Abstract
γδ T cells stimulated by phosphoantigens (pAg) are potent effectors that secrete Th1 cytokines and kill tumor cells. Consequently, they are considered candidates for use in cancer immunotherapy. However, they have proven only moderately effective in several clinical trials. We studied the consequences of pAg-stimulated γδ T-cell interactions with natural killer (NK) cells and CD8+ T cells, major innate and adaptive effectors, respectively. We found that pAg-stimulated γδ T cells suppressed NK-cell responses to “missing-self” but had no effect on antigen-specific CD8+ T-cell responses. Extensive analysis of the secreted cytokines showed that pAg-stimulated γδ T cells had a proinflammatory profile. CMV-pp65–specific CD8+ T cells primed with pAg-stimulated γδ T cells showed little effect on responses to pp65-loaded target cells. By contrast, NK cells primed similarly with γδ T cells had impaired capacity to degranulate and produce IFNγ in response to HLA class I–deficient targets. This effect depended on BTN3A1 and required direct contact between NK cells and γδ T cells. γδ T-cell priming of NK cells also led to a downregulation of NKG2D and NKp44 on NK cells. Every NK-cell subset was affected by γδ T cell–mediated immunosuppression, but the strongest effect was on KIR+NKG2A– NK cells. We therefore report a previously unknown function for γδ T cells, as brakes of NK-cell responses to “missing-self.” This provides a new perspective for optimizing the use of γδ T cells in cancer immunotherapy and for assessing their role in immune responses to pAg-producing pathogens.
See related Spotlight by Kabelitz, p. 543.
Introduction
γδ T cells comprise a small but distinctive subpopulation of T cells that participates in varied immune responses targeting microbial pathogens and cancer (1). In the peripheral blood of healthy adult humans, γδ T cells typically comprise 0.5% to 5% of the total T cells (2). Most of these cells express the Vγ9Vδ2 T-cell receptor (TCR) and, during bacterial and protozoan infections, can expand to represent up to 50% of the total circulating T cells (3). In humans and nonhuman primates, γδ T cells expressing a Vγ9Vδ2 TCR recognize phosphorylated compounds known as phosphoantigens (pAg). Some pAgs, such as hydroxy-methyl-butyl-pyrophosphate (HMBPP), are microbial products, whereas others, such as isopentenyl pyrophosphate (IPP), are endogenously produced through the enzymes of the mevalonate pathway (4–6). pAgs specifically bind to the intracellular B30.2 domain of butyrophilin subfamily 3 member A1 (BTN3A1), which leads to conformational changes in BTN3A1 and to the activation of γδ T cells through the Vγ9Vδ2 TCR (7, 8). Another butyrophilin, BTN2A1, synergizes with BTN3A1 and is essential for pAg sensing through binding of germline-encoded Vγ9 regions (9).
Upon stimulation with pAgs, γδ T cells express a high level of HLA-DR, as well as various ligands that engage costimulatory proteins and other receptors, many of which are also expressed by natural killer (NK) cells (10–12). Thus activated, γδ T cells perform a range of innate immune cell functions. They produce Th1 cytokines, such as IFNγ and TNFα (13, 14). They also proliferate and directly kill some tumor cells (15–17) as well as cells infected with pAg-producing pathogens, such as mycobacterium tuberculosis (3). pAg-stimulated γδ T cells also contribute to the initiation of adaptive immunity. In a manner similar to T follicular helper cells, pAg-stimulated γδ T cells migrate to lymphoid follicles and contribute to humoral immunity by supporting B-cell maturation (10). Moreover, pAg-stimulated γδ T cells have features of professional antigen-presenting cells (e.g., features of dendritic cells; refs. 11, 18).
Some cancer cells, including type I Epstein–Barr virus (EBV)–infected Burkitt lymphoma lines such as Daudi, maintain a high concentration of endogenous IPP, which stimulates γδ T cells (12, 15, 19). Consistent with this, aminobisphosphonate drugs, which are given to some patients with cancer as treatment for osteoporosis and/or bone metastasis, increase intracellular levels of IPP and lead to activation of γδ T cells (20, 21). Aminobisphosphonates or synthetic pAgs, together with IL2, are therefore being tested as potential cancer immunotherapies in several clinical trials (22, 23). However, many of these trials have shown a lower clinical efficacy than was anticipated (24). Identifying the consequences of interactions between pAg-stimulated γδ T cells and other lymphocytes could be a key step in the development of effective γδ T cell–based immunotherapies.
Here, we analyzed the cytokine profile of pAg-stimulated γδ T cells and studied the cross talk of these activated cells with CD8+ T cells and NK cells, major components of the adaptive and innate immune systems, respectively.
Materials and Methods
Cells and cell culture
Blood samples from healthy adult donors were purchased as anonymized leukoreduction system chambers from the Stanford Blood Center. The CMV status of each donor had been determined serologically at the Blood Center. Peripheral blood mononuclear cells (PBMC) were isolated by density gradient centrifugation (Ficoll-Paque PLUS, GE Healthcare; 17144003) and cryopreserved in heat-inactivated FBS (Corning; 35-015-CV) with 10% DMSO (EMD Millipore; 67-68-5). For all experiments, PBMCs were thawed 1 day prior to use and then centrifuged for 10 minutes at 300 × g. Cell pellets were resuspended in RPMI-1640 (Gibco; 11875-093), containing 10% heat-inactivated FBS, and cultured at 37°C overnight. This allowed the PBMCs to recover from the freeze–thaw before use in in vitro experiments. Daudi (ATCC CCL213) is an EBV-positive Burkitt lymphoma cell line that lacks cell-surface HLA class I. Raji (ATCC CCL86) is an EBV-positive Burkitt lymphoma cell line. K562 (ATCC CCL-243) is an EBV-negative erythroleukemia cell line and is also HLA class I–deficient. P815 (ATCC TIB64) is a mouse mast cell line. All cell lines were cultured in RPMI-1640 supplemented with heat-inactivated FBS (10%), penicillin/streptomycin (100 IU/mL; Gibco; 15140-122), and L-glutamine (5 mmol/L; Gibco; 25030-081). All cell lines were obtained from ATCC between 2014 and 2016 and cultured for no longer than 2 months before thawing a new vial from the original stock. Cell lines were not reauthenticated using specific assays but were identified by morphology and growth characteristics. Lymphoblastoid cell lines (LCL) were generated by infecting PBMCs from healthy donors with the B95.8 strain of EBV, as described previously (25). Cell lines were routinely tested for Mycoplasma.
