The receptor tyrosine kinase c-kit provides a critical signal for survival, expansion, and maturation of mouse natural killer cells

Mice with a null mutation in the γ_cchain10 were from the fourth generation backcross to the C57Bl/6 background. RAG2 mice31 were from the 10^th generation backcross to C57Bl/6 (gift of B. Rocha, INSERM U345, Paris, France). Doubly deficient mice (RAG2/γ_c^−/−) were obtained by intercrossing as described.30 We also used breeding stocks of WB-W/+ mice32(gift of Dr H.-R. Rodewald, Basel, Switzerland). Mice were intercrossed to generate the null allele W/W(c-kit^−/−) and the control (c-kit^+/+ or c-kit ^±, hereafter referred to as wt) day 13 embryos. The morning of the vaginal plug discovery was designated as day 0. Fetal liver cell suspensions were obtained by passage of the tissue through a 23-gauge needle, and the genotypes were determined by fluorescence-activated cell sorter (FACS) analysis using an anti–c-kit monoclonal antibody (mAb). RAG2/γ_c^−/− andRAG2^−/− mice (more than 6 weeks of age) were irradiated with 0.3 Gy (gray) from a cobalt source, and 4 hours later they were injected intravenously with 4-8 × 10⁵ FL cells as a source of HSCs. All mice received tetracycline and bactrim in the drinking water for the period following the FL cell transfer.

Single cell suspensions were prepared from spleen, thymus, liver, and BM cells. Erythrocytes were lysed in ammonium chloride, and the cells were resuspended in phosphate-buffered saline (PBS) with 3% fetal calf serum (FCS) and 0.01% sodium azide. The following mAbs (PharMingen, San Diego, CA) were directly conjugated to fluorescein isothiocyanate (FITC), phycoerythrin (PE), Tricolor (TRI), or biotin (biot) and were used for immunofluorescence analysis: CD117 (c-kit), CD122 (IL-2Rβ), CD3, CD4, CD8, TCRαβ, TCRγδ, CD24 (HSA), NK1.1, B220, CD90 (thymine [Thy-1]), DX5, CD2, Ly49A, Ly49C/I, Ly49D, CD43, immunoglobulin M (IgM), and IgD. Biotin conjugates were revealed by streptavidin-TRI (Caltag, San Francisco, CA). Cells (10⁶) were first incubated with anti–fragment(Fc)γRII/III (hybridoma 2.4G2) and then stained with a mixture of biotinylated and fluorochrome-labeled mAbs at saturating concentrations, washed twice, and finally incubated with streptavidin-TRI. Analysis was performed on a fluorescence-activated cell sorter (FACS) flow cytometer (CellQuest software; Becton Dickinson, San Josè, CA). Dead cells were excuded by means of their forward and side scatters, and an electronic gate was set to acquire 5-10 × 10³ lymphoid cells.

A chromium 51 (⁵¹Cr) release assay was used to measure NK lytic activity in vitro, as previously described.21 For natural cytotoxicity, mice were injected intraperitoneally with 200 μg of poly (I:C), and 40 hours later, to deplete macrophages, red cell–depleted splenocytes were incubated on plastic for 1 hour at 37°C. YAC-1 and EL-4 thymoma cells were maintained in complete medium (CM; RPMI-1640 with 10% FCS, 10^-5 mol/L β-ME, 100 μg/mL streptomycin, and 100 units/mL penicillin). Target cells were labeled with 3.7 MBq (100 μCi) ⁵¹Cr (ICN Pharmaceutical, Costa Mesa, CA), and 2.5-5 × 10³cells were incubated with graded numbers of effector cells in 200 units/L (μl) of medium for 4 hours. In some experiments, effector cells were CD122⁺DX5⁺ NK cells isolated from splenocytes by cell sorting.

