|Home | About | Journals | Submit | Contact Us | Français|
Over the last 50 years, mouse models of bone marrow transplantation have provided the critical links between graft-versus-host disease (GVHD) and graft-versus-leukemia (GVL) pathophysiology and clinical practice. The initial insight from mouse models that GVHD and GVL were T cell dependent has long been confirmed clinically. More recent translations from mouse models have included the important role of inflammatory cytokines in GVHD. Newly developed concepts relating to the ability of antigen presenting cell (APC) and T cell subsets to mediate GVHD now promise significant clinical advances. The ability to use knockout and transgenic approaches to dissect mechanisms of GVHD and GVL mean that mouse systems will continue as the predominant preclinical platform. The basic transplant approach in these models, coupled with modern “real-time” immunologic imaging of GVHD and GVL is discussed.
Bone marrow transplantation (BMT) was conceived in the early 1950s out of rodent studies focusing on the effects of radiation. These now classical experiments noted that shielding of the spleen, or infusion of bone marrow from a “donor animal” could protect animals from lethality [1,2]. Interestingly, this protection was initially thought to be because of the transfer of a humoral factor until subsequent chromosomal analysis confirmed the engraftment of donor cells in 1956 . Even at this time, the presence of a graft-versus-leukemia (GVL) effect was mooted . A report of clinical transplantation followed a year later by E. Donnel Thomas and colleagues , and the field was born. Over the subsequent decade the introduction of rudimentary tissue typing paved the way for the classical murine studies of Korngold and Sprent [6,7] in the 1970s, which confirmed the T cell dependence of GVHD. Indeed, much of our current understanding of BMT has its origin in their subsequent studies that determined the role of T cell subsets in GVHD across various MHC  and minor histocompatibility (HA) mismatches .
Current murine models of BMT can be broadly grouped into those in which GVHD is directed to MHC (class I, class II, or usually both) or to isolated multiple minor HA alone. Although multiple minor HA mismatches also exist in the former, their impact is usually limited relative to that induced by full MHC disparities. The GVHD that develops in response to a full (class I and II) MHC disparity is dependent primarily on CD4 T cells, although CD8 T cells can provide additive pathology. These systems, by virtue of their CD4 dependence, result in an inflammatory “cytokine storm,” capable of inducing GVHD in target tissues without the requirement for cognate T cell interaction with MHC on tissue . The BMT models in which GVHD is directed to mutated class I (bm1) or class II (bm12) MHC antigens in isolation represent well-utilized systems where GVHD is mediated only by CD8 or CD4 T cells, respectively. The parent to F1 models (eg, B6 → B6D2F1) offer the advantage of eliminating T cell-dependent host-versus-graft rejection responses (NK-dependent “hybrid resistance” remain), and so are particularly useful if one is attempting to eliminate this as a variable across groups (eg, comparing the intensity of conditioning regimes on GVHD severity independent of differential engraftment kinetics) . The induction of GVHD to multiple minor HA results in a process where either CD8 T cells, CD4 T cells, or both may play a role in disease . In contrast to CD4-dependent GVHD, CD8 T cells induce GVHD primarily by their cytolytic machinery, which requires the TCR to engage MHC on target tissue . Importantly, these CD4 versus CD8-dependent GVHD models will have differing requirements for antigen presentation. Either host or donor antigen-presenting cells (APC) will be able to initiate CD4-dependent GVHD , whereas only host APC will induce GVHD in CD8-dependent systems . Although the minor HA disparate systems may be the most appropriate models of clinical BMT, the MHC disparate systems can also induce the full spectrum of clinically relevant GVHD while permitting greater dissection of immunologic mechanisms because of the enhanced ability to measure immunologic pathways of GVHD both in vivo and ex vivo. Thus, it is critical from the outset to understand what questions are being asked so that the most relevant BMT model can be chosen (summarized in Table 1).
