Introduction

According to the American Cancer Society, rhabdomyosarcoma (RMS) comprises about three percent of childhood cancers, with about 350 new cases occurring annually in the US [1], and it affects slightly more males than females [2]. RMS is a small, round, blue cell tumor usually arising in skeletal muscle tissue, and it is thought to originate from mesenchymal cells likely committed to the skeletal muscle lineage. Consistent with a myogenic origin, RMS tumors express skeletal muscle markers such as skeletal muscle actin and myosin, desmin, myoglobin, Z-band protein, MYOD and often myogenin [3–7]. RMS consists of two major histological subtypes, embryonal and alveolar RMS. The embryonal subtype (ERMS) is thought to be histologically reminiscent of embryonic develo** skeletal muscle [7]. ERMS is the most prevalent of the subtypes, accounting for about 60% of RMS cases [2]. It occurs mainly in children younger than 10 years and is usually associated with a favorable prognosis, with a failure-free survival rate of 81% [8, 53].

Though GLI1 is amplified genetically, the expression of this gene is not always associated with its genetic amplification. When GLI1 is overexpressed in RMS, it has been associated with an undifferentiated subtype rather than ERMS or ARMS, indicating that GLI1 may play a role in tumors that show primitive histopathological features [54]. Thus, GLI1 overexpression cannot be well associated with the ARMS pathology.

MDM2 is perhaps the best candidate oncogene in this region because of its inhibitory effect on p53 function [55]. However, MDM2 is not always included in this 12q13-15 amplification. RH30, an ARMS cell line, lacks amplification of MDM2 but shows amplification and overexpression of SAS, CHOP, GLI1 and A2MR [53]. In addition, the frequency of MDM2 gene amplification specifically may be as low as 10% in ARMS tumors [56]. One study found only 2 of 34 ARMS samples to be highly immunoreactive for MDM2 [12]. Moreover, MDM2 expression shows no association with patient prognosis or other clinicopathologic parameters [12]. Thus, it may be amplification of one of the other genes at this chromosome 12 locus that is the important cooperating mutation with PAX3-FOXO1.

Other alterations in the p53 pathway have been found in ARMS. In ARMS tumor samples mutated p53 was reported in 0 to 22% of cases [12, 56, 57]. RMS cell lines show a significantly higher rate of p53 abnormalities with 60%, indicating establishment of these cell lines through xenograft and cell culture increases the proportion of cell lines with p53 alterations [56]. Looking at p53 and MDM2 expression levels, both are low in ARMS and ERMS. Metastatic ERMS tumors show significantly higher p53 protein expression, indicating that p53 gene alterations are a late event in rhabdomyosarcomagenesis. Again, p53 status did not show any correlation to prognosis [58].

Chromosome 2 has been shown to be amplified at 2p24 in 32 to 60% of ARMS tumors [49, 51, 59, 60]. This region is known to contain the proto-oncogene N-MYC. Two independent studies have shown that a gain in the genomic copy number of the N-MYC gene is associated with an unfavorable disease outcome [59, 61]. In addition, N-MYC is more highly expressed in ARMS cells lines than ERMS lines, despite the fact that it was found to only be genomically amplified in one of the five lines, indicating more than one mechanism of N-MYC overexpression in ARMS. However, in this study no clear relationship in N-MYC expression was seen with regard to primary tumor samples [62].

Another chromosomal region frequently amplified in ARMS is 13q31-32, showing amplification in between 14 and 19% of ARMS tumors [49, 51]. Presence of this amplification is significantly associated with poorer failure-free survival in ARMS [63]. The minimum overlap** region of amplification at this region was originally defined as only containing two genes: GPC5 and C13ORF25. The C12ORF25 gene encodes the micro-RNA cluster miR-17-92 (MIR17HG) in an intron [64]. GPC5 overexpression can increase cell proliferation through the modulation of the growth factor activity of FGF2, HGF and WNT1a [64]. However, more thorough map** of the genetic amplification showed that the entire GPC5 locus was only amplified in 12.5% of 13q31 amplified ARMS tumors, while the minimally amplified region contains only the peptidylprolyl isomerase pseudogene (LOC390419) and MIR17HG. This amplification is particularly prevalent in PAX7-FOXO1-positive ARMS tumors. The miR-17-92 cluster of micro-RNAs has been shown to play a role in a variety of cancer types (for review, see [65]). In PAX7-FOXO1, but not PAX3-FOXO1 expressing ARMS, overexpression of miR-17, -19a, -19b, 20a and 92a is specifically associated with an increased rate of 2-year treatment failure. This indicates a possible pro-tumorigenic interaction between PAX7-FOXO1 and miR-17-92 locus overexpression [63].

