Is Human Cell Therapy Research Caught in a Mousetrap?

Origins

About 65 million years ago, an asteroid the size of a city hit the Yucatan Peninsula, leading to the extinction of the dinosaurs. The impact fortuitously brought evolutionary opportunities for the last common ancestor of humans and mice to set out on the great radiation of mammals that today humans seem to be doing their best to deplete. Meanwhile, the mouse has been widely adopted as a model animal for understanding mammalian biology. Mice are at the forefront of animal research for many good reasons, not least their size, cost, life span, and availability in a myriad of well-defined strains. Furthermore, they come with an encyclopedic blueprint of murine biology down to constituent molecules. The description of the mouse genome was completed only a year after that of the human genome,¹ and mouse studies occupy the pages of the most auspicious scientific journals. So much intellectual property is invested in mouse biology that it has become an important discipline. At the same time, mice have been used as models of human disease and, in the same way as the archetypal guinea pig, as preclinical treatment models.

Mouse research in allogeneic stem cell transplantation

Here we confine ourselves to the role of the laboratory mouse as a model for human diseases and their treatment with cell therapy and allogeneic stem cell transplantation (SCT) in particular. It is not possible to provide an exhaustive list of the contributions made by murine studies to this field, but several areas of research and discovery deserve highlighting. First, mouse studies have underpinned the basis of bone marrow SCT. Historically, the cellular basis of SCT, the first description of graft-versus-host disease (GVHD) and graft-versus-leukemia (GVL) effects were discovered in the mouse.² Today the many well-defined mouse strains available permit the design of experiments in which precise immunological differences between donors and recipients create controlled experimental settings for studying GVHD, rejection, and antitumor effects. Models of diseases susceptible to cell transplantation and gene therapy (e.g., sickle cell disease, muscular dystrophy, and neurodegenerative storage diseases³) have served as important validations for similar treatment approaches in humans. In recent years, the development of gene knockout mice has enabled us to pinpoint molecular mechanisms underlying alloimmune reactivity.⁴ The ability to track cell migration via fluorescent tagging in mice has been widely applied in studies of the role of lymphocyte stem cells and stromal cells in tumor control.⁵ In summary, the laboratory mouse has been the source of many basic immunological principles that are central to the way in which SCT and cell and gene therapy have developed for humans.

Mice as preclinical models

When it comes to modeling human disease and treatment, the very strengths that make the mouse an extraordinary tool for understanding fundamental biological truths can become a hindrance when mice are used as a translational platform for clinical studies. Here the differences (e.g., in body size, longevity, and pharmacokinetics), rather than the similarities, between our two species become an issue. Furthermore, the laboratory mouse is distinct from its wild counterpart—and us—in being genetically uniform, selected for efficient reliable breeding, and brought up in environments that are almost as sterile as the bubbles that we used to use to protect immune-deficient children.

Genetic purity is essential for obtaining clear-cut results in microbiologically “clean” animal stem cell transplant experiments, but this does not mirror human SCT between outbred individuals, who have such individual packages of major and minor histocompatibility antigens and cytokine gene polymorphisms that the outcome of transplant—even between donors and recipients from the same family—is a genetic lottery. Factor in the powerful immune impact of the infectious agents that we are immersed in, and comparisons with the immune responses in the clean laboratory mouse become less sustainable. It is more difficult to generate GVHD in clean laboratory mice. With respect to the specifics of immunotherapy and SCT, there are other obstacles to making reliable comparisons between human and mouse. The experimental tumors and cell lines we typically use as targets for alloimmune responses have no close relationship to the human malignant counterpart, raising the question of whether the GVL effects described are truly comparable with the human.⁶

