Vγ2Vδ2 T-cell co-stimulation increases natural killer cell killing of monocyte-derived dendritic cells

Abstract

Interactions between natural killer (NK) cells and dendritic cells (DC) affect maturation and function of both cell populations, including NK cell killing of DC (editing), which is important for controlling the quality of immune responses. We also know that antigen-stimulated Vγ2Vδ2 T cells co-stimulate NK cells via 4-1BB to enhance the killing of tumour cell lines but we do not know what regulates 4-1BB expression or whether other NK effector functions including DC killing, might also be influenced by NK–γδ T-cell cross-talk. Here we show that antigen-stimulated γδ T cells co-stimulate NK cells through inducible T-cell co-stimulator (ICOS)– ICOS ligand (ICOSL) and this signal increases NK cell killing of autologous DC. Effects of NK–γδ T-cell co-culture, which could be reproduced with soluble ICOS-Fc fusion protein, included increased CD69 and 4-1BB expression, interferon-γ, tumour necrosis factor-α, macrophage inflammatory protein-1β, I-309, RANTES and sFas ligand production, as well as elevated mRNA levels for co-stimulatory receptors OX40 (TNFRSF4) and GITR (TNFRSF18). Hence, ICOS–ICOSL co-stimulation of NK by Vγ2Vδ2 T cells had broad effects on NK phenotype and effector functions. The NK–γδ T-cell cross-talk links innate and antigen-specific lymphocyte responses in the control of cytotoxic effector function and DC killing.

Introduction

The Vγ2Vδ2 (γδ) T-cell receptor found only in human and nonhuman primates, recognizes non-peptidic antigens^1,2 in the absence of MHC restriction.³ Effector functions of γδ T cells include rapid cytokine production (reviewed in ref. 4), direct killing of infected or malignant cells (reviewed in ref. 5), and antibody-dependent cellular cytotoxicity.^6–8 Rapid and potent γδ T-cell responses reflect positive selection for Vγ2-Jγ1.2 rearrangements (designated Vγ9-JγP in an alternate nomenclature),^9,10 which are responsible for recognizing ubiquitous phosphoantigens.¹¹ Circulating Vγ2-Jγ1.2 Vδ2 T cells are mostly central or effector memory cells¹² as a result of chronic antigen exposure.

The capacity for rapid, nearly homogeneous responses to phosphoantigens using rearranged T-cell receptors gives γδ T cells the characteristics of both innate and acquired immunity.¹³ By secreting interferon-γ (IFN-γ), tumour necrosis factor-α (TNF-α) and other cytokines, Vγ2Vδ2 cells help to polarize adaptive T-cell responses^14–16 and promote dendritic cell (DC) maturation.¹⁷ A fraction of γδ T cells provides help to B cells^18–20 and some Vγ2Vδ2 cells can present or cross-present antigens.^21,22 The γδ effector functions are modulated by invariant receptors including NK cell receptors and Killer immunoglobulin-like receptors;^23–27 Fcγ receptor IIIa expression makes them potent effector cells for antibody-dependent cellular cytotoxicity.^7,8 These activities support host immunity against microbial pathogens and cancer⁵ but the full potential of γδ T cells, especially their role(s) in immune regulation, are less known.

We reported previously that direct contact of γδ T cells with natural killer (NK) cells involved the co-stimulatory receptor 4-1BB (CD137) and increased NK cytolysis of tumour cell targets.²⁸ This interaction suggested that antigen-specific responses, such as phosphoantigen stimulation of γδ T cells, may be involved in regulating NK cell effector activities.

Much is known already about NK–DC interactions and how they control immunity. Cross-talk between NK cells and DC depends on the activation status and abundance of each cell type.^29–31 Immature DC activate licensed NK cells through cognate receptor interactions^29,31 and release of soluble factors including interleukin-18 (IL-18).³² In turn, activated NK cells induce DC maturation or kill immature DC in a mechanism termed ‘editing’.^29–31,33 A low ratio of NK : DC favours DC maturation,³¹ which is partly mediated by alarmin HMGB1 released from NK cells,³² whereas a high NK : DC ratio promotes DC editing,³¹ which depends on NKp30²⁹ and the TNF-related apoptosis-inducing ligand (TRAIL)/DR4 pathway.³⁴ Mature DC resist NK killing because they express high levels of MHC Class I,^29,35 which vetoes NK cell recognition. Hence, editing mechanisms select highly immunogenic, mature DC in vivo.³⁶ In HIV⁺ patients the editing process is impaired,^34,35 leading to an accumulation of tolerogenic DC in lymph nodes and rising levels of IL-10 in blood.³⁵ Aberrant maturation of DC³⁵ that resist NK editing³⁴ has been linked to generalized immune dysregulation in HIV disease. Poor DC editing is paralleled by Vγ2Vδ2 T-cell depletion in all HIV⁺ individuals (reviewed in refs ^37,38). It is important to describe NK–γδ T-cell interactions in greater detail to learn how the profound loss of γδ T-cell function affects key mechanisms of innate immunity.

