Rhabdomyosarcoma (RMS) is the most common soft tissue sarcoma in children and represents a high-grade neoplasm of skeletal myoblast-like cells. Decades of clinical and basic research have gradually improved our understanding of the pathophysiology of RMS and helped to optimize clinical care. The two major subtypes of RMS, originally characterized on the basis of light microscopic features, are driven by fundamentally different molecular mechanisms and pose distinct clinical challenges. Curative therapy depends on control of the primary tumour, which can arise at many distinct anatomical sites, as well as controlling disseminated disease that is known or assumed to be present in every case. Sophisticated risk stratification for children with RMS incorporates various clinical, pathological and molecular features, and that information is used to guide the application of multifaceted therapy. Such therapy has historically included cytotoxic chemotherapy as well as surgery, ionizing radiation or both. This Primer describes our current understanding of RMS epidemiology, disease susceptibility factors, disease mechanisms and elements of clinical care, including diagnostics, risk-based care of newly diagnosed and relapsed disease and the prevention and management of late effects in survivors. We also outline potential opportunities to further translate new biological insights into improved clinical outcomes.
Soft tissue sarcoma accounts for ~7% of cancers in children and 1% of cancers in adults1. Approximately half of the population of children with soft tissue sarcoma have rhabdomyosarcoma (RMS), which is a high-grade, malignant neoplasm in which cancer cells have a propensity for myogenic differentiation. Two major RMS subtypes exist, ‘alveolar’ RMS (ARMS) and ‘embryonic’ RMS (ERMS), which are driven by fundamentally different mechanisms. Both subtypes pose substantial clinical challenges because achieving a cure requires controlling the primary tumour, which may arise in a wide variety of anatomical sites, by surgical resection and/or ionizing radiation and eradicating systemic metastatic disease using intensive chemotherapy. The past 30 years have witnessed dramatic improvements in survival for many children with RMS, as evidenced by the development and implementation of sequential clinical trials conducted nationally or internationally as cooperative groups spanning North America and Europe2,3,4,5,6,7. Moreover, advancements in molecular biology and genetics have also enabled more sophisticated understanding of RMS pathogenesis. Those approaches continue to provide a platform to improve diagnosis, disease classification, patient risk stratification and management strategies.
ARMS and ERMS have emerged as the two major RMS subtypes on the basis of light microscopic features of cells distributed around an open central space8 (ARMS) or cells resembling immature skeletal myoblasts (ERMS)9. That distinction was supported by the recognition that ARMS was often associated with balanced chromosomal translocations involving chromosomes 2 or 1 and chromosome 13 (referred to here as t(2;13) and t(1;13)), originally detected by cytogenetics10,11,12. As described in detail below, a small but substantial fraction of patients with ARMS do not harbour one of these translocations, and tumours from those patients are biologically and clinically similar to ERMS.
The WHO also recognizes two rarer RMS subtypes. Pleomorphic RMS is a morphological variant of RMS that typically occurs in adults13. Similar to ERMS, unifying molecular genetic aberrations in pleomorphic RMS are not yet clear. In children, a spindle cell/sclerosing RMS variant is seen; those tumours arising in the head and/or neck region seem to be more likely to carry specific somatic mutations and have a poorer prognosis13.
Disease classification of RMS subtypes has been further refined by the identification of ‘fusion-positive’ RMS (FPRMS) and ‘fusion-negative’ RMS (FNRMS). Molecular biology approaches and next-generation DNA and RNA sequencing have shown ARMS-associated translocations to generate novel fusion proteins involving PAX3 (encoding paired box protein 3) or PAX7 and FOXO1 (encoding forkhead box protein O1)14,15,16. Excluding pleomorphic RMS that occurs in adults, most experts consider childhood RMS to be best described on the basis of fusion status. We adhere to that convention in this Primer, but we use ARMS and ERMS as descriptors when describing pathology reports and earlier research based on those classifiers.
Despite many advances, the chance of cure for children with widely metastatic and recurrent disease remains very low. Moreover, patients experience months of intensive, multifaceted therapies that can bring life-threatening acute toxicities and, in some cases, life-changing late effects.
In this Primer, focusing on paediatric RMS, we reveal new light being shed on RMS biology through the use of next-generation nucleic acid sequencing and the employment of whole-organism-based models of disease. We also discuss how the application of functional approaches for therapeutic target identification and validation should propel further clinical advances.
Global disease burden
Although a rare disease, RMS is a fairly common form of childhood cancer and is the most common soft tissue sarcoma in children. The overall incidence rate of RMS is ~4.5 patients per million individuals aged <20 years17. In the United States, this equates to ~350 new cases per year. On the basis of data from the Surveillance, Epidemiology and End Results (SEER) Program, we know that the incidence of RMS differs by both age and histology18 (Fig. 1). Moreover, the incidence of RMS in Europe appears to be similar to that in the United States. For example, a 2016 report from Sweden indicates an overall annual incidence of 4.9 patients with RMS per million individuals aged <15 years19. Interestingly, the incidence of RMS appears to be lower in parts of Asia, with just over 2 patients per million individuals reported in Japanese, Indian and Chinese populations20.
Influence of age and sex
RMS incidence rates are influenced by several factors intrinsic to the tumour and to individual patients. For example, a recent analysis of SEER Program data showed the diagnosis of ERMS to be ~2.5-fold more frequent than the diagnosis of ARMS18. However, the exact frequency of these two forms is affected by evolving diagnostic criteria (discussed below), especially in North America13,18. Age also influences RMS incidence. ERMS is the most common form in early childhood (Fig. 1a), but some data indicate a second peak in early adolescence for ERMS21 (Supplementary Fig. 1). Bimodal peaks are not evident in children with ARMS; the incidence of ARMS remains constant throughout childhood and adolescence17. Notably, one report showed the median age of children with PAX7–FOXO1-positive disease to be younger than that of those with PAX3–FOXO1-positive RMS (age 6 versus 13 years)22, but that conclusion was from a small, retrospective analysis of a limited number of US institutions and has uncertain significance. In contrast to ERMS and ARMS, WHO-classified pleomorphic RMS primarily occurs in adult males in their sixth decade of life23. ‘Pleomorphic’ RMS in children is typically classified as ERMS with diffuse anaplasia13.
RMS incidence also varies by sex, as male children have a higher incidence of ERMS than female children (male:female ratio of 1.51, 95% CI: 1.27–1.80)21. On the basis of an analysis of pooled cancer registry data from five US states, which focused on the risk of RMS by parental ethnicity, overall notable differences in incidence by ethnicity have not been observed24. The only exception was that the risk of RMS in children was significantly lower when both parents were of Hispanic ethnicity (OR: 0.65, 95% CI: 0.48–0.88).
Using data from the SEER Program, several groups have reported the incidence of ERMS for the period 1975–2005 to have remained fairly stable18,25. By contrast, a significant increase in the incidence of ARMS (annual percentage change of 4.20%, 95% CI: 2.60–5.82) was evident over the same period (Fig. 1b). This apparent increase may be related to fluctuations in diagnostic criteria, such as the proportion of the tumour required to display alveolar features in order to diagnose ARMS. Although the use of a more objective RMS classification scheme based on the presence or absence of the t(2;13) or t(1;13) translocations or expression of PAX3–FOXO1 or PAX7–FOXO1 fusion transcript may clarify this matter, that information has not been captured by the SEER Program or many other large cancer registries.