To generate adherent–lymphokine activated killer (A-LAK) cells, red cell and macrophage-depleted splenocytes were cultured at 5 × 106 cells/mL in CM supplemented with 1000 units/mL of dihydrouridine IL-2 (huIL-2) (Peprotech, Rocky Hill, NJ) or with 0.5 μg of huIL-15 (Immunex, R&D Systems, Minneapolis, MN). After 3-4 days the nonadherent cells were removed, and the A-LAK cells were refed every 2-3 days and cultured until days 8-10. A-LAK cultures produced in this manner routinely contained more than 80% NK1.1⁺CD3⁻cells. For antibody-dependent cell cytotoxicity (ADCC), day 8-10 A-LAK cells were used as effectors, and target cells were EL-4 cells coated with anti-Thy 1.1 mAb, as previously described.21 Radioactivity released into the cell-free supernatant was measured, and the percentage of specific lysis was calculated as the following: 100 × (experimental release − spontaneous release) / (maximum release − spontaneous release).

Day 13 FL cells prepared from normal C57Bl/6 timed pregnancies were plated at 5 × 10³ cells/mL in round-bottom microtitre plates in CM supplemented with 50 ng/mL muIL-7 (Peprotech), 100 ng/mL of huflk2L/flt3L (Immunex), and 100 ng/mL rat SCF (Peprotech). After 8 days, the cytokines were removed, and the cells were cultured with CM containing huIL-2 (1000 units/mL) for an additional 8-10 days.

Purified splenic NK1.1+ c-kit+ and NK1.1+ c-kit− cells were isolated from RAG2^−/−mice by cell sorting. Stringent sorting gates were applied to avoid cross-contamination of the 2 sorted subsets. Sorted cells (greater than 97% pure) were plated at 10⁴ cells/well in round-bottom microtitre plates in 200 μL of CM supplemented with graded doses of huIL-2, in the presence or absence of 100 ng/mL muSCF. Cell viability was evaluated daily by trypan blue dye exclusion, and after 6 days of culture, cells were analyzed for cell-surface phenotype, cell cycle distribution, and cytotoxic activity. Cell- cycle analysis was performed using 7-aminoactinomycin-D (7 AAD) incorporation into saponin-permeabilized cells as described.32 

Data were analyzed (Statview software, Version 4.5, SAS Institute, San Francisco, CA) applying the paired Student ttest. The null hypothesis was rejected, and the difference was assumed significant when P < .05.

We have recently developed a novel mouse strain harboring mutations in both the RAG2 and common cytokine receptorγ_c genes.30,RAG2/γ_c^−/−mice are alymphoid and have severely reduced numbers of BM and thymic lymphoid precursors, thereby making them ideal hosts for genetic complementation experiments analyzing lymphopoiesis. In order to define the role of the c-kit receptor tyrosine kinase in NK cell development, we reconstituted 0.3 Gy irradiatedRAG2/γ_c^−/− mice with c-kit–deficient HSC precursors. W/W mice harbor a splice donor site mutation that causes deletion of the transmembrane region, thereby completely lacking cell-surface expression of c-kit molecules. The absence of c-kit results in severe anemia in these mice and death by day 10 after birth.27 In our experiments, c-kit–deficient embryos were identified by their extreme pallor, and day 13 FL cells were used as a source of HSCs. The genotypes were confirmed by flow cytometry using an anti–c-kit mAb (data not shown). While HSCs express c-kit, previous studies have shown that the content of HSCs in day 13 FL from SCF-deficient mice is no different from control mice.34 We therefore transferred equal numbers of FL cells from W/W or wt day 13 embryos, and 8 weeks posttransfer, analyzed thymic, splenic, and BM lymphoid populations in the RAG2/γ_c chimeras.

