Abstract LB-276: Characterizing tumorigenesis of ovarian cancer across metastatic tumors and circulating, cell-free DNA

Abstract

Abstract Tumor heterogeneity creates significant challenges for the generation and monitoring of treatment strategies. Circulating, cell-free DNA (cfDNA) can facilitate these processes by providing a noninvasive method to detect and track mutations without the need for multiple biopsies. To study both tumor heterogeneity and its relationship to cfDNA, we used samples taken from multiple positions on two metastatic tumors from a stage 3B ovarian carcinosarcoma patient as well as a cfDNA sample that was collected at the time of surgery. To maximize low frequency variant detection from next-generation sequencing (NGS) data, molecular identifiers (MIDs) were employed, allowing accurate distinction of true variants from PCR-induced errors along with sequencing artifacts. The use of MIDs to uniquely label input DNA generates tagged library molecules for the detection and removal of PCR duplicates while simultaneously preserving fragmentation and strand duplicates. By over-sequencing the MID tagged library, sequencing reads from PCR duplicates can be grouped based on their shared MID tag, generating a consensus sequence to identify and subsequently remove errors. To validate this technology, we performed low frequency spike-in experiments to 0.5% and 1% by combining Coriell NA12878 and HG005 genomic DNA as well as different cfDNA samples. Libraries were sequenced to 8000x coverage and a consensus sequence was generated with BMFtools (ARUP labs). All known variants present at 1% and 0.5% allele frequencies were maintained in the resulting data set. Further, true variants were preserved while sequencing and PCR-induced errors were removed, demonstrating improved sensitivity but also improved specificity using MIDs. After validation of the technology, libraries with MIDs were prepared from the ovarian carcinosarcoma FFPE and cfDNA samples. Libraries were sequenced on an Illumina® HiSeq® to a minimum of 13,000x coverage, and we determined data retention after de-duplication with and without the use of MIDs. We observed an increase in data retention for both Covaris-sheared FFPE and cfDNA libraries that led to a 2 to 3-fold increase in coverage using MIDs. Variant calling depicted significant inter-tumor heterogeneity in these samples as well as low frequency variants that represent intra-tumor heterogeneity. Specifically, we identified a pathogenic TP53 mutation present in one tumor sample and the cfDNA but absent from the second tumor. Additionally, we observed a significant loss of heterozygosity for only one of the tumors. We also identified low frequency, pathogenic mutations (as low as 1%) that are either unique to one tumor or unique to one sample from a tumor. This study highlights the use of MID technology to enable low frequency variant calling during disease diagnosis and tracking to facilitate precise, personalized treatment of complex cancers. Citation Format: Ashley Wood, Sukhinder Sandhu, Matthew Dashkoff, Olga Camacho-Vanegas, Laurie Kurihara, Timothy Harkins, John Martignetti, Vladimir Makarov, Peter Dottino. Characterizing tumorigenesis of ovarian cancer across metastatic tumors and circulating, cell-free DNA [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr LB-276. doi:10.1158/1538-7445.AM2017-LB-276