Through complex interplay with APCs, subsets of NK cells play an important role in shaping adaptive immune responses. Bovine tuberculosis, caused by Mycobacterium bovis, is increasing in incidence and detailed knowledge of host–pathogen interactions in the natural host is essential to facilitate disease control. We investigated the interactions of NK-cell sub-populations and M. bovis-infected DCs to determine early innate mechanisms in the response to infection. A sub-population of NK cells (NKp46+CD2−) selectively expressing lymphoid homing and inflammatory chemokine receptors were induced to migrate towards M. bovis-infected DCs. This migration was associated with increased expression of chemokines CCL3, 4, 5, 20 and CXCL8 by M. bovis-infected DCs. Activation of NKp46+CD2− NK cells and secretion of IFN-γ was observed, a response reliant on localised IL-12 release and direct cellular interaction. In a reciprocal manner, NKp46+CD2− cells induced an increase in the intensity of cell surface MHC class II expression on DCs. In contrast, NKp46+CD2+ NK cells were unable to secrete IFN-γ and did not reciprocally affect DCs. This study provides novel evidence to demonstrate distinct effector responses between bovine NK-cell subsets during mycobacterial infection.
Tuberculosis (TB) is a chronic respiratory illness caused by mycobacterial pathogens afflicting a wide range of species. Infection in humans is most commonly as a result of Mycobacterium tuberculosis, whereas cattle are the primary host for Mycobacterium bovis. Whilst the mechanisms that govern protective immunity to mycobacteria have not been completely deciphered, IFN-γ plays an important role enabling containment of infection and Th1 polarisation [1, 2].
NK cells have key roles in bridging innate and adaptive immune responses. In particular, increased numbers of activated NK cells are evident within the lungs following mycobacterial infection in mice [3, 4] and have been shown in humans to lyse M. tuberculosis-infected monocytes . Crucially, NK-cell-derived IFN-γ contributes to early innate resistance to bacterial infection  and the subsequent development of Th1-biased immune responses within draining LNs [6-8]. Fully functional effector responses of NK cells are however reliant upon interplay with accessory cells, particularly DCs. During this process, DCs secrete chemokines to orchestrate the recruitment of effector cells with important roles for CXCR3 and CCR5 expression by NK cells [9, 10]. The subsequent co-localisation enables reciprocal interactions with enhanced secretion of IFN-γ following contact of human NK cells with M. tuberculosis-infected DCs . In addition, DCs are able to trigger NK cells to kill mycobacterially infected APCs or autologous immature DCs [5, 12]. Deciphering the mechanisms by which NK cells and DCs interact has demonstrated that multiple signals are required in the form of both soluble and contact-dependent stimuli. Infected DCs readily secrete IL-12 , which promotes NK-cell-derived IFN-γ production and cytolytic activity [10, 11]. In addition, direct cell contact is required to enable the formation of immunological synapses for polarised cytokine delivery  and receptor-mediated interactions. The maturation status of DCs is positively influenced by interactions with NK cells. Human NK cells stimulated with IL-18, or infected with Klebsiella pneumonia enhanced DC-derived IL-12 secretion and subsequently potentiated CD4+ T-cell-derived IFN-γ production and Th1 polarisation [10, 15].
NK cells are a heterogeneous population defined in humans as CD3− cells with variable expression of CD56 and CD16 . Functionally, distinct subsets of murine NK cells are defined by expression of CD11b and CD27 . Bovine NK cells constitutively express the natural cytotoxicity receptor NKp46 (CD335) and subsets can be differentiated by expression of CD2 [18, 19]. In a similar manner to human CD56bright  or murine CD27high NK-cell subsets , bovine CD2− NK cells readily produce IFN-γ upon cytokine stimulation and are prevalent within the LNs of healthy cattle [18, 21].
Bovine NK cells are responsive during mycobacterial infection, with in vitro evidence demonstrating restriction of M. bovis replication within macrophages, and IFN-γ secretion during interactions with APCs [22, 23]. However, our understanding of the precise roles that individual subsets play particularly during interactions with DCs is limited. We therefore investigated the individual immunological responses of bovine NK-cell subsets to M. bovis infection.