As opposed to osteosarcoma26 and Ewing sarcoma27 (two other fairly common childhood soft tissue sarcomas), there has not been a published genome-wide association study for RMS. Moreover, whole-exome and whole-genome sequencing has identified somatic mutations in RMS28,29,30, but few studies have characterized the role of germline DNA on disease susceptibility. It remains challenging to define risk factors in a rare cancer type with an incidence of 4–5 patients per million individuals. However, a great deal of literature exists to support the hypothesis that genetic susceptibility and environmental factors play a part in RMS development.
Genetic risk factors
Numerous reports highlight that children with certain genetic disorders develop RMS more frequently than unaffected peers. Syndromes that are most commonly seen in children with ERMS include Li–Fraumeni syndrome (germline mutation of TP53, a tumour suppressor)31, neurofibromatosis type I (deletions in the NF1 gene)32,33, Costello syndrome (HRAS mutation)34,35, Noonan syndrome (germline genetic variants activating RAS–MAPK (mitogen-activated protein kinase) pathways)34, Beckwith–Wiedemann syndrome36 and DICER1 syndrome (germline DICER1 mutations)37 (Table 1). However, on the basis of smaller clinic-based studies, only ~5% of patients with RMS are thought to have comorbid germline susceptibility syndromes38. Interestingly, cancer predisposition syndromes appear to be more frequent in patients with ERMS than in those with ARMS33,34,35,39. This finding seems to contrast experimental studies showing that germline loss of specific tumour suppressors facilitates PAX3–FOXO1-driven neoplasia in genetically engineered mouse models40. Importantly, large population-based studies are required to systematically characterize mutations (Table 1) in children with RMS and to evaluate differences by fusion protein status.
Environmental risk factors
Several environmental exposures and other factors have been implicated in RMS risk in children. Many published reports are based on a large epidemiological case–control study of RMS that was enabled through the former Intergroup Rhabdomyosarcoma Study Group (IRSG) and current Children’s Oncology Group (COG), which drive therapeutic studies for 80–85% of all children with RMS in North America41. In that study, 322 patients with RMS aged ≤20 years at the time of diagnosis and 322 control individuals matched by sex, age and ethnicity were enrolled between April 1982 and July 1988. Key findings include prenatal X-ray exposure25, parental recreational drug use42 and several other factors43,44,45, correlating with increased risk of RMS (Table 2).
In contrast to research into osteosarcoma and Ewing sarcoma46,47, the epidemiology of RMS has been understudied. Although clusters of cases have been reported in the literature48,49, it remains unclear whether these patterns are indicative of specific aetiologies or are due to chance alone. A molecular basis for the apparent influence of male sex or Hispanic ethnicity on RMS diagnosis also remains unknown. Additional studies are needed to describe the global distribution of RMS across the age spectrum and to identify risk factors for this childhood cancer, as prevention strategies are not well developed or routinely implemented. Future studies should be conducted in a large-scale setting because some associations identified in smaller analyses have not been confirmed in independent assessments.
The two major subtypes of RMS — ARMS and ERMS — were originally recognized on the basis of light microscopic features. However, the pathogenesis of those two subtypes is distinct, as ARMS tumour cells usually contain a balanced chromosomal translocation generating an oncogenic ‘fusion protein’ that is absent in ERMS. As the fusion protein has considerable biological and clinical implications and not all ARMS cases harbour a fusion protein, most experts feel that RMS is better classified as FPRMS and FNRMS. Many insights into RMS pathogenesis stemmed initially from analyses of tumour-derived DNA and RNA. Genomics approaches are being complemented by functional approaches to study disease mechanisms in various models (Box 1).
Notably, although all subtypes of RMS resemble skeletal myoblasts, the cell of origin is not well characterized. Indeed, elegant genetic models indicate that RMS-like tumours can form when certain oncogenic proteins are expressed in cells from the skeletal myoblast lineage40,50 but also in non-myogenic cells51. It is conceivable that RMS driven by different oncogenic changes or at different anatomical sites may originate in different types of cells that are programmed during tumour formation to express a complement of skeletal myocyte genes.
Fusion proteins generated by translocations
Cytogenetic studies identified recurrent chromosomal translocations, including a frequent t(2;13)(q35;q14) or a variant t(1;13)(p36;q14), in the majority of patients with ARMS52,53 (Fig. 2). These translocations juxtapose PAX3 on chromosome 2 or PAX7 on chromosome 1 with FOXO1 on chromosome 13 (refs14,15,54). PAX3 and PAX7 encode highly related members of the paired box transcription factor family, which are expressed in skeletal muscle progenitors, whereas FOXO1 encodes a widely expressed member of the forkhead transcription factor family. The t(2;13) and t(1;13) translocations respectively generate PAX3–FOXO1 and PAX7–FOXO1 fusion genes that are expressed as fusion transcripts and translated into neomorphic fusion proteins. These fusion genes encode chimeric transcription factors containing an amino-terminal PAX3 or PAX7 region with an intact DNA-binding domain and a carboxy-terminal FOXO1 region containing an intact transcriptional activation domain. In contrast to wild-type PAX3 and PAX7 proteins, the PAX3–FOXO1 and PAX7–FOXO1 fusion proteins have enhanced transcriptional activity, which is attributed to decreased sensitivity of the FOXO1 transactivation domain to the inhibitory effects of amino-terminal PAX3 or PAX7 domains55,56. In addition to the functional effects, there is a higher expression of PAX3–FOXO1 and PAX7–FOXO1 mRNA and protein relative to the wild-type products57. The high level of PAX3–FOXO1 expression results from a copy-number-independent increase in transcription of the fusion gene, whereas the high level of PAX7–FOXO1 expression results from an increased copy number due to in vivo amplification of the fusion gene57.
The expression and/or function of the PAX3–FOXO1 fusion protein, the most well studied of RMS-associated fusion proteins, is also influenced at the post-translational level. Within the carboxy-terminal FOXO1 region, there are multiple phosphorylation and acetylation sites. Some of these sites, such as AKT (serine/threonine kinase 1)-dependent phosphorylation sites at S437 and S500, regulate the subcellular localization and degradation of wild-type FOXO1 but not PAX3–FOXO1 (ref.58). However, the stability of the fusion protein is influenced by post-translational modifications at other sites. Expression of the PAX3–FOXO1 fusion protein is stabilized by S503 phosphorylation mediated by the serine/threonine protein kinase PLK1 and by K426 and K429 acetylation mediated by the histone acetyltransferase KAT2B. Interestingly, pharmacological inhibition of PLK1 or KAT2B in ARMS-derived tumour cells results in ubiquitin-mediated degradation of PAX3–FOXO1, an observation that could ultimately lead to a way to therapeutically target this oncogenic fusion protein59,60. In addition, studies with the pharmacological kinase inhibitor PKC412 revealed a regulatory role for phosphorylation of the amino-terminal PAX3 region on DNA binding and transcriptional function of the fusion protein61. As multiple phosphorylation sites appear to be involved, the specific proteins mediating these events have not been identified.
