Innate and cytokine-driven signals, rather than microbial antigens, dominate in natural killer T cell activation during microbial infection

Studies using only NKT cell hybridomas do not adequately model the NKT cell activation mechanism that may occur in vivo because such systems lack the potential to respond to both antigen and cytokine signals. To investigate the mechanisms of iNKT cell activation by microbes, we used a system with primary mouse iNKT cell lines and BM-derived DCs that is able to respond to a variety of stimuli (Chiba et al., 2009). iNKT cell lines incubated with DCs and stimulated with the CD1d-presented microbial GSL antigen GSL-1, which is found in Sphingomonas spp. (Kinjo et al., 2005; Mattner et al., 2005; Sriram et al., 2005), produced large amounts of IFN-γ (Fig. 1 A, left). This IFN-γ response was dependent on recognition of CD1d by iNKT cell lines, as use of CD1d-deficient DCs resulted in markedly reduced IFN-γ secretion (Fig. 1 A). To determine the requirement for TLR- and cytokine-mediated stimulation for the activation of iNKT cell lines, we performed experiments using DCs deficient in the adaptor protein MyD88 or the production of IL-12, as well as blocking antibodies against IL-12. Stimulation of iNKT cell lines with GSL-1 was not altered when DCs deficient in MyD88 were used (Fig. 1 B) and was only marginally reduced when IL-12–deficient DCs were used or when mAbs to IL-12 were added to the cultures (Fig. 1 C). Thus, iNKT cell activation by a microbial lipid antigen required CD1d expression and was essentially independent of TLR signaling or IL-12 production by APCs. The slightly reduced iNKT cell IFN-γ response to GSL-1 in the presence of IL-12–deficient DCs or after addition of mAb against IL-12 was likely the result of a lack of IL-12–mediated amplification of iNKT cell IFN-γ production that has been noted after CD40L-mediated stimulation of DCs by antigen-activated iNKT cells (Tomura et al., 1999).

In contrast to their direct stimulation by CD1d-presented microbial lipid antigens, iNKT cells can also be stimulated with TLR agonists in the presence of DCs by a self-antigen– and cytokine-driven pathway that does not require the cognate recognition of microbial antigens by iNKT cells (Brigl et al., 2003; Nagarajan and Kronenberg, 2007; Paget et al., 2007; Salio et al., 2007). Indeed, iNKT cell lines incubated with WT DCs and stimulated with the TLR agonists LPS (TLR4) or CpG (TLR9) produced copious amounts of IFN-γ (Fig. 1 A, second and third panel). To determine the requirements for a CD1d–TCR interaction, TLR signaling, and IL-12 stimulation after activation of iNKT cells with TLR agonists, we used DCs from CD1d-, MyD88-, or IL-12–deficient mice and blocking antibodies against IL-12. Using CD1d-deficient DCs, TLR agonist–induced IFN-γ production by iNKT cell lines was significantly reduced in response to LPS or CpG, respectively, compared with stimulation in the presence of WT DCs (Fig. 1 A). Because no exogenous microbial antigens were present under these conditions, recognition of CD1d-presented cellular self-antigens appeared to be required for iNKT cell activation, as has been previously noted (Brigl et al., 2003). When MyD88-deficient DCs were used during stimulation of iNKT cell lines, IFN-γ production in response to LPS or CpG was markedly reduced (Fig. 1 B). Furthermore, iNKT cell activation in response to LPS or CpG was strikingly reduced when DCs deficient in IL-12 production were used and when mAb to IL-12 was added to the cultures (Fig. 1 C). In response to antigens, iNKT cells are also known to secrete IL-4. After stimulation with GSL-1, iNKT cell lines produced IL-4 in a CD1d-dependent and MyD88- and IL-12–independent manner (Fig. 1, D and E). In response to LPS or CpG, no significant amounts of IL-4 were detected (Fig. 1, D and E). Thus, TLR agonist–induced IFN-γ secretion by iNKT cell lines in the presence of DCs was driven by recognition of CD1d-presented self-antigens and IL-12, which are produced by APCs after TLR-mediated activation, whereas antigen-driven activation results in TLR- and IL-12–independent IFN-γ and IL-4 secretion.

To examine the mechanism of iNKT cell activation in response to bacterial infection, we assembled a panel of 11 bacteria, 5 of which are known to express CD1d-presented iNKT cell antigens (Table I). The diverse panel of bacteria was selected to include clinically important pathogens such as the Gram-positive bacteria S. pneumoniae, Listeria monocytogenes and Staphylococcus aureus, the Gram-negative bacteria Pseudomonas aeruginosa, S. typhimurium, and E. coli, the spirochete B. burgdorferi, the mycobacterium Mycobacterium tuberculosis, and the α-proteobacteria Sphingomonas capsulata, Novosphingobium aromaticivorans, and Sphingomonas yanoikuyae. Members of the class of α-proteobacteria are known to cause opportunistic infections in humans and have been associated with the induction of autoimmunity (Mohammed and Mattner, 2009; Ryan and Adley, 2010). Furthermore, iNKT cells are critical in mice for protective immunity to infection with S. pneumoniae, B. burgdorferi, P. aeruginosa, S. capsulata, and L. monocytogenes, and iNKT cells have been implicated in the development of primary biliary cirrhosis in a mouse model after infection with N. aromaticivorans (see Table I for summary and references).

