Rajendra Kumari, Ph.D. CSO PRECOS

Humanizing mouse models of cancer to expand drug development opportunities.

Introduction

The majority of cancer drugs entering Phase I clinical studies fail to reach the market. The success rate in predicting clinical efficacy of anticancer modalities using standard xenograft models has been reported to be only 30-40%,1-2 which is leading to the re-evaluation of both in vivo and in vitro models. Standard xenograft models use cell lines that are maintained in plastic and have adapted to grow independently of the tumor microenvironment, resulting in models with genetic and phenotypic characteristics distinct from those seen in the clinic.3

Many aspects of the tumor microenvironment need to be understood and considered, including growth characteristics, blood supply, hypoxia, metastasis, stromal interactions, immune response, and resistance. It is therefore not surprising that anticancer agents developed using simplified models have not yielded much success in the clinic. In an attempt to reduce drug attrition and improve clinical predictivity, patient-derived xenograft tumors (PDX) are being used to improve and refine preclinical xenograft modeling. These models provide a more relevant heterogeneous system in which human tumor and stromal cells are in close cooperation within a unique microenvironment, thereby testing anticancer agents in the most relevant manner.

The Challenge to Developing Humanized Models of Cancer

Two-dimensional in vitro models together with xenograft systems have a number of limitations, which may misguide outputs of drug efficacy. Establishing models that more fully recapitulate the tumor microenvironment both in vitro and in vivo can be challenging.

PDX models have been reported to sustain molecular, genetic, and histological heterogeneity of the original tumors, and as such data generated from these models closely resembles clinical data with over 90% prediction of tumor sensitivity and resistance.4 However, one challenge that needs to be overcome is that PDX lose human stroma (Figure 1), which can lead to the loss of human paracrine signaling such as c-met:HGF axis and IL-6 signaling. In addition, PDX models are maintained in serial passage in vivo to avoid contact with plastic, which in itself limits the use for in vitro screening.


Figure 1: Loss of human stroma (vimentin staining) in passaged PDX (bottom panel) compared to original patient tissue (top panel).

There is increasing evidence to suggest that the interactions between stroma and tumor contributes significantly to cancer development and growth, with cancer associated fibroblasts reported to account for over 50% of the tumor mass in some tumors. Stromal reactions to tumors have also been linked to drug resistance; e.g., desmoplasia, the buildup of fibrous tissue, protects the tumor from the toxic effects of chemotherapeutics. In order to retain the human stroma in both PDX and cell-line models, human mesenchymal stem cells (MSCs) or patient-derived cancer-associated fibroblasts (CAFs) can be supplemented with the tumor cells prior to implantation. In vivo, tumors implanted at orthotopic sites in the presence of human stroma are poorly differentiated and highly invasive compared to standard xenografts (Figure 2). Using fibroblasts such as MRC5s, which produce high levels of human HGF, c-met paracrine signaling is optimally recapitulated in mouse models. 

Similarly, for in vitro assays, a 3D-tumor growth assay (3D-TGA) can be established using disaggregated PDX material or cell lines, which are admixed with MSCs or CAFs and seeded into 96 or 384-well format. The 3D-TGA allows for a rapid noninvasive in vitro measurement of cancer cell expansion in the presence of multiple tumor-associated cell types or soluble factors and facilitates the medium throughput screening of test agents. The main advantage of the 3D-TGA is that the experimental tumor cultures are established in complex mixtures of tumor-derived factors with both physical and soluble matrix interactions permitted and spatial limitations greatly reduced, thereby allowing the 3-TGA to more accurately reflect the complex in vivo microenvironment and thus providing a more relevant screening system.


Figure 2: Tumor established in vivo at orthotopic site in presence of fibroblasts is more poorly differentiated and invasive in comparison to standard xenograft.

