Molecular Alterations in Pediatric Sarcomas: Potential Targets for Immunotherapy

Purpose/results/discussion. Recurrent chromosomal translocations are common features of many human malignancies. While such translocations often serve as diagnostic markers, molecular analysis of these breakpoint regions and the characterization of the affected genes is leading to a greater understanding of the causal role such translocations play in malignant transformation. A common theme that is emerging from the study of tumor-associated translocations is the generation of chimeric genes that, when expressed, frequently retain many of the functional properties of the wild-type genes from which they originated. Sarcomas, in particular, harbor chimeric genes that are often derived from transcription factors, suggesting that the resulting chimeric transcription factors contribute to tumorigenesis. The tumor-specific expression of the fusion proteins make them likely candidates for tumor-associated antigens (TAA) and are thus of interest in the development of new therapies. The focus of this review will be on the translocation events associated with Ewing's sarcomas/PNETs (ES), alveolar rhabdomyosarcoma (ARMS), malignant melanoma of soft parts (MMSP) (clear cell sarcoma), desmoplastic small round cell tumor (DSRCT), synovial sarcoma (SS), and liposarcoma (LS), and the potential for targeting the resulting chimeric proteins in novel immunotherapies.


Introduction
C hrom osom al abnorm alities are com m on in hum an tum ors w ith m any malignancies exhibiting clonal chrom osomal aberrations. 1 The identi® cation of tum or-spe ci® c chrom osom al translocations aids in diagnosis and serves as a prognostic indicator. 2± 6 W ith an increasing understanding of the effect these events have on normal cellular processes, novel therapies can be developed which have greater speci® city and ef® cacy.
Tw o m ajor consequences of chrom osom al rearrangem ents in tum ors have been identi® ed: the activation of an oncogene, or the creation of a novel oncogenic protein. First, translocations can result in the activation of genes located at or near the breakpoint. O ften, these genes normally function in the prom otion of cell growth and differentiation. T hus, their disruption can affect norm al cell regulation. T his type of alteration, which is m ost com m on in hematological m alignancies, is illustrated by the t (8;14) translocation associated with Burkitt' s lymphom a in which c-M YC is activated by repositioning under the control of the potent Ig enhancer. 1 An alternative consequence of chrom osom al translocations is the generation of functional chim eric genes. This scenario is m ost com m on in solid tum ors and usually involves unrelated genes. O ften, these translocation events affect genes encoding transcription factors, thereby generating chim eric transcription factors with properties of both genes (Table 1). T he fusion proteins often exhibit the D N A-binding speci® city of one gene with the activation dom ain of the other gene. Such fusion proteins activate/repress transcription, exhibit altered D N A binding speci® city or participate in novel protein± protein interactions. Thus, they are thought to play a critical role in the neoplastic transform ation process.
The identi® cation of translocations asso ciated with a group of prim itive sarcom as, and the subsequent cloning of the chrom osomal breakpoint regions, has revealed that a com m on theme in these tum ors is the generation of chim eric transcription factors. The fusion proteins are expressed exclusively in the tum or cells, and function as potent transcription factors where they are thought to contribute to neoplastic transfo rm ation by mediating bZip TF aberrant expression of norm al genes. Several of the chim eric genes have been cloned and found to confer a transform ed phenotype when expressed in vitro. 7± 11 The tumor-spe ci® c expression of the fusion proteins m ake them likely candidates for tumorasso ciated antigens (TAA), in which the junction point creates a neo-antigenic determinant. T he focus of this review will be on the translocation events asso ciated w ith Ew ing' s sarcom as/prim itive neu-  EW S encodes a 656-aa protein, the function of which rem ains unclear. W hile this protein is ubiquitously expressed, expression levels¯uctuate with the cell cycle. 19± 23 EW S contains two m ajor functional dom ains. The ® rst is the N -term inal region (exons 1± 7) consisting of a series of degenerate repeats that resem ble the transactivation dom ains of several transcription factors, such as SP-1 24 while the second region, the C -term inal region, includes a putative RN A-b inding dom ain (exons 11± 13) de® ned by a conserved 80-aa dom ain. 24 W ild-type EW S has been shown to bind RN A in vitro and EW S/GAL4 fusion proteins can activate a reporter gene, suggesting a role for EW S in transcription. 9,21,23 FLI1, a m ember of the ETS fam ily of transcription factors, is the hum an hom ologue of the m urine FLI1 gene and is norm ally expressed in hematopoietic tissues. 25 T he ET S D N A-b inding domain, usually located in the C-term inal portion of the protein, is an 85-aa region that recognizes target genes through a conserved GG AA /T sequence. 26 In FLI1, the E TS dom ain is encoded in the C-term inus, and the N -term inal region contains a dom ain that is functional in reporter gene assays. 9,27 EW S/F LI1 is a potent transcription factor that can transfo rm N IH 3T 3 cells, and studies have show n that sequences in both EW S and FLI1 are essential for transform ation. 7± 9 T o better de® ne the functional regions of the fusion protein, substitutions were m ade in which dom ain 1 of EW S was replaced w ith a strong heterologous activation dom ain. M any of these fusion proteins retained activity, although not all were transfo rm ing. 7,23 D om ain 2 of EW S could also be exchanged with a weak transcriptional activation dom ain from TLS/ FU S without loss of activity. T hus, these data support a m odel wherein the EW S region of EW S/F LI1 confers strong transactivation through dom ain 1 with additional properties (protein± protein interaction) contributed by dom ain 2.
