Because severe forms of the graft-versus-host reaction directed against normal tissues (also termed graft-versus-host disease [GVHD]) also contribute to morbidity and mortality following allogeneic hematopoietic stem cell transplantation, major efforts have focused on strategies to separate GVHD from the potentially beneficial immune reactivity against tumor (also called the graft-versus-tumor [GVT] effect ). This article focuses on the data supporting the contribution of the GVT effect to cure of malignancy, what is known about the biology of the GVT reaction, and, finally, strategies to manipulate the GVT effect to increase the potency of HSCT.
Transplantation of hematopoietic stem cells from either bone marrow, peripheral blood, or umbilical cord blood from one individual to another following the administration of high doses of cytotoxic chemotherapy or radiation therapy (allogeneic hematopoietic stem cell transplantation [allo HSCT]) can cure malignancy. Indeed, this approach is the standard of care for children with acute leukemias at high risk of recurrence following chemotherapy. When allo HSCT was initially developed as a treatment for cancer, the concept was largely based on the opportunity to administer lethally myeloablative doses of cytoxic therapy to eradicate leukemia within the normal bone marrow with rescue provided by the donor hematopoietic stem cells. However, as experience with this procedure progressed, it became apparent the immune reactivity of the donor immune system against the recipient, termed graft-versus-host reaction , was associated with diminished risk of relapse, indicating that the curative potential of allo HSCT for cancer is multifactorial. Because severe forms of the graft-versus-host reaction directed against normal tissues (also termed graft-versus-host disease [GVHD]) also contribute to morbidity and mortality following allo HSCT, major efforts have focused on strategies to separate GVHD from the potentially beneficial immune reactivity against tumor (also called the graft-versus-tumor [GVT] effect ). This article focuses on the data supporting the contribution of the GVT effect to cure of malignancy, what is known about the biology of the GVT reaction, and, finally, strategies to manipulate the GVT effect to increase the potency. Discussions about the biology and treatment of GVHD, and about the closely related topic of tumor-directed immunotherapeutic strategies to treat pediatric cancer, are contained in other articles in this issue.
Evidence for graft-versus-tumor reactivity
Although HSCT was originally conceived as an approach to replace defective or cancer-containing bone marrow and to rescue the patient from lethal doses of chemoradiotherapy, it was eventually recognized that the transplant offered additional advantages in terms of preventing cancer relapse. In fact, the first evidence for an allogeneic antileukemic effect came earlier from murine models where it was observed that radiation alone was not curative in all cases of leukemia, but that the addition of immunologically disparate allogeneic bone marrow could improve the likelihood of leukemia eradication. The earliest publications of large series of allogeneic transplants undertaken for leukemia in the late 1970s noted that the development of acute or chronic GVHD was associated with decreased risk of relapse. The role of T cells in the GVT reaction was demonstrated first in preclinical models and subsequently in humans where increased relapse counterbalanced the beneficial effect of T cell depletion for GVHD in patients with leukemia. The dramatic effect of GVHD on relapse emerged from a retrospective review of a large series of transplants for leukemia performed at multiple centers. In this study, the risk of relapse was shown to be significantly affected by type of transplant or the development of GVHD with the highest relapse risk in patients undergoing transplants from genetically identical twins (also reported in other studies). The next highest risk of relapse was seen in recipients of transplants using T cell–depleted grafts and in recipients of transplants from T cell–containing grafts not developing GVHD. The lowest risk of relapse occurred in patients with acute or chronic GVHD. There appears to be a graft-versus-leukemia (GVL) effect (measured by decreased risk of relapse) even in the absence of clinically evident GVHD, suggesting that these two immunologic aspects of transplant can be separated. Additional evidence for the GVL effect comes from data demonstrating that the level of immune suppression given to prevent GVHD can also affect the risk of relapse. Finally, as will be discussed in more detail, manipulation of the immunologic milieu in the recipient with recurrent leukemia posttransplant through rapid withdrawal of immune suppression or the infusion of donor-derived lymphocytes (donor lymphocyte infusion [DLI]) can induce responses (and even remission in some cases), providing direct evidence for the GVL effect. Indeed, for recurrent chronic myelogenous leukemia (CML) detected posttransplant while still in chronic phase, the likelihood of inducing remission with DLI may be as a high as 70% to 80%, thus representing the most potent form of cancer immunotherapy. A number of recent publications have comprehensively reviewed the GVL effect.