Rhabdomyosarcoma can also be associated with a loss of heterozygosity (LOH) or loss of imprinting (LOI) at 11p15.5 [66, 67]. This region contains several imprinted genes such as IGF2, which is maternally imprinted (paternal allele is expressed), and H19 and p57/Kip2, which are paternally imprinted (maternal allele is expressed) [68–70]. IGF2 expression appears to be specifically upregulated by changes in imprinting or LOH at this locus in RMS. ERMS tumors are associated predominantly with a LOH at the IGF2 locus, though there is some discrepancy in the proportion of ERMS tumors showing this change: 23% according to Anderson et al. [66] and 72% according to Visser et al. [67]. Conversely, IGF2 is upregulated by LOI in 46% of fusion-positive ARMS tumors, while imprinting of H19 is conserved in 93% [66]. This indicates that an increase in IGF2 expression in RMS is important for tumorigenesis, though the mechanism of this upregulation, either LOH or LOI, varies by subtype.

A screen for PAX3-FOXO1-interacting proteins using ARF−/− primary mouse myoblasts expressing PAX3-FOXO1 and an RH30 cDNA expression library identified a gene that could induce tumor formation where ARF−/− myoblasts expressing PAX3-FOXO1 alone did not. The RH30 gene expression library expressed a truncated fragment of this novel gene dubbed IRIZIO, and expression of either this truncated form or the full-length wild-type IRIZIO were protumorigenic in this model [71]. Due to the nature of the screen, and given that abrogation of the p53 and pRb pathways are required for PAX3-FOXO1-driven cell transformation [46, 47], this screen was expected to identify proteins that could abrogate the pRb pathway [71]. The mechanism of the interaction between IRIZIO and pRb, however, has yet to be identified.

PAX3-FOXO1 target genes

Many gene expression studies have been performed by various groups to try to identify genes that are either downstream of PAX3-FOXO1 gene expression in various cell types or are indicative of ARMS tumor gene expression profiles (see Table 1). Only a small proportion of these studies have gone on to further investigate the mechanism of PAX3-FOXO1 regulation of these genes and/or what role these genes may play in ARMS tumorigenesis.

Table 1 PAX3-FOXO1 target genes in ARMS
Full size table

Two of these genes have already been mentioned as cooperating mutations seen in ARMS tumors, N-MYC and IGF2. The N-MYC locus is known to be amplified in a proportion of ARMS tumors, and the IGF2 locus is known to show LOI in ARMS tumors (see cooperating mutations in ARMS tumors). However, these studies also indicate that PAX3-FOXO1 may regulate the gene expression from these loci.

N-MYC expression has been shown to be upregulated in four independent studies using PAX3-FOXO1 targeting siRNA in the ARMS cell line, RH4 [72], PAX3-FOXO1 overexpression in the RD (ERMS) cell line [73, 121, 122].

Lagutina et al. [120] developed a model where PAX3-FOXO1 was knocked into the PAX3 locus. This knock-in locus expressed low amounts of PAX3-FOXO1, which in heterozygous pups was sufficient to result in developmental defects in the heart and diaphragm, leading to congestive heart failure and perinatal death, as well as malformations of some hypaxial muscles. However, neither chimeric adults nor their newborn heterozygous pups developed malignancies. It was hypothesized that PAX3-FOXO1 expression from the PAX3 control sequences was insufficient to cause ARMS formation, and downstream regions of the FOXO1 locus may be required to induce sufficient PAX3-FOXO1 expression to induce tumor development.

A PAX7-FOXO1 model of ARMS was also attempted in Drosophila [123]. Expression of UAS-hPAX7-FOXO1, under control of myosin heavy-chain Gal4, also resulted in developmental defects in the fly, evidenced by disorganized myogenic patterning. Though nothing resembling tumor formation was seen, this group did note dissemination and infiltration of non-native tissue by PAX7-FOXO1 expressing mononucleated cells, indicating an increase in invasive capacity of these cells.

Keller et al. [47] used a conditional PAX3-FOXO1 knock-in into the PAX3 locus, and Myf6-driven Cre expression. This allowed, upon Cre recombination, expression of PAX3-FOXO1 driven by the PAX3 promoter and 3’ FOXO1 genomic sequences that potentially contain cis-regulatory elements, a region absent from previous PAX3 knock-in strategies. This was the first animal model that successfully recapitulated the formation of ARMS, though at the low frequency of approximately 0.4% (1/228) and with latency of over 1 year (383 days). However, this frequency was greatly enhanced, and latency greatly reduced, in homozygote PAX3P3Fa/P3Fa mice also lacking Trp53 or Ink4a/Arf. Subsequently, ARMS tumors have developed in this conditional PAX3-FOXO1 knock-in model with a Pax7 CreER and M Cre (Pax3 hypaxial muscle enhancer) also lacking functional Trp53 [124]. Moreover, histologically diagnosed fusion-negative ARMS tumors have been found to develop in conditional Ptch1+/−Trp53−/− mice when Cre is expressed from Pax7 CreER. The latency and incidence of ARMS tumor development in these different models have yet to be compared.