Several varian ts of the t(11;22)(q24;q12) EW S/ F LI1 gene fusion have been described, 8,19 but m ost include E W S exons 1± 7 and FLI 1 exons 8 and 9. 2,28 T herefore, the am ino term inal portion of EW S is always fused to the carboxy term inal region of F LI1 8,19 w hich suggests that these EW S/F LI1 variants contribute to oncogenesis by sim ilar m echanism s. EW S/FLI1 and FLI1 have sim ilar D N A-binding speci® city and af® nities, 9,29 but EW S/FLI1 is a m ore potent transactivator than FLI1. 9,29,30 In vitro studies suggested that EW S/FLI1 functioned as a transactivator at 10-fold lower concentrations than FLI1. 29 T hus, it is likely that EW S/FL I1 m ediates its transform ing effects, at least in part, by transactivation of F LI1 targets or promoters containing ET S-binding sites.  29 Recent studies suggest that EW S/F LI and FLI1 exhibit som e differences in D N A-bin ding and pro-tein± protein interactions. 31 T herefore, it is possible that EW S/FLI1 also contributes to transfo rm ation by activating genes not norm ally regulated by FLI1. Studies are ongoing to identify the norm al targets of EW S/FLI1 and FLI1. Braun et al. 32 utilized representational difference analysis (RDA) to identify differentially expressed genes from N IH 3T3 cells containing EW S/FLI1 or norm al F LI1. T his approach revealed that several transcripts w ere dependent on the fusion protein for expression, while at least two transcripts were repressed. Strom elysin 1, cytokeratin 15, and a m urine hom olog of cytochrome P-450 F1 are all induced following expression of EW S/FLI1. However, the kinetics of expression argue against the direct upregulation of all of these target genes. The elucidation of such prim ary targets will provide insight into the role of EW S/FLI1 in transfo rm ation. It is likely that the oncogenic properties of EW S/F LI1 results from both the inappropriate expression of FLI1 target genes, as well as novel protein± protein interactions w hich m ay lead to the activation of non-FLI1 target genes. Studies that utilized antisense EW S/FLI1 cD N A to dim inish EW S/FLI1 RN A levels demonstrated m arkedly decreased cell grow th in vitro, thereby im plicating the fusion protein as a key contributor to aberrant growth. 33,34 EW S/F LI1 m ay contribute to oncogenesis is by inhibition or alteration of norm al apoptotic pathways. Yi et al. 35 observed suppression of apoptosis in Ewing' s sarcom a cells expressing EW S/F LI1 and found that expression of the fusion protein antisense RN A increased susceptibility to apop tosis. T hus, EW S/FLI1 m ay contribute to m alignant transfo rmation by alteration of m ore than one gene or gene pathw ays.