Materials and methods

Blood collection and peripheral blood mononuclear cell isolation

This study was approved by the University of Maryland Institutional Review Board. Peripheral blood was obtained from healthy adult volunteers after written, informed consent. Whole blood was diluted with PBS (Lonza, Walkersville, MD) and layered over Ficoll–Hypaque (GE Healthcare, Uppsala, Sweden) density gradients to isolate peripheral blood mononuclear cells (PBMC). Cell viability was assessed by Trypan Blue dye exclusion.

γδ T-cell expansion

To expand Vγ2Vδ2 T cells (γδ), PBMC at 10⁶ cells/ml were stimulated with 0·5 μm Zoledronate (Zometa; Novartis Oncology US, East Hanover, NJ) plus 100 U/ml of human recombinant IL-2 (rIL-2) (Tecin, Biological Resources Branch, NIH, Bethesda, MD). Treated PBMC were cultured at 37° with 5% CO₂ for 14 days in RPMI-1640 medium (GIBCO, Grand Island, NY) supplemented with 10% fetal bovine serum (GIBCO), 1% l-glutamine, 100 U/ml penicillin and 100 U/ml streptomycin (GIBCO). Interleukin-2 was added to γδ cultures on days 3, 7 and 10. Fourteen days after stimulation, 10 U/ml rIL-2 was added and cells were rested with this low concentration of IL-2 for 2 days. On day 16, lymphocytes were harvested and the percentage of γδ T cells was measured by flow cytometry. The proportion of γδ lymphocytes in Zoledronate-expanded cultures ranged between 70% and 85%; cells were not purified further before co-culture with NK cells.

NK cell isolation

Autologous NK cells were isolated from PBMC by magnetic bead separation using the MACS NK cell negative selection kit (MiltenyiBiotec, Auburn, CA) according to the manufacturer’s instructions. NK cell purity, measured by flow cytometry, was always > 95%.

NK–γδ T-cell co-culture

Twenty-four-well tissue culture plates were coated overnight at 4° with human IgG1 (10 μg/well) (Sigma, St Louis, MO) diluted in PBS (Lonza). After washing the wells once with PBS, purified NK cells and autologous expanded γδ T cells were co-cultured for 20 hr at a 1 : 1 ratio (1·5 × 10⁶ cells of each type) in 1 ml of complete RPMI. NK or γδ T cells alone were cultured at 3 × 10⁶ cells/well. In selected experiments, IL-2 (100 U/ml) or soluble human inducible T-cell co-stimulator (ICOS) -Fc chimera (sICOS) (10 μg/ml) (Sino Biological, Beijing, China) was added to the NK–γδ co-culture, NK, or γδ T cells. Soluble human IgG1 (10 μg/ml) was added to control wells. After 20 hr of culturing, supernatants were collected from NK, γδ T or mixed cells, and used for cytokine analyses. Viable cells were counted using the Trypan Blue dye exclusion method and analysed by flow cytometry for activation and co-stimulatory markers or used as effectors in cytotoxicity assays with autologous DC. NK cells (NK*) isolated from NK–γδ co-cultures by the MACS negative selection kit were used as cytotoxic effectors for DC killing.

Phenotyping studies

NK, γδ or NK–γδ co-cultures (3 × 10⁵ cells) were resuspended in RPMI−10% fetal bovine serum and stained at 4° with directly conjugated monoclonal antibodies. After 15 min, cells were washed with PBS and resuspended in PBS with 1% paraformaldehyde. Lymphocytes (3 × 10⁴; gated on the basis of forward and side scatter profiles) were collected for each sample on a FACSCalibur (BD Biosciences, San Jose, CA) and results were analysed with FlowJo software (Tree Star, Ashland, OR).