From among the panel of bacteria, microbial CD1d-presented lipid antigens that are capable of stimulating iNKT cells have been described for S. capsulata, N. aromaticivorans, and S. yanoikuyae (all three GSL-1), B. burgdorferi (BbGL-II), and S. pneumoniae (monoglycosyl- and diglycosyldiacylglycerol; for details and references see Table I). To confirm that bacteria used in this study indeed expressed iNKT cell antigens, we isolated polar lipids from S. capsulata, N. aromaticivorans, S. yanoikuyae, B. burgdorferi, and S. pneumoniae and subjected the lipids to electrospray mass spectrometry analysis. Lipids from S. capsulata, N. aromaticivorans, and S. yanoikuyae yielded ions representing the [M – H]⁻ ions of the antigenic α-glucuronosylceramide GSL-1 (Fig. 2, A–C; and Table S1). For B. burgdorferi lipids, ions corresponding to the [M + CH3COO]⁻ adduct ions of α-galactosyldiacylglycerol, the BbGL-II antigen, were detected (Fig. 2 D and Table S1). In positive-ion mode, the lipids from S. pneumoniae gave rise to [M + Na]⁺ ions corresponding to the α-glucosyldiacylglycerol (GlcDAG), together with ions corresponding to the α–galactosyl GlcDAG (GalGlcDAG; Fig. 2 E and Table S1). The structural assignments for individual lipid species were supported by multiple-stage ion-trap tandem mass spectrometry (Table S1 and not depicted). Further analysis of the bacterial lipids by liquid chromatography, light scattering detection, and mass spectrometry analysis confirmed the presence of the respective antigenic lipid species in these five bacteria and suggested that all bacteria expressed significant amounts of the respective antigens (Fig. S1). Thus, 5 of the 11 bacteria used in this study expressed known iNKT cell lipid antigen species as expected. No lipids capable of stimulating iNKT cells directly have so far been described for the other 6 bacteria.

We next examined the mechanism of iNKT cell activation in response to the 11 bacteria described in the previous section. iNKT cell lines incubated with WT DCs and stimulated with any of the 11 heat-killed bacteria produced substantial amounts of IFN-γ (Fig. 3 A). iNKT cell activation induced by heat-killed bacteria required a CD1d-TCR interaction, as IFN-γ production by iNKT cell lines in the presence of CD1d-deficient DCs and any of the 11 heat-killed bacteria was reduced by a mean of 50–80%, compared with stimulation in the presence of WT DCs (Fig. 3 A). Little or no IL-4 was generated by iNKT cells after stimulation with most of the heat-killed bacteria, whereas stimulation with B. burgdorferi, S. capsulata, N. aromaticivorans, and S. yanoikuyae resulted in production of low amounts of IL-4 (Fig. 3 B and not depicted). The amounts of IL-4 detectable in culture supernatants varied substantially between experiments and were generally reduced after repeated stimulation of iNKT cell lines, and no significant amounts of IL-4 were detected with iNKT cells isolated directly from Vα14 TCR transgenic mice (unpublished data).

Next, we determined if iNKT cell activation by heat-killed bacteria required TLR-mediated signaling by the APCs. iNKT cell lines incubated with MyD88-deficient DCs plus any of the 11 heat-killed bacteria resulted in reduced IFN-γ secretion when compared with WT DCs (Fig. 4 A and Fig. S2). Thus, whether the bacteria used for stimulation expressed known microbial lipid antigens, iNKT cell IFN-γ secretion was critically dependent on TLR-mediated signaling by DCs. This result suggested that iNKT cell activation in response to bacteria expressing microbial iNKT cell antigens might not be mediated by TCR recognition of the microbial lipid antigens and instead might be cytokine-driven.