pAg stimulation of γδ T cells
PBMCs were cultured with Daudi cells for 10 days, at an effector-to-target cell ratio of 10:1, in 24-well plates and passaged upon reaching confluence. For stimulation with 100 µmol/L exogenous IPP (Echelon Biosciences; I-0050), 200,000 PBMCs per well were cultured in 96-well plates and passaged upon reaching confluence. Cells were cultured at a density of 1 × 106 cells/mL, in RPMI supplemented with 5% heat-inactivated human serum (Sigma; H4522), 5% heat-inactivated FBS, and recombinant human IL2 (200 IU/mL). Recombinant human IL2 was provided by the Biological Resources Branch of the NCI, at Frederick, Maryland. The proportion of Vδ2+ T cells in cultures was assessed by flow cytometry after 10 days. The gating strategy used to assess the expansion of Vδ2+ T cells is shown in Supplementary Fig. S1.
Antibodies and analysis by flow cytometry
Antibody clones and their suppliers are listed in Supplementary Table S1. For all experiments, cells were stained with a fixable viability dye (Zombie Aqua; BioLegend; 423101) and antibodies specific for surface markers. For the staining of IFNγ, cells were fixed and permeabilized using a Fixation/Permeabilization solution kit as recommended by the manufacturer (BD Biosciences; 554714). Data were collected with Accuri C6 and LSR II instruments (BD Biosciences) and analyzed using FlowJo software version 10.6.1 (TreeStar Inc.).
γδ T-cell sorting
γδ T cells were negatively sorted from unstimulated PBMCs by magnetic separation, using the TCR γ/δ+ T-cell isolation kit, as recommended by the manufacturer (Miltenyi Biotec; 130-092-892). When sorting stimulated PBMCs, the antibody cocktail from the TCR γ/δ+ T-cell isolation kit was supplemented with 50 µL of biotinylated antibodies specific for NKp46 (9E2), CD4 (OKT4), and CD8 (SK1; all BioLegend). The purity of the sorted γδ T cells was assessed by flow cytometry using antibodies specific for CD3, γδ TCR, Vδ1 TCR, and Vδ2 TCR (Supplementary Table S1); only preparations of nearly 100% γδ T cells, comprising >90% Vγ9Vδ2 T cells and <10% Vδ1 T cells, were studied further.
Luminex assays
Freshly sorted γδ T cells were cultured in cytokine-free RPMI for 3 days, in 96-well plates, at a density of 0.5×106 cells/mL. Supernatants were collected and analyzed for the presence of 76 soluble factors at the Stanford Human Immune Monitoring Center, using a multiplexed fluorescence-based assay (Luminex Bead Array; see for https://iti.stanford.edu/himc/immunoassays.html details). The measured mean fluorescence intensity for each cytokine was normalized by logarithmic transformation. The hierarchical clustering of secretion profiles for each sample was determined by Ward's minimum-variance method. In further analysis, the panel was limited to 49 cytokines, based on functional relevance and quality control measures. As a positive control for inflammatory soluble factors, CMV-pp65–specific CD8+ T cells stimulated with pp65-loaded LCL [LCL preincubated in serum-free RPMI containing 1 µg/mL of NLVPMVATV pp65 (JPT; 495-503), for 1 hour at 37°C] were cultured in cytokine-free media for 3 days, after which culture supernatants were analyzed with the Luminex assay, as described above (Supplementary Fig. S2).
In vitro expansion of CMV-specific CD8+ T cells
PBMCs from CMV-seropositive, HLA-A*02:01+ individuals were suspended in serum-free RPMI containing 1 µg/mL of NLVPMVATV, the HLA-A*02:01–restricted peptide antigen derived from the CMV phosphoprotein pp65. The PBMCs were then incubated for 1 hour at 37°C, at a density of 1.5×106 cells/mL. PBMCs loaded with the peptide were cultured for 15 days with autologous CD8+ T cells at a 1:1 ratio. After 15 days, the proportion of HLA-A*02:01–pp65 restricted CD8+ T cells in the cultures was assessed by flow cytometry. The gating strategy used to assess the expansion of HLA-A*02:01–pp65–specific CD8+ T cells is shown in Supplementary Fig. S1.
Functional assays of CD8+ T cells
HLA-A*02:01–pp65 restricted CD8+ T cells were cultured for 3 days, either alone or in the presence of Daudi- or IPP-stimulated γδ T cells from the same donor. For subsequent challenge of the CD8+ T cells, autologous LCL were suspended in serum-free RPMI with or without 1 µg/mL of NLVPMVATV. After incubation at a density of 1×106 cells/mL for 1 hour at 37°C, the CD8+ T cells were then cultured for 6 hours with LCL alone, or pp65-peptide–loaded LCL, at an effector-to-target cell ratio of 10:1. In some experiments, LCL were loaded with only 0.25 µg/mL of NLVPMVATV, and CD8+ T cells were challenged with LCL at an effector-to-target cell ratio of 3:1. Cultures were established in 96-well plates, with RPMI containing FBS (10%) and IL2 (100 IU/mL). CD107a-specific antibody was added at the start of cultures. After 1 hour, cultures were supplemented with 5 µg/mL brefeldin A (Sigma; 20350-15-6). Cells were then stained for flow cytometry analyses with an HLA-A*02:01–NLVPMVATV dextramer (Immudex, purchased from JPT Peptides) and antibodies specific for CD8 and IFNγ. Information about the antibodies is in Supplementary Table S1.