NKp46+CD2− NK cells transcribe high levels of inflam-matory and lymphoid homing chemokine receptors
NK cells were identified as a small population (1–5%) of CD3− NKp46+ lymphocytes (Fig. 1A and B). Analysis of CD2 expression confirmed the existence of two distinct subsets of NKp46+CD2+ and NKp46+CD2− NK cells (Fig. 1C). These subsets were highly purified for further analysis (Supporting Information Fig. 1).
To gain an insight into the migratory characteristics of bovine NK cells, we analysed the chemokine receptor repertoire of purified NK-cell subsets by qRTPCR [24, 25]. NK cells expressed transcripts for all measured chemokine receptors (Fig. 2). Despite evidence of variability in the chemokine receptor expression levels, no one receptor was transcribed at significantly different levels than the other receptors. In contrast, chemokine receptor expression levels varied significantly between CD2+ and CD2− NK-cell subsets (Fig. 2A and B). NKp46+CD2− NK cells transcribed significantly higher levels of CCR2, CCR5, CCR6, CCR7, CXCR3, CXCR4 and CXCR5 when compared with their CD2+ counterparts. However, significantly higher levels of CCR1, CCR8, CXCR6 and CX3CR1 were found in CD2+ NK cells. In general, CD2− NK cells transcribed higher levels of chemokine receptors with important roles in the inflammatory immune response and lymphoid homing.
Mycobacterium bovis-infected DCs selectively recruit NKp46+ CD2− NK cells
Since the chemokine receptor profiles differed between NK-cell subsets, we utilised chemotaxis assays  to investigate the capacity of M. bovis-infected DCs to recruit NK-cell subsets. Small percentages of CD2+ NK cells migrated to M. bovis-infected DCs, but this was not significant compared with the control conditions. In comparison, M. bovis-infected DCs specifically recruited CD2− NK cells with significantly higher percentages of migrated cells compared with media (p = <0.001), uninfected DCs (p = 0.004) or M. bovis alone (p = 0.042) (Fig. 2C). Furthermore, significantly higher percentages of CD2− NK cells migrated in response to M. bovis alone (p = <0.001) or M. bovis-infected DCs (p = <0.001) when compared with their CD2+ counterparts (Fig. 2C). There was no significant cell death within the migrated fractions and CD2 expression levels remained unaltered confirming selective recruitment of CD2− NK cells by M. bovis-infected DCs.
Mycobacterium bovis-infected DCs up-regulate expression of inflammatory chemokines
To establish the signals that may be associated with selective recruitment of CD2− NK cells, we analysed the expression of selected chemokines by M. bovis-infected DCs. Significant transcriptional changes were observed: inflammatory chemokines CCL3, CCL4, CCL5, CCL20 and CXCL8 were augmented by 3–6 h post infection (Fig. 3A–F). After the initial increase in transcription at 3 h post infection, levels of CCL4 and CCL5 (Fig. 3B and C) remained elevated with no further increase in expression. In comparison, transcription of CCL3 (Fig. 3A), CCL20 (Fig. 3D) and CXCL8 (Fig. 3E) continued to increase over time with levels at 12–24 h significantly augmented when compared with those at 3–6 h post infection, and also compared with uninfected DCs (Fig. 3F). The chemokines CCL2, CCL8, CXCL9 and CXCL10 were not significantly induced by M. bovis infection (up to 24 h post infection) and CCL19 was not detectable (data not shown).
Mycobacterium bovis-infected DCs differentially interact with sub-populations of NK cells
Murine NK cells are instrumental in the polarisation of Th1-mediated immunity through interactions with DCs . To extend these observations to cattle, we investigated reciprocal interactions of bovine NK-cell subsets with DCs. As a measure of activation and cytolytic potential, we assessed the expression of CD25 (Fig. 4) and perforin (Fig. 5), respectively. The vast majority of freshly isolated NK cells did not express CD25. In contrast, M. bovis-infected DCs significantly induced the activation of NK cells with increased percentages of both CD2+ (p = <0.001) and CD2− (p = <0.001) NK cells expressing CD25 (Fig. 4F and H). Analysis of the proportions of CD25 positive NK cells within each subset, expressed as a percentage of the total proportion of CD2+ or CD2− NK cells demonstrated that similar percentages of NK cells within each subset were activated. However, the MFI of CD25 expression on CD2− NK cells was significantly augmented following culture with M. bovis-infected DCs compared with control conditions (p = <0.001) and compared with CD2+ NK cells (p = <0.001) under all conditions (Fig. 4I).
