CTLA-4 Is Expressed by Activated Mouse NK Cells and Inhibits NK Cell IFN-γ Production in Response to Mature Dendritic Cells

Natural killer cells are innate immune effectors that play an important role during antiviral and antitumor immune responses (1, 2). NK cells express an array of activating and inhibitory surface receptors whose engagement determines the outcome of NK cell activation (3). Activating receptors recognize ligands that are upregulated by stressed, transformed, or infected cells. MHC class I molecules, expressed at high levels by healthy cells, make up the main group of ligands engaging inhibitory NK cell receptors. In addition to the engagement of NK cell receptors by respective NK cell ligands, microenvironmental signals, such as cytokines, can trigger and modulate NK cell responses. For instance, IL-12 and IL-18 synergistically induce high levels of IFN-γ production by NK cells (4, 5). On the contrary, NK cell function can be suppressed by factors such as TGF-β (6, 7). Microenvironmental factors also can regulate NK cell responses by modulating the levels of activating and inhibitory surface receptors and, thereby, determine the threshold for NK cell activation.

T cell activation also requires integration of at least two signals (8). Signal one is triggered via the TCR by the recognition of antigenic peptides in the context of MHC molecules. Signal two is delivered by the engagement of costimulatory molecules, such as CD28. Its ligands, B7-1 and B7-2, are primarily expressed by APCs and are engaged by CD28, as well as by CTLA-4 (CD152). CTLA-4 is an inhibitory receptor induced following TCR activation and represents a critical regulator of ongoing T cell responses (9, 10). In T cells, most of the CTLA-4 protein resides intracellularly; only a small fraction is exposed on the cell surface where it can engage its ligands. CTLA-4–deficient mice display extensive proliferation and accumulation of self-reactive T cells, resulting in tissue destruction and premature death of animals (11, 12). The hyperactivation of CTLA-4–deficient T cells can be prevented by interruption of the activating CD28:B7-1/B7-2 pathway (13, 14) or by the concomitant presence of CTLA-4–sufficient T cells (15–17). It was reported that CTLA-4 exerts its function through both cell-intrinsic and cell-extrinsic mechanisms. In this regard, CTLA-4 was shown to compete with CD28 for their shared ligands, inhibit signal transduction downstream of the TCR, deliver inhibitory signals affecting T cell function, and remove B7-1 and B7-2 from the surface of APCs (18–20).

Previous studies (21–24) reported that CD28 was expressed by mouse NK cells, and its triggering induced NK cell proliferation, cytotoxicity, and cytokine secretion. In studies by Chiossone et al. (25) and Terme et al. (26), CTLA-4 transcripts were detected by whole-genome microarray analysis of mouse NK cells, indicating that CTLA-4 protein might also be expressed by NK cells. However, the function of CTLA-4 on NK cells remained unknown. In this study, we show that IL-2–activated NK cells express both receptors: CTLA-4 and CD28. We provide further insight into the regulation of their expression by cytokines. Moreover, our data show that CTLA-4 inhibits IFN-γ release by NK cells upon coculture with mature dendritic cells (mDCs). Notably, expression of both CD28 and CTLA-4 was detected in tumor-infiltrating NK cells in several mouse models of solid tumors, suggesting their involvement in NK cell antitumor responses.

Wild-type (WT) C57BL/6N (Charles River Laboratories), WT C57BL/6-Ly5.1, and Rag2- or Ly5.1-Rag2–deficient mice (both bred in the animal facility of the Deutsches Krebsforschungszentrum) were used at the age of 8–16 wk. CD28-deficient mice and respective C57BL/6J WT control mice were purchased from The Jackson Laboratory and also were used at 8–16 wk of age. WT OT-I–transgenic and OT-I–transgenic CTLA-4–deficient mice (both bred in the animal facility of the Otto-von-Guericke University) were used at the age of 4–8 wk. In all experiments, phenotype and responses of CD28- or CTLA-4–deficient NK cells were compared with their respective WT control cells. All mice were housed under specific pathogen–free conditions. For tumor experiments, mice were inoculated s.c. with 10⁶ RMA-S lymphoma, B16 melanoma, or LL/2 lung carcinoma cells. For analysis of NK cells from blood and tumors, mice were sacrificed when the tumor diameter exceeded 1 cm (typically 1.2–1.6 cm on days 12–16 after tumor cell inoculation). All animal experiments were approved by the “Regierungspräsidium Karlsruhe.”

