Bacterial DNA contains a high frequency of unmethylated CpG motifs that stimulate immune cells via TLR9. NK cells express a low-affinity activating receptor for the Fc portion of IgG (FcγRIIIa), but were not thought to express TLR9 protein. The direct response of NK cells to CpG oligodeoxynucleotides (ODN) in the presence of FcR stimulation was investigated. Human NK cells cultured in the presence of CpG ODN plus immobilized IgG or Ab-coated tumor cells secreted large amounts of IFN-γ (>2000 pg/ml), whereas cells stimulated with Ab alone, CpG ODN alone, or Ab and control ODN produced negligible amounts. Enhanced secretion of IL-8, macrophage-derived chemokine, and MIP-1α was also observed after costimulation. NK cell cytokine production was not the result of interactions with APCs or their cytokine products. Flow cytometric analysis revealed that 36 ± 3.5% of human NK cells expressed basal levels of TLR9. TLR9 expression in human NK cells was confirmed by immunoblot analysis. Only TLR9-expressing NK cells responded to CpG ODN and Ab, because cytokine production was not observed in NK cells from TLR9-deficient mice. Mice receiving CpG ODN and HER2/neu-positive tumor cells treated with an anti-HER2 Ab exhibited enhanced systemic levels of IFN-γ compared with mice receiving either agent alone. TLR9−/− animals reconstituted with TLR9+/+ NK cells secreted IFN-γ in response to CpG ODN and Ab-coated tumor cells. These findings indicate that CpG ODN can directly enhance the NK cell cytokine response to Ab-coated targets via activation of TLR9.
Natural killer cells are bone marrow-derived, large granular lymphocytes that participate in innate immune responses to viruses, bacteria, and neoplastic cells (1). NK cells constitutively express receptors for several cytokines, including IL-10, IL-12, IL-15, and IL-18, that are produced by monocytic cells in response to bacterial or viral infections (2, 3, 4, 5). NK cells have been shown to respond to these factors with the secretion of factors with potent immune modulatory properties (e.g., IFN-γ, TNF-α, and MIP-1α) that aid in the clearance of microbial pathogens (6). Although many of the mechanisms underlying the activation of innate immune cells in response to infection have been elucidated, direct activation of NK cells by microbial products remains less well characterized.
Bacterial pathogens express or shed a number of factors that are recognized by components of the innate immune system, mainly via a class of receptors known as the TLRs. Bacterial DNA can stimulate innate immune cells because it contains a high frequency of unmethylated CpG motifs, which allows the immune system to distinguish microbial from eukaryotic DNA (7). In the hope of stimulating innate immune responses for the treatment of various diseases, investigators have developed synthetic oligodeoxynucleotides (ODN) 4 that contain a high frequency of CpG dinucleotides. These ODN, like intact bacterial DNA, are known to act as agonists for TLR9, one of the key receptors for CpG motifs (8, 9). Unlike the other TLRs, which localize to the cell surface, TLR9 is located intracellularly (10, 11). Bacterial DNA is thought to stimulate B cells and plasmacytoid dendritic cells (12, 13, 14), because these are the only human cell types that have been shown to express functional levels of TLR9. Although Hornung et al. (15) found low levels of TLR9 mRNA within NK cells, the expression of functional TLR9 protein within resting or activated NK cells has not been reported to date. And although NK cell production of IFN-γ in response to treatment with CpG motifs has been described, the overall levels of cytokine secretion seen in these studies were either very low for the large number of cells used (16) or potentially the result of interactions with APCs (17).
In contrast, NK cell activation is known to occur readily in response to FcR stimulation. NK cells are unique in that they constitutively express an activating, low-affinity receptor for the Fc portion of IgG (FcγRIIIa or CD16) that enables them to interact with Ab-coated targets (18). In the current report we examined the ability of CpG-containing ODN to enhance NK cell cytokine secretion in response to FcR stimulation by immobilized IgG or Ab-coated human breast cancer cells. Costimulation of purified NK cells in this manner resulted in potent, synergistic production of proinflammatory cytokines and chemokines, including IFN-γ and MIP-1α. In addition, we have provided evidence that CpG ODN directly activate NK cells via TLR9, a mechanism of action for CpG ODN on NK cells that has not been previously reported. These findings suggest a role for NK cells in the Ab response to bacterial pathogens and indicate a potential mechanism by which concurrent administration of CpG ODN with therapeutic mAbs may mediate antitumor effects against human malignancies.