The following antibodies were purchased from BD Biosciences: anti-human Vδ2 (clone B6), anti-human CD3 (clone SP34-2), anti-human CD3 (clone UCHT1), anti-human CD69 (clone FN50) mouse anti-human 4-1BB/CD137 (clone 4B4-1), mouse anti-human 4-1BBL/CD137L (clone C65-485), CD16 (clone B73.1), and matching isotype controls. The following antibodies were purchased from eBioscience (San Diego, CA): anti-human ICOS/CD278 (clone ISA-3), anti-human B7-H2/ICOSL/CD275 (clone MIH12), anti-human CD80 (clone 2D10.4), anti-human CD86 (clone IT2.2), anti-human CD28 (clone CD28.2), anti-human PD-1/CD279 (clone eBioJ105), anti-human B7-H1/PDL-1/CD274 (clone MIH1), anti-human HVEM/LIGHTR/CD270 (clone eBioHVEM-122), anti-human LIGHT/HVEML/CD258 (clone 7-3), anti-human BTLA/CD272 (clone MIH26). Anti-human CD56 (clone N901) was purchased from Beckman Coulter (Indianapolis, IN).

Cytotoxicity

Cytolytic NK and γδ T-cell activity was measured by 4-hr ⁵¹Cr-release assays. Human monocytes purified by negative selection (EasySep; StemCell Technologies, Vancouver, BC, Canada) were differentiated into immature IL-4 DC or IL-15 DC as described previously,³⁹ then used as autologous targets for cytotoxicity. DC (5 × 10⁵ cells) were labelled with 50 μCi ⁵¹Cr (Na₂⁵¹CrO₄; Perkin Elmer, Waltham, MA) for 1 hr at 37° in 5% CO₂. After incubation, DC were washed twice with RPMI and incubated for an additional 30 min at 37° in 5% CO₂ to reduce spontaneous release. DC were then plated at 10⁴ cells per well in a volume of 100 μl using 96-well V-bottom plates. Effector cells consisted of CD3⁻ CD56⁺ NK, NK* (NK purified from co-cultures) NK–γδ co-cultures (mixed effectors), and expanded γδ T cells. Effector cells (2 × 10⁵ cells) were added to autologous DC in triplicate wells to give a final volume of 200 μl per well and an effector : target ratio of 20 : 1. DC in medium alone were used to determine spontaneous ⁵¹Cr release. Maximum release was determined by lysing DC in medium with 2% Triton X. After 4 hr of incubation, 50 μl supernatant was harvested and mixed with 150 μl Optiphase Supermix Scintillant (Perkin Elmer) in a 96-well disposable gamma-plate (Perkin Elmer). The ⁵¹Cr was measured in a Wallac Micro-β gamma counter. Specific lysis was calculated using the formula: 100 × (experimental release − spontaneous release)/(maximum release − spontaneous release).

Cytokine and chemokine levels

Supernatants collected from NK–γδ co-cultures, NK cell, or γδ T-cell cultures were analysed for cytokines (IFN-γ, TNF-α, IL-4, IL-10, IL-13), chemokines [macrophage inflammatory protein-1β (MIP-1β), regulated on activation, normal T cell expressed and secreted (RANTES/CCL5), I-309/CCL-1, macrophage-derived chemokine (MDC/CCL22)] and apoptosis-related factors (sFAS, sFASL, TRAIL) using Luminex technology (EMD; Millipore, Billerica, MA).

Gene expression studies

RNA was extracted from 1 × 10⁶ to 2 × 10⁶ NK cells using an RNeasy mini Kit (Qiagen, Valencia, CA), as described by the manufacturer. Before RNA extraction, NK cells were incubated for 20 hr in IgG1-coated plates with sICOS (10 μg/ml) or sIgG1 (10 μg/ml). RNA was reverse transcribed with an oligo-dT primer containing the T7 RNA polymerase promoter. The cDNA was used for in vitro transcription in the presence of biotin-NTP to generate single-stranded RNA (cRNA). After hybridization the biotinylated cRNA was detected with streptavidin-Cy3 and quantified using an iScan System scanner. The image data were compiled with GenomeStudio software, version 1.0 (Illumina, San Diego, CA). All data were log₂ transformed and used to calculate differences in signal intensities and significance of the data.