NK:γδ T-cell cocultures and functional assays
NK cells were negatively sorted from PBMCs with an NK-cell isolation kit, according to the manufacturer's instructions (Miltenyi Biotec; 130-092-657). Only preparations comprising >97% NK cells were studied further. The purity of the sorted NK cells was assessed by flow cytometry using antibodies specific for CD3 and CD56 (Supplementary Table S1). They were then cultured for 3 days, either alone or with Daudi- or IPP-stimulated γδ T cells from the same donor, at a γδT:NK ratio of 1:10. Cultures were established in 96-well plates, with RPMI containing FBS (10%) and IL2 (200 IU/mL). Where stated, IL2 was replaced with IL12, IL15, and IL18 (Thermo Fisher Scientific; PHC1124, PHC9151, and PHC0186, respectively), at a concentration of 5 ng/mL each. In some experiments, neutralizing antibodies against CEACAM1 (clone 18/20; produced in house), PD-L1 (clone 29E.2A3, BioLegend), CD86 (clone BU63, BioLegend), HLA-DR (clone Tü39, BioLegend), galectin-9 (clone 9M1-3, BioLegend), and HLA-A/B/C (clone W6/32, produced in house) were used to block interactions between these proteins, expressed by γδ T cells, and their inhibitory receptors on NK cells. To assess the requirement of BTN3A1 in the cross-talk between NK cells and γδ T cells, in some experiments, NK cells were treated either with an antagonistic BTN3A1-specific antibody (clone 103.2; produced in house) or with an IgG1 isotype control (BioLegend; MG1-45). For Transwell assays, γδ T cells were placed in Transwell inserts (0.4 µm; Corning), and NK cells were cultured in the wells beneath the inserts. For NK-cell priming with unstimulated γδ T cells, unstimulated B cells, or LCL, the same protocol was followed, with pAg-stimulated γδ T cells being substituted by freshly sorted γδ T cells and B cells of the same donor, or LCL generated from PBMCs of the same donor. B cells were negatively sorted from PBMCs with a B-cell isolation kit, according to the manufacturer's instructions (Miltenyi Biotec; 130-101-638). The purity of the sorted cells was assessed by flow cytometry using the anti-CD19 antibody (Supplementary Table S1). Only preparations comprising >97% B cells were studied further.
After 3 days of coculture, NK cells were purified using the NK-cell isolation kit according to the manufacturer's instructions (Miltenyi Biotec). All cultures, including those consisting of NK cells alone, were subjected to the isolation process. NK cells were then cultured for 6 hours with HLA class I–deficient target cells, Daudi or K562, at an effector-to-target cell ratio of 10:1. Cultures were established in 96-well plates, with RPMI containing FBS (10%) and IL2 (200 IU/mL). CD107a-specific antibody was added at the start of culture. After 1 hour, cultures were supplemented with 5 µg/mL brefeldin A. Cells were then stained with antibodies specific for IFNγ and various surface proteins (CD3, CD20, CD56, KIR2DL1/DS1, KIR3DL1/DS1, and NKG2A) and then assessed by flow cytometry for their degranulation and IFNγ production. Information on the antibodies is given in Supplementary Table S1.
The reduction in NK-cell degranulation after priming with γδ T cells was calculated as follows:
For antibody-dependent cellular cytotoxicity (ADCC) assays, Raji cells (1 × 106/mL) were precoated with rituximab (Genentech) at 10 µg/mL for 30 minutes. The Raji cells were then washed in RPMI-1640 and mixed with effector cells at an effector-to-target cell ratio of 10:1. For redirected lysis, NK cells were mixed with P815 cells precoated with anti-NKG2D (clone 1D11; BioLegend) at 10 µg/mL for 30 minutes.
Cell conjugate assay
Sorted NK cells and Daudi-stimulated γδ T cells, as well as K562 and P815 target cells, were labeled with cell tracking dyes: NK cells were labeled with 2 µmol/L Celltracker Green (Thermo Fisher; C2925); γδ T cells, K562 cells, and P815 cells were labeled with 0.2 µmol/L Celltracker Deep Red (Thermo Fisher; C34565). NK cells were then combined with γδ T cells, HLA class I–deficient K562 cells (positive control) or P815 mouse myeloma cells (negative control) at a 2:1 ratio. Cells were transferred into FACs tubes containing 300 µL ice-cold 0.5% PFA, after a 0-, 1-, or 4-hour incubation in RPMI containing FBS (10%) and IL2 (200 IU/mL). The formation of cell conjugates was then assessed and quantified using flow cytometry.
Confocal microscopy
Chamber slides (Lab-Tek, Nunc) were coated with 0.01% poly-L-lysine (Sigma; A-005-C) for 30 minutes at room temperature, then washed and dried overnight. Isolated NK cells and Daudi-stimulated γδ T cells were cultured at a 2:1 ratio in the prepared poly-L-lysine–coated chamber slides for 1 hour, at 37oC. Cells were fixed with 4% PFA for 20 minutes at room temperature. Samples were permeabilized with 0.1% Triton-X-100 for 5 minutes, washed in PBS and then incubated in PBS containing 1% BSA and 2% human serum for 1 hour. Cells were then incubated for 1 hour with antibodies specific for CD3 (UCHT1) and HLA-DR (L243) and for a further 20 minutes with an F-actin stain (Alexa Fluor 594 Phalloidin, Life Technologies; A12381). Confocal imaging was performed on the Leica SP8 confocal system with white light laser, using a 63× oil immersion objective. Images were processed and analyzed using ImageJ (26).