Materials and Methods
ODNs and Abs
The following CpG-containing ODN were used: CpG 2336 (GGGGACGACGTCGTGGGGGGG, activating for NK cells and plasmacytoid dendritic cells), CpG 2006 (TCGTCGTTTTGTCGTTTTGTCGTT, activating for B cells), and CpG 2395 (TCGTCGTTTTCGGCGCGCGCCG, broadly activating) (19). A non-CpG-containing ODN (2243) was used as a control (GGGGGAGCATGCTGGGGGGG). All ODN were purchased from the Coley Pharmaceutical Group. Polyclonal human IgG (huIgG) and polyclonal murine IgG (muIgG) were purchased from Sigma-Aldrich. Herceptin, a humanized anti-HER2 mAb, was provided by Genentech.
Isolation of NK cells
NK cells (>95% CD56+) were isolated directly from fresh peripheral blood Leukopacks (American Red Cross) by 30-min incubation with RossetteSep mixture (Stem Cell Technologies) before Ficoll Hypaque (Sigma-Aldrich) density gradient centrifugation. Human NK cells were cultured in RPMI 1640 medium supplemented with 10% heat-inactivated pooled human AB serum (HAB; C-six Diagnostics), 100 U/ml penicillin, 100 μg/ml streptomycin, and 0.25 μg/ml amphotericin B (10% HAB medium) (20). Murine NK cells were harvested from whole splenocytes by magnetic labeling. Splenocytes were incubated with PE-labeled anti-DX5 and anti-NK1.1 Abs (Beckman Coulter) for 30 min at 4°C. Cells were then labeled with anti-PE magnetic microbeads (Miltenyi Biotec) for an additional 30 min at 4°C and passed over an LS+ magnetic column (Miltenyi Biotec). Isolated murine NK cells were routinely >95% DX5+/NK1.1+ by FACS analysis.
In vitro costimulation assays
For immobilized IgG experiments, wells of a 96-well, flat-bottom plate were coated with 100 μg/ml huIgG or muIgG in cold PBS overnight at 4°C, washed with cold PBS, and then plated with immune cells (2 × 105 cells/well) and 1 μM CpG ODN. At the indicated time points, cell-free culture supernatants were harvested and analyzed for levels of various cytokines and chemokines by ELISA (R&D Systems). For the in vitro coculture assay, wells of a 96-well, flat-bottom culture plate were seeded with either the HER2-overexpressing human breast cancer cell line SKBR3 or the HER2-negative human breast cancer cell line MDA-468 (American Type Culture Collection) at a density of 5 × 104 tumor cells/well. Tumor cells were grown to confluence overnight and then treated with 100 μg/ml Herceptin for 1 h at 37°C, as previously described (20). After washing off unbound Herceptin, purified NK cells were then added at 2 × 105 cells/well in 200 μl of 10% HAB medium containing 1 μM CpG ODN. Control conditions consisted of NK cells treated with medium alone, Herceptin-coated tumor alone, or CpG ODN alone. Culture supernatants were harvested after 72 h and analyzed by ELISA.
Flow cytometry for TLR9 expression
Resting NK cells from normal donors (n = 20) were harvested, washed twice in cold flow buffer (5% FBS in PBS), and fixed and permeabilized using a commercially available kit (BD Pharmingen). Cells were then blocked with normal rat IgG (Sigma-Aldrich) for 10 min and stained for intracellular TLR9 for a 30-min incubation with a PE-labeled TLR9 mAb (eBioscience). CD19+ B cells were isolated by magnetic labeling (Miltenyi Biotec) and stained for TLR9 as a positive control. The expression of TLR9 in human NK cells was confirmed by immunoblot analysis as previously described (20).