Inbred mouse strains are variably sensitive to radiation, and so maximal tolerated doses (primarily delivered in 2 split fractions within a day at <150 cGy/min) differ from strain to strain. As a general rule, B6 are more resistant that BALB/C mice, and F1 more resistant than parental strains. Thus, the maximal tolerated total body irradiation (TBI) dose (ie, that which will allow universal survival following transplantation of syngeneic or T cell-depleted (TCD) allogeneic grafts) is approximately 900 cGy for a BALB/C mouse relative to 1500 cGy for a B6D2F1 mouse. Generally, the higher the TBI dose, the earlier and greater the intensity of the inflammatory arm of GVHD . By contrast, BMT models utilizing low TBI doses and high donor T cell doses will result in GVHD dominated by later onset T cell-dependent pathology. This is particularly important when studying inflammatory mediators of GVHD, as inhibitors may have little effect in BMT models utilizing low TBI and high donor T cell doses, potentially giving rise to false negative results. Conversely, this system may be ideally suited to studying inhibitors of T cell effector function. BMT models using non-TBI-based conditioning have been limited, primarily because of the ease of TBI delivery and its clear clinical relevance. Nevertheless, cyclophosphamide, fludarabine, and busulphan can be delivered in mouse systems, and this may become important to model clinical nonmyeloablative transplantation. Certainly, the addition of these agents to TBI in mouse models can cause a dramatic enhancement of GVHD severity and alteration of effector pathways .
Unlike aGVHD models, the induction of clinically relevant cGVHD in BMT models using (nonmutated) inbred strains has been challenging. Perhaps the most relevant is the B10.D2→BALB/C system, originally described as a model of scleroderma in which grafts are transplanted following sublethal (600 cGy) doses of TBI [16,17]. In our experience, a similar and more penetrate spectrum of disease also develops using lethal TBI doses in conjunction with low doses of (CD4+) T cells. A second equally relevant model is the LP/J → B6 system, which also induces profound scleroderma . The DBA/2 → B6D2F1 and BALB/c→(BALB/c × A) F1 models (without conditioning) induce antibody-dependent lupus nephritis [19–21]. The B6→B6 × BALB/C F1 model (without conditioning) has been described  as inducing aGVHD before switching to a chronic (lupus nephritis) formthereafter. Unfortunately, these lupus models have only limited relevance to clinical cGVHD. Common to these models is CD4-dependent GVHD in the absence of a significant inflammatory milieu, together with limited CD8-dependent cytotoxicity. Subsequent models using MHC-deficient hosts or perforin deficient donors have also been decribed [23,24].
Perhaps the clearest advances in transplant immunology have come about through the use of mutant animals as BMT donors or recipients. Although there are some caveats to the use of these mice (predominantly controlling for developmental abnormalities), their use allows elucidation of specific cell populations or molecules therein on alloreactive responses that cannot be achieved by other means. Recently, models have been developed using immunodeficient mice that permit engraftment of human hematopoietic cells. This allows the screening of various drugs on human leukemia in vivo and also permits the study of “humanized” reagents in vivo. These animals also develop also develop xenogeneic GVHD if human T cells and APC are engrafted, although the ineffective cognate interaction between host (mouse) APC and donor (human) T cells and vice versa  does present limitations as a disease model.
GVHD and graft-versus-tumor (GVT) reactions are complex biologic processes that require in vivo modeling for the greatest understanding. Murine models have been extremely effective because of the understood genetics of many different mouse strains that are available for evaluation, as well as the similarity between murine GVHD and the human counterpart. Models across both major and minor histocompatibility mismatches have been well established in the literature. Typically, the endpoint of experimental studies is animal survival, weight loss, and gross symptoms such as diarrhea, hunchback, and fur ruffling, which are used for evaluating GVHD progression and severity. In addition, pathologic scoring systems have been established for assessing quantitatively the degree of GVHD severity; however, these latter assays require animal sacrifice and cannot be performed serially. To overcome these limitations, we have used novel concepts of bioluminescent-based imaging to track donor-derived cell populations in murine models . The basis of in vivo bioluminescence imaging (BLI) is to track cell populations that have been modified to express a bioluminescent marker gene. A number of bioluminescent molecules are available for such purposes. In the majority of studies, the North American firefly luciferase (luc) gene has been utilized because of well-characterized properties, as well as the use of a known substrate, namely luciferin, for generating light. In this ATP, magnesium and oxygen-dependent reaction luciferin is oxidized, resulting in the emission of a single photon of light. This light can be captured using sensitive CCD cameras to create an image of the light emission over time . The specific advantages of this approach are that the approach is noninvasive, quantitative, and can be performed serially in individual animals. Therefore, far fewer numbers of animals are required, and the bioluminescent signals can be used as guides to further assess the biologic processes at specific times and in special orientations.
We have utilized this approach of BLI to model GVHD (Figure 1) . A particular challenge has been to introduce the luciferase gene into cell populations of interest. Luc expression can be accomplished using cell transfer techniques; however, it is laborious and poorly reproducible. To overcome this limitation we have generated transgenic mouse strains that constitutively express the firefly luciferase driven by the chicken beta actin promoter . To further enhance utility of the model system luc has been coupled to green fluorescence protein (GFP) to create a dual bioluminescent and fluorescent based imaging system. Using these transgenic animals, any cell population of interest can be isolated and used as a donor cell population. We have tracked both conventional CD4 and CD8 cells, as well as regulatory T cells using this system [28,30].