IL-12, IL-18, and type I IFNs have been implicated in activating iNKT cells after TLR-mediated stimulation of DCs in the absence of iNKT cell recognition of CD1d-presented microbial lipid antigens (Brigl et al., 2003; Mattner et al., 2005; Nagarajan and Kronenberg, 2007; Paget et al., 2007; Salio et al., 2007). We determined if any of these cytokines were generated by DCs after stimulation with the panel of heat-killed bacteria by measuring the concentration of IL-12, IL-18, and IFN-β in supernatants from DCs cultured with TLR agonist, GSL-1, or any of the 11 heat-killed bacteria. Significant quantities of IL-12p40 and IL-12p70, but not of IL-18 or IFN-β, were detected in supernatants of DCs exposed to LPS, CpG, or any of the 11 heat-killed bacteria (Fig. 4 B and not depicted). In contrast, no IL-12p40 or IL-12p70 was detected in supernatants of DCs cultured in the presence of the GSL-1 antigen. Then, we determined whether IL-12 was required for IFN-γ secretion by iNKT cell lines in response to stimulation with heat-killed bacteria. Indeed, incubation of iNKT cell lines with IL-12–deficient DCs in the presence of any of the 11 heat-killed bacteria resulted in substantially reduced IFN-γ production compared with stimulation in the presence of WT DCs (Fig. 4 C and Fig. S2). Similarly, addition of blocking mAb against IL-12 reduced iNKT cell IFN-γ secretion after stimulation with any of the 11 heat-killed bacteria (Fig. 4 C). In contrast, stimulation of iNKT cell lines with IL-18–deficient DCs or after stimulation of iNKT cell/DC co-cultures in the presence of antibodies blocking type I IFN signaling did not show a requirement for IL-18 or type I IFNs, respectively, in response to any of the 11 heat-killed bacteria (Fig. S3, A and B). As described, MyD88- and IL-12–deficient DCs were capable of presenting the purified GSL-1 antigen to stimulate iNKT cell lines but were not capable of eliciting IFN-γ secretion by iNKT cell lines in response to the TLR agonists LPS and CpG (Fig. 1). We next determined if expression of the early activation marker CD25 on iNKT cells after stimulation was dependent on IL-12. Expression levels of CD25 were increased after stimulation with GSL-1, TLR agonists, or heat-killed bacteria (Fig. 4 D and Fig. S4). Stimulation with TLR agonists or heat-killed bacteria, but not with GSL-1, resulted in decreased expression on CD25 when DCs deficient in IL-12 were used. Similar results were obtained for the early activation marker CD69 (unpublished data). iNKT cell IFN-γ production and CD25 expression were also both reduced on day 2 after stimulation with antigen, TLR agonists, or heat-killed bacteria, suggesting that iNKT cell activation in the absence of IL-12 was not delayed (Fig. S5, A and B). The weak IL-4 responses observed after stimulation with Borrelia and Sphingomonas spp. showed a trend toward being reduced when MyD88-deficient DCs were used and toward being increased when IL-12–deficient DCs were used (Fig. 4 E). iNKT cells freshly isolated from WT mice by tetramer sorting also showed a strong dependence of the iNKT cell IFN-γ production on MyD88 and IL-12, whereas little or no IL-4 was generated (Fig. S6, A and B). The CD1d-dependence of TLR agonist- and bacteria-induced activation of freshly isolated iNKT cell was reduced compared with what was observed using iNKT cell lines, which is likely the result of stimulation of iNKT cells using TCR-β antibodies and tetramer for their purification by cell sorting, as has been observed previously (Matsuda et al., 2003; Nagarajan and Kronenberg, 2007).

To determine if alterations in CD1d expression levels by the APCs may contribute to the observed iNKT cell activation, CD1d expression levels were determined on WT DCs after exposure to heat-killed bacteria. Stimulation with the TLR agonists LPS and CpG and all 11 of the heat-killed bacteria, but not GSL-1, resulted in increased expression of CD1d by DCs (Fig. S7, A and B). The increased expression of CD1d after stimulation was absent in the presence of MyD88-deficient DCs, suggesting that TLR-mediated signaling was required. Thus, iNKT cell activation by a wide variety of heat-killed bacterial pathogens in vitro, including bacteria which expressed CD1d-presented microbial lipid antigens, such as Sphingomonas spp., B. burgdorferi, M. tuberculosis, or S. pneumoniae, was dominantly dependent on TLR signaling and IL-12–driven stimulation.

To further investigate the mechanism of iNKT cell activation in response to DCs infected with live bacteria, we used an in vitro live infection model with three microorganisms shown to express iNKT cell antigens (B. burgdorferi, S. capsulata, and N. aromaticivorans; Table I) and three microorganisms for which iNKT cell antigens have not been described (P. aeruginosa, S. typhimurium, and L. monocytogenes). DCs were infected with bacteria for 3 h and infected DCs were then co-cultured with iNKT cell lines, resulting in production of substantial amounts of IFN-γ for all six bacteria tested (Fig. 5 A). To determine the relative requirements for a CD1d–TCR interaction, TLR signaling, and IL-12 stimulation after activation of iNKT cells with infected DCs, we used DCs from CD1d-, MyD88-, or IL-12–deficient mice. Infection of DCs deficient in CD1d, MyD88, or IL-12 production with P. aeruginosa, S. typhimurium, L. monocytogenes, B. burgdorferi, S. capsulata, or N. aromaticivorans resulted in reduced IFN-γ production by iNKT cell lines, compared with infection of WT DCs (Fig. 5 A). Live in vitro infection with B. burgdorferi, and to a lesser extent S. capsulata, resulted in the production of IL-4 (Fig. 5 B). This IL-4 production was largely MyD88 dependent but IL-12 independent. Thus, similar to the results obtained after stimulation with heat-killed bacteria, iNKT cell IFN-γ production in response to DCs infected with live bacteria was dependent on CD1d recognition, TLR-signaling, and IL-12 stimulation, irrespective of the expression of iNKT cell antigens by the bacteria. This suggested that iNKT cell activation in response to infected DCs was dominantly cytokine driven and that TLR-mediated signals were required, whereas CD1d-mediated signals played a variable role.

