Throughout their decades of use in medical research, there exist numerous examples where mice have failed to adequately parallel human biology, leading to considerable wastes of money, time and effort. Nevertheless, questioning the adequacy of mice as human surrogates remains controversial, in large part due to irrefutable examples where rodents have helped to progress our understanding of biology and disease and have aided in the discovery of novel therapeutics.
Mice are not humans. This seemingly obvious statement often serves as a blanket explanation for discrepancies observed between data from humans and mice. It critiques mice as imperfect tools due to their inherent dissimilarity to humans. But what if mice could be more like humans?
In efforts to improve upon the human relevance of mouse-based studies, investigators are increasingly turning to models with genetically introduced human components. These humanized models range from mice expressing single human proteins to mice engrafted with human immune systems and/or human tissues. By expanding the human relevance of the rodent model system, mice with human components are helping to progress scientific discovery.
Human genes extend the utility of mouse models
Driving the expression of human genes in rodents is not a novel approach. Indeed, expressing a human gene or gene variant in rodents is one classical means to demonstrate a link between a particular gene and a phenotype. However, it is common in science to see modern applications of a classical approach, especially in response to advances in a particular field.
The rise of biopharmaceuticals over the last decades provides one example of an advancement that has impacted the rationale for expressing human genes in rodents. As opposed to expressing a human gene in a mouse to confer or reveal a phenotype, the expression of a putative human therapeutic target can aid the downstream development of biopharmaceuticals.
Monoclonal antibodies are a class of biopharmaceuticals where, during development, it is often ideal to have a mouse model that expresses the human target protein. In models that manifest a human disease phenotype and express a putative human target, monoclonal antibodies can be tested for their ability to reverse the disease phenotype through a highly specific interaction with the target human protein.
One example illustrating the benefits of expressing a human target protein is found in the TNFα mouse, a transgenic mouse that over-expresses human TNFα and spontaneously develops arthritis. The TNFα mouse was instrumental to illustrating a central role for TNFα in arthritis progression and to the development of infliximab, the first monoclonal antibody that targeted TNFα for the treatment of arthritis.
The human TNFα transgenic mouse’s contribution to the development of a monoclonal therapy began over two decades ago, yet this mouse is still highly relevant today. Models like the TNFα mouse have utility upstream and downstream in the drug development pipeline. Since its first described use, the TNFα mouse has been improved to better model disease progression and is still used in current arthritis drug development investigations.
The tg2576/APPSWE mouse model of familial Alzheimer’s disease has a developing story that is a modern parallel to that of the TNFα transgenic mouse. This model manifests Alzheimer β-amyloid (Aβ) brain plaques due to expression of a transgene coding for a human Aβ precursor protein carrying a Swedish familial mutation.
Like the expression of human TNFα in the TNFα transgenic mouse, expression of human Aβ in tg2576/APPSWE mice has helped to extend the model’s utility. This model was created two decades ago, yet its expression of human Aβ confers to it a target protein that has been exploited as recently as this year for the development of aducanumab, a Phase 3 monoclonal antibody that targets Aβ in Alzheimer’s patients. Thus, the TNFα transgenic mouse and tg2576/APPSWE mice are classical and modern examples of how expression of a putative human therapeutic target can aid the downstream development of biopharmaceuticals.
Human genes are improving mice for studying drug metabolism and disposition
The benefits of expressing human proteins also extend to the development of small molecule-based therapeutics. For these class of therapies, understanding how a drug is absorbed, distributed, metabolized and excreted is particularly important. Interspecies differences in these properties are generally caused by concomitant differences in the underlying composition and expression of enzymes, transporters and xenobiotic receptors.
Humanized mouse models expressing proteins that influence drug metabolism and disposition can be developed through “knock-in” approaches or by crossing a transgenic mouse with a mouse knockout for the particular transgene. These humanized mouse models have already provided mechanistic insights that can improve prediction of xenobiotic risks in humans. However, the ultimate goal is to overcome species differences that impede accurate prediction of drug disposition and metabolism in humans. To this regard, humanized models will serve as platforms for future progress.