Conventional T cells labeled with luciferase can be initially seen within 24 hours following adoptive transfer in secondary lymph nodes including Peyer’s patches, mesenteric lymph nodes, and the spleen. There these cells undergo rapid replication and upregulate certain important molecules such as α4β7 and other chemokine receptors, which allow entry into GVHD target organs. Between days 3 and day 5, donor-derived alloreactive cells are primarily located in nodal sites and then leave these sites. CD4 cells appear to infiltrate first, followed by CD8 cells. At subsequent time points the proliferating alloreactive T cells migrate out of these areas into GVHD target organs such as the skin, liver, and gut. A major question is to what are the requirements at these nodal sites for GVHD induction. We have attempted to block GVHD pathology by adding various monoclonal antibodies that block entry into GVHD target organs such as MADCAM1 or L-selectin. Both of these molecules are known to be required for entry into secondary lymph nodes. Using this approach, GVHD could only be blocked if the addition of monoclonal antibodies (mAb) was accompanied by splenectomy. In addition, further studies developing genetic strains of animals that were devoid of either Peyer’s patches or lymph nodes also indicated that there was redundancy in the activation sites such that the spleen could compensate for a loss in these other priming sites. Therefore, blocking entry of T cells into GVHD entry sites is unlikely to be a practical approach for control of GVHD because splenectomy was also required (Beilhack et al., submitted).
Additional approaches have been explored including the addition of regulatory T cell populations, as well as developing T cells that lack the ability to induce GVHD because of their isolation and manipulation in culture. CD4+CD25+FoxP+ regulatory T cells (Treg) have been particularly effective in blocking GVHD pathophysiology . In this setting, Treg are capable of homing to many of the same sites as conventional T cells, yet appear to block conventional T cell proliferation at these sites. The timed addition of Treg 48 hours prior to conventional CD4 and CD8 cells allows for a significant reduction in cell dose, which could have practical clinical implications . The mechanism of action of regulatory T cells is not known with certainty; however, they appear to require homing to secondary lymph node sites because isolation of Treg, which express CD62L, are much more effective in vivo than their CD62L− counterparts, which are both functional in vitro [32,33]. These studies provide important insights into the biologic basis of GVHD induction and also clues to the development and strategies for control of GVHD in the clinic.
Other promising approaches to enhancing GVT without GVHD includes the isolation of certain cell populations that lack GVHD inducing potential. Such examples could include the use of ex vivo culture cytokine-induced killer cells  or the isolation of memory CD4 cells [35,36], which appear to have much less GVHD-inducing potential compared to naive CD4 cells. In addition, the isolation of specific T cell populations that are cytolytic to tumor cells yet not to normal host tissues also have great promise in the clinical application of enhancing GVT without GVHD .
These bioluminescent models also demonstrate that the current approach to GVHD control may be limited by assessing the availability of drugs to block GVHD once it has already been clinically established. Clearly, important events have occurred at nodal tissues that allow these T cells to become activated and then infiltrate GVHD target organs, and additional methodologies are required to develop strategies to image T cell activation and predict GVHD induction in the clinic.
The concept of leukemia/tumor destruction by the immune system was originally envisioned by Ehrlich and later refined by Burnet . Clinical observation of such a process was made by Coley , who noted regression of tumors in patients that developed infections. The seminal demonstration of superior leukemia elimination by adoptive transfer of allogeneic cells than by syngeneic cells was made by Barnes et al.  employing murine models nearly 50 years ago. This was attributed to an immunologic reaction by the donor cells against the nonself-host type leukemia. Barnes noted that extrapolation of this observation from mouse to man for treatment of leukemia might be possible under certain contexts. Progress in clinical hematopoietic stem cell transplantation (HSCT) made it increasingly obvious that although the intensity of the conditioning regimen was important, reconstitution of immune cells from the donor graft is also critical for optimal leukemia/tumor elimination: a process called GVL/tumor effect, a term first used in murine models . It is now widely appreciated that GVL effect following allogeneic HCT is the most potent and clinically successful form of immunotherapy . The observation of the tight link between GVHD and GVL, their temporal association, and that lymphohematopoietic tissue is the primary target of GVH reaction made from mouse models . These models were also critical for the initial demonstration of the possibility for separation of GVHD from GVL . Mouse models thus have been pivotal in making the early and fundamental observations on the presence and potency of both GVHD and GVL. Other models, too, such as the canine, nonhuman primate, and rat models, have and continue to play an important roles in studying the immunobiology of allogeneic HSCT. Nonetheless, the presence of well-characterized in-bred strains, availability of knock-out and transgenic animals, easy availability of reagents, and the relative low cost have made mouse models the most utilized systems for investigating the mechanisms of GVL responses. This review will briefly discuss some of the most commonly used mouse models of GVL, the recent insights gained, and the potential pitfalls. Additionally, it is important to note that the caveats that are applicable to studying GVHD in these mouse models are also germane to GVL, which were reviewed earlier in the article.