Next, we analyzed iNKT cell activation in a mouse model of Sphingomonas infection. Mononuclear cells were isolated from WT mice 18 h after intravenous infection with S. yanoikuyae and iNKT cells were identified using CD1d tetramers. CD1d tetramer–positive iNKT cells from livers of uninfected mice stained at intermediate intensity for the early activation marker CD69 (mean fluorescence intensity [MFI], 465 ± 13) and, 18 h after infection, expression of CD69 increased to an MFI of 2,990 ± 639 (Fig. 6 A). IFN-γ secretion by iNKT cells showed that a mean of 41 ± 14% of the CD1d tetramer–positive iNKT cells were positive 18 h after infection, compared with a mean of 2 ± 0.1% in uninfected controls (Fig. 6 B). Thus, iNKT cells became rapidly activated after S. yanoikuyae infection. To assess whether IL-12 was required for the in vivo activation of iNKT cells during S. yanoikuyae infection, as suggested by our in vitro experiments, we compared IFN-γ secretion and CD69 expression by liver iNKT cells in WT versus IL-12–deficient mice after infection. 18 h after infection, IL-12 deficiency resulted in reduced IFN-γ secretion and CD69 expression (Fig. 6, C and D).

To compare the requirement for IL-12 in iNKT cell IFN-γ secretion after infection with S. yanoikuyae bacteria that express the GSL-1 antigen with iNKT cell activation induced by the GSL-1 antigen, we analyzed iNKT cell activation after stimulation with purified GSL-1 antigen. After injection of a range of doses of GSL-1 antigen, IFN-γ secretion by CD1d tetramer–positive liver iNKT cells was similar in WT or IL-12–deficient mice (Fig. 6 E). IL-4 production by iNKT cell in vivo was readily observed after injection of the α-galactosylceramide (GalCer) antigen; however, during infection with S. yanoikuyae, S. capsulata, or N. aromaticivorans no significant amounts of IL-4 were detected in either WT or IL-12–deficient mice (Fig. 6, F and G; and not depicted). Thus, iNKT cell activation after infection with a GSL-1–expressing microbe is dependent on IL-12, whereas iNKT cell activation with a microbial antigen alone does not require IL-12.

Next, we examined the role of iNKT cells after infection with S. pneumoniae. Infection of WT and iNKT cell–deficient Jα18^−/− mice with S. pneumoniae showed reduced survival of iNKT cell–deficient Jα18^−/− mice (Fig. 7 A). Between days 6 and 12 after infection, >50% of Jα18^−/− mice succumbed to infection, whereas none of the WT animals died. Similar results were obtained in CD1d-deficient animals (Fig. S8 A). Bacterial burden in lungs of infected WT or Jα18^−/− animals were comparable on day 3 after infection. However, on day 6 after infection, bacterial burden in lungs of Jα18^−/− were ∼1,000-fold higher when compared with WT animals (Fig. 7 B). Thus, iNKT cells were critically important for protective immunity after pulmonary infection with S. pneumoniae.

To examine iNKT cell activation after pulmonary infection with S. pneumoniae, lymphocytes were isolated from lungs of infected and uninfected animals and iNKT cells were analyzed by flow cytometry for surface expression of the activation marker CD69 and for secretion of IFN-γ. Expression of CD69 was intermediate on iNKT cells from uninfected mice and increased on days 2, 4, and 7 after infection (Fig. 7 C). Thus, iNKT cells became activated rapidly after S. pneumoniae infection. Cytokine secretion assays showed significant IFN-γ secretion by iNKT cells on days 2 and 4 after infection that returned to baseline by day 7 after infection (Fig. 7 D). No significant production of IL-4 by iNKT cells was observed in either WT or IL-12–deficient mice during infection at any of the time points examined (Fig. 7 D and not depicted). Thus, iNKT cells became activated rapidly and expressed IFN-γ early during pulmonary infection with S. pneumoniae.

Next, we determined whether IL-12 was required for the activation of iNKT cells during S. pneumoniae infection, as was the case in the in vitro experiments described in the previous paragraphs. IL-12 deficiency resulted in reduced IFN-γ secretion and CD69 expression on days 2 and 4 after infection (Fig. 7, E and F; and Fig. S8, B and C). Thus, most of the iNKT cell activation and IFN-γ production during S. pneumoniae infection in vivo was critically dependent on IL-12.

Responsiveness to IL-12 correlates with expression of IL-12 receptor (IL-12R) on NK and activated T cells (Presky et al., 1996; Trinchieri, 2003). To determine the expression of IL-12R by iNKT cells, we took complementary approaches. First, expression of mRNA for the two IL-12R subunits, IL-12Rβ1 and IL-12Rβ2, was determined by microarray analysis in iNKT cells and other lymphocyte subsets. IL-12Rβ1, which is responsible for binding IL-12, was expressed at very low levels by naive CD4 and CD8 T cells, memory CD8 T cells, γδT cells, and NK cells, whereas slightly higher levels were detected in memory CD4 T cells (Fig. 8 A). In contrast, much higher levels of IL-12Rβ1 mRNA were detected in both CD4⁺ and CD4⁻ iNKT cells isolated from spleen or liver. IL-12Rβ2, which is required for IL-12R–mediated signaling, was expressed at very low levels by naive CD4 and CD8 T cells, memory CD8 T cells, and γδT cells, whereas slightly higher levels were expressed by memory CD4 T cells (Fig. 8 A). In comparison, much higher levels of IL-12Rβ2 mRNA were detected in NK cells. Intermediate levels of IL-12Rβ2 were expressed by CD4⁺ and CD4⁻ iNKT cells isolated from spleen, and high levels, similar to those detected on NK cells, were detected in CD4⁺ and CD4⁻ liver iNKT cells. Second, we determined if the expression levels of IL-12R mRNA correlated with protein expression. Using CD1d tetramers to detect iNKT cells, we found that expression of IL-12Rβ1 could be detected exclusively on tetramer⁺/CD3⁺ lymphocytes (iNKT cells) but not on tetramer⁻/CD3⁻ lymphocytes (mainly NK cells) or tetramer⁻/CD3⁺ (containing MHC-restricted CD4 and CD8 αβT cells and γδT cells; Fig. 8 B).

