ID #1091 The genomic landscape of extrachromosomal amplifications across pediatric brain tumors
Sunita Sridhar, Owen Chapman, Eugene Chow, Rishaan Kenkre, Aditi Dutta, Shanqing Wang, Miguel Brown, Jens Luebeck, Daisuke Kawauchi, Vineet Bafna, Kevin Yip, Megan Paul, Jill Mesirov, Lukas ChavezAbstract
Extrachromosomal DNA (ecDNA) has been linked to oncogene amplification, evolution of drug resistance, and poor outcomes in various human cancers including medulloblastoma, the most common malignant pediatric brain tumor. However, the extent to which ecDNA plays a role in oncogenesis of other pediatric brain tumors has yet to be elucidated. In addition, the relative importance of amplification versus extrachromosomal amplification remains a topic of contention in pediatric and adult cancers alike. To address these questions, we have reanalyzed whole genome sequencing data to identify ecDNA sequences in a large retrospective cohort of 3,003 pediatric tumors from the St. Jude Cloud and the Pediatric Brain Tumor Atlas. Among pediatric CNS tumors, we find ecDNA most frequently in embryonal tumors with multilayered rosettes (ETMR), pediatric high-grade gliomas (pHGG) and medulloblastomas, and examples of ecDNA+ tumors belonging to ultra-rare subtypes of pineoblastoma and ependymoma. Recurrently ecDNA-amplified genomic loci frequently included well-established oncogenes including MYC family proto-oncogenes, cell cycle regulators, and growth factor receptors. By applying variant calling and annotation to WGS data, we examined the distribution of somatic small variants and found that those targeting TP53 were significantly depleted in ecDNA amplified versus chromosomally amplified tumors. Crucially, multivariate Cox regression identifies additive effects of amplification and extrachromosomal amplification which contribute independently to poor outcomes in pediatric tumors. These results constitute a comprehensive map of the genetic and phenotypic diversity of extrachromosomal amplification in pediatric tumors, and underscore the utility of large genomic data resources to drive clinically relevant discoveries.