Neutrophil depletion impairs natural killer cell maturation, function, and homeostasis

We conducted an N-ethyl N-nitrosourea (ENU) mutagenesis program in mice, using a functional test to screen and identify genes involved in NK cell responsiveness. We selected a new mouse pedigree, Genista, with hyporesponsive NK cells (Fig. 1, A and B). This hyporeactivity in vitro was revealed by testing blood NK cell degranulation (i.e., surface exposure of CD107a) after stimulation with YAC-1 target cells (Fig. 1, A and B), and was accompanied by an impaired capacity to reject MHC-I–deficient splenocytes in vivo (Fig. 1 C). We used Genista splenocytes to analyze NK cell functions in more detail. The fraction of spleen NK cells responding to stimulation with YAC-1 cells was lower than that in WT mice, as demonstrated by determinations of the percentages of CD107a⁺ and IFN-γ⁺ NK cells, confirming our initial observations on blood (Fig. 1 D). Genista NK cells also responded less strongly than WT cells to stimulation of the NK1.1-, NKp46-, or NKG2D- and Ly49D-activating receptors (Fig. 1 E and not depicted). This hyporeactivity was not caused by down-regulation of the surface expression of these receptors (Fig. 1 F and not depicted). We then used phorbol 12-myristate 13-acetate (PMA) and ionomycin stimulation, which induces NK cell activation by bypassing cell surface receptor engagement. In these conditions, the vast majority of NK cells from Genista and WT mice responded similarly, both in terms of the percentage of responding NK cells and in terms of the ability of IFN-γ production per cell (Fig. 1 G). Thus, NK cell hyporeactivity of Genista mice was not caused by a permanent inability to degranulate or to produce cytokines.

We further dissected the mechanisms involved in the NK cell functional defect in Genista mice by determining whether the hyporesponsive phenotype was NK cell intrinsic or extrinsic. We transferred purified spleen NK cells from CD45.1⁺ WT donors into CD45.2⁺ Genista recipients or CD45.2⁺ WT recipients as a control and analyzed the reactivity of CD45.1⁺ WT donor cells 7 d after adoptive transfer. By stimulating NK1.1 and NKp46, we showed that WT NK cells transferred into Genista mice became hyporesponsive, displaying weaker responses than WT NK cells transferred into WT recipients (Fig. 1 H). The exposure of spleen WT NK cells to a Genista environment thus modified their responsiveness, demonstrating the involvement of an NK cell–extrinsic factor inducing NK cell hyporeactivity in Genista mice. In control, when stimulated by PMA and ionomycin, WT and Genista NK cells showed similar responsiveness (Fig. 1 H). It has recently been reported that splenic WT NK cells become hyporeactive when transferred into a MHC-I–deficient environment (Elliott et al., 2010; Joncker et al., 2010). We therefore assessed the expression of MHC class I molecules on the cell surface in Genista mice and found no difference between these mice and WT mice (unpublished data).

In studies performed in parallel to our NK cell–oriented screen, Genista mice were found to lack mature neutrophils (unpublished data). Analysis of the neutrophil compartment in the blood, spleen, and BM (Fig. 2 A), as well as in the liver, lungs, and lymph nodes (not depicted) showed that mature CD11b⁺Ly6G^high neutrophils were selectively absent from Genista mice. The NK cell hyporesponsive phenotype and the neutropenia were perfectly correlated in the colony of F2 animals obtained from the cross of Genista and WT mice. Genetic analysis identified a point mutation leading to an amino-acid substitution in the third zinc finger of the growth factor–independent-1 (Gfi-1) transcription factor in Genista mice (unpublished data). Gfi-1 has already been implicated in neutrophil development, as patients with mutations in GFI-1 and Gfi-1 KO mice are severely neutropenic (Karsunky et al., 2002; Zarebski et al., 2008). As previously observed in Gfi-1 KO mice (Karsunky et al., 2002), Genista mice display an accumulation of atypical myeloid precursors (Ly6G^low, Ly6C^high, CD11b⁺) in the BM, but we did not detect any major modification in the percentages of monocytes at the periphery (Fig. 2, A and B). The dissection of the monocyte compartment in the spleen of Genista mice showed normal numbers of inflammatory (CD115⁺, CD11b⁺, Ly6C⁺) and resident (CD115⁺, CD11b⁺, Ly6C⁻) monocytes as compared with WT (Fig. 2 D). In addition, percentages and numbers of DC subpopulations (conventional, plasmacytoid and CD8α) were comparable between WT and Genista in the spleen as well as in the cutaneous lymph nodes (Fig. 2, C and D; and not depicted). As Gfi-1 has also been described as a critical regulator of DC versus macrophage differentiation (Rathinam et al., 2005), we sought to test the ability of BM cells from Genista mice to differentiate into DCs or macrophages in vitro. After 7 d in culture with M-CSF, Genista BM cells normally differentiated into BM-derived macrophages (BMMs) as judged by the up-regulation of F4/80 and CD11b (Fig. 2 E). The overnight stimulation with LPS induced a comparable up-regulation of the co-stimulatory molecules CD80, CD69, and CD86 at the surface of BMMs from WT and Genista (Fig. 2 F). In vitro DC differentiation assays lead to the generation of normal MHC-II⁺ CD11c⁺ BM-derived DCs (BMDCs) from WT and Genista BM cells (Fig. 2 G). Overnight activation with LPS showed a comparable up-regulation of CD80, CD69, CD86, and MHC-II on WT and Genista-derived BMDCs (Fig. 2 H). Collectively, these data show that, beside the lack of neutrophils, Genista mice do not display any major defect in NK cell accessory cells.

