum or cell in® ltration rem ains unsolved. T o assess this problem , im m unohistochem ical studies m ay prove to be superior to RT -PC R. W hile M IC2 m ay serve as a valuable surface marker in tum or diagnosis, its exploitability for tum or cell detection in hem atopoietic tissue is lim ited. 124 T he recently discovered expression of gastrin-releasing peptide (GRP) by all EFT cell lines and about half of the prim ary tumors tested (Lawlor et al., unpublished) m ay provide a m arker that, in conjunction w ith M IC2, m ay allow the identi® cation of EFT cells in blood and bone m arrow w ith increased speci® city. T he study of E W S± FLI1 transcriptional targets m ay result in the identi® cation of tum or cell-restricted im m unohistochem ical m arkers. In order to detect positively staining cells with very low abundance on a routine basis, automated m icroscopic screening and consequently sophisticated technical equipm ent is w arranted.

C onclusions
In this review, I have sum m arized evidence for the im portance of studying EFT-spe ci® c genetic alterations in an authentic cellular background. Since the histogenesis of EF T is still enigm atic and no experim ental evidence for E W S± FLI-m ediated tum origenesis has been reported from transgenic mouse m odels so far, EF T cell lines rem ain the only available system for such investigations. In this institution, cell lines could be established from 12 EFT patients with well docum ented clinical course. If a cell line could be expanded from the prim ary tumor, all subsequent tum or sam ples also gave rise to a cell line. All but one patient died from the disease suggesting that establishing a cell line selects for patients with adverse prognosis. In fact, non-type 1 E W S± ETS gene fusions, p16 deletions and p53 m utations were clearly increased in EF T cell lines from 23 patients investigated. They m ay, therefore, represent the m ost therapy-resistant subpopulation of tumor cells despite variable in vitro sensitivity to cytotoxic agents. T hus, EFT cell lines m ay serve as a pool for the identi® cation of putative bad prognostic m arkers. However, only large cooperative clinical studies and m ultivariate statistical analysis will help to address the question: how far can the identi® cation of such m arkers translate into clinically useful criteria for treatm ent strati® cation? W hen com paring a wide spectrum of EF T-derived cell lines for marker expression and response to either differentiation inducing agents, growth factors or cytotoxic compounds, an im m ense variability was observed. C onsequently, in order to sort out the biological defects com m on to all E FT, a large panel of genetically well de® ned cell lines w ill have to be investigated. In the long term , such studies w ill result in the identi® cation of EW S± ETS-sp eci® c target genes and in a m ore detailed knowledge of the m echanism of m alignant conversion of the enigm atic E FT precursor cell. H opefully, for the bene® t of the patients, this knowledge will potentially provide novel targets for therapeutic intervention.

A cknowledgem ents
I thank D rs Enrique de Alava, Peter Am bros, D ave Aryee, M arc Ladanyi, Poul Sorensen, Laszlo Tora and Jeff T oretsky for perm itting m e to refer to their before publication results.