Interactions between Human NK Cells and Macrophages in Response to Salmonella Infection¹

Salmonella enterica is an intracellular Gram-negative bacterial pathogen capable of infecting humans by food or water poisoning. Human and animal salmonellosis presents a wide range of clinical syndromes, from gastroenteritis to severe bacteremia and typhoid fever. During bacteremia, local defenses can be insufficient to limit bacterial dissemination to the intestinal tract and Salmonella can spread to lymph nodes, liver, and spleen. It is well established in mice that resident macrophages play a crucial role in clearing Salmonella infection by controlling bacterial intracellular growth and by activating neighboring cells. NK cells are important immune effectors for preventing microbial invasion and dissemination (1, 2). They are found in blood as well as in peripheral nonlymphoid tissues and secondary lymphoid organs, and they are recruited to the site of inflammation (3, 4). Activation of NK cytotoxicity relies on a complex balance of inhibitory and activating receptors. In early host responses, NK cells exert two main functions, namely, cytotoxicity and the secretion of a range of cytokines. Among the secreted cytokines, IFN-γ plays a key role in enhancing immune responses, in particular by modulating macrophage activation, which enhances killing of intracellular bacteria (5, 6, 7). Interestingly, patients deficient in one part of the IL-12/IFN-γ axis show marked susceptibility to salmonellosis (8, 9). The preponderant IL-12-dependent, IFN-γ-producing cells are NK, T, and NK-like T cells, consistent with involvement of at least one of these cell types in the human anti-Salmonella response. The roles of these different subpopulations of cells, and particularly NK cells, have been well explored in some virus and parasite infections (10, 11, 12, 13).

A potent role for NK cells in controlling Salmonella in mice has been previously suggested, but little is known in human infection (14, 15, 16, 17). A potential problem with previous experiments is that they were performed either by injection of Salmonella into mice, or they used mice that were treated with antibiotics. Therefore, we first sought to confirm a role of NK cells in infection by gavage, in antibiotic-free mice. Data from the mice then led us to examine the human situation. Salmonella dissemination is directly linked to its ability to replicate within macrophages (18), and emerging studies call for crosstalk of NK cells with macrophages (11, 19, 20). These two facts led us to question if these two cell types work in concert in the establishment of an efficient antimicrobial immune response and clearance of Salmonella.

Experiments were performed using 6- to 8-wk-old female C57BL/6 mice in accordance with regulations. At day −1, anesthetized mice were injected i.v. with 50 μg of NK1.1 mAb (PK136) in 200 μl of PBS to deplete NK cells. Mice in the control group were injected with PBS. Mice were deprived of food for 24 h, anesthetized again, and infected (day 0) by oral gavage with 0.1 ml of bacteria suspended in PBS. Mice were injected again with control or NK1.1 Ab at day 5 and followed for mortality up to 12 days. Statistical analyses were performed using the GraphPad Prism software (log rank test).

S. enterica serovar Typhimurium 12023 wild-type (ATCC 14028) and attenuated (sifA mutant) strains were used for in vitro mice infection and mouse infection experiments, respectively. Stationary-phase Salmonella were opsonized in RPMI 1640 containing 20% autologous serum for 30 min. Macrophages were infected at a multiplicity of infection of ≈50:1. Cells were washed with PBS and then fresh culture medium containing gentamicin (50 μg/ml) was added for 1 h, and the antibiotic concentration was reduced to 5 μg/ml for the remainder of the experiment.

For enumeration of intracellular bacteria, cells were washed in PBS and lysed with 0.1% Triton X-100 in PBS to release bacteria. Dilution series in PBS were plated onto Luria-Bertani agar plates, incubated overnight at 37°C, and colonies from triplicates were counted. When indicated, the CFUs were normalized to the sample of infected macrophages in the absence of other cells (macrophages plus Salmonella = 100%).

PBMCs were from Buffy coats (British National Transfusion Service). PBMCs were isolated by Lymphoprep (Axis-Shield) density gradient and washed in PBS. Autologous serum was collected, filtered, and heat-inactivated. PBMCs were resuspended in PBS for further purification or were maintained in liquid nitrogen.