Statistical analysis

Statistical analyses were performed using the software GraphPad Prism (GraphPad Software, La Jolla, CA). For each measured variable, a D’Agostino & Pearson omnibus normality test was performed to assess whether values were normally distributed. Differences between means or medians of paired groups were evaluated using a paired t-test or a one sample t-test for normally distributed variables, and Wilcoxon rank test for non-Gaussian distributions.

Results

NK-γδ T-cell co-culture increases expression of activation and co-stimulatory molecules

γδ T cells in PBMC were expanded (using Zoledronate plus IL-2) for 14 days and rested for two additional days. Zoledronate (an aminobisphosphonate drug) inhibits farnesyl synthase in myeloid cells, causing accumulation of the stimulatory phosphoantigen isopentenyl pyrophosphate.⁴⁰ Cultures with > 70% γδ T cells after expansion were used for co-culture experiments. NK cells were isolated from autologous PBMC by negative selection using magnetic beads. CD56⁺ CD3⁻ (NK) cells were > 95% pure after magnetic bead separation (see Supporting information, Fig. S1). NK and γδ T cells in a ratio of 1 : 1, were co-cultured for 20 hr in plates pre-coated with human IgG1. After co-culture with γδ T cells, 70% of NK (NK*) cells expressed CD69 compared with only 33% of NK cells incubated in IgG1-coated plates without γδ T cells (Fig.1a,b). NK cells activated with IL-2 (instead of immobilized IgG1) also up-regulated CD69 (64%, Fig.1a,b) and co-culture with γδ cells further increased CD69 expression (Fig.1a,b). More than 80% of γδ cells expressed CD69 in all conditions, because of the previous zoledronate/IL-2 stimulation (Fig.1a).

Figure 1.

Natural killer (NK)–γδ T-cell cross-talk enhances NK cell activation and up-regulates 4-1BB expression. Expression of CD69 or 4-1BB was detected after 20 hr of incubation with plastic-immobilized IgG1 (or interleukin-2) for NK cells, γδ cells cultured alone, or NK and γδ cells after co-culture (NK* and γδ*). (a) Dot plots show CD69 and 4-1BB expression profiles for a representative donor. (b–d) Box plots compare 25th, 50th, 75th centiles, and mean values (n = 6 to n = 11) after culture for (b) CD69 expression on NK cells, (c) 4-1BB expression on NK cells and (d) 4-1BB expression on γδ cells. (e) The dot plot shows 4-1BB expression relative to CD56 expression for γδ* cells (one donor representative of 11). Differences between groups were analysed by a paired t-test or Wilcoxon matched-pairs signed rank test. *P < 0·05; **P < 0·01; *** P < 0·001.

Consistent with previous results,²⁸ the percentage of NK cells expressing 4-1BB increased from approximately 10% up to 34% when NK were co-cultured with γδ T cells (Fig.1c). Additionally, the percentage of γδ T cells expressing 4-1BB increased from 16% to 22% after co-culture (γδ*). Fewer γδ* T cells expressed 4-1BB after co-culture compared with NK* (P = 0·005). Within the γδ* population, CD56⁺ cells preferentially expressed 4-1BB (Fig.1e). 4-1BB expression required exposure to immobilized IgG1; NK or γδ cells did not up-regulate 4-1BB alone or in co-culture, when treated with IL-2 in the absence of IgG1 (Fig.1a,c,d).

Effect of NK-γδ T lymphocyte co-culture on immune regulatory factors

We asked whether NK–γδ T-cell co-culture altered the expression of cytokines (including IFN-γ, TNF-α, IL-4, IL-10), chemokines (including MIP-1β, RANTES, I-309, MDC) or apoptosis-related factors (FAS, FASL, TRAIL). γδ T-cell cultures accumulated higher levels of soluble factors compared with control NK cells (Table1). NK-γδ co-cultures contained levels of IFN-γ, TNF-α and MIP-1β similar to γδ control cultures (Table1), despite having only half the numbers of γδ cells. Other factors, including the chemokines I-309 (CCL1) and MDC, were released in co-cultures at half the amounts secreted by γδ T cells alone. When we normalized soluble factor levels according to the γδ cell count, IFN-γ, TNF-α and MIP-1β levels were higher in co-cultures than in control γδ cultures (Fig.2a); similar results were obtained for soluble FASL. Normalized levels of I-309 (as well as MDC, RANTES, FAS and TRAIL, data not shown) were the same in co-cultures and control γδ cells. T helper type 2 cytokines IL-4 (Fig.2a) and IL-13 (data not shown) were lower in co-cultures compared with γδ T cells cultured alone, suggesting that co-culture suppressed T helper type 2 cytokine responses. In most donors, IL-10 was below the limits of detection in control cultures, and was present only at low levels in co-cultures (data not shown).