Transcriptional and epigenetic effects
As a transcription factor, the PAX3–FOXO1 fusion protein usually functions in ARMS tumour cells as an activator that increases the expression of downstream target genes by binding to PAX3-binding sites near these genes. Chromatin immunoprecipitation followed by sequencing (ChIP–seq) studies demonstrated that most regions bound by PAX3–FOXO1 are situated >2.5 kb distal to the nearest transcription start site and were associated with active enhancer chromatin marks, such as acetylation of histone H3K27 (refs62,63). Furthermore, these regions often contain E-box DNA-binding motifs in addition to PAX3-binding sites. PAX3–FOXO1 binds to these regions along with the E-box-specific transcription factor N-Myc (encoded by MYCN) and myogenic basic helix–loop–helix transcription factors MYOD1 and myogenin (encoded by MYOD1 and MYOG, respectively) and generates super-enhancers near a subset of target genes, including ALK (encoding anaplastic lymphoma kinase), FGFR4 (encoding fibroblast growth factor receptor 4) and MYCN, MYOD1 and MYOG. In addition, PAX3–FOXO1 also interacts directly or indirectly with chromatin-related proteins, including chromatin remodelling proteins bromodomain-containing protein 4 (BRD4) and the chromodomain helix DNA-binding protein 4 (CHD4)63,64. By interacting with these transcription factors and chromatin-related proteins, PAX3–FOXO1 probably reprogrammes the chromatin landscape and establishes super-enhancers that associate with target gene promoters by 3D looping. The co-dependence of ARMS cells on the fusion protein as well as these other collaborating proteins is demonstrated by the high sensitivity of ARMS cells to small interfering RNAs (siRNAs) and small molecules targeting each of these proteins63,64.
The PAX3–FOXO1 and PAX7–FOXO1 fusion proteins function as oncoproteins by dysregulating multiple cellular pathways (Table 3). Following gene transfer into several mammalian cell types, exogenous PAX3–FOXO1 expression is associated with transforming activity, including loss of contact inhibition of proliferation and gain of anchorage independence65,66. Repression of PAX3–FOXO1 expression in patient-derived human ARMS cell lines with siRNA or short hairpin RNA (shRNA) constructs results in decreased proliferation and survival, increased differentiation and decreased motility and invasion67,68. Moreover, a mouse knock-in model that conditionally expresses a Pax3–Foxo1 fusion in cells of the myogenic lineage develops ARMS-like tumours, and tumour susceptibility is greatly increased in animals lacking tumour suppressors encoded by either the Trp53 or Cdkn2a (encoding cyclin-dependent kinase inhibitor 2A) genes40. The oncogenic effect of these fusion proteins on growth, survival, differentiation and other pathways is mediated through activation of numerous downstream target genes such as the aforementioned ALK, FGFR4, MYCN and MYOD1 as well as genes encoding the CXC-chemokine receptor 4 (CXCR4) and the Met proto-oncogene receptor tyrosine kinase (MET)62.
Molecular differences between subtypes
RMS-specific fusion genes can be detected in clinical biopsy samples by reverse transcription PCR (RT-PCR) and fluorescent in situ hybridization (FISH) assays. These assays have revealed that 60% of patients with ARMS express PAX3–FOXO1, 20% express PAX7–FOXO1 and 20% are FN69. A small subset of patients with ARMS lack detectable PAX3–FOXO1 or PAX7–FOXO1 fusion proteins but have novel variants, such as PAX3–FOXO4 or PAX3–NCOA1 (NCOA1 encodes nuclear receptor coactivator 1)70,71. The clinical or biological consequences of these variants are not clear. Nucleic acid sequencing has shown that patients with FN ARMS do not express fusion proteins but instead show genetic changes in tumour cells that are similar to ERMS tumours, such as whole-chromosome gains, recurrent point mutations and 11p15.5 allelic loss28,30,72,73,74. In addition, studies of genome-wide mRNA expression revealed that ERMS and fusion-negative ARMS tumours have very similar expression profiles, which are distinct from the expression profiles in PAX3–FOXO1-positive and PAX7–FOXO1-positive ARMS tumours71,72,75. These studies thus provide genetic evidence for the combination of ERMS and fusion-negative ARMS tumours into a single FNRMS subset and the combination of PAX3–FOXO1-positive and PAX7–FOXO1-positive ARMS tumours into a distinct FPRMS subset.
Aberrant signalling pathways
In FNRMS, various mutations have been identified that largely converge on a limited number of pathways. Interestingly, these pathways are also perturbed in FPRMS through upregulation of downstream targets of the fusion proteins and/or genomic amplification, indicating some commonality in the molecular driving forces in RMS (Fig. 3).
Mutations in RAS–PI3K pathway components
Aberrations in various genes associated with RAS pathway signalling are predominant in FNRMS. Approximately one-third of patients with FNRMS are reported to have activating mutations in key components of the RAS pathway, including mutations in NRAS, KRAS and HRAS. NRAS mutations are predominant in adolescents and KRAS and HRAS mutations are more frequent in infants aged <1 year28,74,76. Mutations in genes encoding proteins associated with RAS pathway intracellular signalling, such as protein tyrosine phosphatase, non-receptor type 11 (PTPN11), NF1, the B-Raf proto-oncogene BRAF (encoding a serine/threonine kinase) and phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic-α (PIK3CA; protein is also known as PI3Kα), are also described. Overall, >50% of patients with FNRMS harbour a mutation expected to impact RAS–RAF–MAPK and/or PI3K–AKT–mTOR (mechanistic target of rapamycin) pathways28,30,74,76,77. Activation of these pathways in a large fraction of FNRMS is also supported by gene expression analyses78,79,80. These findings are further supported by immunohistochemistry analyses of phosphorylated AKT, MAPK and ribosomal protein S6 kinase-β1 (S6K1) that show >80% of RMS biopsy samples show activation of the PI3K pathway, with co-activation of the MAPK pathway in over one-third of ARMS biopsy samples and nearly one-half of ERMS biopsy samples81. Various cell surface receptor tyrosine kinases (RTKs) can signal through these pathways, including RTKs that are induced by the fusion protein62, and these RTKs are implicated in RMS development and progression.
Activating mutations in FGFR4 occur in ~7% of patients with FNRMS. The K535 and E550 FGFR mutants activate RAS and STAT (signal transducer and activator of transcription) signalling pathways and induce tumour growth and metastatic behaviour in mouse tumour cells expressing the human protein82. In FPRMS, expression of wild-type FGFR4 and FGFR2 is also elevated as these genes are downstream transcriptional targets of PAX3–FOXO1 (refs62,63) and they contribute to tumour behaviour83,84,85 (J.S., unpublished observation). Similarly, insulin-like growth factor receptor 1 (IGF1R) is highly expressed in FPRMS, occasionally via genomic amplification86. Very frequent high expression of the ligand IGF2 also occurs in RMS, resulting from loss of heterozygosity or loss of imprinting of the 11p15.5 locus that encodes the IGF2 and H19 genes. This molecular lesion is more common in patients with FNRMS28. IGF signalling is pro-survival and anti-apoptotic in RMS tumour cells and increases tyrosine phosphorylation of insulin receptor substrate 1 (IRS1)87. In poor-prognosis patients with RMS (that is, those with advanced or metastatic disease), IRS1 activation appears refractory to normal negative feedback mediated by increased phosphorylated mTOR and S6K1 (ref.88), and inhibiting IGF1R activity can promote compensatory regulatory activity of the SRC family kinase YES89. The epidermal growth factor RTK ERBB2 is frequently expressed90,91 and upregulated with associated MAPK signalling in response to IGF1R inhibition92. Furthermore, knockdown of PIK3CA leads to increased expression of PIK3CA isoforms and elevated RAS pathway signalling in cell line models81. Changes in feedback loops and dynamic signalling after targeting specific pathway components suggest that RMS may need to be treated with combination therapeutic strategies.
