The following mAbs produced in our laboratory were used in this study: anti-HLA class-I (A6/136, IgM), anti-2B4 (CO54, IgM), anti-NTBA (MA127, IgM), anti-CD48 (CO202, IgM), anti-CD9 (M1B16 IgM), anti-DNAM-1 (F5, IgM), anti-NKp30 (F252, IgM), anti-NKp46 (KL247, IgM), anti-KIR3DL1/L2-S1 (AZ158, IgG2a), anti-KIR2DL2/L3 (GL183, IgG1), anti-KIR2DL1/S1 (11PB6 IgG1), anti-NKG2A (Z199, IgG2b), anti-p75 (QA79, IgG1), anti-IRp60 (E59/126, IgG1), anti-LFA-1 (ECM17/120, IgM), anti-LFA-3 (TS2/9, IgG1), anti-CD16 (c127, IgG1), anti-HLA-DR (D1.12, IgG2A), anti-PVR (M5A10, IgG1), anti-Nectin-2 (L14, IgG2a), anti-MIC-A (BAM195, IgG1), anti-ICAM-1 (7E22, IgG1), anti-CD69 (c227, IgG1), anti-CD25 (MAR93, IgG1), anti-NKp44 (Z231, IgG1), anti-CD86 (FM95, IgG1), anti-CD1a (FM184, IgM).

The following commercial mAbs were also used: anti-CD62L (clone DREG-56, IgG1) mAb, anti-CCR3 (clone 61828, IgG2A) mAb, PE-conjugated IgG2A-specific goat anti-rat secondary reagents (BD Biosciences, San Jose, CA, USA); anti-CXCR1 (IgG1) (Santa Cruz, CA, USA); anti-CCR4 (IgG1) (BD Pharmingen); anti-CXCR4 (IgG2b) (R&D); anti-ICAM2 (clone B-T1), anti-ICAM3 (clone BR1) (Diaclone); anti-CD32 (IgG2a) (Beckman Coulter); anti-ULBP1 (clone M295), anti-ULBP2 (clone M310) and anti-ULBP3 (clone M550) (Amgen Inc., Seattle, WA, USA). Anti-PD-L1 and anti-PD-L2 (IgG1) were kindly provided by Prof. Daniel Olive (Aix Marseille Université, France).

Annexin V-FITC was purchased from Bender MedSystems (Vienna, Austria, Europe). ToPro3 Iodide was purchased from Invitrogen (Eugene, OR, USA). Cytofluorimetric analysis of eosinophlis was performed by gating on Annexin V⁻/ToPro3⁻ cells. Anti-B7-H6 (IgG1) was kindly provided by Prof. Eric Vivier (Centre d’Immunologie de Marseille-Luminy, France). Anti-human IFNγ was purchased from R&D Systems Inc. (Minneapolis, MN, USA). Cytofluorimetric analysis was assessed by flow cytometry FACSCalibur; Becton Dickinson & Co. (Mountain View, CA, USA).

Buffy coats from healthy donors were obtained from the Immunohematology and Transfusion Center at the S. Martino Hospital (Genova, Italy).

Approval was obtained by the ethical committee of IRCCS S. Martino-IST (39/2012) of Genova (Italy). Informed consent was provided according to the Declaration of Helsinki.

Buffy coats were mixed at ratio 1:1 with 2% Dextran T500 (Pharmacosmos, Holbaek, Denmark). After sedimentation of red blood cells, the upper phase was separated into granulocytes and mononuclear cells by density gradient centrifugation. Residual erythrocytes in the pellet were gently lysed in water to yield a pure population of granulocytes. To obtain a pure population of eosinophils from granulocytes, we used the eosinophil isolation Kit (Miltenyi Biotec, Bergisch Gladbach, Germany) according to the manufacturer’s instruction. The purity of eosinophils was greater than 98% (defined as CD16⁻/2B4⁺/NTBA⁺ granulocytes, as shown in Figure S1 in Supplementary Material).

Notably, healthy donors were selected based on the percentage of eosinophils in peripheral blood and on their phenotype after separation. In particular, we discarded donors with a percentage of eosinophils more than 4% and with a phenotype indicating, according to the information found in the literature, a possible activation or sensitization of eosinophils (e.g., expression of CD69).