Table 1.

Release of soluble factors (pg/ml) after culture with immobilized IgG1 (n = 6 to n = 10)

	IFN-γ	TNF-α	MIP-1β	FASL	I-309
NK	215·8 ± 215·6	135 ± 80·9	2845 ± 1129	69·1 ± 49·3	8·6 ± 7·9
NK-γδ	7337 ± 2791	1878 ± 791·1	9679 ± 1720	263·6 ± 143·8	1573 ± 979
γδ	7818 ± 3069	2273 ± 807	8401 ± 2458	282·2 ± 153·2	3594 ± 2002*

Natural killer (NK), γδ T or 1 : 1 ratio NK–γδ cells were cultured for 20 hr in plates coated with human IgG1. Differences between NK–γδ co-cultures and control γδ cells were analysed by a paired t-test or Wilcoxon matched-pairs signed rank test

^*P < 0·05.

IFN-γ, interferon-γ; TNF-α, tumour necrosis factor-α; MIP-1β, macrophage inflammatory protein 1β; FASL, FAS ligand.

Figure 2.

Natural killer (NK)–γδ T-cell co-culture enhances cytokine production and NK cytotoxicity against autologous immature dendritic cells (DC). (a) Soluble factors were measured by luminex technology in culture supernatants. For each donor (n = 6 to n = 10), soluble factor levels in γδ cultures (γδ) or NK–γδ co-cultures (γδ*) were normalized according to the number of γδ cells, and paired values (γδ* versus γδ) were compared using a single sample t-test (H₀: for γδ* ÷ γδ, μ₀ = 1). (b) Cytotoxicity against interleukin-4 (IL-4) or IL-15 DC with NK, NK* (purified after co-culture with γδ T) NK–γδ co-cultures (mixed effectors), and γδ cells tested as effectors in Cr release assays. Box plots show 25th, 50th, 75th centiles, and mean specific lysis for each type of effector cell (n = 6 to n = 8). Differences between groups were analysed by a paired t-test or Wilcoxon matched-pairs signed rank test. *P < 0·05; **P < 0·01.

γδ T cells co-stimulate NK cell cytotoxicity

We next tested whether culture conditions affected NK cell cytotoxicity. After 20 hr of co-culture, NK cells were negatively selected (NK*) and used as effectors in a cytotoxicity assay (4-hr Cr release) with autologous monocyte-derived DC as targets. As controls, we used NK cells cultured alone or NK–γδ co-culture cells (mixed). DC were generated in the presence of granulocyte–macrophage colony-stimulating factor with either IL-4 (IL-4 DC) or IL-15 (IL-15 DC). We reported previously that co-culture with γδ T cells enhanced NK cell cytotoxicity against tumours.²⁸ Co-culturing γδ and NK cells also increased NK-mediated killing of autologous immature DC (Fig.2b). Interestingly, NK* or NK–γδ co-culture cells killed DC to comparable extents (Fig.2b). Either the number of effectors was at a plateau with the 1 : 20 E : T ratio or co-cultured γδ T cells also acquire the capacity to kill autologous DC. To test whether γδ T cells have increased lytic effector function after co-culture, it would be necessary to define γδ purification schemes that do not use anti-T-cell receptor or anti-CD3 antibodies, which super-activate expanded γδ lymphocytes.