Consistent with previous reports,33 we found that B-cell development was permissive in the absence of c-kit (Figures1A-C). No obvious defects in BM B-cell development were observed, and the absolute numbers of pre-B (B220⁺IgM⁻) and mature B cells (B220⁺IgM⁺) were not significantly different between wt(more than RAG2/γ_c) and W/W (more thanRAG2/γ_c) chimeras (Table1, Figure 1B). In contrast, we found a 10- to 20-fold reduction in total thymocyte numbers in the absence of c-kit (Table 1). The diminished thymopoiesis after transfer of c-kit–deficient FL cells likely reflects the important role for c-kit/SCF interactions in the expansion of early thymic double-negative precursors, as described by Rodewald et al.35 Once past this critical stage, intrathymic development in the absence of c-kit appeared normal, and all mature thymocyte subsets could be identified in W/W chimeras, including single positive CD4 and CD8 αβ T cells, γδ T cells, and NK T cells (Figure 1A; data not shown). We confirmed previous results of Takeda et al33 demonstrating that c-kit–deficient FL cells fail to reconstitute the thymus of irradiatedRAG2^−/− hosts (data not shown). The ability of W/W HSCs to repopulateRAG2/γ_c^−/− but notRAG2^−/− hosts can be explained by the severe reduction in both thymic and BM lymphoid precursors inRAG2/γ_c^−/−mice30 (Table 1).RAG2/γ_c^−/− mice thereby offer a less competitive host microenvironment thanRAG2^−/− mice, where mutant (eg,c-kit^−/−) HSCs can further differentiate. These results attest to the utility of theRAG2/γ_c strain for reconstitution experiments and the analysis of genes affecting HSCs and their immediate progeny.

IgM+/IgD+ B cells and CD4+ and CD8+ αβ T cells seed the peripheral lymphoid organs in W/W chimeras more thanRAG2/γ_c chimeras (Figure 1C; data not shown). Total splenic lymphocyte numbers were slightly reduced in the absence of c-kit, and this decrease was largely confined to the T-cell compartment. Thus, splenic CD3⁺ cells were 6.5 ± 1.8 × 10⁶ in wt (n = 7) and 4.6 ± 0.7 × 10⁶ in W/W chimeras (n = 7; P < .05) (Figure 1C). However, compared to the dramatic decrease in thymopoiesis (5%-10% of normal), peripheral T-cell numbers were about 70% of controls, suggesting that peripheral T-cell expansion had occurred. Together these results suggest that T lymphocytes do not require continual c-kit/SCF interactions for peripheral homeostasis.

Since all NK cells in RAG2/γ_c chimeras are donor-derived,30,36 any abnormalities in NK cell differentiation in W/W chimeras result from cell intrinsic defects due to lack of c-kit signaling. The ability of W/W FL cells to fully replenish total BM lymphoid cellularity and BM and splenic B-cell compartments suggests that HSC function in the absence of c-kit was not a limiting factor.34 NK cells were detected in the BM and spleens of W/W FL chimeras (Figure2A and B), but absolute numbers of BM NK cells were only 40% of normal in the absence of c-kit. The difference was found to be statistically significant (P < .05). Thus, c-kit/SCF interactions are dispensable for NK lineage commitment but are required for normal NK cell differentiation (Table 1, Figure 2A). Total numbers of splenic NK cells were similarly decreased inW/W chimeras (Figure 2B), which suggests that either peripheral expansion is not a mechanism to regulate NK cell homeostasis or that this process normally relies on c-kit signaling pathways. In agreement with previous reports,1 we found that about 10% of peripheral NK1.1+ CD3-NK cells from wt chimeras expressed c-kit, while W/W NK cells were c-kit−, as expected (Figure 2C).

In light of the existence of a bipotent NK/T thymic precursor, the development of thymic NK cells and NK-T cells in the absence of c-kit was also evaluated. Both of these minor thymic lymphoid subsets could be detected in W/W chimeras, although their absolute numbers were clearly reduced compared to control chimeras (NK:wt = 5.95 ± 3.0 × 10⁴ cells versusW/W = 0.5 ± 0.3 × 10⁴ cells; NK-T:wt = 15.7 ± 5.6 × 10⁴ cells versusW/W = 3.4 ± 1.7 × 10⁴ cells; n = 4 for each genotype; P < .05). The reduction in absolute cell numbers of thymic NK and NK-T cells paralleled the overall reduction in thymic cellularity. These results suggest that thymic T, NK, and NK-T cells develop via a c-kit–dependent early thymic precursor, and the results argue against a major pathway of NK cells generated from a putative NK/T bipotent precursor.