In vivo costimulation assay
A BALB/c murine colon carcinoma line overexpressing human HER2/neu, CT-26HER2/neu, was obtained from Dr. P. Kaumaya (Ohio State University, Columbus, OH). CT-26HER2/neu cells were incubated at 4°C in flow buffer (cell density, 1 × 107 cells/ml) with either Herceptin or normal huIgG (both at 1 mg/ml) for 45 min. Cells were then washed twice in sterile PBS, resuspended in PBS, and injected i.p. into female BALB/c mice, aged 5–7 wk (The Jackson Laboratory). Mice also received a separate i.p. injection of 10 μg of CpG ODN. Control groups (n = 8) received injections of IgG-treated tumor alone, Herceptin-coated tumor alone, or IgG-treated tumor and CpG ODN. Serum was harvested 6 h postinjection and analyzed for IFN-γ levels by ELISA (Pierce Endogen). In some experiments, mice were depleted of NK cells through i.p. injections of an anti-asialo-GM1 Ab (Wako BioProducts) three times a week for 1 wk before coadministration of the tumor cells and CpG ODN (200 μg/injection). Control mice received injections of PBS. This technique resulted in >93% depletion of NK cells in the peripheral blood and spleen, as determined by flow cytometric analysis with an anti-murine DX5 Ab.
In vivo costimulation after adoptive transfer of NK cells
An in vivo costimulation assay was performed on TLR9−/− mice that had been reconstituted with NK cells from wild-type (i.e., TLR9+/+) animals before coadministration of Ab-coated tumor cells and CpG ODN. Female, 4- to 6-wk-old TLR9−/− mice on a C57BL/6 background were obtained from Coley Pharmaceuticals. Age-matched, wild-type C57BL/6 mice were used as controls (The Jackson Laboratory). Murine NK cells were isolated as described above. Purified TLR9+/+ NK cells (1 × 106) were injected i.p. into TLR9−/− recipient mice. Control groups consisted of TLR9+/+ mice receiving TLR9−/− NK cells, TLR9+/+ mice receiving TLR9+/+ NK cells, and TLR9−/− mice receiving TLR9−/− NK cells. The following day, P815 cells murine lymphoma cells (American Type Culture Collection) were harvested and incubated for 45 min at 4°C in flow buffer (cell density, 1 × 107 cells/ml) with 1 mg/ml rabbit anti-mouse lymphocyte IgG as the Ab source (Accurate Chemical and Scientific). Cells were washed twice in sterile PBS and injected i.p. into recipient mice with or without a concurrent injection of 10 μg of CpG ODN. Serum was harvested 6 h after injection and analyzed for IFN-γ content by ELISA.
ELISA cytokine levels were analyzed by Student’s t test, with p < 0.05 considered significant.
Human NK cells costimulated with CpG ODN and immobilized huIgG exhibit enhanced cytokine production
Costimulation of NK cells with CpG ODN 2336 (1 μM) and immobilized IgG led to potent IFN-γ production, whereas NK cells exposed to CpG ODN alone released smaller amounts of IFN-γ over the 72-h culture period (Fig. 1,A). Costimulation with a control ODN lacking CpG motifs resulted in minimal cytokine secretion. A time-course study revealed that NK cell production of IFN-γ began within 12 h of costimulation and peaked at 72 h (Fig. 1,B). The NK cell cytokine response was CpG ODN dose dependent (range, 0.10–2.5 μM), with concentrations in the 1-μM range resulting in optimal synergy (Fig. 1,C). Costimulation of purified NK cells with CpG ODN and immobilized IgG also resulted in synergistic production of the chemotactic factors MIP-1α, IL-8, and macrophage-derived chemokine (MDC), but not MCP-1, GRO-α, or MIP-1β (Fig. 1,D and data not shown). Also, NK cell production of GM-CSF and TNF-α was not enhanced in the presence of CpG ODN (data not shown). The effects of CpG ODN were sequence specific, because NK cells produced lesser amounts of IFN-γ in response to stimulation with CpG motifs known to be activating to B cells or to a variety of cells (19) (Fig. 1,E). IFN-γ secretion was maximal at a 1-μM concentration of each ODN tested (ODN 2336, 2395, and 2006; Fig. 1 C and data not shown). Because NK cell IFN-γ secretion was highest in response to stimulation with CpG ODN 2336, this ODN was used throughout this study.
