Granulocyte/macrophage colony-stimulating factor (GM-CSF) promotes the survival, proliferation, and differentiation of myeloid lineage cells and regulates chemotaxis and adhesion. However, mice in which the genes encoding GM-CSF (Gmcsf) or the β common subunit of the GM-CSF receptor (βc) are inactivated display normal steady-state hematopoiesis. Here, we show that host GM-CSF signaling strongly modulates the ability of donor hematopoietic cells to radioprotect lethally irradiated mice. Although bone marrow mononuclear cells efficiently rescue Gmcsf mutant recipients, fetal liver cells and Sca1+ lin−/dim marrow cells are markedly impaired. This defect is partially attributable to accessory cells that are more prevalent in bone marrow. In contrast, Gmcsf-deficient hematopoietic stem cells demonstrate normal proliferative potentials. Short-term survival is also impaired in irradiated βc mutant recipients transplanted with fetal liver or bone marrow. These data demonstrate a nonredundant function of GM-CSF in radioprotection by donor hematopoietic cells that may prove relevant in clinical transplantation.
Hematopoietic stem cell (HSC) transplantation (HSCT) is a front-line treatment for many hematologic disorders. Most conditioning regimens administer myeloablative doses of radiation and/or chemotherapy; however, how the response of the host microenvironment influences donor cell repopulation remains poorly understood. Transplantation protocols utilize bone marrow, mobilized peripheral blood stem cells (PBSCs), and umbilical cord blood as sources of HSCs, and intrinsic variations in HSCs derived from different sources have been reported previously (1). In clinical practice, the source of HSCs modulates the duration of posttransplant cytopenia with cytokine-mobilized PBSCs inducing the most rapid recovery and umbilical cord blood cells inducing the slowest (2, 3).
GM-CSF promotes the proliferation and differentiation of myeloid progenitors and their progeny (4). Although GM-CSF has not been directly implicated in HSC engraftment, treatment with recombinant GM-CSF accelerates myeloid recovery in patients undergoing autologous marrow transplantation (5). Moreover, some patients with posttransplant graft failure show hematopoietic recovery after GM-CSF treatment (6). The mechanisms underlying these effects are unknown. Primitive hematopoietic cells express low to undetectable levels of the GM-CSF receptor, suggesting that GM-CSF does not act directly on HSCs, but rather on lineage-committed cells (7). The GM-CSF, IL-3, and IL-5 receptors share a common β subunit (βc) that associates with unique α chains to mediate biological responses to these cytokines (8).
Homozygous Gmcsf and βc mutant mice (Gmcsf−/− and βc−/−) maintain normal blood counts, and bone marrow from these animals restores hematopoiesis in irradiated WT recipients (9–12). The primary abnormality found in both strains is protein accumulation in the pulmonary alveoli due to defective macrophage function (9–12). We previously observed poor survival of irradiated Gmcsf−/− mice that were transplanted with fetal liver cells doubly mutant at the Gmcsf and Nf1 loci (13). Here, we show that donor accessory cells and host GM-CSF signaling strongly modulate radioprotection. These data, which define the first essential role for GM-CSF in hematopoiesis, may prove relevant for enhancing HSCT.
Results And Discussion
Defective radioprotection of
Gmcsf−/− mice by fetal liver cells
Consistent with previous observations, 5 × 105 Gmcsf−/− bone marrow cells efficiently repopulated irradiated WT hosts (9, 12). However, Gmcsf−/− recipients that received either Gmcsf−/− or WT bone marrow showed a modest reduction in survival (Figs. 1 A and Fig. S1). In contrast, Gmcsf−/− mice that were transplanted with fetal liver cells demonstrated markedly reduced survival (Fig. 1 A). In addition, WT mice that received Gmcsf−/− fetal liver cells had reduced survival compared with the recipients of WT cells (Fig. 1 B). Recipient mice succumbed 10–30 d after adoptive transfer with signs of hematopoietic failure (Fig. 1 B). Various numbers of WT fetal liver cells were injected to observe if the cell dose influenced survival rates. All WT mice that received ≥5 × 105 donor cells survived, whereas transferring up to 2 × 106 of the same cells did not rescue additional Gmcsf−/− animals (Fig. 1 C). These data reveal defective radioprotection of irradiated Gmcsf−/− mice, which is highly dependent on whether donor cells are derived from fetal liver or bone marrow. We also observed an independent effect of donor fetal liver cell genotype on the survival of WT recipients.
