|Home | About | Journals | Submit | Contact Us | Français|
There is potential for a variety of stem cell populations to mediate repair in the diseased or injured CNS; in some cases, this theoretical possibility has already transitioned to clinical safety testing. However, careful consideration of preclinical animal models is essential to provide an appropriate assessment of stem cell safety and efficacy, as well as the basic biological mechanisms of stem cell action. This article examines the lessons learned from early tissue, organ and hematopoietic grafting, the early assumptions of the stem cell and CNS fields with regard to immunoprivilege, and the history of success in stem cell transplantation into the CNS. Finally, we discuss strategies in the selection of animal models to maximize the predictive validity of preclinical safety and efficacy studies.
The loss of cells of the CNS is a hallmark of neurological disorders and traumatic neural injury, such as Parkinson’s disease, Huntington’s disease, Alzheimer’s disease, multiple sclerosis, amyotropic lateral sclerosis, spinal cord injury (SCI), traumatic brain injury and stroke. Currently, there are very few therapeutic interventions to ameliorate these conditions in the human patient population. However, stem cell therapy has potential to provide methods of repairing, regenerating or protecting the damaged CNS. Successful efficacy using human stem cells in animal models has already led to multiple human clinical trials [1–3], which have been sponsored by several sources (Geron, StemCells, Inc., and Advanced Cell Technology).
A stem cell is a cell that possesses the ability to both self-renew and differentiate into multiple cell types. Embryonic stem cells are pluripotent, capable of differentiating into all cell types of the body . Recent advances in the stem cell field have resulted in the development of induced pluripotent stem cells reprogrammed from somatic cells, which have been shown to have a remarkably similar fate potential to embryonic stem cells . Proceeding down the lineage tree of development, later stem cell populations become more restricted to the tissue from which the cells were derived. For example, neural stem cells of the CNS have a restricted fate potential and are capable of only differentiating into neurons and glia . For the purposes of our review of the neurological disease and injury literature (Tables 1–11), we will focus on human embryonic, fetal and adult-derived cells, in particular highlighting the issues associated with neurotransplantation of these populations.
Transplantation of stem cells into the diseased or injured CNS allows a unique replacement therapy not afforded by pharmacological therapeutics. Stem cells can provide benefit by differentiating and integrating into the host to restore functional and behavioral deficits that result from the loss of host CNS cells [2,7,8]. However, stem cells can also provide trophic support or deficient factors to the host tissue, reducing cell loss or potentially promoting host regeneration/plasticity mechanisms to restore function [9,10]. In some cases, the benefits of stem cell transplantation may derive from the short-term neurotrophic/neuroprotective effects during the acute phase postinjury/transplantation. However, the risk: benefit ratio of a cellular therapeutic that is neurotrophic/neuroprotective in nature in a human clinical population may not be advantageous due to increased risk factors to the patient deriving from this method of delivery. These may include tumorigenesis and graft rejection. This is especially true when alternatives that offer similar mechanistic recovery are available, such as conditioned media, neutralizing antibodies, or neuroprotective pharmacological approaches. Thus, we focus here on mechanisms of action in which long-term stem cell engraftment and survival are of critical importance to the regenerative medicine field.
The goal of this review is to broadly look at the role of the choice of animal models in testing proof of concept and safety for the clinical translation of stem cell transplantation strategies, with a particular focus on the CNS. We discuss the available animal model systems for stem cell transplantation into the diseased or injured CNS, identify and discuss key criteria for engraftment and cell survival that should be considered in the experimental setting in the evaluation of efficacy and/or safety studies, and review a brief history of the animal models selected by different arms of the regenerative medicine field. Finally, we review the existing CNS stem cell transplantation literature in the context of cell engraftment and survival criteria, summarizing findings in the field of regenerative medicine and addressing the critical need for more refined criteria of success in animal models utilizing xenotransplants.
The success of laboratory and clinical transplantation is largely defined by the immunological compatibility between donor and host tissue/cells. Autografts are transplants where the donor tissue comes from the recipient. A common example of an autograft is when skin or bone is taken from one part of a patient’s body to reconstruct another . Syngeneic transplants, or isografts, are those in which the donor and recipient are either genetically identical or sufficiently identical to allow for complete immunological compatibility. An example of syngeneic graft is a kidney transplant between identical twins . Allografts are transplants where the donor and recipient are nongenetically identical members of the same species. Differences in MHC antigens, specifically HLA in humans, determine the successful integration of the graft. In the clinical setting this is a complicated issue, as HLA/androgen-binding proteins match stringency can be dependent on the organ/tissue transplanted. However, it is known from organ transplant research that MHC matches, or near matches, reduce the likelihood of graft rejection . Regardless of the level of histocompatibility, immunosuppression is required to overcome the immune response against alloantigens. Finally, a xenograft is a transplant in which the donor and recipient are from different species. Transplantation between closely related species, such as mice and rats, is classified as a concordant xenograft. Transplantation between distantly related species, such as humans and rodents, is classified as a discordant xenograft. The level of immune response, and hence the risk of an immunorejection response, increases in magnitude when moving from an autologous transplant, to a syngeneic transplant, to a matched allograft, to a near-matched allograft, to an unmatched allograft, to a concordant xenograft and finally to a discordant xenograft (Figure 1). Therefore, the type of transplantation paradigm is a crucial factor in establishing a clinically relevant model system for stem cell therapy.