Gfi-1 is expressed in mouse NK cells, and might thus play a role in NK cell function. To address this question, we transferred purified spleen NK cells from Genista mice in WT recipients and monitored NK cell reactivity 7 d after transfer. In a WT environment, Genista NK cells became as responsive as host WT NK cells (Fig. 3 A), showing that the Gfi-1 mutation in NK cells was not sufficient to affect their functional capacities at steady-state. We thus focused on the extrinsic factors present or absent in Genista mice that led to NK cell hyporesponsiveness. With the exception of their NK cell phenotype, the only apparent defect in Genista mice was neutropenia. In WT animals, immunohistological analysis of spleen and lymph node sections showed that the two cell types were localized in close proximity in the red pulp of the spleen or in the medulla next to the lymphatic vessels of the lymph nodes (Fig. 4 A). Furthermore, in vitro experiments assessing the ability of NK cells to form conjugates with neutrophils revealed the strong propensity of these two cell types to interact, as compared with the lack of conjugates detected between neutrophils and T or B cells (Fig. 4 B). We thus directly tested the hypothesis that neutrophils represented the NK cell extrinsic factor required to promote NK cell reactivity by using mAbs to deplete neutrophils from WT animals. As the use of neutrophil-depleting mAbs is still a matter of debate (Daley et al., 2008), we performed these depletion experiments using anti-Gr1 (RB6-8C5) or anti-Ly6G (1A8) mAb. 6 d after mAb injection, RB6-8C5 treatment resulted in the depletion of neutrophils, but also of some monocytes (Ly6C^highCD11b^low), whereas 1A8 treatment had induced the selective depletion of neutrophils, as previously reported (Daley et al., 2008; Fig. 5 A). Remarkably, the depletion of neutrophils by either treatment was sufficient to induce NK cell hyporeactivity in WT mice (Fig. 5 B). The intensity of the NK cell hyporesponsiveness in these conditions was lower than that in Genista mice, which is consistent with the lower severity of the neutropenia induced by mAb treatment than of that observed in Genista mice. Indeed, anti-Gr1 (RB6-8C5) or anti-Ly6G (1A8) mAb treatment induced neutrophil depletion in the blood and in the spleen, but not in the BM (Fig. 5 C and not depicted). A CD4 cell-depleting control mAb had no effect on NK cell function, indicating that the hyporeactivity induced by the injection of neutrophil-depleting mAb resulted directly from the lack of neutrophils, rather than from eventual side effects of the depleting antibody treatment (Fig. 5, D and E). Neutrophils are thus essential for NK cell function at steady state.

We next investigated the mechanisms by which neutrophils affect NK cell function by analyzing the effects of these cells on NK cell maturation, a process that is classically studied by monitoring the cell surface expression of CD27 and CD11b. Immature CD27⁻CD11b⁻ NK cells first acquire CD27 expression, to become CD27⁺CD11b⁻, and then acquire CD11b, yielding double-positive (DP) CD27⁺CD11b⁺ NK cells, which eventually lose their CD27 expression to become fully mature CD27⁻CD11b⁺ NK cells (Kim et al., 2002; Hayakawa and Smyth, 2006; Chiossone et al., 2009). An investigation of these markers showed that NK cell maturation was severely affected in Genista mice, with most NK cells blocked at the DP stage (Fig. 6 A). An accumulation of DP NK cells was also observed in WT animals depleted of neutrophils (Fig. 6 B). Neutrophils therefore appeared to be required for the final stages of NK cell maturation, corresponding to the transition from the DP stage to the most mature CD27⁻CD11b⁺ stage. We characterized these maturation stages further by studying CD43, which is up-regulated during the down-regulation of CD27 in fully mature WT NK cells (Kim et al., 2002; Yokoyama et al., 2004). The percentage of NK cells expressing CD43 was much lower in Genista mice than in WT mice (Fig. 6 C). These findings were confirmed in studies in which mAbs were used to deplete neutrophils in WT mice (Fig. 6 D). We interpreted the lower decrease in CD43 expression in WT mice depleted of neutrophils as compared with Genista mice as a consequence of the milder severity of the neutropenia induced by mAb treatment, as mentioned in the previous paragraph. Importantly, the percentage of CD43⁺ WT NK cells 7 d after transfer into Genista mice was also lower than that monitored after transfer into a WT recipient (Fig. 6 E). We investigated the relationship between this impaired maturation in the absence of neutrophils and the hyporesponsive phenotype of NK cells by comparing the responsiveness of CD43⁻ and CD43⁺ NK cells in WT mice. After the stimulation of NK1.1 and NKp46, both the percentage of IFN-γ⁺ NK cells and the mean fluorescence intensity of IFN-γ⁺ NK cells were lower in the CD43⁻ NK cell population than in the CD43⁺ NK cell population (Fig. 6, F–H). CD43-expressing NK cells thus responded more strongly than CD43⁻ NK cells in these conditions of stimulation. The lower percentage of CD43-expressing NK cells may thus account for the hyporeactivity of NK cells in neutropenic mice.

