Mouse models can predict cancer therapy

In this issue of Blood, Chesi et al show that their genetically engineered mouse model of multiple myeloma can predict positive or negative activity of drugs previously tested in clinical trials.¹ 

In the past decade a technological revolution has led to a remarkable increase in the molecular characterization of human tumors.²  Initially, gene expression profiling provided for a molecular classification of tumors that often were not distinguishable by standard pathologic procedures, and sometimes identified expression signatures that predicted prognosis or response to standard therapies. More recently, second-generation sequencing has spawned procedures that have enabled a nearly complete description of the genome and transcriptome for many kinds of tumors.

At the same time, investigators have developed high-throughput strategies to identify tumor-specific therapeutic targets. There also is continued interest in developing immunomodulatory therapies, as well as therapeutic strategies that target the critical cooperative interactions of the tumor cell and its microenvironment. It seems increasingly apparent that combinations of 2 or more therapies will be necessary, particularly in view of intratumor genetic heterogeneity and the potential for rapid generation of genetic variants.

Despite these major advances in our understanding of tumor biology and the ongoing identification of new therapeutic strategies, the cost and time required for validation in clinical trials is prohibitive. There is a pressing need to develop better preclinical tumor models—or to better use existing models—that can enable the design of more efficient and productive clinical trials.^2,3 

The use of genetically engineered mouse models (GEMMs) of cancer is an approach that has shown significant promise. Singh et al used KRAS-driven GEMMs for non-small cell lung cancer (NSCLC) and pancreatic ductal adenocarcinoma, both of which model the multistage tumor progression from early to advanced disease.⁴  Importantly, the tumor evolves in an immunocompetent animal and apparently in the same microenvironment as the corresponding human tumors. They assessed responses to existing standard-of-care chemotherapeutics alone and together with other inhibitors. Using standard clinical end points (overall survival and progression-free survival) and noninvasive imaging, they concluded that both GEMMs are able to predict outcomes in human clinical trials while providing insights into mechanisms of therapeutic response and resistance. More recently, another group of investigators conducted a “co-clinical” trial with standard-of-care chemotherapy plus a MEK inhibitor in a KRAS-driven GEMM of NSCLC that mimicked an ongoing human clinical trial in patients with KRAS-mutant lung cancers.⁵  They concluded that their co-clinical trials “not only anticipate the results of ongoing human clinical trials, but also generate clinically relevant hypotheses that can inform the analysis and design of human trials.”⁵ 

Some of the preclinical models that have been used for multiple myeloma (MM) are summarized in Table 1. Xenogenic transplants into different strains of immunodeficient mice are useful for human myeloma cell lines (HMCLs) and some extramedullary MM tumors but rarely for intramedullary MM tumors.^6,7  A better albeit imperfect mimic of the tumor microenvironment is provided by “humanized” immunodeficient mice, perhaps accounting for the successful transplantion of intramedullary MM tumors in this model.⁸  The recent description of the SCID-syn-hu model appears to open the possibility of manipulating the microenvironment by co-injecting different combinations of human bone marrow cells with the synthetic bone matrix before injecting human tumor cells.⁹  A syngeneic transplant model uses 2 different spontaneous C57/BL6 MM tumors that home to the bone marrow, and sometimes extramedullary sites, in immunocompetent mice.¹⁰ 

Chesi et al developed their Vk*MYC GEMM of MM in the C57/BL6 mouse strain that spontaneously develops a premalignant monoclonal gammopathy of uncertain significance (MGUS) tumor, which is similar to the MGUS tumor that universally precedes MM in humans. Their immunocompetent C57/BL6 Vk*MYC model sporadically activates the MYC transgene in germinal center B cells, and uniformly develops MM in aged mice, a result that models the increased MYC expression in human MM compared with human MGUS. The slowly expanding MM tumors in this model appear to faithfully reflect the biology of human MM, including bone marrow localization and presumptive interaction with the microenvironment, low fraction of proliferating tumor cells, and secondary abnormalities such as bone disease and anemia. When these MM tumors are transplanted into normal C57/BL6 mice, the MM tumor becomes more proliferative and extends to extramedullary sites, and thus is a model for relapsed refractory human MM. Serum protein electrophoresis is used to detect the amount of tumor-specific monoclonal immunoglobulin. Therefore, it is possible to monitor the amount of tumor during the course of disease even when the tumor is dispersed at different intra- and extramedullary locations.

Using the Vk*MYC GEMM of MM, Chesi et al have determined the effects of single drugs with both known and unknown clinical activity on MM tumors that have arisen spontaneously. Significantly, they find that 4 of 6 clinically effective drugs (67%) are active in this model. Equally important is that 7 of 8 clinically ineffective drugs (88%) are inactive in this model, even though many of these ineffective drugs have activity against HMCLs. In addition, they provide evidence that the more aggressive disease that models end-stage drug-resistant MM responds only to combinations of drugs with single-agent activity against de novo Vk*MYC MM.

Similar to the other GEMM models described above, the Vk*MYC GEMM model of MM is extremely promising for identifying both single agents and combinations of existing and new therapeutic agents that are more likely to be effective in clinical trials designed to investigate different disease stages. One caveat is that the MM tumors in this model may not fully model some kinds of MM tumors (eg, tumors with t(4;14) translocations). However, unlike most of the other preclinical models used for MM (Table 1), these tumors appear to be localized in the appropriate microenvironments in immunocompetent hosts. Therefore, the Vk*MYC GEMM model also appears to be suitable for identifying therapies that target the interaction of tumor cells with the microenvironment, and immunomodulatory therapies.