The phenotypic maturation of splenic NK cells from wt andW/W chimeras was characterized by flow cytometry using a panel of cell-surface markers (Figure 3). NK cells that were c-kit–deficient expressed normal levels of CD2, IL-2Rβ, NK1.1, DX5, and CD90 (Thy-1). Percentages and levels of Ly49 family members (including Ly49A, Ly49C/I, and Ly49D) were normal inW/W chimeras, which suggests that NK cell “education” with respect to inhibitory receptors for MHC class I molecules does not critically involve c-kit. Interestingly, c-kit–deficient NK cells showed markedly reduced expression of the B220 antigen (Figure 3). Similar results were obtained during analysis of BM NK cells (data not shown). As B220 expression increases on NK cells upon activation,16 we further compared the functional capacity of wt and c-kit–deficient W/W NK cells ex vivo.

Previous studies had shown that unfractionated splenocyte populations from mice with a viable c-kit mutation (W/W^v) had a decreased level of natural killing.28 However, since absolute NK cell numbers were not determined in those studies, it was not possible to precisely define the nature of the functional defect. We found that total spleen cell preparations from W/W chimeras demonstrated less than one-third of the natural killing lytic activity as compared with wtsplenocytes at equivalent effector:target ratios (n = 5; data not shown), confirming the initial data by Seaman and Talal.28In order to determine the lytic ability ofc-kit^−/−NK cells on a per-cell basis, we purified wt and W/W splenic NK cells by cell sorting. As shown in Figure 4, c-kit–deficient W/W NK cells showed less than half of the lytic capacity of wt NK cells against YAC-1 targets on a per-cell basis, thus confirming an intrinsic defect in natural killing of NK cells in the absence of c-kit.

Combined signaling through c-kit and IL-2/IL-15 receptors has been shown to synergistically enhance NK cell expansion and lytic activity both in vitro and in vivo.8 We therefore tested whether exposure of c-kit–deficient NK cells to IL-2 or IL-15 in vitro (thereby generating A-LAKs) could correct their functional and phenotypic defects. In contrast to freshly isolated W/W NK cells, A-LAK cells from W/W chimeras lysed YAC-1 targets in a fashion indistinguishable from controls (Figure5A). A-LAK cells that were c-kit–deficient also mediated normal levels of ADCC. In parallel with the restoration of natural killing, A-LAK cells from W/W chimeras also upregulated B220 surface expression (Figure 5B). Similar results were obtained with IL-15 (data not shown). Taken together, these results suggest that the abnormal differentiation of c-kit–deficient NK cells in vivo can be corrected in vitro by signaling through the IL-2/IL-15 receptor.

Our observations in W/W chimeras suggest that c-kit/SCF interactions in vivo are essential for generation of normal numbers of NK cells and for full differentiation of this lymphoid subset. To gain more insight into the stages at which c-kit signaling operates in NK cell development, the 2-step culture system described by Williams and colleagues2 was used to compare the cytokine requirements for differentiating NK cells from wt andW/W day 13 FL cells. When cultured in a mixture of IL-7, SCF, and flk2L/flt3L followed by high-dose IL-2, W/W FL cells developed similar frequencies of IL2Rβ+ NK1.1− and IL2Rβ+ NK1.1+ cells as compared with wt FL cells (Table2, Figure 6). However, total cell numbers were 3- to 4-fold reduced in W/Wcultures before the addition of IL-2 (data not shown). At the end of the culture period, c-kit^−/−W/WNK cells were found to be less numerous than wt NK cells (Table2). These results confirm that SCF/c-kit interactions are not required for NK commitment, but they show that SCF provides an essential signal for the expansion of NK precursors, which possibly occurs during the acquisition of IL-15 responsiveness (up-regulation of IL2Rβ). NK cells generated from W/W FL cultures were as cytolytic aswt-derived NK cells (Table 2), which confirms that stimulation with high doses of IL-2 (or IL-15) in vitro can compensate for the lack of c-kit signaling and suggests that SCF may synergize with IL-15 in vivo to induce full lytic competence.