The specific cellular effects of IL-12 are mainly a result of its ability to induce STAT4 activation. To determine if the constitutive expression of high levels of IL-12 receptor enabled prompt STAT4 activation in iNKT cells after IL-12 stimulation, we purified naive and memory CD4 T cells, NK cells, and iNKT cells by cell sorting and stimulated the purified cell populations with IL-12 for 1 h. Intracellular phosphorylation of STAT4 was assessed by antibody staining and flow cytometry. No STAT4 phosphorylation was detected in naive and memory CD4 T cells (Fig. 8 C). In contrast, phosphorylation of STAT4 was detected in a large number of NK cells and iNKT cells. Thus, unique among T cells but similar to NK cells, high constitutive expression of functional IL-12 receptor appears to provide a mechanistic explanation for the ability of iNKT cells to rapidly secrete IFN-γ in response to IL-12–mediated stimulation.

Last, we sought to address how IL-12–mediated co-stimulation alters iNKT cell activation by CD1d-presented self- and microbial lipids. The addition of recombinant IL-12 to iNKT cell lines cultured in the presence of WT DCs resulted in notably increased IFN-γ secretion, as compared with the addition of IL-12 to iNKT cell lines co-cultured in the presence of CD1d-deficient DCs (Fig. 8 D). As no exogenous antigens were added in this system, the stimulation provided by CD1d recognition was likely a result of recognition of CD1d-presented self-lipids. Next, we determined how co-stimulation with IL-12 altered iNKT cell responses to microbial antigens. A cross-titration of GSL-1 and IL-12 revealed that in particular low concentrations of GLS-1 and low concentrations of IL-12 resulted in a synergistic stimulation of IFN-γ by iNKT cells (Fig. 8 E). Thus, weak TCR-mediated stimulation of iNKT cells with CD1d-present self- or microbial antigens was significantly amplified by IL-12, resulting in potent secretion of IFN-γ by iNKT cells.

iNKT cells have been shown to play important protective roles during several bacterial, parasitic, and viral infections. However, the central question of how they are activated by microbes is not fully explained. Because viruses lack lipid antigens and antigens have only been described for a few pathogens so far, it is important to determine how many diverse pathogens can activate iNKT cells. To investigate the mechanism of iNKT cell activation in response to infection, we used a large panel of diverse bacteria, including several that express known CD1d-presented iNKT cell lipid antigens. Unexpectedly, we found that, irrespective of the expression of CD1d-presented lipid antigens by the microbes, microbe-induced iNKT cell activation and IFN-γ production in vitro and in vivo was critically dependent on IL-12 released by DCs after TLR-mediated activation. Thus, our data do not support a recently proposed model in which iNKT cell activation during infection is driven by TCR-mediated recognition of CD1d-presented microbial antigens (Mattner et al., 2005; Tupin et al., 2007). Instead, our data suggest that innate cytokine-driven activation is the dominant pathway for iNKT cell activation in response to virtually all infectious agents examined to date that induce the production of IL-12 by APCs after TLR-mediated activation (Fig. 9).

In addition to IL-12 as a critical mediator for stimulating iNKT cell responses to microbes, other studies have suggested a role for IL-18 and type I IFN in the iNKT cell response to E. coli LPS or the TLR9 agonist CpG, respectively (Nagarajan and Kronenberg, 2007; Paget et al., 2007). However, our in vitro experiments did not show a significant role for IL-18 or type I IFNs in response to any of the 11 bacterial pathogens tested. Instead, our in vitro and in vivo experiments repeatedly implicated IL-12 as the critical cytokine required for iNKT cell activation and IFN-γ secretion in response to all microbes tested. Underscoring the critical and dominant role of IL-12 for rapid activation of iNKT cells, we detected that constitutive expression of high levels of IL-12Rβ1 and IL-12Rβ2 chains by iNKT cells and IL-12 stimulation rapidly led to STAT4 activation in iNKT cells. In contrast, MHC-restricted T cells lack constitutive expression of components of the IL-12 receptor and are only known to express functional IL-12 receptor after activation and Th1 differentiation (Presky et al., 1996; Trinchieri, 2003). In agreement with previous studies, we found that naive and memory CD4⁺ T cells were not able to activate STAT4 after IL-12 stimulation in the absence of additional stimulation or differentiation. Recently, MAIT (mucosal-associated invariant T) lymphocytes, another lymphocyte population characterized by expression of an invariant TCR-α chain, have been shown to become activated by microbes independently of TLR- and cytokine-mediated stimulation, suggesting that these innate T cells require recognition of cognate microbial antigens for their activation (Le Bourhis et al., 2010). Thus, among T cells, iNKT cells appear to be uniquely equipped to promptly secrete IFN-γ upon primary encounter of an inflammatory condition that results in the production of IL-12.