Trogocytosis assay
NK cells and pAg-stimulated γδ T cells were cultured together for 60 minutes, with either NK cells or γδ T cells prelabeled with the lipophilic dye PKH67 (Sigma; PKH67GL), following the manufacturer's recommendations. Cocultures were made at a NK:γδ T-cell ratio of 3:1, in 96-well plates containing RPMI with FBS (10%) and IL2 (200 IU/mL). After 60 minutes incubation at 37°C, cells were placed on ice for further staining with anti-CD3, anti-CD56, and anti-Vδ2TCR. Data were collected using flow cytometry.
HLA genotyping
Blood donors’ HLA genotypes were determined as described (27). Briefly, genomic DNA was isolated from PBMC using the QIAamp DNA Blood Mini kit, as described by the manufacturer (QIAGEN; 51104). Genomic DNA fragments encoding HLA-A, -B, and -C were amplified using oligonucleotide probes and sequenced using a Miseq next-generation sequencer with reagent kit v3 (Illumina, MS-102-3003), as recommended by the manufacturer
Statistical analysis
Statistical analyses were performed in GraphPad Prism and R. P values were calculated using the Student t test and ANOVA. Only P < 0.05 were considered significant. The normal distribution of the samples was confirmed using Shapiro–Wilk test.
Data availability
Raw data for this study were generated at the Stanford Human Immune Monitoring Center, the Stanford Shared FACS Facility and the Stanford Cell Sciences Imaging Facility. Derived data supporting the findings of this study are available from the corresponding author upon request.
Results
pAg-stimulated γδ T cells secrete proinflammatory cytokines and chemokines
Peripheral blood γδ T cells expand rapidly in response either to pAgs produced by pathogens (3) or to pAgs produced by transformed B cells maintaining type I EBV infection (12, 15, 28). Most of the responding T cells express the Vγ9 and Vδ2 TCR chains. They also express several NK-cell receptors (12), HLA-DR, and an extensive array of adhesion molecules and ligands for costimulatory receptors (11). To characterize the cytokine secretion profile of pAg-expanding γδ T cells, we used a multiplexed immunoassay that detects up to 76 soluble factors. Analyses were performed on cultures of unstimulated γδ T cells and of γδ T cells stimulated for 10 days with either IPP or Daudi, a type I EBV–infected B cell line.
Figure 1A shows signal intensities for a variety of secreted proteins including cytokines, chemokines, and soluble adhesion molecules. These data were grouped by hierarchical clustering based on overall secretion profiles. Data from unstimulated and stimulated γδ T cells clustered separately, indicating that their secretion profiles were distinct. In addition, the fact that the data from Daudi- and IPP-stimulated γδ T cells clustered together indicated that they had similar patterns of secretion. Indeed, Daudi- and IPP-stimulated γδ T cells secreted more FAS ligand, FAS receptor, TRAIL, IFNγ, MIP1α, MIP1β, CCL5, and sICAM1 than unstimulated γδ T cells (Fig. 1B). Daudi-stimulated γδ T cells also exhibited a significant increase in their secretion of TNFα and CXCL1 (Fig. 1B). In contrast, Daudi- and IPP-stimulated γδ T cells secreted significantly less CCL2, IL8, and IL12p40 than unstimulated γδ T cells. The anti-inflammatory cytokine IL10 was also reduced in the supernatants from IPP-stimulated γδ T cells (Fig. 1B).
Overall, these data show that γδ T cells stimulated with either Daudi or IPP exhibited a similar secretion profile, one characterized by increased production of proinflammatory cytokines and chemokines. Consistent with these increases, production of the anti-inflammatory cytokine IL10 was decreased in IPP-stimulated γδ T cells. From these observations, we predicted that pAg-stimulated γδ T cells would enhance the immune response of other cytotoxic lymphocytes.
pAg-stimulated γδ T cells have little effect on antigen-specific CD8+ T-cell responses
CD8+ T cells are important effectors of adaptive immunity during infections with pAg-producing pathogens. Under these conditions, γδ T cells become activated to display a proinflammatory phenotype (Fig. 1A–B). Therefore, we developed an in vitro model in which CMV-specific CD8+ T cells were used to study the effect of pAg-stimulated γδ T cells on antigen-specific CD8+ T-cell responses.
We expanded HLA-A*02:01–pp65–specific CD8+ T cells from HLA-A*02:01–positive, CMV-seropositive healthy donors. To obtain high numbers of CMV-specific cytotoxic T cells (CTL) in culture, CD8+ T cells were sorted and incubated for 15 days with autologous PBMCs loaded with NLVPMVATV, an HLA-A*02:01–restricted peptide antigen derived from the CMV phosphoprotein pp65 (Fig. 2A). Thus expanded in vitro, the HLA-A*02:01–pp65–restricted CTLs were then cultured, either alone or in the presence of IPP- or Daudi-stimulated γδ T cells for 3 days. CTLs from all cultures were challenged with autologous LCLs (effector-to-target cell ratio of 10:1), either loaded or not with 1 µg/mL of NLVPMVATV.
