PA01 is a clinical strain of PA (no. 15692) (11) whose genome has been fully sequenced. It expresses most of the documented virulence factors, including the T3SS also known as the “needle complex” and its effectors: Exoenzymes (Exo) S, T, and Y released in targeted cells through T3SS. Three isogenic deleted strains were used: PA-ΔS (ExoS deletion), PA-ΔT (ExoT deletion), and PA-ΔT3SS (deletion of the needle complex). PA-ΔS and ΔT were a gift from Dr. Andrew Y. Koh Laboratory at the University of Texas Southwestern Medical Center in Dallas, TX, USA. PA expressing the Green Fluorescent Protein (PA-GFP) was a gift from Dr. Wu at the University of North Dakota. PA-ΔT3SS (also called ΔPscC) carries a truncated PscC gene leading to a non-functional protein. PscC is a secretin-like constitutive protein of the outer membrane forming a channel enabling needle growth. Without the functional pscC protein, the needle in the T3SS cannot protrude to the cell surface and, as a result, the bacteria cannot inject Exo in the host cell cytoplasm (12). This strain came from Dr. Donald Moir at microbiotix^INC in Worcester, MA, USA. The PCR study confirmed the phenotype of each deleted strain (see Figure S1 in Supplementary Material). The isogenicity between each deleted strain was confirmed by pulsed-field gel electrophoresis (see Figure S2 in Supplementary Material).

MEK/ERK kinase inhibitor (PD98059) were purchased from Sigma Aldrich (France). Inhibitor was diluted according to the manufacturer instructions with dimethyl sulfoxide (DMSO) at 0.04%. Before the PA challenge, NK cells were incubated for 1 h at 37°C with 10 µM of inhibitor (14). Cultures with DMSO 0.04% under inhibitor-free conditions were also prepared to control its potential effect on cytokine response.

We used our validated pneumonia model (4, 15, 16). Six-week-old female Swiss mice (20–24 g) were anesthetized with isoflurane. A 24-G transtracheal feeding needle was inserted to inject 75 µl of bacterial suspension adjusted to 10⁸ CFU/ml. The same anesthesia procedure and treatment with 75 µl of saline buffer was applied to control mice (SHAM). Mice were maintained on a 12-h light/dark cycle. All experimental protocols were approved by the Committee of Animal Ethics of the Pays de Loire (CEEA-2012-233) and all methods were carried out in accordance with the guidelines and regulations.

Antibodies were purchased from BD Biosciences unless otherwise stated. Data were collected with four-color FACSCalibur (BD Biosciences) and LSRII cytometer (Benton Dickinson, Le Pont de Claix, France) and analyzed using FlowJo 6.2 software (Ashland, OR, USA). For PBMCs, NK cell gating was performed with anti-CD56-APC (NCAM16.2, #341026), anti-CD3-PerCP (SK7, #345766), and the corresponding isotype-matched control mAb. Cytolytic activity (CD107a membrane expression) was assessed with CD107-FITC (H4A3, #555800) after 5 h of in vitro incubation. To study IFN-γ production, the cells were treated with Brefeldin A (Sigma) at 10 µg/ml for 5 h. IFN-γ intracellular staining was then performed with anti-IFN-γ-PE (B27, #554701) after cell permeabilization with PFA 4% (Sigma) at 4°C overnight followed by Saponin 0.1% (Sigma). Cell viability among NK cell line was assessed by APC-Fixable Viability Dye Kit eFluor 780 staining (eBioscience).

In mouse pneumonia model, cell suspensions were obtained by mechanical and collagenase D digestion (1 h at 37°C) of lungs collected 24 h postinfection. NK cell gating was performed with anti-NK1.1-BV 421 (#562921) and anti-CD3-APC (#553066). For IFN-γ intracellular staining, after red blood cells lysis (RBC lysis buffer, Ozyme), 70 µm filtered cells were cultured 5 h in RPMI 1640 medium supplemented with 2% FCS with GolgiPlug, washed twice, and then stained for surface markers. Fixation and permeabilization was performed following manufacturer instructions (BD Cytofix/Cytoperm kit, BD Bioscience). Anti-IFN-γ AlexaFluor 488 (#557724) antibody or its rat IgG1κ isotype control were incubated overnight at 4°C. Cells were washed twice before analysis on a LSRII flow cytometer (BD Bioscience).

For confocal microscopy, cells were stained with primary rabbit anti-human NCR2 antibody (#133668, Abcam) for 30 min at 4°C (1/100) and secondary goat anti-rabbit Alexa 568 antibody (#11011, Life Technologies) for 20 min at 4°C (1/400). NK92 cells were seeded onto glass cover-slips with 2-octyl Cyanoacrylate DERMABOND™, Ethicon, and underwent 2 min of centrifugation at 1,500 rpm. Infection was performed with PA-GFP immediately before confocal visualization (Nikon A1 RSi) with Plan APO 60× objective with a numerical aperture of 1.40. Stack acquisition was performed at 30-s intervals, scan size was 512 × 512, 5× zoom, pinhole 2 (Airy unit), and 2 µm step sizes. Images were not processed after acquisition.

