Characterization of extrachromosomal DNAs in tumors through computational analysis of single-cell and bulk sequencing data

项目总结 染色体外DNA(EcDNA)存在于40%的肿瘤中，但在正常细胞中很少发现。重要的是 它们含有并表达来自染色体序列的扩增癌基因。与之相对的是 染色体，ecDNA在细胞分裂过程中不平等地分离到子细胞，因此可以在 肿瘤内单个细胞的高拷贝数。这导致了肿瘤内的异质性(ITH)， 可以使肿瘤细胞亚群具有选择性生长优势，并使其对癌症治疗具有抵抗力。而当 以前的研究集中在染色体突变的ITH如何对肿瘤进化做出贡献，很少有研究 了解ecDNA可能如何影响肿瘤的进化和患者的预后。要解决ecDNA如何 有助于ITH和肿瘤进化，Aim 1将确定来源于以下来源的细胞系的ecDNA的ITH 单细胞DNA测序(scDNA-seq)的患者匹配的原发和复发的胶质母细胞瘤 之前已经生成了标准的批量短读全基因组测序(WGS)数据。至 克服在scdna-seq数据中检测单个ecDNA的技术挑战，我们将采用 在scDNA-seq的高拷贝数片段之间使用“断点”的另一种监督方法 数据作为ecDNA断点的替代，并将这些数据与识别的ecDNA断点相交 参考集中的序列。这种方法将使我们能够研究ecDNA驱动的ITH和 单个细胞系之间的细胞系来源于纵向胶质母细胞瘤。我们还将应用此 利用现有的scDNA-seq数据集评估ecDNA的存在。当前的计算工具 用于预测标准批量短读WGS数据中的ecDNA确定ecDNA的能力有限 断点；因此，我们预计我们提出的方法虽然在概念上很简单，但将具有 对提高我们对ecDNA如何在肿瘤细胞内进化的理解有重大影响。在目标2中，一个 代表多种癌症类型的大量可公开获得的肿瘤批量WGS数据集将是 用来更广泛地描述ecDNA及其对肿瘤进化的影响。我们将表演 利用多种肿瘤的ecDNA和其他基因组特征的综合分析来表征ecDNA和 推断它们形成的潜在分子机制。我们将构建一个机器学习分类器 可以使用非WGS数据(即，完整外显子组和RNA测序)来预测ecDNA的存在 已经是对患者肿瘤进行排序的主要策略，因此，比起 WGS。我们还将对许多单时间点和纵向肿瘤样本进行系统分析，以 表征ecDNA对肿瘤进化选择压力的影响。总体而言，完成这些工作 AIMS将极大地促进我们对ecDNA在肿瘤进化中的理解，从而揭示 EcDNA影响患者的预后，并最终为新的癌症疗法奠定基础。

PROJECT SUMMARY Extrachromosomal DNAs (ecDNAs) are found in 40% of tumors but rarely found in normal cells. Importantly, they contain and express amplified oncogenes derived from chromosomal sequences. In contrast to the chromosomes, ecDNAs segregate unequally to daughter cells during cell division and thus can accumulate at high copy numbers in individual cells within a tumor. This contributes to intratumor heterogeneity (ITH), which can give subsets of tumor cells a selective growth advantage and enable resistance to cancer treatment. While previous studies have focused on how ITH of chromosomal mutations contributes to tumor evolution, little is known about how ecDNAs might impact tumor evolution and patient outcomes. To address how ecDNAs contribute to ITH and tumor evolution, Aim 1 will determine the ITH of ecDNAs for cell lines derived from patient-matched primary and recurrent glioblastoma tumors for which single-cell DNA sequencing (scDNA-seq) and standard bulk short-read whole-genome sequencing (WGS) data have been previously generated. To overcome the technical challenge of detecting individual ecDNAs in scDNA-seq data, we will employ an alternative supervised approach of using `breakpoints' between high-copy number segments in the scDNA-seq data as surrogates for the ecDNA breakpoints and intersect these with the identified ecDNA breakpoint sequences in the reference sets. This approach will enable us to study ecDNA-driven ITH and evolution in single cells between the cell lines derived from the longitudinal glioblastoma tumors. We will also apply this approach to existing scDNA-seq datasets to assess the presence of ecDNAs. Current computational tools used to predict ecDNAs in standard bulk short-read WGS data have limited ability to determine the ecDNA breakpoints in single cells; thus, we anticipate that our proposed approach, while conceptually simple, will have a major impact on improving our understanding of how ecDNAs evolve within the cells of a tumor. In Aim 2, a large cohort of publicly available tumor bulk WGS datasets representing multiple cancer types will be leveraged to characterize ecDNAs more broadly and their effects on tumor evolution. We will perform integrative analysis of ecDNAs and other genomic features using many tumors to characterize ecDNAs and to infer the potential molecular mechanisms underlying their formation. We will build a machine learning classifier that can predict the presence of ecDNAs using non-WGS data (i.e., whole-exome and RNA sequencing) that have been a primary strategy for sequencing patient tumors, and therefore, are more widely available than WGS. We will also systematically analyze many single time point and longitudinal tumor samples to characterize the effects of ecDNAs on evolutionary selection pressures in tumors. Overall, completion of these Aims will greatly advance our understanding of ecDNAs in tumor evolution, thereby shedding light on how ecDNAs impact patient outcomes and ultimately establishing a basis for novel cancer therapeutics.