Dendritic cell phenotype and function vary depending on whether they are differentiated with IL-15 or IL-4.⁴¹ The DC differentiated with IL-15 produce more IL-17-promoting factors, including IL-1β, IL-6 and IL-23 in response to Toll-like receptor agonists.³⁹ When comparing cytolysis of DC differentiated with IL-15 or IL-4, both NK* cells and co-culture (mixed) effectors killed IL-4 DC or IL-15 DC targets to similar extents (Fig.2b). Immature DC were more susceptible to NK* cytolysis compared with mature DC (data not shown). Preliminary experiments with γδ* effectors (γδ cells negatively purified after co-culture) show that they kill immature and mature DC with equal potency, but are more cytotoxic against IL-15 DC (data not shown). This is not surprising, since IL-15 DC produce more IL-15, which enhances γδ T cell cytotoxicity.⁴²

ICOS-ICOSL are important for NK–γδ cross-talk

Co-stimulatory molecules expressed by NK or γδ T cells are important for effector activation. We focused on the ICOS that was expressed on pre-activated γδ T cells (Fig.3), and ICOSL that was expressed on NK after treatment with immobilized IgG1. When NK cells were cultured with autologous γδ T cells, ICOS and ICOSL levels were increased in two of four donors (γδ versus γδ*, and NK versus NK*, Fig.3a,b). We also identified a subset of expanded γδ T cells that expressed CD16, the low-affinity Fcγ receptor III A/B on both CD56⁺ and CD56⁻ subpopulations (Fig.3c). Four of all the donors used in this study were tested for CD16 expression on γδ T cells (Fig.3d). The percentage of positive cells varied from 26% (donor 3) to 47% (donor 1) with a mean of 34%. Knowing the expression of CD16 is important because the soluble ICOS reagent is an Fc fusion protein and could bind and signal through CD16 present on γδ T or NK cells in culture. We monitored the expression of several co-stimulatory molecules (Fig.3e) that may be relevant for NK-γδ cross-talk, but we did not perform functional tests to assess their roles in co-stimulation and the control of effector functions.

Figure 3.

Expression of co-stimulatory molecules after culture. Histograms show results for one representative donor of four that were studied (a, b). Cells were stained for (a) inducible T-cell co-stimulator (ICOS) or (b) inducible T-cell co-stimulator ligand (ICOSL) on natural killer (NK) or γδ cells or NK-γδ cell co-culture (NK* and γδ*) after incubation on IgG1-coated plates. Expanded γδ T cells were stained for CD56 and CD16 (Fcγ receptor III A/B) (c, d). After gating on Vδ2 for two representative donors, a substantial fraction of cells expressed CD16 (c). Within the CD16⁺ subpopulations, the majority of cells also expressed the cytotoxicity marker CD56 (NCAM-1). The proportion of total Vδ2 T cells that expressed CD16 for four donors used in this study (d) ranged between 26% and 47%. NK* and γδ* expressed multiple co-stimulatory molecules (e) that have been detected but not yet tested for functional significance.

We then tested whether ICOS-ICOSL binding was involved in NK–γδ cross-talk. We added soluble, recombinant ICOS-Fc chimera (sICOS) to NK cells, γδ cells or NK–γδ co-cultures. We expected this reagent to compete for cell surface ICOSL and block co-stimulation. Surprisingly, sICOS up-regulated CD69 and 4-1BB on NK cells, alone or in co-culture (Fig.4a,b). It is possible that sICOS delivered a co-stimulatory signal to NK cells or blocked an unknown negative signal. Consistently, sICOS did not affect 4-1BB levels on γδ cells (Fig.4b), because they do not express ICOSL (Fig.3b). As NK cells and γδ T cells express Fcγ receptor IIIA/B (CD16), ICOS Fc might stimulate either cell type through this activating receptor and this is a technical challenge in proving the role for ICOS–ICOSL signalling in our co-culture experiments.

Figure 4.

Soluble inducible T-cell co-stimulator (sICOS) increases CD69 or 4-1BB expression and soluble factor production by natural killer (NK) cells. (a) Proportions of CD69 or 4-1BB expressing cells were determined for NK or γδ cells cultured alone or after NK–γδ co-culture (NK* and γδ*), in the presence of sICOS (10 μg/ml) or soluble IgG1 (10 μg/ml). Box plots show 25th, 50th, 75th centiles, and mean values for each culture condition (n = 7). Differences between groups were analysed by the Wilcoxon matched-pairs signed rank test. *P < 0·05. (b) Soluble factor levels were measured (pg/ml) for NK cells cultured in IgG1 coated plates with ICOS (10 μg/ml) or soluble IgG1 (10 μg/ml). Box plots show 25th, 50th, 75th centile, and mean values for six soluble factors detected with cells from six independent donors. Differences between groups were analysed by Wilcoxon matched-pairs signed rank test. *P < 0·05. (c) Differences in soluble factor levels for NK ICOS versus NK control (IgG1 only) were evaluated as a ratio (NK ICOS : NK control) for each donor (n = 6), then analysed with a one-sample t-test (H₀ for NK ICOS : NK control, μ₀ = 1). A bar graph shows mean values and standard deviations for six soluble factors.