Single-cell suspensions from peripheral blood were prepared after RBC lysis with buffered ammonium chloride potassium phosphate solution (ACK buffer) for 2 min at room temperature. The tumor tissue was cut into small pieces and treated with digestion buffer (PBS with 0.5 mg/ml hyaluronidase V and 0.5 mg/ml DNase; Sigma-Aldrich) for 30 min at 37°C. Subsequently, viable cells were collected after centrifugation over a Lympholyte-M gradient (Cedarlane Laboratories). In experiments comparing the expression of surface molecules on NK cells from blood and tumor, both tumor tissue and single-cell suspensions from blood were treated with digestion buffer.

NK cells were obtained from Rag2-deficient mice by expansion of nonadherent splenocytes in the presence of recombinant human (rh)IL-2 (National Institutes of Health). NK cells were isolated from WT and gene-deficient mice by MACS separation (Miltenyi Biotec) comprising depletion of CD3⁺ cells, followed by positive selection of DX5⁺ cells. Cells were maintained in RPMI 1640 medium supplemented with 10% FCS, 2 mM l-glutamine, 100 U/ml penicillin/streptomycin, 1 mM sodium pyruvate, 1% nonessential amino acids, 0.25 mM 2-ME (all from Sigma-Aldrich or Invitrogen), and 1700 U/ml rhIL-2. The purity of NK cell cultures on days 7–10 of expansion was ≥99% CD3⁻NK1.1⁺. Dendritic cells (DCs) were generated from bone marrow cells in medium (DMEM with 10% FCS, 2 mM l-glutamine, and 100 U/ml penicillin/streptomycin) supplemented with GM-CSF (10% supernatant of GM-CSF–producing X6310 cell line). On day 6, DCs were cultured with 200 ng/ml LPS (Sigma-Aldrich) for 18–24 h to generate mDCs. Cytokines used in NK cell cultures were purchased from PeproTech (IL-15, IL-12, and TGF-β) and MBL (IL-18).

Cells were incubated for 30 min at 4°C with FcR-blocking reagent (10% supernatant of αCD16/CD32-producing hybridoma 2.4G2), followed by incubation with fluorochrome-labeled mAbs. Dead cells were excluded by labeling with 7-aminoactinomycin D (7-AAD), propidium iodide, Zombie Aqua (BioLegend, BD Biosciences), or the LIVE/DEAD Fixable Violet Dead Cell Stain Kit (Molecular Probes). CD28 expression was assessed by cell surface staining using monoclonal anti-CD28 Abs (clone 37.1 and clone E18; BioLegend). For intracellular staining of CTLA-4, cells were stained with mAbs directed against surface Ags, followed by fixation, permeabilization (Foxp3 Staining Buffer Set; eBioscience), and staining with anti–CTLA-4 mAb (clone UC10-4F10-11 from BD Bioscience and clone UC10-4B9 from BioLegend). For detection of CTLA-4 on the cell surface, fluorochrome-labeled anti–CTLA-4 mAb was added to NK cell cultures at 37°C for 4 h before cell harvesting, and cells were analyzed by flow cytometry. FACSCalibur and FACSCanto were used for FACS acquisition, and FACSDiva, FACSAria, and Influx were used for cell sorting. For the preparation of tumor-infiltrating NK cells, C57BL/6-Ly5.1 mice were inoculated with RMA-S tumor cells expressing Ly5.2. Total tumor-infiltrating leukocytes were enriched using MACS separation by positive selection of Ly5.1⁺ cells. NK cells were sorted as Ly5.1⁺CD3⁻NK1.1⁺7-AAD⁻ cells by FACS (purity > 98%). Data were analyzed using CellQuest Pro (BD Bioscience) and FlowJo (Tree Star) software.