Without stimulation, small percentages (5–15%) of both CD2+ and CD2− NK cells expressed perforin (Fig. 5C and H). Upon culture with uninfected DCs (Fig. 5D and H), a small increase in the percentage of NK cells expressing perforin was observed and this was significant for CD2− NK cells (p = <0.001). Furthermore, a significantly higher proportion of both CD2+ (p = 0.013) and CD2− (p = <0.001) NK cells expressed perforin in response to M. bovis alone signifying a level of activation without DC involvement (Fig. 5E and H). However, following co-culture with DCs infected with M. bovis (Fig. 5F and H) expression was highly augmented with 50–70% of both CD2+ (p = <0.001) and CD2− (p = <0.001) NK cells expressing perforin. This was coupled with a significant increase in perforin intensity for both CD2+ (p = <0.001) and CD2− (p = <0.001) NK cells when compared with all control conditions (Fig. 5I). No significant differences in perforin intensity were observed between CD2+ and CD2− subsets. Within NK cell mediated killing assays the increase in perforin expression observed following interactions with M. bovis-infected DCs was coupled with increased capacity to kill P815 target cells (Supporting Information Fig. 2).
Activated NK cells produce immunomodulatory cytokines which contribute to the early priming of the immune response. We determined the expression of cytokines following interactions of bovine NK cells and DCs. No detectable TNF-α was observed. In contrast, CD2− NK cells produced significantly elevated levels of IFN-γ but only after contact with M. bovis-infected DCs: very low levels were produced by CD2− NK cells alone or after culture with uninfected DCs or M. bovis (p = <0.001) (Fig. 6A). By contrast, CD2+ NK cells did not produce IFN-γ in response to M. bovis-infected DCs. However, this did not reflect a complete inability of CD2+ NK cells to secrete IFN-γ as high levels were secreted upon stimulation with IL-12 and IL-18 (Fig. 6A).
Differential requirement of NK-cell subsets for cell–cell contact with DCs
To determine the extent to which the interaction between NK cells and DCs was reliant on cell-to-cell contact, co-culture assays were performed with culture well inserts to prevent cell contact but to retain cytokine induced responses. Analysis of CD25, perforin and IFN-γ levels revealed a significant dependence on cell–cell contact for NK-cell activation by M. bovis-infected DCs. Separation of NK cells from M. bovis-infected DCs resulted in a significant reduction in the percentage of CD25 expressing CD2+ (p = 0.008) and CD2− NK cells (p = 0.001) (Fig. 4G and J) and significantly decreased the intensity of CD25 on CD2− NK cells (p = 0.007) (Fig. 4K). Comparison of NK-cell subsets showed that the reduction in the percentage of CD25-expressing CD2+ (40.8 ± 13.3%) and CD2− (38.7 ± 9.5%) NK cells was similar; however, there were significantly higher residual numbers of CD2− NK cells expressing CD25 (p = 0.012) compared with CD2+ NK cells (Fig. 4J). Therefore, loss of cell–cell contact had a more profound effect on the activation of CD2+ NK cells. Similarly, significantly lower percentages of perforin positive NK cells were observed after separation of M. bovis-infected DCs from CD2+ (p = <0.001) and CD2− NK cells (p = 0.045) (Fig. 5G and J). Whilst there was also a loss in the intensity of perforin expression this was not significant (Fig. 5K). Comparison of NK-cell subsets revealed that in a similar manner to CD25 expression, separation from M. bovis-infected DCs had a greater effect on CD2+ NK cells with a larger reduction (42.8 ± 4.7%) in the percentage of perforin positive NK cells (Fig. 5J).
The consequence of transwell separation on NK-cell-derived IFN-γ production was also investigated. As previously observed, CD2+ NK cells did not produce IFN-γ. IFN-γ secretion by CD2− NK cells was highly dependent on contact with M. bovis-infected DCs with a highly significant loss (97.1 ± 1.5%) (p = <0.001) of detectable cytokine upon separation (Fig. 6B).
