THE PREFERRED SOURCE OF stem cells for allogeneic transplantation has become a central question among the transplant community. In the early 1990s, it was found that granulocyte colony-stimulating factor (G-CSF)– or granulocyte-macrophage colony-stimulating factor (GM-CSF)–mobilized peripheral blood stem cells (PBSCs) led to speedier granulocyte and platelet recovery after autologous transplantation than seen with marrow; given the practical and economic benefits of more rapid recovery, use of PBSCs quickly became the community norm, despite the lack of randomized trials measuring the impact of PBSC use on survival in specific disease states. However, there was hesitation in applying this technology to the allogeneic setting, because unmodified growth factor-mobilized PBSC collections contain, on average, 1 log more T cells than a standard marrow collection and murine studies have demonstrated a close relationship between the number of T cells in a graft and the development of acute graft-versus-host disease (GVHD). In 1995, 3 pilot studies were published in a single issue of Blood, each demonstrating in a small number of patients that use of PBSCs for matched sibling transplantation resulted in the same rapid recovery seen in the autologous setting and, surprisingly, no dramatic increase in acute GVHD over that seen with allogeneic marrow.1-3These pilots were followed by several larger phase II studies that likewise demonstrated accelerated granulocyte and platelet recovery with PBSCs and no apparent increase in acute GVHD.4,5However, several of these studies have suggested an increase in chronic GVHD.6 A recent summary of registry data is consistent with these phase II studies and, in addition, suggests a survival advantage during the early posttransplant period with the use of PBSCs compared with marrow, particularly for high-risk patients.7 These data all come from nonrandomized studies, but several large randomized trials of PBSCs versus bone marrow for matched sibling transplantation are well underway and should be nearing completion soon.
The report by the Essen group in this issue of Blood is the first to suggest that use of PBSCs may influence relapse rates, at least in the case of chronic myelogenous leukemia (CML).8 In a nonrandomized comparison between 29 patients with CML transplanted using PBSCs and 62 recipients of bone marrow, the incidence of molecular and cytogenetic relapse posttransplant was far greater in the marrow recipients. At this stage in an introduction, it is customary for the author to tell the reader about the weaknesses of the cited study and, to be sure, there are some in the Essen report. The study involves only a limited number of patients, and they were not randomized according to stem cell source. Most significantly, there was a higher degree of HLA-mismatching among PBSC recipients, with 45% of them being mismatched for a single class I or II antigen with their family member donor compared with 18% of bone marrow recipients. Nonetheless, the data appear to be convincing. The magnitude of the difference in relapse rates is substantial (44% with bone marrow compared with 7% with PBSC) and is similar in recipients of HLA-identical and single antigen mismatched grafts. The difference persists after statistical analyses accounting for all known confounding factors. And finally, the findings make sense.
The investigators offer 2 explanations for these findings. The first explanation they offer is that their observations are “consistent with a stem cell competition effect by which a rapidly expanding normal progenitor cell compartment can inhibit or displace residual clonogenic leukemia cells” and cite as further evidence the finding of improved survival in unrelated donor transplants if high bone marrow cell doses are used. However, in the unrelated donor study they cite, the increased cell dose was associated with a decrease in nonrelapse mortality, but had no effect on relapse rates.9Furthermore, although it might be possible and, in fact, interesting to study the question, there are few, if any, convincing animal models supportive of a nonimmunologically based stem cell competition effect capable of eradicating established leukemia. It is much more likely that the important observation of the Essen group is yet another example of the potent effect of donor T cells against CML. Although the investigators argue that neither acute nor chronic GHVD had a significant influence on residual molecular or cytogenetic disease in the study, the data would argue otherwise. The incidence of molecular relapse was more than twice as high (38%) in those without acute GVHD as those with (15%), and a similar magnitude of difference was seen with and without chronic GVHD. That the P values were not significant in the statistical analysis (P = .2 for acute GVHD and P = .052 for chronic GVHD) speaks more to the size of the study than a new biologic principle. Further, a graft-versus-leukemia effect can be operative without clinically evident GVHD. As the clearest example, CML relapse rates are much higher in recipients of T-depleted transplants than in recipients of nonmodified allogeneic marrow who do not develop clinically evident GVHD.