Other RTKs implicated in RMS development include ephrin receptors93. High PAX7 expression sustains migration and invasiveness in ERMS cells by upregulating the ephrin receptors A3 (EPHA3) and A1 (EFNA1)94. Moreover, ERMS cell lines deficient in EPHA2 and EPHB2 have reduced migratory capacity and are induced to differentiate into a myogenic-like phenotype95. Similarly, increased signalling via MET in FNRMS contributes to invasive tumour growth96,97, and PAX3–FOXO1-driven MET expression in FPRMS promotes motility, mainly through MAPK1 (ref.98), and blocks myogenic differentiation99,100. Platelet-derived growth factor receptor-α (PDGFRA) gene expression is driven by the fusion protein in FPRMS, and PDGFRA is occasionally mutated in FNRMS, whereas PDGFRB is primarily expressed in the vascular stroma of RMS28,62,101. Together, PDGFR activity regulates cancer cell stemness, differentiation, senescence and apoptosis, with the stromal compartment providing a supportive role102.
Loss of PTEN, TP53 and CDKN2A
Mutation or promoter methylation of the tumour suppressor PTEN can occur in FNRMS and negatively regulates PI3K signalling30. In a recent targeted resequencing study of tumour cells from 631 patients with FNRMS, TP53 mutations were identified in 12% of tumours, which is a higher incidence than had been previously reported28,103. High expression of MDM2 occurs in 10% of patients with RMS28,30 and negatively regulates p53 (encoded by TP53). In addition, promoter methylation, allelic loss and mutation of the tumour suppressor CDKN2A is frequent. The finding of p53 pathway disruption or loss of CDKN2A in RMS is concordant with model systems in which these changes are synergistic and critical for increasing penetrance in PAX3–FOXO1 transgenic mouse models30,40,104,105 (Box 1).
Involvement of developmental pathways
The β1-catenin gene, CTNNB1, is commonly mutated in FNRMS28,30,74, and β-catenin and other proteins associated with canonical Wnt signalling (a key pathway involved in development) are expressed in a high proportion of FNRMS and FPRMS tumours106. Notably, inhibition of Wnt family member 3a (WNT3A) protein or glycogen synthase kinase 3β (GSK3β) in cultured RMS cells results in the nuclear translocation and transcriptional activation of β-catenin in RMS cells, followed by decreased proliferation and the induction of differentiation, which suggests a potential therapeutic approach106,107. Analyses of primary tumours, cell lines and mouse models support other developmental pathways contributing to RMS development and progression, including Hedgehog51,108,109,110 (Box 1), Notch111 and Hippo signalling pathways79,112,113. Crosstalk between these developmental pathways and RAS signalling114 creates an integrated signalling network that supports the development of RMS111,115.
Myogenic regulation and epigenetics
Mutations in MYOD1 occurring together with mutations in genes of the PI3K–AKT pathway define a particularly aggressive form of FNRMS in children and adults116. Interestingly, DNA-binding sites of L122R-mutant MYOD1 are similar to those of the MYC proto-oncogene protein117, potentially explaining a switch from differentiation to proliferation. In addition, mutations may lead to decreased MYOD1 binding adjacent to critical myogenic genes, which is thought to contribute to reduced expression of the myogenic programme in RMS118. In FPRMS, the histone methyltransferase KMT1A is highly expressed and associates with MYOD1, thereby adding repressive marks to histone H3K9 residues at MYOD1-regulated target genes, thus blocking differentiation119. These findings add to our understanding of how MYOD1, which usually promotes the differentiation of skeletal myocytes, contributes to the undifferentiated phenotype of RMS. Similarly, PAX3–FOXO1-driven expression of JARID2 (a regulator of histone methyltransferase complexes) recruits PRC2 (a repressive histone methyltransferase) via EZH2 (the catalytic subunit of PRC2) to the MYOG and MYLH genes, also contributing to failed myogenic differentiation120,121. Patients with RMS may have mutations in BCOR that inactivate the BCL6 co-repressor, which interacts with histone deacetylases28,122. Mutations also occur in the gene encoding ARID1A, a protein involved in the SWI/SNF chromatin remodelling complex28. Together with global and specific changes in DNA methylation and histone marks, and expression of microRNAs and other non-coding RNAs30,63,123, there is increasing evidence for epigenetic regulation shaping RMS development and progression.
Cellular and physiological processes
From a physiological and cellular level, metastasis is particularly important and often the lethal component of RMS124. Early studies of FPRMS suggested that the presence of the PAX3–FOXO1 fusion transcript correlated with greater metastasis risk than PAX7–FOXO1 (ref.69), but that initial finding has not been consistently validated in clinical trials. Certain experimental models implicate cannabinoid receptor 1 (Cnr1) or the cytoskeleton-associated protein ezrin as potential mediators of metastasis induced by the PAX3–FOXO1 fusion protein125 or the sine oculis gene (SIX1)126, respectively. Although PAX3–FOXO1 is clearly important in the human disease, whether SIX1 represents a potentially actionable vulnerability or biomarker for metastasis in RMS is not clear. In many epithelial cancers, the epithelial-to-mesenchymal transition (EMT; a complex series of molecular and cellular changes enabling epithelial cells to mobilize during normal organism development or in healing of certain epithelial wounds) plays a critical role to increase cancer cell mobility and drive metastasis127,128. As a mesenchymal cancer, it remains unclear whether RMS undergoes an EMT-like process. Interestingly, genes well known to induce EMT, SNAI1 (refs129,130) and SNAI2 (refs131,132), have been implicated as oncogenic drivers for RMS. Expression of SNAI1 and SNAI2 derails the myogenic differentiation programme129,133, which is associated with cell proliferation arrest. However, the underlying cellular mechanisms for metastasis control in RMS remain to be elucidated.
Evading immune surveillance is also widely recognized as a key process in human cancer134,135,136. The immune landscape of RMS is understudied. Indeed, little is known about interactions between RMS cells and the tumour stroma and non-stromal elements of the tumour microenvironment, including immune cells. Large next-generation sequencing studies of RMS have focused on characterizing tumour cell-intrinsic features, such as single-nucleotide variants and gene copy-number alterations, not the tumour microenvironment or stroma28,29. Similarly, transcriptome profiling identified gene expression signatures correlating with RMS subtype and prognosis75,78,137 but did not reveal substantial insight into stromal elements, immune surveillance escape mechanisms or metastasis. Interestingly, gene expression analysis of RMS occurring on the extremities138 indicated that higher expression of immune genes (for example, β2-microglobulin, complement genes and major histocompatibility complex genes) in the primary tumour correlated with regional lymph node metastasis risk. Defining the tumour stroma and immune cell landscape in RMS may reveal new insights into disease biology. The potential for immune-based therapeutics could still be limited by the fairly low mutational burden in RMS28,29. Obviously, the fusion proteins driving FPRMS represent neoantigens that, in principle, could be amenable to immune targeting. That possibility is supported by results from a small pilot study suggesting improved survival for children with FPRMS (and also Ewing sarcoma) who received post-cytoreductive consolidation using dendritic cells pulsed with peptides derived from the relevant fusion protein139.
