儿童癌症患者来源异种移植模型中染色体外DNA的保留与克隆行为

Background: Extrachromosomal DNA (ecDNA) is a class of structural variants linked to poor prognosis in pediatric cancers. Patient-derived xenograft (PDX) models are crucial tools for basic and translational cancer research, as they are believed to recapitulate the molecular features and intratumoral heterogeneity present in patient tumors. However, ecDNA demonstrates unique evolutionary dynamics under selective pressure, and the behavior of ecDNA during PDX model development and propagation remains largely uncharacterized. This study investigates the fidelity of PDX models in representing ecDNA from primary tumors. By analyzing ecDNA sequence composition and copy number conservation across a variety of pediatric solid cancers, we evaluate the extent to which PDX models recapitulate the ecDNA landscape observed in human tumors. Methods: Amplicon Architect, a tool that reconstructs focally amplified DNA regions to understand cancer genome architecture, was used to analyze the whole-genome sequencing (WGS) of 338 PDX models and 127 corresponding primary tumors. ecDNA status, sequence, copy number, and associated genes were compared between PDX models and their matched human tumors. Additionally, multiome RNA and ATAC single-cell sequencing of a PDX tumor enabled comparison of ecDNA intratumoral heterogeneity relative to similar data previously obtained from the primary tumor. Results: ecDNA in PDX models largely recapitulated the spectrum of oncogene amplifications observed in human tumors, with MYCN being the most frequently amplified. ecDNA status remained unchanged for a majority of the PDX models (106/127, 83%) compared to the primary tumors, with 20% of previously ecDNA-negative cases acquiring ecDNA during PDX development. Consequently, ecDNA was more prevalent in the PDX models than in their corresponding human tumors (McNemar's test, p = 0.0014). Detailed examination of ecDNA sequences in tumor-PDX pairs showed substantial conservation (67% with >90% sequence overlap) but variable breakpoint concordance. Single-cell analysis demonstrated that rare ecDNA-positive cells from the primary tumor preferentially drive PDX tumor development. Conclusion: This study highlights the prevalence, oncogenic content, and conservation of ecDNA in PDX models relative to pediatric patient tumors. We observed that ecDNA frequently recapitulates oncogene amplifications found in human cancers, is generally preserved during PDX establishment, and reflects subtype-specific patterns across tumor types. These findings support the utility of PDX models in studying ecDNA biology and their implications for the study of pediatric cancer progression and treatment. Longitudinal sampling during PDX tumor growth and under therapeutic pressure could provide valuable insights into the dynamics of molecular evolution, clonal selection, and ecDNA-driven therapy resistance.