The study raises 3 large questions (at least). First, why is there not more acute GVHD with the use of PBSCs containing so many more T cells than marrow? Although in murine models there is a dose-response relationship between the number of T cells infused and the incidence of acute GVHD, it is possible that, in the clinic, once more than approximately 1 × 105 CD3+ cells/kg are transplanted, a critical threshold has been exceeded and further increases do not necessarily translate into more acute GVHD. It is also possible that the faster engraftment achieved with PBSCs may decrease the incidence and severity of infection, which, in turn, may have a protective effect on GVHD. Prior studies in mice and humans have shown that transplantation in protected environments can diminish the incidence of GVHD.10,11 Several studies have shown that G-CSF treatment may cause a shift in the population of T cells in the periphery towards CD4+ Th2 cells, a change expected to diminish acute GVHD.12,13 Finally, it has recently been demonstrated that administration of G-CSF favors mobilization of type II dendritic cells, which, in turn, should favor the development of Th2 T cells.14 If this last explanation is correct, then a study of the continued administration of G-CSF in the posttransplant period to further diminish the incidence of acute GVHD would seem to be warranted.
A second major question again raised by this study concerns why CML is so susceptible to an allogeneic reaction. A better understanding of this question would suggest where else to expect an advantage of PBSCs over bone marrow and, in those cases in which an effect is less apparent, possibly how to create it. The power of an allogeneic effect in CML may relate, in part, to the tumor’s growth rate. The number of leukemic cells eliminated by an allogeneic reaction is likely to be rate-limited and may be outstripped by a fast growing tumor. In fact, donor lymphocyte infusions have been more successful in chronic phase CML than in accelerated phase or blast crisis and more successful in other slower growing hematologic malignancies (chronic lymphocytic leukemia, multiple myeloma, and myelodysplasia) than in very rapidly growing ones.15 Other aspects of the target cell may determine the impact of an allogeneic effect. CML cells constitutively express high levels of class I antigens as opposed to acute lymphoblastic leukemia (ALL) blasts, a setting in which donor lymphocyte infusions are less effective. There may be other molecules expressed on the cell surface or secreted into the immediate microenvironment around CML cells (interleukin-12?) that favor an immunologic effect; and the reverse may also be true, ie, there may be molecules that suppress such an effect in those diseases in which an allogeneic impact is less apparent. The anatomic location of the tumor may also be important. CML is largely limited to blood, marrow, and spleen and rarely involves immunologically privileged sites. The spleen is an ideal place for donor T cells to encounter antigen. It would be of interest to know if the success of donor lymphocyte infusions is diminished in CML patients who have had a prior splenectomy.
The final question is whether the findings presented here by Elmaagacli et al8 should change our standard approach to transplantation. The answer, I would argue, is not yet. It is still unknown if the benefits of PBSCs, including more rapid engraftment, perhaps a diminution in early nonrelapse mortality (at least in high-risk patients), and now a potential reduction in posttransplant relapse (at least in CML), are worth the price of more chronic GVHD. There are, after all, relatively effective ways of dealing with molecular or cytogenetic posttransplant recurrence of CML, including the addition of α-interferon, withdrawal of immunosuppression, and use of donor lymphocyte infusions, and chronic GVHD can be a difficult disease. The results of the large randomized studies being performed will help better define this balance. There will also almost certainly be improvements in the PBSC product. In the studies cited above, no effort was made to limit the total number of T cells in the collections, and it may be that some limitation could preserve the improved antileukemic effect without an increase in chronic GVHD. The long-term goals of the allogeneic transplant community include providing a better definition of the populations of stem cells and immune effector cells contained in the transplant product and manipulating those cells to provide faster engraftment, more protection against infections and disease recurrence, and less GVHD. Recent advances with the use of growth factor-mobilized stem cell products have helped move this goal from the periphery to center stage.
Address reprint requests to Frederick R. Appelbaum, MD, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave N, D5-310, PO Box 19024, Seattle, WA 98104; e-mail: firstname.lastname@example.org.