RNA was isolated from FACS-sorted NK cells (purity ≥ 99%) with the RNeasy Mini Kit (QIAGEN), according to the manufacturer’s instructions. Contaminating DNA was removed using the TURBO DNA-free kit (Ambion Life Technologies). Similar amounts of RNA were used for cDNA synthesis (ProtoScript M-MuLV First Strand Synthesis kit; New England Biolabs). As a negative control, a reaction containing all components with the exception of reverse transcriptase was carried out (−RT control). Quantitative real-time PCR (qPCR) was performed with LightCycler 480 SYBR Green I Master mix, reaction plates, LightCycler 480 instrument, and software (Roche Diagnostics). The reaction mix contained 1× LightCycler 480 SYBR Green I Master mix, 0.5 μM forward primer, and 0.5 μM reverse primer in a total volume of 10 μl. Relative expression of CTLA-4 compared with a housekeeping gene (HPRT) was calculated using the Δ-crossing point method. The following primers were used: CTLA-4-f: 5′-GCTTCCTAGATTACCCCTTCTGC-3′, CTLA-4-r: 5′-CGGGCATGGTTCTGGATCA-3′, HPRT-f: 5′-CTTTGCTGACCTGCTGGATT-3′, and HPRT-r: 5′-TATGTCCCCCGTTGACTGAT-3′.

NK cells were incubated for 8 h with plate-bound control EphA4-IgG or B7-1–IgG fusion protein (both from R&D), coated overnight at 4°C in 50 μl of 2 μg/ml fusion protein solution in PBS, or for 24 h with mDCs (1:2 NK/DC ratio), in the presence of 1700 U/ml rhIL-2. A total of 20 μg/ml control-IgG or CTLA-4–IgG fusion proteins (R&D Systems) were used in blocking experiments. Tumor-infiltrating and blood NK cells were stimulated in the presence of 200 U/ml rhIL-2 for 24 h at 37°C. IL-2–expanded NK cells were stimulated with 50 ng/ml PMA and 750 ng/ml ionomycin (Sigma-Aldrich) for 8 h or with 10 ng/ml PMA and 150 ng/ml ionomycin for 24 h at 37°C in the presence of 1700 U/ml rhIL-2. Levels of IFN-γ in culture supernatants were determined by ELISA (BD Biosciences). For intracellular staining of IFN-γ, a protein-transport inhibitor (GolgiStop; BD Biosciences) was added to the cell culture at 37°C for 4 h before harvesting. IFN-γ was detected (mAb clone XMG1.2 from BioLegend or eBioscience) after cell fixation and permeabilization using the Foxp3 Staining Buffer Set (eBioscience).

Statistical significance was assessed using the Student t test and GraphPad Prism software (GraphPad Software, La Jolla, CA). The differences between groups were considered significant when p ≤ 0.05.

To elucidate expression and regulation of CTLA-4 in mouse NK cells, we first monitored its expression by intracellular staining upon activation with IL-2. We did not detect CTLA-4 expression in freshly isolated splenic NK cells, but culture of NK cells in the presence of IL-2 resulted in an induction of CTLA-4 (Fig. 1A). We also confirmed expression of CTLA-4 mRNA in IL-2–activated, but not freshly isolated NK cells by qPCR (Fig. 1B). Expression of CTLA-4 protein increased over time during culture in IL-2 (Fig. 1C). In parallel, we observed that freshly isolated splenic NK cells expressed very low levels of CD28 (Fig. 1A). In concordance with previously published reports (21, 23, 24), culture in IL-2 led to its upregulation. The increase in CD28 surface expression and CTLA-4 intracellular levels occurred with similar kinetics (Fig. 1C) and was dependent on the dose of IL-2 present in the cultures (Fig. 1D). Of note, IL-2 activation of NK cells not only modulated expression of CD28/CTLA-4, it induced global changes in NK cell phenotype, including downregulation of activating receptors (NKp46, NK1.1, NKG2D), adhesion molecules (CD11b, ICAM-1), and B7-1 and B7-2 (Supplemental Fig. 1). As shown in Fig. 1E, CD28 expression also was upregulated on NK cells in the presence of IL-15, whereas, under our experimental conditions, the presence of IL-15 did not induce expression of CTLA-4.

