Abstract 3019: An ultra-accurate NGS assay reveals great variation in mouse ratios in xenograft tumors by mouse strain and cancer type

Background: Tumors are infiltrated by stromal cells including fibroblasts, myofibroblasts, macrophages, lymphocytes, mast cells, endothelial cells, etc. Heterotypic interactions, such as ligand-receptor signaling, between the two support the growth and metastasis of cancer cells. Xenograft models are useful in studying such interactions because mouse stromal cells quickly replace the human counterparts associated with human tumors implanted in immunocompromised mice, and can be readily detected. There is, however, a lack of accurate and large-scale studies of human/mouse ratios in xenograft tumors, although there are several ways of estimating mouse ratios in xenograft tumors, including RT-PCR, RNA-seq and whole-exome sequencing (WES). None of these methods is accurate due to some systematic bias. For example, RT-PCR compares the expression of one or a few genes, e.g. GAPDH, between human and mouse by using species-specific primers that have inequivalent amplification efficiency and large errors. We have developed a NGS assay to quantify human/mouse ratios in human-mouse mixtures more accurately than the methods currently employed. Methods: Our NGS assay simultaneously amplified 108 human and mouse homologous segments, each with identical flanking sequences in human and mouse, so that a pair of common primers was designed to have equal hybridization efficiency in the two species. A total of 5000 tumors from our patient-derived xenograft (PDX) collection were profiled across 1674 models in 24 cancer types. Results: Several interesting observations were identified. (a) The average mouse content across the collection of PDX models was 23%, but with great variation, ranging from zero to 93%. (b) PDXs have significantly more mouse content in BALB/c than in NOD/SCID mice (31.9% vs 22.1%, p value