Diagnosis, screening and prevention
RMS is a global problem, and many factors exist that influence delivery of care to childhood patients with cancer, including those with RMS (Box 2). The following section focuses on how the disease presents and is managed in what could be construed as an idealized setting.
The diagnosis of RMS has been traditionally based on recognizing the features of skeletal myoblast-like tumour cells using light and, in some cases, electron microscopy, and the use of immunohistochemical (IHC) staining for skeletal muscle proteins13,140. RMS can arise in virtually any anatomical site; however, the most common sites depend on the histological subtype: ERMS most commonly arises in the head and neck, including the eye socket, or in genitourinary sites; ARMS typically arises at extremity sites, with a smaller fraction arising in the head and/or neck or torso17. Radiographic or clinical evidence for distant metastatic disease is present in ~20% of children at diagnosis1. Metastases arise by both lymphatic and haematogenous routes, and spreading to the lung, bone and bone marrow is fairly common141.
Few comprehensive studies exist of presenting signs or symptoms in children with RMS. Signs and symptoms are typically associated with soft tissue mass and are often described as painless masses found in the extremities or head and/or neck region142,143; they can also be associated with signs and symptoms due to mass effect on adjacent organs or neurovascular tissues or associated with a visible mass protruding from an orifice (for example, a ‘grape-like’ mass in botryoid RMS of the vagina)142,144,145. Orbital primary sites typically present as a unilateral, space-occupying lesion with proptosis146.
The diagnosis of RMS requires the direct analysis of tumour tissue from either an incisional or excisional biopsy sample or core needle biopsy sample that is subjected to a series of histology and molecular pathology studies. The WHO had recognized three histological variants of RMS — ARMS, ERMS and pleomorphic RMS — with ARMS and ERMS being the most frequent childhood forms. A recent update from the WHO now includes spindle cell/sclerosing RMS as a distinct entity13,147. Morphologically, RMS cells are of heterogeneous shapes including undifferentiated and round cells, ovoid cells, ‘tadpole-like’ cells, spindle-shaped cells and fully differentiated rhabdomyoblasts148,149.
As previously mentioned, RMS cells usually display some evidence for skeletal muscle lineage specification and/or differentiation evident by light or electron microscopy and/or IHC or molecular evidence for expression of skeletal muscle gene products, such as muscle-specific actin and myosin, desmin, myoglobin, Z-band protein and MYOD1 or MYOG. Notably, a diagnosis of RMS can be made even if only a minority of tumour cells display detectable expression of skeletal muscle proteins. Such a finding reflects the failed terminal differentiation that is central to RMS. However, cellular morphology and architecture must also be considered for accurate diagnosis because myogenic proteins can be expressed in other childhood neoplasms, such as Wilms tumour and malignant triton tumour149.
ARMS and ERMS
ARMS and ERMS can be subdivided further on the basis of histological or molecular features. For example, recognized histological variants include botryoid RMS and spindle cell/sclerosing RMS, which are generally associated with a superior prognosis13. In addition, ARMS can be subdivided on the basis of the presence or absence of the PAX3–FOXO1 or PAX7–FOXO1 fusion transcript72. ARMS is typically composed of densely packed, small, round cells lining septations that are reminiscent of pulmonary alveoli, whereas ERMS comprises immature ‘rhabdomyoblasts’ in a less dense, stroma-rich background without an alveolar pattern150 (Fig. 4). It should be noted that some RMS specimens can display intratumour heterogeneity with a ‘mixed’ phenotype of both alveolar and embryonal features; fusion gene detection in tumour classification is particularly valuable in this context151,152.
Pleomorphic RMS is a rare adult variant of RMS that has distinct histological features (Box 3) that are similarly composed of cells displaying evidence of skeletal muscle lineage commitment or differentiation. Pathological review of biopsy samples from 38 primary tumours revealed three histological subtypes — classic, round cell and spindle cell — on the basis of general morphology. The pleomorphic rhabdomyoblasts that define this subtype express a range of skeletal muscle proteins detected by immunohistochemistry, which helps to discriminate it from adult pleomorphic sarcoma23,153. Pleomorphic RMS generally arises in older adults, more commonly in males and often involves the lower extremities154. Owing to its rarity, little is known about the pathogenesis of pleomorphic RMS. For example, one report describes very complex cytogenetic and/or chromosomal abnormalities, similar to that seen in children with FNRMS155. Defining molecular abnormalities are not clear for this disease.
The advent of tools for molecular diagnostics has greatly aided the identification of RMS. In particular, ARMS is more precisely diagnosed as FPRMS on the basis of detecting the presence of PAX–FOXO1 fusion in tumour cells, observed using FISH, or by detecting the fusion transcript by RT-PCR assays156,157 (Fig. 5). Systematic application of molecular diagnostic tests to biopsy samples from patients with pathologically diagnosed ARMS reveals that ~20% of these patients are negative for the fusion transcript150. This finding is particularly important because fusion-negative ARMS has molecular features reminiscent of ERMS and the clinical outcome of children with fusion-negative ARMS parallels that of children with ERMS158,159.
A molecular diagnosis based on fusion status clarifies some confusing issues regarding the routine pathology of RMS. For example, the fusion transcript can be detected in many patients with ARMS tumours that have a solid morphology, with rather few alveolar spaces and in some cases with ‘mixed’ alveolar and embryonal features, clarifying the diagnosis150. With the exception of pleomorphic RMS in adults, all patients with FNRMS are felt to have an equivalent disease to ERMS. However, spindle/sclerosing RMS and botryoid subtypes of ERMS disease can have distinct clinical features and may also carry distinct molecular defects, such as MYOD1 mutation. Finally, trisomy of chromosome 8 and loss of heterozygosity at 11p15 are also fairly common in FNRMS28,29. Because these chromosomal aberrations are found in other childhood cancers (including some patients with FPRMS), they are not routinely employed to make a diagnosis of FNRMS.
The clinical relevance of fusion status over histology underlines the importance of evaluating fusion gene status to complement traditional pathology assessment, which will enable the diagnosis of FPRMS to be made using a molecular assay. A recent study evaluated the potential for FPRMS and FNRMS to be distinguished by a panel of IHC stains, rather than FISH, a more accessible approach for smaller clinical programmes. Strong IHC staining for MYOG, AP2β and NOS1 can identify FPRMS, whereas HMGA2 expression is consistent with FNRMS, and the diagnosis may be made using an algorithmic approach160. Ongoing COG trials are incorporating fusion gene status into patient risk stratification (NCT02567435); however, diagnosis using only fusion gene detection has not yet risen to routine practice, nor has the use of IHC staining as a surrogate for evaluating fusion status.
Surveillance and prevention
Surveillance strategies for RMS, similar to most childhood cancers, are still in early stages of development, and preventive strategies are not clear. In principle, the recognition of genetic syndromes that predispose to RMS (and other cancers) should provide a foundation on which to focus RMS surveillance strategies. Indeed, at least one report shows the potential value of biochemical and imaging-based surveillance for children with Li–Fraumeni syndrome161. This research led to a consensus statement advocating for cancer surveillance in children with germline p53 variants162 and to recognition of the need for further study163. Conceivably, advances in next-generation sequencing enabling sensitive detection of circulating, cell-free tumour DNA may lead to affordable and minimally invasive surveillance for RMS in at-risk populations164.