In T cells, surface-exposed CTLA-4 is rapidly internalized and either degraded or recycled back to the cell surface (9). To investigate whether CTLA-4 protein can also be detected on the surface of NK cells, we stained IL-2–activated NK cells with the anti–CTLA-4 mAb at 37°C for 4 h, which enables labeling of surface-exposed CTLA-4 over the time period of incubation with mAb (27). Using this method, very small amounts of CTLA-4 protein were detectable on the cell surface of IL-2–activated NK cells (Fig. 1F).

Next, we aimed to identify additional stimuli regulating CTLA-4 and CD28 protein expression in IL-2–activated NK cells. CTLA-4 expression in NK cells was first observed 48 h after initiation of culture with IL-2 (data not shown). Upon addition of IL-12 and IL-18 to IL-2–supplemented NK cell cultures, a small percentage of CTLA-4–expressing NK cells was already detected after 24 h (Supplemental Fig. 2A). In NK cells expanded for 1 wk in IL-2, addition of IL-12 24 h before harvesting increased surface levels of CD28 and intracellular levels of CTLA-4 (Fig. 2A, 2B). Stimulation with IL-12 and IL-18 resulted in an upregulation of CTLA-4 expression, whereas CD28 was slightly reduced. On the contrary, the presence of TGF-β increased CD28 expression, while leading to a reduction in CTLA-4. Of note, addition of TGF-β significantly reduced CTLA-4 expression in NK cells, even in the presence of IL-12 and IL-18 (Fig. 2A, 2B). Upregulation of CTLA-4 and CD28 was detected 18 h after IL-12 addition and increased over time (Fig. 2C). Similarly, TGF-β–mediated downregulation of CTLA-4 was detected 18 h after cytokine addition to IL-2–supplemented cultures (Fig. 2C). An increase in CD28 and decrease in CTLA-4 expression also were observed in the presence of PMA and ionomycin, which activate protein kinase C and Ca²⁺ signaling, respectively (Supplemental Fig. 2B). Effects of added cytokines on CD28/CTLA-4 expression also were observed using NK cells generated from Rag2-deficient mice, ruling out the possibility that the observed changes were due to NKT or T cell contamination of our cultures. Taken together, our data show that different cytokines differentially regulate the expression of CD28 and CTLA-4 in IL-2–activated NK cells, which could potentially affect NK cell responses controlled by these receptors.

Previous studies (21–23) indicated that CD28 triggering supported NK cell proliferation, cytotoxicity, and cytokine secretion. Because IL-2–activated NK cells expressed both CD28 and CTLA-4, we investigated how these molecules regulate NK cell responses to their shared ligand, B7-1. Although B7-1 and B7-2 were detected on a subpopulation of freshly isolated splenic NK cells, these molecules were downregulated during expansion with IL-2 and were barely detectable on day 6 after initiation of culture (Supplemental Fig. 1B). Thus, IL-2–activated NK cells express CD28/CTLA-4, but, in parallel, B7-1/2 expression is absent/low, suggesting that it is unlikely that these molecules interact on activated NK cells under our experimental conditions. Upon stimulation with plate-bound B7-1–IgG fusion protein (B7-1–Fc), IL-2–expanded NK cells produced increased amounts of IFN-γ (Fig. 3A). As shown in Fig. 3B, high levels of intracellular IFN-γ correlated positively with high CD28 expression. In parallel, higher levels of IFN-γ were detected in CTLA-4^low/− NK cells compared with the CTLA-4⁺ subset. Among IFN-γ–producing CD28⁺ NK cells, the cells expressing CTLA-4 produced less IFN-γ compared with CTLA-4^low/− cells. These data suggest that IFN-γ production in response to B7-1 is positively regulated by CD28 and negatively regulated by CTLA-4. Accordingly, IFN-γ production in response to B7-1–Fc was completely abrogated in CD28-deficient NK cells, indicating that it was mediated by CD28 (Fig. 3C, left panel). We also observed that, in concordance with previously published data (22), B7-1–transduced RMA-S lymphoma cells also induced higher amounts of IFN-γ compared with a vector control–transduced cell line upon coculture with NK cells (Supplemental Fig. 3A). This increase was abrogated by preincubation of tumor cells with a blocking anti–B7-1 mAb or when CD28-deficient NK cells were used, indicating the functional interaction between B7-1 expressed by tumor cells and NK cell–expressed CD28. Of note, B7-1 expression did not influence the sensitivity of RMA-S cells to NK cell killing (Supplemental Fig. 3B). Moreover, CD28-deficient NK cells exerted similar cytotoxic responses to B7-1–expressing B16 cells and B7-1–expressing RMA-S cells, as well as to the prototypic NK cell target cell line, YAC-1 (Supplemental Fig. 3B), compared with WT NK cells.

