Cutting Edge: The Mechanism of Invariant NKT Cell Responses to Viral Danger Signals¹ Free

Invariant NK T (iNKT)⁶ cells are a subpopulation of innate-like T lymphocytes that express an invariant T cell Ag receptor α-chain. iNKT cells rapidly mount a response to synthetic glycolipid Ags presented by CD1d that is characterized by the production of large amounts of IFN-γ and IL-4 (1, 2). Additionally, iNKT cells respond to several bacteria that have glycolipids that engage their invariant TCRs. iNKT cells probably also can be activated by microbes that do not encode Ags for their TCRs (3). For example, iNKT cells have been reported to influence the course of some viral infections. In fact, several viruses down-regulate human CD1d expression, suggesting a key role for iNKT cells in antiviral defenses (4, 5).

It is well established that stimulation of APCs with bacterial products or the TLR ligand LPS activate iNKT cells in the absence of foreign Ag for their TCR (6, 7). Until recently, however, it was not known whether other TLR ligands, such as those that detect viral danger signals, also activate iNKT cells. Therefore, in this study we investigated the response of iNKT cells to a TLR9 ligand, CpG oligodeoxynucleotide (ODN), and the response to virus infection with mouse CMV (MCMV) (8). MCMV is a β-herpesvirus that is a widely studied model with viral control dependent on innate immune responses by stromal cells, dendritic cells (DCs), and NK cells (9, 10). Despite studies assessing the potential contribution of iNKT cells in innate control of MCMV (11), direct analysis of iNKT activation, and the cytokine requirements for this activation, has not been performed. In the present study we show that the mechanism of iNKT cell activation by MCMV is similar to that of in vitro activation by CpG ODN-stimulated APCs.

C57BL/6 (B6), B6.129S1-Il12b^tm1Jm/J (IL-12p40^−/−), and B6.129P2-Il18^tm1Aki/J (IL-18^−/−) mice were purchased from The Jackson Laboratory. CD1d^−/− mice were a gift from Dr. L. Van Kaer (Vanderbilt University, Nashville TN). Tlr9^cpg1 mice were a gift from Dr. B. Beutler (The Scripps Research Institute, La Jolla CA) (8). All mice were housed in specific pathogen-free conditions.

Liver lymphocytes were isolated as described previously (12). For preparation of DCs, bone marrow was cultured in medium supplemented with 100 ng/ml recombinant human fms-like tyrosine kinase-3 ligand (hFlt3L; Amgen) for 8 days. DCs were treated with 10 μg/ml type B CpG ODN 1826 or control 1982 ODN (Alexis Biochemicals), and infected with MCMV Smith strain at a multiplicity of infection of 3, or mock infected, for 2 h. DCs were cultured with iNKT cells for 48 h before cytokine detection (7). iNKT cells were obtained from spleen by selecting for NK1.1⁺ TCRβ⁺ cells by flow cytometry as previously described (7). The purity of the cells collected, typically >98%, was assessed using CD1d tetramers loaded with α-galactosylceramide (αGalCer; provided by Kirin Pharma).

A standard sandwich ELISA was performed to measure mouse IFN-γ, TNF, and IL-4 following the manufacturer’s instructions (R&D Systems).

Salivary gland extract stocks of Smith strain MCMV were prepared as previously described (13). Mice were infected with 5 × 10⁴ PFU of virus in 500 μl of PBS via i.p. injection.

To block CD1d in vitro, 30 μg of anti-mouse CD1d clone 1B1 mAb was added to cultures. For blocking in vivo, mice received 300 μg of mAb on day −1, 200 μg on day 0, and 100 μg on day 1 of 1B1 mAb or rat IgG controls (eBiosciences) in PBS via i.p. injection (14).