The EW S gene is also involved in several other tum or-associated translocations. For exam ple, a m inority of PN ETs present with a varian t t (21; 22) translocation that fuses E W S to the E RG gene. 3, 28,36 Like FLI1, ERG is a m ember of the ET S fam ily of transcription factors and m ay regulate sim ilar target genes. 32 Studies are underway to identify ERG target genes. Several lines of evidence suggest EW S/ ERG m ay contribute to neoplastic transfo rm ation by the sam e or sim ilar m echanism s as EW S/F LI1. First, PN ET s containing EW S/FLI1 or EW S/ERG are phenotypically and clinically indistinguishable. 2,36 As is seen in EW S/F LI1, EW S/ERG fusions include EW S exons 1± 7, with ERG sequences encoding the ETS dom ain. 3, 28,36 The fusion protein also functions as a transcription factor and requires the sam e regions for transactivation de® ned in EW S/ FLI1 studies. 21 Furtherm ore, cells expressing EW S/ ERG have a decreased ability to undergo apoptosis. T hese cells could be m ade susceptible to apoptosis by the expression of EW S/ERG antisense RN A. 35 T herefore, it is likely that EW S/ERG fusions contribute to oncogenesis in a m anner sim ilar to EW S/ FLI1.
A rare, third varian t, t(7:22)(p22;q12) has been described 37 in which EW S is fused to E TV I, the hum an hom olog of the m urine ET S gene ER81. It is likely that EW S/ET V1 contributes to m alignant transform ation by m ediating aberrant transcription and/or repressing expression of regulatory genes. H owever, RD A analysis of EW S/ET V1 revealed that only one of eight EW S/FLI1 target genes was upregulated by EW S/ETV 1. T his suggests that EW S/ETV 1 activates only a portion of the EW S/ FLI1 transform ation pathway, requiring other alterations for tumorigenesis, or that EW S/ETV 1 plays a m inor role in transform ation. Further studies are needed to de® ne the effect of EW S/ET V1 on norm al gene expression.
Recently, Peter et al. identi® ed a new m em ber of the ET S fam ily fused to EW S in Ewing' s sarcoma, the FEV gene. 38 FEV, which m aps to chrom osom e 2, encodes a 238-aa protein. Its expression is highly restricted with protein being detected only in adult prostate and sm all intestines, but not in other fetal or adult tissues. F EV contains an ETS D N A binding dom ain closely related to that of ERG and FLI1; however, in contrast to these proteins, FEV has a sm all N -term inal region of only 42 aa which suggests that it lacks important transcription regulatory dom ains present in other ETS fam ily proteins. It is unclear whether or not EW S/FEV alters transcription of sim ilar target genes than other EW S fusion proteins. F urther studies are needed to elucidate this fusion protein' s role in the pathogenesis of ES.
The com m on denom inator of these tum ors is that all are prim itive neuroectoderm al sarcom as occurring in children and young adults, and the evidence strongly im plicates EW S fusions as key m ediators of m alignant transform ation. There is also strong evidence to suggest that these fusion proteins contribute to oncogenesis by aberrant expression of target genes (activation and repression), as well as altering the expression of genes not norm ally regulated by the native transcription factors. 32 F urtherm ore, these genes m ay effect norm al growth regulation by interfering with apoptotic pathw ays. 35

A lveolar rhabdom yosarcom a (A RM S)
Rhab dom yosarcom a is the m ost com m on soft tissue sarcom a in pediatric patients, w ith approxim ately 250 cases per year in the United States. Roughly 20% of these cases are of the alveolar m orphological type (ARM S) w hich is characterized by alveolar-like spaces form ed by ® brovascular septa. T hese spaces are ® lled w ith m alignant cells that are distinguished by their eosinophilic cytoplasm . Approxim ately 80% of ARM S express a translocation involving the long arm s of chrom osom es 2 and 13 t(2;13)(q35;q14), w hich results in the juxtapositioning of a truncated PAX 3 gene of chrom osom e 2 to the 39 -terminal region of the FKHR gene of chrom osom e 13. 39± 43 The PAX fam ily of transcription factors play important roles during em bryonic developm ent, particularly in m orphogenesis and pattern formation. 44 T hese genes contain a paired-box (PB) D N Abinding dom ain and som e also contain a hom eobox (H B) D N A-binding dom ain. O verexpression of these genes can result in oncogenic transformation 10,11 and loss of function m utations has been observed in several genetic diseases, including W aardenburg syndrom e. 45 FKH R, form ally known as ALV, 41 is a m em ber of the fork-he ad dom ain (FD ) fam ily of transcription factors which contain a conserved D N A-binding m otif related to the Drosophila region-sp eci® c hom eotic gene fork-head. This fam ily of transcription factors norm ally functions during em bryogenesis. T he FKHR gene is ubiquitously expressed and functions as a transcription factor.
