NKT Cells Are Required for Complete Freund’s Adjuvant-Mediated Protection from Autoimmune Diabetes

Type 1 diabetes (T1D) is an autoimmune disease in which pancreatic islet β-cells are targeted and destroyed by the host immune system (1). Studies of the NOD mouse, a model of spontaneous T1D, have demonstrated that autoaggressive CD4 and CD8 T cells mediate β-cell destruction (2). Although the precise factors responsible for initiating pathogenesis are unknown, it has been proposed that defects in T and B lymphocyte selection (central tolerance) and regulatory lymphocyte function (peripheral tolerance) involving FOXP3-expressing CD4 T cells, NKT cells, and NK cells underlie the development of T1D (3–5). Moreover, investigations of the NOD mouse suggest that NKT cells play a pivotal role in the development of T1D (6–16).

Invariant NKT (iNKT) cells compose a unique subset of T cells that expresses an invariant TCR together with surface markers characteristically found on NK cells (17–19). iNKT cells use a limited repertoire of TCRs, a constant TCR α-chain (Vα14Jα18), and a small set of Vβ-chains to recognize hydrophobic lipid and glycolipid Ags presented by the non-polymorphic MHC class I-like molecule CD1d (17–19). Thymic CD1d expression is essential for the positive selection of iNKT cells because CD1d^−/− mice completely lack this lymphocyte subset (17–19). As NKT cells can rapidly secrete large amounts of Th1 and Th2 cytokines upon TCR stimulation (17–19), they are thought to act as important immunoregulators, activating cells of the innate and adaptive immune systems, including NK cells, dendritic cells (DCs), macrophages, and MHC-restricted T cells and B cells (18–20).

The injection of mycobacterial preparations, such as bacillus Calmette-Guérin (BCG; Mycobacterium bovis) or CFA, containing inactivated Mycobacterium tuberculosis prevents the onset of autoimmune diabetes in NOD mice and BB diabetic rats (21–23). Previously, we have shown that CFA rapidly activates NK cells and that they are required for protection from disease in the NOD mouse model (24, 25). However, the mechanism by which CFA activates NK cells and prevents diabetes onset remains unknown. In this study, we demonstrate that NKT cells are a critical element of the adjuvant effects of CFA and are essential in CFA-mediated NK cell activation and protection from diabetes.

Female C57BL/6J (B6) and NOD mice were purchased from The Jackson Laboratory (Bar Harbor, ME) and bred in a specific pathogen-free facility at the Child and Family Research Institute. CD1d^−/− NOD and CD1d^−/− B6 mouse strains, generated independently (12, 26), have been back-crossed at least 11 and 12 generations onto their respective backgrounds and were generously obtained from Dr. Mark A. Exley (Harvard University) and Dr. Terry L. Delovitch (University of Western Ontario), respectively. MyD88^−/− mice, bred at least 10 generations onto NOD background, have been described elsewhere (27) and were kindly provided by Dr. Marc S. Horwitz (University of British Columbia). All studies followed guidelines set by the Animal Care Committee at the University of British Columbia in conjunction with the Canadian Council on Animal Care.

Abs recognizing CD3ε (145-2C11), TCR-β (H57-597), CD69 (H1.2F3), DX5 (HMα2), NKp46 (29A1.4), and B220 (RA3-6B2) were purchased from eBioscience (San Diego, CA). Anti-NKG2D (191004) Ab was acquired from R&D Systems (Minneapolis, MN). Murine CD1d tetramers, loaded with α-galactosylceramide (αGC) and conjugated to PE or allophycocyanin, were provided from the National Institutes of Health Tetramer Core Facility (Emory University). Single-cell suspensions were generated from spleens and treated with RBC lysis buffer containing NH₄Cl. For isolation of hepatic lymphocytes, livers were perfused with PBS via the portal vein, mashed through a metal mesh, and enriched through Ficoll-Paque gradient centrifugation. Cells were fluorescently labeled by incubation with the indicated Abs in FACS buffer (PBS containing 0.3% BSA) for 30 min on ice. Subsequently, samples were washed and suspended in PBS containing 1% FCS and 2.5% paraformaldehyde. Data were acquired on a FACSCalibur using CellQuest software (BD Biosciences) or BD FACSCanto with BD FACSDiva software (BD Biosciences) and analyzed with FlowJo (Tree Star, Ashland, OR).

