Background: We aimed to systematically determine the spectrum of fusion-transcripts and their potential relevance for neuroblastoma pathogenesis.
Methods: We analyzed 498 primary tumors by RNA-sequencing. Fusion-transcripts were validated by dideoxynucleotide sequencing. Genomic rearrangements were determined in eleven samples by whole-genome-sequencing. Ectopic FGFR2 expression and FGFR2 knockdown was performed in JoMa1 neural crest progenitor cells and NBL-S neuroblastoma cells, respectively.
Results: We detected a total of 97 fusion-transcripts in 78 tumors, involving 169 genes. In 16/20 cases, genomic rearrangements corresponding to the fusion-transcripts were identified, indicating that they mostly represent fusion-genes. We found that chromosomes 2, 12 and 17 were significantly enriched for genes involved in fusion-transcripts (p<0.05). Fifty-two and 45 fusion-transcripts were predicted to be potentially protein-coding and truncating, respectively. We observed that the 3’-partners of the fusion-genes had significantly higher expression levels in the index cases than their wild-type counterparts in the remaining cohort, suggesting that transcriptional dysregulation is a major effect of the underlying genomic event. Similar to SNV patterns in neuroblastoma, the spectrum of fusion-transcripts was heterogeneous, with most fusion-genes being private events. We noted, however, that fusion-transcripts were significantly enriched for genes annotated in the COSMIC database (p<0.001), and comprised genes previously associated with neuroblastoma (e.g., FOXR1, HACE1 and TERT), pointing towards their potential mechanistic relevance. As an example, we evaluated the functional effect of FGFR2, the full-length coding sequence of which was involved in a KLHL13-FGFR2 fusion. Ectopic expression of FGFR2 resulted in augmented clonogenic growth, while FGFR2 knock-down significantly impaired proliferation. Finally, we observed that the outcome of patients whose tumors harbored fusion-transcripts was significantly worse than that of patients without fusion-transcripts, both in the entire cohort and in patient subgroups, such as MYCN-amplified or high-risk patients.
Conclusions: Our data suggest that fusion-transcripts resulting from genomic rearrangements contribute to an unfavorable neuroblastoma phenotype.