Evidence for direct effects of SCF on mature murine NK cells has not been provided to date. Previous studies1,21 have shown that a subset of mature murine and human NK cells express c-kit, and human purified c-kit+ NK cells have been reported to show a lower cytotoxic activity, which suggests that this subset is somewhat less mature.21 In line with these studies, we found that freshly isolated c-kit+ murine NK cells have lower cytotoxic activity (70%-80%) than the c-kit–NK subset (data not shown). Studies by Caligiuri and colleagues8 have demonstrated that in vivo injection of SCF can enhance IL-2–mediated expansion of NK cells, which presumably acted at the level of NK cell precursors.

To characterize the potential mechanisms by which c-kit/SCF interactions act during NK cell differentiation, we analyzed the ability of SCF to promote survival, proliferation, and cytolytic activity of ex vivo purified c-kit+ and c-kit− NK1.1⁺cells from the pool of splenic NK cells. When c-kit+ NK cells were incubated with high doses of huIL-2 (1000 units/mL), the addition of SCF resulted in a modest (1.5- to 2-fold) increase in cell yield (Figure 7, Table3), although the cell viability (greater than 90%) and the percentage of cycling cells (about 30%) of c-kit+ cells cultured in high doses of IL-2 were not significantly different in the presence or absence of SCF (Table 3). In addition, the maximal lytic activity was comparable between c-kit+ and c-kit− NK cells cultured under conditions of high IL-2 concentrations (Table 3). These results are consistent with the ability of high doses of IL-2 to stimulate normal lytic activity in W/W NK cells.

In contrast, when c-kit+ or c-kit− NK cells were cultured in a suboptimal concentration of IL-2 (100 units/mL), SCF had a clear effect on survival, cell cycle progression, and lytic activity of c-kit+ NK cells (Table 3). Survival in the presence of SCF was 70%, while in the absence of SCF, few c-kit+ (23%) or c-kit− (18%) NK cells were viable after the 6-day culture period. In the absence of SCF, fewer c-kit+ NK cells were cycling (11%), and more cells were apoptotic (12%), based on their DNA content. In contrast, a normal rate of cycling cells (25%) and apoptotic cells (4%) were detected in c-kit+ NK cells cultured in the presence of SCF. This results in the highest absolute cell yield (Table 3, Figure 7). Taken together, the results suggest that SCF can provide survival signals to c-kit+ mature NK cells.

A number of soluble factors that play important roles in the generation of NK cells from hematopoietic precursors have been identified.1 It is now generally accepted that IL-15 is a dominant cytokine for NK cell development. Triggering of the IL-2Rβ/γ_c receptor complex by IL-15 (or mimicking this signal using high doses of IL-2) is required in vitro for NK cell differentiation from HSCs, while mice lacking IL-15 receptor components are NK deficient.3,10-12,14-16,18 Still, IL-15 alone only poorly promotes in vitro NK cell development,2,3 and additional factors may intervene in this process. The ligands for the receptor tyrosine kinases c-kit and flk2/flt3 (SCF and flk2L/flt3L, respectively) are 2 growth factors that play a role in the differentiation of multiple hematopoietic cell lineages.23,37 In vitro studies have shown that both SCF and flk2L/flt3L can significantly enhance the expansion of NK cells from HSCs in combination with IL-2 or IL-15.3,4Williams et al2 and Yu et al5 have demonstrated that flk2L/flt3L induce expression of the IL-2Rβ chain in murine and human stem cell precursors. Together, these results suggest a 2-stage model for NK cell differentiation, requiring first, c-kit and flk2/flt3 signaling, and second, IL-15 receptor signaling.1 