Autologous NK cells and monocytes were isolated from PBMCs using a negative isolation kit (Dynal Biotech) (purity of 90–98% and 88–97%, respectively). Monocytes were differentiated into macrophages by culturing 7–10 days in RPMI 1640, 1% autologous serum, and 50 ng/ml M-CSF (PeproTech). NK cells were cultured in RPMI 1640, 10% autologous serum,with 100 U/ml IL-2 (R&D Systems) before use. When stated, autologous NK cells were freshly purified from thawed PBMCs.

PBMCs, pure resting NK cells, or IL-2-cultured NK cells were cocultured with Salmonella-infected or uninfected macrophages at a macrophage/NK cell ratio of 1:2 in RPMI 1640 media, 10% autologous serum, for 20–24 h. When stipulated, cells were separated by transwells or activated with exogenous IL-2 (100 U/ml) for 24 h. In neutralizing assays, mAbs used were as follows: anti-NKG2D, anti-IL-2, anti-IL-12, anti-IL-15, anti-IFN-γ (R&D Systems), anti-IL-18 (MBL International). Anti-NKG2D (ON72), anti-NKp30, anti-NKp44, anti-NKp46 (kind gifts from A. Moretta, Genova, Italy), anti-2B4 (Immunotech), anti-NKp80 (kind gift from A. Steinle, Tübingen, Germany), anti-CD11a (Abcam), anti-CD18 (Endogen), anti-ICAM-1 (15.2, Serotec; RR1/1, Alexis Biochemicals), and anti-Fas-L (R&D Systems) were added to NK cells/PBMCs at 4°C, 20 min before incubation with macrophages. Respective isotype controls were added for each Ab in the same conditions.

For all FACS analyses, cells were incubated with 5% autologous serum and 5% FCS, 2 mM EDTA in PBS (FACS buffer) at 4°C. They were then incubated with Abs in FACS buffer on ice and washed twice before fixation in 2% formaldehyde, and cells were then analyzed on a FACScan (BD Biosciences). The percentages of CD69 expression were normalized to the sample of NK cells/PMBCs with infected macrophages and in the absence of mAbs (NK cells/PBMCs plus macrophages plus Salmonella = 100%).

For intracellular staining, brefeldin A (10 μg/ml; Sigma-Aldrich) was added to the cells for 4 h. The cells were resuspended in FACS buffer and fluorochrome-conjugated Abs were added on ice. The cells were washed and fixed for 15 min at room temperature in PBS paraformaldehyde (3%; Sigma-Aldrich). Cells were then permeabilized in FACS buffer, 0.1% saponin (Sigma-Aldrich) along with FITC-conjugated perforin, granzyme A, or IFN-γ (all from BD Pharmingen). Cells were washed in FACS buffer before analysis. When indicated the percentages of IFN-γ expression were normalized to the sample of NK cells/PMBCs plus infected macrophages, minus mAbs (NK cells/PBMCs plus macrophages plus Salmonella = 100%).

Directly labeled Abs were from BD Pharmingen, except for the following mAbs: PE anti-NKp30, NKp44, NKp46, and NKG2D (Beckman Coulter); mAbs PE anti-NKp80 and anti-CD69 (R&D Systems); and anti-CD11a, anti-CD18, anti-ICAM-1, anti-2B4, and anti-AICL (kind gift from A. Steinle, Tübingen, Germany). If unlabelled, appropriate secondary Abs were used.

Monocyte differentiation occurred directly on sterile coverslips. Pure IL-2-cultured NK cells were added for 4 h before fixation. IL-12Rβ2 Ab (goat polyclonal; eBioscience) was used in PBS before permeabilization and intracellular staining. For cytotoxic granule staining, cells were permeabilized with 0.1% saponin, stained with anti-granzyme A and anti-perforin (BD Pharmingen). Cells were then washed and incubated with secondary Ab (phalloidin was used to label actin). Coverslips, mounted in Mowiol (Sigma-Aldrich), were viewed under a Zeiss LSM 510 laser scanning confocal microscope. Images were analyzed using the LSM (Zeiss) and Adobe Photoshop software.

