Stochastic dynamics of extra-chromosomal DNA

Nuclear extra-chromosomal DNA (ecDNA) is highly prevalent in human tumours. ecDNA amplifications can promote accessible chromatin, oncogen over-expression and imply a worse clinical prognosis. Yet, little is known about the evolutionary process of ecDNA in human cancers. Here, we develop the theoretical foundation of the ecDNA somatic evolutionary process combining mathematical, computational and experimental approaches. We show that random ecDNA segregation leads to unintuitive dynamics even if ecDNA is under very strong positive selection. Patterns of inter- and intra-tumour ecDNA and chromosomal heterogeneity differ markedly and standard approaches are not directly applicable to quantify ecDNA evolution. We show that evolutionary informed modelling leads to testable predictions on how to distinguish positively selected from neutral ecDNA dynamics. Our predictions describe the dynamics of circular amplicons in GBM39 cell line experiments, suggesting a 300% fitness increase due to circular extra-chromosomal EGFRvIII amplifications. Our work lies the basis for further studies to quantitate ecDNA somatic evolutionary processes.