The HLA-A*02:01–pp65–restricted CTLs obtained from these cultures were highly cytotoxic. They also secreted a substantial quantity of IFNγ in response to pp65-loaded LCL. Thus, the proportion of CTL expressing both CD107a and IFNγ increased from 0.0% to 0.8% after challenge with LCL to 62.7% to 87.3% after exposure to pp65-loaded LCLs (Fig. 2B and C). Moreover, priming of the CTLs with either Daudi cell–stimulated or IPP-stimulated γδ T cells caused only modest change to CD107a expression and IFNγ production compared with CTLs cultured alone (Fig. 2D and E). Only priming with IPP-stimulated γδ T cells gave a statistically significant increase in IFNγ production: from 77.9%–92.9% to 82.6%–96.3% (Fig. 2E, right bottom).
As the proportion of CD8+ T cells that respond to the immunodominant pp65 epitope is close to 100%, it is possible that γδ T cells have an effect on CD8+ T-cell functions that is not revealed by the dynamic range of our system. To test this hypothesis, we challenged CD8+ T cells with LCL loaded with a lower concentration of NLVPMVATV (0.25 µg/mL), at a lower effector-to-target cell ratio (3:1). Under these experimental conditions, the proportion of CD8+ T cells that degranulated and produced IFNγ in response to pp65-loaded LCL was reduced (25%–65%). However, no effect of γδ T cells was observed on the CD8+ T-cell response to pp65 (Fig. 2F).
We find that priming HLA-A*02:01–pp65–restricted CTLs with Daudi- or IPP-stimulated γδ T cells had little effect on their immune response to pp65. This suggests that, despite their proinflammatory phenotype, pAg-stimulated γδ T cells do not promote an antigen-specific CD8+ T-cell response.
pAg-stimulated γδ T cells suppress NK-cell responses to HLA class I–deficient cells
NK cells are major effectors of innate immunity. In contrast to CTLs, which activate on recognition of specific peptide antigens bound to HLA class I, NK cells detect loss of HLA class I by unhealthy cells. Here, we studied the influence of pAg-stimulated γδ T cells on the capacity of NK cells to kill target cells having no surface expression of HLA class I.
NK cells were cultured alone or with Daudi-stimulated γδ T cells for 3 days. They were then sorted and challenged with HLA class I–deficient target cells, either K562 or Daudi. Priming NK cells with Daudi-stimulated γδ T cells substantially reduced the NK-cell response to these HLA class I–deficient targets. Thus, the degranulation response to Daudi cells was reduced from 15.3%–62.1% to 6.9%–33.5% (Fig. 3A, left). A similar reduction was observed in the NK-cell response to K562 cells (Fig. 3A, right). Such suppression of the NK-cell response by Daudi-stimulated γδ T cells also affected IFNγ production, which declined from 4.9%–15.6% to 2.0%–10.1% (Fig. 3B). Furthermore, NK cells did not recover their effector functions in response to HLA class I–deficient target cells when allowed to rest for 72 hours after coculture with Daudi-stimulated γδ T cells (Supplementary Fig. S3).
We investigated if IPP-stimulated γδ T cells also suppressed the NK-cell response to “missing-self.” On priming NK cells with IPP-stimulated γδ T cells, the NK-cell degranulation response to Daudi cells was reduced from 15.3%–65.7% to 8.7%–41.0% (Fig. 3C, left). Similarly, the response to K562 cells was reduced from 11.4%–65.7% to 9.9%–41.0% (Fig. 3C, right). Priming with IPP-stimulated γδ T cells also suppressed NK-cell production of IFNγ, which was reduced from 6.9%–14.8% to 1.8%–6.8% (Fig. 3D).
To determine if priming with pAg-stimulated γδ T cells also affected the capacity of NK cells to mediate ADCC, we assessed the response of NK cells primed with Daudi-stimulated γδ T cells to Raji cells precoated with anti-CD20. Neither NK cell–mediated cytotoxicity nor IFNγ production were affected by priming with γδ T cells (Fig. 3E). These results demonstrate that NK cells primed with pAg-stimulated γδ T cells had suppressed responses to “missing-self” but retained the capacity to kill target cells by the mechanism of ADCC. Similarly, redirected NK-cell lysis induced by anti-NKG2D against the FcR+P815 cell line was not affected by priming with Daudi-stimulated γδ T cells (Fig. 3F).
To test the effect of pAg-stimulated γδ T cells in the presence of exogenous proinflammatory cytokines, we performed our cocultures of NK cells and pAg-stimulated γδ T cells in the presence of IL12, IL15, and IL18. Treatment with these cytokines prevented the immunosuppression of NK cells by the pAg-stimulated γδ T cells (Fig. 3G).
Next, we investigated the effect of unstimulated γδ T cells on NK-cell responses to “missing-self.” On priming with freshly sorted γδ T cells, a slight but consistent reduction in the response of NK cells to Daudi cells was observed for all studied donors (Fig. 3H, left). In contrast, no reduction in NK-cell response was observed after priming with freshly sorted B cells (Fig. 3H, right) or autologous LCL (Supplementary Fig. S4). These results indicate that unstimulated γδ T cells also immunosuppressed NK cells but to a lesser extent than pAg-stimulated γδ T cells.
Previously we demonstrated that all human individuals make one of two distinctive innate immune responses to EBV infection. Group 1 individuals make strong NK-cell and γδ T-cell responses to EBV, whereas Group 2 individuals combine a strong NK-cell response with a weak γδ T-cell response (12). We therefore compared the immunosuppression mediated by Daudi-stimulated γδ T cells on NK cells from the two donor groups. No difference was found between them (Supplementary Fig. S5).