Informative model systems that have good predictive value of the human clinical transplantation setting are necessary to increase the chances of success in translating advances in stem cell therapy of neurological diseases from bench to bedside. From a stem cell transplantation perspective, there are two major components of a model system: the host species in which the neurological disease is being studied and the host species from which the stem cells are derived. Clearly, from both an ethical and regulatory standpoint, human subjects are not an appropriate starting point for testing stem cell therapies. However, there are a wealth of animal models that closely mimic some of the pathological and behavioral deficits associated with different neurological disorders and types of neurotrauma. While some groups have performed experiments in larger animal models such as nonhuman primates [14–16], the vast majority of preclinical research using human stem cells is conducted in rodent models due to the lower cost, lack of nonhuman primate models of neurological disease and injury (versus the abundance of transgenic/knockout mice), and difficulty of achieving adequate immunosuppressive regimens in nonhuman primates. In fact, the latter may be such a significant barrier that, in many cases, stem cell transplantation into nonhuman primate models cannot practically be employed to inform clinical translation. For the purposes of this article, we will focus on rodent models of neurological disease and trauma.
If we accept rodent models as the basis for the study of neurological disease/trauma, what is the most appropriate stem cell source: human stem cells or animal stem cells? A large advantage to using specifically rodent stem cells for preliminary proof-of-concept studies is that researchers can perform syngeneic or allotransplants, which can bypass many of the immunological hurdles of xenografts and more closely mimic some aspects of the clinical setting (i.e., matched human tissue grafted into a human patient). However, while nonhuman cells can provide preliminary proof-of-concept data, the proposed population of human cells should be tested in preclinical studies for regulatory submission in optimal and appropriate animal models. Furthermore, long-term safety data must be established using the clinical grade of stem cells in the target tissue and in potentially both naive and disease/injury conditions. Consequently, there is a need in the regenerative medicine field for preclinical model systems with good predictive value; this need necessitates the use of human stem cells.
In the case of a discordant xenograft, such as human stem cells into a mouse host, the main hurdle is avoiding or sufficiently minimizing the rejection response by the host immune system in order to achieve successful engraftment and survival of the transplanted human stem cells. Moreover, the presence of an active immunorejection response itself may significantly alter the efficacy, and critically, the safety profile of the cell therapy candidate. In this regard, a valid assessment of the safety profile can be argued to be particularly dependent on conditions that enable or encourage maximal theoretical engraftment and cell survival. In the absence of maximal theoretical engraftment conditions, a valid analysis of the tumorigenic potential of a cell therapy candidate may not be possible. An example of this would be the failure to develop teratomas after embryonic stem cell transplantation into immunocompetent versus immunodeficient hosts [17,18]. By employing animal models that recapitulate the key elements of human pathogenesis, permit sensitive evaluation of disease-modifying activity and permit successful engraftment and long-term survival of transplanted human stem cells, researchers can improve the predictive validity of preclinical safety and efficacy studies, and the likelihood of success of translation to clinical trials.
Recognition of the difficulty of achieving significant long-term engraftment and survival after transplantation is not new; by contrast, it is an issue grounded in the history of skin and organ grafting, which provided the foundations for our understanding of immunological tolerance, as well as the trial-and-error history of hematopoietic cell transplantation [19,20]. Although now regarded as a clinically-acceptable therapy for leukemia, the evolution of hematopoietic cell transplantation spans half a century and culminates in the conclusion that addressing the multidimensionality of the immune response is critical in order to achieve successful cell engraftment and survival.
Human hematopoietic cell transplantation originated from a number of preliminary animal experiments that demonstrated the importance of histocompatibility between donor and host. Murine studies have shown that successful engraftment of allogeneic marrow cells could trigger an immune reaction against the host  (reviewed in ), now termed graft-versus-host disease and known to be mediated by T cells derived from donor tissue. Critically, the severity of the immune reaction was regulated by genetic factors . Successive animal experiments revealed the importance of histocompatibility between donor and host tissue [24–29]. These data suggested that T-lymphocyte-mediated immune responses could be triggered as a result of MHC mismatch. Accordingly, subsequent experiments employing methotrexate  and cyclophosphamide , both of which attenuate T-cell responses, achieved more favorable outcomes in animal models of bone marrow transplantation. However, the success rate in translating these advancements for the treatment of human leukemias via hematopoietic cell transplantation remained low . Conversely, hematopoietic cell transplantation in cases where patients exhibited severe combined immunological deficiency [32–34] resulted in high levels of engraftment, and in some cases, patient survival for more than 25 years . With the discovery of a spontaneous mutation in mice leading to a similar form of severe combined immunodeficiency (SCID) , efforts were undertaken to recapitulate the human findings in this new mouse model. Approximately 5 years later, Mosier and colleagues performed the first successful hematopoietic stem cell xenograft using the recently discovered immunodeficient SCID mouse, in which the human hematopoietic stem cell engrafted successfully and constituted long-term repopulation of the mouse immune system . As a result of this and later experiments using immunodeficient models developed to target multiple components of immunorejection (see later), the use of immunodeficient animals in the field of hematopoietic stem cell transplantation permitted increased engraftment, cell survival and rapid advancement of knowledge in this area [38–40].