Several strain combinations of mice and a variety of tumors have been utilized to analyze GVL. They can be broadly categorized into MHC disparate models (BALB/c or FVB → B6, B6 → AKR), parent → F1 (B6 → B6D2F1), single MHC class I or II disparate (B6 → B6bm1 and B6 → B6bm12, respectively) and MHC matched but minor histocompatibility antigen disparate strains that are either dependent on CD8+ T cells (C3H.SW → B6, B10.BR → CBA/J) or CD4+ T cells (B10.D2 → BALB/c and B6 → BALB.B) for GVH reaction. A variety of murine leukemia/ lymphoma cell lines that exhibit the classic 6 hallmarks of cancer and have spontaneously evolved (C1498, A20, and BCL-1) or viral or chemically transduced (MBL, MMB3.19, p815, EL-4), and are syngeneic to recipient animals and have been utilized .
In these models small numbers of leukemic cells are infused at the time of transplant or a few days preceding HSCT to mimic minimal residual disease. In a few studies, leukemic cells are infused days after HSCT. The HSCT protocols, conditioning regimen (type and intensity: various doses of radiation are most commonly employed), the source and purity of T cells (spleen, lymph node or BM), vary between the studies.
It is important to note that the variations in strains or tumors of the HSCT protocols can cause significant alterations in the outcomes of the studies. Nevertheless, recent studies utilizing the above models have enhanced our understanding of GVL responses. For example, studies have suggested that depending on the model and protocol, cytokines that favor either Th1 or Th2 polarization can preserve GVL responses . Data also showed that cognate interactions between T cells and tumors and T cell cytotoxic (CTL) pathways are critical for GVL . Studies with donor T cells that are deficient in specific CTL effector functions (perforin, TRAIL, FasL, TNFR1, or TNF-α) or by retrovirus expressing chronic or blast crisis phase CML resistant to specific T cell effector pathway have deepened our understanding of the GVL effector mechanisms [47,48]. Recent reports have demonstrated a potentially exciting prospect for separating GVHD from GVL by utilizing subsets of T cells that are polyclonal. Memory T cell subsets or addition of natural or adaptive regulatory T cells from naïve donor have been shown to preserve GVL despite reducing GVHD [31,49]. Novel immune cell subsets other than donor T cells—NK, NKT, and host dendritic cells—have been shown to modulate GVL [50–52]. Several reagents that alter T cell trafficking, costimulation, or survival have shown promise in these models . Recent experimental data from mouse models have also enhanced our understanding of the role of immunodominant epitopes, epitope spreading, the impact of mixed chimerism, antileukemia effects of DLI, the importance of alloantigens and crosspresentation on professional antigen presenting cells in GVL [54–56]. These studies have thus collectively provided for novel perspectives in understanding the overall framework of GVL responses.
Despite the long track record and the promise of these models in studying GVL mechanisms, it is important to consider some caveats before extrapolating the observations to human context . First, as noted above, results might vary depending on the type of HCT protocol, the timing, and dose of leukemia induction. Second, the immune responses in inbred mice that are housed under sterile conditions and are transplanted without upfront immunosuppression might not reflect those following alloHCT in outbred, immunosuppressed humans. Third, the tumor cell lines that are commonly utilized, despite exhibiting the classic hallmarks, show variability in subcloning efficiency and outgrowth of dominant clones that might likely alter the nature, immunogenicity, and density of relevant antigens. Moreover, the immune repertoire of leukemia and lymphomas that develop in these models, unlike in humans, is more homogenous, as they had not been sculpted by the process of immunoediting . Fourth, the kinetics of tumor growth and allo-responses in most mouse models is rapid, and there are no models that closely reflect chronic leukemia and cGVHD in humans.
In summary, murine model systems have provided a wealth of sound experimental data, which, when interpreted with appropriate caveats, are highly instructive for understanding the biology of GVL.