Rearrangement bursts generate canonical gene fusions in bone and soft tissue tumors

INTRODUCTION Gene fusions are often disease-defining events in cancer. The mutational processes that give rise to fusions, their timing relative to initial diagnosis, and whether they change at relapse are largely unknown. Mutational processes leave distinct marks in the tumor genome, meaning that DNA sequencing can be used to reconstruct how fusions are generated. A prototypical fusion-driven tumor is Ewing sarcoma (ES), a bone cancer predominantly affecting children and young adults. ES is defined by fusions involving EWSR1 , a gene encoding an RNA binding protein, and genes encoding E26 transformation-specific (ETS) transcription factors such as FLI1 . We sought to reconstruct the genomic events that give rise to EWSR1-ETS fusions in ES and chart their evolution from diagnosis to relapse. RATIONALE We studied the processes underpinning gene fusions in ES using the whole-genome sequences of 124 primary tumors. We determined the timing of the emergence of EWSR1 fusions relative to other mutations. To measure ongoing mutation rates and evolutionary trajectories of ES, we studied the genomes of primary tumors, tumors at relapse, and metastatic tumors. RESULTS We found that EWSR1-ETS , the key ES fusion, arises in 42% of cases via complex, loop-like rearrangements called chromoplexy, rather than by simple reciprocal translocations. Similar loops forming canonical fusions were found in three other sarcoma types. Timing the emergence of loops revealed that they occur as bursts in early replicating DNA, as a primary event in ES development. Additional gene disruptions are generated concurrently with the fusions within the loops. Chromoplexy-generated EWSR1 fusions appear to be associated with an aggressive form of the disease and a higher chance of relapse. Numerous mutations present in every cell of the primary were absent at relapse, demonstrating that the primary and relapsed diseases evolved independently. This divergence occurs after formation of an ancestral clone harboring EWSR1 fusions. Importantly, we determined that divergence of the primary tumor and the future relapsed tumor occurs 1 to 2 years before initial diagnosis, as estimated from the number of cell division–associated mutations. CONCLUSION Our findings provide insights into the pathogenesis and natural history of human sarcomas. They reveal complex DNA rearrangements to be a mutational process underpinning gene fusions in a large proportion of ES. Similar observations in other fusion-defined sarcoma types indicate that this process operates more generally. Such complex rearrangements occur preferentially in early replicating and transcriptionally active genomic regions, as evidenced by the additional genes disrupted. EWSR1 fusions arising from chromoplexy correlated with worse clinical outcomes. Formation of the EWSR1 fusion genes is a primary event in the life history of ES. We found evidence of a latency period between this seeding event and diagnosis. This is in keeping with the often-indolent nature of symptoms before clinical disease presentation.