Mice were injected s.c. with a single 100-μl dose of CFA emulsion (contains heat-killed and dried M. tuberculosis strain H37Ra, ATCC 25177; No. F5881, Sigma-Aldrich, St. Louis, MO) into the bases of their tails. For αGC stimulations, 10 μg αGC, kindly provided by Dr. Huib Ovaa of The Netherlands Cancer Institute, was dissolved in 100 μl PBS and injected i.v. via the tail vein as previously described (20). At the indicated time points, lymphocytes from livers and spleens were harvested, labeled with surface markers directly ex vivo, or treated with GolgiStop (contains monensin; BD Biosciences) for 4 h at 37°C and stained intracellularly with anti-mouse IFN-γ (XMG1.2) or anti–IL-4 (11B11) Abs (BD Biosciences).

RBC-depleted bone marrow cells were seeded at 3 × 10⁶ cells/ml in complete medium supplemented with GM-CSF (10 ng/ml) plus IL-4 (10 ng/ml) for 5 d. Subsequently, bone marrow-derived dendritic cells (BMDCs; 2.5 × 10⁵ cells/well) were incubated in medium alone or 100 μg/ml heat-killed M. tuberculosis (H37RA; BD Biosciences) or 10 ng/ml αGC in a 96-well plate for 24 h. DN32 NKT cells (2.5 × 10⁵ cells/well), a mouse Vα14Jα18 invariant TCR-positive T cell hybridoma (28), were subsequently added into each well and incubated for 24 h at 37°C. For blocking TCR–CD1d interactions, 10 μg/ml anti-CD1d Ab (clone 1B1; BioLegend) were added per well. ELISAs were based on capture and detection anti–IL-2 Abs JES6-1A12 and JES6-5H4 from BD Biosciences.

Five- to six-week-old NOD and CD1d^−/− NOD mice were given a single dose of CFA as described previously (24), and blood glucose was monitored weekly using test strips (LifeScan, Milpitas, CA).

Student t tests (parametric) and Mann–Whitney U tests (non-parametric) were used to calculate statistical significance as indicated. Differences in diabetes incidence between mouse groups were deemed significant using a two-sided Fisher's exact test.

NK cells are rapidly activated when mice are treated with heat-killed M. tuberculosis (24, 25, 29) or infected with live M. tuberculosis (30, 31). Observations describing rapid NK cell activation after αGC treatment of mice have suggested the possibility that CFA-induced NK cell activation may be mediated through the action of iNKT cells (20, 32). To test the hypothesis that the ability of M. tuberculosis to trigger NK cell activation is dependent on iNKT cells, we measured CD69 expression and IFN-γ production by NK cells in wild-type and CD1d^−/− B6 mice 16 h posttreatment with adjuvant containing (CFA) or control lacking (IFA) M. tuberculosis. These experiments revealed that hepatic and splenic NK cells in wild-type mice induced CD69 expression and produced IFN-γ after CFA treatment, whereas those derived from CD1d^−/− mice did not (Fig. 1A, 1B). By contrast, IFA treatment did not affect NK cell expression of CD69 or IFN-γ regardless of mouse genotype (data not shown). The inability of CFA to activate NK cells in CD1d^−/− mice is not likely due to an intrinsic NK cell defect because NK cell number and development appear normal in these mice (33). Furthermore, a previous study has shown that NK cells from CD1d^−/− mice respond normally to polyinosinic-polycytidylic acid, proliferating and acquiring effector functions at a level comparable with that of NK cells in wild-type mice (32). Our findings show that CFA activates NK cells through an NK cell-extrinsic mechanism that is dependent on CD1d expression.

