|Home | About | Journals | Submit | Contact Us | Français|
Invariant NKT (iNKT) cells are a unique subset of T lymphocytes that rapidly carry out effector functions following activation with glycolipid Ags, such as the model Ag α-galactosylceramide (αGalCer). Numerous studies have investigated the mechanisms leading to Th1- and Th2 cytokine production by iNKT cells, and the effects of the copious amounts of cytokines these cells produce. Less is known, however, about the mechanisms of iNKT cell cytotoxicity. Here we investigated the effect of antigen availability and strength, as well as the molecules involved in iNKT cytotoxicity. We demonstrate that the iNKT cell cytotoxicity in vivo correlates directly with the amount of CD1d expressed by the targets as well as the TCR affinity for the target glycolipid Ag. iNKT cells from spleen, liver and thymus were comparable in their cytotoxicity in vitro. Surprisingly, we show that the antigen-specific cytotoxicity of iNKT cells in vivo depended almost exclusively on the interaction of CD95 (Fas) with CD178 (FasL), and that this mechanism can be efficiently utilized for tumor protection. Therefore unlike NK cells, which rely mostly on perforin/granzyme mediated mechanisms, the antigen-specific cytotoxicity of iNKT cells in vivo is largely restricted to the CD95/CD178 pathway.
Invariant NKT (iNKT) cells are a unique subset of T lymphocytes characterized by the expression an invariant Vα14-Jα18 TCR rearrangement (Vα14 iNKT cells) and the recognition of CD1d, a non-polymorphic MHC class I homolog. CD1d binds lipid structures and one of the best studied Vα14 iNKT cell antigens is α-galactosylceramide (αGalCer), a glycolipid originally isolated from a marine sponge, or perhaps more likely, from a Sphingomonas microorganism associated with the sponge (1, 2). Activation with this model antigen leads to rapid induction of effector functions by Vα14 iNKT cells. Indeed the identification of αGalCer was largely based on the discovery of its anti-tumor activity (1, 2), and numerous studies describe the strong iNKT cell- and CD1d-dependent anti-tumor properties of αGalCer.
Previous studies on iNKT cell effector functions have focused principally on their strong and rapid production of Th1- and Th2 cytokines (3–6). However, less attention has been given to their cytotoxic potential. Some reports demonstrate NK cell like cytotoxicity of iNKT cells following activation with IL-12 (7–14) or αGalCer (1, 2, 15–27). Yet few studies addressed the antigen-specificity of this cytotoxicity. We previously demonstrated antigen-specific cytotoxicity by transgenic, non-Vα14i-expressing CD8+ NKT (tgNKT) cells (28). However, little is known about the properties of and requirements for antigen-specific cytotoxicity by iNKT cells (24, 29, 30).
To achieve a better understanding of the cytotoxic potential of iNKT cells in vitro and in vivo, we utilized an antigen-specific in vivo cytotoxicity assay (28, 31). We investigated the effect of antigen availability and strength on iNKT cell cytotoxicity and found a positive correlation between these two parameters. Furthermore, we demonstrate that the antigen-specific iNKT cell cytotoxicity in vivo depends on the interaction of CD95 and its ligand CD178. In addition, we demonstrate that the direct antigen-specific cytotoxicity by iNKT cells can be utilized for anti-tumor responses.