DC-derived IL-12 is required for optimal secretion of IFN-γ by NKp46+ CD2− NK cells
DC-derived IL-12 is essential to the control of NK-cell activation. As previously reported  M. bovis-infected DCs secreted IL-12, with very low levels of IL-10. These were not significantly altered after culture with NK cells (data not shown). To investigate the extent to which IL-12 secreted by M. bovis-infected DCs contributed to the IFN-γ release by CD2− NK cells, an IL-12 specific blocking mAb  was added to co-cultures. This induced effective blocking and significantly reduced levels of IL-12 (49.3 ± 14.7% reduction). This resulted in a significant (62.5 ± 18.8%; p = 0.017) reduction in IFN-γ produced by CD2− NK cells (Fig. 6C).
NKp46+ CD2− NK cells reciprocally influence the maturation of DCs
To ascertain whether bovine NK cells were able to reciprocally activate DCs, expression of MHC class II (MHC II), CD40, CD80 and CD86 was measured. DCs were identified in co-cultures as high forward scatter cells (Fig. 7A) which expressed SIRPα (Fig. 7B). Whilst there was no significant alteration in the percentages of DCs expressing co-stimulatory molecules, M. bovis infection induced an increase in CD40 intensity (data not shown) as previously reported . These levels remained unaltered in the presence of NK cells. However, when M. bovis-infected DCs were in contact with CD2− NK cells (Fig. 7D and E; light grey bars), a proportion of DCs significantly up-regulated MHC II with increased intensity of expression compared with DCs infected with M. bovis (p = 0.042) in the absence of NK cells (DCs + M.bo, black bar), or with uninfected DCs cultured with CD2− NK cells (p = 0.002, DCs, light grey bar). This was also significant compared with the response observed when M. bovis-infected DCs were cultured with CD2+ NK cells (p = 0.044, Fig. 7E, dark grey bars) highlighting a selective reciprocal effect between NKp46+CD2− NK cells and DCs.
Understanding the early immune events which occur following exposure to infectious agents, and which are pivotal in determining whether disease or immunity ensues will pinpoint potential areas for targeted control strategies including vaccination. We analysed NK-cell subsets and defined key interactions with DCs in cattle, the natural host for M. bovis infection. Bovine TB is a disease of increasing incidence, large economic burden and with zoonotic potential; hence defining early immune mechanisms is of significant importance. In addition, a number of reports have highlighted the closely related immunology and pathogenesis of bovine and human TB (reviewed in ). Conclusions arising from studies in cattle are likely therefore to be of direct relevance to human disease.
Bovine NK-cell subsets differentially express CD2. Boysen et al.  demonstrated that the CD2−/low subset of bovine NK cells was more functionally responsive with augmented cytolytic activity, cytokine production and faster proliferation following IL-2 stimulation. As such, parallels have been drawn between bovine NKp46+CD2− NK cells and the CD56brightCD16− subset of human NK cells which demonstrate similar functionality and anatomical localities. Here, we provide novel evidence to demonstrate distinct effector responses between subsets of NK cells during mycobacterial infection, with differential requirements for DC contact or cytokine-mediated activation.
The effective containment and development of protective immunity against mycobacterial pathogens requires the recruitment of leukocytes. Chemokines are central to this process and are produced in a highly complex and redundant system that functions to orchestrate the localisation of immune effector cells bearing cognate receptors. We hypothesised that differences in chemokine receptor expression patterns would affect the capacity of NK-cell subsets to migrate towards DCs and that this would be significant in determining the in vivo interactions of these cells. Quantitative PCR methods and migration assays were utilised enabling us to correlate RNA expression profiles with functional capacity. When NK-cell subsets were compared, CD2− NK cells expressed significantly higher levels of inflammatory and lymphoid homing chemokine receptors including CCR2, CCR5, CCR6, CXCR3 and CCR7, demonstrating greater potential for migration towards sites of infection or vaccination.