To determine if the effect of γδ T cells on NK-cell responses required physical contact between the two cell types, cultures were made in a Transwell system that prevented contact between the γδ T cells and the NK cells, but that did not impede the movement of soluble factors. Under these conditions, Daudi-stimulated γδ T cells had no significant effect on the NK-cell response to HLA class I–deficient cells (Fig. 4A). To further investigate dependence on direct cell contact, we assayed the formation in culture of stable cell conjugates between NK cells and pAg-stimulated γδ T cells. NK cells and pAg-stimulated γδ T cells were labeled with cell tracking dyes prior to coculture, during which the formation of stable conjugates was indicated by the presence of a doubly stained population. NK cells were also cultured with K562 cells, as a positive control for stable conjugate formation, and with P815 mouse myeloma cells as a negative control. During 1 hour of coculture, the NK cells formed stable conjugates with pAg-stimulated γδ T cells. The extent of conjugate formation was comparable with that seen with K562 target cells (Fig. 4B and C). Indeed, 25% to 30% of the NK cells were found in conjugates with γδ T cells, compared with only 5.5% to 13.5% of NK cells in conjugates with P815 cells (Fig. 4B and C). These results are consistent with the formation of stable interactions between γδ T cells and NK cells, at early times in their coculture.
Further experiments were performed to assess the formation of conjugates between NK cells and pAg-stimulated γδ T cells, using confocal microscopy. The cells were cocultured for 1 hour on glass slides coated with nonactivating poly-L-lysine, then stained with CD3- and HLA-DR–specific antibodies, as well as a phalloidin label for F-actin. The CD3- and HLA-DR–specific antibodies allowed us to distinguish between the NK cells and the highly activated γδ T cells. This analysis revealed direct interactions between γδ T cells and NK cells across multiple imaging fields (Fig. 4D). The γδ T cells were seen to have an enlarged and elongated morphology compared with the NK cells, consistent with their greater extent of activation (Fig. 4D). Together, all these data demonstrate the occurrence of a physical cross talk between NK cells and pAg-stimulated γδ T cells, causing NK-cell suppression.
Our finding that inhibition of NK cells by γδ T cells required physical contact between the two cell types pointed to a role for interactions between the cell-surface proteins of NK cells and γδ T cells. We therefore sought to identify ligands for NK-cell inhibitory receptors that are upregulated on the surface of pAg-stimulated γδ T cells. Significant increases in the expression of HLA-DR (11), HLA-E, HLA-G, CD86 (11), galectin-9, CEACAM1, PD-L1, and PD-L2 were observed (Fig. 5A). HLA-DR is known to inhibit T-cell and NK-cell activation via the LAG-3 or Fcrl6 receptors (29–31). HLA-E and HLA-G are ligands for the inhibitory receptors NKG2A and LILRB1, respectively (32, 33). CD86, a ligand for CTLA-4, is known to inhibit NK-cell cytokine production in mice (34). CEACAM-1 inhibits NK-cell cytotoxicity through homotypic interactions (35). CEACAM1 and galectin-9 inhibit NK-cell activity through interactions with TIM-3 (36, 37). PD-L1 and PD-L2 are ligands of PD-1 (38), which is expressed by subsets of NK cells (39). Despite the increased expression of these ligands for inhibitory NK-cell receptors, blocking particular receptor–ligand interactions with ligand-specific antibodies did not restore the strength of the NK-cell response to cancer cells, after priming with pAg-stimulated γδ T cells (Supplementary Fig. S6A). On the other hand, blocking multiple ligands together in a single experiment controlled with antibody isotype controls was not informative as it led to an overall activation of NK cells in all cultures (Supplementary Fig. S6B).
We assessed the immunosuppressive effects mediated by pAg-stimulated γδ T cells on NK-cell populations expressing either KIR or NKG2A, the receptors that determine the NK-cell response to “missing-self.” After priming with Daudi-stimulated γδ T cells, the immunosuppression of KIR+NKG2A– NK cells was greater than that of KIR–NKG2A+ NK cells (Fig. 5B). This difference implicates KIR, NKG2A, or both receptors in the interactions of NK cells with activated γδ T cells. An alternative interpretation of these results is that KIR+NKG2A– NK cells, which represent the mature NK-cell subsets, express other proteins involved in the cross-talk between NK cells and γδ T cells. Previously, we found that the methionine/threonine (M/T) dimorphism at position −21 of the HLA-B leader sequence influences NK-cell education and function, with M/T and M/M individuals having more educated NKG2A+ NK cells than T/T individuals (40). Here, we show that this dimorphism did not affect the magnitude of the NK-cell immunosuppression mediated by either Daudi-stimulated γδ T cells (Supplementary Fig. S5B, left) or IPP-stimulated γδ T cells (Supplementary Fig. S5B, right).
To assess the functional contribution of BTN3A1 to the γδ T cell–mediated immunosuppression of NK cells, we examined the effect of an antagonistic antibody specific for BTN3A1 (clone 103.2). BTN3A1 is expressed by NK cells (41). Thus, NK cells were treated with either anti-BTN3A1 or control mouse IgG and then cultured with pAg-stimulated γδ T cells for 3 days. NK cells were then sorted and challenged with HLA class I–deficient target cells. Unlike the control IgG, anti-BTN3A1 prevented the reduction in NK-cell response to “missing-self” (Fig. 5C). These results demonstrated an essential role of BNT3A1 in the cross-talk between NK cells and γδ T cells that leads to NK-cell immunosuppression.
Further analyses of NK-cell phenotype revealed a consistent reduction in NK-cell expression of the activating receptor NKG2D after priming with Daudi-stimulated γδ T cells (Fig. 5D). However, as previously shown (Fig. 3F), this change in NK-cell phenotype did not affect redirected lysis triggering NKG2D. We then assessed if there were changes involving other activating NK-cell receptors. NK cells were cocultured with Daudi-stimulated γδ T cells for 3 days and assayed for their expression of additional activating receptors. We found a slight decrease in NKp44 expression by NK cells, after coculture with Daudi-stimulated γδ T cells (Fig. 5E). In contrast, the exposure of γδ T cells to NK cells led to a considerable increase in γδ T-cell expression of NKp44 and CD16 (Fig. 5F).