Purified eosinophils were resuspended in RPMI 1640 medium supplemented with 2 mM glutamine, 50 µg/ml penicillin, 50 µg/ml streptomycin, and 10% heat-inactivated FCS (Sigma-Aldrich, Taufkirchen, Germany), in the absence or in the presence of cytokines (IFNγ 500 U/ml, TNFα 200 U/ml, IL12 1 ng/ml, IL15 20 ng/ml, GM-CSF 50 ng/ml, or IL5 50 ng/ml, all purchased from Peprotech Inc., London, UK) or in the presence of NK cells. Importantly, in all cytofluorimetric analyses, we only considered live eosinophils (Annexin V⁻ and ToPro3⁻ cells).

Myeloid DC were generated from monocytes purified using CD14 MicroBeads human Isolation Kit (Miltenyi Biotec) from PBMC of healthy donors. Monocytes were cultured in RPMI 1640 containing 10% FCS, in the presence of IL4 and granulocyte/macrophage colony-stimulating factor (GM-CSF) (Pepro Tech, London, UK) at final concentrations of 20 and 50 ng/ml, respectively. After 6 days of culture, cells were characterized by the CD14⁻CD1a⁺CD83⁻ phenotype corresponding to immature DCs (iDCs). To generate CD83⁺CD86⁺ mature DCs, iDCs were stimulated overnight (o.n.) with LPS (Sigma-Aldrich) at a final concentration of 1 µg/ml.

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 greater than 98% NK cells (defined as CD56⁺/CD3⁻).

Such freshly purified NK cells were resuspended in RPMI 1640 medium, supplemented with 2 mM glutamine, 50 µg/ml penicillin, 50 µg/ml streptomycin, and 10% heat-inactivated FCS (Sigma-Aldrich, Taufkirchen, Germany), in the presence of either 1 ng/ml of IL12 (purchased from Peprotech Inc., London, UK) or 20 ng/ml of IL15 (purchased from Peprotech Inc., London, UK). These pro-inflammatory cytokines were selected for their known activating properties on NK cells, and because, unlike IFNγ and GM-CSF, they do not affect eosinophil survival nor eosinophil phenotype (Figure S2 in Supplementary Material) (38–40).

Cells were plated at 10⁵ cells/ml in round-bottom 96-well tissue culture plates (Costar, Corning Corp.). After overnight culture (o.n.), NK cells were washed and incubated o.n. with purified eosinophils. Then, NK cells were harvested and assessed for surface phenotype, cytolytic activity and cytokine production. For cytofluorimetric analyses, NK cells and eosinophils were first identified on the basis of their size difference (FSC) and granularity (SSC) and then of different surface markers. In particular, NK cells were identified as CD56⁺/CD3⁻ cells, whereas eosinophils were identified as CD16⁻/2B4⁺ granulocytes. Dead cells were defined as Annexin V⁺/ToPro3⁺ cells, thus, cytofluorimetric analysis of eosinophlis was always performed by gating on Annexin V⁻/ToPro3⁻ cells.

In some experiments, the same NK cells, after exposure to eosinophils, were cocultured o.n. with iDCs. Then, DCs were harvested and maturation markers were assessed by flow cytometry. This analysis was performed by gating on CD1a⁺/ToPro3⁻ cells.

For transwell (TW) experiments, NK cells and eosinophils were placed in 24-well TW (0.3 µm pore size; Corning Costar), upper and bottom chamber, respectively. To obtain activated polyclonally expanded NK cells (bulk), freshly isolated NK cells were cultured on irradiated feeder cells in the presence of 100 U/mL recombinant human IL2 (Proleukin; Chiron) and 1.5 ng/ml phytohemagglutinin (PHA) (GIBCO Ltd.).

Plasmids utilized for expression of human NKp30-Fc*, NKp46-Fc*, and DNAM-1-Fc* recombinant molecules were prepared as previously described (41–43), utilizing pRB1-2B4Fcmut vector (kindly provided by M. Falco, Istituto G. Gaslini, Genova, Italy) that contains a mutagenized sequence coding for a human IgG1 portion that does not bind to Fc receptors. Soluble receptors were produced in HEK293T cell line (human embryonic fibroblast) and purified by affinity chromatography using Protein A-Sepharose 4 Fast Flow (Amersham Biosciences) (41–43). Eosinophils (1 × 10⁵ cells) were incubated with 2 µg of NKp30-Fc* and NKp46-Fc* 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. Flow cytometry was performed using a FACSCalibur flow cytometer (BD Biosciences) and cells were analyzed with CellQuest Pro software (BD Biosciences).