Next-Generation PDX Modeling

These new approaches have paved the way for the development of next-generation PDX models, which offer greater insights and predictive accuracy. Fully profiled PDX models can be used to mimic human clinical trials in mice known as patient avatar trials or human surrogate trials. These models provide a clinically relevant environment to interrogate new drugs as well as identify suitable biomarkers of response, which can be applied to the design of Phase II clinical trials to increase the success.

Human surrogate trials or patient avatar trials are a game changer for oncology drug discovery programs and the study of human disease. The most significant cost driver for drug development is the high failure rate in late-stage clinical development. The need to reduce drug attrition is especially acute in the field of oncology, where drugs often fail not because of toxicity but rather lack of efficacy. As such, one of the main goals of translational research for oncology is to identify the likely responsive patient populations, particularly via discovering biomarkers which predict clinical efficacy. The successful development of drugs (like trastuzumab, imatinib, and gefitinib) has demonstrated the critical need to identify biomarkers in order to select patients who are most likely to benefit from the drug treatment, thus tailoring treatment towards a personalized approach.

Models of Acquired Resistance

Development of resistance to radiotherapy, chemotherapy, and targeted agents is a critical area in oncology patient care. Resistant models are needed to explore the additional mechanisms of resistance and for testing new agents and/or combination strategies to delay/combat the emergence of resistance. These models can also be used to develop resistant isolates to new targeted agents and analyze the mechanisms of resistance involved, identify relevant biomarkers and rationally-designed combination studies to delay/overcome the resistance issues, and generate evidence-based hypothesis testing suitable for clinical trials.

Relevant PDX and CDX models of acquired resistance present more important opportunities in identifying key targeted combination studies and identifying new intellectual property around existing patents.

Conclusion

Preclinical models that closely replicate the human tumor microenvironment and heterogeneity are imperative in order to guide clinical strategies and patient selection in the quest to optimize novel cancer therapeutics. The tumor microenvironment is important in resistance and disease progression. It is critical to understand the mechanism of resistance in a given patient’s tumor in order to identify the most appropriate follow-on therapy as part of a personalized medicine strategy. These models not only provide a greater understanding into the tumor microenvironment and improving the efficiency of oncology drug discovery, but also offer the opportunity for new target identification and validation.

By leveraging genomically characterized PDX assets to discovery biomarkers and identify patient responder and nonresponder profiles, it is possible to accelerate drug development and improve chances of success in the clinic with a new candidate. Conducting a clinical trial style study (“patient avatar trial”) offers a potentially powerful tool to predict clinical response and enables biomarker discovery. Such findings can then be validated in the clinical setting and then used for patient stratification in trials and in the treatment practice. Our capacity to perform large-scale studies screening large numbers of PDX models in one study means it is possible to perform a trial in months, not years. Human surrogate trials have the potential to revolutionize the drug development process in the field of personalized oncology.

For more on the tumor microenvironment, be sure to check out “Corruption of the Tumor Microenvironment” from our June 1 issue.

Rajendra Kumari, Ph.D. ([email protected]), is CSO at Precos, a Crown Bioscience company.

References:
1 J I Johnson, S Decker, D Zaharevitz, L V Rubinstein, J M Venditti, S Schepartz, S Kalyandrug, M Christian, S Arbuck, M Hollingshead, and E A Sausville.  Relationships between drug activity in NCI preclinical in vitro and in vivo models and early clinical trials. Br J Cancer. 2001 May; 84(10): 1424–1431.
2 Fricker, J. Time for reform in the drug-development process. The Lancet Oncology Volume 9, Issue 12, December 2008, Pages 1125–1126.
3 BC Giovanella, JS Stehlin, ME Wall, MC Wani, AW Nicholas, LF Liu, R Silber, M Potmesil. DNA topoisomerase I-targeted chemotherapy of human colon cancer in xenografts. Science 1989, Vol. 246 no. 4933 pp. 1046-1048.
4 Feibig et al, EJC 40;802, 2004

Previous articleTop 14 Molecular Diagnostics Firms
Next articleSISCAPA and Automation for High-Throughput Protein Bioanalysis