At the molecular level, murine and human cells differ, leading to differences in immune function.⁷ Killer immunoglobulin-like receptor molecules on murine natural killer cells are sugars whereas humans have proteins, limiting specific comparisons of natural killer function in the two species. Regulatory T cells in mice and humans have important differences; they are more readily isolated in mice than they are in humans,⁸ and murine mesenchymal stromal cells do not always behave like their human counterparts.⁹ When dealing with defined antigens of human infectious agents and tumors that are linked to specific major histocompatibility molecules, the divergences become too great to use the mouse for the development of human vaccines outside the constraints of the HLA-A2 transgenic humanized immune-deficient mouse.¹⁰

These considerations, however, have not prevented the successful prediction of some murine models for clinical strategies. Early on, gastrointestinal decontamination was found to reduce severe bacterial and fungal infection and to control GVHD.^11,12,13 The potential of gastrointestinal decontamination to reduce GVHD was also demonstrable in humans, especially when targeting oral microflora¹⁴ and anaerobes.¹⁵ However, it has proved difficult in practice to reliably render humans sufficiently germ-free. A recent example of successful translation of mouse to human is the creation of the sickle cell mouse and the development in this model of a durable, low-intensity stable mixed chimeric transplant, a strategy that has led directly to successful clinical studies.^16,17

Lost in translation

The problems arise when murine studies are not congruous with human clinical models. Failure of translation from mouse to human falls into several categories. First, results of murine models may dissuade investigators from moving into the clinic. A recent example of such divergent findings is the high incidence of lethal GVHD in mice receiving a lymphodepleting regimen followed by syngeneic cells transduced with genes encoding T-cell receptors (TCRs). Rosenberg and colleagues treated more than 100 patients with syngeneic T cells transduced with TCR genes using several retroviral vectors without any evidence of GVHD.¹⁸ Second, while results of such murine studies might have precluded regulatory approval for clinical trials, other studies generate false optimism, as was the case with interleukin-11 (IL-11). Murine experiments indicated that IL-11 reduced transplant-related mortality, preventing GVHD while conserving GVL.¹⁹ By contrast, a subsequent phase I/II trial of IL-11 in allogeneic human SCT in which recombinant human IL-1 was added to GVHD prophylaxis was terminated early because of multiorgan failure and severe fluid retention,²⁰ indicating that IL-11 in this setting did not, as anticipated from the murine studies, reduce transplant-related mortality. Another disappointment was the potential application of antibody to CD154 as a single-infusion antirejection agent in renal cell transplantation. Following animal models indicating efficacy and safety, the attempt to use antibody to CD154 in the clinic as a means of protection against organ graft rejection resulted in dangerous side effects from hypercoagulability because, as it emerged, CD154 is present on human platelets.²¹

A third problem that confounds the direct translation of murine studies to the human is the strain specificity of some observations, especially with cytokines, that render uncertain the prediction of the outcome in a human trial. One example is the complicated function of IL-2. Initial studies indicated that IL-2 impaired acquired tolerance²² and accelerated GVHD.²³ Subsequently, Sykes et al. first described a protective effect of IL-2 in the prevention and treatment of GVHD.^24,25,26,27 In later studies, however, they concluded that IL-2 had varied effects depending on the mouse strain used²⁸ and concluded that the specific degree of histoincompatibility between donor and host is determined by the type of CD4 activity mediating GVHD. In humans, IL-2 has been used both to amplify the effector functions of T cells²⁹ and to prevent GVHD.³⁰ If IL-2 is to be used to advantage in human SCT, clinical trials in specific transplant situations will be needed to identify appropriate doses and schedules. Finally, there are situations in which the mouse cannot work as a preclinical model; for example, mice do not develop a disease comparable to human chronic GVHD.⁶