All ELISA kits were purchased from eBioscience.

Total RNA was isolated using the RNeasy kit (Qiagen) and treated for 45 min at 37°C with DNase (Promega). RNA (1 µg) was reverse-transcribed with superscript III reverse transcriptase (Life Technologies). The cDNA was subjected to RT-qPCR in a Bio-Rad iCycler iQ system using the QuantiTect SYBR Green PCR kit (Qiagen). See primer sequences in the online supplementary table (see Table S4 in Supplementary Material). Relative gene expression was normalized on GAPDH and calculated using the 2^−ΔΔCt method with samples from the IL-12 free, non-infected group as calibrators.

Cell pellets were suspended in cold RIPA buffer, protease inhibitor, and 1% phosphatase cocktails (Sigma), mixed with the SDS sample buffer, boiled, and separated by SDS-PAGE (10% TGX Precast Gel, Bio-Rad). Proteins were transferred onto a Trans-Blot^® Turbo™ membrane. Membranes were successively probed with primary antibodies, DyLight™ 680 or 800 secondary antibodies and revealed on an infrared imager (LICOR ODYSSEY). The following primary antibodies were used (cell signaling unless otherwise stated): anti-Phospho-Stat4 (Tyr693) (D2E4) rabbit mAb, anti-Phospho-p44/42 MAPK (Erk1/2) (9101S) rabbit mAb, anti-p44/42 MAPK (Erk1/2) (9102S) rabbit mAb or anti-Actin (A5441) mouse mAb (Sigma).

Statistical analyses were performed with GraphPad prism software (La Jolla, CA, USA). Continuous non-parametric variables were expressed as the median (25th to 75th percentile). The Kruskal–Wallis test was used to compare multiple groups. The post hoc Dunn’s test was used to perform multiple comparisons. Survival curves were compared to a log-rank test. P < 0.05 was considered to be statistically significant.

The datasets generated and/or analyzed during the current study are available from the corresponding author on request.

The close interaction between NK cells and PA has been reported previously (4, 17, 18). We first focused on the determinant of IFN-γ production by NK cells following PA-infection. We performed PA infection in PBMC to assess NK cells response in a physiological microenvironment. NK cells are heterogeneous with different subsets specialized in either cytokine or cytotoxic activities (8), thus we aimed to select the main subset of NK cells specialized in IFN-γ production. After gating on CD56^pos and CD3^neg cells, we analyzed intracellular IFN-γ staining and CD107a membrane expression (surrogate marker of degranulation) among CD56^bright and CD56^dim subsets (Figure 1A). After PA-WT infection, the proportion of CD56^bright NK cells increased and exhibited higher IFN-γ activity and lower cytotoxic activity compared to the CD56^dim subset (Figures 1B,C). To confirm preferential IFN-γ activity of CD56^bright subset, we sorted NK cells from PBMC by cytometry according to their subset (CD56^bright or CD56^dim) and subsequently 24-h infected each subset with or without IL-12 stimulation. IL-12 produced by dendritic cells (DC) is critical for NK cell activation (19). As previously described, without IL-12 priming, NK cells produced low levels of IFN-γ in response to PA. After IL-12 stimulation, PA significantly increased IFN-γ as compared to non-infected cells. Moreover, CD56^bright NK cells produced higher level of IFN-γ than CD56^dim, inciting us to study CD56^bright subset to precise intracellular pathways leading to IFN-γ production (Figures 1D,E).

To further explore microenvironment influenced on IFN-γ response after PA infection, we used a human NK cells line (NK92) specialized in cytokine production and sharing the CD56^bright NK cells receptor repertory (CD56^bright KIR^neg CD16^neg) (8) (see detailed phenotype of NK92 in supplemental Figure S3). Similarly to sorted human NK cells, without IL-12 priming, NK92 cells released low level of IFN-γ after PA infection (Figure 1F). IL-12 triggered IFN-γ production and PA further increased IFN-γ level after IL-12 priming. As compared to IL-15 or IL-21, also reported to participate in cytokine response of NK cells, IL-12 stimulation gave rise to higher IFN-γ production (see Figure S5 in Supplementary Material).

