Generation of orthotopic patient-derived xenografts from gastrointestinal stromal tumor

Background: Gastrointestinal stromal tumor (GIST) is the most common sarcoma and its treatment with imatinib has served as the paradigm for developing targeted anti-cancer therapies. Despite this success, imatinib-resistance has emerged as a major problem and therefore, the clinical efficacy of other drugs has been investigated. Unfortunately, most clinical trials have failed to identify efficacious drugs despite promising in vitro data and pathological responses in subcutaneous xenografts. We hypothesized that it was feasible to develop orthotopic patient-derived xenografts (PDXs) from resected GIST that could recapitulate the genetic heterogeneity and biology of the human disease. Methods: Fresh tumor tissue from three patients with pathologically confirmed GISTs was obtained immediately following tumor resection. Tumor fragments (4.2-mm(3)) were surgically xenografted into the liver, gastric wall, renal capsule, and pancreas of immunodeficient mice. Tumor growth was serially assessed with ultrasonography (US) every 3-4 weeks. Tumors were also evaluated with positron emission tomography (PET). Animals were sacrificed when they became moribund or their tumors reached a threshold size of 2500-mm(3). Tumors were subsequently passaged, as well as immunohistochemically and histologically analyzed. Results: Herein, we describe the first model for generating orthotopic GIST PDXs. We have successfully xenografted three unique KIT-mutated tumors into a total of 25 mice with an overall success rate of 84% (21/25). We serially followed tumor growth with US to describe the natural history of PDX growth. Successful PDXs resulted in 12 primary xenografts in NOD-scid gamma or NOD-scid mice while subsequent successful passages resulted in 9 tumors. At a median of 7.9 weeks (range 2.9-33.1 weeks), tumor size averaged 473 +/- 695-mm(3) (median 199-mm(3), range 12.6-2682.5-mm(3)) by US. Furthermore, tumor size on US within 14 days of death correlated with gross tumor size on necropsy. We also demonstrated that these tumors are FDG-avid on PET imaging, while immunohistochemically and histologically the PDXs resembled the primary tumors. Conclusions: We report the first orthotopic model of human GIST using patient-derived tumor tissue. This novel, reproducible in vivo model of human GIST may enhance the study of GIST biology, biomarkers, personalized cancer treatments, and provide a preclinical platform to evaluate new therapeutic agents for GIST.