Total cellular RNA was extracted from eosinophils and from HEK293T cell line (human embryonic fibroblasts) using an RNAeasy Micro Kit (Qiagen, Hilden, Germany). Oligo (dT)-primed cDNA was prepared by standard technique using Transcriptor (Roche, Monza, Italy). Amplification of B7-H6 cDNA (nt. 247-708, Accession N° NR_026750) was performed with Platinum Taq (Life Technologies Paisley, UK) for 35 cycles (30 s at 95°C, 30 s at 58°C, and 1 min at 68°C) using the following primers: H6 for 2 5′ TGCTGTGGGCGCTGACGA and H6 rev2 5′ GGTAGAACCCACTTGACTCA. β-actin and CD63 amplifications were performed for 30 cycles using the same conditions and served as internal controls; primers used were: β-actin up 5′ ACTCCATCATGAAGTGTGACG; β-actin dw 5′ CATACTCCTGCTTGCTGATCC; CD63 up 5′ CAGCCATGGCGGTGGAAG and CD63 dw 5′ CCACTCCCCCAGATGAGG. PCR products were run on a 0.8% agarose gel and visualized by ethidium bromide staining (16, 44).

NK cells that had been exposed to IL12 or IL15 and then cocultured with eosinophils were tested for cytolytic activity against various NK cell-susceptible target cells, including K562 and allogeneic iDCs, in a classical 4-h ⁵¹Cr-release assay as previously described (22). In other experiments, the cytolytic activity of resting or cytokines-primed NK cells was evaluated against autologous or allogeneic freshly isolated eosinophils in a 4-h ⁵¹Cr-release assay (22). The concentration of mAbs used for masking experiments was 10 µg/ml. The E/T ratios are indicated in the figure legends.

ELISA kits were used for measuring IFNγ assessment in the supernatants of NK cells stimulated with eosinophils (BioSource Int. Inc., CA, USA). Ab-mediated blocking experiments were performed adding saturating amounts of purified anti-NKp46, anti-NKp30, anti-2B4, and anti-LFA1 mAbs at the onset of the cell cultures.

Independent samples t-test was employed for evaluating quantitative variables. The test is a statistical technique that is used to analyze the mean comparison of two independent groups. The statistical level of significance was preset at 0.05. Graphic representation and statistical analyses were performed using the PASW Statistic version 18.0 software (formerly SPSS Statistics) (IBM, Italy) and GraphPad Prism 6 (GraphPad Software, La Jolla, CA, USA).

In agreement with previous studies, we found that eosinophils from non-atopic healthy donors express 2B4, NTBA, and IRp60 receptors (45–47), while they do not express CD16 (Figure S1A in Supplementary Material) (48) and CD69 (49). Moreover, eosinophils expressed different chemokine receptors, including CD62L as well as CCR3 (50) and the ligands for the programmed cell death protein 1 (PD-L1 and PD-L2) (Figure S1B in Supplementary Material). To assess the possibility that NK cells and eosinophils may interact with each other, we analyzed the phenotype of resting eosinophils also for the surface expression of specific ligands for NK cell receptors. As shown in Figure 1A, eosinophils did not express ligands for NKG2D (MICA, ULBPs) or for the DNAM-1 receptor (PVR and Nectin-2), although some variations in Nectin-2 expression could be observed among different donors. On the other hand, eosinophils expressed the ligands for the activating coreceptors 2B4 (CD48), CD2 (CD58), and NTBA (NTBA itself) (Figure 1A; Figure S1B in Supplementary Material) (43). Regarding the LFA-1 ligands, freshly isolated eosinophils expressed ICAM-3, whereas they were negative for ICAM-1 and ICAM-2 adhesion molecules (Figure S1B in Supplementary Material).

In order to assess the expression of NCR ligands on eosinophils, we used soluble NKp46-Fc* and NKp30-Fc* molecules. As shown in Figure 1B, both molecules react, albeit weakly, with eosinophils, thus indicating that these cells may express one or more ligands for NKp46 and NKp30.

