The purpose of this study was to develop a method of classifying cancers to specific diagnostic categories based on their gene expression signatures using artificial neural networks (ANNs). We trained the ANNs using the small, round blue-cell tumors (SRBCTs) as a model. These cancers belong to four distinct diagnostic categories and often present diagnostic dilemmas in clinical practice. The ANNs correctly classified all samples and identified the genes most relevant to the classification. Expression of several of these genes has been reported in SRBCTs, but most have not been associated with these cancers. To test the ability of the trained ANN models to recognize SRBCTs, we analyzed additional blinded samples that were not previously used for the training procedure, and correctly classified them in all cases. This study demonstrates the potential applications of these methods for tumor diagnosis and the identification of candidate targets for therapy.

Neuroblastoma is a malignancy of the developing sympathetic nervous system that often presents with widespread metastatic disease, resulting in survival rates of less than 50%1. To determine the spectrum of somatic mutation in high-risk neuroblastoma, we studied 240 cases using a combination of whole exome, genome and transcriptome sequencing as part of the Therapeutically Applicable Research to Generate Effective Treatments (TARGET) initiative. Here we report a low median exonic mutation frequency of 0.60 per megabase (0.48 non-silent), and remarkably few recurrently mutated genes in these tumors. Genes with significant somatic mutation frequencies included ALK (9.2% of cases), PTPN11 (2.9%), ATRX (2.5%, an additional 7.1% had focal deletions), MYCN (1.7%, a recurrent p.Pro44Leu alteration), and NRAS (0.83%). Rare, potentially pathogenic germline variants were significantly enriched in ALK, CHEK2, PINK1, and BARD1. The relative paucity of recurrent somatic mutations in neuroblastoma challenges current therapeutic strategies reliant upon frequently altered oncogenic drivers.

Despite gains in survival, outcomes for patients with metastatic or recurrent rhabdomyosarcoma (RMS) remain dismal. In a collaboration between the National Cancer Institute, Children's Oncology Group, and Broad Institute, we performed whole-genome, whole-exome and transcriptome sequencing to characterize the landscape of somatic alterations in 147 tumor/normal pairs. Two genotypes are evident in RMS tumors; those characterized by the PAX3 or PAX7 fusion and those that lack these fusions but harbor mutations in key signaling pathways. The overall burden of somatic mutations in RMS is relatively low, especially in tumors that harbor a PAX3/7 gene fusion. In addition to previously reported mutations of NRAS, KRAS, HRAS, FGFR4, PIK3CA, CTNNB1, we found novel recurrent mutations in FBXW7, and BCOR providing potential new avenues for therapeutic intervention. Furthermore, alteration of the receptor tyrosine kinase/RAS/PIK3CA axis affects 93% of cases providing a framework for genomics directed therapies that might improve outcomes for RMS patients.

The majority of neuroblastoma patients have tumors that initially respond to chemotherapy, but a large proportion of patients will experience therapy-resistant relapses. The molecular basis of this aggressive phenotype is unknown. Whole genome sequencing of 23 paired diagnostic and relapsed neuroblastomas showed clonal evolution from the diagnostic tumor with a median of 29 somatic mutations unique to the relapse sample. Eighteen of the 23 relapse tumors (78%) showed mutations predicted to activate the RAS-MAPK signaling pathway. Seven events were detected only in the relapse tumor while the others showed clonal enrichment. In neuroblastoma cell lines we also detected a high frequency of activating mutations in the RAS-MAPK pathway (11/18, 61%) and these lesions predicted for sensitivity to MEK inhibition in vitro and in vivo. Our findings provide the rationale for genetic characterization of relapse neuroblastoma and show that RAS-MAPK pathway mutations may function as a biomarker for new therapeutic approaches to refractory disease.

Alveolar rhabdomyosarcoma is a life-threatening myogenic cancer of children and adolescent young adults, driven primarily by the chimeric transcription factor PAX3-FOXO1. The mechanisms by which PAX3-FOXO1 dysregulates chromatin are unknown. We fi nd PAX3-FOXO1 reprograms the cis -regulatory landscape by inducing de novo super enhancers. PAX3-FOXO1 uses super enhancers to set up autoregulatory loops in collaboration with the master transcription factors MYOG, MYOD, and MYCN. This myogenic super enhancer circuitry is consistent across cell lines and primary tumors. Cells harboring the fusion gene are selectively sensitive to small-molecule inhibition of protein targets induced by, or bound to, PAX3-FOXO1-occupied super enhancers. Furthermore, PAX3-FOXO1 recruits and requires the BET bromodomain protein BRD4 to function at super enhancers, resulting in a complete dependence on BRD4 and a signifi cant susceptibility to BRD inhibition. These results yield insights into the epigenetic functions of PAX3-FOXO1 and reveal a specifi c vulnerability that can be exploited for precision therapy.SIGNIFICANCE: PAX3-FOXO1 drives pediatric fusion-positive rhabdomyosarcoma, and its chromatinlevel functions are critical to understanding its oncogenic activity. We fi nd that PAX3-FOXO1 establishes a myoblastic super enhancer landscape and creates a profound subtype-unique dependence on BET bromodomains, the inhibition of which ablates PAX3-FOXO1 function, providing a mechanistic rationale for exploring BET inhibitors for patients bearing PAX-fusion rhabdomyosarcoma. Cancer Discov; 7(8);