The potency of graft-versus-tumor varies by malignancy
Multiple studies assessing response to DLI have indicated that the potency of the GVT reaction differs depending on the type of malignancy for which the transplant was performed. As indicated above, success is greatest when the GVT effect is used to treat recurrent disease posttransplant in patients with CML detected at low levels. However, as the disease burden increases and patients with CML are treated in more advanced stages, the response rate declines dramatically with only 10% to 30% of patients treated in blast crisis achieving remission. The response rate to DLI in acute myelogenous leukemia (AML) approximates that observed in CML treated in blast crisis, suggesting that a major factor determining the effectiveness of the GVT reaction is the pace of the disease.
Although disease burden and pace clearly contribute to the effectiveness of the GVT reaction, other factors certainly contribute as well, as indicated by the lower response for the GVT effect in acute lymphoblastic leukemia (ALL), the most common form of pediatric leukemia. There is definite evidence for the existence of a GVT effect in ALL, particularly in children. However, the potency of this effect is certainly lower than that for CML and, probably, AML, although in the latter case studies have been conflicting. Because of this apparent low potency, clinicians have shown little enthusiasm for treating recurrent ALL posttransplant with discontinuation of immune suppression or DLI, although reports of remission being achieved using these maneuvers have been reported. Regardless, the poor response rate indicates a clear need for strategies to enhance the GVT reaction in ALL.
In terms of solid tumors, the best evidence for the GVT effect comes from studies in lymphoma in adults (reviewed by Grigg and Ritchie). As with leukemia, there are likely to be differences in potency, depending on the type of lymphoma. Non-Hodgkin lymphoma histology is less diverse in pediatrics, with the vast majority of lymphomas being high grade. Given the small numbers of pediatric lymphomas treated with allo HSCT, it is not possible to draw firm conclusions. However, recent retrospective analyses suggest the presence of a graft-versus-lymphoma effect in B and T lymphoblastic lymphoma. Hodgkin disease may also be targeted by an allogeneic GVT effect. Finally, in terms of nonhematologic malignancies, there have been multiple reports of a GVT effect in adult renal cell carcinoma and some reports of a GVT effect in other solid tumors. However, although there are case reports and small series of allo HSCT in pediatric solid tumors, definitive data are lacking. Allo HSCT in pediatric solid tumors is the focus of several ongoing studies.
The potency of graft-versus-tumor varies by malignancy
Multiple studies assessing response to DLI have indicated that the potency of the GVT reaction differs depending on the type of malignancy for which the transplant was performed. As indicated above, success is greatest when the GVT effect is used to treat recurrent disease posttransplant in patients with CML detected at low levels. However, as the disease burden increases and patients with CML are treated in more advanced stages, the response rate declines dramatically with only 10% to 30% of patients treated in blast crisis achieving remission. The response rate to DLI in acute myelogenous leukemia (AML) approximates that observed in CML treated in blast crisis, suggesting that a major factor determining the effectiveness of the GVT reaction is the pace of the disease.
Although disease burden and pace clearly contribute to the effectiveness of the GVT reaction, other factors certainly contribute as well, as indicated by the lower response for the GVT effect in acute lymphoblastic leukemia (ALL), the most common form of pediatric leukemia. There is definite evidence for the existence of a GVT effect in ALL, particularly in children. However, the potency of this effect is certainly lower than that for CML and, probably, AML, although in the latter case studies have been conflicting. Because of this apparent low potency, clinicians have shown little enthusiasm for treating recurrent ALL posttransplant with discontinuation of immune suppression or DLI, although reports of remission being achieved using these maneuvers have been reported. Regardless, the poor response rate indicates a clear need for strategies to enhance the GVT reaction in ALL.
