Human NK Cells Induce Neutrophil Apoptosis via an NKp46- and Fas-Dependent Mechanism

Neutrophils are the most abundant leukocyte population in human peripheral blood and constitute an essential part of the innate immune system. During the early stages of tissue damage secondary to infection, neutrophils are rapidly recruited from the blood to the tissue where they play an important role in the inflammatory responses and in the clearance of microbes. Neutrophils ingest pathogens and kill them using oxygen radicals and an array of other toxic substances and enzymes that are stored in preformed granules in the cytosol (1, 2). These substances are toxic not only to pathogens but also to the surrounding tissue. Hence as much as the neutrophil inflammatory response is an invaluable asset in the immediate phase of a limited infection, massive or prolonged neutrophil accumulation and activation can result in extensive damage and inflammation in the tissue (3). The intracellular content of highly toxic substances in neutrophils calls for a timely and vigilant apoptotic program for neutrophils to ensure minimum leakage to the surrounding tissue and to enable resolution of inflammation (4). The presence of microbes and certain proinflammatory cytokines favors neutrophil survival, but less is known about factors that promote neutrophil apoptosis after microbial clearance (5, 6).

NK cells are large granular lymphocytes with the ability to kill aberrant cells without prior sensitization (7). NK cell function is pivotal for a favorable outcome in several forms of cancer and is crucial for the defense against viral infections (8, 9). In addition, in recent years, NK cells have been shown to display immunomodulatory functions (9, 10). Thus, NK cells release cytokines, such as IFN-γ and TNF-α, that induce dendritic cell (DC) maturation and promote Th1 responses (11, 12). The intimate interaction with DCs also leads to NK cell activation and acquisition of the capacity to home to lymph nodes, where they can further promote adaptive immune responses (13). NK-mediated immunomodulation is also conveyed via cytotoxicity (14). NK cells display cytotoxic activity against LPS-activated macrophages and immature DCs (14–18), whereas mature DCs are resistant to lysis. Resistance is related to the higher expression of HLA class I, which engages inhibitory receptors on NK cells (15).

In this study, we investigated whether NK cells could be involved in the modulation of neutrophil function. We show that NK cells can significantly accelerate neutrophil apoptosis, and that this process is dependent on cell–cell contact and mediated by the activating NK cell receptor NKp46 and the Fas pathway. Our results are suggestive of a novel aspect of NK cell-mediated immunoregulation, which may have implications for resolution of acute inflammation.

The following mAbs produced in our laboratory were used in this study: anti-CD56 (A6.220, IgM), anti-HLA class I (A6.136, IgM), anti-2B4 (CO54, IgM), anti-NTBA (MA127, IgG1), anti–DNAM-1 (F5, IgM), anti-NKp30 (F252, IgM), anti-NKp44 (KS38, IgM), anti-NKp46 (KL247, IgM), anti-NKp80 (Cer 1, IgM), anti-poliovirus receptor (M5A10, IgG1), anti–Nectin-2 (L14, IgG2a), and anti–MIC-A (BAM195, IgG1). Anti-ULBP1 (clone M295), anti-ULBP2 (clone M310), and anti-ULBP3 (clone M550) were obtained from Amgen (Seattle, WA). Anti-Fas ligand (anti-FasL; clone 10F2) and anti-Fas (clone ZB4) were purchased from AbD Serotec (Oxford, U.K.) and MBL International (Woburn, MA), respectively. Anti-CD18 (clone 7E4, IgG1) was obtained from Beckman Coulter (Fullerton, CA). PE-conjugated anti-CD107a (IgG1) was purchased from BD Biosciences, PE-conjugated isotype-specific goat anti-mouse secondary reagents were from Southern Biotechnology, and PE-conjugated F(ab′)₂ goat anti-human IgG was obtained from Jackson Immunoresearch Laboratories (West Grove, PA).

Buffy coats were obtained from healthy blood donors and mixed in a 1:1 ratio with 2% Dextran T500 (Pharmacosmos, Holbaek, Denmark). After sedimentation of RBCs, the upper phase was separated into neutrophils and mononuclear cells by density gradient centrifugation. Residual erythrocytes in the pellet were lysed in water to yield a pure population of granulocytes. Lymphocytes were obtained by countercurrent centrifugal elutriation of the mononuclear cells as described previously (19).