IL-12 was detected in cell supernatants using a commercial ELISA kit (PeproTech). IL-18 was detected using a FlowCytomix human IL-18 kit (Bender MedSystems). All samples were analyzed in duplicate or triplicate as indicated for mean and SD calculations.

NK cell degranulation was assessed by FACS using surface mobilization of CD107a. GolgiStop and FITC-anti-CD107a mAb (BD Biosciences) were added to the NK cell/macrophage coculture for 4 h. NK cells were harvested and stained for flow cytometry. FITC-positive cells, scored by FACS, revealed those NK cells that had effected cytotoxicity (21). When indicated, the percentages of NK cells exhibiting CD107a expression were normalized to the sample of NK cells in the presence of infected macrophages, but in the absence of mAbs (NK cells plus macrophages plus Salmonella = 100%).

Graphic representation and statistical analyses of NK cell activation were performed using the Graphpad Prism software. Comparisons of distributions were performed using a Wilcoxon matched pairs test with 95% confidence intervals with p values of ≤0.05 considered to be significant.

As mentioned above, production of IFN-γ by IL-12 stimulation of NK, T, and NK-like T cells is a critical arm of resistance to pathogens. To explore this in regard to Salmonella infection, we first examined the role of NK and NK-like T cells in an in vivo infection protocol. Bacteria were inoculated by a physiologic route (oral infection, gavage) in antibiotic-untreated mice in which NK1.1-dependent NK and NK-like T cells had been depleted. As shown in Fig. 1,A, no differences in survival between NK1.1-depleted and nondepleted mice were noticed at high bacterial doses. However, NK1.1-depleted mice were marginally more susceptible to oral infection with lower doses of Salmonella (Fig. 1,B). We took advantage of an attenuated strain of Salmonella, mutated for the well-characterized virulence factor sifA (22). Compared with control mice, NK1.1-depleted animals were particularly sensitive to infection with Salmonella (Fig. 1 C). The results are consistent with other studies where asialo-GM1-dependent depletion (which depletes NK, NK T, and CD8⁺ T cells and some macrophages) and nonphysiologic routes of infection were used, either by direct injection into the peritoneal cavity or directly into the bloodstream (14, 15). Using the anti-NK1.1-based depletion assay and a physiologic oral infection pathway, the data demonstrate that mice depleted of both NK and NK-like T cells are more sensitive to infection, confirming the involvement of one or both of these cell types in controlling Salmonella growth in vivo.

We next turned to human material. Patients with deficiencies in the IL-12/IFN-γ axis present with extraintestinal nontyphoidal Salmonella infections (23, 24). IL-12-dependent sources of IFN-γ are NK, T, or NK-like T cells, but the cooperation and contribution of these lymphocyte subsets to IFN-γ production in response to salmonellosis in humans are largely unknown. To approach this, we used an ex vivo system where human monocytes were differentiated into macrophages before infection with Salmonella, then cocultured 24 h with autologous total PBMCs. We measured production of IFN-γ by lymphocytes, using intracellular immunostaining and FACS analysis, in the presence or absence of autologous macrophages, both infected and uninfected. Only in coculture with infected macrophages did a small fraction of lymphocytes within the PBMCs express IFN-γ (Fig. 2^,A). By assaying CD3 and CD56 markers, we found that CD3⁻CD56⁺ NK cells were the main source of IFN-γ from PBMCs (Fig. 2^,B). Using transwells, which permit physical separation of PBMCs and macrophages without impairing soluble protein transfer, NK cell activation was found to be dependent on contact of PBMCs with infected macrophages (Fig. 2^,C). CD25 and CD69 activation markers were highly expressed on CD3⁻CD56⁺ gated NK cells, in a contact-dependent manner, when PBMCs were cocultured with Salmonella-infected macrophages, consistent with activation (supplemental Fig. S1A).³ After 24 h of coculture with autologous infected macrophages, only a very small proportion of T cells produced IFN-γ (Fig. 2,D). Interestingly, a significant proportion of NK-like T cells (CD56⁺ T cells) produced IFN-γ in response to Salmonella-infected macrophages, in a contact-dependent manner, suggesting that these cells may also contribute to the responses against Salmonella (Fig. 2 E). Because NK cells were the main IFN-γ-producing cells in our Salmonella infection model, compared with T and NK-like T cells, we chose to study them in more detail.