NK cell cytokine production in response to Ab-coated tumor cells is enhanced in the presence of CpG ODN
Based on these results, we hypothesized that CpG ODN would enhance the immune response of NK cells to Ab-coated tumor targets. SKBR3 (HER2-overexpressing) and MDA-468 (HER2-negative) human breast cancer cells were treated with an anti-HER2 mAb (Herceptin) and cultured with purified NK cells in the presence or the absence of CpG ODN. Costimulation of NK cells with CpG ODN and Herceptin-coated SKBR3 cells resulted in synergistic production of IFN-γ (Fig. 2,A). In contrast, there was minimal production of IFN-γ by NK cells in response to HER2-negative cells, regardless of the presence of CpG ODN. Costimulation of NK cells also resulted in the synergistic production of the chemokines MIP-1α, IL-8, and MDC, but not MCP-1, GRO-α, or MIP-1β (Fig. 2,B and data not shown). Although IFN-γ secretion in response to CpG ODN and control Ab-treated tumor cells was minimal, this stimulation induced NK cells to secrete significant amounts of MIP-1α and IL-8 (Fig. 2,B and data not shown). In addition, NK cells exposed to Ab-coated tumor cells (or immobilized IgG) alone consistently produced low, but measurable, levels of IFN-γ (Fig. 1, A–C, and Fig. 2,A) and MIP-1α (Fig. 2 B), suggesting that FcR stimulation primes the NK cell for cytokine secretion.
NK cells are the primary source of IFN-γ produced in response to CpG ODN and immobilized Ab
To determine the effect of APCs on NK cell cytokine secretion, the responses of purified NK cells and whole PBMCs to costimulation with CpG ODN and immobilized IgG were compared. Whole PBMCs or purified NK cells from the same donor were cultured in the presence or the absence of CpG ODN on plates precoated with human IgG. IFN-γ production by 2 × 105 PBMCs (containing ∼15% NK cells) in response to costimulation was minimal compared with that of an equal number of purified NK cells (Fig. 3,A). When the response of 2 × 105 PBMCs was compared with that of 3 × 104 NK cells from the same donor (i.e., ∼15% of the PBMC cell number), both populations consistently produced the same amount of IFN-γ after costimulation with CpG ODN and IgG (data not shown). CD56+ NK cells, CD19+ B cells, CD3+ T cells, and CD14+ monocytes (all >95% pure) were then isolated from a single donor and cultured in the presence or the absence of CpG ODN on IgG-coated plates. Purified NK cells, but not purified T cells, B cells, or monocytes, responded to CpG ODN stimulation with the secretion of IFN-γ (Fig. 3,B). The culture supernatants were also assayed for the presence of IL-6, because B cell secretion of IL-6 in response to CpG ODN is well documented (13, 17). Human B cells, but not the other cell types, responded to stimulation with CpG ODN with the secretion of large quantities of IL-6 (Fig. 3 C), providing a positive control for the functionality of B cells within our immobilized IgG assay compared with their lack of IFN-γ secretion. Of note, B cell secretion of IL-6 in response to CpG ODN did not increase in the presence of immobilized IgG. Collectively, this evidence demonstrated that NK cells are the primary source of IFN-γ produced by PBMCs in response to CpG ODN and immobilized Ab.
NK cells respond directly to CpG DNA
To confirm that the effect of CpG ODN on NK cells was direct and not induced by IL-12, IFN-α, or TNF-α secreted by APCs possibly contaminating the purified NK cell preparations, NK cell supernatants from the CpG/immobilized IgG assays were analyzed for these factors by ELISA. IL-12 and IFN-α levels in these supernatants were consistently undetectable by ELISA (≤20 pg/ml). TNF-α levels, although detectable, were the same across all four conditions (∼200 pg/ml). Furthermore, addition of neutralizing Abs against IL-12, IFN-α, or TNF-α to NK cell cultures did not diminish NK cell IFN-γ secretion in response to CpG ODN and immobilized IgG (data not shown). This evidence suggested that the synergistic production of IFN-γ by purified NK cells in response to costimulation with CpG ODN and IgG proceeded independently of these APC-derived cytokines.
Based on the potential for interaction between B cells or monocytes and NK cells through cytokine signaling and/or direct cell contact, depletion experiments were conducted in the context of the costimulation assay. The removal of CD19+ B cells or CD14+ monocytes from the PBMC population did not reduce IFN-γ secretion in this system (data not shown). Next, experiments were performed in which NK cells were mixed with B cells, monocytes, or whole PBMCs at different ratios before their use in the assay (NK:APC ratios of 1:1, 1:2, and 1:4). Addition of these cell populations to NK cells did not enhance NK cell IFN-γ secretion in response to CpG ODN and immobilized IgG at any ratio examined (data not shown). Collectively, this evidence demonstrated that NK cells were able to respond directly to CpG ODN with the secretion of IFN-γ, and that this response was not enhanced in the presence of APCs or their cytokine products.