Lymphocytes were stained with αGalCer/CD1d tetramers labeled with streptavidin-allophycocyanin, anti-NK1.1-PerCp PE-cyanin (PECy)5, anti-CD8-PECy7, anti-CD11b-PECy7, anti-TCRβ-allophycocyanin-AF750, and CD25-FITC. All Abs and isotype controls, except anti-TCRβ-allophycocyanin-AF750 from eBioscience, were purchased from BD Biosciences. As a positive control for iNKT cell responses, mice were injected with 2 μg of αGalCer. Cells were fixed and permeabilized using Cytofix/Cytoperm buffer and stained for intracellular IFN-γ with PE-labeled clone XMG1.2. The data were collected on a LSR II flow cytometer (BD Biosciences) and analyzed using FlowJo software (Tree Star).

Previous studies have demonstrated that mouse iNKT cells were activated following stimulation of DCs with the TLR4 ligand LPS (7). To further investigate the potential role of iNKT cells in regulating the immune response to viral danger signals, we evaluated the ability of TLR9-stimulated or MCMV-infected DCs to activate iNKT cells in vitro.

Because triggering the TLR9 pathway with CpG-ODN serves as a model for innate recognition of viral DNA, we tested the ability of DCs stimulated with CpG ODN to activate iNKT cells in culture. DCs derived from bone marrow by culture with hFlt3L were exposed to CpG ODNs and cocultured with highly purified splenic iNKT cells. High levels of IFN-γ, but no IL-4 or TNF, were detected in the stimulated cocultures (Fig. 1,A and data not shown). Similar results were obtained when the DCs were stimulated with TLR7 agonists (data not shown). The production of IFN-γ in the absence of IL-4 was similar to previous results obtained from analyzing iNKT cell responses to LPS (7), but it contrasts with the results obtained with iNKT cell activation by TCR agonists, which elicits all three cytokines. IFN-γ secretion by iNKT cells was dependent on TLR9 activation of the DCs, as nonstimulated DCs or DCs stimulated with an inactive CpG ODN did not induce it. Additionally, hFlt3L-derived DCs from TLR9^CpG1 mutant mice were unable to induce IFN-γ secretion by iNKT cells (Fig. 1 B). IFN-γ production in these cultures was dependent on iNKT cells, although TLR9-stimulated DCs in the absence of iNKT cells produced IL-12p70 (data not shown).

To define the mechanism of iNKT cell activation by CpG ODN-treated DCs, DCs were generated from mice deficient for IL-12p40, IL-18, or CD1d. Similar to our previous results with LPS-stimulated DCs, iNKT cells did not produce IFN-γ when the TLR9-stimulated DCs were made from IL-12p40^−/− mice. Somewhat surprisingly, IL-18^−/− DCs treated with CpG ODN did induce IFN-γ production by iNKT cells, contrasting with the results from LPS stimulation (7). DCs derived from CD1d-deficient mice were equally effective at causing IFN-γ secretion, and similar results were obtained when a CD1d-blocking mAb was added to cocultures containing wild-type (WT) DCs or CD1d^−/− DCs to block the contribution of CD1d expression by the iNKT cells themselves. These results indicate that iNKT cells can be activated in response to IL-12 induced by endosomal TLR ligands independently of either foreign or self-lipid Ag presentation by CD1d.

Our results are consistent with those from several recent reports showing iNKT cell activation following exposure to TLR9 agonists. In one of these, however, activation in vivo did not occur unless the CpG ODN was encapsulated in liposomes (15). In studies using human iNKT cells, activation marker expression was increased but cytokine release was not observed (16), whereas in another study TLR3, 7, and 9 agonists were shown to cause the in vitro expansion of human iNKT cells in cultures of PBMCs (17). Furthermore, a recent report demonstrated that liver-derived mouse iNKT cells were activated by DCs exposed to TLR7 or TLR9 agonists. However, in that study iNKT cell activation was dependent upon type I IFN secretion, not IL-12, despite the production of both of these cytokines by the activated DCs (18). Furthermore, iNKT cell IFN-γ secretion was found to be at least partially dependent upon CD1d expression and the presentation by CD1d of a charged glycosphingolipid autologous Ag of undefined structure (18). The reasons for this discrepancy are not known but could be due either to differences in DCs generation or to the use of liver mononuclear cells as opposed to purified splenic iNKT cells. In our studies, hFlt3L cultured DCs were used whereas the previous study used 14-day cultured GM-CSF DCs. As described below, hFlt3L-cultured DCs may more appropriately model in vivo infection with virus.

