(P 287) ESTABLISHMENT AND GENOMIC VALIDATION OF NOVEL PATIENT-DERIVED XENOGRAFT MODELS FOR DRUG DISCOVERY IN GASTROINTESTINAL STROMAL TUMOR

Objective: Gastrointestinal stromal tumors (GIST) are the most common mesenchymal tumors in the gastrointestinal tract. GIST patients often develop clinical resistance to approved tyrosine kinase inhibitors (TKI). The development of effective treatment strategies for refractory GIST requires identification of novel therapeutic targets for investigational agents and well-characterized preclinical models to test these agents. Patient-derived xenograft (PDX) models have potential to better preserve the cellular architecture and molecular complexity of parent tumors. GIST PDX models published thus far have been shown to mirror patient tumors in cell morphology and driver mutation status, but none of these models have been comprehensively profiled at the genome-wide level. Recent studies in other cancers indicate that PDX models may exhibit more nuanced changes at the genomic level upon grafting and passaging in mice. Here, we performed a comprehensive genomic analysis of six KIT-mutant GIST PDX models, all derived from patients previously treated with TKIs.

Methods: Six PDX models were established from recurrent metastatic refractory GIST and maintained in vivo over multiple passages in mice. Along with KIT/PDGFRA genotype determination, histochemical and immunochemical analyses and comparisons of patient to PDX passaged tumors, we performed genome-wide copy number (CN) analysis to define and compare chromosomal aberrations, as well as targeted next-generation sequencing (NGS) profiling with Illumina TruSight Oncology 500 (TSO500), to detect additional mutations in KIT and

Results: Pairwise comparison of copy number variation (CNV) profiles for all samples showed a high degree of correlation between PDX and matched parental tumors. However, four models demonstrated moderate to strong change in number of CNV events compared to primary tumor. Aberrations in cancer associated genes detected with TruSight Oncology 500 assay showed stability of primary KIT driver mutations, with minor fluctuations in allelic frequencies of acquired mutations across passages, suggesting initial heterogeneity of GIST tumors and potential enrichment of sub-clonal populations through passaging. Preclinical drug testing of these PDX models at various passages is ongoing and will be reported at the meeting.

Conclusion: We have successfully developed six GIST PDX models with diverse treatment histories and genotypes which recapitulate the frequent polyclonal tumors that arise following treatment with multiple lines of TKIs. These tumors represent an unmet clinical need and novel treatment strategies are needed to address this clinical obstacle. Our GIST PDX models preserve a high level of genomic similarity to their corresponding primary tumors in all key molecular features. These fully characterized PDX models have the potential to improve the quality of preclinical testing of novel treatment strategies and to provide further insights into resistance mechanisms to these agents.