Based in large part on work in the hematopoietic cell transplantation field, the role of T lymphocytes in immunorejection has become much better understood over the past 50 years. T-cell activation is dependent on the recognition of MHC class I and II antigens. MHC I is expressed on almost all cells, while MHC II is generally expressed in association with antigen-presenting cells (APCs), including dendritic cells, B cells and macrophages, as well as microglia and astrocytes in the CNS. MHC class I is principally associated with CD8 cytotoxic T-cell activation, whereas MHC class II is principally associated with CD4 T-cell activation; both lymphocyte subtypes participate in immunorejection responses. In addition to a requirement for the recognition of an MHC/antigen complex displayed by an APC by the T-cell receptor, T-lymphocyte activation also depends on exposure to an array of costimulatory ligands (e.g., B7). If T-cell receptor binding to an antigen/MHC complex takes place in the absence of a costimulatory activation signal, a T-cell can be rendered unable to respond to that antigen (anergic), a mechanism that has been suggested as a means of achieving tolerance .
Increasing evidence suggests that other aspects of the immune system contribute to both allogeneic and xenogenic rejection, including natural killer (NK) cells, the complement cascade and the lymphatic system [42–44]. In this regard, it is not only the expression, but also the lack of expression, of MHC I that can invoke an immune response by NK cells [45,46]. As the immunological barrier grows, the role of these additional mechanisms of rejection may increase, becoming greater for xenotransplantation than in the case of allotransplantation . Immunorejection in the allograft setting is predominantly mediated by the adaptive immune response via T cells and B cells. By contrast, xenorejection in discordant species combinations occurs at three relatively distinct phases. Within minutes of transplantation, the humoral immune response is activated during hyperacute rejection . Within days, infiltration of inflammatory mediators, such as host mononuclear cells and neutrophils, mediate acute rejection. Finally, in the delayed xenograft rejection response, cellular mechanisms (both NK and T cell-mediated) modulate xenograft rejection . Understanding these cellular mechanisms is critical, because while T-cell expansion can be controlled by conventional immunosuppressant agents that target calcineurin signaling, such as cyclosporin A  and FK506 (Prograf®/tacrolimus) , these agents do not affect NK cells or other rejection mechanisms. In fact, transplantation of human cells into immunosuppressed rodents often results in eventual graft failure within the first few weeks after injection [52–57], suggesting that pharmacological immunosuppression is ineffective at preventing the delayed xenograft rejection response. These data suggest that immunorejection in the xenotransplantation setting is complex and requires multidimensional intervention (i.e., the administration of immunosuppressive agents targeting T-cell expansion combined with anti-NK cell), and/or anti-costimulation factor agents, to sufficiently attenuate the immune response in order to obtain long-term engraftment in immunocompetent animal models [58,59].
In accordance with the immunological barrier presented by xenotransplantation, several approaches have been utilized in an attempt to achieve high levels of transplant engraftment and cell survival in the organ and bone marrow transplantation fields. First, high-dose combinatorial treatment with multiple conventional pharmacological immunosuppressant drugs. High dose and combinatorial pharmacological immunosuppression has associated toxicity concerns in human [60,61] and animal models [62–64]. A key question to be considered is whether there may be exogenous, alternative and/or unanticipated effects of conventional immunosuppressant agents on transplanted cell populations. In this regard, multiple studies have also shown that immunosuppressive agents can interact with, and influence, various cell populations, altering cell proliferation, fate, migration and perhaps secretomes [65–67].
Second, humanized rodent models have been developed to lower the xenotransplantation barrier. Humanized mice or mouse–human chimeras are immunodeficient animals that are reconstituted with human-derived hematopoietic cells or tissues to minimize HLA mismatches with subsequent human-derived cell populations . While in many ways this model may be considered the ultimate goal for regenerative medicine research, its use is significantly confounded by several major constraints, including the low efficacy of immune reconstitution, the time required for generation of animals, specialized equipment and significant cost. Thus, humanized rodent models of cell transplantation are rarely utilized.
Third, constitutively immunodeficient animal models, in which components of adaptive and/or innate immunity are compromised or deficient, can be utilized to improve the success of xenotransplantation. Immunodeficient animals provide an environment in which the immune response is suppressed endogenously, rather than via exogenous treatment, which allows for direct hypothesis evaluation without complications that may arise from the immunosuppressive treatments necessary to support sufficient engraftment in immunocompetent animals. Historically, evidence for long-term (up to 6 months post-transplant) tolerance of cellular xenografts in immunodeficient animal models is supported by experiments demonstrating prolonged survival of human fetal tissue and blood cells [37,69], and later, pig and human islet cells in constitutively T-cell deficient mice and rats [70,71], suggesting that a lack of functional T cells at least partially circumvents the barriers of chronic rejection. However, the survival of at least mouse–mouse allografts of embryonic stem cells transplanted into heart  or muscle , and human–rat xenografts of neural stem cells transplanted into the spinal cord  has also been shown to be significantly greater in immunodeficient models compared with immunosuppressed models. While a wide variety of immunodeficient/immunocompromised rodents are available for xenotransplantation studies [74–76], not all constitutively immunodeficient animal models achieve equal levels of immunodeficiency; identification of a model that is ‘sufficiently immunodeficient’, meaning that it achieves 100% engraftment and long-term survival, is therefore essential.