Our in vitro experiments showed that full iNKT cell activation in response to all pathogens tested also required CD1d-dependent signals and that increased expression of CD1d by bacteria-stimulated APCs may contribute to iNKT cell activation. However, in the absence of co-stimulation by IL-12, the CD1d-mediated signals were not sufficient to induce IFN-γ secretion by iNKT cells. Our in vitro studies showed that iNKT cell responses to weak self-antigens or to low concentrations of microbial antigen were amplified by exogenously added IL-12. Therefore, although not able to activate iNKT cells through TCR-mediated signaling alone, it seems likely that even small numbers of CD1d–microbial antigen complexes or presentation of more potent self-antigens may contribute to or modulate iNKT cell activation during infection in the presence of IL-12–mediated co-stimulation. Furthermore, the requirement for CD1d-TCR–mediated signals may ensure that IL-12–amplified iNKT cell activation occurs only in close contact with CD1d-expressing APCs such as monocytes, DCs, macrophages, and B cells. The lack of dependence on specific recognition of CD1d-presented microbial antigens found in this study suggests that the exceptional potency of the pharmacological iNKT cell agonist α-GalCer, which is much greater than that of any naturally occurring iNKT cell antigen found to date, may have led to overestimation of the importance of microbial antigens in iNKT cell activation.

iNKT cell responses to all of the microbial antigens described so far result in the production of both IFN-γ and IL-4 by iNKT cells both in vitro and in vivo. Most of the bacteria tested in our studies resulted in a bias toward production of IFN-γ, using iNKT cell lines, freshly isolated iNKT cells, and during in vivo infection. Interestingly, the occasionally observed IL-4 induced by Sphingomonas spp. in vitro was increased in the absence of IL-12, suggesting that IL-12 may play an important role in polarizing iNKT cell responses to infection. However, in vivo infection with Sphingomonas spp. in IL-12–deficient mice did not result in a notable increase in the number of iNKT cells producing IL-4 (unpublished data). Stimulation of iNKT cells with B. burgdorferi resulted in IL-4 production by iNKT cells in our studies, which is consistent with previous studies detecting IL-4 production by a small number of iNKT cells in spleen after B. burgdorferi infection (Tupin et al., 2008). The in vitro IL-4 production observed in our studies appeared to be IL-12 independent but MyD88 dependent, suggesting that innate signals, rather than microbial antigens, are responsible for the iNKT cell IL-4 production in response to this bacterium. Further studies will be required to understand the nature of the signals responsible for bacteria-induced IL-4 production by iNKT cells.

Activation of iNKT cells during infection driven by innate signals and cytokines, rather than microbial antigens, provides several advantages. As an amplification loop for innate signals, cytokine-mediated activation enables the activation of iNKT cells in response to virtually any microorganism that induces IL-12 production, irrespective of the expression of iNKT cell ligands, and thus allows iNKT cells to overcome their restricted TCR repertoire and conserved mode of antigen-recognition. Such a mechanism of activation may be critical for iNKT cell responses to viral infection, because viruses are not known to encode enzymes for the production of unique microbial lipids, and for responses to other microorganisms that lack expression of CD1d-presented iNKT cell ligands. For the organisms that contain iNKT cell lipid antigens, synthesis, uptake, and loading of antigens into CD1d molecules, and subsequent transport of antigen–CD1d complexes to the surface of APCs, may be inefficient and slower than the IL-12–driven iNKT cell response.

The rapid innate cytokine-driven activation of iNKT cells provides the evolving immune response with T cell effector functions early during the course of infection that otherwise would only be available after the expansion and differentiation of antigen-specific MHC-restricted T cells (Brigl et al., 2003; Chiba et al., 2008). During infection, iNKT cell–secreted IFN-γ appears to serve several critically important functions. This includes a critical role in activation of macrophages and in neutrophil-mediated clearance of pathogens (Nieuwenhuis et al., 2002; Nakamatsu et al., 2007). In addition, the early IFN-γ provided by iNKT cells has been shown to be critical for the development of adaptive Th1 immune responses and may affect the magnitude and quality of MHC-restricted CD4 and CD8 T cell responses during infection (Tupin et al., 2007; Cohen et al., 2009). Together, iNKT cells, and the IFN-γ they produce, have emerged as critical regulators of the early immune response during infection.