The Ewing sarcoma family of tumors (EFT) is a group of highly malignant small round blue cell tumors occurring in children and young adults. We report here the largest genomic survey to date of 101 EFT (65 tumors and 36 cell lines). Using a combination of whole genome sequencing and targeted sequencing approaches, we discover that EFT has a very low mutational burden (0.15 mutations/Mb) but frequent deleterious mutations in the cohesin complex subunit STAG2 (21.5% tumors, 44.4% cell lines), homozygous deletion of CDKN2A (13.8% and 50%) and mutations of TP53 (6.2% and 71.9%). We additionally note an increased prevalence of the BRCA2 K3326X polymorphism in EFT patient samples (7.3%) compared to population data (OR 7.1, p = 0.006). Using whole transcriptome sequencing, we find that 11% of tumors pathologically diagnosed as EFT lack a typical EWSR1 fusion oncogene and that these tumors do not have a characteristic Ewing sarcoma gene expression signature. We identify samples harboring novel fusion genes including FUS-NCATc2 and CIC-FOXO4 that may represent distinct small round blue cell tumor variants. In an independent EFT tissue microarray cohort, we show that STAG2 loss as detected by immunohistochemistry may be associated with more advanced disease (p = 0.15) and a modest decrease in overall survival (p = 0.10). These results significantly advance our understanding of the genomic and molecular underpinnings of Ewing sarcoma and provide a foundation towards further efforts to improve diagnosis, prognosis, and precision therapeutics testing.

Summary Neuroblastoma is a pediatric malignancy that typically arises in early childhood and is derived from the developing sympathetic nervous system. Clinical phenotypes range from localized tumors with excellent outcomes to widely metastatic disease where long-term survival is approximately 40% despite intensive therapy1. A previous genome-wide association study (GWAS) identified common polymorphisms at the LMO1 gene locus that are highly associated with neuroblastoma susceptibility and oncogenic addiction to LMO1 in the tumor cells2. Here we sought to discover the causal DNA variant at this locus and the mechanism by which it leads to neuroblastoma tumorigenesis. We first imputed all possible genotypes across the LMO1 locus and then mapped highly associated single nucleotide polymorphism (SNPs) to areas of chromatin accessibility, evolutionary conservation, and transcription factor binding sites. SNP rs2168101 G>T was the most highly associated variant (combined P=7.47×10-29, Odds Ratio 0.65, 95% CI: 0.60-0.70) and resided in a super-enhancer defined by extensive acetylation of histone H3 lysine 27 within the first intron of LMO1. The ancestral G-allele that is associated with tumor formation resides in a conserved GATA transcription factor binding motif. We show that the newly evolved protective TATA allele is associated with decreased total LMO1 expression (P=0.028) in neuroblastoma primary tumors and ablates GATA3 binding (P<0.0001). We demonstrate allelic imbalance favoring the G-containing strand in tumors heterozygous for this SNP as demonstrated both by RNA sequencing (P<0.0001) and reporter assays (P=0.002). These findings show that a recently evolved polymorphism within a super-enhancer element in the first intron of LMO1 influences neuroblastoma susceptibility through differential GATA transcription factor binding and direct modulation of LMO1 expression in cis, and this leads to an oncogenic dependency in tumor cells.

Loss of 1p36 heterozygosity commonly occurs with MYCN amplification in neuroblastoma tumors, and both are associated with an aggressive phenotype. Database searches identified five microRNAs that map to the commonly deleted region of 1p36 and we hypothesized that the loss of one or more of these microRNAs contributes to the malignant phenotype of MYCNamplified tumors. By bioinformatic analysis, we identified that three out of the five microRNAs target MYCN and of these miR-34a caused the most significant suppression of cell growth through increased apoptosis and decreased DNA synthesis in neuroblastoma cell lines with MYCN amplification. Quantitative RT-PCR showed that neuroblastoma tumors with 1p36 loss expressed lower level of miR-34a than those with normal copies of 1p36. Furthermore, we demonstrated that MYCN is a direct target of miR-34a. Finally, using a series of mRNA expression profiling experiments, we identified other potential direct targets of miR-34a, and pathway analysis demonstrated that miR-34a suppresses cell-cycle genes and induces several neural-related genes. This study demonstrates one important regulatory role of miR-34a in cell growth and MYCN suppression in neuroblastoma.