Owing to a loss-of-function mutation in the mouse PRKDC gene preventing full T- and B-cell development , CB-17 SCID mice, which were used in the original hematopoetic stem cell xenografts performed by Mosier et al. in 1988 , lack functional T- and B-cells . However, SCID mice retain high levels of innate (NK cell) immunity , which precludes complete avoidance of immune rejection; the increase in graft failure in initial hemopoietic stem cell transplant studies highlights this limitation . To avoid the shortcomings observed in these early SCID models, alternative immunodeficient animal strains have been generated to further improve graft survival . Nonobese diabetic (NOD)-SCID mice, which in addition to the T- and B-cell deficiencies of CB-17 SCID models, also display reduced hemolytic complement levels, reduced dendritic cell function and defective macrophage function , as well as reduced NK cell activity , have been used extensively in a multitude of different stem cell and transplantation studies with great success . Additional SCID variants include β2 microglobulin-deficient (B2Mnull) mice, which display limited amounts of MHC class I (classical and nonclassical) on the cell surface and therefore prevent CD8 T-cell development , and recombinase activating gene 1- and 2-deficient (Rag1null and Rag2null) mice, which, similar to PRKDC mutant mice, do not have the ability to generate fully mature T- and B-cell lymphocytes due to failure of DNA strand break V(D)J recombination [82,83]. Furthermore, recent development [84,85] of genetic variants with nearly complete ablation of T-, B-, and NK cell activity offer even more effective options in a xenograft transplantation setting . These include NOD-SCID IL2RG, and Rag2null IL2RG mice, which include a null mutation in the gene encoding the IL-2 receptor γ chain (IL2Rγ), which prevents cell surface signaling to several interleukins as well as NK cell differentiation . Additionally, larger rodent models lacking certain components of the immune response also exist and may be utilized in experiments where smaller laboratory mice are not an appropriate choice; the principle example is the athymic nude rat, which lacks a normal thymus and functionally mature T cells . However, caution should be exercised when considering nude rodent models, as evidence suggests normal to increased levels of NK cell activity , which may be sufficient to induce graft rejection [87,89]. Accordingly, selection of an immunodeficient mouse (or rat) model should be considered based on the known combination of deficits in the immune response and resulting engraftment characteristics.
In contrast to the hematopoietic transplantation field, in which constitutively immunodeficient animal models rapidly gained widespread use because they enabled the study of both normal and malignant hematopoietic repopulation , the neurotransplantation field has not followed this path. In fact, neurotransplantation research was heavily directed in its early foundations by a small body of data suggesting that the CNS is immunoprivileged, which led to the widespread belief that achieving engraftment in the CNS was a relatively easy task. Billingham and Boswell first suggested the term ‘immunologically privileged’ in 1953 , in a paper in which they described evidence of longer tissue graft survival in some sites (e.g., the cornea) in comparison to others (e.g., skin). The concept of the CNS as an ‘immunoprivileged’ site was extended later based on similar tissue grafting studies using brain tissue . Several mechanisms explaining the relative immunoprivilege of the CNS were hypothesized, including the tight nature of the blood–brain barrier, the absence of professional APCs (which are required to mount a T-cell-mediated adaptive immune rejection response) in the CNS, the lack of MHC expression in the CNS, reports of high levels of factors with immunomodulatory properties in the CNS (e.g., TGF-β), and the absence of traditional lymphatic drainage in the brain as an organ [41,93]. Combined, these factors were thought to render the immune system incapable of mounting an effective rejection response within the CNS.
More recent data has made it clear that, while the CNS may be ‘immunologically quiescent’ , it is not at all immunologically incompetent. In fact, the ability of T cells to conduct surveillance via migration across even an intact blood–brain barrier is now recognized as a normal part of immune system function . Similarly, we now know that there is breakdown of the blood–brain barrier and evidence for MHC I/II upregulation in CNS injury/disease, the capacity for microglia, infiltrating macrophages/monocytes, and astrocytes to express MHC II and function as APCs , and lymphatic drainage from the CNS to the cervical lymph nodes, which may serve as sites for antigen-mediated activation of the adaptive immune response .
It is not only the CNS itself that has been assumed to possess an immunoprivileged status. Like mesenchymal cells, embryonic stem cells and their derivatives, as well as neural stem cells, have been ascribed a mix of immunoprivilege, immunosuppressive and immunomodulatory properties , and have even been reported to fail to induce an immune response in immunocompetent animals [98,99]. Initial reports from work with embryonic stem cells suggested that these cells express very low levels of MHC I and II [100,101], which were not sufficient to stimulate T-cell proliferation in vitro, and that their inherent immunoprivileged state could contribute instead to the induction of host tolerance to embryonic stem cell-derived tissues . Similarly, neural stem cells have been reported to express low levels of MHC [103,104], and to lack the expression of co-stimulatory molecules required for T-cell activation . The combination of CNS and stem cell immunoprivilege led to the assumption, in some cases, that immunosuppression would be unnecessary for clinical stem cell allograft transplantation, or even for proof-of-concept xenograft studies in animal models.
However, as in the case of the immunoprivileged status of the CNS, there is a wealth of more recent data contradicting the existence of immunoprivilege in stem cell populations [41,106–109], from which several common themes emerge. First, embryonic stem cells, their differentiated progeny and fetal neural precursor populations can all upregulate MHC expression in vitro and upon in vivo transplantation [108,110], and thus, permit immunorejection . Second, MHC I expression in both embryonic and neural stem cells is dramatically upregulated following cytokine exposure, notably exposure to IFN-γ [101,112]. IFN-γ is expressed at high levels in many CNS disease and injury states, a point that should be noted in combination with the potential role of disease and/or trauma-induced inflammatory responses to augment immunorejection . Third, NK cells exhibit the capacity to target both embryonic and neural stem cells in vivo and/or in vitro [45,112]. Taken together, the current literature for human embryonic and neural stem cells does not support the capacity to transplant these cell populations with immunological impunity.