Over >30 years of study in large, cooperative group clinical trials, the 5-year overall survival of paediatric RMS has improved substantially, such that it now exceeds 70%2,5,6,7,165,166,167 (Fig. 6a). Notably, management and outcome for adults with RMS considerably differs (Box 3). Several factors contribute to this improved survival, including the use of multifaceted therapies that typically include surgical resection of the primary tumour, ionizing radiation to the primary tumour and multi-agent, intensive chemotherapy; the development of clinical and pathological staging systems enabling risk-adapted therapy; and the systematic evaluation of new therapeutic approaches in multi-institutional clinical trials conducted on a national or international scale.
The three main cooperative groups dedicated to RMS are the COG Soft Tissue Sarcoma Committee in North America1,168; the European paediatric Soft Tissue Sarcoma Study Group (EpSSG), which involves many European countries as well as Argentina, Brazil and Israel169; and the Cooperative Weichteilsarkom Studiengruppe der GPOH (CWS) group, which includes predominantly German-speaking countries in Europe170. Cooperative group trials of different childhood soft tissue sarcomas often include the majority of children with that specific disease. For example, a recent EpSSG study comparing the number of children enrolled in EpSSG clinical trials with the expected RMS incidence between 2008 and 2015 showed that 77% of patients with RMS aged <14 years were enrolled169. Notably, the rate of enrolment for older adolescents and young adults was somewhat lower, which highlights an opportunity to focus future trials on this population.
Curing RMS first requires eradication of the gross primary tumour (that is, local disease), which is often accomplished using a combination of surgery and/or external beam ionizing radiation. Curative therapy has also included systemic chemotherapy to eradicate disseminated disease that is assumed or proven to exist in most children with RMS. That most, if not all, patients have disseminated disease is based, in part, on observations from the 1960s and 1970s, including a Children’s Cancer Study Group A (CCSGA) trial in the United States showing common regional and distant recurrence in 8 of 15 children treated without chemotherapy171. Further, sensitive molecular biology tools such as RT-PCR have been reported to detect RMS cells in peripheral blood or bone marrow in 12 of 16 children with RMS, even those with grossly localized disease172. The COG Soft Tissue Sarcoma Committee (and its predecessors, the CCSGA and IRSG) developed and systematically tested a wide variety of approaches to achieve both goals (reviewed in refs1,168) over >40 years, and some differences in approaches have evolved over that time. The current standards of care — where they exist — and ongoing research goals in the field are outlined in more detail below.
A major advancement in RMS management is the capacity to define risk groups on the basis of clinical, pathological and, increasingly, molecular features1,173. Improvements in risk stratification have enabled tailored therapy. The IRSG first introduced the concept that RMS disease status can be described in terms of both ‘stage’ and clinical ‘group’168 (Tables 4,5). RMS stage, from 1 to 4, depends on the anatomical site of the primary tumour (with involvement of the bladder and/or prostate, or the extremities, indicating a more advanced stage), tumour size, presence or absence of regional lymph node involvement and presence or absence of distant metastasis174. Clinical Group, from I to IV, applies surgical and/or pathological features, including the degree of resection of localized or regionally spread tumour and the presence or absence of distant metastasis175. Recent recognition that molecularly defined RMS (that is, FPRMS or FNRMS)152,158,159 and more detailed consideration of clinical features, including the number of metastatic sites141, improves our ability to predict outcome may enable further refined stratification. Furthermore, numerous ‘metagene’ expression signatures may have prognostic value, especially in FNRMS137,158. The most robust of these is a five-gene metagene (MG5) signature, which was first defined as a prognostic variable in FNRMS in a European cohort, and has been validated using additional cohorts, including one from the COG176. Although promising, the MG5 signature has not yet been widely incorporated in clinical practice.
Risk stratification is used to determine treatment allocation in clinical trials; however, the details of risk stratification differ in North American and European groups. For example, a recent EpSSG protocol (RMS 2005) defined low-, standard-, high- and very-high-risk groups for children with RMS on the basis of tumour histology, anatomical site and degree of surgical resection of the primary tumour and presence or absence of detectable metastatic disease177. By contrast, the most recent series of COG trials (ARST0331, ARST0431 and ARST0531) incorporated only three risk groups1 (Table 6; Fig. 6). The overlap between the EpSSG high-risk and COG intermediate-risk groups is particularly noteworthy because children with histologically defined ARMS and regional lymph node involvement are classified as very high risk by EpSSG but intermediate risk by COG1. Thus, developing informatics infrastructure to enable comparisons across these two cooperative groups represents a challenge and an opportunity to establish a global resource for collection, storage and sharing of RMS clinical trial data.
Intermediate-risk and high-risk disease
Systemic therapy for RMS continues to be based on a foundation of intensive, alkylating-based, multi-agent chemotherapy administered at intervals over 6–9 months. In North America, the combination of vincristine, actinomycin D and cyclophosphamide (known as VAC) is the standard backbone therapy for RMS, whereas in Europe, the standard is ifosfamide, vincristine and actinomycin D (known as IVA). Notably, a randomized comparison of either VAC or IVA as the initial therapy (followed by VAC for all patients) showed similar outcomes2.
The EpSSG RMS 2005 trial evaluated two randomized questions. First, the efficacy and safety of doxorubicin in patients with grossly localized, but otherwise high-risk RMS was studied after an initial pilot study demonstrated activity and safety of IVA plus doxorubicin (IVADo)178. However, the EpSSG RMS 2005 study failed to show that IVADo improved event-free survival177. Second, this group studied the efficacy of 24 weeks of maintenance therapy with low-dose, continuous chemotherapy of oral cyclophosphamide and vinorelbine in children with localized RMS who achieved a complete radiographic response after 27 weeks of intensive therapy. Previous evaluation of the maintenance therapy in a pilot study had demonstrated activity in patients with recurrent RMS179. In the EpSSG RMS 2005 trial, the addition of maintenance therapy improved overall survival (87.3% in maintenance arm versus 77.4% in standard treatment arm at 3 years post-randomization, P = 0.011), with a marginal improvement in disease-free survival (78.4% versus 72.3%, P = 0.061)180.
In North America, the COG conducted trials to test the addition of a camptothecin drug to the VAC backbone in randomized studies of children with intermediate-risk disease. The D9803 trial, conducted between 1999 and 2005, failed to show improved event-free survival with the addition of topotecan5. The subsequent ARST0531 trial compared standard VAC versus VAC alternating with vincristine–irinotecan, on the basis of the particularly high activity of vincristine–irinotecan in a phase II ‘window’ study181, and the 4-year event-free survival and overall survival were similar in the two arms182. Presently, the VAC backbone for RMS therapy remains a commonly accepted, North American standard for those with intermediate-risk disease, whereas IVA is the European standard for those with localized disease, which includes patients similar to the COG intermediate-risk stratum177,183 (Table 6).