Pure populations of NK cells were obtained from PBMC or lymphocytes using the NK cell isolation kit (Miltenyi Biotec, Bergisch Gladbach, Germany), according to the manufacturer’s instruction. In some experiments, MACS CD15 Micro beads were added to further improve depletion of granulocytes. The purity of NK cells was >95% NK cells (defined as CD56⁺/CD3⁻). CD56-depleted lymphocytes were obtained using MACS CD56 Micro beads and a MACS LD column. The NK cell content of depleted lymphocytes was <1%.

Neutrophils were incubated with NK cells or CD56-depleted lymphocytes (37°C, 5% CO₂) and assayed for apoptosis after 4 h of coculture. If not otherwise stated, short-term activated allogeneic NK cells (1 ng/ml IL-12, 20 h) were used as effector cells in the assay. In some experiments, neutrophils were preincubated with caspase inhibitors benzyloxycarbonyl-Val-Ala-Asp (OMe) fluoromethylketone (z-VAD-fmk; Calbiochem) or benzyloxycarbonyl-Ile-Glu(OMe)-Thr-Asp(OMe) fluoromethylketone (z-IETD-fmk; BD Biosciences) for 1 h before coculture with NK cells. Neutralizing Abs against the Fas receptor (10 μg/ml) and its ligand (FasL; 10 μg/ml) were used in some experiments.

To determine the percentage of apoptotic cells, cells were stained with FITC-conjugated Annexin V (BD Biosciences) according to the manufacturer’s instructions with the addition of To-Pro-3 iodide (0.5 μM; Invitrogen, Eugene, OR). Neutrophil apoptosis was determined using a BD FACSCalibur or an Accuri C6 flow cytometer.

CD107a (LAMP-1) found in lytic granules can be detected on NK cell surface shortly after stimulation and degranulation (20). NK cells were incubated with target cells (ratio 1:1) in the presence of a PE-conjugated anti-human CD107a (IgG1; BD Biosciences). After 1 h, monensin (GolgiStop; BD Biosciences) was added. K562 cells were used as a positive control for NK cell degranulation.

Neutrophils were incubated in the presence or absence of short-term activated NK cells and GM-CSF (100 U/ml). After 4 h, cells were stained with a fluorescent caspase inhibitor (FLICA; Immunochemistry Technologies) (21), fixed, and FACS sorted directly onto poly-l-lysine–coated slides (Sigma-Aldrich). Cells were mounted in Prolong Gold Antifade with DAPI (Invitrogen, Carlsbad, CA). Photomicrographs were acquired with a Zeiss LSM700 confocal microscope (Jena, Germany) using a 63×/1.40 oil differential interference contrast objective.

To prepare the plasmids for the expression of dimeric NKp30-Fc and NKp46-Fc soluble receptors, we obtained the sequence coding for human IgG1 hinge region by annealing (4 min at 95°C, 10 min at room temperature) the following primers: 5′-ACGCGTCGACCCCAAATCTTGTGACAAAACTCACACATGCCCACCGTGCCCGGATCCCG-3′ (Hinge SalI/BamHI forward), 5′-CGGGATCCGGGCACGGTGGGCATGTGTGAGTTTTGTCACAAGATTTGGGGTCGACGCGT-3′ (Hinge SalI/BamHI reverse); the reaction product was digested with SalI/BamHI restriction enzymes and subcloned into the SalI/BamHI digested pRB1-DNAM1Fcmut vector in frame with the sequence coding for the mutagenized human IgG1 Fc portion (1).