Up to this point, we have discussed the necessity of starting with an animal model that mimics the clinical pathology for the neurological disease of interest and supports engraftment as well as long-term survival, explored lessons learned from the hematopoietic transplantation field and mechanisms of allogenic and xenogeneic rejection, and reviewed early classifications of the CNS, as well as embryonic and neural stem cells, as ‘immunoprivileged’. We now present a comprehensive review of the literature regarding achieving engraftment and cell survival with xenografts in the CNS. We focused our literature review on studies xenotransplanting human cells into rodent models of neurological diseases/trauma or into the normal brain. For purposes of this review, we define ‘engraftment’ as the percentage of animals that demonstrate surviving cells (total number of animals at sacrifice with human cells still present divided by the total number of initially transplanted animals × 100). In parallel, we define ‘cell survival’ as the total number of cells present at sacrifice. The total number of cells must have been assessed either via unbiased stereology (typically via optical fractionator) or bioluminescence for the percentage cell survival report to be included in our tables. Unbiased stereological techniques permit rigorous quantitative analysis of tissue, including accurate volume-corrected estimates of cell number; because changes in tissue and structure volume due to disease/injury pathogenesis can be a significant experimental confound, stereological analysis is the gold standard for determining cell number, lesion volume and other variables vulnerable to these artifacts (see [114,115] for a review of the use of stereology in neuroscience). A limitation to the cell survival data generated by stereological analysis in the CNS is how the region of interest is defined. If the analysis is confined to an anatomically defined region (e.g., the striatum) , but the transplanted cell population has migrated beyond this artificial boundary, the determination of estimated cell number will only be accurate within the striatum, and the total number of surviving cells will be underestimated. Because of this limitation, papers that have this constraint in stereological data collection have been included in the overall tables, indicated by a symbol, and in the calculation of percentage engraftment, but have not been included in the calculation of percentage cell survival. Bioluminescence also permits quantitative analysis of cell survival within tissue, however, there are two critical limits for cell detection. First, sensitivity; while stereology permits an estimate of total cell number based on the detection of every cell visualized by immunoperoxidate, immunofluorescence, or promotor driven fluorescence in a transplanted tissue, bioluminescence has a clear threshold for detection that is affected by many factors, including transplantation site/depth. In the neural stem cell transplantation field, at least one study has demonstrated that the number of luciferase-expressing cells necessary to generate a detectable bioluminescence signal is in the order of 1000 . Coupled with this detection threshold limit, the propensity of transplanted cell populations for migration will significantly affect the accuracy of total cell survival quantification by this method. Finally, the long-term stability of luciferase expression has not been established, and decrements in signal may, in some cases, result from promoter downregulation . Owing to the limitations of bioluminescence in providing accurate total cell survival quantification, papers using this method of quantification have been included in the overall tables, indicated by a symbol, and in the calculation of percentage engraftment, but have not been included in the calculation of percentage cell survival.
Literature searches for this analysis of xenotransplantation in the CNS were performed in January 2011 with the keywords ‘human stem cell’ in combination with other keywords in series: ‘transplantation’, ‘brain’, ‘CNS’, ‘spinal cord’, ‘spinal cord injury’, ‘stroke’, ‘middle cerebral artery occlusion’, ‘ischemia’, ‘brain injury’, ‘brain trauma’, ‘multiple sclerosis’, ‘Parkinson disease’ and ‘Huntington disease’. Additional references were added when found cited in the initial round of papers retrieved via MedLine. No papers were excluded from our analysis, a priori. Using these criteria, we found 133 unique, relevant papers. It should be noted that the primary focus of any given paper need not have been the key variables discussed in this article (i.e., engraftment and cell survival). Rather, many papers compared a cell line with and without an experimental treatment or other component in an injured environment, or the effects of a cell line on functional outcome, and did claim to assess either engraftment or quantify total surviving cells. We grouped these papers into three primary categories based on common features of the model: models of normal neonatal or adult brain; models of acute/traumatic injury (SCI, traumatic brain injury or stroke); and models of chronic/atraumatic injury (Parkinson’s disease, Alzheimer’s disease, Huntington’s disease, amyotropic lateral sclerosis, allodynia or demyelination). Within each primary category, references were subdivided into those using immunocompetent animals with immunosuppression, those using immunocompetent animals without immunosuppression and those using immunodeficient animals. A summary of the primary categories and the number of papers using immunocompetent versus immunodeficient animals is shown in Table 1.
We recorded 25 variables for each paper: citation, URL, model, primary category of model, paradigm or method, cell type, host species/sex, immunocompetent or immunodeficient host, location of injury, transplant time (time post-injury), final dose of cells, volume of injection, route/location of transplant, immunosuppressant, dose and duration, detection method for human cells, quantification method (stereology or number of sections), treatment groups, survival time (post-transplant), total number of transplanted animals in a group, numbers of animals with cells at sacrifice, percentage engraftment, and behavioral outcome. Factors that were not applicable for a given model or not reported were noted as well. In order to accurately assess the influence of immunocompetent (with or without immunosuppression) or immunodeficient models on successful engraftment and cell survival, we included only those studies using cell suspensions and excluded studies using solid grafts where total cell number (dose) or surviving cell number could not be determined. Individual tables are provided for each of the three primary categories (models of normal neonatal or adult brain, models of acute/traumatic injury and models of chronic/atraumatic injury, as defined previously); these primary categories are further subdivided by whether cells in the identified studies were transplanted into immunocompetent animals without immunosuppression, immunocompetent animals with immunosuppression, and immunodeficient animals (Tables 3–11).