In order to find out whether the NKp30 ligand expressed on eosinophils is the B7-H6 molecule [whose expression has been recently described on neutrophils/monocytes under inflammatory conditions (51)], we used specific anti-B7-H6 mAbs and performed the analysis at the mRNA level by RT-PCR. These experiments indicate that fresh eosinophils, similar to iDCs, do not express B7-H6 mRNA nor surface B7-H6 protein (Figure 1C and not shown), thus suggesting the existence of alternative (additional) NKp30 ligand/s on these cells. Unfortunately, we could not evaluate the expression of BAG6, due to the unavailability of specific reagents. Collectively, these data demonstrate that eosinophils express several ligands for NK receptors, suggesting that the two innate cell types may interact and influence each other.

We next analyzed whether eosinophils could modulate the NK cell phenotype and effector function. In these experiments, NK cells, either resting or short-term primed with IL12 or IL15, were cocultured o.n. with fresh autologous or allogeneic eosinophils. After o.n. coculture, NK cells were assessed for the expression of the early activation marker CD69. As shown in Figure 2A, the surface density and the percentage of CD69⁺ NK cells were strongly upregulated on IL12-conditioned NK cells, but not on resting NK cells, in the presence of autologous or allogeneic eosinophils. The same results were obtained after stimulation of NK cells with IL15 (not shown). Importantly, CD69 upregulation was mainly detected following cell-to-cell contact, although a slight increase of CD69 expression was detectable also when NK cells and eosinophils were separated by a TW membrane (Figure 2B). In order to identify molecules that may be involved in the interaction between NK cells and eosinophils, cocultures were performed in the presence of mAbs specific for different NK receptors, including the adhesion molecule LFA-1 and the activating receptors NKp46, NKp30, 2B4, and DNAM-1. As shown in Figure 2C, the eosinophil-induced upregulation of CD69 on NK cells was reduced by the combined antibody-dependent blockade of NKp46, NKp30, and LFA-1, but not by masking individual receptors (Figure 2C).

Regarding the expression of other classical activation markers (CD25 and NKp44), no major differences between IL12-conditioned NK cells and IL12-conditioned NK cells in the presence of eosinophils could be detected, although these molecules were weakly increased in NK cells that had been cultured with eosinophils (data not shown).

The level of surface expression of the other molecules analyzed in these experiments (including NKG2D and 2B4) remained substantially similar in NK cells cultured either in the absence or in the presence of eosinophils (data not shown).

In the same set of experiments, NK cells were used as effector cells in cytolytic assays against K562 (a classical NK-susceptible tumor target). As shown in Figure 3A, IL12-stimulated NK cells displayed significantly increase of cytotoxicity after coculture with eosinophils (right). In contrast, under the same culture conditions, resting NK cells did not acquire a higher cytolytic activity against the same target cells (left).

Notably, the stimulatory effect on NK cell cytotoxicity was dependent on cell-to-cell contact, as no increase in cytotoxic activity was observed in TW experiments (Figure 3B).

In order to determine whether eosinophils could also promote the production of pro-inflammatory cytokines by NK cells, coculture supernatants were evaluated for the presence of IFNγ (Figure 4). In these experiments, eosinophils did not induce IFNγ production by resting NK cells. However, high amounts of this cytokine was detected in NK cells pre-activated o.n. with IL12. In line with the results above, there was no detectable IFNγ production by NK cells in the absence of NK/eosinophil direct contact (Figure 4A). In all these experiments, resting NK cells that had been exposed o.n. to IL12 plus IL18 were used as positive control.

In experiments aimed at defining the molecular interactions involved in the eosinophil-NK cells cross talk, cocultures were performed in the presence of mAbs specific for different NK receptors. As shown in Figure 4B, antibody-mediated masking of NKp46 and even more when used in combination with anti-LFA-1 mAb inhibited IFNγ release by NK cells. Remarkably, the maximal effect of inhibition occurred again upon simultaneously masking of NKp46, LFA-1, and NKp30 (Figure 4B). By contrast, masking of other receptors (e.g., 2B4) had no substantial effect (Figure 4B and not shown).

Thus, both TW and masking experiments pointed to a critical role for receptor/ligand interactions during the NK/eosinophil cross talk, resulting in amplification of NK cell activation, as determined by the upregulation of CD69 expression and by the increases of cytotoxic activity and IFNγ production.