The sequences coding for the extracellular portions of NKp30 and NKp46 were amplified starting from the corresponding cDNA inserted in the pcDNA3.1TOPO-V5 plasmid using the following primers: 5′-CAGGGCATCTCGAGTTTCCGACATGGCCTGGATGCTGTTG-3′ (NKp30 XhoI up) and 5′-CCGCTCGAGATGTGTACCAGCCCCTAGCTGAGGATG-3′ (NKp30 XhoI down); 5′-CAGGGCATCTCGAGTCTGAGCGATGTCTTCCACACTCC-3′ (NKp46 XhoI up) and 5′-CCGCTCGAGATGATTCTGGGCAGTGTGATCC-3′ (NKp46 XhoI down). Amplification was performed with Platinum Taq High Fidelity (Invitrogen, Carlsbad, CA) for 25 cycles (30 s at 95°C, 30 s at 60°C, and 1 min at 68°C). The PCR products were digested with XhoI restriction enzyme and subcloned into the SalI digested pRB1-HingeFcmut vector. The nucleotide sequences of the constructs were checked using a BigDye Terminator v3.1 Cycle Sequencing Kit and an ABI PRISM 3100 Genetic Analyzer (Applied Biosystems, Foster City, CA).

Soluble receptors were produced in HEK293T cell line (human embryonic fibroblasts) and purified from supernatants by affinity chromatography using nProtein A-Sepharose 4 Fast Flow (GE Healthcare). The purity of purified fusion proteins was checked by SDS-PAGE followed by silver staining and by ELISA using mAbs specific for the different receptors. NTB-AFc soluble receptor was prepared as described previously (22).

Neutrophils (1 × 10⁵ cells) were incubated with 2 μg NKp30Fc, NKp46Fc, or NTB-AFc soluble receptors for 30 min at 4°C, washed, and stained with PE-conjugated F(ab′)₂ goat anti-human IgG (Jackson Immunoresearch Laboratories) for 30 min at 4°C. FITC-conjugated Annexin V and To-Pro-3 iodide were added to exclude dead cells from the analysis. Flow cytometry was performed using a FACSCalibur flow cytometer (BD Biosciences), and cells were analyzed with CellQuest Pro software (BD Biosciences).

Skin blisters were induced on the volar side of the forearm of healthy volunteers by a suction chamber coupled to a portable vacuum pump for 2 h (23). At different time points, the blister fluid (∼20 μl/blister) was collected. Cells were stained and the immune cell content and viability were determined by flow cytometry. Informed written consent was obtained from all volunteers included in the study.

For multiple comparisons within a data set, one-way ANOVAs with Bonferroni’s post tests were performed. For single comparisons, paired sample t tests were used. All reported p values are two-sided: *p < 0.05, **p < 0.01, and ***p < 0.001.

Neutrophil apoptosis is a critical step in the termination of acute inflammatory responses. However, incomplete information is available on the proapoptotic signals leading to apoptosis (5, 6). Previous studies have addressed NK cell cytotoxicity against other innate immune cells, such as myeloid DCs and polarized macrophages (14, 16, 18, 24). However, little is known about the NK cell capacity to kill neutrophils. In pilot experiments, freshly isolated neutrophils were incubated with short-term IL-12–activated NK cells and assayed for apoptosis after 4 h of coculture. As shown in Fig. 1A, NK cells induced significant apoptosis in neutrophils (p < 0.0001; n = 18). In contrast with tumor cell lines that are highly susceptible to NK cell-mediated lysis, the majority of dying neutrophils remained intact (Annexin V⁺, To-Pro-3⁻; Fig. 1B).

We next evaluated whether also resting NK cells, both autologous and allogeneic, were able to induce neutrophil apoptosis. As shown in Fig. 1C and 1D, in both cases, resting NK cells triggered neutrophil apoptosis, even though allogeneic NK cells appeared to be slightly more efficient. The small difference between the effect mediated by autologous or allogeneic NK cells was likely due to the fact that polymorphonuclear neutrophils (PMN) display low expression of surface HLA class I molecules (Supplemental Fig. 1). Accordingly, blocking of HLA class I on neutrophils did not enhance apoptosis induction by resting NK cells (data not shown). To determine whether the ability to induce neutrophil apoptosis was confined to NK cells or whether it was a general capacity shared by other lymphocytes, we used a CD56-depleted lymphocyte population in apoptosis experiments. As shown in Fig. 1E, CD56-depleted lymphocytes failed to compromise neutrophil survival, suggesting that neutrophil apoptosis is specifically induced by NK cells.