To further investigate the effect of CTLA-4 on NK cell IFN-γ production, IL-2–expanded NK cells derived from WT or CTLA-4–deficient OT-I–transgenic mice were stimulated with plate-bound B7-1–Fc or control Fc. CTLA-4–deficient NK cells released significantly more IFN-γ compared with the respective WT NK cells (Fig. 3C, right panel). Of note, expansion in the presence of IL-2 led to comparable induction of CTLA-4 in NK cells derived from both CD28-deficient and WT mice, and CD28 was upregulated in a similar manner in OT-I–transgenic WT and CTLA-4–deficient NK cells (Supplemental Fig. 4A). In parallel, expression of a panel of activating and inhibitory NK receptors, as well as of NK cell maturation markers, was similar between age-matched gene-deficient and respective WT NK cells (Supplemental Fig. 4B). IL-2–expanded NK cells from respective WT and CD28- or CTLA-4–deficient mice produced comparable amounts of IFN-γ after stimulation with PMA and ionomycin (Supplemental Fig. 4C), implying that, in general, NK cells derived from these mice are able to respond at levels comparable to their respective control WT NK cells. Thus, our results indicate that IFN-γ release by NK cells in response to B7-1 is mediated via CD28 and is negatively regulated by CTLA-4.

B7-1 and B7-2, the ligands of CD28 and CTLA-4, are primarily expressed by APCs, such as DCs. Upon activation, DCs undergo maturation and upregulate expression of both B7-1 and B7-2 (28). mDCs were shown to activate NK cells to secrete IFN-γ (29). Thus, we investigated whether the CD28/CTLA4:B7-1/2 axis contributes to the NK cell/mDC cross-talk. Upon coculture with LPS-matured DCs, IL-2–activated NK cells produced increased amounts of IFN-γ (Fig. 3D). IFN-γ production was partially abrogated by pretreatment of mDCs with CTLA-4–Fc, which binds to the ligands, B7-1 and B7-2, and prevents their interaction with the cognate receptors expressed on NK cells (Fig. 3D, left panel). Furthermore, upon coculture with mDCs, mouse CD28-deficient NK cells released significantly less IFN-γ, whereas CTLA-4–deficient NK cells secreted more IFN-γ compared with WT NK cells (Fig. 3D, middle and right panels). We also determined DC apoptosis in the presence of IL-2–activated NK cells, as measured by annexin V staining. In these experiments, similar frequencies of apoptotic DCs were observed, regardless of whether they were cocultured with CD28-deficient cells or with their respective WT NK cells (data not shown). These data demonstrate that the interaction of CD28, expressed by NK cells, with its ligands on mDCs contributes to the NK cell IFN-γ release induced by mDC, whereas CTLA-4 negatively regulates this response.