All mice were housed under SPF conditions at the animal facilities of the LIAI (La Jolla, CA) and the Scripps Research Institute (La Jolla, CA) in accordance with the Institutional Animal Care Committee guidelines. Experiments were performed according to animal experimental ethics committee guidelines. C57BL/6J mice (#000664), CD45.1 congenic B6.SJL mice (#002014) and mice deficient for perforin (#002407), CD95/Fas (#000482) and CD178/FasL (CD95L, #001021) on the C57BL/6 background were purchased from the Jackson Laboratories (Bar Harbor, ME). B6.129-Tcra-Jtm1Tgi (Ja18−/−) mice and CD1d-deficient mice (CD1d−/−) on the C57BL/6 background were the kind gift of Dr. M. Taniguchi (RIKEN Institute, Yokohama, Japan) and Dr. Luc Van Kaer (Vanderbilt University, Nashville, TN) respectively. The B lymphoma A20 (BALB/cAnN, #TIB-208) and the melanoma B16-F10 (C57BL/6, #CRL-6475) were purchased from American Type Culture Collection (ATCC) (Manassas, VA) and were virally transfected to stably express CD1d as previously described for A20 (32), resulting in the lines A20-CD1d and B16-CD1d.
α-galactosylceramide (αGalCer; (2S,3S,4R)-1-O-(a-D-galactopyranosyl)-N-hexacosanoyl-2-amino-1,3,4-octadecanetriol) and OCH were obtained from the Kirin pharmaceutical research corporation (Gunma, Japan). C-glycoside and GalA-GSL (GSL-1’) were obtained from the NIH tetramer core facility (Emory University, Atlanta, GA). CFDA-SE was obtained from Invitrogen (Carlsbad, CA). Monoclonal antibodies (mAbs) against the following mouse antigens were used in this study: CD1d (1B1), CD3ε (145.2C11, 17A2), CD4 (GK1.5, RM4-5), CD8α (53-6.7, 5H10), CD11b (M1/70), CD11c (HL3), CD19 (1D3, 6D5), CD21/CD35 (7G6), CD23 (B3B4), CD24 (M1/69), CD25 (PC61.5), CD44 (IM7), CD45.1 (A20), CD45.2 (104), CD45R (B220, RA3-6B2), CD69 (H1.2F3), CD95/Fas (Jo2), CD178/FasL (MFL3), TCRβ (H57-597), Thy1.2 (30-H12, 53-2.1), NK1.1 (PK136), Ly6CG (Gr1), perforin (CB5.4, eBioOMAK-D), granzyme B (GB11), TNFα (MP6-XT22), IL-4 (11B11) and IFNγ (XMG1.2). Antibodies were purchased from Abcam (Cambridge, MA), BD Biosciences (San Diego, CA), BioLegend (San Diego, CA), eBioscience (San Diego, CA) or Invitrogen (Carlsbad, CA). Antibodies were biotinylated or conjugated to Pacific Blue, Alexa Fluor 405, Pacific Orange, V500, FITC, Alexa Fluor 488, PE, PE-TexasRed, PerCP, PerCP-Cy5.5, PE-Cy7, APC, Alexa Fluor 647, Alexa Fluor 700, APC-Cy7, APC-AlexaFluor750 or APC-eFluor780. Anti-mouse CD16/32 used for Fc receptor blocking was isolated in our laboratory. Purification of mouse CD1d and preparation of PE- and APC-conjugated αGalCer loaded CD1d tetramers were produced as described (33).
Single-cell suspensions were prepared from the liver, spleen and thymus. Prior to extraction, the liver was perfused with PBS via the portal vein until opaque and meshed through a 100 µm cell strainer (BD Biosciences, San Diego, CA) and washed. Total liver cells were then resuspended in a 40% isotonic Percoll (Amersham, Piscataway NJ) solution and underlaid with a 70% isotonic Percoll solution. After centrifugation for 20 min at 900 g, mononuclear cells were isolated from the 40/70% interface. The cells were washed once with medium. For the spleen and the thymus, lymphocytes were filtered through a 70 µm cell strainer (BD Biosciences). With splenocytes indented for in vitro cultures red blood cells were lysed (Red Cell Lysis Buffer (RCLB) from Sigma Aldrich, St. Louis MO) and white blood cells were washed once with medium (RPMI 1640 (Invitrogen, Carlsbad, CA) supplemented with 5% (v/v) fetal calf serum (FCS) (Mediatech, Manassas, VA), 1% (v/v) Pen-Strep-Glutamine (10.000 U/ml penicillin, 10.000 µg/ml streptomycin, 29.2 mg/ml L-glutamine (Invitrogen) and 50 µM β-mercaptoethanol (Sigma, St. Louis, MO)). For the in vivo cytotoxicity assay, spleen B cells were purified by positive selection with αCD19 conjugated magnetic beads (Miltenyi Biotec, Auburn, CA) according to the manufacturer’s instruction. For enrichment of Vα14i NKT cells from splenocytes and thymocytes, cells were incubated with PE conjugated NK1.1 (PK136) followed by positive selection with αPE magnetic beads (Miltenyi Biotec). Liver associated lymphocytes were used directly.