CCR5 is a promiscuous receptor and expression confers responsiveness to multiple inflammatory chemokines including CCL3, CCL4 and CCL5. These chemokines are induced following M. tuberculosis infection of DCs  and within our experiments, along with CCL20, were highly up-regulated as early as 3 h post M. bovis infection. Kang et al. demonstrated an important role for CCR5 in the clustering of NK cells observed following Listeria infection of DCs resulting in early IFN-γ secretion and restriction of pathogen replication . Cella et al.  described the expression of CCR6 on lymphoid resident NK cells responsive to CCL20, a chemokine which is readily detectable during TB infection, and Martin-Fontecha et al. reported that DC-mediated recruitment of murine NK cells was dependent on CXCR3 expression with early NK-cell-derived IFN-γ essential for Th1 polarisation . Higher expression levels of CXCR3 alongside the classical lymphoid homing receptor CCR7 may explain the observation that CD2− bovine NK cells are selectively enriched within lymphoid regions .
Bovine CD2− NK cells were selectively recruited by M. bovis-infected DCs. This was associated with increased secretion of inflammatory chemokines CCL3, CCL4, CCL5 and CCL20, receptors for which were found to be highly expressed by CD2− NK cells. The involvement of chemokine-receptor pairings in NK-DC crosstalk will be fully deciphered as bovine antibodies become available.
Co-culture studies demonstrated that M. bovis-infected DCs potentiated the activation and cytolytic status of NK cells by enhancing expression of CD25 and perforin. The observed cytolytic response was not restricted to either subset of NK cells. In contrast, following interactions with infected DCs, IFN-γ was selectively produced by CD2− NK cells which also demonstrated enhanced CD25 expression. In a similar manner, CD56bright CD16− human NK cells have been shown to proliferate and secrete IFN-γ during interactions with DCs . We showed the presence of a distinct sub-population within the CD2− NK-cell subset which expressed CD25 at a higher intensity following interactions with M. bovis-infected DCs. The increased expression of CD25 would confer enhanced capacity to respond to IL-2 stimulation. Horowitz et al.  demonstrated that within an adaptive recall response, human NK-cell-derived IFN-γ secretion was partially reliant on CD4+ T-cell-derived IL-2. Discrete sub-populations of human NK cells were shown to respond to IL-2 stimulation and expand during mycobacterial infection . The augmented expression of CD25 by a sub-population of CD2− NK cells may therefore confer an enhanced capacity to respond to IL-2 stimulation enabling participation within adaptive immune responses.
Control of mycobacterial infection is reliant on the capacity of migratory DCs to prime and induce multi-functional T-cell responses. Through reciprocal interactions with DCs, a ‘helper’ role has been ascribed to NK cells whereby adaptive immune responses are indirectly modulated. In particular, DCs produced elevated levels of IL-12, increased expression of CCR7 and induced stronger T-cell-mediated IFN-γ responses following contact with NK cells [6, 8]. The depletion of human NK cells prior to M. tuberculosis infection reduced the cytolytic capacity and frequency of IFN-γ producing CD8+ T cells . In addition, DCs from mice lacking NK cells demonstrated reduced capacity to directly prime or induce Ag-specific CD4+ T-cell responses as a consequence of aberrant maturation and reduced IL-12 secretion . We demonstrated that during M. bovis infection, CD2− NK cells selectively enhanced the maturation of DCs with elevated cell surface expression of MHC II. A number of studies have shown that activated NK cells select more immunogenic DCs during a protective immune response through selective killing of immature DCs (reviewed in ). Recently, Morandi et al.  provided evidence for perforin-dependent in vivo editing of DCs by NK cells. Importantly, the remaining DCs displayed significantly enhanced capacity for T-cell stimulation . Thus, the selective killing and maturation of sub-populations of DCs by NK cells, whilst contributing to the control of infection, may also translate into an augmented capacity to present Ag and influence T-cell responses. Further studies are underway in our laboratory to determine whether bovine NK-cell subsets differ in their capacity to kill M. bovis-infected DCs.