The engagement of endosomally expressed TLR9 by CpG ODN in DCs is a model for some of the early events leading to innate immune activation following infection with a DNA virus. To test this, we evaluated the in vitro iNKT cell response to DCs infected with MCMV. Infected hFlt3L DCs were cocultured with purified iNKT cells, and the supernatants were measured for cytokine production (Fig. 2, A and B). MCMV infection induced rapid production of IFN-γ but not IL-4 (Fig. 2,A). IFN-γ secretion required MCMV infection, as uninfected DCs (Fig. 2,A) or infection with a mouse-adapted strain of Dengue virus (19) (data not shown) did not lead to IFN-γ production. MCMV-infected DCs from TLR9^CpG1 mutant mice or from IL-12p40^−/− mice also did not induce IFN-γ production (Fig. 2 B). By contrast, MCMV-infected, GM-CSF-cultured DCs did not activate iNKT cells (data not shown).

MCMV has been shown to provoke innate immune responses that lead to IL-12 secretion and NK cell activation (9, 20). Our data suggest that the recognition of MCMV by TLR9 induces IL-12 production by DCs, leading to IFN-γ production by iNKT cells. Unlike in vitro stimulation with CpG-ODN, however, optimal production of IFN-γ by cocultures with iNKT cells was dependent on the ability of the DCs to produce IL-18 (Fig. 2,B). In no case did the absence of CD1d expression reduce the amount of IFN-γ produced by iNKT cells to the levels observed with DCs deficient for TLR9 or IL-12. In the experiment shown, IFN-γ release was reduced by ∼30% when WT DCs were treated with an anti-CD1d mAb or when mAb blockade was combined with the use of CD1d^−/− DCs (Fig. 2 B). Comparing multiple experiments, the reduction in IFN-γ when CD1d was blocked or deleted, although variable, was generally minor and in several experiments there was little or no effect of inhibiting the interaction with CD1d (data not shown).

Upon Ag recognition or exposure to the combination of IL-12 and IL-18, iNKT cells rapidly become activated in vivo (21). As early as 12 h postinfection with MCMV, tetramer⁺ iNKT cells showed increased surface expression of CD25 (Fig. 3,A). The increased expression peaked by 12–18 h and rapidly declined by 36 h in both the liver and spleen (Fig. 3,B). A similar temporal expression profile was observed when CD69 expression was analyzed (data not shown). This early activation of iNKT cells parallels very closely the induction kinetics of initial innate cytokine production by stromal cells upon MCMV infection (13). Liver and spleen iNKT cells produced IFN-γ 36 h postinfection when analyzed directly ex vivo without restimulation (Fig. 3,C and data not shown). No production of IL-4 or TNF was observed. A previous study demonstrated that a TCR⁺NK1.1⁺ population produced IFN-γ at this time point (8), but in the absence of tetramer staining these data could not unequivocally show the activation of iNKT cells. A time course analysis revealed that iNKT cells were capable of producing IFN-γ as early as 24 h postinfection (Fig. 3 D), a time preceding the viral spread within infected organs. Therefore, similar to results in vitro, the activation of iNKT cells post-MCMV infection in vivo leads to early CD25 and CD69 up-regulation, as well as IFN-γ production somewhat later.