B16 and B16-CD1d melanoma cells were either loaded with 250 ng/ml αGalCer (37°C, 90 min) or mock treated, washed twice with PBS, and 1 × 105 tumor cells were injected i.v. into C57BL/6 mice as indicated. 14 days after challenge the numbers of metastatic nodules on the lung surface were counted. 400 tumor nodules were established as upper limit for the counting, as at higher densities discrete tumor nodules could no be separated accurately any more.
For staining of cell surface molecules, cells were suspended in staining buffer (PBS, 1% BSA, 0.01% NaN3) and stained with fluorochrome-conjugated Ab at (0.1–1 µg/106 cells) for 15 min in a total volume of 50 µl. FcγR-blocking antibody αCD16/32 (2.4G2) and unconjugated rat IgG (Jackson ImmunoResearch, West Grove, USA) were added to prevent non-specific binding. If biotin-conjugated Ab were used, cell bound Ab was detected with streptavidin conjugates (1:200) in a second incubation step. Staining of T cells with α-galactosylceramide (αGalCer) loaded CD1d tetramers (34) was performed as described previously. In brief, cells were stained with the tetramer together with other surface mAbs in staining buffer at 4°C for 30 min. For analysis of intracellular cytokines, cells were fixed and permeabilized using the Cytofix/Cytoperm reagents (BD Biosciences, San Diego CA) for 10 min at 37°C. Cells were washed twice and incubated for 30 min with fluorochrome-conjugated Ab and unconjugated rat and mouse IgG in Perm/Wash solution (BD Biosciences), which was followed by additional 5 min incubation in Perm/Wash solution without mAb. For in vitro experiments intended for intracellular staining GolgiPlug and GolgiStop (BD Bioscience) were added for the last 4 h of incubation. Cells were analyzed with FACSCalibur, FACSCanto or LSR II (BD Bioscience) and data were processed with CellQuest Pro (BD Bioscience) or Flow Jo (Tree Star Inc., San Carlos, USA) software. Graphs derived from digital data are displayed on a ‘logical scale’ (35).
In vivo cytotoxicity assays were performed according to (31) with minor alterations. Splenic B cells were purified as described and either pulsed with the indicated glycolipids (250 ng/ml, 1 h at 37°C) and labeled with a high concentration of CFDA-SE (1 µM, 15 min at 37°C; CFSEhigh cells) or were mock treated and labeled with a low concentration of CFDA-SE (0.15 µM; CFSElow cells). Cells were washed three times with PBS and equal numbers of cells from each population were injected intravenously (total of 1 × 107 target cells). Animals were sacrificed after indicated times and the presence of target cells in spleen and liver was determined by flow cytometry. To calculate specific lysis of the in vivo cytotoxicity assay, the following formula was used: percentage specific cytotoxicity = 100 - (100 × (CFSEhigh/CFSElow)C57BL/6 / (CFSEhigh/CFSElow)iNKT cell deficient). For in vitro cytotoxicity assays the tumor cells were either pulsed with αGalCer (250 ng/ml, 1 h at 37°C) or were mock treated, mixed at equal ratio and 1 × 105 cells were incubated with iNKT cell enriched lymphocytes at indicated ratio for 4 h. Tumor cells utilized were either A20 and A20-CD1d or RMA together with B16 or B16-CD1d. Later pairings were labeled with CFDA-SE (0.15 µM; 15 min at 37°C). A20/A20-CD1d targets cells were distinguished based on FSC/SSC characteristics and CD1d expression. RMA/B16 and RMA/B16-CD1d targets cells were distinguished based on FSC/SSC characteristics and TCRb expression.