Humanized immune systems for immuno-oncology and beyond
For decades, immunodeficient mice have enabled the engraftment of human tumors for the study of cancer biology. Now, on the heels of recent progress in the field of immuno-oncology, investigators are turning to models that can readily accept foreign cancer cells and also recapitulate aspects of the human immune system.
The NOD-SCID mouse’s role as the preeminent immunodeficient model for human tissue engraftment is rapidly being usurped by its modern successors. In particular, NOD-SCID models that also carry partial or complete deletions of the IL-2 receptor gamma (IL2rγ) chain (e.g., NOG and NSG mice) are increasingly preferred. Contributing to their rapid adoption is the elimination of thymic lymphomas, which manifest in about 70 percent of NOD-SCID mice and are prevented by the loss of functional IL2rγ in NOG/NSG mice.
NOG and NSG mice readily accept human hematopoietic stem cells, which differentiate in vivo to reconstitute much of the human lymphoid immune system. These humanized immune system mice also accept cancer xenografts, allowing investigators to test drugs with the potential to harness the immune system to combat a tumor. Indeed, NOG mice engrafted with a human immune system and a tumor have demonstrated appropriate responses to checkpoint inhibitors, including the target-specific expansion of human T cells.
One caveat to human immune system engraftment in NOG and NSG mice is they develop an incomplete human immune system that lacks much of the myeloid component. Myeloid cells are believed to play important roles in cancer progression. Furthermore, checkpoint inhibition mechanisms extend to myeloid lineage cells. Thus, humanized immune system models that reconstitute both lymphoid and myeloid cells would be beneficial for cancer therapy development, including combinatorial checkpoint inhibitor studies.
Next-generation NOG and NSG models (huNOG-EXL and NSG-SGM3, respectively) for human immune system engraftment have been engineered that can reconstitute the myeloid lineage. These models introduce transgenes expressing myeloid-differentiating human cytokines onto NOG and NSG backgrounds. The human myeloid cells that differentiate are capable of homing to engrafted tumors and display additional functional aspects, such as eliciting a human allergy response.
There remain additional limitations in human immune system mice that investigators are working toward overcoming. Typically, overcoming these limitations involves additional layers of humanization. For example, co-implantation of human fetal liver and thymus with hematopoietic stem cells can improve T cells by allowing their maturation in human thymic tissue. Another example comes from the MISTRG mouse, a unique model that reconstitutes human myeloid and lymphoid lineages through the expression of four human cytokine genes knocked into their respective mouse loci.
It’s important to point out that no mouse model perfectly reproduces the human immune system. Indeed, models that improve certain aspects of humanization tend to reveal new limitations. Co-implantation of human fetal tissues in humanized immune system mouse models results in the eventual onset of graft vs. host disease, and the MISTRG mouse suffers from fatal anemia several weeks after engraftment. Nevertheless, these models illustrate how humanization is an approach that increases the translational relevance of mouse models.
The future of mouse models: conclusions
We should continue to expect improvements in mouse models mediated by humanization of proteins, cells and tissues. These improvements should help to create versatile models with long-term utility, and should diminish the discrepancies observed between experiments in mice and humans.
However, mouse models are research tools that will always have limitations. It’s important for a model to be carefully chosen based upon a defined set of research goals. The burden is on the investigators to understand the limitations of a given model within their experimental aims. Humanization is certainly improving our mouse models, but we cannot expect humanized mice to solve all of the past issues we’ve observed with rodents. After all, mice are not humans.
Michael Seiler, Ph.D., is portfolio director of genetically engineered and humanized immune system models at Taconic Biosciences. He earned his doctoral degree in cell and molecular biology at the Baylor College of Medicine and completed postdoctoral training at the University of Chicago. Seiler published 15 peer-reviewed manuscripts while in academia, then when on to serve as vice president of a startup firm focused on commercializing gene editing technology before joining Taconic in 2014.