In addition to their defective maturation, NK cells proliferated more in the absence of neutrophils. Indeed, the percentage of Ki67⁺ NK cells in Genista was significantly higher than that in WT mice, whereas there was no difference in the T and B cell compartments (Fig. 7 A and not depicted). This phenotype was neutrophil-dependent, as neutrophil depletion in WT mice led to the specific proliferation of NK cells (Fig. 7 B). This conclusion was also supported by experiments involving the adoptive transfer of WT NK cells. Indeed, 7 d after the transfer of CFSE-labeled WT NK cells into Genista recipients, 80 ± 13.2% of the WT NK cells had proliferated, whereas no proliferation was observed after transfer into WT recipients (Fig. 7 C). Given this extensive proliferation of WT NK cells after transfer into Genista, we hypothesized that the hyporeactive NK cell phenotype might be displayed only by new NK cells generated by the division of donor cells. We tested this hypothesis by comparing the responsiveness of undivided (CFSE^high) WT NK cells after transfer into WT or Genista mice. We showed, by stimulation of NK1.1 and NKp46, that there was a change in NK cell responsiveness after transfer of WT cells in Genista mice, even for NK cells that did not divide (Fig. 7 D), ruling out the possibility that NK cell hyporeactivity in the absence of neutrophils was caused by NK cell proliferation alone. Nevertheless, NK cells that had divided (CFSE^diluted) had lower levels of activity than undivided (CFSE^high) cells (Fig. 7 D). NK cell proliferation in the absence of neutrophils was therefore associated with NK cell hyporeactivity, but did not fully account for the phenotype. The higher level of NK cell proliferation in Genista mice was not associated with an increase in the number of NK cells. Indeed, the overall percentages of NK cells in the blood and spleen were unchanged (blood, 3.71 ± 0.85% for WT and 3.05 ± 1.04% for Genista, P = 0.3417; spleen, 2.98% ± 0.87 for WT and 2.4% ± 1.01 for Genista, P = 0.3002), whereas it was significantly reduced in the BM of Genista mice as compared with WT mice (BM, 0.93 ± 0.35 for WT and 0.49 ± 0.33% for Genista, P = 0.0021), suggesting that NK cell proliferation was accompanied by a decrease in NK cell survival. Indeed, Genista NK cells displayed higher levels of cell death in vitro than WT NK cells, and this difference was not abolished by overnight treatment with IL-15 or IL-2 (Fig. 7 E). Thus, in mice lacking neutrophils, NK cells displayed hyperproliferation and poor survival and were blocked at an immature stage associated with hyporesponsiveness.

These data revealing a new role for neutrophils in NK cell function in the mouse prompted us to assess NK cell maturation and function in patients suffering from neutropenia. Patients with severe congenital neutropenia (SCN) have no terminally differentiated neutrophils because of the blockage of maturation at the promyelocyte stage in the BM (Donadieu et al., 2011; Klein, 2011). Mutations of the neutrophil ELANE gene, encoding the serine protease elastase, are the most common cause of SCN (Bellanné-Chantelot et al., 2004; Horwitz et al., 2007), but mutations targeting the gene involved in Wiskott-Aldrich syndrome (Ancliff et al., 2006; Devriendt et al., 2001), HCLS1-associated protein X-1 (Klein et al., 2007), G6PC3 (Boztug et al., 2009), or GFI-1 (Person et al., 2003) have also been described. We analyzed a cohort of 22 SCN patients, 11 of whom had a mutation in ELANE; the mutations in the other 11 patients were unidentified (Table 1). Some of them were treated with granulocyte colony-stimulating factor (G-CSF) to increase neutrophil counts. Blood NK cell counts were similar in SCN patients and healthy controls (Fig. 8 A), but using CD56 down-regulation as a marker of terminal NK cell maturation, the frequency of CD56^bright NK cells was higher in SCN patients than in controls, indicating that NK cells of SCN patients were less mature (Fig. 8 B). We assessed the possible functional impairment of NK cells from SCN patients, by monitoring the percentage of IFN-γ⁺ NK cells generated in response to stimulation with K562 tumor target cells or antibody-coated P815 cells, as a means to evaluate antibody-dependent cell-mediated cytotoxicity (ADCC). Under both types of stimulation, the percentage of responding NK cells was much lower in SCN patients than in healthy controls (Fig. 8, C and D). This hyporesponsive phenotype was confirmed by determining the percentage of lysed target cells (Fig. 8, E–F). NK cells from patients lacking neutrophils thus displayed a maturation and functional defect similar to that observed in neutropenic mice, demonstrating that the key role of neutrophils in NK cell maturation and function is conserved between humans and mice.