CpG ODN indirectly enhances NK-mediated cytotoxicity in vitro
The effect of CpG ODN on NK cell lytic activity was examined. Purified human NK cells were cultured overnight in medium supplemented with various concentrations of CpG ODN (1–50 μM). These cells were then used as effectors in a standard 4-h chromium release assay in which NK-sensitive K562 cells served as targets (20). In Ab-dependent cellular cytotoxicity (ADCC) assays, chromium-labeled BT-474 human breast cancer cells that had been coated with Herceptin served as targets (20). No enhancement of natural cytotoxicity or Herceptin-specific ADCC activity was observed for purified NK cells, even at the highest concentration of CpG ODN tested (Fig. 4,A). Based on previous reports in which CpG ODN had been shown to enhance the lytic activity of murine NK cells within a splenocyte population (21, 22), we isolated whole PBMCs from normal human donors, activated them overnight with increasing concentrations of CpG ODN (1–50 μM), and analyzed their cytotoxic activity as described above. Although CpG ODN did not enhance the lytic activity of purified NK cells, overnight treatment with CpG ODN greatly enhanced the natural cytotoxicity and ADCC activity of NK cells within a PBMC population. Lytic activity was maximal when PBMC were activated with 10 μM CpG ODN (Fig. 4,B). In addition, large amounts of IL-12, IFN-α, and TNF-α were detected within culture supernatants from CpG-activated PBMCs, but not purified NK cells (Fig. 4 C). These results suggested that CpG ODN treatment of whole PBMCs can enhance NK cell lytic activity indirectly, via the induction of APC-derived cytokines. Indeed, CD56dim NK cells isolated from whole PBMCs that had been treated overnight with 10 μM CpG ODN exhibited increased ADCC and natural cytotoxicity compared with NK cells derived from control-treated PBMCs, indicating that NK cells could be activated for lytic activity by these soluble factors.
Human NK cells express TLR9
The expression of TLR9 within human NK cells was assessed. Purified NK cells from 20 normal donors were stained for intracellular TLR9. Results from a representative donor are shown in Fig. 5,A. Resting peripheral blood CD19+ B cells were used as a positive control for TLR9 expression, because they are known to constitutively express this receptor (13, 14). On the average, 36 ± 3.5% (range, 15–83%) of primary human NK cells expressed basal levels of TLR9. TLR9 protein was detected in NK cells from all donors tested. The distribution of TLR9 expression in normal donor NK cells is shown in Fig. 5,B. Protein expression of TLR9 on resting NK cells and B cells was confirmed by immunoblot analysis (Fig. 5 C). Of note, cross-linking of FcγRIIIa with an anti-CD16 mAb (clone 3g8; Mederex) did not lead to increased TLR9 expression at 24 h, as measured by intracellular flow cytometry (data not shown). Overnight treatment with CpG ODN (1 μM) also failed to increase NK cell expression of TLR9 (data not shown).
Human NK cells can be divided into two functional subsets on the basis of CD56 surface density expression. CD56dim NK cells express high levels of CD16, but do not produce IFN-γ in large amounts. In contrast, CD56bright NK cells express low levels of CD16 (CD16dim or CD16neg), but are known to be potent producers of IFN-γ in response to monokine stimulation (23). We therefore compared the responses of CD56bright and CD56dim NK cells to the combination of CpG ODN and immobilized IgG. When NK cells were sorted on the basis of their CD56 expression and stimulated separately with immobilized IgG and CpG ODN, IFN-γ production was restricted to the CD56dim NK cell population (Fig. 6,A). When NK cells were costained for CD56 and TLR9, both populations of NK cells were found to express TLR9 (Fig. 6, B and C), indicating that the inability of CD56bright NK cells to secrete IFN-γ in response to IgG and CpG ODN was probably due to low CD16 expression rather than lack of TLR9 within these cells.