Proinflammatory cytokines, such as GM-CSF, favor prolonged neutrophil survival (6), and supernatants recovered from stimulated NK cells were recently proposed to prolong neutrophil survival (25). To investigate whether the proapoptotic effect of NK cells could override survival signals mediated by cytokines, we exposed neutrophils to short-term activated NK cells in the presence or absence of exogenously added GM-CSF or IL-12, which may trigger cytokine release from NK cells. As expected, exogenously added GM-CSF reduced the rate of spontaneous neutrophil apoptosis (p < 0.05; Fig. 2A). However, NK cells induced significant apoptosis also in GM-CSF–protected neutrophils (p < 0.001). Presence of IL-12 during NK–neutrophil cocultures slightly enhanced NK cell-induced neutrophil apoptosis (p < 0.05, n = 10, data not shown).

In the next series of experiments, neutrophils were preincubated with the pan-caspase inhibitor z-VAD-fmk before coincubation with NK cells. As seen in Fig. 2B, apoptosis was reduced in the presence of z-VAD-fmk in a dose-dependent manner. Next, we used confocal microscopy to visualize the activation of caspases in neutrophils. Activation of caspases was visualized using a fluorescent inhibitor of caspases (FLICA) that was added to the cell suspension (21). This cell-permeable, carboxyfluorescein-labeled reagent binds to activated caspases and can be detected inside apoptotic cells. As shown in Fig. 2C, in the presence of NK cells, neutrophils displayed green cytosolic staining, indicating caspase activation and ongoing apoptosis.

In a series of experiments, we examined whether the NK cell-mediated apoptosis induction was dependent on soluble factors or whether cell–cell contact was needed. Conditioned media, obtained from NK cell or NK–neutrophil cocultures, were collected and added to freshly isolated neutrophils. As shown in Fig. 2D, none of these conditioned media contained soluble mediators capable of inducing apoptosis as determined after 4 h of incubation. These results suggest that NK cell-induced neutrophil apoptosis is not mediated through soluble factors, and that cell–cell contact is important for apoptosis induction.

NK cell cytotoxicity is regulated by the balance between inhibitory and activating signals emanating from interactions between target cell ligands and NK cell receptors. In a series of experiments, we characterized the expression of NK cell receptor ligands on the neutrophil cell surface. As shown in Fig. 3A, neutrophils displayed very low levels of ligands for the activating receptors 2B4 (CD48) and NKG2D (MICA, ULBP 1-3). DNAM-1 ligands, Nectin-2 and poliovirus receptor, were expressed at higher levels, although some variation was observed among different individuals (Fig. 3A).

To assess the cell surface expression of ligands to the natural cytotoxicity receptors (NCRs), NKp46 and NKp30, on neutrophils, we determined the binding of NCRFc fusion proteins by flow cytometry. As shown in Fig. 3B, neutrophils displayed weak but significant binding of a soluble NKp46 receptor. Also, NKp30Fc stained neutrophils but to a lower extent, whereas no binding of NTB-AFc was observed, confirming previous reports stating that NTB-A ligands are specifically expressed in eosinophilic granulocytes (26). These data provide, to our knowledge, the first evidence that neutrophils constitutively express cell surface ligands for the NCRs NKp30 and NKp46.

Next, we investigated the role of different activating NK cell receptors for NK cell-mediated apoptosis induction in human neutrophils. As shown in Fig. 3C, masking of NKp46 with a specific mAb significantly protected neutrophils from NK cell-induced apoptosis, whereas Abs directed to other activating NK cell receptors failed to significantly protect neutrophils. The absence of a suitable IgM Ab against NKG2D precluded analysis of the role of NKG2D in NK cell-induced neutrophil apoptosis. These data suggest that a hitherto unidentified structure on neutrophils interacts with the activating NK cell receptor NKp46 to trigger NK cell activation and subsequently neutrophil apoptosis. Alternatively, NKp46 binding to neutrophil-encoded ligand may generate an apoptosis-promoting signal in neutrophils.

To address this possibility, we determined NK cell degranulation after coincubation with neutrophils. When NK cells become activated upon interaction with a target cell, lytic granules fuse with the plasma membrane to release perforin and granzymes to kill the encountered cell (27, 28). Because these granules are specialized secretory lysosomes, they contain lysosomal membrane proteins (e.g., CD107a) that can be detected on the cell surface after granule fusion (20, 29). As shown in Fig. 4A, short-term activated NK cells incubated with neutrophils displayed significant degranulation, albeit not of the same magnitude observed with the classical NK-susceptible target cells K562.