Cytokines such as IL-2, IL-12, IL-15, IL-18, and type I IFN have been implicated in NK activation (13). We therefore asked if these soluble factors were involved in NK activation by Salmonella-infected macrophages. We tested a range of neutralizing mAbs for blocking activation of NK cells cocultured 20–24 h with infected macrophages. IL-12 and IL-18 were important for IFN-γ production by NK cells (Fig. 2,F), consistent with other infection models (for review, see Ref. 13). Our data were consistent with a synergistic effect of IL-12 and IL-18 (Fig. 2,F). These cytokines were also involved, but to a lesser extent, in induction of the activation markers CD25 and CD69 on NK cells (supplemental Fig. S1B). However, type I IFN did not appear to be necessary (data not shown). IFN-γ production by NK-like T cells in response to Salmonella-infected macrophages was similarly impaired by IL-12 and IL-18 neutralizing mAb (Fig. 2 E).

IL-2 and IL-15 had only a minor involvement in NK cell activation (Fig. 2,F). Since IL-2 and IL-15 have biologically redundant activities, we neutralized both cytokines simultaneously (25). At least one of them was required for expression of IFN-γ cell upon Salmonella infection, confirming that these two cytokines play a redundant role in this process (Fig. 2 F).

To dissect further the direct influence of infected macrophages on NK cell activation, we purified monocytes and differentiated them into macrophages, and then we cocultured them with autologous, freshly purified NK cells. In this scenario, IFN-γ production by resting NK cells was barely detectable, even when the cells were cocultured 24 h with infected macrophages (Fig. 3,A and data not shown). By adding exogenous recombinant IL-2 (100 U/ml) during the 24 h of coculture, IFN-γ production dramatically increased in the cocultures with infected, but not uninfected, macrophages. As shown for PBMCs in Fig. 2, this activation still required direct contact of the NK cells and infected macrophages (Fig. 3 A). NK cell activation was initiated by infected macrophages, but consequent IFN-γ production was dependent on priming by the cytokine environment, especially as regards IL-2. These findings are consistent with three components in responses to infected cells: a “ménage à trois” of NK cells, infected macrophages, and IL-2/IL-15-secreting cells. Presumably the source of the IL-2 or IL-15 in our ex vivo experiments, and in vivo, is another partner cell, probably a dendritic cell or T cell.

To define the molecular mechanisms involved in the NK-macrophage interaction, we used IL-2-stimulated NK cells in subsequent experiments. This protocol allowed us to avoid any variables associated with the partner cells supplying IL-2 and/or IL-15. Using this method, we showed that Salmonella-infected, but not uninfected, macrophages, without other cells, induced IFN-γ production by NK cells (Fig. 3 B). This IFN-γ production was completely impaired by preventing NK-macrophage contact using transwells, confirming that cell-to-cell interaction was required for NK cell activation.

We then investigated the requirement of two well-characterized adhesion molecules: ICAM-1 and LFA-1. Two Abs against ICAM-1 (clone 15.1 and RR1/1) and an Ab against LFA-1 (anti-CD11a) partially blocked NK cell activation (Fig. 4,A). The data suggest that ICAM-1/LFA-1 interaction contributes, at least partially, to the contact dependence. Direct contact between these two cell types was then confirmed by confocal microscopy. Synapse formation was visualized by actin polarization (labeled by phalloidin) at the contact zone. As shown in Fig. 4,B, LFA-1 (labeled by an anti-CD18) and ICAM-1 were relocalized at the contact zone between infected macrophages and NK cells. Finally, we showed that ICAM-1 and LFA-1 were highly expressed on both the NK cell and the macrophage cell surface (Fig. 4, C and D). Furthermore, ICAM-1 was significantly up-regulated on infected macrophages, providing a mechanism for enhancing contact with NK effectors (Fig. 4, C and D).