The ability of either flk2L/flt3L or SCF to promote NK cell development in concert with IL-15 would be in agreement with presumed redundant roles for these 2 receptor tyrosine kinases in NK cell differentiation. However, while the in vitro data are compelling, little is known about the essential roles of these 2 pathways for NK development in vivo. Preliminary observations in mice with a null mutation in the flk2L/flt3L indicated an absence of NK1.1^highcells,26 although the nature of this defect has not yet been fully explored. The NK cell deficiency inflk2L/flt3L^−/−mice would suggest, however, that c-kit and flk2/flt3 signaling pathways are not entirely redundant in this regard. Seaman and Talal analyzed NK cells in viable (W/W^v) c-kit mutant mice and suggested that c-kit was involved in the acquisition of natural cytotoxicity.28However, these studies raised the question of whether c-kit was essential for NK cell development, since the W^vmutation retains significant kinase activity.29 We report here that c-kit^−/−W/W FL cells generate reduced numbers of NK cells in vitro and only partially repopulate the NK cell pool in alymphoid RAG2/γ_c hosts. In addition, the W/W NK cells generated in vivo show defective killing. These results provide clear evidence for a nonredundant role of c-kit in the in vivo differentiation of murine NK cells. Together, the available data support the notion of unique (and perhaps additive4) roles for c-kit and flk2/flt3 signaling in NK cell differentiation.

While the precise action of SCF on developing NK precursors remains to be defined, a major role for c-kit/SCF interactions in the induction of IL-15 responsiveness appears unlikely, as IL2Rβ expression was comparable on wt and W/W NK cells both in vivo and in vitro (Figures 3 and 6). Assuming that this event restricts cells to the NK cell lineage, we can conclude that c-kit/SCF interactions are dispensable for NK cell commitment. Consequently, flk2L/flt3L might act as the dominant factor in NK cell determination in vivo. In contrast, c-kit signaling is required for generating normal numbers of NK cells. We would propose that c-kit/SCF interactions augment absolute numbers of BM and splenic NK cells by acting as a survival factor or (less likely) by inducing proliferation at the level of NK cell precursors. In this way, triggering of flk2/flt3 and c-kit on early lymphoid progenitors would provide distinct yet synergistic signals for NK cell differentiation.

Previous studies from Fehniger and colleagues8 suggested that SCF could synergize with IL-2 to promote proliferation of mature c-kit+ human NK cell subsets, although it did not appear that SCF had a direct effect on NK cell proliferation. The same group showed that in vivo SCF administration in mice also resulted in higher yields of NK cells, but studies done on bulk splenic populations did not show any direct effect of SCF on NK cell survival or proliferation.8Our in vitro data demonstrate that under conditions where other growth factors may be limiting, SCF can provide an essential signal for survival and cell cycle progression of the c-kit+ cells among the pool of mature NK cells. These c-kit/SCF interactions appear to act in an anti-apoptotic fashion and provide protection from cell death in part through the up-regulation of Bcl-2 levels.38 Taken together, these studies support a role for c-kit/SCF interactions in the survival of a subset of mature NK cells and their precursors in both mice and man.

The reduction in peripheral NK cell function in the absence of c-kit is noteworthy. Both lytic activity of freshly isolatedc-kit^−/− W/W NK cells and their activation status (as evidenced by B220 expression) were reduced. These results would suggest that c-kit signals are required for the final maturation of NK cells, and they are in line with previous observations showing that human c-kit+ NK cells appear to be less mature.21 The functional defects inc-kit^−/−NK cells could be reversed by exposure to IL-2 in vitro, which suggests that this phenotype could be overcome by compensatory pathways. These results are consistent with previous reports which have demonstrated that IL-2 and SCF can act synergistically to enhance NK cell expansion and lytic activity both in vitro and in vivo.8 We would propose that SCF not only expands the number of NK cell precursors, but that it also acts in synergy with IL-15 to drive full maturation of NK cells in vivo.

In summary, c-kit/SCF interactions are essential for the normal expansion of NK cell precursors and for the full maturation and survival of mature NK cells. In view of a role for c-kit/SCF interactions in generation of NK cell mediated functions, we are currently investigating the role of c-kit–deficient NK cells during in vivo responses to infectious pathogens and during tumor rejection.

We would like to thank Dr Hans Reimer Rodewald (Basel, Switzerland) for discussions and for kindly providing WB−W/+ mice; Drs Ana Cumano (Paris, France) and Emma Colucci-Guyon (Paris, France) for critically reviewing the manuscript; and the anonymous reviewers for their helpful comments.