Our results also show that sICOS triggers reverse signalling though ICOSL, thus enhancing the NK cell activation induced by immobilized IgG1 and contributing to NK–γδ T-cell cross talk. Soluble ICOS treatment increased cytokine/chemokine release by NK cells alone (Fig.4c,d), while γδ control cells and co-culture cells were not affected (see Supporting information; Fig. S2). We suspect that pre-activated γδ T cells do not require NK co-stimulation to secrete soluble factors, so sICOS would minimally (if at all) affect their ability to produce cytokines in co-culture. NK cells require γδ co-stimulation, but sICOS behaves like an agonist, enhancing rather than blocking NK-mediated cytokine release. Overall there was no measurable effect of sICOS on cytokine production by co-culture cells.

Gene expression profiling by microarray analysis revealed that sICOS up-regulated mRNA for NK genes related to cell activation (Table2). Consistent with previous reports, NK cells treated with sICOS had higher transcription levels of OX40⁴³ and GITR⁴⁴ mRNA, two members of the TNF receptor superfamily, suggesting the potential for a cascade of co-stimulatory interactions among NK and γδ T cells beyond the 4-1BB and ICOS-dependent steps we tested. Also TIM-3, a receptor associated with NK-cell activation⁴⁵ or maturation,⁴⁶ which can increase IFN-γ secretion upon binding galectin-9,⁴⁷ was higher after sICOS treatment. Adding sICOS also elevated mRNA for PIM1, a kinase that phosphorylates the RelA/p65 subunit of nuclear factor-κB, prevents its degradation and stabilizes nuclear factor-κB-mediated signalling.⁴⁸ Finally, I-309 (CCL1) gene expression was increased by sICOS, consistent with the increased I-309 release measured by ELISA. Conversely, TNF and MIP-1β transcripts were both lower after sICOS treatment, even though the levels of these soluble factors in NK cell supernatants were higher after sICOS treatment than in controls. It is possible that TNF and MIP-1β transcription was suppressed while stored cytokine continued to be released, a phenomenon we observed for RANTES in freshly activated γδ T cells.⁴⁹ suppressor of cytokine signaling 1, which was up-regulated in two of four donors, may support negative feedback loops where secreted cytokines negatively regulate their own mRNA accumulation.^50,51

Table 2.

Differential gene expression in sICOS-treated natural killer cells

Gene ID	Alternate name	Accession number	Fold increase1	P-value	Donors2
PIM1		NM_002648	2	8·22E-05	3/4
CCL1	I309	NM_002981	2·09	4·50E-07	3/4
TNFRSF4	OX40	NM_003327	2·2	8·72E-05	4/4
TNFRSF18	GITR	NM_005092	2·51	1·39E-06	4/4

Gene ID	Alternate name	Accession number	Fold decrease	P-value	Donors
CCL2	MCP1	NM_002982	3·13	8·70E-04	4/4
TNF	TNF	NM_000594	2·56	1·63E-03	3/4
CXCL10	IP10	NM_001565	2·44	1·44E-02	3/4
CCL4L1	MIP-1β	NM_002984	2·33	4·49E-04	4/4
IL1RN	IL1-R Antagonist	NM_000577	2·13	1·49E-02	3/4

¹Gene expression levels were compared for natural killer (NK) cells cultured in IgG1-coated plates with or without soluble inducible T-cell co-stimulator (sICOS). Only genes that had a twofold increase or decrease with sICOS compared with IgG1 are listed here.

²Only genes that were significantly modulated in at least three of four donors are shown.