NK cell activation is dependent on a subtle balance between activating and inhibitory receptors. A wide range of activating receptors has been described, so we investigated their potential role in our system. Previous studies have shown that the levels of various activating receptors are modulated during contact with target cells. We analyzed the expression of these receptors on NK cells by flow cytometry after 24 h of coculture with macrophages, infected macrophages, or infected macrophages separated by transwells (supplemental Fig. S2A). Only minor differences in expression of NKp30, NKp44, NKp46, NKp80, or 2B4 molecules were found. However, NKG2D expression was significantly influenced by 24 h of coculture with infected macrophages. This effect was impaired when the cells were separated by transwells, consistent with a possible NKG2D-ligand interaction (supplemental Fig. S2A).

We next examined whether the range of surface markers on Salmonella-infected macrophages differed from those on uninfected cells. Minor differences were noted for MHC class I and for three NKG2D ligands: ULBP2, ULBP3, and MICA/B (supplemental Fig. S2B). Up-regulation of three ligands for activating receptors was observed: ULBP1 (ligand of NKG2D), which was highly induced, CD48 (2B4 ligand), and AICL (NKp80 ligand).

To probe the role of activating receptors further, we incubated NK cells with different blocking mAbs and then assessed IFN-γ production after 24 h of coculture with infected macrophages. IFN-γ production was not blocked by two different mAbs, in each case, to NKG2D, NKp30, NKp44, and NKp46. Anti-2B4, CD80/CD86 coreceptors, CD40L, and NKp80 reagents were similarly ineffective (supplemental Fig. S3, A and C). None of these neutralizing mAbs affected the CD69 activation marker, which was overexpressed during infection (supplemental Fig. S3B). So far, none of the activating receptors tested seemed to be involved in NK cell activation by infected macrophages.

The data in Fig. 2 showed, using unfractionated PBMCs, that IL-12 and IL-18 are involved in NK cell activation. We addressed further the source of these two cytokines responsible for activation of purified NK cells. Blocking Abs against IL-12 and IL-18 extinguished IFN-γ production from NK cells induced by infected macrophages (Fig. 5,A). The two cytokines were also implicated in CD69 expression by NK cells (Fig. 5 B).

To determine whether infected macrophages enhance the cytotoxicity of NK cells, we measured the level of degranulation by the presence of the lysosomal protein CD107a (Lamp-1) at the NK cell surface. Coculture overnight with infected cells, before the degranulation experiments, resulted in a significant increase of CD107a staining at the surface of NK cells (Fig. 5,C). Interestingly, this induced degranulation overlapped with dramatically lowered numbers of infected macrophages, suggesting that activated NK cells triggered death of infected macrophages (Fig. 5,D). To investigate if activating receptors were involved, we ran the degranulation assay in the presence of various mAbs against NKG2D, NKp30, NKp44, NKp46, NKp80, and 2B4. None of these Abs had a significant effect on NK cell degranulation (supplemental Fig. S3D). However, in contrast to the IFN-γ assay, neutralization of IL-12 and IL-18 with mAbs partially abrogated NK cell degranulation (Fig. 5,E). Importantly, consistent with these results, macrophages infected with Salmonella were the source of significant amounts of secreted IL-12 and IL-18 (Fig. 5 F).

Thus, IL-12 and IL-18, secreted by Salmonella-infected cells, are a prerequisite for degranulation and IFN-γ production by NK cells in response to bacterial growth in macrophages.

It is well established that efficient cytokine-dependent crosstalk between immune cells, such as DCs and NK cells, needs synaptic delivery of cytokines. The key involvement of the cytokines IL-12 or IL-18 in NK cell activity in the antibacterial process, described in earlier experiments, prompted us to test for the presence of cytokine receptors at the surface of NK cells. As shown in Fig. 6,A, IL-12Rβ2 was slightly up-regulated at the surface of NK cells cocultured with infected macrophages. Interestingly, IL-12Rβ2 colocalized with actin at the contact zone with infected macrophages (Fig. 6 B), suggesting a possible synergistic signaling role of a ligand-receptor contact, or a synaptic delivery of IL-12 and IL-18 by infected macrophages, as has been reported for DCs (26, 27).