We further investigated whether CD28 and CTLA-4 expression in NK cells also was detectable under pathological conditions, such as in malignant disease. For that purpose, we inoculated syngeneic WT BL/6 mice s.c. with 10⁶ RMA-S lymphoma, B16 melanoma, or LL/2 carcinoma cells, which led to progressive tumor growth. Our data demonstrate that a subset of NK cells infiltrating s.c. RMA-S lymphoma expressed CD28 and CTLA-4 (Fig. 4). CTLA-4 also was detected in B16 melanoma– and LL/2 carcinoma–infiltrating NK cells, whereas CD28 expression was mostly low or undetectable. Of note, CTLA-4 expression was detected only in tumor-infiltrating cells, but not in NK cells from blood (Fig. 4) or spleen (data not shown) of tumor-bearing mice. Moreover, our microarray analyses of NK cells from RMA-S tumor–bearing mice revealed elevated CTLA-4 transcript levels in tumor-infiltrating, compared with blood NK cells, indicating transcriptional regulation of CTLA-4 (A. Stojanovic, L. Li, N. Gretz, and A. Cerwenka, unpublished observations). CTLA-4 expression was not detected in NK cells infiltrating small tumors (data not shown), but it was observed in NK cells isolated from tumors ≥10 mm in diameter, suggesting that its expression might correlate with tumor progression. Importantly, NK cells purified from RMA-S tumor tissue responded to stimulation with plate-bound B7-1–Fc with increased IFN-γ production compared with blood NK cells (Fig. 5A). Thus, functional receptors for B7-1 on tumor-infiltrating NK cells exist that might modulate NK cell responses to B7-1–expressing cells in the tumor tissue. Analysis of tumor cells from the established s.c. tumors revealed that B7-1 and B7-2 were not detectable on RMA-S, B16, and LL/2 tumor cells (Fig. 5B). In parallel, higher levels of B7-1 and B7-2 were detected on the myeloid fraction of tumor-infiltrating immune cells, such as the CD11b⁺F4/80⁺ and CD11b⁺CD11c⁺F4/80⁻ subsets, comprising mainly macrophages and DCs, respectively (Fig. 5B).

NK cell activation is the result of the integration of activating and inhibitory signals delivered via activating and inhibitory receptors (3). Many of these receptors and their ligands important for NK cell/tumor cell interaction and NK cell cross-talk with other immune cells have been identified. In this study, we provide evidence that additional receptors, CTLA-4 and CD28, upregulated upon activation of mouse NK cells with cytokines, are involved in the regulation of NK cell function. We show that CD28/CTLA-4 differentially modulated IFN-γ release by IL-2–activated NK cells in response to their ligand, B7-1, and upon interaction with mDCs. Because both CD28/CTLA-4–expressing NK cells and B7-1/2–expressing myeloid cells are present in the tumor tissue of different tumor entities, the NK cell responses mediated via CD28 and CTLA-4 could be of importance during malignant disease and/or therapy regimens based on CD28 and CTLA-4 targeting.

Our in vitro experiments revealed that IL-2 induced expression of CTLA-4 in mouse NK cells, whereas, in concordance with other studies (21, 23, 24), CD28 expression was upregulated. Most importantly, NK cells infiltrating certain solid tumors expressed CTLA-4 and CD28. We observed that tumor-infiltrating NK cells also expressed increased transcript and protein levels of IL-2Rα (CD25) compared with blood NK cells (A. Stojanovic et al., unpublished observations), which might result in their greater responsiveness to IL-2 and lead to CTLA-4 induction and CD28 upregulation. T cells are considered the main producers of IL-2 during ongoing immune responses (30); thus, they are likely to produce IL-2 in tumor tissue. Accordingly, we did not detect CTLA-4 expression in NK cells infiltrating RMA-S tumors in Rag2-deficient mice that lack T cells (data not shown). In concordance with a report by Hunter et al. (24), our study revealed that IL-15, closely related to IL-2 as a member of the common γ-chain cytokine family, induced CD28 upregulation on NK cells. In parallel, under our experimental conditions, we did not detect CTLA-4 expression in NK cells upon IL-15 culture. It is unknown whether IL-15 can modulate CTLA-4 expression in concert with other cytokines or microenvironmental stimuli.