SCN is currently treated by G-CSF injections, which increase the rate of granulopoiesis and result in the mobilization of neutrophils to the periphery. Neutrophil count restoration is variable among patients and depends on the dose of G-CSF used (Donadieu et al., 2011). In some patients, these neutrophils have been shown to be only partially functional (Donini et al., 2007). Some studies have indicated a potential role of G-CSF in NK cell function (Miller et al., 1997). We therefore compared the reactivity of NK cells between SCN patients with and without G-CSF treatment. NK cell reactivity did not differ between G-CSF–treated and untreated patients (Table 1; and Fig. 8, C–F). An increase in the frequency of CD56^bright cells was also observed in SCN patients treated with G-CSF compared with controls (Fig. 8 B). Thus, G-CSF treatment is not sufficient to restore NK cell reactivity and maturation in SCN patients.

The BM is the main reservoir for neutrophils at steady state. We found that neutrophils were still present in the BM of neutrophil-depleted WT mice in which NK cells were hyporesponsive (Fig. 2 A; and Fig. 5, B and C). In this model, the NK cell hyporesponsiveness was milder than that in Genista mice, which lack neutrophils in both the BM and the periphery, suggesting a contribution from, but not a requirement for, BM neutrophils (Fig. 1, E and F; and Fig. 5 B). We therefore investigated whether BM neutrophils or neutrophils circulating at the periphery were required for NK cell function in humans. We addressed this question by analyzing a second cohort of patients with autoimmune neutropenia (AIN; Table 1). AIN patients have circulating antibodies targeting neutrophils, leading to a lack of mature neutrophils at the periphery. Unlike SCN patients, AIN patient display normal neutrophil development in the BM. In AIN patients, NK cell counts were not statistically different from the control individuals (Fig. 9 A), whereas NK cell maturation was affected with a trend toward an accumulation of immature CD56^bright NK cells (Fig. 9 B). Despite the high variability among individuals and the limited number of AIN patients, NK cells from these patients were also less reactive to antibody-coated cells, as shown by the percentage of IFN-γ⁺ cells (Fig. 9 C) or direct target cytotoxicity (Fig. 9 D). A down-regulation of NK cell reactivity in response to K562 target cells was also observed in blood samples from AIN patients (Fig. 9, E and F), but this trend was not statistically significant for IFN-γ secretion, probably because of the small size of our cohort of AIN patients. Thus, despite the presence of neutrophils in the BM, peripheral NK cells from AIN patients were hyporesponsive, like NK cells from SCN patients, although the observed defect was milder. Interestingly, one of the AIN patients recovered from neutropenia 8 mo after the first analysis. This increase in neutrophil counts was associated with the recovery of NK cell functions (Fig. 9, G and H). Altogether, these results indicate that neutrophils contribute to NK cell development and functions in humans.

We report here a new role for neutrophils as nonredundant regulatory cells ensuring the terminal maturation of NK cells and the acquisition of their full effector functions in steady-state conditions in both humans and mice.

Using a forward genetic approach we identified a mutant mouse, Genista, with hyporesponsive NK cells coupled to neutropenia. The hyporeactivity of Genista NK cells was observed in vivo against MHC-I–deficient splenocytes and in vitro upon stimulation with the tumor cells YAC-1 or after triggering of the NK1.1, NKp46, NKG2D, and Ly49 receptors. The surface expression of these receptors was unchanged, suggesting that a wide range of activating pathways coupled to various signaling adaptor molecules, such as FcRγ, DAP12, and DAP10, are affected in NK cells in the absence of neutrophils. The hyporesponsive phenotype was associated with a block in NK cell maturation before the up-regulation of cell surface CD43, a marker that accompanies NK cell responsiveness in WT mice.

In humans, we studied several subtypes of neutropenia with distinct clinical profiles and pathophysiology backgrounds. First, we analyzed SNC patients. 11 of them presented a mutation in the ELANE gene, whereas the genetic etiology was unidentified for the others. Like in the mouse model, NK cell maturation and responsiveness were affected in these patients. These results first established the conservation of the major role of neutrophils on NK cell functions across species. In addition, they also indicated that the mechanism of NK cell/neutrophil cellular cooperation was independent of the gene involved in neutropenia (Gfi-1 in the mouse, ELANE in humans). The role of neutrophils in NK cell biology was further confirmed by the depletion of neutrophils with antibodies in WT mice and by studying AIN patients. In these patients, AIN is transient and is related to the presence of antibodies directed against epitopes of neutrophil membranes: NA1a, b, and c in most cases (Bux and Stroncek, 2002). In conclusion, whatever the cause of the neutropenia is, the phenotype of NK cell hyporesponsiveness is present both in humans and in mice.