Table 1 shows that of the 133 unique papers, six papers examined xenotransplantation in normal brain (5%), 95 papers examined xenotransplantation in acute/traumatic models (71%), and 32 papers examined xenotransplantation in chronic/atraumatic models (24%). A total of 79 papers (59%) reported neither engraftment nor survival, while only 11 papers (8%) reported both engraftment and survival. For example, in the primary category ‘acute/traumatic’, 37 out of 95 papers (40%) used no immunosuppression in immunocompetent animals. Seven of these 37 papers reported the percentage of animals engrafted, only one reported the percentage of cell survival based on stereology, no papers reported both engraftment and percentage of cell survival, and 30 reported neither variable. In this category (acute/traumatic), the one paper to report cell survival (based on stereology) in a stroke model without the use of immunosuppression reported that only 35% of the initial cell dose survived 8 weeks post-transplant across four treatment groups  (see Table 7 for additional details).
Several key conclusions can be drawn from Tables 1 & 2. First, the majority of CNS xenografts have used immunocompetent animals coupled with immunosuppression (n = 69 or 52%). The next most common paradigm is to use immunocompetent animals and no immunosuppression (n = 47 or 35%). Only 19 papers (or 14%) used immunodeficient animals in CNS xenograft studies using human cells. Second, the percentage of animals engrafted (when reported) is usually highest in immunodeficient animals. Third, the paucity of studies in the normal CNS and chronic/atraumatic cohorts that have employed immunodeficient models and quantitatively assessed engraftment/cell survival makes acute/traumatic models the only category in which these variables can be compared between immunocompetent and immunodeficient animals, and makes this the most robust category from which to draw relative conclusions for xenotransplantation success. In the acute/traumatic/cohort, the highest percentage of cell survival is in immunodeficient animal models (263%, n = 4). Fourth, although the studies available for comparison are limited, the data suggest that an uninjured niche (normal brain) may be no better than an injured niche in terms of engraftment or cell survival. Finally, caution should be exercised in interpreting papers that report surviving cell numbers. Many papers extrapolate total cell number from a limited number of histological sections. Moreover, even when systematic random sampling and stereological assessment of total cell number is performed, the number of animals assessed can be insufficient to yield interpretable numbers. For example, Suzuki et al. report the stereological assessment of total cell number in an amyotropic lateral sclerosis model where 114% of the initial dose of cells was detected 6 weeks post-transplant; but quantification was performed in only one animal .
Looking at those papers that reported either human cell engraftment or survival numbers, several interesting issues are evident. First, in immunocompetent animals given immunosuppression, cell survival decreases over time. Yashuhara et al. transplanted green fluorescent protein (GFP)-expressing human HB1.F3 fetal-derived neural stem cells in a 6-OHDA toxicity model of Parkinson’s disease and quantified GFP+/human nuclear antibody (hNuc)+ cells in every fifth section through the entire striatum . Animals received 200,000 cells in 3 μl phosphate-buffered saline into the ipsilateral striatum immediately following the 6-OHDA lesion. Sets of animals (n=8 per time point) were sacrificed 3, 7, 14 or 28 days post-transplant. All animals received daily intraperitoneal injections of cyclosporin A (10 mg/kg). Cell survival within the striatum was 11.2, 4.7, 1.7 and 1.1% across the four time points, respectively. Although quantification was not a true stereological assessment (neither random counting frames nor stereological dissectors were used), these cell survival numbers are directly comparable within the study and demonstrate that a host rejection response was likely active and involved in rapidly killing the human cells over time. Notably, it is possible that GFP expression may have also been downregulated over time in conjunction with a rejection response, resulting in the observed reduction in total cell numbers over time; however, antibodies to a human nuclear antigen were also used to detect the transplanted cells .
Second, while direct comparisons within a single study between groups of immunocompetent animals receiving either immunosuppression or no immunosuppression are rare, not using immunosuppression in immunocompetent animals significantly reduces successful engraftment. Wennersten et al. transplanted 210,000 human fetal-derived neural stem cells into a contusion model of SCI immediately postinjury . Animals received cyclosporin A for 3 or 6 weeks post-transplant (4 mg/kg Monday and Wednesday, and 8 mg/kg on Friday), while a third group received no cyclosporin A post-transplant. The three groups were sacrificed 6 weeks post-transplant and the presence or absence of humans cells was confirmed using hNuc immunohistochemistry. All animals receiving cyclosporin A, regardless of length of administration, were successfully engrafted, but only one of six animals (16.7%) contained human cells in the no cyclosporin A group. A fourth group of animals (n = 8) received cyclosporin A for 3 weeks but was allowed to survive 6 months instead of 6 weeks; five out of eight (62.5%) exhibited successful engraftment at 6 months. Quantification of cell numbers was not performed . While this study suggests that transient immunosuppression may suffice to achieve long-term engraftment, it also demonstrates that using no immunosuppression in immunocompetent animals significantly reduces the rate of successful engraftment. Unfortunately, without quantification of total cell numbers in such a study, it is impossible to ascertain the effect of short versus long-term immunosuppression on cell survival.