Upon addition of IL-12 and IL-18 to our in vitro cultures, CTLA-4 expression increased further. IL-12 and IL-18 are mainly produced by activated macrophages and DCs and are cytokines shown to be present in the tumor microenvironment (31, 32). Thus, it is possible that IL-12 and IL-18 also might have been involved in the induction of CTLA-4 in tumor-infiltrating NK cells. IL-12 and IL-18, produced by activated DCs, are also potent inducers of IFN-γ production in NK cells. Because our data demonstrate that CTLA-4, expressed by NK cells, negatively regulated IFN-γ production in response to mDCs, the induction of CTLA-4 by IL-12 and IL-18 on NK cells could represent a negative-feedback loop limiting IFN-γ production in response to DC-expressed B7-1 and B7-2. The presence of IL-12 increased CD28 expression in IL-2–activated NK cells, whereas, upon exposure to IL-18, CD28 expression remained unchanged. Of note, addition of the combination of IL-12 and IL-18 did not result in increased CD28 expression, suggesting that signals delivered by IL-18 could counteract upregulation of CD28 mediated by IL-12. Furthermore, CD28 expression might be regulated indirectly by additional soluble factors released upon stimulation with IL-12/18 (33). Addition of TGF-β to in vitro NK cell cultures downregulated IL-2–induced CTLA-4 expression in NK cells. TGF-β was shown to have multiple effects on NK cells, including the downregulation of activating receptors (7) and the inhibition of IFN-γ release induced by IL-12 and IL-18 (6). Thus, the final outcome of NK cell exposure to TGF-β would depend on the signal integration in pathways of NK cell activation (e.g., by cytokines or via activating receptors) and inhibition (e.g., CTLA-4), which are both differentially affected by TGF-β. Our data show that TGF-β reduced CTLA-4 expression in IL-2–expanded NK cells. TGF-β is present in the microenvironment of many tumor entities, and its role in immune suppression is very well documented (34–36). Whether TGF-β has a similar effect on tumor-infiltrating NK cells in concert with other microenvironmental factors in the tumor remains unknown. In the tumor tissue, the effect of TGF-β on CTLA-4 expression might depend on its concentration relative to other cytokines that might differentially accumulate during tumor progression. Our experiments revealed that CTLA-4 was detectable on NK cells in tumor tissues, despite the possible presence of TGF-β, and could play an important role in NK cell effector responses and during immunotherapy applying anti–CTLA-4–blocking Abs.

Our data show that RMA-S cells overexpressing B7-1 also induced enhanced levels of IFN-γ production by NK cells compared with vector control–transduced RMA-S cells. This response was abrogated in NK cells derived from CD28-deficient mice. Preincubation of target cells with a blocking anti–B7-1 mAb reduced NK cell IFN-γ production to the levels induced by control cells, indicating that the increased IFN-γ production was dependent on the presence of B7-1 on target cells. However, expression of B7-1 on target cells or the presence of CD28 on NK cells did not influence the cytotoxic response of IL-2–activated NK cells. It is well documented that DCs also can be killed by activated NK cells (37–39). Although we observed increased apoptosis of DCs in the presence of IL-2–activated NK cells, this effect was independent of CD28 expression on NK cells (data not shown). Thus, in our experimental system, the CD28:B7-1 axis controls IFN-γ release, but is not involved in NK cell cytolytic responses. Several studies investigated the NK cell responses to B7-1–expressing targets in mice. Although the involvement of CD28 in these responses was addressed in certain studies, the role of CTLA-4 remains undetermined. Kelly et al. (22) reported that NK cell proliferation, IFN-γ production, and cytotoxicity were increased in response to B7-1–expressing RMA-S cells compared with vector control cells. Similarly, P815 B7-1⁺ cells induced higher levels of IFN-γ in IL-2–expanded NK cells compared with control cell lines (24). Evidence exists that B7-1, expressed by the target cell, can costimulate the responses triggered through other NK cell receptors, such as Ly49D (40). Geldhof et al. (21) showed that CD28 played a crucial role in the cytotoxic response of IL-2/IL-12–expanded NK cells to B7-1⁺ syngenic, but not B7-1⁺ allogenic, tumor cells. Finally, Chambers et al. (41) reported that the enhanced lysis of B7-1–expressing RMA-S cells by syngenic NK cells compared with B7-1⁻ RMA-S cells did not depend on CD28 or CTLA-4, implying the existence of an additional receptor for B7-1 on NK cells. Together, these data indicate that the contribution of CD28/CTLA-4 to the NK cell response to B7-1–expressing targets greatly depends on the mode of NK cell activation, expression of other activating/inhibitory ligands, and the type of tumor targets used in the studies. In human NK cells, controversial data generated with different anti-CD28 mAbs were reported with regard to the expression of CD28 (42, 43). Depending on the target cells used, either activation of human NK cells (44–46) or the lack of increased responses to B7-expressing targets (47) was observed. Using different anti-CD28 mAbs, we were unable to detect CD28 on human IL-2–activated NK cells, whereas our preliminary data showed that activation with IL-2 increased expression of CTLA-4. In addition, stimulation with IL-12 and/or IL-18 for 24 h further increased CTLA-4 levels in human NK cells (data not shown). These results indicate that expression of CTLA-4 might be regulated by similar mechanisms in mouse and human NK cells. Its function in human NK cells remains to be investigated.