Importantly, one of the AIN patients of our cohort spontaneously resolved his neutropenia and NK cell functions were restored in parallel, reinforcing the intimate link between the two cell subsets. In contrast, no restoration of NK cell reactivity was detected in patients responding to G-CSF treatment by an increase in neutrophil count (above the threshold of 1,500 cells/µl). Similarly, 1 wk of G-CSF injections in Genista mice mobilized immature, Ly6G⁻ neutrophils to the periphery, but did not restore NK cell function (unpublished data). Thus, neutrophils that are induced by G-CSF treatment do not acquire the capacity to potentiate NK cell function, prompting further dissection of the differences between these neutrophil populations and normal neutrophils. Along this line, it was shown in some patients that G-CSF treatment reverses neutropenia without correcting all the functional deficiencies of neutrophils (Donini et al., 2007). This issue remains to be revisited in larger cohorts of SCN patients with different genetic etiology.

NK cell homeostasis was also modified in mice lacking neutrophils. We observed an increased sensitivity to cell death ex vivo and an increased proliferation of NK cells in vivo. As we did not observe any accumulation of NK cells in neutropenic mice, it could indicate that NK cell proliferation is counterbalanced in vivo by NK cell death. NK cell proliferation induced in neutropenic mice differed from NK cell homeostatic proliferation occurring in a lymphopenic environment, which is associated with higher NK cell reactivity (Sun et al., 2011). We have shown in transfer experiments of WT NK cells in neutropenic recipients that the acquisition of the hyporesponsive NK cell phenotype did not require proliferation, even if hyporesponsiveness was increased in proliferating cells. It is possible that NK cells and neutrophils make use of the same trophic factors and that the absence of neutrophils modifies the concentration of these molecules, with an impact on the NK cell compartment. However, we do not favor this hypothesis because a large number of neutrophils remained present in the BM when peripheral neutrophils were depleted in mAb-treated WT animals and in AIN patients. In the two species, the presence of BM neutrophils was not sufficient to promote NK cell functions even if, in these conditions, the NK cell defect was milder. These findings support a model in which neutrophils are mandatory for NK cell maturation from the CD43⁻ CD27⁺CD11b⁺ toward the CD43⁺ CD27⁻CD11b⁺ stage. A similar block in NK cell maturation has been previously reported in T-bet KO mice (Townsend et al., 2004; Soderquest et al., 2011), but we did not detect any change in T-bet or eomesodermin expression by intracellular staining on Genista NK cells as compared with WT (unpublished data). The lack of neutrophils may induce an increased turnover of NK cells that cannot reach a mature functional state. Neutrophil-induced NK cell maturation may occur not only in the BM, in which NK cells develop, but also at the periphery, suggesting a constant requirement for neutrophils, to promote NK cell function. Along this line, we also showed that spleen NK cells from neutropenic Genista mice reacquire a functional phenotype 7 d after transfer in a WT environment.

To further dissect the mechanisms at work in this cellular cooperation, we described direct NK cells/neutrophils interactions in vitro and in vivo. The two cell subsets are located in the same compartments in lymphoid organs (spleen and LN) of uninfected mice and have a high propensity to form conjugates in vitro. Yet, the short-term co-culture of purified WT neutrophils with Genista NK cells was not sufficient to restore NK cell function (unpublished data). In addition, we were not able to restore NK cell functions in Genista mice by injecting purified WT BM neutrophils. However, despite repeated injections (every day for 6 d), the number of neutrophils in recipient mice remained well below that observed in WT mice (unpublished data). Both in vivo and in vitro, these experimental conditions are limited by the fragility and very short half-life of neutrophils. Therefore, these negative results do not rule out a direct role of neutrophils on NK cells. Alternatively, neutrophils might cooperate with a third cell type to potentiate NK cell activity. Along this line, cross talk between human neutrophils and NK cells has been reported after stimulation with LPS and IL-2 or IL-15/IL-18 (Costantini et al., 2011). This crosstalk involves 6-sulfo LacNAc⁺ DCs (slanDC), through both direct cell–cell interactions dependent on CD18, ICAM-1, and ICAM-3, and indirect interactions mediated by cell-derived cytokines, such as IL-12 and IFN-γ (Costantini et al., 2011). It is impossible to address the role of slanDC in vivo, as no counterparts of these cells have been identified in the mouse (Costantini et al., 2011). Interestingly, we previously found that CD18 on NK cells was required for NK cell maturation and function (Crozat et al., 2011). CD18 associates with CD11a, CD11b, and CD11c proteins to generate functional integrin receptors (Luo and Springer, 2006). CD11/CD18 integrins bind to several receptors, including ICAM-1, ICAM-2, ICAM-3, inactivated C3b, and fibrinogen (Luo and Springer, 2006). We investigated the involvement of ICAM-1 in the acquisition of NK cell functions. In ICAM-1–deficient mice, NK cell functions, assessed by determining CD107a exposure and IFN-γ production upon stimulation with YAC-1 target, NK1.1, and NKp46, was found to be similar to that in WT controls (unpublished data). Moreover, NK cell maturation was not affected in ICAM-1 KO mice, as shown by the expression of CD43, CD11b, and CD27 markers (unpublished data). Thus, ICAM-1 is not required for neutrophil-induced NK cell maturation, indicating that if CD18 is involved in this process in vivo, it can act through interactions with other ligands. The mechanisms at work in steady-state conditions may therefore at least partially differ from the NK-neutrophil-slanDC “ménage à trois” described as being initiated by neutrophil activation with human cells in vitro (Costantini et al., 2011).