In terms of solid tumors, the best evidence for the GVT effect comes from studies in lymphoma in adults (reviewed by Grigg and Ritchie). As with leukemia, there are likely to be differences in potency, depending on the type of lymphoma. Non-Hodgkin lymphoma histology is less diverse in pediatrics, with the vast majority of lymphomas being high grade. Given the small numbers of pediatric lymphomas treated with allo HSCT, it is not possible to draw firm conclusions. However, recent retrospective analyses suggest the presence of a graft-versus-lymphoma effect in B and T lymphoblastic lymphoma. Hodgkin disease may also be targeted by an allogeneic GVT effect. Finally, in terms of nonhematologic malignancies, there have been multiple reports of a GVT effect in adult renal cell carcinoma and some reports of a GVT effect in other solid tumors. However, although there are case reports and small series of allo HSCT in pediatric solid tumors, definitive data are lacking. Allo HSCT in pediatric solid tumors is the focus of several ongoing studies.
Biology of graft-versus-tumor
Multiple lines of evidence support the notion that the immune system can prevent the development of cancer, can contribute to controlling the growth of cancer, and can be harnessed to treat established cancer. A detailed discussion of these topics is beyond the scope of this article but has been extensively reviewed elsewhere, as well as in other articles of this issue. Thus, the concept that the donor immune system that recovers following allogeneic bone marrow transplant can target the recipient’s tumor is consistent with a large body of data on the ability for the autologous immune system to recognize cancer. However, it has also been recognized that the growing tumor often acquires features that enable it to escape the patient’s own immune system, a development termed immunologic tolerance . Thus, the most simplistic explanation for the GVT effect is the replacement of an immune system that has been rendered tolerant by the tumor by an immune system that has not been made tolerant. While this is certainly a major contributor to the potency of the GVT reaction, there are many other layers of complexity. Perhaps the most obvious is the presence of multiple genetic differences between donor and recipient giving rise to new antigenic targets for the donor immune system to recognize. Indeed, it has long been recognized that genetic mismatch at the major histocompatibility loci (also termed tissue rejection antigens ) can result in rapid rejection of solid organs due in part to a high precursor frequency of lymphocytes (particularly T cells) recognizing proteins derived from nonself versions of these highly polymorphic genes. However, even when a donor fully matched at these genes can be identified (as in the case of a matched sibling), multiple minor antigenic differences (eg, proteins derived from the Y chromosome in sex-mismatched transplants) remain and can serve as targets for GVHD or the GVT effect. Because even minor antigenic differences are still recognized as nonself, they can induce more potent immune reactivity than can autologous immune responses recognizing self antigens on tumors. Thus, genetic disparity between donor and recipient likely serves as a major contributor to the potency of the GVT effect. Indeed, it has recently been demonstrated that when allo HSCT is performed from a major histocompatibility complex–mismatched donor (termed haploidentical as in a parental donor for a child), leukemic relapse can be associated with loss of expression of the disparate major histocompatibility complex haplotype under immunologic pressure from the donor. In addition, the recent report of the association of GVT responses with specific HLA antigen disparities in a large series of patients has further contributed to our developing understanding of the mechanism of the GVT effect.