NK cells respond to CpG ODN via TLR9
To establish the involvement of TLR9 in mediating the NK cell response to CpG ODN, additional experiments were conducted using the TLR9−/− mouse (8). Figs. 7–9 represent murine studies. Splenic NK cells were isolated from TLR9−/− mice or age-matched wild-type control mice and costimulated with 1 μM CpG ODN and immobilized murine IgG. NK cells from wild-type mice secreted modest amounts of IFN-γ in response to CpG ODN alone, and this was enhanced >5-fold upon costimulation with immobilized IgG. In contrast, NK cells from the TLR9−/− mice were unable to respond to CpG alone or to costimulation with CpG ODN and muIgG (Fig. 7). When NK cells derived from TLR9−/− mice were costimulated with muIgG and IL-12 or IL-12 and IL-18, these cells secreted comparable amounts of IFN-γ as TLR9+/+ mice (data not shown) (20, 24), demonstrating that NK cells from TLR9-deficient animals were able to secrete IFN-γ in response to other stimuli, and that their lack of responsiveness to CpG DNA was TLR9 specific and was not due to a generalized deficit in IFN-γ production.
CpG ODN enhance in vivo IFN-γ production in response to Herceptin-coated tumor cells
We next performed an in vivo costimulation experiment in which CpG ODN and Herceptin-coated, human HER2-overexpressing murine tumor cells were delivered to naive mice via the i.p. route. Coadministration resulted in increased serum levels of IFN-γ, whereas mice receiving CpG ODN and a control Ab-treated tumor produced significantly lower amounts of this cytokine (Fig. 8,A). IFN-γ secretion was markedly inhibited when mice were depleted of NK cells before coadministration of Herceptin-coated tumor and CpG ODN (Fig. 8 B). To confirm the functional effect of NK depletion, the lytic activity of splenocytes from mock-treated and NK-depleted animals was compared in a classical 4-h chromium release assay against the NK-sensitive cell line, YAC-1 (American Type Culture Collection). No significant cytotoxic activity was observed using splenocytes from NK-depleted animals, confirming the efficacy of the NK depletion (data not shown). These results demonstrate that in vivo administration of CpG ODN can enhance the NK cell cytokine response to tumor cells that have been treated with a therapeutic mAb.
NK cells respond directly to CpG ODN in vivo
To confirm the direct action of CpG ODN on NK cells in vivo, we performed an adoptive transfer experiment in which TLR9−/− mice were reconstituted with NK cells from wild-type (i.e., TLR9+/+) mice before the coadministration of Ab-coated tumor and CpG ODN. Control groups consisted of TLR9+/+ animals receiving TLR9−/− NK cells, TLR9+/+ animals receiving TLR9+/+ NK cells, and TLR9−/− animals receiving TLR9−/− NK cells. As shown in Fig. 9, TLR9+/+ mice receiving TLR9−/− NK cells secreted large amounts of IFN-γ in response to the combination of Ab-coated tumor and CpG ODN. This IFN-γ secretion was slightly enhanced within TLR9+/+ mice that had received additional TLR9+/+ NK cells. TLR9−/− mice receiving TLR9−/− NK cells did not secrete IFN-γ in response to coadministration of Ab-coated tumor and CpG ODN. However, TLR9−/− mice that had received wild-type NK cells responded to the combination of Ab-coated tumor and CpG ODN with the secretion of IFN-γ. Because NK cells were the only TLR9+/+ cell type in these animals, these results demonstrate that NK cells can respond directly to CpG ODN with the secretion of IFN-γ.
We have demonstrated that human NK cells respond directly to synthetic ODN containing unmethylated CpG motifs by secreting IFN-γ and several chemokines. This cytokine response was significantly enhanced in the presence of Ab-coated targets, both artificial (immobilized polyclonal IgG) and with clinical relevance (mAb-coated human breast cancer cells). Coadministration of CpG ODN with mAb-coated tumor cells to mice resulted in enhanced circulating levels of IFN-γ. Studies using primary human NK cells as well as NK cells derived from TLR9-deficient mice confirmed the involvement of TLR9 in the NK cell cytokine response to CpG ODN. These data are the first to demonstrate the expression of TLR9 by human NK cells and its ability to enhance the secretion of cytokines by NK cells in response to Ab-coated targets.