In summary, our findings indicate that during infection with a variety of bacterial pathogens, including bacteria which express known iNKT cell antigens, iNKT cell activation is driven predominantly by IL-12, rather than by TCR-mediated recognition of CD1d-presented microbial antigens (Fig. 9). In contrast to the dominant role observed for foreign antigen recognition in the activation of MHC-restricted T cells, the lack of dependence on specific recognition of cognate microbial antigen defines a different role for the TCR in innate T cell activation. For the activation of iNKT cells during infection, CD1d–TCR interactions instead appear to determine the cell–cell interaction between iNKT cells and APCs and allow T cell activation to become manifest. This may control such innate cytokine-driven activation so that it is localized to sites of APC-produced IL-12. The constitutive expression of high levels of IL-12 receptor uniquely equips iNKT cell for immediate cytokine-driven responses. Our results suggest a unified dominantly innate cytokine-driven model of iNKT cell activation that explains how these cells become activated efficiently and rapidly in response to highly diverse microbial pathogens.

C57BL/6, IL-12p40^−/−, IL-12p35^−/−, and IL-18^−/− mice were obtained from The Jackson Laboratory. Jα18^−/− (Cui et al., 1997) and CD1d^−/− mice (Exley et al., 2003) were provided by M. Exley (Harvard Medical School, Boston, MA), Vα14Jα18 TCR transgenic mice (Bendelac et al., 1996) by A. Bendelac (University of Chicago, Chicago, IL), and MyD88^−/− (Adachi et al., 1998) by K. Kobayashi (Harvard Medical School, Boston, MA). All mice were on the C57BL/6 background, were housed in specific pathogen-free conditions or a biosafety level 2 animal facility after in vivo infections, and female mice at 6–14 wk of age were used for experiments. All animal studies were approved by the Dana-Faber Cancer Institute Animal Care and Use Committee.

iNKT cell lines were generated and maintained in complete RPMI medium (RPMI supplemented with 10% FBS [Gemini], Hepes [Invitrogen], L-glutamine, penicillin/streptomycin, and 2-ME) supplemented with 20 U/ml of recombinant mouse IL-2 (PeproTech) and 10 ng/ml IL-7 as previously described (Chiba et al., 2009). DCs were generated from BM of mice cultured in complete RPMI supplemented with 10 ng/ml of recombinant mouse GM-CSF and 1 ng/ml IL-4 for 5–8 d and purified with CD11c magnetic beads (Miltenyi Biotec). iNKT cell lines (10⁵/well) were cultured with DCs (2 × 10⁴/well) in 96-well plates (Costar; Corning) in complete RPMI. Freshly isolated iNKT cells were obtained from spleens of WT B6 animals after depletion of CD19⁺ and CD8a⁺ cells using magnetic beads (Miltenyi Biotec), followed by cell sorting using CD1d tetramers and TCR-β staining on a FACSAria IIu instrument (BD). GSL-1 (National Institutes of Health [NIH] Tetramer Core Facility), LPS (Salmonella enterica serotype Typhimurium; Sigma-Aldrich), CpG (ODN1826; InvivoGen), recombinant mouse IL-12 (PeproTech), or bacteria were added and cultures were incubated for 16–48 h at 37°C. Culture supernatants were analyzed by ELISA with matched antibody pairs (IFN-γ and IL-4 [BD], sensitivity 10 and 2 pg/ml, respectively; IL-18 and IFN-β [R&D Systems]). Blocking anti–IL-12 (BD) and anti–mouse IFN-α/β R1 antibodies (R&D Systems) were added to cultures at 5–10 µg/ml. Expression of CD25 was determined on iNKT cells by flow cytometry (gated by FSC/SSC). Surface expression of CD1d on CD11c⁺ BM DCs was determined using CD1d Abs (BD) after incubation with various stimuli.

S. pneumoniae (for strains and references see Table I; provided by K. Kawakami (University of the Ryukyus, Nishihara, Okinawa, Japan), cultured in THY [BD] or BHI [Oxoid] medium), L. monocytogenes (BHI medium), S. aureus (LB medium; BD), P. aeruginosa (LB medium, provided by R. Blumberg, Harvard Medical School, Boston, MA), S. typhimurium (LB medium), E. coli (LB medium), S. capsulata (MH medium, 30°C; Oxoid), N. aromaticivorans (MH medium), S. yanoikuyae (TSB), and B. burgdorferi (BSK-H medium, 33°C; provided by L. Bockenstedt, Yale School of Medicine, New Haven, CT; Sigma-Aldrich) were grown in culture medium at 37°C, unless otherwise noted, until mid- to late-log phase, washed 3× in LPS-free PBS (Invitrogen), and heat inactivated at 65°C for 45 min. B. burgdorferi were inactivated at 48°C for 30 min. M. tuberculosis were inactivated by γ radiation (Colorado State University, Fort Collins, CO). All bacterial preparations, except gram-negative bacteria, tested negative for LPS by Limulus amebocyte lysate test (limit of detection 0.03 EU/ml; Associates of Cape Cod).