For many disparate antigens between donor and recipient, broader expression on multiple normal tissues can make it challenging to separate GVHD from the GVT effect. Indeed, multiple clinical studies have shown that the severity of GVHD is linked to the risk of relapse. However, recent retrospective data have demonstrated that the GVT effect from matched unrelated donors (predicted to have greater genetic disparity at minor histocompatibility antigens than siblings) was not more potent than in related donor transplants. Furthermore, there is evidence for the GVT effect even in the absence of clinical GVHD. Restricted expression of minor histocompatibility antigens on the cancer (or at least in the compartment in which the cancer resides, such as bone marrow in the case of leukemia) provides the potential for some selectivity of the GVT reaction. Finally, since the donor immune system will not have been exposed to the malignancy, tolerance to tumor-restricted antigens that can serve as targets of an autologous antitumor immune response may not have occurred. Indeed, donor-derived responses to the leukemia-associated antigens derived from the Wilms tumor 1 ( WT1 ) and proteinase 3 ( PR3 ) genes have been identified in allo HSCT recipients and responses to these antigens are associated with reduced risk of relapse. Several potential targets of an allogeneic GVT response are summarized in Fig. 1 . Ultimately, the potency of the GVT effect is likely to be derived from the composite immune reactivity against histocompatibility antigens (minor or major, depending on the degree of donor-recipient matching) and tumor antigen-directed immunity. Thus, manipulating the relative contribution of tumor-directed reactivity and responses to more broadly expressed antigens may be a potential strategy to increase the selectivity of the GVT effect (as discussed elsewhere in this issue).
While T cells mediate immunity, other immune cells also contribute to the GVT reaction. Recent data have demonstrated that, as with T cell responses, clinical responses to the GVT effect are associated with the development of antibody responses to tumor-specific antigens derived from immunoglobulin gene rearrangements in lymphoid malignancies or from more broadly expressed minor histocompatibility antigens, such as HY.
Adaptive immune responses (T cell or B cell) clearly play a major role in the GVT effect as depletion of donor T cells from the infused allograft and poor lymphocyte reconstitution following allo HSCT can result in increased risk of relapse, including pediatric ALL. However, activation of the nonspecific innate immune system can modulate the potency of an adaptive immune response (reviewed by Zitvogel and colleagues). Following allo HSCT, heightened innate immune activation is induced by inflammation from the preparative regimen and is a central feature of the pathophysiology of GVHD. Thus, it is likely that the potency of antigen-specific adaptive GVL responses may be modulated by increased innate immunity present in the post–allo HSCT environment and the impact can likely be both beneficial or detrimental, depending upon the setting.
Natural killer (NK) cells are also capable of mediating a potent GVT effect, particularly early after HSCT. Unlike T and B lymphocytes that generate an antigen receptor that recognizes a specific target, NK cells mediate cell killing. This characteristic is based in part on lack of self (discussed elsewhere in this issue). Thus, depending on degree and type of matching, donor NK cells can have varying degrees of reactivity. Indeed, multiple retrospective studies have shown that, for AML, NK reactivity is associated with a markedly decreased risk of relapse. Interestingly, although NK cell–mediated GVT effect was originally associated with AML and believed to have a minimal effect in ALL, recent data have suggested that pediatric ALL may be more susceptible to NK cells than adult ALL. However, the NK cell effect, while potent, is transient and limited to early post–allo HSCT before T cell recovery. Thus, the optimization of the GVT effect to prevent relapse long term may require the memory associated with adaptive immune responses.
Strategies to enhance graft-versus-tumor
The clinical GVT effect plays an important role in curing children with acute leukemia treated with allo HSCT. In recent years, different approaches aimed at augmenting or inducing an immune GVL effect have been used to treat relapsed leukemia after allogeneic bone marrow transplant. Such approaches include abrupt cessation or rapid tapering of immune suppression ; administration of cytokines, such as interleukin 2 ; and DLI with or without cytokines. As discussed above, although the benefit of immunotherapy for CML is well documented, there are fewer reports of success in patients with acute leukemia. Furthermore, clinical response to immunotherapy has usually been associated with measurable GVHD, possibly caused by high doses of donor cells given to patients in frank hematological relapse. However, there is evidence that low doses of donor T cells may also induce an effective GVT effect and produce long-term remission in patients whose leukemia burden is small.