In the current model for the mechanism of action of CpG ODN on immune cells, TNF-α, IL-12, or type I IFNs derived from CpG-stimulated B cells or dendritic cells are thought to secondarily activate NK cell IFN-γ production (25). In support of this model, several reports have shown indirect enhancement of NK cell cytokine secretion and lytic activity after exposure of splenocytes or PBMCs to CpG ODN (17, 22, 26, 27). In these studies, depletion of macrophages or the addition of neutralizing Abs to IL-12 and/or TNF-α to these cellular preparations abolished the NK cell cytokine response. Although PBMC-derived TNF-α and IL-12 may activate NK cells (27), the present data suggest that these stimuli are not absolutely necessary for the induction of NK cell cytokine secretion in response to CpG ODN. Indeed, synergistic IFN-γ secretion by purified NK cells costimulated with CpG ODN and immobilized IgG occurred in the presence of undetectable levels of IL-12 and IFN-α and equivalent levels of TNF-α among the different assay conditions. Furthermore, TLR9−/− animals reconstituted with TLR9+/+ NK cells had enhanced serum levels of IFN-γ after coadministration of Ab-coated tumor cells and CpG ODN. The present data confirm that direct stimulation of human and murine NK cells with CpG ODN leads to the secretion of modest amounts of IFN-γ and demonstrates that simultaneous engagement of TLR9 and FcγRIIIa on NK cells triggers synergistic release of this cytokine, independently of APC-derived factors.
It has been suggested that endogenous IFN-γ produced in response to FcγRIIIa ligation could sensitize NK cells to CpG ODN in an autocrine fashion (16). However, IFN-γ-dependent activation of STAT1 did not appear to be involved in the synergistic response of NK cells to costimulation via CD16 and TLR9, because the addition of an anti-IFN-γR mAb to the cultures did not inhibit NK cell cytokine production. Similarly, NK cells from STAT1-deficient mice exhibited the expected synergistic production of IFN-γ when costimulated with CpG ODN and immobilized IgG (data not shown). Our group continues to investigate the mechanism of synergy to ligands to CD16 and TLR9.
Other evidence has recently emerged indicating that NK cells can be activated directly by microbial products and that this can contribute to adaptive immune responses. For example, murine NK cells have been shown to recognize a mouse CMV protein, m157, via the activating receptor LY49H (28). In addition, Schmidt et al. (29) have recently reported that viral dsRNA directly activates human NK cells through TLR3, and Becker et al. (30) have shown that human NK cells directly recognize the cell surface lipophosphoglycan of the parasite Leishmania via TLR2. TLR2- and TLR3-stimulated human NK cells were found to secrete IFN-γ in both these reports. We show in this study that CpG DNA can stimulate NK cell cytokine secretion via TLR9. TLR9 protein expression was detected on resting human NK cells via flow cytometry and immunoblot analysis. Although NK cells expressed TLR9 at relatively low levels, and the percentage of NK cells expressing TLR9 varied between donors, TLR9 protein was detected in NK cells from all individuals tested. Hornung et al. (15) previously used real-time PCR to examine transcript levels of TLR1 through TLR10 within cellular subsets of human PBMCs. The authors did detect TLR9 mRNA within NK cells, although the level of transcript was extremely low compared with that of other cell subtypes (∼10 copies of TLR9 mRNA/1000 copies of a housekeeping gene in NK cells, compared with >100 copies in dendritic cells). Our studies with TLR9-deficient mice indicate that low level expression of TLR9 on NK cells is sufficient to permit activation by CpG motifs.
Direct induction of NK cell-derived IFN-γ by CpG DNA has been reported previously, although no mechanism of action was proposed. Klinman et al. (17) showed by ELISPOT analysis that purified murine NK cells secreted IFN-γ in response to in vitro stimulation with CpG ODN. Subsequently, Iho et al. (16) showed induction of IFN-γ when purified human NK cells were cultured with specific palindromic DNA sequences containing CpG dinucleotides. However, the overall levels of IFN-γ secretion in these studies were found to be quite low, suggesting that additional stimuli (e.g., FcR-dependent mechanisms) might be required for optimal NK cell cytokine production. Indeed, Iho et al. (16) described increased production of IFN-γ when NK cells were costimulated with an agonistic anti-CD16 Ab. Although both CD56dim and CD56bright NK cells were found to express TLR9, neither cell population secreted significant levels of IFN-γ in response to CpG ODN alone. CD56bright NK cells are known to be potent producers of IFN-γ in response to stimulation with monokines such as IL-12 and IL-18 compared with CD56dim NK cells (23). However, differential IFN-γ secretion by NK cell subsets in response to other stimuli, such as bacterial components, has not been well characterized. The finding that CD56dim NK cells (but not CD56bright NK cells) could be induced to secrete IFN-γ in response to CpG ODN though simultaneous triggering of FcγRIIIa and TLR9 indicates a potential mechanism by which the CD16-positive CD56dim NK cell subset can also be induced to secrete substantial levels of immunomodulatory cytokines. Thus, in the presence of CpG ODN, the response of the CD56dimCD16bright NK cell subset to Ab-coated targets may not be limited to cytotoxic activity.