Notably, the outcome for children with metastatic disease is very poor, with overall survival at 3 years of ~25–30%, despite intensive systemic therapy, even therapy including high-dose chemotherapy and autologous stem cell rescue1,141,184. In an attempt to improve these poor outcomes, the COG recently conducted a phase II study (ARST0431) testing whether an ‘interval compression’ strategy, in which the interval between chemotherapy doses was decreased, utilizing doxorubicin, ifosfamide, etoposide and irinotecan to the VAC backbone. ARST0431 showed an improved event-free survival for a subset of children with metastatic RMS compared with historical control individuals185.
In contrast to children with metastatic disease, outcomes are excellent for children with low-risk disease, generally considered to be those with localized, histologically confirmed ERMS (in favourable anatomical sites, localized and grossly resected ERMS and ERMS localized to the orbit only; Tables 4,5)173. The most recent low-risk study in the COG, ARST0331, investigated two strategies to reduce the burden of therapy without compromising survival. In one subset of patients with stage 1 or 2–group I or II ERMS, or stage 1–group III orbital ERMS, therapy was shortened to 24 weeks and total cyclophosphamide dose was decreased186. This less intensive therapy resulted in excellent 3-year failure-free survival (survival without relapse of disease) and overall survival (89% and 98%, respectively). The second subset of patients included those with stage 1–group III, non-orbital or stage 3–group I or II ERMS; in these patients, total cyclophosphamide dose was decreased but vincristine and actinomycin chemotherapy was kept at a standard duration of 48 weeks. Notably, radiation therapy was also omitted for female patients with vaginal tumours achieving a complete response with or without surgical resection187. Although the 3-year overall survival was still very good (92%), the failure-free survival was 70% overall and 57% for girls with genital tract tumours. COG investigators concluded that this failure-free survival is suboptimal if the primary goal is to prevent disease recurrence and avoid the even more intensive therapy that is probably needed to achieve ultimate cure.
It is worth noting that the ‘philosophy’ in the European cooperative groups has been primarily focused on reducing morbidities associated with local control while retaining excellent overall survival. As such, a higher incidence of local disease recurrence has often been tolerated, with the understanding that a more intensive therapy may need to be employed at recurrence but that the majority of individuals would not be exposed to that therapy6. Studying the relative benefits and costs of the North American and European approaches would be valuable, especially for those in which the chance of survival is high; therefore, concern for acute and late toxicities and cost-effectiveness become more important.
Special management considerations
Local disease control
Local control of RMS depends on surgical resection and/or use of ionizing radiation. RMS represents a cancer that is initially sensitive to cytotoxic chemotherapy and radiation; therefore, surgical approaches are generally limited to those that are not compromising to form or function188,189. Nevertheless, the improved survival seen in patients with group I and II versus group III disease (Fig. 6b) supports the importance of primary tumour resection when possible3,175,190. Historically, ionizing radiation was classified as a crucial adjunct to surgical resection to optimize local disease control. However, in current practice, the dose is adjusted on the basis of the degree of previous surgical resection to reduce radiation-related sequelae, such as skeletal muscle and/or soft tissue changes, joint stiffness, axial and appendicular skeletal growth problems and secondary malignancy191,192,193. In European trials, radiation therapy is omitted in certain circumstances in an attempt to spare the majority of patients from adverse treatment effects as long as effective salvage therapy is available.
RMS in very young children
Children aged <3 years, especially infants up to 12 months old, pose a management challenge because their outcome is often worse than in older children194,195. Reports suggest that the worse outcome might arise in part from hesitancy to offer the same local disease control or chemotherapy dose intensity that is typically applied to older children owing to toxicity concerns. Also conceivable is the fact that the pathophysiology of RMS may differ according to age. Indeed, a recent report showed variant gene fusions involving VGLL2 fusion to either CITED2 or NCOA2 in infants with spindle cell/sclerosing RMS196. As next-generation sequencing analyses move forward, correlation of molecular genotype with clinical prognostic features (such as anatomical site, tumour stage and clinical group) and age may provide additional evidence to support this possibility.
Management of recurrent RMS
Recurrent RMS has a very poor prognosis, and a standard of care has not been widely accepted. Overall survival following RMS relapse depends on several factors, including the stage of disease at original diagnosis, the site of relapse and tumour histology. In general, patients undergoing more intensive therapy at original diagnosis had worse outcomes following relapse than patients receiving less intensive therapy at the time of original diagnosis. For example, individuals with relapse after treatment for stage 1–group I ERMS disease displayed a 5-year event-free survival after relapse of 52% compared with ~20% for those with stage 3–group III disease and ~12% for those with stage 4 disease at original diagnosis197. In addition, children originally diagnosed with stage 1–group I disease who had a local recurrence also had a better prognosis than those with regional or distant recurrence197. Unfortunately, post-relapse treatment data were not included in this analysis.
Given the bleak outlook, cooperative groups in North America have been studying novel therapeutic approaches in children with relapsed RMS (and other sarcomas). Phase II trials of promising agents, such as the anti-IGF1 receptor antibody R1507 (ref.198) and the multi-RTK inhibitor sorafenib199, were recently conducted within the COG, but unfortunately neither study showed evidence for significant, single-agent activity. The addition of the mTOR inhibitor temsirolimus to the anti-IGF1 receptor antibody cixutumumab also failed to significantly improve outcomes despite promising preclinical data200. The most recent COG trial for relapsed RMS (ARST0921) employed a backbone of vinorelbine and cyclophosphamide, taking a cue from a previous European study showing that these agents had some activity as low-dose, maintenance therapies179. To that backbone, ARST0921 randomized the addition of either the vascular endothelial growth factor inhibitor bevacizumab or temsirolimus in a phase II pilot study to select one of the two agents for further study168. This pilot study showed relatively more benefit from temsirolimus201 and informed the COG Soft Tissue Sarcoma Committee’s decision to study it in a randomized, phase III study of children with intermediate-risk RMS (ARST1431). Because overall survival was still poor for those children with relapsed RMS, despite the use of temsirolimus, investigators continue to study new therapeutic approaches in children with relapsed RMS.
Quality of life
For almost 40 years, the focus of North American and European clinical trials groups has been on improving survival. Some consideration has been given to quality of life (QOL) issues, such as the general philosophy to avoid form-compromising or function-compromising surgical interventions; however, few QOL data have been collected on those trials. As the majority of children with RMS will be long-term, disease-free survivors, greater consideration should be given to QOL data during and after RMS therapy.
Children with cancer are known to experience high levels of suffering and ongoing symptom burden related to their cancer and its treatment202. Toxicity reporting in paediatric cooperative group studies relies on symptom descriptions documented by health-care providers and then extracted from medical records by researchers. As such, these data may not reflect the experiences of patients. Prospective collection of patient-reported outcomes (PROs) could be used to explore symptom experience by children, from diagnosis and throughout initial therapy. Owing to the wide range of patient ages and the widely ranging sites of disease, collection of PROs is a challenge with current clinical trial models.
Despite the historical challenges, some studies suggest progress in defining symptom and toxicity data and in examining age-based comparisons. For example, in the COG ARST0431 study of children with high-risk RMS, adolescents were less likely to complete therapy (63% of individuals aged >13 years versus 76% of children aged ≤13 years) and more likely to have unplanned dose modifications outside of protocol guidelines (23% versus 2.7%) than younger children203. Furthermore, nausea and vomiting (17% versus 4%) and pain (20% versus 6%) were more likely to be reported in adolescents than in younger patients203. These findings should be interpreted with caution, as symptom reporting may be more effective in adolescents than in younger children. It remains unclear whether PRO collection will overcome that obstacle. In settings in which clinical balance exists, differences in PROs may help guide treatment decisions for an individual patient.