Clearly the problems that have arisen during the development of an animal model for ARMS indicate that the timing and the cell lineage targeted for PAX3-FOXO1 expression are very important for the development of ARMS tumor formation and for avoiding developmental defects. In a review [125] following the publication of the animal model [47], Keller et al. discuss the possibilities for the cell of origin for ARMS; because Keller et al. achieved the formation of ARMS tumors in their mouse model using Myf6-Cre-driven conditional PAX3-FOXO1, and Myf6 is usually expressed in differentiating skeletal muscle myotubes, they propose a potential dedifferentiation mechanism for ARMS development. However, the formation of a fusion gene such as PAX3-FOXO1 suggests that the cell of origin for ARMS should express both PAX3 and FOXO1 at the time that the translocation occurs, given that open chromatin is likely required for these two genomically distinct regions to translocate. Anecdotal evidence for this includes that the genome translocations that occur in many different cancer types occur between genes that are expressed in the cell type of origin. A recent study by Osborne et al. [126] showed that the MYC and IGH genes, which are involved in a chromosomal translocation common in Burkitt lymphoma, are colocalized at the same transcription factories more often in activated B-cells, the originating cell of Burkitt lymphoma, than resting B cells. This colocalization at the same transcription factory allows for close proximity of these gene loci in euchromatin, providing the circumstances where these genes would be in close association, facilitating the specific translocation event. PAX3 is rapidly downregulated upon myoblast differentiation, so it would be unlikely that the PAX3 loci would be expressed in a nascent myotube expressing Myf6, making it difficult to understand how translocation could occur in nascent myotubes and therefore cast some doubt on whether the dedifferentiation model is feasible. However, it is possible that Myf6 expression does rarely occur in a small subset of undifferentiated myogenic cells in conjunction with PAX3. This could allow for this model to produce ARMS tumors and account for the low frequency at which these tumors are seen as well as the requirement for homozygous PAX3-FOXO1 knock-in alleles [47].

From these animal models it is apparent that the timing of PAX3-FOXO1 expression is critical for ARMS development. Too early and widespread expression of PAX3-FOXO1 expression can result in developmental defects and no apparent tumor development [120–123], whereas later expression of PAX3-FOXO1, via a Myf6-driven Cre recombinase, does cause disease, though at a low frequency [47]. Perhaps inducible expression, driven by various myogenic genes with carefully characterized expression profiling, would result in an increased frequency of disease and help to narrow down the exact stage in which PAX3-FOXO1 expression drives ARMS tumorigenesis. Nevertheless, the cell of origin for ARMS is yet to be identified, and animal models of ARMS will no doubt play an important role in its identification.

Conclusion

To date, numerous factors (outlined in Figure 2) have been identified that contribute to ARMS tumor development and its aggressive clinical phenotype. These consist of both PAX3/7-FOXO1 target genes, such as N-MYC, IGF2, MET, CXCR4, CNR1, TFAP2B, FGFR4 and P-cadherin, and PAX3/7-FOXO1 cooperating factors, such as the abrogation of the p53 pathway, IGF2 deregulation, N-MYC and miR17-92 amplification, and IRIZIO expression. Future ARMS research will continue to discover the mechanisms by which ARMS tumorigenesis occurs. This will involve the identification of more PAX3-FOXO1 target and cooperating genes; more importantly, the mechanisms by which these genes contribute to tumorigenesis will be elucidated. It is critical that we develop a mechanistic understanding of how these factors contribute and interact to perpetrate ARMS tumorigenesis. This will allow new opportunities to develop specifically targeted therapies for this aggressive pediatric disease.

Figure 2
figure 2

Review summary: Fusion gene regulated genes contributing to alveolar rhabdomyosarcoma. Rhabdomyosarcoma develops from an unknown cell of origin from the mesodermal lineage that may be skeletal muscle specified. This cell likely expresses both PAX3/7 and FOXO1 and may also express Myf6. A gene fusion event resulting in a PAX3/7 DNA-binding domain fused to a more potent transcriptional activation domain occurs. This fusion transcription factor is capable of inducing a group of PAX3-FOXO1-regulated genes that contribute to ARMS development in conjunction with other genetic lesions.

Full size image