Third, in the few studies that conducted direct comparisons, human cell engraftment was shown to be higher in immunodeficient animals than immunocompetent ones. Deng et al. transplanted human olfactory ensheathing cells into a hemisection model of SCI using either athymic nude (immunodeficient) or Sprague–Dawley (immunocompetent) rats . Their data demonstrate that survival of human olfactory ensheathing cells transplanted into immunocompetent animals was minimal at 24 h post-transplant and no surviving cells were identified by 7 days post-transplant; robust macrophage infiltration was found at the injection site by 7 days and engraftment was 0%. Conversely, human olfactory ensheathing cells transplanted into athymic nude rats survived and migrated away from the site of injection at 24 h and 7 days post-transplant. Engraftment was observed in 40% of the athymic nude rats, while no engraftment was observed in Sprague–Dawley rats . Although this study transplanted olfactory ensheathing cells rather than a strictly defined stem cell population, the proliferative properties of olfactory ensheathing cells in vitro are well known, suggesting that activation of the host immune system initiates a rejection response.
The goal of this comprehensive literature review was to assess the status of the field in achieving adequate engraftment and survival in xenotransplantation models to predict clinical translation. As noted earlier, it should be acknowledged that the principal end point of a given study may not have been the assessment of cell survival per se, and such studies can still contribute meaningful data to the literature. This review of 133 xenotransplantation papers shows that the field of regenerative medicine has focused heavily on the administration of cyclosporin A alone in immunocompetent animals as a strategy for proof-of-concept experiments, resulting in both poor engraftment and low to very low cell survival (when reported). Furthermore, this summary shows that xenografted stem cells retain proliferative capacity in immunodeficient versus immunosuppressed models within the acute/traumatic category. While other differences in the acute/traumatic CNS niche could contribute to differences in the retention of proliferative capacity, it is likely that these differences provide insight into the capacity to initiate immunorejection mechanisms in these conditions. Again, given that acute/traumatic models represent the only category in which multiple studies have quantified engraftment and survival in immunodeficient animals, these data represent the most robust category from which to draw relative conclusions for xenotransplantation success. Combined with our historical survey of the broader xenotransplantation field, this analysis clearly suggests that it will be necessary to administer a multimodal course of pharmacological immunosuppression to achieve meaningful engraftment of a transplanted human cell population when using immunocompetent animals. Alternatively, immunodeficient animals yield much higher engraftment and cell survival numbers than using immunocompetent animals (with or without immunosuppression).
It seems evident that, ideally, preclinical testing of safety and efficacy should have the goal of achieving a human–mouse xenograft that is as comparable as possible to the human clinical setting (i.e., a human–human allograft). In this regard, we begin by considering what features one can reasonably predict to be associated with human–human allografts in the CNS.
First, as we have shown in this article, one would expect that the immunological barrier associated with a human–human allograft would be less than that associated with a human–rodent xenograft, and be principally T-lymphocyte mediated, and therefore require less in the way of multifaceted immunosuppression protocols. Accordingly, the administration of drugs targeting T-cell proliferation and expansion via calcineurin signaling may be sufficient to maintain long-term cell survival in the clinical setting, especially under conditions in which the partial restriction of access of the immune system to CNS parenchyma maintained by the blood–brain barrier is restored over time. By contrast, we know that this is not the case for xenotransplantation, and at least NK cells must also be suppressed (e.g., as in the case of SCID beige mice or athymic nude rats). Second, one would expect that a more successful pharmacological immunosuppression protocol could be achieved in human–human allografts than in human–rodent xenografts. Human immunosuppression protocols are understandably considerably better characterized and designed in terms of the pharmacokinetics of drug delivery and metabolism, and achieving target circulating drug levels in man is more precise, particularly when one considers the side effects of subcutaneous delivery and variability in oral delivery of immunosuppressants in rodent models. Moreover, the pharmacokinetics of immunosuppressant metabolism are almost never monitored or accounted for in rodent models, resulting in both reduced efficacy and increased toxicity. In this regard, for example, there are known differences in the peak/trough levels of immunosuppressant metabolism and effective dosing strategies . Furthermore, rodent and human T cells can exhibit different levels of responsiveness to calcineurin inhibitors such as cyclosporin A . Third, one can predict that several factors will gain momentum in the future, further reducing the immunological barrier of human clinical allotransplantation : MHC I matching by virtue of the generation of cell banks for candidate clinical cell therapeutics [124,125]; the development of clinical strategies to achieve allograft tolerance [126,127] and/or an eventual shift to autologous cell transplantation, for example using induced pluripotent cells from somatic sources [5,128]. As a result, one would predict both higher levels of engraftment and a greater degree of cell survival in human–human allografts than in human–rodent xenografts under immunosuppressive therapy, particularly under optimal immunosuppression and MHC matching conditions, and certainly in the event of successful immunotolerance strategies and/or autologous transplantation.