The hybrid gene which results from the t(2;13)(q35;q14) translocation encodes a fusion protein containing the am ino term inal portion of the PAX3 protein including the PB and H B dom ains joined to the carboxyl region of the FK HR protein that is truncated within the winged helix D N Abinding region, but retains a putative transactivation dom ain. Evidence suggests that the D N A-binding speci® cty of PAX3/FKH R is contributed by PAX3, m ost likely through the PB and H B dom ains, while F KHR contributes the transactivation region. Although the D N A-binding activity of PAX3/FKH R is less than wild-type PAX3, the fusion protein is a m ore potent transactivator. 46± 49 O verexpression of m urine PAX3 transfo rm s N IH 3T 3 cells 11 and the PAX 3/FK HR fusion protein transfo rm ed chicken em bryo ® broblasts. 10 O ne possible m echanism of transform ation is through a gain of function, not only by increased transactivation potency, but also through constitutive and increased expression. 49,50 Interestingly, a recent study which utilized antisense technology to downregulate PAX3/FKH R in ARM S tum or cells demonstrated reduced cell viability, which led to the conclusion that PAX3/F KHR m ay contribute to m alignant transfo rm ation through suppression of apoptotic processes which would norm ally cause cell death. 51 Interestingly, 10± 20% of ARMS tum ors contain a variant translocation, t(1;13)(p36;q14), that results in the in-fram e fusion of 59 PAX7 to 39 F KHR. PAX 7 and PAX3 are highly hom ologous in the PB and H B domains, suggesting that they m ight recognize similar target genes. 40± 43,52 Furtherm ore, the PAX 3/FK HR and PAX 7/FK HR chim eric proteins share structural sim ilarities in that they both contain intact N -term inal PB and HB regions fused to the acidic and proline-rich C -term inal region of FK HR. 41,42,52 Therefore, it is likely that these translocations create sim ilar chim eric transcription factors that contribute to transform ation by altering expression of a comm on group of target genes. 50± 52

M alignant melanom a of soft parts (M M SP) or clear cell sarcom a (CC S)
M alignant melanom a of soft parts (M M SP), also known as clear cell sarcom a (C CS), is a rare, but aggressive soft tissue sarcom a of m uscle tendons and aponeuroses that occurs m ost frequently in young adults between the ages of 15 and 35 years. 53 Over 95% of M M SP cases occur in the extremities, and only rarely (less than 2% ) occur in the head and neck region. Although M M SP is a m elanin-producing tum or, there is no evidence to suggest that these tum ors are directly related to m alignant m elanoma. M M SP is thought to have neuroectoderm al origins 54 and expresses neural antigens, as w ell as m arkers of m elanin production, such as HM B-45. A t(12;22) (q13;q12) translocation event is present in m ore than 70% of these tumors 55,56 and m olecular analysis of the breakpoint reveals an EW S/ATF1 fusion. T his chim eric protein joins the 59 RN A-b inding region of the EW S gene and the 39 region of the ATF1 gene, a m ember of the CREB/transcription factor fam ily of leucine zipper transcription factors that has a bZIP dom ain for D N A binding and protein± protein interaction. 57 This fam ily of transcription factors mediates transcription through AT F-binding sites. The expression of these genes is induced by cAM P, and they are activated by phosphorylation by cAM P-dependent protein kinase A (PK A). 58,59 The t(12; 22) translocation fuses the N -terminal portion of EW S to the C-term inal region of ATF1, retaining the bZIP dom ain. However, the PKA regulatory phosph orylation site is lost. 58 T hus, it is likely that EW S/ATF1 could exhibit the D N A-b inding speci® city of ATF1, and dimerize with CREB, but would not be cAM P-inducible. EW S/ATF 1 does activate prom oters with AT F1 binding sites, although not all such prom oters were activated, 60 and som e promoters were found to be repressed by EW S/ATF 1. T herefore, EW S/ATF 1 m ay contribute to m alignant transform ation by several m echanism s. F irst, EW S/ATF 1 m ay constitutively activate ATF 1 target genes that are norm ally induced by cAMP, or it m ay repress genes that norm ally function in growth control. Alternatively, EW S/AT F1 m ay activate novel genes, perhaps genes regulated by other C REB/ATF fam ily m embers.
