Development of Patient Derived Xenograft Models of Overt Spontaneous Breast Cancer Metastasis: A Cautionary Note

Several approaches are being evaluated to improve the historically limited value of studying transplanted primary tumors derived by injection of cells from established cell lines for predicting subsequent cancer therapy outcomes in patients and clinical trials. These approaches include use of genetically engineered mouse models (GEMMs) of spontaneous tumors, or patient tumor tissue derived xenografts (PDXs). Almost all such therapy studies utilizing such models involve treatment of established primary tumors. An alternative approach we have developed involves transplanted human tumor xenografts derived from established cell lines to treat mice with overt visceral metastases after primary tumor resection. The rationale is to mimic the more challenging circumstance of treating patients with late stage metastatic disease. These metastatic models entail prior in vivo selection of heritable, phenotypically stable variants with increased aggressiveness for spontaneous metastasis; they were derived by orthotopic injection of tumor cells followed by primary tumor resection and serial selection of distant spontaneous metastases, from which variant cell lines having a more aggressive heritable metastatic phenotype were established. We attempted to adopt this strategy for breast cancer PDXs. We studied five breast cancer PDXs, with the emphasis on two, called HCI-001 and HCI-002, both derived from triple negative breast cancer patients. However significant technical obstacles were encountered. These include the inherent slow growth rates of PDXs, the rarity of overt spontaneous metastases (detected in only 3 of 144 mice), very high rates of tumor regrowths at the primary tumor resection site, the failure of the few human PDX metastases isolated to manifest a more aggressive metastatic phenotype upon re-transplantation into new hosts, and the formation of metastases which were derived from de novo mouse thymomas arising in aged SCID mice that we used for the experiments. We discuss several possible strategies that may be employed to overcome these limitations. Uncovering the basis of the failure to detect a high rate of overt spontaneous distant metastases having a heritable phenotype in PDX models may reveal new insights into the biology and treatment of advanced metastatic disease.