Finally, we assessed whether NK cells, activated by the presence of infected macrophages, had a direct role in controlling replication of Salmonella in macrophages. Infected macrophages were cultured 24 h with or without autologous NK cells, and Salmonella survival was measured using CFU assays. Measured at 24 h, there was a dramatic decrease (>60%) in bacterial replication in the presence of NK cells (Fig. 7, A and B), showing that human NK cells directly control Salmonella replication in macrophages. This effect was contact-dependent, confirming that direct crosstalk between these two cell types was a prerequisite for an efficient antibacterial response. Using blocking Abs against Fas-L (CD95-L) or IFN-γ, we showed that the decrease of bacterial replication is independent of IFN-γ secretion by NK cells and of Fas/Fas-L interaction (Fig. 7,C). Granzyme A and perforin are two of the major components of NK cell cytotoxic granules (28). NK cells were incubated for 24 h with macrophages as described in Fig. 7,D, and expression of granzyme A and perforin was assessed by intracellular staining. As shown in Fig. 7,D, only a minor increase of perforin and granzyme A was observed in NK cells cocultured with infected macrophages, compared with NK cells cultured alone. Consistent with the increased degranulation observed (Fig. 5), granzyme A and perforin were polarized at the contact zone between NK cells and infected macrophages (Fig. 7 E). It is well established that, with both NK cells and CTLs, killing is mediated by redistribution of cytolytic granules containing granzyme A and perforin at the contact zone with the target cell (29, 30). Our results suggest therefore that the decreased bacterial survival observed in the presence of NK cells is independent of IFN-γ or Fas-L, but may be due to direct killing of infected macrophages mediated by granzyme A and perforin.

In peripheral blood, T cells, NK-like T cells, and NK cells may all produce IFN-γ. In our experiments, using human autologous lymphocytes cocultured with Salmonella-infected macrophages, the main source of IFN-γ was NK cells. The response to infected phagocytic cells could be divided into two identifiable stages. First, resting human NK cells required IL-2- and/or IL-15-dependent priming. The importance of these cytokines has been established in responses to other pathogens, such as Plasmodium, suggesting that NK cell-activating receptors require prior priming, by IL-2 or IL-15, for efficient signaling (12, 13). It is well established that murine dendritic cells and plasmacytoid dendritic cells produce IL-2 and enhance NK cell activation (31). Moreover, monocytes, dendritic cells, and T cells constitutively provide IL-15 (32, 33). Recently, an in vivo study demonstrated that IL-15 production by dendritic cells is an important factor in priming naive NK cells into effector cells (34). Interestingly, contact dependency and exogenous IL-2 are required for NK cell activation, suggesting that transpresentation of IL-15 by IL-15R, if it occurs, does not take place on infected macrophages. The second step of NK cell activation involves direct contact with the IL-12/IL-18-secreting infected macrophages. We confirmed the key role of IL-12/IL-18 in NK cytotoxicity in our experimental set-up (35, 36). Additionally, we demonstrated that IL-12 and IL-18, by activating NK cells, are implicated in the control of bacterial replication by producing IFN-γ and by inducing infected cell death.

Contrary to the studies on another intracellular bacterium, Mycobacterium, we did not find evidence of direct involvement of NKp46 and NKG2D by using inhibitory mAbs (37). Incidentally, we found NKG2D to be highly up-regulated on NK cells, and one of its ligands, ULBP-1, was strongly expressed on infected macrophages. These phenotypes suggest that NKG2D could be involved in NK activation (38). None of the published activation receptors we tested was apparently involved, although other activation receptors cannot be ruled out. However, we showed a contact dependency on NK cell activation, which involves the engagement of the adhesion molecules ICAM-1 and LFA-1. Adhesion molecules and IL-12R β2 were enriched in the contact zone between NK cells and macrophages. It is well established that, during dendritic cell/NK cell crosstalk, directed relocation of IL-12/IL-18 from the target is crucial for efficient NK activation (26, 27). Similar conclusions can be made for crosstalk of macrophages with primed NK cells during bacterial infection, as close contact between these two cell types is needed. An alternative hypothesis is that IL-12/IL-18 and ligand/receptor signals act synergistically to efficiently activate NK cells. A recent study showed that IL-12R and FcγRIIIa, when relocalized to lipid-raft domains, led to IFN-γ production by NK cells. Moreover, it was shown that migration of IL-12R to microdomains acts as an enhancer of signaling activity (39). We found that IL-12R is concentrated at the contact zone between NK cells and macrophages, similarly to LFA-1 (CD18), suggesting that IL-12R is relocalized to the synapse during macrophage recognition by NK cells. In this case, the synaptic relocalization event could concentrate signaling molecules, as is the case with FcγRIIIa cross-linking, a feature that might increase the IL-12 sensitivity. This hypothesis is consistent with the strict contact- and IL-12/IL-18-dependent NK activation we observed. Higher levels of ICAM-1, as well as other adhesion molecules, could contribute to this process by enhancing the stability of the IL-12R-rich domains.