In m ost M M SP tum ors, two hybrid transcripts are generated and expressed by the t(12;22) (p13;q12) translocation. T he expression pro® le of the fusion gene on der(12) chrom osome is compatible w ith the ubiquitous expression of ATF. H owever, this out-of-fram e fusion results in a product consisting of the ® rst 65 N -term inal am ino acids of AT F1, w hich is unlikely to bind D N A or dim erize, m aking its role in transform ation unclear. It is unlikely that expression of the der(12) transcript is essential in transform ation given reports that 30% of M M SP lack expression. 56

D esmoplastic small round cell tum or (DSRC T)
D esm oplastic sm all round cell tumor (DSRC T) is an aggressive sm all round cell tum or that occurs predom inantly in abdom inal serosal surfaces and has a predilection for young males. 61 The tum or is a prim itive sm all round cell w ith features of divergent differentiation, co-expressing epithelial, neural and m yogenic m arkers. T he origin of this tumor rem ains unclear, but it is m ost likely derived from the m esothelium. Alm ost 100% of these tumors contain a t(11;22)(p13;q12) translocation that fuses the 59 region of the EW S gene to the 39 region of W T1, a tum or suppressor gene involved in a subset of W ilm s' tum ors. 62± 66 W T1 binds D N A through a series of zinc ® ngers and represses the transcription of certain genes. T hese zinc ® ngers are essential for transcriptional repression. T he chim eric protein contains the N -term inal region of EW S fused to the W T1 D N A-binding domain. Given that both the w ild-type EW S gene and EW S fusion proteins are known to participate in transcriptional com plexes, it is likely that EW S/W T 1 functions as a transcription factor, possibly through W T 1 targets. Therefore, unlike the loss of function m utation in W ilm ' s tum or, the loss of the zinc ® nger region of W T 1 in EW S/W T1 serves to convert W T 1 from a repressor of transcription to a dom inant transcriptional activator oncogene. 67

Synovial sarcom a (SS)
Synovial sarcom a is an aggressive soft-tissue m alignancy which occurs primarily in the extrem ities near m ajor joints (e.g. ankle, knee) of adolescents and young adults. Virtually all synovial sarcom as contain a translocation of chrom osom es X and 18 68 with app roximately 70% involving t(X;18)(p11.2;q11.2). T his translocation event generates a fusion protein from the 59 region of the SYT gene and the 39 region of SSX1 or SSX 2. 69± 71 T here is no evidence of a transcript being expressed by the reciprical hyb rid der (18). 71 T he function of the SYT gene is unknown, and sequence analysis reveals no classical structural m otifs asso ciated with D N A-binding or transcriptional regulation. However, the presence of SH2 and SH3 dom ains suggests that SYT m ight function through protein± protein interaction. T he recent isolation of the m ouse hom olog of SYT revealed that SYT is expressed ubiquitously during early embryogenesis, 69 but expression is restricted later in developm ent to cartilage tissue, speci® c neuronal cells and som e epithelial-derived tissues. SYT was also detectable in prim ary sperm atocytes. Several studies suggested that SS contained two distinct X chrom osom e breakpoint sites. However, the identi® cation of two closely related genes at X p11.2 established the involvement of distinct coding regions. D espite being 2 M b apart, SSX 1 and SSX 2 share 80% hom ology. 70 Both encode a 188-aa protein with an N -term inal Kruppel-assoc iated box (KRAB) that is thought to function as a transcription repressor dom ain. 72,73 Although these proteins lack zinc ® nger m otifs, the presence of the KRAB sequences suggest a role in transcription. However, this dom ain is not present in the chim eric protein, which suggests that SSX 1 and SSX 2 sequences contribute to transform ation through novel protein± protein interactions or som e other function. SSX3, another KRAB protein, is not im plicated in t(X; 18)-positive SS, 74 but has high hom ology to SSX1 and SSX2 (95 and 90%, respectively). The study of this gene m ay provide insight into the function of SSX 1 and SSX2.