Gmcsf−/− fetal liver cells show equivalent repopulating potentials
Next, we performed competitive repopulation experiments to assess the short- and long-term repopulating abilities of Gmcsf−/− and WT hematopoietic cells in recipients of either genotype. Gmcsf−/− and WT bone marrow cells contributed equally to hematopoiesis in irradiated recipients (Fig. 2 A). Because WT fetal liver cells failed to efficiently radioprotect Gmcsf−/− recipients (Fig. 1), bone marrow–derived competitors were used to assess the repopulating potential of fetal liver test cells. Gmcsf−/− fetal liver cells produced similar levels of chimerism as WT cells in recipients assessed 1 and 4 mo after transfer (Fig. 2 B). To exclude the possibility that GM-CSF production by competitor cells might mask an intrinsic defect, we crossed the Gmcsf−/− mutation onto the B6.BoyJ background and used CD45.1+ bone marrow cells from these mice as a source of competitors. The repopulating potentials of Gmcsf−/− and WT fetal liver cells were equivalent in the absence of GM-CSF production by either competitor cells or irradiated recipients (Fig. 2 C).
Impaired survival of
Gmcsf−/− recipients injected with bone marrow–derived Sca1+ lin−/dim cells
The normal short- and long-term repopulating potentials of Gmcsf−/− fetal liver cells suggested that the failure to radioprotect Gmcsf−/− recipients might be due to accessory cells that are present in bone marrow but are deficient in fetal liver. The Sca1+ lin−/dim fraction, which is enriched for hematopoietic cells with high repopulating potential, comprised a similar percentage of cells in Gmcsf−/− and WT mice (unpublished data). Sca1+ lin−/dim bone marrow cells were isolated from WT adult mice and injected into irradiated WT and Gmcsf−/− recipients (Fig. 3 A). Although 12% of WT recipients that were transplanted with 1,000 Sca1+ lin−/dim cells survived, increasing the number of cells injected to 3,000 or more cells rescued 65–70% of irradiated mice (Fig. 3 B and Fig. S2). In contrast, only 3 out of 32 Gmcsf−/− mice that were injected in parallel with 3,000–20,000 of the same Sca1+ lin−/dim donor cells survived for 1 mo (P < 0.0005).
Irradiated accessory cells and exogenous GM-CSF enhance the survival of
Because the percentage of cells expressing the Mac1 integrin complex was much lower in fetal liver versus bone marrow (∼5 versus ∼45%; unpublished data), we asked if accessory function provided by bone marrow–derived Mac1+ cells contributes to radioprotection. To test this hypothesis, hematopoietic subpopulations were isolated from Gmcsf−/− bone marrow by cell sorting, irradiated with 1,800 cGy, mixed with fetal liver cells, and transferred into irradiated recipients. Table I and Fig. S3 display the results of two independent experiments. Consistent with our previous data, the same WT donor fetal liver cells rescued 8 out of 9 WT recipients, but only 1 out of 11 Gmcsf−/− mice. Irradiated Sca1− Lin+Gmcsf−/− bone marrow cells did not rescue irradiated hosts (unpublished data). However, cotransplanting 2 × 105 or 106 irradiated Sca1− Lin+Gmcsf−/− cells with 2 × 106 WT fetal liver cells produced durable engraftment in 8 out of 11 Gmcsf−/− recipients. Furthermore, coinjecting 106 irradiated Mac1+ cells derived from Gmcsf−/− bone marrow with fetal liver cells also rescued hematopoiesis in all 10 Gmcsf−/− recipients (Table I). These studies identify an accessory function within the differentiated fraction of bone marrow cells that facilitates the ability of fetal liver cells to radioprotect Gmcsf−/− mice.
We also asked if irradiated Gmcsf−/− accessory cells could restore the ability of bone marrow–derived Sca1+ lin−/dim cells to radioprotect Gmcsf−/− recipients. In these studies, irradiated Gmcsf−/− bone marrow cells (5 × 106 per mouse) were coinjected with 7,500 WT or Gmcsf−/− Sca1+ lin−/dim cells. Gmcsf−/− recipients that received Sca1+ lin−/dim cells of either genotype died with graft failure within 14 d of transfer (n = 10). In contrast, 6 out of 10 animals that received Sca1+ lin−/dim cells with irradiated bone marrow cells survived for >3 mo (Fig. S2), including 3 out of 5 transplanted with Gmcsf−/− Sca1+ lin−/dim cells. These data confirm the existence of accessory cells in murine bone marrow that cooperate with Sca1+ lin−/dim donor cells to radioprotect irradiated Gmcsf−/− mice. Furthermore, this accessory cell population achieves rescue through a mechanism that appears to be independent of both proliferation and GM-CSF production.