We have documented that administration of CFA to NOD mice results in the rapid induction of NK cell proliferation that is associated with a rise in their frequency in the blood and spleen (25). Next, we tested whether CFA-mediated mobilization of NK cells is dependent on CD1d-restricted NKT cells by treating a second cohort of mice as described above and measuring hepatic and splenic NK cell frequencies (Fig. 1C, 1D). Consistent with our previous findings in NOD mice (25), CFA treatment of wild-type B6 mice induced a sharp increase in the proportion of NK cells (defined as TCR-β⁻ DX5⁺ NKG2D⁺ events) among hepatic lymphocytes, whereas no change was observed upon IFA application relative to naive mice (2-fold; CFA 20.9 ± 2.9% versus IFA 10.4 ± 1.5%; p < 0.01; Fig. 1C and data not shown). By contrast, CFA administration produced a less dramatic rise in NK cell frequencies among splenocytes (1.2-fold change; CFA 2.5 ± 0.2% versus IFA 2.0 ± 0.1%; p < 0.05; Fig. 1D and data not shown). In contrast to wild-type mice, NK cell frequencies in CD1d^−/− mice were not found to change after CFA injection (Fig. 1C, 1D).

Observations that NKT cells can downmodulate their TCR expression after activation raises the possibility that the recorded rises in NK cell frequencies upon CFA administration may be attributable, at least in part, to the contamination of our NK cell (NK1.1⁺ CD3⁻) gate with NKT cells. Moreover, the discrimination of NK from NKT cells can be convoluted because these two lineages share many surface markers. To more carefully delineate NK from NKT cells, we performed additional stains using the Abs directed at NKp46 and DX5, two markers that are expressed by the vast majority of NK cells but only a scant fraction of iNKT cells (34, 35). Regardless of IFA or CFA treatment, we found that most (>94%) NK cells (NK1.1⁺ TCR-β⁻ events) exhibited strong expression of NKp46 and DX5, whereas most (>97%) NKT cells (αGC/CD1d tetramer⁺ TCRβ⁺ events) did not (Fig. 2A, 2B). Using this more rigorous NK cell gating strategy (TCR-β⁻ NK1.1⁺ DX5⁺ NKp46⁺), CFA was found to cause a 2-fold increase in NK cell frequencies among hepatic lymphocytes relative to IFA or untreated mice (CFA 22.2 ± 0.8%, IFA 9.1 ± 0.3%; Fig. 2C and data not shown), recapitulating our observations in Fig. 1C. Collectively, these studies demonstrate that the M. tuberculosis-derived components of CFA cause a rise in hepatic NK cell frequency that is dependent on CD1d expression.

Previous work has shown that NKT cells are activated by and proliferate in response to BCG infection in mice (36). Combined with our findings that CD1d expression is critical for NK cell activation by mycobacterial components of CFA (Fig. 1), we investigated whether NKT cells are activated by exposure to CFA. Wild-type B6 mice were injected with CFA for the indicated time periods and the immune activation of hepatic NKT cells assessed by measuring expression levels of the early activation marker CD69 (Fig. 3A). Our temporal analyses demonstrated that hepatic NKT cells elevated CD69 expression by as little as 6 h after CFA treatment (mean fluorescence intensity = 160.3 ± 17.5 versus 65.3 ± 2.8 in naive mice) and that their levels peaked by 16 h postinjection (mean fluorescence intensity = 213 ± 54). By contrast, CFA was found to have minimal effect on surface levels of CD69 expressed by splenic NKT cells (data not shown). Because activated NKT cells have been shown to produce both Th1 and Th2 cytokines, including IFN-γ and IL-4 (37), we next considered whether CFA could induce NKT cells to secrete either of these two cytokines (Fig. 3B). Subcutaneous CFA injection resulted in a marked increase in the proportion of NKT cells synthesizing IFN-γ (7.4-fold change; 6.7 ± 1.0% versus 0.9 ± 0.1% in naive mice). However, we were unable to detect any IL-4 production by NKT cells after CFA injection (data not shown). By contrast, NKT cells from mice injected intravenously with the powerful NKT cell agonist αGC were found to produce high levels of IFN-γ and IL-4 (Fig. 3B and data not shown). Together, these experiments indicate that CFA injection rapidly induces NKT cells to express CD69 and to produce IFN-γ but not IL-4.