Solutions

If we are to optimize the rate of advance in cell and gene therapy diseases, we must bridge the gap between the world of mouse transplant immunology and translational research in humans. In the area of cell therapy we are at a stage at which experiments related to fundamental immunological processes requiring carefully selected knockout mice, although still important to our understanding of basic immunology, have not kept pace with technological developments. Mesenchymal stromal cells, regulatory T cells, induced pluripotent stem cells, and gene-transduced immune cells, to name a few, are part of a new armamentarium of potent cells to treat cancers, infections, and autoimmune disease. We need tools to study their distribution, homing, survival, and function in humans. This means that priority should be given to designing mouse models that are as realistic as possible to the human condition and, if they are not, that the deficiencies are recognized and alternative approaches sought. The use of outbred large-animal models, nonhuman primates, and humanized immune-deficient mouse models may serve better to re-create human cell–cell interactions and human therapies. The dog has been a good model for some time-honored clinical strategies, such as GVHD prophylaxis,² and nonhuman primates have been used for preclinical safety studies of cells and cytokines. However, we should be prepared for the eventuality that even what appear to be the closest models may fail to predict the human outcome, as occurred when a primate safety model using antibody to CD28 resulted in life-threatening cytokine storms in initial preclinical trials in humans.³¹

A greater challenge, but ultimately a more direct route, would be to improve our technologies for human studies. Our abilities to study human immunology are now much more sophisticated than they were a decade ago, making possible cellular events using multicolor flow cytometry to analyze nuances in cell subsets and realistic functional assays to characterize the cells we infuse.

Furthermore, it is possible to co-opt computational methods that can address multiple networks and interactions between cells and cytokines that characterize the arcane in vivo interactions between immune cells and their targets.³² Only by identifying key interactions at the molecular level can we begin to address the diversity of responses between one individual and the next. It is to be hoped that we will reach a degree of sophistication in the technology that allows us to perform many more studies directly in humans without the need to first check out imperfect animal models.

Conclusions

Humans and mice are similar but different. Similarities have allowed us to learn fundamental truths about mammalian biology as it applies to humans, but differences complicate the use of the mouse as a preclinical model for safety testing and detailed design of clinical strategies. Nevertheless, regulatory bodies still insist on preclinical testing in mice despite limited predictive value for the clinic. Such testing causes delays and extra expense and serves more as a fig leaf to protect the authorities against recklessness than as a step with meaningful value.^33,34 If there are any conclusions to be drawn from preclinical models, it should be to expect anything. Sometimes predictions work, but—particularly when interpreting the function of cells, antibodies, and cytokines—no assumptions can be made about the likely effect of similar treatments in humans. In the end, therefore, there seems to be no substitute for an ethically sound phase I clinical trial in humans. Fortunately, we now have powerful tools to study events at the molecular level of interaction in humans. Understanding how human biology functions in cell transplantation and gene therapy is still a work in progress, but these new technologies and multifunctional analyses make it more possible today than ever before. Sadly, there is still evidence that human and mouse research activities exist in separate worlds. Consider, for example, the lack of synchrony between a 2008 review defining a proof of concept for CD8⁺ T-cell therapy for cytomegalovirus disease in the mouse,³⁵ published one and a half decades after the pioneering studies by Greenberg, Riddell, and colleagues defining the therapeutic function of CD8⁺ cytomegalovirus-specific T cells in humans.^36,37

Another example in which the clinical trial preceded the mouse model is the generation and clinical testing of Epstein–Barr virus (EBV)–specific cytotoxic T cells described in humans in 1995 (ref. 38) and in mice in 1996 (refs. 39,40). A recent review on therapy with EBV-specific cytotoxic T lymphocytes states, “It is somehow surprising, for example, that the clinical transfer of anti-EBV adoptive immunotherapy has advanced very rapidly, bypassing a rigorous animal preclinical evaluation.”⁴¹ The implication that bypassing the mouse model represents improper scientific procedure should not be allowed to pass unchallenged. If opportunities arise in which it is ethically acceptable and economically preferable to move directly into human preclinical work and related clinical trials, we should not be constrained by regulatory authorities to use time and resources on experiments that are not central to treatment development. It is to be hoped that investigators continue to recognize the differences between the two species that limit complete reliance on the mouse as a preclinical model and avoid the pitfall of trying to model humans on mice.