We administered a supernatant prepared from B16 melanoma cells that had been engineered to produce murine GM-CSF (500 μl of supernatant containing ∼73 ng/ml of GM-CSF) or a control B16 supernatant to ask if exogenous GM-CSF could improve the survival of Gmcsf−/− recipients transplanted with fetal liver cells. All mice were injected intraperitoneally on the day before adoptive transfer, the day they were irradiated and transplanted with 106 WT fetal liver cells, and on the day after transfer. In the first experiment, the mice were treated daily for the next week. In the other, a single additional injection of B16/GM-CSF or control parental B16 supernatant was administered 1 wk after adoptive transfer. In these two experiments, only one out of eight Gmcsf−/− recipients that were injected with the control supernatant survived for 14 d after adoptive transfer, whereas six out of eight animals that received B16/GM-CSF supernatant were alive after 1 mo.
Defective survival of irradiated β
Our studies of Gmcsf−/− mice suggested that signaling through the βc chain in recipient stromal cells modulates the ability of donor cells to rescue short-term hematopoiesis. To further investigate this idea, we transferred WT bone marrow or fetal liver cells (5 × 105 per recipient) into irradiated βc−/− or control mice. βc−/− mice that received WT fetal liver cells showed impaired survival that was similar to Gmcsf−/− hosts (Fig. 4 and Fig. S4). However, this deficit was not restricted to donor fetal liver cells, but was also seen in recipients that were transplanted with WT bone marrow.
These studies identify a nonredundant role of GM-CSF signaling in radioprotection that is strongly modulated by donor accessory cells. The high rate of lethality in Gmcsf−/− or βc−/− recipients transplanted with fetal liver cells and the ability of exogenous GM-CSF to improve survival in Gmcsf−/− mice revealed a major influence of host genotype in radioprotection. Similarly, depleting accessory cells from donor bone marrow or adding these cells to fetal liver modulated the survival of Gmcsf−/− recipients, demonstrating a strong and independent effect of accessory cells in facilitating engraftment. Although we examined short-term radioprotection as the primary endpoint, Kaplan-Meier survival analysis of recipients observed beyond 1 mo gave similar results (Figs. S1–S4). Together, our data indicate a complex interaction between host GM-CSF signaling and donor accessory cells in radioprotection and underscore the importance of both factors in this phenotype.
Myelo-erythroid and common myeloid progenitors are responsible for radioprotection, whereas long-term reconstitution requires engraftment by HSCs (14). However, myelo-erythroid progenitors, common myeloid progenitors, and HSCs do not express the GM-CSF receptor (15) and βc mutant cells efficiently repopulate wild-type recipients (10, 11). These data infer that GM-CSF facilitates radioprotection by a mechanism that does not require direct effects on repopulating donor cells. GM-CSF is concentrated in the bone marrow microenvironment where it is membrane anchored or immobilized within the extracellular matrix (16), which allows it to signal through receptors on target cells in a spatially localized manner. Irradiation induces GM-CSF production by cultured bone marrow stromal cells (17) and by stromal and endothelial cell lines (18). Furthermore, whereas GM-CSF mRNA levels are up-regulated within the bone marrow and spleen of mice 2 d after irradiation, there is no increase in serum levels (19). These data support the idea that GM-CSF is produced by stromal cells in response to radiation and acts locally. Although it is not known how GM-CSF signaling in recipient mice enhances radioprotection, GM-CSF has antiapoptotic effects in some cell types (20, 21). Alternatively, the observation that Mac1+ cells from Gmcsf−/− mice demonstrate defective phagocytosis of apoptotic cells in vivo (22) suggests that impaired clearance of dead or dying cells within the microenvironment might contribute to the radiation sensitivity of Gmcsf and βc mutant mice.