Many candidate clinical cell therapeutics under current investigation are partly committed stem/progenitors, rather than terminally differentiated replacement cells; accordingly, transplanted cells retain some proliferation capacity. Taking the transplantation of human neural stem cells into constitutively immunodeficient animals as an example, it is clear that these cells, at least initially, retain the capacity for proliferation. At the simplest level, this is evident because there a greater number of surviving cells at the time of sacrifice than initially transplanted, albeit in the absence of tumorigenesis (Tables 1 & 3–11). We suggest that preclinical animal models in which the number of surviving cells is a fraction of the number contained in the initial transplant cannot provide informative data regarding proliferation potential, and therefore, cannot provide adequate data for predictive validity in the human clinical setting. The limits of such data are particularly relevant for establishing a risk versus safety profile for a candidate clinical cell therapeutic, as tumorigenesis is known to be dependent on both dose and cell survival, and is greatly attenuated in immunosuppressed immunocompetent versus constitutively immunodeficient animal models [17,18]. However, while it might be argued that efficacy data would only be enhanced by increased cell survival, therefore supporting the use of immunosuppressed immunocompetent animal models, we suggest that this is an assumption with little or no supporting empirical data. Cells respond to both the microenvironment into which they are transplanted, and the conditions of that microenvironment. The potential effects of a microenvironment in which there is an active immune response, or conversely, an increase in cell survival due to a lack of immune-mediated cell death, on cell fate and migration are essentially unknown, and are likely to be different for each candidate clinical cell therapeutic. Furthermore, the potential for an active immunorejection response to contribute to disease modifying activity cannot be ruled out. Owing to the known effect of dose and immune rejection on tumorigenesis, and the unknown effect of dose and immune status on fate, migration and/or disease modifying activity, simply scaling up the initial cell dose administered in an immunosuppressed immunocompetent animal model is not adequate to address either tumorigenesis or efficacy. Accordingly, preclinical studies should be designed to reduce immunorejection of transplanted cells in order to optimize cell engraftment and survival, and preclinical transplantation models should seek to achieve maximal theoretical engraftment in order to provide informative safety and efficacy information. We therefore suggest that immunodeficient animal models should be the model of choice for preclinical testing of safety and efficacy for candidate clinical cell therapeutics.
There are two principal issues in considering this approach. One key question is whether there are exogenous, alternative and/or unanticipated effects of conventional immunosuppressant agents on the transplanted cell population. For example, on overall cell engraftment, or cell fate/migration that could be critical in a clinical human–human allotransplantation setting where these agents will have to be administered post-transplantation, at least for a period of time. Many investigators have used this rationale to continue with experiments in xenograft models using pharmacological immunosuppressive therapy. However, we suggest that it is time for the field of regenerative medicine to re-evaluate this concept. We suggest that a more informative approach would be to establish data from two models. First, investigation of engraftment, tumorigenesis, fate and migration in immunodeficient animal models treated with the planned clinical immunosuppressive therapy. In this paradigm, any potential direct effects of immunosuppressant treatment on the candidate clinical cell population would be more likely to be revealed, given the increased overall cell survival, and reduced likelihood of the inadvertent selection against either the primary stem cell population, or precursor/progenitor populations, providing an improved assessment of tumorigenesis and safety. Second, the effect of immunosuppressant withdrawal on target organ integrity, engraftment, tumorigenesis, fate and migration in an immunocompetent animal model treated with the planned clinical immunosuppression therapy for a transient period following cell transplantation.
A second key question is whether predictive models of neurological disease/injury for testing cellular therapeutics can be established in immunodeficient versus immunosuppressed models. It is increasingly clear that there is a role for both the innate and adaptive immune responses in many types of neurological disease and injury. In the case of autoimmune diseases of the CNS, this role can be pivotal to the development of pathology (e.g., the generation of autoreactive T cells to myelin epitopes in multiple sclerosis); while the specific role of adaptive immune responses and T-cell activation are less clear, it appears certain that Alzheimer’s disease, Parkinson’s disease, stroke, traumatic brain injury and SCI all include an inflammatory component in their pathogenesis [129–134]. Furthermore, in the case of traumatic injury models such as SCI, access of the immune system to the CNS is greatly enhanced by breakdown in the blood–brain barrier, and there may be profound, prolonged and diverse effects of immune activation on pathogenesis and functional outcome . Understandably, the generation of transgenic neurological disease models backcrossed onto constitutively immunodeficient mouse strains is a difficult process. Furthermore, the effect of constitutive ablation or attenuation of T-cell and NK cell responses associated with immunodeficient models, such as the NOD-SCID mouse, on innate immune responses, inflammation and the essential characteristics of lesion pathogenesis would have to be tested as a part of validating a predictive animal model . At least in the case of traumatic SCI, the host macrophage/microglia response, neutrophil response and evolution of the central lesion are overtly unaltered in comparison with other mouse strains . As noted earlier, however, the immune and inflammatory microenvironment, composed of a host of complement proteins, cytokines and chemokines, may affect transplanted cell populations in ways that have yet to be recognized, and the potential for an immunorejection response to influence stem cell populations is equally as great as the potential for alteration of the immune microenvironment presented by the disease state to do so. Critically, however, in order to achieve engraftment of a candidate therapeutic cell population at any level for the purpose of assessment of safety/efficacy in animal models, it is necessary to impair the functional immune status of the host, particularly with regard to activation of T-cell- and NK cell-mediated immune responses; the difference between immunosuppressed and immunodeficient lies then in the fact that, at least under some conditions, treatment with immunosuppressants is not sufficient to achieve levels of xenoengraftment that may be comparable to human–human allotransplantation.
In summary, we suggest the following criteria for preclinical safety and efficacy studies for the application of human stem cell populations in a clinical setting:
For reprint orders, please contact: moc.enicidemerutuf@stnirper
Financial & competing interests disclosure
The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.
No writing assistance was utilized in the production of this manuscript.
Papers of special note have been highlighted as: