The State of Modeling Preclinical Immuno-Oncology Pharmacology

The field of cancer immunology or immuno-oncology (I/O) dates back to the 1890’s with efforts by William Coley to treat malignancies by infection with certain bacterial strains.¹ Fast forward 120 years, and we have observed the clinical success of multiple checkpoint inhibitor antibodies against cancer specific T cell negative regulatory pathways in multiple cancer types.² Because of these clinical successes with antibodies targeting CTLA-4, PD-1, and PD-L1 (e.g., ipilimumab, nivolumab, pembrolizumab, atezolizumab, avelumab), vast resources in biotech and large pharma are being directed toward preclinical I/O research in an effort to develop novel immunotherapies.

A persistent challenge for oncologists and immunologists attempting to develop immunotherapies is the requirement for faithful preclinical models in which to test them. By definition, these models require an intact immune system; so traditional human cell line xenografts that drive much of preclinical oncology research are not suitable. What options does that leave for preclinical immuno-oncology research? In Figure 1, the most commonly used models for testing novel I/O agents are described along with some of their positive traits and drawbacks.

Fig. 1: Modeling Immuno-Oncology in the Mouse

Syngeneic Mouse Models

By far the most utilized models for preclinical I/O pharmacology are the syngeneic mouse tumor cell lines. These models were developed from either spontaneously arising tumors in older mice or from carcinogen induction. These lines can be implanted into their genetically syngeneic hosts to simulate the ability to generate an immune response against the tumor via its tumor antigens. While this doesn’t precisely replicate the dynamics of an immune system evolving with a developing malignancy, it does allow researchers to test novel immunotherapies that activate tumor-specific T cells and inhibit immuno-suppressive pathways that prevent effective anti-tumor immunity. These models are easily accessible for quick and efficient studies. Further, there is a large repository of published literature using these models for testing novel immunotherapies to turn to for comparison of one’s novel agent, effectively making them the “industry standard.”

While syngeneic mouse models are ideal for many I/O drug development programs, they do come with some significant drawbacks. One of the primary issues is that biological therapeutics must be cross-reactive to the mouse ortholog, if it even exists, or a mouse-reactive surrogate must be developed. This is often a big hurdle as it can significantly increase the cost and time associated with the development of a candidate therapeutic. A second drawback is that they often have a neo-antigen load, the result of non-synonymous mutations, that is significantly higher than that found in most human cancers.^3,4 One of the greatest concerns about the use of syngeneic mouse tumor lines for immunotherapy is their perceived lack of predictive power for clinical translation.⁵ An evaluation of the many cases where these models predict and fail to predict clinical response could fill multiple blogs; however, drug developers have to remain cognizant of the differences between the mouse immune system and the human immune system when modeling the biology of their target.

GEM Models

The genetically engineered mouse (GEM) cancer models represent an alternative approach to modeling the pharmacology of novel immunotherapies. These models rely on introduction of clinically relevant driver oncogenes and tumor suppressor mutations/losses to induce multi-stage carcinogenesis in the tissue type of interest. This allows targeted therapies against specific oncogenic drivers (e.g., Braf) to be tested in combination with immunotherapies. Since the tumors develop de novo in concert with an intact immune system, they appear to better reproduce the cellular dynamics between the immune system and cancer cells in humans.

While GEM models have many advantages, they have some notable issues as well. The stochastic nature of de novo carcinogenesis in autochthonous GEM models makes experiments very long and challenging to stage without a rolling enrollment approach. This results in tremendous cost in maintaining mouse colonies and monitoring long studies. Use of GEM-derived transplantable fragments provides a more tractable approach to I/O pharmacology because studies can be implanted and staged as with mouse cell line models. However, GEM-fragments no longer simulate an “experienced” immune system because they are implanted into naïve syngeneic hosts. Finally, GEM models have a characteristically low mutational burden due to the use of strong oncogenic driver(s) and tumor suppressor losses to enable carcinogenesis. Thus, additional mutations are not required for tumor development.⁴ This is a major barrier to eliciting tumor antigen specific T cell responses because there are very few neo-antigens available for the immune system to recognize.

HIS Models

Humanized immune system (HIS) mice are another in vivo approach to model I/O pharmacology. There are a considerable number of models that fall under the classification of HIS mice; however, most of these are reconstituted in the non-obese diabetic (NOD)-scid IL-2Rgamma^null (NSG or NOG) mouse backgrounds. The two most widely utilized models are the human peripheral blood mononuclear cell (PBMC) reconstituted mice (huPBMC-NSG/NOG) and the CD34⁺ hematopoietic stem cell reconstituted mice (huCD34-NSG/NOG). HIS mice allow investigators to test antibodies against the human target without requiring rodent cross-reactivity or surrogate antibodies. Further, one can combine a novel agent with clinically approved checkpoint inhibitor antibodies since the immune system is humanized. Implantation of patient derived xenograft (PDX) models into huCD34-NSG mice enables testing of immunotherapies against tumors that retain much of the human cancer microenvironment.

Although HIS mice are useful when either a mouse-reactive test agent or a mouse ortholog for a target don’t exist, they do not have a fully functional human (or mouse) immune system. The huPBMC-NSG mice only engraft with CD4⁺ and CD8⁺ T cells.⁶ Further, those T cells will cause a fatal graft versus host disease (GVHD) in the host mice between 4-5 weeks post-transplant. The human T cells that mature in the huCD34-NOG mice have several functional deficiencies including accelerated activation induced cell death, attenuated proliferative capacity, reduced interleukin-2 (IL-2) production and impaired CD8⁺ T cell persistence.^7,8 The human NK cells that arise in huCD34-NSG mice are also functionally inert.⁸ Lastly, when used with allogeneic human PDX tumors or human cell lines, these models are more accurately allograft rejection models than faithful autologous anti-tumor immunity models. Efforts to achieve more complete human immunity using human cytokine transgenic mice and transplantation of additional human tissue are ongoing.⁹ Nonetheless, these mice provide an opportunity to drug the actual human target, as long as investigators are aware of their limitations.

Conclusion

While no model will ever perfectly replicate the biology of the human immune system, the cancerous cells or the response to a candidate therapeutic, being aware of the advantages and disadvantages of the various model options allows I/O pharmacologists to make the choice that best informs their research. Contact us today to discuss the optimal model for your therapeutic candidate.

¹Coley WB. The treatment of malignant tumors by repeated inoculations of Erysipelas, with a report of ten original cases. Am. J. Med. Sci. 1893; Vol. 105:487–511.

²Blumenthal GM and Pazdur R. Approvals in 2016: the march of the checkpoint inhibitors. Nat. Rev. Clin. Oncol. 2017; Vol 14(3):131-132.

³Chalmers ZR et al. Analysis of 100,000 human cancer genomes reveals the landscape of tumor mutational burden. Genome Med. 2017; Vol 9(1):34.

⁴McCreery MQ and Balmain A. Chemical Carcinogenesis Models of Cancer:  Back to the Future. Annu. Rev. Cancer Biol. 2017; Vol 1:295–312.

⁵Lechner MG, Immunogenicity of murine solid tumor models as a defining feature of in vivo behavior and response to immunotherapy. J. Immunother. 2013; Vol 36(9):477–489.

⁶van Rijn RS et al. A new xenograft model for graft-versus-host disease by intravenous transfer of human peripheral blood mononuclear cells in RAG2−/− gammac−/− double-mutant mice. Blood.  2003; Vol 102:2522–2531.

⁷Watanabe Y et al. The analysis of the functions of human B and T cells in humanized NOD/shi-scid/gcnull (NOG) mice (hu-HSC NOG mice). International Immunol. 2009; Vol. 21(7):843–858

⁸Andre MC et al. Long-Term Human CD34+ Stem Cell-Engrafted Nonobese Diabetic/ SCID/IL-2Rgnull Mice Show Impaired CD8+ T Cell Maintenance and a Functional Arrest of Immature NK Cells. J. Immunol. 2010; Vol 185:2710–2720.

⁹Rämer PC et al. Mice with human immune system components as in vivo models for infections with human pathogens. Immunol Cell Biol. 2011; Vol 89(3):408-16.