Our objective was to identify the pathways involved in IFN-γ production after IL-12 stimulation and the influence of PA infection on these pathways. For this purpose, PCR analysis was performed on NK92 cell line after PA-WT infection with or without IL-12 stimulation. The protein and mRNA IFN-γ followed the same trends (Figures 1F and 2A) in IL-12-treated NK 92 cells infected with PA, suggesting that PA infection regulates IFN-γ at a pre-transcriptional level. STAT-4 is the main transcriptional factor involved in both IL-12 receptor (IL-12R) signaling and IFN-γ mRNA transcriptional activity (20). Thus, we compared the activation of STAT-4 after 2-h infection with or without IL-12 stimulation in NK92 cell line. Compared to IL-12-treated NK cells, PA did not affect the phosphorylation of STAT-4, which is induced by IL-12 treatment (Figures 2B–D). We concluded that PA could modulate the production of IFN-γ in IL-12-treated NK cells through a pathway independent from IL-12R.

Live confocal microscopy recorded immediately after PA-GFP infection suggested direct bacteria-to-cell contact (Video S6 in Supplementary Material). When NK92 cells were cultured under a filter (preventing any direct contact with PA), the infection failed to increase the production of IFN-γ in IL-12-treated NK cells (Figure 3A). We concluded that direct bacteria-to-cell contact was involved in the cytokine activity modulation. During infection, PA uses a complex T3SS to inject effector proteins (Exoenzymes S, T, and Y) into host cells (21), these proteins interfere with the intracellular signaling pathways (22), the function and viability of target cells (21). Exoenzymes effects on the cytokine response in NK cells had not been investigated to date. For this purpose, sorted human NK cells were challenged with three PA isogenic strains deleted for T3SS or its effectors and compared to PA-WT (Figure 3B) (see Isotype control for intracellular in Supplemental Figure S8). The exoenzymes deletion did not modify bacterial growth (Figure S9 in Supplementary Material) or the survival of NK cells during infection (Figure S10 in Supplementary Material). Needle complex (PA-ΔT3SS) deletion reduced IFN-γ production in IL-12-treated NK cells as compared to PA-WT infection. Contrary to PA-ΔS (expressing ExoT), infection with a strain lacking ExoT (PA-ΔT) decreased IFN-γ activity as needle complex deletion did, suggesting that ExoT is a determinant of IFN-γ activity in NK cells. These results demonstrate that the production of IFN-γ by NK cells is not solely driven by IL-12 stimulation, but that exoenzymes can also modulate cytokine production.

Pseudomonas aeruginosa exoenzymes were already reported to interfere with Ras family proteins which control ERK phosphorylation (23). In view of the singular role of ExoT (Figure 3B), the next step was to investigate the pathway(s) involved in ExoT-induced IFN-γ production under IL-12 stimulation. Thus, we studied intracellular IFN-γ staining in sorted human NK cells after either PA-WT (full set of exoenzymes) or PA-ΔS (expressing ExoT but not ExoS) infection (Figure 3C) with or without ERK inhibitor. ERK inhibitor induced a major reduction in PA-ΔS and PA-WT whereas did not affect the percentage of IFN-γ⁺ cells in NI condition after IL-12 stimulation. Cell viability study after infection with PA-WT or deleted strains with or without ERK inhibitor ensured that cells mortality did not explain these differences (see Figure S10 in Supplementary Material). Western Blot experiments in NK92 cell line (Figures 3D,E) confirmed ExoT-dependent ERK phosphorylation after 2-h infection (see the membrane with unphosphoryled form of ERK in Figure S7 in Supplementary Material). Comparable phosphorylation of ERK with or without IL-12 suggest that PA may activate NK cells through ERK phosphorylation independently of IL-12 stimulation but that IL-12 priming remains a prerequisite for IFN-γ activity. As a result, in our model, ERK is specifically involved in IFN-γ production after PA infection but not in IL-12 dependent IFN-γ production.

We have already reported the critical role of NK cells in controlling infection and producing IFN-γ in a lethal mouse PA-pneumonia model (4). Using the same model, we assessed the role of T3SS and its effectors on mouse mortality and IFN-γ production in mice lungs. The deletion of T3SS or its effectors reduced the mortality rate in infected mice (Figure 4A) irrespective of the bacterial load in lungs 24 h after infection (Figure 4B). These data demonstrate in vivo the critical role of T3SS and its effectors. The IFN-γ level in lungs followed the same trends as observed in our in vitro model, with significantly higher IFN-γ activity in PA-expressing ExoT (PA-WT and PA-ΔS) than in PA-ΔT pneumonia (Figure 4C). Cytometry analysis in lungs after PA pneumonia confirmed that IFN-γ is mainly produced by NK cells (see Figure S11 in Supplementary Material). After PA pneumonia, the absolute number of NK cells was not different compared to sham condition (Figure 4D) but the percentage of NK cells was reduced (Figure 4E). There was no difference between PA-WT and deleted strains regarding NK cells percentage. PA-ΔS (Expressing ExoT) led to higher percentage of IFN-γ⁺ NK cells as compared to PA-ΔT (Figure 4F). These data confirmed the key role of ExoT in vivo on IFN-γ activity modulation.