Preliminary experiments showed that adding sICOS to NK–γδ co-cultures reduced cytotoxicity against autologous DC by 30–40% without affecting NK* cytotoxicity. Conversely, adding 4-1BB Fc did not affect cytotoxicity (data not shown). This suggests that ICOS–ICOSL interactions enhance cytotoxicity against immature DC. Lower DC killing by NK–γδ co-culture is likely to be the result of inhibition of γδ cell-mediated DC killing, either through a regulation of lytic potential or cold target inhibition by the NK cells.

Discussion

Cross-talk between NK and γδ T cells enhances NK cell effector function, including cytolysis of autologous dendritic cells. Antigen-stimulated γδ T cells provide co-stimulation for NK activation via ICOS–ICOSL cognate interactions. Engagement of this receptor pair increases 4-1BB–4-1BBL expression and signalling. Microarray analyses identified other co-stimulatory receptors that may be expressed later in the cross-talk process as part of a cell–cell signalling cascade. Whether these interactions impact cytolysis or regulate other effector phenotypes of NK cells remains to be determined. Our cross-talk model defines a specific role for adaptive γδ T cells in the innate effector response by NK cells.

The ICOS–ICOSL interaction represents a key early event in NK–γδ T-cell cross-talk. ICOS expressed on γδ T cells binds to ICOSL on NK cells to deliver an activating signal. NK cells were effectively stimulated by soluble ICOS, possibly because sICOS is a human Fc fusion protein that binds Fc receptors present on both NK and γδ T cells. However, we cannot rule out that sICOS acts by inhibiting an unknown negative signal. The main consequence of ICOS–ICOSL engagement among other downstream events, is the induction of pro-inflammatory cytokines and TNF receptor superfamily members (including OX40 and GITR) in both cell types. A likely model for the NK–γδ T-cell interaction starts with activation by immobilized IgG1 (NK) or antigen (γδ T) to increase expression of ICOS and ICOSL. Engaging this receptor pair up-regulates 4-1BB and 4-1BBL and provides additional co-stimulation. Later steps may involve OX40 and GITR signalling. Whether some of the co-stimulatory signals are restricted to specialized NK or γδ T-cell subsets remains to be determined.

Not all co-stimulation signals were required for cytolytic activity. When 4-1BB–4-1BBL binding was blocked by s4-1BB Fc, NK* cytotoxicity against DC was unaffected (data not shown). Previously we reported that 4-1BB–4-1BBL binding up-regulated NKG2D,²⁸ a key receptor for NK killing of tumour cells. However, NKG2D is not essential for NK cytotoxicity against autologous DC, which instead depends on NKp30²⁹ and the TRAIL/DR4 pathway.³⁴ In our system, immature or mature DC did not express NKG2D ligands such as MHC class I chain-related A and B (MICA/MICB) and UL16 binding proteins (ULBP); adding anti-NKG2D or anti-ULBP neutralizing antibodies to cytotoxicity assays did not affect DC killing (data not shown). While the 4-1BB–4-1BBL signal is important for NKG2D expression, which mediates NK cytotoxicity against tumour cells, ICOS–ICOSL interaction is important for increased NK-mediated cytolysis of DC. Perhaps we will uncover other examples where individual co-stimulatory molecules control NK receptor expression to impact target cell specificity.

It was already known that NK cells or γδ T cells individually interact with DC.^17,33 Now we understand that reciprocal interactions are established between all three cell types. Having a role for γδ T cells in NK cell activation links innate cytotoxicity of NK cells to antigen-specific responses by Vγ2Vδ2 T lymphocytes, a T-cell subset that is positively selected⁵² and abundant in healthy individuals.⁹ Antigen-specific γδ T cells interact with NK cells; the resulting activation may promote DC editing to select for immunogenic DC.^35,36 The NK–γδ T-cell interaction would occur when both cell types are present at high levels and might also constitute a feedback mechanism to limit further activation or maturation of cytotoxic lymphocytes. Hence, antigen-stimulated γδ T cells may impact adaptive T-cell responses in three ways: directly acting as antigen-presenting cells,^21,22 inducing DC maturation¹⁷ and promoting NK-mediated DC editing.

Our study highlights new functions for human γδ T cells that impact the capacity for immune control of infectious or malignant diseases. Perhaps, strategies for tumour therapy or treatment of HIV/AIDS based on NK cell activation might include direct stimulation of γδ T cells with clinically approved bisphosphonate drugs to co-stimulate critical effector functions of NK cells.

Disclosures

The authors declare that there are no conflicts of interest.