Patient-Derived Xenografts for Investigation of Acquired Resistance in Oncogene-Driven Cancers: Building a Better Mousetrap.

Abstract

The search for preclinical models that accurately reflect the complexity and heterogeneity of human cancer genomics, and moreover are predictive of sensitivity and resistance patterns observed in subsequent clinical trials of new drug therapy, has largely been disappointing. In fact, of the investigational anticancer agents that are demonstrated to be active in existing preclinical models, only a small percentage achieves approval status. In the article that accompanies this editorial, Stewart et al describe their experience developing and testing a panel of epidermal growth factor receptor (EGFR) –mutated patient-derived xenografts (PDXs), with emphasis on comparison of treatment outcomes between the host patient and the corresponding PDX. These data add to the growing body of literature that suggest that this PDX methodology, perhaps above all other preclinical models, most accurately reflects the underlying biology and heterogeneity of individual human lung cancers, and moreover, as discussed here, that PDX treatment faithfully reproduces the clinical outcomes observed in the individual patient donors. The authors are to be congratulated on this work and the insight that it provides into gene signatures that are favorable to engraftment. Nevertheless, this article and others recently published raise as many questions as they answer about both the opportunities and the challenges of using PDXs in drug testing algorithms in the laboratory and even in the clinic. How does the experience of Stewart et al compare with experiences from other PDX data sets? Overall, the data compare favorably with other PDX outcomes emerging in non–small-cell lung cancer (NSCLC) and other tumor types. However, some results differ and are worthy of comment. One potential concern is the low take rate described: only 18% (six of 33) tumors successfully passaged beyond passage one, raising the question of whether these PDXs are really reflective of the larger population of EGFR-mutated cancers. Overall, this rate seems lower than that described in other PDX studies. Although on the surface, this may seem disappointing, it is likely explained by three factors. One is that EGFR-mutated lung cancer tends to have a more indolent biology (eg, EGFR mutation is a favorable prognostic factor in surgically resected disease), and a slower proliferative rate may predispose to reduced success with xenotransplantation. In fact, an earlier article by the same group describes an overall engraftment rate of approximately 40% in NSCLC, with an especially high take rate in KRAS-mutated cancers, mimicking our own experience. Second, the patient population from which these PDXs were derived had primarily early-stage NSCLC (48.5% had stage I disease). Our data suggest that advanced stage NSCLC, both EGFR-mutated and nonmutated, has a higher rate of successful PDX formation. Furthermore, this take rate seems similar between specimens acquired locally and those transported overnight by express mail, establishing transportability in PDX creation. Last, it is probable that the take rate in the nonobese diabetic severe combined immune deficient (NODSCID) mouse model used by Stewart et al is lower than that of its cousin, the NOD-SCID gamma mouse model (Jackson laboratory), although this variable is less well defined. Regardless, refinements in PDX methodology to enhance engraftment rates of earlier stage tumors and those with more indolent biology are needed. Does treatment of PDX models accurately reflect the biology and outcomes of the host patient? Perhaps the most important result of the article by Stewart et al is careful annotation that treatment outcomes of individual PDX models mimic that of the patient from whom that PDX model was derived. In each of the four models tested, histologic and molecular features were similar between the PDX and the donor patient tumor, and response to treatment accurately reflected patient response to EGFR tyrosine kinase inhibitors. However, the ability of PDX models to predict sensitivity and resistance to therapeutic intervention has not yet been fully vetted and requires further validation. An ongoing trial at University of California, Davis (Serial Patient Derived Xenograft Models to Eliminate Cancer Therapy Resistance [SPIDER]) is enrolling patients with cancer (including EGFRmutated NSCLC) whose tumors have a known molecular driver to undergo tumor biopsy before initiation of targeted therapy and again at the time of treatment resistance. Biopsies are used to generate preand post-therapy PDX models that subsequently are treated with the same targeted therapy that the patient received. We anticipate that this study will confirm similar tumor response and genomic characterization in both the patient and the PDX model, before and after development of acquired resistance. What is the future of PDX research? One of the most exciting new directions in lung cancer therapy is that of checkpoint immunotherapeutics, as reflected by the recent US Food and Drug Administration approval of nivolumab, a programmed death-1 (PD-1) –directed drug. Until now, preclinical models have been limited in testing these JOURNAL OF CLINICAL ONCOLOGY E D I T O R I A L VOLUME 33 NUMBER 26 SEPTEMBER 1