QOL assessments in survivors of RMS therapy are important, yet knowledge gaps still exist. Genitourinary and sexual and/or reproductive health are two areas in which issues may arise in this population; however, those QOL data are mostly from small-series studies. For example, one study of the effects of proton therapy in bladder and/or prostate disease was described in only five children cured of their disease with problems ranging from none to enuresis204. In another small study of four patients >14 years of age and previously treated with multimodality therapy for prostate or bladder RMS, sexual function was assessed using a modified International Erectile Function Index and found that two of the four patients had moderate- or good-quality scores each, respectively, after pelvic radiation therapy and cystectomy205. In a study of 13 patients (median age: 20 years) who had survived pelvic RMS, erectile dysfunction (assessed by the Self-Estimation Index of Erectile Function-No Sexual Intercourse (SIEF-NS) and the Erection Hardness Scale) was present in 10 men, 3 of whom had radiation therapy, whereas 3 men without radiation therapy had no evidence of erectile dysfunction206. In a separate study, 13 of 18 survivors who previously underwent partial cystectomy and/or partial prostatectomy, followed by low-dose-rate interstitial brachytherapy, were found to have normal QOL, assessed by a questionnaire derived from the International Workshop on bladder and/or prostate RMS207. In this study, although all pubertal patients reported having normal erections, the validated SIEF-NS tool was not used. In the largest review to date, medical records of 164 individuals who had survived bladder and/or prostate RMS were retrospectively reviewed to identify 44 who had undergone partial urinary bladder cystectomy and had medical record data regarding urinary continence. Of these 44 patients, 12 were described to have had urinary incontinence, frequency or urgency208. How best to target ionizing radiation to large fields, such as the pelvis for those with bladder–prostate RMS, continues to be highly debated209. If survival is similar with and without radiation therapy138,210, prospective collection of data on bladder and sexual function becomes particularly important to guide treatment decisions in those likely to be cured of their disease.
Over the past 30 years, our understanding of the pathophysiology of RMS has become increasingly sophisticated, and these insights have enabled more robust clinical diagnostics and prognostic assessments, which have contributed to the development of more precise therapeutic approaches. These intensive or experimental approaches are aimed at improving the outcome of patients with a poor prognosis and also attempt to reduce harsh acute and lasting treatment-associated effects in individuals likely to be cured by standard approaches.
The next step towards precision medicine entails the development of robust, objective and accessible biomarkers that are highly predictive of response to targeted therapy. One major challenge for RMS is highlighted above — the paucity of highly recurrent and targetable protein coding genes, especially in fusion-driven RMS28,29,30. Current efforts for targeted resequencing of primary RMS biopsy samples will more precisely define the frequency of previously identified somatic variants; correlation of these variants with clinical features, such as age, anatomical site and outcome, should lead to better understanding of tumour biology and enable better risk stratification. However, it seems likely that integrative computational analyses, especially those considering gene expression changes driven by epigenetic reprogramming, will be required to find targetable oncogenic drivers that were not unmasked by nucleic acid sequencing.
In parallel, as the field progresses more clearly into a post-genomics era (Box 1), functional approaches and high-throughput screening also represent important laboratory-based paths towards better therapy. CRISPR–Cas9-based gene editing provides an opportunity for targeted (for example, kinome-wide) or completely unbiased searches for vulnerabilities in RMS211, especially if utilized in well-characterized patient-derived xenograft models that are increasingly available212. Organism-based models in zebrafish213,214 and Drosophila215 are particularly intriguing ways to conduct high-throughput chemical or genetic screens for targetable cancer drivers, especially for fusion-oncogene-driven disease. Those non-mammalian systems have potential pitfalls, but findings there could be confirmed in the elegant mammalian models already employed40,216,217,218 (Box 1).
Another incompletely evaluated approach towards ‘personalized’ RMS therapy includes immunotherapy. Immunotherapy, in which the patient’s dendritic cells are purified, exposed in vitro to neoantigens from the patient’s tumour and returned to the patient, has shown some promise for high-risk sarcoma139,219. However, because immune checkpoint inhibitors such as ipilimumab220 have demonstrated little single-agent activity in RMS, research should focus on identifying predictive biomarkers. In addition, researchers should investigate combination therapies, including coupling checkpoint blockade with ionizing radiation, to increase neoantigen presentation and have a potential abscopal effect on systemic disease221.
Clinical and translational research
Clinical and translational research efforts should continue to enable the development of personalized medical approaches. One short-term goal should be to prospectively validate improved prognostic signatures, such as prospectively validating the MG5, which has already been applied to several independent, retrospective cohorts158,176. Moreover, efforts should be made to develop an MG5 score that could be used in treatment assignment for individual patients with RMS to further aid in risk stratification. In addition, translational studies should focus on whether prognostic biomarker signatures can be enhanced by additional molecular characteristics, such as the presence or absence of RAS gene mutations. These mutations are fairly common and potentially a marker for a more aggressive disease course in those with FNRMS29.
Another direction for future research should be into molecularly targeted therapies, which may further improve outcomes, especially in children with metastatic or recurrent disease. The outcomes for these children remain bleak, and numerous attempts to intensify therapy by the addition of new cytotoxic agents or by interval compression have failed. One ongoing COG trial (ARST1431) for children with intermediate-risk disease is a randomized, phase III study, adding the mTOR inhibitor temsirolimus to the VAC–VI therapy backbone used in North America. This trial is based on a previous study that found that the addition of temsirolimus to vinorelbine and cyclophosphamide improved survival in children with relapsed RMS compared with the addition of bevacizumab to vinorelbine and cyclophosphamide201. As new, molecularly targeted therapies are identified, they could be studied in the context of a ‘basket’, such as the Pediatric MATCH trial, supported in North America by the COG and the National Cancer Institute222. This ‘basket’ trial provides a mechanism to test numerous agents — such as larotrectinib, selumetinib and palbociclib, targeting TRK (also known as NTRK1), MEK, and CDK4–CDK6, respectively — in parallel and in several different types of tumours, including RMS. This type of trial may identify an agent active against RMS, which could then be pursued in a more focused study.
Finally, effort should be devoted to building improved informatics systems to better capture, assimilate and curate the ever-increasing quantity of clinical data arising from cooperative group studies. Moreover, these clinical data can be paired with banked biopsy samples; pathology descriptions and/or actual digital photomicrographic images of pathology specimens, radiographic images, and molecular genetic, epigenetic and proteomic data could greatly facilitate efforts to improve outcomes for children with RMS. Building such a data commons for childhood cancer — or even just for RMS — could enable clinical and translational research discovery in many ways, especially by taking advantage of ever-increasing artificial intelligence capabilities223. Extending this type of effort across North America and Europe will greatly enable trans-Atlantic (or even global) analyses of what will likely become progressively smaller, clinically and molecularly defined RMS subgroups. In addition, expanding the scope of such work to include QOL and PRO analyses may be progressively more important as the likelihood of patient survival continues to improve.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Nature Reviews Disease Primers thanks P.-L. Lollini, B. Schäfer, T. Triche and the other anonymous referee(s) for the peer review of this work.