Consistent with in vitro results, MCMV infection of TLR9^CpG1 mice did not result in a marked activation of iNKT cells to increase CD25 expression at 18 h or to secrete IFN-γ at 36 h after infection (Fig. 4). Also similar to the in vitro results, IL-12p40^−/− (Fig. 4) or IL-12Rβ2^−/− mice (data not shown) did not induce up-regulation of CD25 in iNKT cells shortly after infection (18 h) or IFN-γ production at 36 h. IL-18^−/− mice also had reduced CD25 expression, but IFN-γ production was not reduced, suggesting that in vivo IL-18 may be dispensable for the induction of iNKT cell effector function in response to MCMV.

To determine whether CD1d was required in vivo for viral activation of iNKT cells, WT mice were treated with a CD1d-blocking mAb and infected with MCMV. As a control for the efficacy of blocking, the mAb was administered to WT mice that received the potent agonist αGalCer, and IFN-γ synthesis by liver and splenic iNKT cells was measured directly ex vivo 2 h later. Anti-CD1d-treated mice showed a 50% reduction in iNKT cell IFN-γ production (Fig. 4 E). Treatment with the CD1d-blocking mAb did not alter the induction of iNKT cell CD25 expression following MCMV infection, however, and although the treated mice trended lower for IFN-γ production, this decrease was not statistically significant (p = 0.063). Self-Ags presented by CD1d during viral infection are likely to be much less potent than αGalCer, and thus the response to these Ags should be more effectively inhibited. Therefore, we conclude that CD1d-mediated Ag presentation is not likely to be a key player in MCMV-induced activation of iNKT cells in vivo.

Although iNKT cells are early responders to LPS, they are not the first responders because they depend on IL-12 and IL-18 produced by other cell types. Likewise, the kinetics of the iNKT cell response indicates that these cells also are not the first responders to MCMV infection. Previous work has shown that the initial IFN-αβ response to i.p. infection with MCMV peaks at ∼8 h postinfection in the spleen and is dependent upon B cells expressing lymphotoxin that cross-talk with type I IFN-producing stromal cells (13). A second wave of TLR-dependent activation occurs at ∼36 h postinfection in the spleen, shortly after MCMV completes its first round of replication (13, 22). Given the time course of increased activation marker expression and IFN-γ production by iNKT cells, we propose a two-stage activation process, with the initial response to viral danger signals being sufficient to induce increased surface marker expression by iNKT cells. However, a second TLR-dependent wave of activation and cytokines may be necessary to induce the relatively high levels of IFN-γ secretion needed to detect cytokine production by iNKT cells directly ex vivo. This hypothesis is consistent with data demonstrating that iNKT cells express slightly increased CD69 and CD25 expression in TLR9^CpG1 mice 12 h postinfection, but this was reduced by 18 h, and we were unable to detect iNKT cell IFN-γ production at later times (Fig. 4 and data not shown).

There are multiple pathways for the foreign Ag-independent activation of iNKT cells, which requires, to varying extents, IL-12, IL-18, and type I IFN, as well as CD1d presentation of self-Ags. We speculate that the pathway used will depend on the nature of the inciting stimulus and the type(s) of APCs. Our findings indicate, however, that the mechanism of iNKT cell activation following MCMV infection in vivo is similar to that we observed with TLR9 ligands in vitro or with virus infection in vitro using Flt3L DCs. The data suggest that the iNKT cell response to viral danger signals in vivo from MCMV infection is mediated by TLR9 engagement leading to IL-12 production and that this response is not highly dependent upon CD1d Ag presentation. Clearly, this pathway, along with the others described that do not depend on foreign Ag, endow innate-like iNKT cells with the ability to respond to diverse pathogens despite a highly constricted TCR repertoire.

We acknowledge A. Khurana for assistance with the generation of CD1d tetramers. We acknowledge Dr. Sujan Shresta for help with experiments and advice. This is publication no. 1033 from La Jolla Institute for Allergy and Immunology.

The authors have no financial conflict of interest.