The activation of NKT cells in vivo through any of anti-TCR Ab, αGC, IL-12, or BCG induces an apparent rapid loss of hepatic NKT cells (36, 38–40), likely the result of activation-induced cell death, NK1.1/TCR downregulation, and/or trafficking of these cells to other sites. Regardless of the mechanism(s) responsible for their disappearance, NKT cell numbers return to baseline a short time later demonstrating their tight homeostatic control (38–40). Thus, we examined whether CFA-mediated NKT cell activation affects the distribution of NKT cells found in the liver and spleen (Fig. 3C). Timed analysis revealed that CFA triggered a 3-fold reduction in NKT cell frequencies within the liver by 6 h postinjection (11.1 ± 1.6% in CFA-treated mice versus 31.9 ± 3.9% in naive mice), whereas the proportions did not change substantially in the spleen (Fig. 3C and data not shown). Remarkably, the effect of CFA on the proportion of NKT cells among hepatic lymphocytes was transient, resembling frequencies observed in naive mice by 24 h posttreatment (26.5 ± 2.1% versus 31.9 ± 3.9% in naive mice) and closely mirroring the timing of activation observed above (Fig. 3A, 3B). These experiments show that CFA recapitulates effects of well-established NKT cell agonists to activate NKT cells and decrease their proportion among hepatic lymphocytes.

A few possible mechanisms exist by which mycobacterial Ags in CFA may activate NKT cells. For instance, NKT cells may be affected directly, through TCR recognition of foreign-lipid Ags presented by CD1d molecules on the surface of an APC (18, 19, 41–43), or indirectly, via presentation of endogenous lipids and production of cytokines, such as IL-12 and IL-18, by APCs upon TLR-induced activation (44, 45). It is also possible that NKT cells may be directly activated through cell-intrinsic recognition of TLR agonists, a phenomenon recently suggested for conventional CD8 T cells (46). That heat-killed M. tuberculosis is the active component of CFA and contains several TLR2 agonists (47, 48) suggests that CFA may activate NKT cells through a TCR-independent, TLR-dependent pathway.

To investigate whether the activation of NKT cells by M. tuberculosis is mediated through TLR signaling, we used mice and cells lacking MyD88, a signaling adapter molecule required for signal transduction and inflammatory cytokine production induced by TLR2 as well as other TLRs (49). Wild-type and MyD88^−/− NOD mice were injected with CFA or IFA, and the frequency and activation status of hepatic NKT cells was assessed at 16 h (Fig. 4A, 4B). Treatment with CFA was found to induce similar levels of CD69 and IFN-γ expression by hepatic NKT cells in MyD88^−/− mice relative to wild type. Corresponding with NKT cell activation, hepatic NKT cell frequencies in wild-type and MyD88^−/− mice declined to a similar extent after administration of CFA (Fig. 4C). By contrast, the proportion of NKT cells found in the liver of IFA-treated mice did not fluctuate regardless of genotype, conforming with those observed in naive animals (Fig. 4C and data not shown). These observations demonstrate that components of M. tuberculosis activate NKT cells through a MyD88-independent mechanism.