Liposarcom as (LPS)
Liposarcom as (LS) are soft tissue tum ors that occur prim arily in the extrem ities and retroperitoneum . T hese tum ors are from prim itive m esenchym al cells and they resem ble fetal adipose tissue. Several characteristic cytogenetic aberrations have been identi® ed for adipose tumors. The m ost com m on LS are m yxoid round cell liposarcom as, and greater than 90% of m yxoid liposarcom as contain the t(12;16)(q13;p11) translocation in which C HOP on the long arm of chrom osom e 12 is fused to FUS/TLS. 22,75± 77 H owever, this translocation event has not been detected in other adipose tum ors and, therefore, m ay provide interesting insight into the transform ation process of this subset of tum ors. FU S/TLS is structurally sim ilar to EW S ( . 50% am ino acid identity) 75 and is expressed at high levels in all tissues exam ined. 22 T LS binds RN A and encodes a strong transcriptional activation dom ain in the N -term inal region. 78 Therefore, like EW S, F US/T LS m ay function as a nuclear RN A-binding protein.
CH OP, also called GAD D 153, is a m em ber of the CC ATT /enhancer-binding protein (C/EBP) fam ily of leucine zipper transcription factors that regulate adipocyte differentiation. C HO P is expressed at low levels in adipocytes; however, m RN A levels increase during conditions of stress such as D N A dam age. O verexpression of C HO P in N IH 3T 3 cells results in growth arrest at G1/S. 79 T hus, C HO P is thought to function as a dom inant negative growth regulator. 80 In the TLS/CHO P fusion protein, the N -terminal portion of TLS is joined to the entire CH O P coding region. 75,76 TLS/C HO P can transform N IH 3T 3 cells and studies indicate that transfo rm ation requires sequences from both T LS and CH OP. 78 T he requirem ent for the C -term inal leucine zipper dom ain of CH OP for transform ation suggests a crucial role for C/EBP protein dim erization. Although it is unclear whether normal wild-type CHO P activation requires D N A-binding, the potential D N A-binding region, a basic region of the bZIP dom ain, is required for transform ation. T he role of TLS sequences in transfo rm ation m ay be m ore than that of a strong transactivator, since substitution of this region with other potent transactivating dom ains did not m ediate transfo rm ation. However, substitutions w ith EW S sequences were transform ing. 78 T herefore, TLS/C HO P m ay contribute to transfo rm ation by m echanism s sim ilar to those previously discussed in EW S fusion proteins.

P otential imm unotherap eutic approaches for the treatm ent of pediatric sarcom as
Although m ulti-m odality therapy has im proved survival rates for the pediatric sarcom as described in this review, patients often relapse, at which time responses to m ulti-agent chem otherapy are brief or non-existent. Furtherm ore, patients who present w ith m etastatic disease at diagnosis do very poorly in spite of aggressive m ulti-m odality therapy. T herefore, efforts are needed to develop novel treatments, such as im m unotherapies. Studies over the past decade have provided evidence that treatm ents based on the m anipulation of the im m une system can m ediate regression of established m etastatic cancer. M ore speci® cally, cell-m ediated im m unity can play a critical role in tum or regression.
T lym phocytes are m ost often categorized as C D 8 1 cytotoxic lym phocytes (C TL) or CD 4 1 helper lym phocytes (Th), and both types of T cells are know n to play a role in tum or regression. Our understanding of antigen processing, presentation, and recognition has increased considerably in the last tw o decades and has been expertly reviewed elsewhere. 81 Brie¯y, T cells recognize antigens as short peptides that are bound to the cell surface in the context of m ajor histocom patibility (M HC) m olecules. 81,82 In the case of CD 8 1 C TL, the T cell receptor (TC R) recognizes short peptides (8± 10 am ino acids) bound to M HC class I m olecules. T hese peptides are derived from endogenously expressed proteins which undergo proteolytic processing in the cytosol by large proteosom e com plexes. Peptide fragm ents are then transported into the lum en of the endoplasm ic reticulum (ER) by specialized transporters of antigen processing (TAP). O nce inside the E R, peptides associate with an app ropriate M HC class I m olecule that is asso ciated with beta-2-m icroglobulin (b 2 m ), an invariant subunit w hich is thought to enhance ef® cient M HC folding, optim ize M HC/peptide binding, and increase stability of the M HC /peptide com plex during transport to and expression on the cell surface. Following peptide/M HC binding, the peptide/ M HC /b 2 m com plexes transverse the E R and G olgi app aratus, and are displaye d on the cell' s surface where they are subject to surveillance by CT L. In the case of CD 4 1 Th cells, the T CR recognize slightly larger peptides (10± 25 aa) in the context of M HC class II m olecules. T hese peptides are typically derived from m aterial or organism s which have undergone endo/phagocytosis by APC. Thus, in general, C D 8 1 C TL recognize intracellular (endogenous) peptides while CD 4 1 T cells recognize external (exogenous) protein fragm ents.