The significant increase in NKp44 and CD16 expression by γδ T cells after coculture with NK cells suggested there was a transfer of membrane-containing activating receptors from NK cells to γδ T cells by trogocytosis. To test this hypothesis, we incubated Daudi-stimulated γδ T cells with NK cells prelabeled with the green lipophilic dye PKH67 for 60 minutes and assessed membrane transfer using flow cytometry. No green fluorescence was observed on γδ T cells after culture with PKH67-labeled NK cells (Fig. 5G, left). This indicates that γδ T cells did not acquire NKp44 and CD16 from NK cells. In contrast, a similar assay of NK cells incubated with PKH67-labeled γδ T cells showed a slight but consistent increase in the green fluorescence emitted by NK cells (Fig. 5G, right). This result suggests there was a transfer of cell membrane from γδ T cells to NK cells.
In conclusion, our investigation shows that NK-cell priming with pAg-stimulated γδ T cells significantly reduced NK-cell responses to “missing-self” but did not affect the ability of NK cells to mediate ADCC. This γδ T cell–mediated immunosuppression of NK cells was BTN3A1 dependent, could be prevented by exogenous cytokines and did not occur if direct contact between NK cells and γδ T cells was prevented in Transwell cultures. Flow cytometry and confocal imaging further demonstrated direct cell-surface interactions between NK cells and γδ T cells, pointing to a role for cell-surface proteins in the suppressive effect. We report increased expression of several ligands for inhibitory NK-cell receptors on the surface of γδ T cells. However, blocking individual receptor–ligand interactions failed to restore NK-cell responses after priming with pAg-stimulated γδ T cells. Our data show that KIR+NKG2A– NK-cell responses to “missing-self” were strongly, but not exclusively, affected by priming with pAg-stimulated γδ T cells. Coculture of NK cells with pAg-stimulated γδ T cells led to a decrease in the expression of NKG2D and NKp44 on NK cells and an increase in the expression of NKp44 and CD16 on γδ T cells. In summary, we clearly demonstrate that despite their proinflammatory phenotype, pAg-stimulated γδ T cells strongly suppress NK-cell responses to “missing-self.”
Discussion
In responding to infection or cancer, γδ T cells secrete cytokines and chemokines and kill infected or malignant cells. γδ T cells also function as antigen-presenting cells that initiate adaptive immune responses (10, 11). Here we report an additional function of pAg-stimulated γδ T cells, as regulators of NK-cell responses to “missing-self.” We find that priming NK cells with either Daudi- or IPP-stimulated γδ T cells strongly suppresses the subsequent NK-cell response to HLA class I–deficient target cells but does not affect the ability of NK cells to mediate ADCC.
NK cells were not killed by autologous pAg-stimulated γδ T cells and vice versa. Moreover, NK cells were purified after the 3-day coculture with γδ T cells, and the same number of live NK cells was used in functional assays with HLA class I–deficient targets. Thus, the observed effect is not due to a reduced number of live NK cells from the cultures supplemented with γδ T cells.
Previously, we demonstrated that the human population comprises two distinctive groups that differ in their γδ T-cell response to type I EBV-infected B cells, such as Daudi cells. Group 1 individuals have an abundance of γδ T cells with the JγP complementarity determining region 3 (CDR3s), which vigorously proliferate in response to EBV-infected cells. Group 2 individuals have significantly fewer γδ T cells with JyP CDR3s, and consequently produce a smaller population of activated effector cells in response to EBV (12, 28). For both donor groups, most γδ T cells responding to type I EBV–infected B cells are Vδ2 T cells that express a Vγ9JγP CDR3 (28). In the current study, equal numbers of Daudi-stimulated γδ T cells from groups 1 and 2 individuals were compared for their capacity to suppress NK-cell responses to “missing-self.” As we expected, we found that γδ T cells from both donor groups exert equivalent effect on NK cells. In vivo, however, the number of activated γδ T cells can vary between groups 1 and 2 individuals, in certain inflammatory settings such as infectious mononucleosis (12). This variation could influence the likelihood of direct interactions between γδ T cells and NK cells, and thus their effect on NK-cell functions.
It is striking that γδ T cells can cause a strong immunosuppression of NK cells but not of CD8+ T cells. Indeed, we observed little effect of γδ T cells on antigen-specific CD8+ T-cell immunity. There was even a small increase in the CD8+ T-cell response to CMV after priming with pAg-stimulated γδ T cells.
The contrast between the modest immune stimulation of CD8+ T cells and the strong immunosuppression of NK cells by γδ T cells points to the latter having specific interactions with NK cells. That NK-cell immunosuppression by γδ T cells requires contact between the two cell types points to a functional cross-talk between NK cells and γδ T cells that involves cell-surface proteins. pAg-stimulated γδ T cells can form stable immune synapses with tumor cells, facilitating uptake of lipids and proteins from the target cells (42). NK cells can also form both activating and inhibitory synapses with target cells (43) and regulatory synapses with other immune cells (44, 45). However, the formation of stable immune interactions between γδ T cells and NK cells has previously not been described. We find that stable conjugates between γδ T cells and NK cells are formed during 1 hour of coculture, and these are detected by both flow cytometry and confocal microscopy. Our investigation demonstrates that interactions between cell-surface proteins of γδ T cells and NK cells are responsible for the immunosuppression of NK cells by γδ T cells.