Neutrophils have been described as critical activators of NK cells in mice, acting in an IL-18/MyD88-dependent manner against Legionella pneumophila infection (Spörri et al., 2008), supporting our data on the implication of neutrophils in NK cell reactivity. Yet, our present findings extend the role of these cells to steady-state conditions in humans and mice. In addition, the mechanisms involved during the course of this microbial infection are different from those at work at steady state, as we previously showed that NK cells in MyD88 KO mice display no defect in IFN-γ secretion upon stimulation of NK1.1, NKp46, or YAC-1 (Chaix et al., 2008), in contrast to the situation in mice lacking neutrophils.

Altogether, our data thus rule out the requirement of signaling via IL-18–dependent and other MyD88-dependent receptors in neutrophil-induced NK cell maturation at steady state.

It has been reported recently that a subset of Ly6C⁻ myeloid-derived suppressor cells that develop during chronic inflammation associated to tumors was capable of impairing NK cell development and function (Elkabets et al., 2010). We observed an increased number of Ly6C⁻ atypical myeloid precursors in the BM of Genista mice (Fig. 2, A and B). We do not know if these cells are related to those appearing in such pathological conditions, but we cannot exclude that they could contribute in the induction of the NK cell phenotype in Genista mice. However, such an accumulation is not observed in the BM of WT mice depleted of neutrophils, excluding their major role in the function on NK cells that we report here.

NK cells cooperation with monocytes/macrophages and DCs is well documented (Dalbeth et al., 2004; Baratin et al., 2005; Welte et al., 2006; Lucas et al., 2007; Nedvetzki et al., 2007; Tu et al., 2008; Bellora et al., 2010; Soderquest et al., 2011). Our present study reveals a novel aspect of the role of neutrophils on NK cells. The possibility of a reciprocal cross talk between the two cell types could be envisaged. Along this line, it has been reported that IFN-γ regulates survival of neutrophils in the context of Mycobacterium tuberculosis infections (Nandi and Behar, 2011). As NK cells are important producers of IFN-γ, it would be interesting to analyze neutrophils in NK-deficient mice both in steady-state and inflammatory conditions.

Finally, high incidence of myelodysplasia and acute myeloid leukemia has been reported in neutropenic patients (Donadieu et al., 2005; Rosenberg et al., 2006). Risk factors for myelodysplasia and acute myeloid leukemia include the severity of neutropenia and the exposure of high doses of G-CSF (Donadieu et al., 2005). In light of the data presented here, the defect in NK cell immunosurveillance in these patients might also impact on the development of these malignancies. Thus, our data not only reveal a new regulatory pathway in innate immunity, but also provide insight into the previously unappreciated role of NK cell functional deficiency in patients suffering from neutropenia-associated diseases. This should prompt a reanalysis of immune deficiencies involving neutrophils or NK cells to further unravel the role of the cross talk between these cells in immunity.

The ENU mutagenesis leading to the isolation of the Genista pedigree was performed in the C57BL/6J (Charles River) background (Georgel et al., 2008). C57BL/6J CD45.1 mice were purchased from Charles River. C57BL/6J, Genista, and NKp46^iCre/wtR26R^eYFP/wt (Narni-Mancinelli et al., 2011) mice were bred and maintained under specific pathogen–free conditions at the animal facility of CIML and the Centre d’Explorations Physio-pathologiques Avancées RIO platform in Marseilles. Experiments were conducted in accordance with institutional guidelines for animal care and use. Protocols were approved by the Direction Départementales des Services Vétérinaires des Bouches du Rhône.