Deciphering the precise components of NK cell and DC interaction has revealed roles for soluble stimuli alongside contact-dependent receptor–ligand interactions. Bastos et al.  demonstrated that NK-cell-derived IFN-γ secretion was reliant on contact with BCG-exposed DCs in addition to sub-optimal concentrations of IL-18. We showed herein that NK-cell subsets possess differential requirements for soluble and contact-mediated signals in the activation of effector responses. Thus, CD2− NK cells were reliant on both cytokines (IL-12) and contact-dependent stimuli from M. bovis-infected DCs to produce IFN-γ whilst retaining CD25 and perforin expression with cytokines alone. In comparison, the CD2+ NK cells were unable to secrete IFN-γ and were entirely dependent on cytokines and cell–cell contact to become activated and maintain cytolytic capacity. Emerging evidence suggests that such variations in NK-cell effector responses may in part be explained by the existence of a threshold of activation determined by the balance of expression between activatory and inhibitory receptors . Cattle are the only species outside of primates to have expanded polymorphic KIR genes as well as functional polymorphic Ly49 [41, 42]. Our preliminary (unpublished) data showed differential expression levels of Ly49 receptor expression between bovine NKp46+ CD2− and CD2+ NK cells that may partially explain differences in functionality between these subsets. Additional studies on the KIR repertoire and expression of other important receptors including NKp30 will provide further insight and identify factors that govern responsiveness of NK-cell subsets during mycobacterial infection.
As a consequence of the growing body of evidence implicating a role for NK cells in the development of adaptive immunity, the potential for targeting the NK-DC interaction for vaccine-induced protective immunity is gaining precedence [38, 43, 44]. This study provides evidence that the precise roles of NK-cell subsets during interactions with DCs differ substantially as a consequence of distinct requirements for DC-derived cytokines and contact-mediated interactions. Notably, M. bovis-infected DCs preferentially recruited NKp46+CD2− NK cells which selectively produced IFN-γ and reciprocally augmented the maturation of DCs. The capacity to modulate adaptive immune responses during M. bovis infection may therefore be specific to this sub-population of NK cells. Bovine NKp46+CD2− NK cells may therefore represent a novel target by which Th1-mediated immune responses may be potentiated in the induction of protective immunity to mycobacterial infection.
Materials and methods
Experiments were carried out on British Holstein-Friesian calves housed at the Institute for Animal Health according to Home Office guidelines and with full ethics approval. The herd has been certified free of bovine TB for over 10 years. Animals were below 6 months of age and age-matched within experimental groups. Blood was collected by jugular venepuncture into sodium heparin (Leo Pharma, UK).
Purification of NK-cell subsets
NK cells were positively selected from PBMCs by labelling with mAb recognising bovine NKp46 (anti-CD335; AbD Serotec, UK) and Rat Anti-Mouse IgG1 MicroBeads (Miltenyi-Biotech, UK) to purities >97%. To isolate NK-cell subsets (NKp46+ CD2+ and NKp46+ CD2−), NKp46+ cells were labelled with the CD2 specific mAb IL-A43  coupled to IgG2a-FITC (Zenon® labelling kit, Invitrogen, UK) and purified using the FacsAria™ (BD, Germany) to >98% (Supporting Information Fig. 1)
Culture of monocyte-derived DCs
Monocytes were isolated from PBMCs with anti-human CD14 beads (Miltenyi Biotech)  and cultured for 3 days at 37°C with 5% CO2 in media (RPMI 1640 – Gibco, UK), FCS (TCS Cell Works, UK), 2-mercaptoethanol (Sigma, UK) and gentamicin (Sigma) supplemented with recombinant bovine GM-CSF and IL-4 (Dr G. Entrican, Moredun Research Institute) to obtain DCs.
Isolation of RNA and synthesis of cDNA
Cells were lysed with 4 M guanidine thiocyanate prior to RNA extraction using RNeasy kits (Qiagen, Germany). Contaminating DNA was removed using DNAfree kits (Ambion, UK) and RNA reverse transcribed using SuperScript® III Reverse Transcriptase (Invitrogen). Quality and concentration was assessed using the NanoDrop spectrophotometer (Thermo Scientific, USA).
Quantitative real-time PCR
The primer and probe sequences for bovine chemokines and receptors have previously been reported [24-26]. For CCL20, we used the following:
5′-TCAGAAGCAAGCAACTTCGAC-3′, Rev 5′- GATGTCACA GGCTTCATTGG-3′, Probe 5′-CCAGCTGCTGTGTGAAGCCCA-3′. Quantitative real-time PCR analysis was carried out using the TaqMan FAST Universal PCR Master Mix on an ABI Prism 7500 Fast Real-Time PCR System (Applied Biosystems, UK). All results were quantified against a plasmid DNA standard curve and data normalised to the housekeeping gene RPLPO (bovine homologue for large fragment of RNA polymerase)  .
