Background. Patient derived orthotopic xenograft (PDOX) models are an important tool for cancer research, including personalized preclinical testing of targeted therapies. PDOX models of CNS tumors faithfully replicate the invasive nature of patient tumors, resulting in variable contamination of the harvested xenograft tissues by mouse brain cells. This presents technical challenges to genomic and transcriptomic analyses, including correct alignment of sequencing reads to human or mouse genomes, and can result in identification of false positive tumor mutations. Improved bioinformatic methods for systematic correction of such errors are needed.

Methods. Whole exome sequencing (WES) was performed on 32 diverse childhood CNS tumors (mean of 109 M reads/sample) and matching PDOX models (mean of 62 M reads/sample), including atypical teratoid rhabdoid tumor (ATRT), glioblastoma (GBM), and medulloblastoma (MB). Patient-matched blood DNA was also subjected to WES. Whole transcriptome sequencing (WTS) was performed on a subset of tumors and PDOX models (mean of 94 M reads/sample). Integrated WES and WTS analysis pipelines were constructed to unequivocally identify each sequencing reads’ species-of-origin prior to genomic and transcriptomic variant calling, expression profiling, and fusion gene analysis. Xenograft NGS reads were competitively mapped to a hybrid reference created by merging the human (hg19) and mouse (mm10) genomes, with reads segregating to their respective genomes based on mapping score. Reads that mapped exclusively to the human reference were selected for analysis.

Results. Mouse DNA contamination varied widely between PDOX tumors (1% to 94%; mean 40%). Hybrid mapping resulted in 2.4-fold higher combined sensitivity and specificity for identification of somatic variants as compared to direct human reference mapping followed by mouse polymorphism subtraction. WES analysis confirmed the maintenance of somatic driver mutations in PDOX models, including SMARCB1 in ATRT, TP53 and NF1 in GBM, and DDX3X in MB. Interestingly, 37% of tumor and 55% of PDOX somatic mutations exhibited significant allele fraction changes between the primary tumor and PDOX model. WTS analysis based on hybrid mapping resulted in an increase in positive predicted value (PPV) for correlation of FPKM values between tumor and PDOX model, with greater PPV in cases with high levels of mouse contamination, albeit with lower sensitivity due to decreasing coverage.

Conclusions. Hybrid mapping based WES and WTS analysis identifies fewer false positives than direct hg19 mapping and represents a more streamlined process of species-of-origin sequencing read classification and filtering at the alignment stage. This method has the potential to improve molecular characterization of PDOX models and inform the design of rational preclinical studies.

Citation Format: Oliver A. Hampton, Michael C. Gundry, Mari Kogiso, Lin Qi, Yuchen Du, Yulun Huang, Frank K. Braun, Huiyuan Zhang, Sibo Zhao, Holly Lindsay, Sarah G. Injac, David A. Wheeler, Xiao-Nan Li, D. William Parsons. A hybrid mapping approach improves genomic and transcriptomic analysis of patient derived orthotropic xenograft (PDOX) models of pediatric CNS tumors [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 3841. doi:10.1158/1538-7445.AM2017-3841