Although all subsets of NK cells were immunosuppressed to some degree by activated γδ T cells, the greater immunosuppression was observed for KIR+NKG2A– NK cells. This points to a role for KIR, NKG2A, or both receptors in the cross-talk. However, similar priming of NK cells with autologous LCL did not prevent subsequent NK-cell responses to “missing-self.” This result is consistent with NK cells having specific interactions with γδ T cells. Moreover, the fact that LCL also express self HLA ligands for KIR and NKG2A points to these receptors not contributing directly to the cross-talk. Alternatively, KIR+NKG2A– NK cells, which represent a more mature NK-cell subset (46), could have higher expression of the surface proteins involved in the cross talk with γδ T cells.
To define cell-surface proteins that contribute to the suppression of NK cells, we profiled the surface of pAg-stimulated γδ T cells for expression of ligands recognized by inhibitory NK-cell receptors. We found that pAg-stimulated γδ T cells upregulate several ligands for inhibitory NK-cell receptors, including HLA-DR, HLA-E, HLA-G, CD86, galectin-9, CEACAM1, PD-L1, and PD-L2. However, blocking the interactions of these ligands and their receptors, individually, did not restore NK-cell responses to the levels achieved by NK cells cultured alone. These results suggest that NK-cell suppression by pAg-stimulated γδ T cells is a consequence of combinations of receptor–ligand interactions. Supporting this model, a previous study showed cooperative activity between different NK-cell receptors, which each have lesser functional impact when stimulated alone (47). Alternatively, the immunosuppression of NK cells by γδ T cells could be mediated by proteins that have yet to be described.
Previous research showed that pAg-stimulated PBMCs enhance NK-cell killing of HLA class I–deficient tumor cells and monocyte-derived dendritic cells, via CD137 and ICOS, respectively (48, 49). In those systems, the effect of γδ T cells required NK-cell priming with IgG1. Moreover, γδ T cells were not isolated from the pAg-stimulated PBMC cultures, which could also explain the difference between our findings and those analyses (48, 49).
Immunosuppression of NK cells by pAg-stimulated γδ T cells could contribute to the low efficacy of cancer treatment with ex vivo– or in vivo–expanded γδ T cells (24). Further investigation of NK-cell function should help to understand the clinical outcomes after in vivo expansion of γδ T cells or adoptive transfer of pAg-stimulated γδ T cells. We show that NK cells primed with pAg-stimulated γδ T cells have reduced expression of NKG2D and NKp44. The essential functions of these receptors in NK cell–mediated antitumor responses are well documented (50, 51) and indicate their loss could dampen NK-cell functions. Moreover, our analysis demonstrated a transfer of membrane by trogocytosis from γδ T cells to NK cells. It is likely that in addition to the loss of NKG2D and NKp44, NK cells obtain one or several inhibitory proteins from γδ T cells that further impair NK-cell responses to HLA class I–deficient target cells.
These findings provide a novel perspective on infections with pAg-producing pathogens, in which γδ T cells are a high proportion of the total lymphocytes in peripheral blood. Indeed, although it is generally accepted that γδ T cells have a beneficial role in the immune defense against these pathogens (3), it is important to study the effect of their interactions with other lymphocyte subsets. For instance, a decrease in NK-cell activity has been observed in patients with mycobacterium tuberculosis infection (52).
We have shown that pAg-stimulated γδ T cells strongly suppress NK-cell responses to “missing-self,” while exerting no such effect on NK cell–mediated ADCC or antigen-specific CD8+ T-cell responses. Such a regulatory role for γδ T cells could illuminate the clinical outcomes of cancer therapy with pAg-stimulated γδ T cells and open new avenues of investigation to define innate immune responses to infections with pAg-producing pathogens. An important next step is to determine if this NK-cell immunosuppression by γδ T cells occurs in vivo. If that is the case, it could set the stage for cancer immunotherapy based on γδ T cells and their interactions with NK cells.
Authors' Disclosures
K. Walwyn-Brown reports grants from NIH during the conduct of the study; other support from McCann Health Medical Communications outside the submitted work. J. Pugh reports grants from NIH during the conduct of the study; grants from NIH outside the submitted work; and currently employed in industry, but all contributions to this work were completed long before employment, with no overlap. L.A. Guethlein reports grants from NIH during the conduct of the study. Z. Djaoud reports grants from NIH and the National Institute of Allergy and Infectious Diseases during the conduct of the study. No disclosures were reported by the other authors.
Authors' Contributions
K. Walwyn-Brown: Investigation, methodology, writing–original draft. J. Pugh: Investigation, methodology. A.T.H. Cocker: Investigation, methodology. N. Beyzaie: Methodology. B.B. Singer: Methodology. D. Olive: Methodology. L.A. Guethlein: Funding acquisition, methodology. P. Parham: Supervision, funding acquisition, writing–original draft, writing–review and editing. Z. Djaoud: Conceptualization, data curation, formal analysis, supervision, funding acquisition, validation, investigation, methodology, writing–original draft, writing–review and editing.
Acknowledgments
The multiplexed fluorescence-based assay (Luminex) for analysis of soluble factors in cell supernatants was performed at the Stanford Human Immune Monitoring Center. Flow cytometry analysis was performed in the Stanford Shared Flow Cytometry Facility. Confocal imaging was performed at the Stanford Cell Sciences Imaging Facility; the Leica SP8 system is supported by Award Number 1S10OD010580 from the National Center for Research Resources (NCRR). The contents of this report are solely the responsibility of the authors and do not necessarily represent the official views of the NCRR or the NIH. This work was supported by NIH and the National Institute of Allergy and Infectious Diseases grant AI136592 (P. Parham).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.