To determine whether NKT cells could be directly activated by M. tuberculosis-stimulated DCs and whether MyD88 modulates this interaction, we designed an in vitro assay using the NKT cell line DN32, a murine CD1d-restricted, Vα14Jα18 invariant TCR-expressing T cell hybridoma (28). DN32 cells were stimulated with BMDCs from wild-type and MyD88^−/− mice treated with medium alone (untreated), heat-killed M. tuberculosis (strain H37Ra from CFA), or αGC (positive control) after which NKT cell activation was assessed through measurements of CD69 expression and IL-2 secretion (Fig. 4D, 4E). In these experiments, both wild-type and MyD88^−/− DCs were equally capable of inducing DN32 cells to elevate CD69 and IL-2 expression. In addition to components activating TLR pathways, reports have described the peptidoglycan constituent muramyl dipeptide (MDP), recognized by the intracellular NOD2 receptor, as one of the major adjuvant activities of M. tuberculosis (50, 51). As a consequence, we investigated whether MDP can activate NKT cells by incubating DN32 cells with treated BMDCs in vitro. However, two different formulations of MDP (Sigma or Invivogen) were found to be incapable of activating NKT cells despite their ability to induce DCs to upregulate CD80/CD86 expression and secrete IL-1β (data not shown). Together, these findings suggest that M. tuberculosis-treated DCs directly activate NKT cells through MyD88- and NOD2-independent pathways.

NKT cells are activated during a variety of microbial infections through the recognition of exogenous- or endogenous-lipid Ags in a CD1d-dependent manner (52). To determine if M. tuberculosis-induced NKT cell activation requires TCR recognition of CD1d molecules, the DN32 NKT cell line was stimulated with BMDCs as described earlier (Fig. 4D, 4E) except in this instance, BMDCs were generated from wild-type or CD1d^−/− B6 mice, and the stimulations were performed in the presence of control rat or blocking anti-CD1d Abs (Fig. 5). Similar to findings obtained using NOD DCs (Fig. 4D, 4E), DN32 cells that were stimulated with M. tuberculosis-treated DCs from B6 mice upregulated surface CD69 expression and produced IL-2, unlike NKT cells incubated without Ag (Fig. 5). In addition, M. tuberculosis-induced elevations of CD69 and IL-2 levels by DN32 cells were sensitive to the addition of blocking anti-CD1d Abs and were lost when DC stimulators lacked CD1d expression, implying that NKT cell activation depends on TCR–CD1d interaction. As expected, the activation of DN32 cells with the superagonist αGC was also dependent on CD1d expression and blocked by anti-CD1d Abs (Fig. 5). Together, these results indicate that M. tuberculosis-induced NKT cell activation requires TCR recognition of CD1d molecules.

NK cells and their capacity to secrete IFN-γ have been shown to be critical in CFA-mediated protection of NOD mice from diabetes (24, 25). This, in association with our observations indicating that NK cell activation by CFA is dependent on CD1d (Fig. 1), led us to hypothesize that CFA-mediated protection of NOD mice from diabetes requires CD1d expression. To address this hypothesis, wild-type and CD1d^−/− female NOD mice were injected with a single dose of CFA, IFA, or PBS and blood glucose levels monitored weekly to detect the onset of diabetes (Fig. 6). As expected (21, 25, 53), treatment of wild-type NOD mice with CFA at 5–6 wk of age prevented the development of disease, whereas mice treated with IFA or PBS had no effect on the natural history of disease (Fig. 6A, 6B and data not shown). In accord with other groups (11, 13) but in contrast to another in which CD1d-deficiency exacerbated disease (12), analysis of CD1d^−/− female NOD mice revealed no significant differences in disease incidence or age of diabetes onset compared with wild-type NOD females. Remarkably, the incidence of diabetes in the CD1d^−/− cohort was unaffected by administration of CFA (Fig. 6C). These experiments indicate that CD1d is essential for CFA-mediated protection of NOD mice from diabetes.