CT L can distinguish self from non-self peptides associated with MH C class I m olecules, so that expression of viral proteins or altered cellular proteins w ill be re¯ected in the peptide/MH C complexes displayed on the cell surface. Although the tum or-sp eci® c fusion proteins described in this review function as nuclear transcription factors, they are still subject to the proteolytic processing and presentation pathways described. T here is experim ental evidence that tum or-associated nuclear proteins, such as m utant p53, can induce imm une responses. 83± 88 .
The identi® cation of T AA and an increased understanding of the requirem ents for the induction of cell-m ediated im m une responses (Table 2) has led to advances in im m unotherapy. 89 W hile a number of T AA have been identi® ed for several tumor types, 90± 93 it is unclear whether all T AA will be effective tumor regression antigens. Ideally, one would like to identify and target T AA which play a key role in neoplastic transfo rm ation, so that they cannot be lost without loss of m alignancy. T he tum or-associated translocations identi® ed for a num ber of pediatric sarcom as such as ES and AR m ay very well be such antigens, since they generate functional chim eric transcription factors known to contribute to abb errant gene expression. M ore Tab le 2. Immunothe rapeutic approaches using tum or-associated antigens Passive imm unotherapy with anti-tumor lymphocytes generated in vitro Generation of CT L using imm unodominant peptide-pulsed APCs Generation of Th by coincubation of APC with antigenic peptides speci® cally, the breakpoint junctions are likely neoantigens. Further, it should be possible to avoid autoimmune responses by focusing on m inim al peptides corresponding to the sequences which span the breakpoint, since these would not be present in norm al cells. This hypothesis was tested in animal m odels using synthetic peptides corresponding to the breakpoint junctions in ES and ARM S as imm unogens. In these studies, peptide-pulsed APC adm inistered intravenously, generated C D 8 1 CT L responses capable of lysing peptide-pulsed tum or cells in vitro as w ell as tumor cells transfected to express the full-length fusion protein. F urtherm ore, these responses were able to reduce or irradicate tum or in vivo. T hese data demonstrate that the chim eric fusion products resulting from chrom osom al translocations can serve as neoantigens.
Because the translocation events are tumor speci® c, therapies targeting the resulting fusion proteins w ould be highly speci® c and potentially less toxic. C linical trials are currently underway in patients w ith ES and ARM S to evaluate the generation of anti-tum or responses using a sim ilar approach. In addition, studies are ongoing to not only identify additional T AA , but also to gain an understanding as to which T AA m ay serve as tum or rejection antigens. Since it is clear that the im m une system does not react against all possible antigenic determ inants, characterization of the im m unodom inant peptides in the tum or regression antigens w ill further aid in the development of effective treatments. 94 The identi® cation of T AA and the cloning of the genes which encode them provides num erous opportunities for the developm ent of cancer therapies (Table 2). T herapies could utilize the T AA protein either alone or with adjuvants. Alternatively, the adm inistration of peptides derived from the T AA protein adm inistered alone, with adjuvants or in com bination with helper peptides, has certain advantages in that this approach has been demonstrated to generate T cell responses while having m inim al risk in the induction of unw anted and potentially dangerous autoimm une reactions. Antitum or responses generated by peptide vaccination m ay be augmented by m anipulation of the route/ m ode of adm inistration. T he cloning of genes encoding TAA will facilitate their expression in high-ef® ciency expression system s, such as recom binant viruses or bacteria. These vectors can be engineered to express the T AA alone or in conjunction with cytokine genes or genes encoding costim ulatory molecules. Furtherm ore, direct injection into m uscle of D N A encoding antigens or the use of gene guns' in w hich D N A is attached to sm all particles that are m echanically propelled into cells is also an effective m ethod of inducing im m une responses. 95± 100 Anti-tumor responses have been generated by in vitro sensitization of peripheral blood lymphocytes (PBL) to peptide-pulsed AP C or irradiated tum or cells. Repeated in vitro sensitization using imm unodom inant peptides from m elanom a antigens pulsed onto autologous peripheral blood m ononuclear cells in the presence of IL-2 resulted in the expansion of C TL (10,000-fold ) over a 6-week period. Cells generated by this approach showed imm une reactivity 50± 100 times greater than corresponding tum or in® ltrating lymphocytes (TIL) 101 and speci® cally recognized the approp riate im munodom inant peptide as w ell as tum or cells as m easured by lysis and cytokine release. Studies in experim ental anim al m odels suggest that speci® c tum or recognition as determined by lysis and cytokine secretion assays correlated highly w ith in vivo anti-tum or effects. 102 T hese correlates have also been observed in patients treated with autologous T IL . 103,104 In several other studies, T cells stim ulated in vitro were capable of recognizing and lysing target cells pulsed with peptides known to bind to a particular M H C class I m olecule; however, these sam e T cells were often incapable of recognizing and lysing the low levels of processed peptides expressed by tumor cells. 105 Thus, there is considerable heterogeneity in anti-tum or responses.