NK cells also use FasL to kill target cells; upon binding of FasL, the cytoplasmic part of Fas recruits death domain-containing adaptor proteins that interact with caspase-8 to form the death-inducing signaling complex (30). Within this complex, caspase-8 becomes active and promotes a series of events leading to apoptosis. In a next series of experiments, we examined whether caspase-8 was involved in NK cell-induced neutrophil apoptosis. As shown in Fig. 4B, preincubation with the caspase-8 inhibitor, z-IETD-fmk, significantly protected neutrophils from NK-induced apoptosis, suggesting a role for caspase-8 in NK cell-mediated induction of neutrophil apoptosis. However, granzymes released from NK cells share the substrate specificity with caspase-8. Thus, to discriminate between the granzyme and Fas pathways, we performed experiments with neutralizing Abs to Fas and FasL. As shown in Fig. 4C, blocking of FasL or Fas resulted in significantly decreased apoptosis of neutrophils (p < 0.001; n = 4), strongly suggesting a key role for the Fas pathway in NK cell-induced neutrophil apoptosis.

To investigate whether NK cell–neutrophil encounters may occur in inflammatory lesions in vivo, we studied cells from a human aseptic inflammatory in vivo model. Skin blisters were induced on the forearm of healthy volunteers by a suction chamber coupled to a portable vacuum pump (23). At different time points after blister formation, the blister fluid was collected and the presence of immune cells in the blister fluid was determined by flow cytometry. As shown in Fig. 5, neutrophils started to accumulate already after 2 h. At a later time point, NK cells (CD3⁻ NKp46⁺), together with other lymphocytes and monocytes, were detected in the blister fluid (Fig. 5A, 5B). Interestingly, the appearance of NK cells coincided with an increased percentage of apoptotic neutrophils in the blister fluid (Fig. 5C).

Neutrophils are the most abundant leukocyte in human peripheral blood, and a vast number of mature neutrophils are released from the bone marrow every day (31, 32). However, they are not long-lived, and in the absence of an inflammatory insult, they enter an apoptotic program. In contrast, when neutrophils are recruited to a site of inflammation, their life span is extended. At these sites, neutrophils recognize and phagocytose microbes and participate in the inflammatory response. As long as the threat remains, proinflammatory mediators, such as GM-CSF, IFN-γ, IL-1β, IL-3, IL-6, TNF-α, and LPS enhance neutrophil survival (6). Activated neutrophils produce chemokines that attract more neutrophils, as well as other immune cells, to the lesion (33). For example, NK cells express receptors for many of the mediators produced or generated by activated neutrophils, such as CXCL8 (IL-8), CCL3 (MIP-1α), and chemerin (33–36). The recruitment of more neutrophils is an important amplification loop in the immediate response to infection. Ingested pathogens are killed using oxygen radicals and an array of other toxic substances and enzymes that are stored in preformed granules in the neutrophil cytosol (37, 38). These substances are not only toxic to pathogens but also to the surrounding cells and tissue (39, 40). Thus, after microbial clearance, neutrophils should undergo apoptosis and subsequently be removed by tissue macrophages before their toxic content leaks out in the tissue (41). In this study, we show that NK cells trigger neutrophil apoptosis, suggesting that NK cells may play a role in limiting the inflammatory response and terminating acute inflammation. In line with this, in an in vivo model of acute inflammation, NK cell appearance in skin blister fluid coincided with elevated neutrophil apoptosis (Fig. 5C). At this time point, the NK/neutrophil ratios were significantly lower in blister fluid than in the in vitro experiments. However, a potential lethal encounter does not necessarily occur in the blister fluid; for example, in synovitis, neutrophils accumulate in the synovial fluid, whereas NK cells are mainly found in the synovial tissue (42). It is thus possible that the ratio is quite different at the tissue–fluid interface.