We also showed that Fas and IFN-γ are not directly involved in the decrease of Salmonella replication observed, whereas degranulation occurs. Cytotoxic responses may be crucial for controlling intramacrophagic bacterial replication. However, macrophage/NK cell cooperation leads to a significant IFN-γ production, which might be important for clearance and control of Salmonella dissemination in vivo.

In support of our findings, in vivo studies have established the importance of synergistic IL-12 and IL-18, as well as IFN-γ, in protection against various pathogens, including S. typhimurium (40, 41, 42, 43, 44). In contrast, studies of patients deficient in one part of the IL-12 pathway suggested that IL-12 is redundant in human immune responses against most microorganisms, with the exception of mycobacterial infection and salmonellosis (8, 9, 23, 24). A study analyzing data from 154 of these patients deficient in a component of IL-12 highlighted the fact such patients have a higher prevalence of salmonellosis compared with patients with IFN-γ pathway deficiency (23). These findings suggest that Salmonella prevalence in IL-12/IL-12R deficiency is partially mediated by IFN-γ and might, taking our results into account, also be dependent on IL-12-mediated NK cell degranulation. A recent study showed that IL-12/IL-12R-deficient patients possess very low numbers of CD56⁺ T cells. Their NK cells are hyporesponsive for IFN-γ production and for cytotoxicity, consistent with a role of IL-12/IL-23 in priming human NK effector function as well as NK-like T cell differentiation and survival (45). These results are consistent with our findings showing that NK and NK-like T cells are similarly activated by human infected macrophages. The relative contributions of these two cell types are difficult to determine, as salmonellosis has not been documented in cases of human NK cell deficiency. It is not possible to draw any reliable conclusions, as the number of cases is very rare and patients suffer from multiple virus infections or virus-dependent disorders (46, 47).

Future work using conditional NK cell-deficient and NKT cell-deficient mice may permit dissection of the relative contribution of these cells in immune responses against Salmonella. Experiments performed by a physiologic route of infection, in antibiotic-untreated mutant mouse strains, should also confirm in vivo the importance of cytokines (IL-12, IL-15, and IL-18) or adhesion molecules (LFA-1) in response to Salmonella. Moreover, the murine system might provide crucial information about where infected macrophages and NK cells may encounter each other in vivo.

Our data highlight the key role of human NK cells in the response to infected macrophages, leading to clearance of resident Salmonella. Our favored interpretation for this process is that it involves co-occurrence of three components; that is, a “ménage à trois” of IL-2- and/or IL-15-secreting cells, Salmonella-infected macrophages, and NK cells. As part of innate immunity to Salmonella infection, activated NK cells can respond directly to infected macrophages by cytotoxicity and IFN-γ production, and they may also modulate other components of the host immune response.

We thank Drs. M. Espeli, N. Young, R. Eagle, J. Kaufmann, H. Reyburn, P. Boutet, C. Chang, A. Barrow, D. Jones, and R. Apps for helpful discussions and comments. We also thank Drs. A. Moretta and A. Steinle for providing Abs. We thank N. Miller and Dr. M. Bowen for technical assistance with the FACS cytometry and confocal microscopy analyses, respectively.

The authors have no financial conflicts of interest.