The monoclonal antibodies used for flow cytometry and activation were: purified anti-NKp46 (29A1.4) conjugated with Alexa Fluor 647 or PE; anti-NK1.1 (PK136) conjugated with APC and PerCP-Cy5.5 and purified; anti-CD3 (145-2C11) conjugated with PE, FITC, PerCP-Cy5.5, and APC; anti-CD11b (M1/70) conjugated with V450; anti-CD27 (LG.3A10) conjugated with PE; anti-CD43 (S7) conjugated with PE or FITC; anti-CD45.1 (A20) conjugated with Pacific blue and Pe-Cy7; CD45.2 (104) conjugated with PerCP; anti–IFN-γ (XMG1.2) conjugated with APC and Alexa Fluor 647; anti-CD107a (1D4B) conjugated with FITC; Annexin V-FITC; anti-CD80 (16-10A1); anti-CD86 (GL1); anti-CD69 (H1.2F3); anti–mI-A/I-E (M5/114.15.2); and anti-CD19 (AD3). All antibodies were purchased from BD. Anti-Ki67 (20Raj1) antibody conjugated with PE and anti-CD45.2 (104) antibody conjugated with Alexa Fluor 700 were purchased from eBioscience. Anti-CD115 (AFS98) was purchased from BioLegend. Samples were analyzed with FACSCanto II, LSRII (BD), and FlowJo software (Tree Star).

Blood lymphocytes isolated with a Lympholyte gradient for mammalian cells (TEBU) or from spleen cell suspensions after red blood cell lysis were incubated with YAC-1 tumor targets or dispensed into a 96-well 2HB Immulon plate coated with purified antibody against NK1.1 (25 µg/ml), NKp46 (10 µg/ml), or isotype controls. Cells were activated in the presence of monensin (GolgiStop; BD), GolgiPlug (BD), and anti-CD107a antibody conjugated with FITC (where indicated) in complete medium (RPMI-1640 [Invitrogen] supplemented with 10% fetal calf serum, 1 mM sodium pyruvate, 10 mM Hepes, 100 U/ml penicillin, and 100 µg/ml streptomycin). As a positive control, we stimulated the cells with a mixture of 200 µg/ml PMA and 5 µg/ml ionomycin. The cells were incubated for 4 h at 37°C, and then cell surface staining was performed. For intracellular IFN-γ staining, cells were fixed with 2% paraformaldehyde (PFA) and permeabilized with Perm/Wash solution (BD).

This method for the quantitative assessment of in vivo killing was adapted from a method described in a previous study (Oberg et al., 2004). In brief, splenocytes from WT or β2mKO mice were labeled with 2 and 0.2 µM CFSE (Invitrogen), respectively, and mixed at a 1:1 ratio. The ratio of the various populations before injection was determined by FACS analysis and compared with the ratio in the spleen 48 h after injection.

Splenocyte suspensions from CD45.2⁺ WT or Genista donor mice were obtained by mechanical disruption of the spleen of the mouse, in complete medium, on a cell strainer with 70-µm pores (BD). Red blood cells were lysed in RBC lysis buffer (eBioscience). The preparation was then enriched in NK cells by negative depletion with mAbs against CD4, CD5, CD8, Ter119, and IA/IE, as previously described (Chiossone et al., 2009). NK cell purity was ∼85%. Donor NK cells were labeled with 1 µM CFSE (Invitrogen) and 2–3 × 10⁶ NK cells/mouse were injected i.v. into WT or Genista CD45.1⁺ recipient mice.

BM cells were flushed from tibiae and femurs of WT or Genista mice and cultured in complete RPMI medium supplemented with M-CSF at 10 ng/ml. Adherent BMMs were obtained after 7 d in culture. BMDCs were obtained and cultured, as previously described (Inaba et al., 1992). 7 d after culture of BMDCs or BMMs, 10 ng/ml of LPS was added to the culture and up-regulation of co-stimulatory molecule expression was monitored by FACS.

Neutrophil depletion was achieved by injecting 500 µg of 1A8/mouse (BioXCell), 100 µg of RB6-8C5/mouse (BioXCell) or isotype controls i.p. into mice on days 0 and 2. CD4⁺ T cells were depleted by two i.p. injections of 200 µg of GK1.5/mouse (BD) on days 0 and 1. The analysis was performed on day 6.

The BM was flushed from the tibia and femur of WT mice and separated on a 65% Percoll (GE Healthcare) gradient. The pellet was retained and the red blood cells were lysed in RBC lysis buffer (eBioscience). The resulting preparation contained >80% neutrophils, which were then stained with PKH26 (Sigma-Aldrich), according to the manufacturer’s protocol. Purified neutrophils were then incubated at a 1:1 ratio, with WT splenocytes previously stained with anti-NK1.1 and anti-CD3 antibodies. After 10 min at 37°C, cells were fixed with 1% paraformaldehyde and analyzed by FACS. The conjugates correspond to PKH26⁺ NK1.1⁺ CD3⁻ cell doublets.