Our finding that donor accessory cells facilitate hematopoietic recovery after irradiation is consistent with a report that examined engraftment of nonobese diabetic/severe combined immunodeficiency (NOD/SCID) mice by human Lin−CD34+CD38− cord blood cells (23). Furthermore, CD26 expression on donor hematopoietic cells reduces engraftment efficiency and it has been suggested that accessory cells partially rescue this phenotype (24). Accessory cells may also modulate the effects of other molecules such as CXCR4 on homing and engraftment. In studies using fetal liver donor cells, we found that a relevant accessory population is contained within the Mac1+ fraction. However, we have not formally excluded the possibility that other cell fractions contribute to radioprotection. The ability of Gmcsf−/− bone marrow–derived accessory cells to improve the survival of Gmcsf−/− mice transplanted with fetal liver or with bone marrow–derived Sca1+ lin−/dim cells is consistent with the observation that these recipients are efficiently rescued by Gmcsf−/− bone marrow. However, it is intriguing that this accessory activity is radioresistant, particularly because the same populations are present in recipient bone marrow. The apparent paradox might be explained by differential effects of in vivo versus ex vivo radiation on accessory function, or may be modulated by the number of accessory cells injected. Consistent with the latter possibility, WT donor fetal liver cells provided better radioprotection than Gmcsf−/− fetal liver cells but had equivalent long-term repopulating potential. These data suggest that GM-CSF production by donor accessory cells becomes important at reduced cell numbers. Limit dilution experiments comparing WT versus Gmcsf−/− accessory cells are required to definitively address this issue. Potential mechanisms through which irradiated donor accessory cells might facilitate radioprotection include guiding cells with short- and long-term repopulating potential to appropriate niches, facilitating adherence, and/or clearing apoptotic cells and debris from the irradiated host microenvironment.
Our data raise the possibility that a relative deficit in accessory cell function contributes to delayed engraftment in patients who are transplanted with umbilical cord blood (2, 3). Recombinant GM-CSF has not proven superior to G-CSF in recipients of mobilized PBSCs and is no longer used in most transplant centers because of a higher incidence of side effects. However, investigating how GM-CSF and other growth factors modulate the responses of host bone marrow microenvironment to irradiation and how cytokines act upon accessory cells that are infused could uncover strategies for enhancing the safety and efficacy of HSCT. Furthermore, manipulating donor cell populations ex vivo to isolate cells with high repopulating potential could paradoxically increase the risk of graft failure if critical accessory cell populations are eliminated during processing.
Materials And Methods
Gmcsf and βc mutant mice (9, 10) were backcrossed six generations onto the C57BL/6 strain (CD45.2+). WT C57BL/6 mice (CD45.2+) and congenic B6.SJL-PtrcaPep3b/BoyJ (B6.BoyJ) mice (CD45.1+) were purchased from The Jackson Laboratory. The University of California San Franscisco Committee on Animal Research approved the experimental procedures.
Pregnant WT, Gmcsf−/−, and βc−/− females were killed by CO2 inhalation at E14 and fetal liver cells were prepared as described previously (13). Bone marrow cells were collected by flushing tibias with IMDM supplemented with 20% FCS (Hyclone Laboratories).
Adoptive transfer and competitive repopulation.
Recipients were irradiated with a single fraction of 850 cGy, which was uniformly lethal in the absence of donor cells. Fetal liver or bone marrow cells were injected into the dorsal tail vein of 10–12-wk-old recipients after irradiation. Short-term engraftment was defined as survival for 30 d after adoptive transfer, and radioprotection was defined as survival beyond the 12–18 d critical window of bone marrow failure. Competitive repopulation experiments were performed essentially as described by coinjecting CD45.2+ test cells with CD45.1+ competitors (25). Recipients received prophylactic oral antibiotics for 3 wk after irradiation.
Blood leukocytes were analyzed for chimerism using antibodies to CD45.1-PE and CD45.2-FITC. In some experiments, cells were counterstained with the lineage specific antibodies CD3-TC, B220-TC, Gr1-PE, and Mac1-PE (BD Biosciences). CD45 chimerism analysis was performed by flow cytometry with 10,000 events collected using a FACScan (Becton Dickinson) and analyzed using CELLQuest software. For the Sca1+ lin−/dim sort, cells were stained with the following mix of antibodies: Sca-1–FITC, B220-PE, CD3-PE, Mac1-PE, TER119-PE, and Gr-1–PE. The cells were labeled with anti-FITC MACS beads to select for Sca1+ cells on an autoMACS instrument and the Sca-1–enriched product was sorted using a BD FACS Vantage SE cell sorter. A rectangular sorting gate was drawn around the FITC+ PE−/dim population and these cells were collected for injection (Fig. 3 A).
Online supplemental material.
We are indebted to A. McMillan for assistance with statistical analysis; to R. Murray for providing βc mutant mice; and to F. Appelbaum, G. Begley, H. Geiger, A. Leavitt, C. Lowell, and G. Spangrude for insightful comments.
This work was supported by National Institutes of Health grant nos. CA72614, CA74886, CA66996, and CA092625; by the American Cancer Society (ACS) grant no. LBC-RSG-96-104-06; and by the Peterson Family Foundation. T.R. Katsumoto was a Howard Hughes Medical Institute Medical Student Research Fellow and J. Duda was an ACS Postdoctoral Scholar.
The authors have no conflicting financial interests.
J. Duda's present address is Stanford University Medical Center, Stanford, CA 94305.