B7-1 is endogenously expressed only by certain tumor cells (48–50), but it is often highly expressed by activated APCs. In this context, our data demonstrate that CD28/CTLA-4 differentially modulate IFN-γ release by NK cells in response to mDCs. Addition of a CTLA-4–Fc to DCs, which blocks B7-1 and B7-2 binding to corresponding receptors on NK cells, significantly inhibited NK cell IFN-γ production in response to mDCs. Of note, in our experiments, the reduction in IFN-γ production upon CTLA-4–Fc treatment was incomplete. Thus, it is likely that other factors could be involved in the NK/DC cross-talk. Several cell surface receptors and soluble factors, such as IL-12, were shown to mediate the NK/DC cross-talk in humans and mice (21, 46, 51–55). These factors, in addition to CD28/CTLA-4, might also be involved in regulating NK cell IFN-γ production in response to mDCs in our experimental system. Our data demonstrate that both CTLA-4 and CD28 can be expressed by tumor-infiltrating NK cells and that their corresponding ligands were detectable on other tumor-infiltrating immune cells. Several factors might contribute to the regulation of CD28/CTLA-4 expression in tumor-infiltrating NK cells. In this context, we observed that cytokines, such as IL-2, IL-12, IL-18, and TGF-β, which might be present at different levels in the tumor microenvironment, modulated CD28/CTLA-4 expression on NK cells in vitro. Thus, the interplay of these factors and their relative amounts in the tumor microenvironment might determine the relative levels of CD28 and CTLA-4 expression in tumor-infiltrating NK cells and their response to B7-1/2. Moreover, our data indicate that, in certain tumors, CTLA-4 can be expressed in tumor-infiltrating NK cells while CD28 is absent. It is possible that CTLA-4 exerts functions in addition to controlling CD28-dependent responses in NK cells. Furthermore, by acting in a cell-extrinsic manner (20), NK cell–expressed CTLA-4 could influence the responses of other tumor-infiltrating cells, such as DCs and T cells.

Blockade of CTLA-4 by mAbs was shown to enhance T cell effector functions and to delay tumor progression in several experimental tumor models (56–59). Our data demonstrate that tumor-infiltrating NK cells also represent potential targets of anti–CTLA-4 mAb-based immunotherapy and, thus, might contribute to orchestrating efficient immune responses upon CTLA-4 blockade. Combinatorial strategies incorporating CTLA-4 blockade and NK cell–based therapies might lead to clinical benefits in the treatment of cancer.

We thank the Deutsches Krebsforschungszentrum Animal Laboratory for animal care and Flow Cytometry Core Facility for cell sorting, Dr. Markus Feuerer and Dr. Jan Hettinger for providing qPCR primers, Dr. Uta E. Höpken and Dr. Michaela Kern for collaboration and helpful discussion, and Dr. Margareta Correia for critical reading of the manuscript and helpful discussion.

The authors have no conflicts of interest.