S um m ary
T he generation of chim eric transcription factors is a com m on consequence of chrom osom al translocations in solid tumors. The resulting fusion proteins have been show n, in several cases, to have transform ing activity. Chim eric oncoproteins m ay function through several m echanisms. F irst, a strong activation dom ain from one gene may be fused to the D N A-bin ding speci® city region of another gene, leading to dysregulated expression of target genes. T he fusion proteins asso ciated with M M SP, ARM S, and PN ET s are exam ples of this m echanism . H owever, in m yxoid liposarcom a, the F US/C HO P gene product appears to m ediate its effect on transcription through protein± protein interactions and m ay not require D N A-binding. Second, a fusion partner m ay contribute m ore than an activation dom ain. F or exam ple, the EW S/FLI1 fusion protein of ES seem s to com bine the transactivation dom ain of EW S with the D N A-binding region of FL I1. H owever, the fusion protein appears to m ediate novel protein± protein/protein± nucleic acid interactions. Also, the chim eric oncoprotein m ay heterodim erize w ith other transcription factors. For exam ple, the heterodimerization of T LS/CH O P with C/EBP with C /EBP fam ily m em bers regulates adipocyte grow th in a dom inant-negative m anner. Finally, chim eric genes m ay be overexpressed as a result of a strong promoter region from one of the partner genes. H owever, this m echanism has not been observed in solid tumors, but may be relevant in hem atopoietic m alignancies. N onetheless, it is likely that expression of hybrid proteins in solid tumors dysregulates the transcription of key growth control genes or pathw ays, thereby promoting tumorigenesis.
W hile fusion proteins are likely to invoke a com bination of the afo rem entioned mechanism s, the redundancy of their role in oncogenesis is noteworthy. T he multiple interchange of functional dom ains from related genes such as FLI1 and ERG in PNE Ts, PAX 3 and PAX7 in ARM S and SSX 1 and SSX 2 in SS result in sim ilar tum or phenotypes. 78 D om ain-sw ap experim ents involving EW S for T LS in T LS/CH O P show ed that substitutions can be m ade with little change in m orphology. However, other experim ents in which FLI1 was exchanged for C HO P in fusions with TLS or EW S had an effect on cell m orphology, such that the m orphology in som e cases w as dependent on the DN A-b inding region of the chim eric transcription factor. Finally, of note is the early onset of m any of these tum ors. T his suggests that the genes involved in sarcom a-associated translocations have speci® c patterns of developmental regulation, and that dysregulation of this temporal regulation has profo und effects.
Attem pts at developing new therapeutic approaches to the treatm ent of these tum ors have included im m unotherapy. How ever, successful imm unotherapeutic stratagies m ust m eet several criteria, the ® rst of which is the expression of TAA that are recognized by T lym phocytes. In the case of the sarcom as presented in this review, the chim eric transcription factors represent potential TAA. Studies in experim ental anim als suggest that the translocation breakpoints in ES and ARM S represent neoantigens which can be recognized by CT L. Furtherm ore, these response were suf® cient to m ediate in vivo tum or regression in animal m odels. Clinical vaccine studies are ongoing to evaluate the ability of these T AA to serve as tumor regression antigens. Finally, identi® cation of the im m unodom inant epitopes in tum or regression antigens will favor the induction of effective anti-tum or responses. Screening vaccines and various delivery system s (peptides or proteins in adjuvants or on dendritic cells, D N A, viruses) in anim als, such as HLA-transgenics, w ill help to identify the m ost prom ising vaccines for use in clinical trials.