Guimond and colleagues demonstrated that mixed hematopoietic chimerism (mixed donor and recipient cells) in T and NK cell subpopulations can frequently distinguish pediatric patients with leukemia relapse from children in remission, but such mixed hematopoietic chimerism is not useful for making the similar distinctions among adult patients. These findings support the hypothesis that a state of mixed hematopoietic chimerism may reduce the clinical GVT effect of alloreactive donor-derived effector cells in patients with acute leukemia and may facilitate the proliferation of residual malignant cells that may have survived the preparative regimen. Barrios and colleagues could show in 133 patients with acute leukemia that patients with increasing mixed hematopoietic chimerism had a significantly elevated risk of relapse. Based on these studies, several consecutive trials were initiated evaluating the hypothesis that relapse of acute leukemia could be prevented by preemptive immunotherapy on the basis of chimerism analysis. A large, prospective multicenter trial including patients with ALL demonstrated that (1) serial analysis of chimerism reliably identifies patients with highest risk of relapse and (2) that overt hematological relapse of ALL can primarily be prevented by withdrawal of immune suppression (cyclosporine A [CSA]) or administration of low-dose DLI based on chimerism analyses. As indicated earlier, the presence of a GVT reaction in patients with ALL has been suggested by the higher incidence of relapse in the absence of GVHD or with the use of T cell–depleted allografts. These results are consistent with the finding that the highest incidence of increasing mixed hematopoietic chimerism was detected in those patients who received T cell–depleted stem cells. This supports a model wherein that T cell depletion substantially reduces the alloreactive potential of the graft, facilitates the recurrence of autologous hematopoiesis, and allows the reexpansion of underlying disease. Molecular evidence of persisting or reappearing recipient cells may be a reflection of either survival of leukemic cells, or of survival of normal host hematopoietic cells, or of a combination of both. Surviving host normal hematopoietic cells may, in turn, facilitate the reemergence of a malignant cell clone by inhibiting immune competent donor effector cells. Fig. 2 illustrates the posttransplant course of a patient suffering from pre–B cell ALL transplanted in second remission from a matched unrelated donor. The patient developed decreasing donor chimerism posttransplant without the detection of leukemic cells or minimal residual disease (MRD). Consequently the patient became MRD positive and finally, demonstrated definitive relapse. This outcome is in concordance with patients treated for CML where it has been clearly demonstrated that reappearance of host hematopoietic cells in the mononuclear cell fraction precedes hematological relapse. Based on these findings, the development of mixed hematopoietic chimerism has been considered to reduce the potency of the GVT effect.
As mentioned above, support for the GVL effect in patients with ALL was documented in a study from Locatelli and colleagues, who reported that lower-dose CSA reduced the risk of relapse in children with acute leukemia receiving allo HSCT transplants from human leukocyte antigen–identical siblings. Slavin and colleagues reported the first successful implementation of DLI in a patient with ALL relapsed after allo HSCT. Since that time, experience with immunotherapy had shown that DLI initiated during frank hematological relapse induced complete remission in fewer than 10% of patients with ALL and in fewer than 25% of patients with AML. If tumor burden was reduced by chemotherapy before DLI, the rate of complete response significantly improved (to 33% in ALL and 37% in AML). These results suggested that immunotherapy offered the greatest benefit to patients with acute leukemia if administered before hematological relapse occurs. Fig. 3 illustrates the posttransplant course of a patient suffering from common-ALL transplanted in second remission from a matched unrelated donor. An MRD level of 1E-03 (1 in 1000 bone marrow cells) was detected before transplant. Posttransplant, the patient was 95% donor chimera. CSA was withdrawn, and a DLI administered to augment the GVT effect, which was associated with return to full donor chimerism. On day 60, after an MRD level of greater than 1E-04 (1 in 10,000 bone marrow cells) was detected, a second DLI was given and the patient developed GVHD grade 1. This was associated with eradication of MRD followed by long-term remission.