Organs were harvested and fixed by incubation in 0.05 M phosphate buffer supplemented with 0.1 M l-lysine, pH 7.4, 2 mg/ml NaIO₄, and 10 mg/ml paraformaldehyde for 12 h. They were then washed in phosphate buffer and dehydrated by incubation in 30% sucrose in phosphate buffer. Spleens from NKp46^iCre/wtR26R^eYFP/wt mice were snap-frozen in Tissue-Tek (Sakura Finetek). We cut 20-µm frozen sections and stained them as previously described (Bajénoff et al., 2003). Anti-B220 (RA3-6B2) antibody conjugated with Alexa Fluor 647 (BD), anti-MMP9 rabbit polyclonal antibody (Abcam) detected with a Pacific blue–conjugated anti–rabbit antibody (Life Technologies), anti-Lyve1 rabbit polyclonal antibody (Acris GmbH) detected with A647-conjugated anti–rabbit antibody (Invitrogen), and an anti-GFP Alexa Fluor 488 antibody (Invitrogen) were used for staining. Immunofluorescence confocal microscopy was performed with a Leica SP5 confocal microscope. Separate images were collected for each fluorochrome and overlaid to obtain a multicolor image. Final image processing was performed with Imaris software (Bitplane) and Photoshop software (Adobe).

PBMCs were isolated by the centrifugation on Ficoll (EuroBio) gradients of whole blood samples obtained from healthy volunteer donors and from the SCN and AIN patients described in Table 1. Patients were recruited from departments of Pediatric Hematology of Trousseau Hospital (Paris), Necker Enfants Malades Hospital (Paris), and La Timone Hospital (Marseille) if they present a chronic neutropenia classified according to the literature (Donadieu et al., 2011). Patients with congenital neutropenia were included in the French Severe Chronic Neutropenia cohort, which has been reported elsewhere (Donadieu et al., 2005). The patients or their parents provided their written informed consent for genetic testing and inclusion in the register approved by the Commission Nationale de l’Informatique et des Libertés (number 01–1084). Genomic DNA was extracted from blood with standard procedures and ELANE, G6PC3, HAX1, WASP, and SBDS mutations were screened (Bellanné-Chantelot et al., 2004; Donadieu et al., 2011). In addition to patients with congenital neutropenia, AIN patients, diagnosed if autoantibodies against neutrophils were detected (Bux et al., 1998), were also included in the study.

PBMCs were incubated in the presence of GolgiStop (1/1500; BD) with MHC class I⁻ human erythroleukemic K562 target cells or P815 mouse mastocytoma cells coated with rabbit anti–mouse lymphocyte antibodies (BioValley). After incubation for 4 h at 37°C, cells were stained with anti-CD3 (SK7), PercPCy5.5-conjugated antibody (BD) and with anti-CD56 (NKH-1) APC-conjugated antibody (Immunotech). Intracellular IFN-γ staining was performed after fixation in 2% paraformaldehyde and permeabilization in Perm/Wash solution (BD). For NK cytotoxicity analysis, K562 or P815 cells incubated with rabbit anti–mouse lymphocyte antibodies were stained with 0.5 µM CFSE (Invitrogen) for 10 min, washed, and incubated with PBMCs at a 50:1 ratio. After 4 h at 37°C, cells were stained with 2.5 µg/ml propidium iodide (Invitrogen) and analyzed by flow cytometry.

All P values were determined with Prism software (GraphPad Software Inc.) for nonparametric unpaired or paired tests (two-tailed), as indicated. We considered p-values <0.05 to be significant and the degree of significance is indicated as follows: *, P < 0.05; **, P < 0.01; ***, P < 0.001.

We thank Nicolas Chevrier for experiments on ICAM-1 KO mice, and the CIML animal facility and flow cytometry facilities.

This work was supported by European Research Council advanced grant (E. Vivier and S. Ugolini); MUGEN Network of Excellence and MASTERSWITCH Integrating Project from European Communities (B. Malissen and M. Malissen); Fondation pour la Recherche Médicale and ARC (D. Ordoñez-Rueda); Agence Nationale de la Recherche (S. Ugolini); Ligue Nationale contre le Cancer (E. Vivier); Axa Research Fund (B.N. Jaeger); and institutional grants from Institut National de la Santé et de la Recherche Médicale, Centre National de la Recherche Scientifique and Université de la Méditerranée to the CIML.

E. Vivier is a cofounder and shareholder of Innate-Pharma. The other authors have no competing financial interests.

Author contributions: B.N. Jaeger, E. Vivier, and S. Ugolini designed and analyzed the experiments and wrote the paper. M. Malissen and B. Malissen initiated and conducted the early phases of the ENU mutagenesis screen, isolated the neutropenic Genista mice, and determined the corresponding genetic defect. D. Ordoñez-Rueda analyzed the neutropenic Genista phenotype; C. Bernat identified the NK cell functional defect present in Genista; B.N. Jaeger performed and analyzed the experiments on mice; C. Cognet performed and analyzed the experiments on human patients; M. Bajénoff performed immunostaining experiments; A. Fenis was responsible for mouse handling and breeding; E. Narni-Mancinelli participated in the experimental design; J. Donadieu, N. Mahlaoui, and V. Barlogis provided blood samples from neutropenic patients. J. Donadieu is the scientific secretary of the French SCN register. C. Bellanné-Chantelot was responsible of the human genetic analysis, B. Beaupain monitored the French SCN register.