Mycobacterial culture and infections
All experiments utilising M. bovis were conducted under ACDP CL3. Cultures of M. bovis AF2122 were obtained as previously described . Titres were calculated by serial dilution on Middlebrook 7H11 agar plates.
DCs were infected with M. bovis (MOI 1) within 24-well chemotaxis plates (Corning Life Sciences, UK). After 18 h, purified NK cells were placed on top of the transwell membrane (5 μm/6.5 mm) suspended above the infected DCs and incubated at 37°C + 5% CO2 for 4 h. Following incubation, the migrated and input cells were diluted to equivalent concentrations and fixed with 2% paraformaldehyde. Cells were counted using the FACSCalibur flow cytometer (BD). Cell death was determined by labelling with propidium iodide (PI) prior to fixation.
Co-cultures of NK-cell subsets and DCs
DCs were infected with M. bovis (MOI 1:1) and purified NK cells added to the co-culture (NK:DC ratio 5:1). Where indicated, anti-IL-12 mAb CC326  was added at pre-determined optimal concentrations. In contact-dependent assays, NK cells were separated from infected DCs using a transwell insert (0.2 μm) (Nunc, Denmark). Recombinant bovine IL-12  and human IL-18 were included at pre-determined optimal concentrations as a positive control. Supernatants and/or cells were retrieved following 18 h incubation at 37°C + 5% CO2. ELISAs were conducted to detect the presence of bovine IFN-γ, TNFα, IL-12 and IL-10 .
NK-cell-mediated killing assays
Murine P815 cells were used as targets . Following 18 h of co-culture with M. bovis-infected DCs, NK cells were incubated with CFDA-SE (Invitrogen) labelled P815 cells at effector target ratio 12.5:1. Cells were incubated for 3 h at 37°C + 5% CO2. Killing was analysed by labelling with PI and analysis of CFDA-SE/PI expression by flow cytometry.
Multi-colour immunofluorescent labelling
NK-cell subsets (NKp46+ CD2+ and NKp46+ CD2−) were identified by labelling with CD335-PE (Serotec) and CC42-AF647 (conjugated using AF647, Invitrogen). CD25 expression was determined by labelling with mAb CACT108A (VMRD, Washington) transiently coupled with IgG2a-FITC (Zenon® labelling kit, Invitrogen) and comparing with isotype control mAb AV37 (anti-chicken, Dr C Butter, IAH). For intracellular staining of perforin, cells were pre-incubated with media supplemented with Brefeldin A (Sigma) and expression determined by labelling using human perforin-FITC (BD Pharmingen) . DCs (high forward scatter cells, SIRPα+) were identified by labelling with mAb CD172a 647/PE (AbD Serotec). Cell surface expression of MHC II (CC158 – IAH), CD40 (IL-A156), CD80 (N32/52–3) and CD86 (IL-A190) (ILRI Kenya, N McHugh) was compared with isotype control mAbs AV37 and AV20 (anti-chicken, Dr C Butter) using fluorochromes IgG2a-FITC and IgG1-FITC (Southern Biotech, USA). Flow cytometric analysis was conducted using the FACSCalibur (BD) and collection of a minimum of 20 000 events.
Data were analysed using; Excel 2007 (Microsoft), FCS Express v3 (De Novo Software) and 7500 Fast SDS Software (Applied Biosystems). Statistical analysis was carried out using Minitab v16 (Minitab Inc., UK). Data were transformed appropriately prior to assessment for normality and variance. The statistical methods employed included paired t-tests, 1-way ANOVA's and general linear models, with Tukey's post hoc comparisons at 95% confidence intervals. p values <0.05 were considered significant.
The authors gratefully acknowledge the skilled assistance of the animal services team at the Institute for Animal Health. We are very grateful to Dr. Jenny Piercy and Professor Tracey Coffey for assistance with qPCR assays for bovine chemokines. The work was funded by the Biotechnology and Biological Sciences Research Council, UK.
Conflict of interest
The authors declare no financial or commercial conflict of interest.