Recent studies suggest that cytokine-stimulated NK cells can produce IFN-γ and GM-CSF to activate neutrophils and enhance neutrophil survival (25, 43). Furthermore, in the study from the Cassatella group (43), blocking the physical interaction between NK cells and neutrophils using an Ab against CD18 enhanced the survival-promoting effect of NK cell-derived cytokines. Thus, just like interactions between NK cells and immature DCs or macrophages, which can either result in mutual activation or in the induction of NK-mediated killing of these innate cells (9, 14–16, 18, 44), NK–neutrophil interactions can have two opposite outcomes.

Because neutrophils die spontaneously in vitro, it could be argued that NK cells are only capable of “killing” neutrophils that are already dying. However, in our hands, NK cells could induce neutrophil apoptosis even in the presence of the antiapoptotic cytokine GM-CSF. In contrast, neutrophils were fully protected against NK cell-induced apoptosis if pretreated with the pan-caspase inhibitor z-VAD-fmk. Interestingly, neutrophils were strikingly resistant to NK cell-dependent lysis, and the majority of dying neutrophils remained unstained by To-Pro-3. Thus, traditional assays of NK-mediated cytotoxicity, such as chromium release assays and flow cytometry-based assays using only cell-impermeant nuclear stains, detected little NK cell-mediated cytotoxicity against freshly isolated neutrophils. The reason for neutrophil resistance to lysis was not addressed in this study, but it is tempting to speculate that it may reflect the need to keep neutrophils intact in vivo because of their toxic intracellular content.

NK cell cytotoxic activity is tightly regulated. Upon encounter with a potential target cell, activating and inhibitory receptors on NK cells interact with their corresponding ligands, and the outcome is determined by the balance between activating and inhibitory signals in the NK cell (45). By Ab-mediated blocking of activating or inhibitory receptors on NK cells, or their ligands on the target cell, this balance can be shifted to inhibition or activation of the cytotoxic response, respectively. In this study, an Ab to the NKp46 receptor was found to significantly protect neutrophils from NK-induced apoptosis (Fig. 3). In line with these data, neutrophils displayed significant binding of a soluble NKp46Fc fusion protein. To our knowledge, this is the first report to show NCR ligand expression in neutrophils. Altogether, these data suggest that NKp46 binding to a still undefined ligand on neutrophils is of importance in NK cell-induced neutrophil apoptosis.

Previous studies have shown that CD18 is involved in NK cell interactions with target cells (46, 47), and blocking CD18 enhanced the neutrophil prosurvival effects of NK cell-derived cytokines (43). In line with these data, we found neutrophil apoptosis induced by NK cells to be significantly reduced by a blocking Ab to CD18 (data not shown). In contrast, blocking HLA class I molecules on neutrophils with a specific Ab did not increase NK cell-induced apoptosis, probably because of the moderate expression of HLA class I molecules on neutrophils.

NK cell activation results in the fusion of lytic granules with the plasma membrane and the release of perforin and granzymes into the immunological synapse formed between the NK cell and the target cell (27, 28). Whether FasL is present only in lytic granules or is also stored in distinct intracellular structures is still a matter of debate (48, 49). Our data suggest that degranulation occurs when NK cells interact with neutrophils even though only a small fraction of NK cells upregulated CD107a expression after coincubation with neutrophils as compared with coincubation with K562 cells. However, K562 cells represent an exceptional target cell line in their ability to induce NK cell activation and degranulation. In this context, it should be noted that several target cell lines that are susceptible to NK cells in chromium release assays, such as several EBV-infected cells, also fail to induce substantial degranulation as determined in a CD107a assay (A. Moretta, unpublished observations). Pretreatment with the caspase-8 inhibitor, z-IETD-fmk, or neutralizing Abs to FasL or Fas significantly protected neutrophils from NK cell-induced apoptosis, strongly suggesting that NK cells trigger PMN apoptosis by use of the Fas pathway.

In conclusion, in this study, we show that human NK cells are capable of triggering apoptosis in neutrophils. NK cell-induced apoptosis was contact and Fas dependent, and mediated by the activating NK cell receptor, NKp46. These data expand the role for NK cells as modulators of immune responses and suggest that NK cells could contribute to the safe removal of neutrophils, and thus play an important part in the resolution of inflammation.

We thank Michela Falco (Istituto Giannina Gaslini, Genova) for supplying the NTB-AFc fusion protein.

The authors have no financial conflicts of interest.