Although simultaneous stimulation of purified NK cells via TLR9 and FcγRIIIa led to enhanced IFN-γ production, direct TLR9 ligation did not appear to augment the NK cell cytotoxic response to K562 cells or to Ab-coated tumor cells. In contrast, overnight treatment with CpG ODN greatly enhanced the natural cytotoxicity and ADCC activity of NK cells within a PBMC population. This finding is consistent with previous reports that treatment of whole splenocytes with CpG ODN enhances NK cell cytotoxicity (21, 22). In addition to IFN-γ, high levels of IL-12, TNF-α, and IFN-α were detected in culture supernatants of PBMCs pretreated with CpG ODN. These results suggested that CpG-mediated NK cell activation could occur via two separate pathways. As reported previously, CpG ODN can induce APCs to secrete type I cytokines that secondarily promote NK cell cytotoxicity. In addition, NK cells can be directly activated by CpG ODN through ligation of TLR9 located within the NK cell. Stimulation of NK cells in this fashion induces them to secrete IFN-γ and other cytokines, an effector function that is markedly increased in the presence of immobilized Ab. However, NK cytotoxic activity is unaffected by direct ligation of TLR9, and additional signals from CpG-activated APCs are apparently required for these cells to gain cytolytic function.
The secretion of IFN-γ and an array of chemotactic factors by NK cells costimulated with CpG ODN and IgG suggests a role for these effectors in the clearance of Ab-coated targets in vivo and may reflect a normal pathway by which NK cells can aid in the control of microbial infections. The observation that NK cells are able to respond to a variety of ODN sequences supports this contention. The administration of CpG ODN in conjunction with therapeutic mAbs could theoretically enhance the innate immune response to Ab-coated tumor cells via the induction of NK cell-derived cytokines. The observed pattern of NK cell cytokine secretion would probably have effects on tumor cell proliferation and Ag presentation (IFN-γ) (31), T cell activation and proliferation (MIP-1α) (32, 33), and T cell chemotaxis (MIP-1α, IL-8, and MDC) (34). Notably, van Ojik et al. (35) have recently demonstrated the importance of NK cells in mediating tumor regression in mice bearing experimental HER2-overexpressing tumors after treatment with CpG ODN and an anti-HER2 mAb. Depletion of NK cells significantly decreased the survival benefit afforded by combination therapy, suggesting a central role for this cell type in the antitumor response. Additional studies in murine models of breast cancer will help to define the role of NK cell-derived cytokines in mediating the antitumor effects of CpG ODN.
In conclusion, these data demonstrate that CpG DNA can enhance the NK cell cytokine response to Ab-coated targets. NK cells produce large amounts of IFN-γ and an array of chemokines in response to costimulation with CpG ODN and immobilized Ab. This unique cytokine profile is critically dependent upon NK cell expression of TLR9 and may reflect a normal pathway by which bacterial DNA can activate innate immune effectors in the context of microbial infections. Furthermore, these data provide a rationale for the administration of CpG ODN in combination with therapeutic mAbs for the treatment of malignant disease.
We thank Dr. Susheela Tridandapani (Ohio State University, Columbus, OH) for insightful discussions.
The authors have no financial conflict of interest.
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.
This work was supported by National Institutes of Health Grants P30CA16058, CA86016, and P01CA95426.
Abbreviations used in this paper: ODN, oligodeoxynucleotide; ADCC, Ab-dependent cellular cytotoxicity; CpG ODN, CpG-containing ODN; HAB, human AB serum; huIgG, human IgG; MDC, macrophage-derived chemokine; muIgG, murine IgG.