NKT cells have an important role in bridging innate and adaptive immune responses by providing direct or indirect help to various subsets of immune cells including NK cells (54). Our previous studies have shown that CFA rapidly activates NK cells and that they are required for protection of NOD mice from diabetes (24, 25). However, the cellular and molecular mechanisms by which CFA activates NK cells and prevents disease remains unknown. NK cell activation has been shown to be dependent on NKT cells and, at least in part, their release of IFN-γ after αGC treatment or microbial infection (20, 32, 55). However, NKT cells are not required for NK cells to function, as steady-state levels of NK cell cytotoxicity are similar between NKT cell-deficient (CD1d^−/− or Jα281^−/−) and wild-type mice (32). As well, NK cells from wild-type and NKT cell-deficient mice have equivalent proliferative and effector function when stimulated in vivo with the NK cell activator polyinosinic-polycytidylic acid (32). In this study, we investigated the role that NKT cells play in the activation of NK cells after CFA administration. Unexpectedly, we observed that NKT cells are rapidly activated upon CFA administration through a MyD88-independent mechanism and are also required for NK cell activation and prevention of diabetes by this adjuvant. These results reveal that CFA-mediated protection from diabetes links CFA-dependent mobilization of NKT cells to sequential activation of NK cells and subsequent regulation of anti–β-islet cell autoimmune responses.

The classical adjuvant properties of CFA and its ability to induce proinflammatory immune responses have been ascribed to the capacity of M. tuberculosis to stimulate the innate immune system through TLR signaling and subsequently Ag-specific adaptive immunity. However, Gavin et al. (56) have shown that mice lacking the essential TLR signal transduction adapter proteins, MyD88 and TRIF, maintain their ability to mount robust Ab responses to T cell-dependent Ags in response to CFA, as well as three other commonly used adjuvants. Therefore, the adjuvant activity of CFA must include at least one mechanism that is independent of TLR signaling. Consistent with this conclusion, our experiments show that CFA activates NKT cells through a MyD88-independent mechanism. However, because TLR3 signaling makes exclusive use of the TRIF adapter protein and TLR4 uses both TRIF and MyD88 adapters (49), we cannot rule out the possibility that TLR3 or TLR4 is responsible for mediating the inflammatory effects of mycobacterial Ags. Currently, no TLR3 or TLR4 agonist activities have been ascribed to heat-killed M. tuberculosis although viable bacteria have been shown to trigger both TLR2 and TLR4 signaling (47, 48). In addition, Wen et al. (27) have shown that CFA administration to TLR2^−/−, TLR3^−/−, or TLR4^−/− NOD mice prevents onset of diabetes, arguing that neither of these pathways by themselves is essential for disease protection. Nevertheless, it is possible that other innate immune receptors besides TLRs recognize pathogen-associated molecular patterns present in heat-killed M. tuberculosis and are responsible for mediating NKT cell activation (57).

CD1d has been shown to mediate microbial activation of NKT cells through one of two distinct mechanisms (18, 19). First, microbial lipid Ags may directly activate NKT cells through CD1d-Ag recognition by TCR. Examples of exogenous NKT cell Ags (besides the prototypical but nonphysiological αGC) include glycosphingolipids from Gram-negative α-Proteobacteria and diacylgycerols from Borrelia burgdorferi (41–43). Mycobacterial lipids also contain phosphatidylinositol mannosides (natural Ags for NKT cells present within BCG) that can activate a subset of both human and mouse NKT cells in vitro in a CD1d-dependent manner (58, 59). Second, microbial products may indirectly activate NKT cells through their potential to induce the proinflammatory cytokines, IL-12 and IL-18, thus amplifying NKT cell reactivity for self-lipid Ags (44, 45). Thus, in the model of CFA-mediated diabetes protection reported in this study, microbial products from M. tuberculosis may tune down anti-islet β-cell immune responses by acting as cognate Ags for NKT cells and/or by serving to stimulate cytokine expression and self-lipid Ag presentation. In conclusion, this study demonstrates that heat-killed M. tuberculosis stimulates NKT cells leading to NK cell activation and the prevention of autoimmune diabetes. The modulation of NKT cell function by M. tuberculosis-derived products may represent a therapeutic option for protection against the onset of T1D.

We thank Dr. Lenka L. Allan (University of British Columbia) for critical reading of the manuscript, Eva Germann (University of British Columbia) for assistance with statistical analyses, James Wetherill (Braun Geotechnical) for reagents, the National Institutes of Health Core Tetramer Facility for preparation of CD1d tetramers, and the Child & Family Research Institute animal facility for animal husbandry.

The authors have no financial conflicts of interest.