Next, we investigated whether coculture with eosinophils could promote NK cell-mediated killing of iDCs. In agreement with previous data, exogenous IL12-conditioned NK cells were able to kill iDCs (22); however, this activity was significantly increased after coculture with eosinophils, as shown in Figure 5A.

Next, we evaluated whether the interaction with eosinophils could influence the NK cell capability of promoting DC maturation. To this end, IL12-conditioned NK cells and eosinophils were cocultured o.n., then harvested, washed, and cultured with iDCs. After 24 h, the expression of CD86, HLA class-I and HLA class-II (i.e., HLA-DR) molecules on DCs was determined by cytofluorimetric analysis. As shown in Figure 5B, substantial increments in the mean expression of HLA molecules (both class I and II) and in the percentage of CD86-expressing cells were detected when iDCs were cocultured with NK cells plus eosinophils, as compared to iDCs cocultured with NK or eosinophils alone. In some experiments, eosinophils were removed before culturing NK cells with iDCs and also under these conditions DCs could undergo maturation (not shown).

Next, we investigated whether phenotypic changes in eosinophils occurred following interaction with NK cells. To this end, fresh eosinophils were cocultured with NK cells (either resting or conditioned with IL12/IL15). At the end of the culture period, eosinophils were harvested and analyzed for the expression of a number of informative markers, including CD69, ICAM-1, HLA molecules, and CD62L. A significant de novo surface expression of CD69, ICAM-1, and HLA-DR molecules and a marked upregulation of HLA class-I and CD62L molecules was detected on eosinophils cocultured with IL12-conditioned NK cells (Figures 6A,B). The same results were obtained by pre-treating NK cells with IL15. Resting NK cells did not induce any substantial effect (not shown).

Experiments were also performed in the presence of blocking anti-IFNγ mAbs to understand whether the de novo expression/upregulation of the above surface molecules could be induced by IFNγ. As shown in Figures 6A,B, when cells were cultured in the presence of anti-IFNγ mAbs, the expression of most of the above markers was reduced but not abolished (Figure 6B).

This indicates that the modification of eosinophil phenotype induced by cytokine-treated NK cells is in part, but not exclusively, due to the production of IFNγ during coculture. Next, we determined if the same receptor/ligand interactions responsible for eosinophil-mediated induction of NK cell effector functions were also involved in the events leading to eosinophil activation. To this end, neutralizing mAbs specific for NKp46, NKp30, or LFA-1 were added, alone or in combination, to cocultures and the expression of CD69 on eosinophils analyzed. As shown in Figure 6C, mAbs, added individually, did not (or only modestly) inhibit CD69 expression. In contrast, CD69 expression was significantly decreased in eosinophils when the neutralizing mAbs specific for NK receptors were used in combination. Particularly strong inhibition was obtained when anti-NKp46, anti-NKp30, and anti-LFA-1 mAbs were simultaneously added to the coculture.

To investigate whether eosinophils could represent possible targets for NK cell cytotoxicity, fresh (not-activated) allogeneic eosinophils were exposed either to resting or to IL2-activated NK cells (bulk) in ⁵¹Cr-release cytolytic experiments. As shown in Figure 7A, IL2-activated, but not resting NK cells, displayed a strong cytotoxic activity toward allogeneic eosinophils. In order to evaluate the contribution of one or another activating NK receptor, cytolytic assays were performed in the presence of mAbs specific for major activating NK receptors. As shown in Figure 7B, NKp46 and NKp30 mainly contributed to the killing of eosinophils, since mAb-mediated masking of these receptors, resulted in significant inhibition of lysis, while mAbs directed to other activating NK receptors had no substantial effect.

Similar results were obtained in cytolytic assays performed in an autologous setting using IL12- or IL15- short-term-primed NK cells as effector cells. As shown in Figure 8A, both types of cytokine-activated NK cells displayed similar levels of cytotoxicity against autologous or allogeneic eosinophils, while NK cells pre-cultured with IL4 or IL18 did not display any cytotoxicity against eosinophils (data not shown) (22). Notably, mAb-mediated disruption of inhibitory receptors/HLA class-I interactions did not result in increases of cytotoxicity (Figure 8A, left). These data suggest that HLA class-I molecules do not provide substantial protection to eosinophils from NK cell-mediated cytotoxicity. In agreement with these results, eosinophils displayed a low expression of surface HLA class-I molecules as compared to other innate cells known to interact with NK cells (Figure 8B).