Lipids were extracted and analyzed by thin layer chromatography, as described elsewhere (Tatituri et al., 2007), or analyzed by high performance liquid chromatography (HPLC) using a AutoPurification HPLC system coupled to a mass detector with both ES and APCI capabilities, evaporative light scattering detector, and photodiode array detector (Waters). MS analysis, including low-energy CAD MSn experiments, were conducted on a linear ion-trap (LIT) mass spectrometer (Finnigan; Thermo Fisher Scientific) with Xcalibur operating system. Lipid extracts dissolved in chloroform/methanol (1/2) were continuously infused (2 µl/min) to the ESI source, where the skimmer was set at ground potential, the electrospray needle was set at 4.5 kV, and temperature of the heated capillary was 300°C. The automatic gain control of the ion trap was set to 2 × 10⁴, with a maximum injection time of 100 ms. Helium was used as the buffer and collision gas at a pressure of 10⁻³ mbar (0.75 mTorr). The MSn experiments were performed with an optimized relative collision energy ranging from 18–25%, an activation q value at 0.25, and the activation time at 30–50 ms to leave a minimal residual abundance of precursor ion (around 10%). Mass spectra were accumulated in the profile mode, typically for 3–10 min for MSn (n = 2, 3, and 4) spectra. The mass resolution of the instrument was tuned to 0.6 D at half peak height. Synthetic GSL-1 was provided by the NIH tetramer facility and P.B. Savage (Brigham Young University, Provo, UT; Mattner et al., 2005). Synthesis of α-GalCer has been previously described (Veerapen et al., 2009).

Bacteria were grown to mid-log phase, as described in Bacteria, and added to DCs cultured in antibiotic-free complete RPMI. After a 3-h incubation at 37°C, infected DCs were washed 3× in complete RPMI. Infected DCs (5 × 10⁴/well) were co-cultured with iNKT cell lines (10⁵/well) in complete RPMI medium in 96-well plates for 16–24 h.

S. pneumoniae was grown as described in Bacteria and mice were anesthetized with i.p. ketamine/xylazine or by inhalation isoflurane and inoculated intranasally or intratracheally, respectively, with 2–3 × 10⁶ bacteria in PBS. CFUs were determined on blood agar plates. For flow cytometric analysis, mononuclear cells were isolated from lungs perfused with PBS followed by digestion with collagenase/DNase. S. yanoikuyae was grown in TSB supplemented with 5% sheep blood and passaged in SCID mice. Mice were injected i.v. with 4 × 10⁹ bacteria in 200 µl PBS. Cells were stained with antibodies against CD45, CD19, TCR-β, PBS57-loaded CD1d tetramers (NIH Tetramer Core Facility), CD69, or isotype-matched control antibodies. Secretion of IFN-γ or IL-4 was determined by cytokine secretion assay (Miltenyi Biotec) according to the manufacturer’s instructions. Flow cytometric data were collected on a LSR II flow cytometer (BD) and analyzed with FlowJo software (Tree Star).

iNKT cells were purified from spleen or liver on a MoFlo cell sorter (Dako) gating on CD19⁻, B220⁻, Ter119⁻, CD11b⁻, CD11c⁻, LyG/Gr-1⁻, CD45⁺, TCR^int, CD1d-tetramer⁺, CD4⁺, or CD4⁻ iNKT cells, followed by RNA extraction with Trizol (Invitrogen), RNA amplified, and hybridized to Affymetrix 1.0 ST MuGene arrays within the NIAID/NIH ImmGen project (protocols and raw data available at http://www.immgen.org; Heng and Painter, 2008). For flow cytometric analysis, cells were stained with CD1d tetramers and antibodies against IL-12Rβ1 (BD).

Naive (CD62L^hiCD44^lo) and memory (CD62L^loCD44^hi) TCRβ⁺CD4⁺ T cells, NK cells (NK1.1⁺TCR-β⁻), and iNKT cells (as in the previous section) were isolated by negative selection and cell sorting. 1–5 × 10⁵ purified cells were cultured in completed RPMI and stimulated with recombinant IL-12 for 1 h at 37°C and subsequently stained with antibodies against pSTAT4 (BD) according to manufacturer’s protocols.

Fig. S1 shows mass spectrometry analysis of microbial lipids separated by liquid chromatography. Fig. S2 shows representative dose titrations for bar graphs depicted in Fig. 4 (A and C). Fig. S3 shows IL-18 and type I IFN independence of iNKT cell activation in response to microbes. Fig. S4 shows representative FACS plots for bar graphs depicted in Fig. 4 D. Fig. S5 shows IL-12 dependence of IFN-γ secretion and CD25 expression by iNKT cells on day 2 after activation with microbial products. Fig. S6 shows cytokine responses to microbial stimulation of freshly isolated iNKT cells. Fig. S7 shows MyD88-dependent CD1d expression by DCs after stimulation with microbial products. Fig. S8 shows survival of WT and CD1d-deficient mice after S. pneumoniae infection and in vivo activation of iNKT cells on day 4 after S. pneumoniae infection.

The authors would like to thank the NIH Tetramer Core Facility for mouse CD1d tetramer and GSL-1, A. Bendelac, M. Exley, and K. Kobayashi for providing mice, K. Kawakami, L. Bockenstedt, and R. Blumberg for providing bacteria, P.B. Savage for providing GSL-1, P. Brennan for critically reading the manuscript, and S. Moskowitz for help with the artwork.

This work was supported by NIH grants AI077795 (M. Brigl), AI028973, and AI063428 (both M.B. Brenner) and grants P41-RR-00954, R37-DK-34388, and P60-DK-20579 (Mass Spectrometry Center, Washington University).

The authors have no conflicting financial interests.