Results are expressed as mean ± standard error of the mean (SEM). Comparisons were drawn using a two-tailed Student t-test (Excel, Microsoft Corporation, Redmond, USA) or ANOVA test (GraphPad Prism, GraphPad Software, San Diego, CA). Each experiment was repeated at least twice. Graphs were generated with GraphPad Prism (GraphPad Software).
We applied an in vivo cytotoxicity assay that has been used to study conventional CD8+ T cells (31) and tgNKT cells (28) for the investigation of the antigen-specific cytotoxicity of iNKT cells. To generate targets, whole splenocytes were loaded in vitro with αGalCer or incubated with medium as a control, differentially labeled with CFSE, injected i.v. and 16 h later the cytotoxicity was analyzed. We detected antigen-specific cytotoxicity of 30.4% +/− 1.7% against the αGalCer-loaded splenocytes when compared to control targets (Fig.1A and B). As reported previously (29), this in vivo cytotoxicity was dependent on iNKT cells, as it was not detected in mice deficient for Jα18 (Fig. 1A) or CD1d (data not shown). Using congenic and surface markers, we then dissected the different splenocyte subsets to determine if there was preferential elimination of certain cell types by iNKT cells in vivo. As shown in figure 1B, the in vivo cytotoxicity of iNKT cells varied significantly for different cell populations, ranging from no cytotoxicity against NK cells (0.4% +/− 2.8%) to almost complete elimination of marginal zone B cells (87.3% +/− 0.3%). Based on the hypothesis that cytotoxicity is likely influenced by the amount of antigen presented on the cell surface of the target cell, and because it is known that marginal zone B cells have very high expression levels of surface CD1d (36, 37), we analyzed CD1d expression by different splenic cell populations. The CD1d expression levels indeed varied between the different lymphoid and myeloid populations (Fig. 1C), and they correlated directly with the observed in vivo cytotoxicity by iNKT cells (Fig. 1D, R2 = 0.912). Purified CD19+ B cells were thereafter used as target cells to maximize cytotoxicity, and consequently the sensitivity of the in vivo assay. This optimization based on target cell type allowed us to detect in vivo cytotoxicity of 40%–60% within four hours of incubation in vivo.
The highest proportion of iNKT cells within the lymphocyte compartment is found in the liver (4, 38, 39), and liver iNKT cells have been reported to be more effective in the in vivo response to a methylcholanthrene induced sarcoma (40). Therefore we were interested to compare the in vivo cytotoxicity of splenic versus hepatic iNKT cells. Interestingly, when we analyzed the antigen-specific iNKT cell cytotoxicity against αGalCer-loaded B cells in the liver we obtained values that were either very low or in most cases negative (Fig. 2A and data not shown). The local immune system of the liver has previously been shown to actively promote tolerance rather than immunity (41, 42). Therefore it could be argued that the tolerogenic environment of the liver was blocking the cytotoxicity of the liver iNKT cells in situ. However, due to the measured negative values of the in vivo cytotoxicity assay, we hypothesized that the αGalCer loaded target cells might be trapped in the liver, thus distorting the ratio of αGalCer loaded to unloaded B cell targets. Therefore we analyzed the ratio of both target populations in liver and spleen over time. As expected the ratio of antigen loaded target cells to control target cells decreased continuously in the spleen (Fig. 2B), in line with the increased cytotoxic removal of antigen-loaded target cells over time. By contrast, this ratio increased rapidly in the liver and was followed by an antigen-specific reduction (Fig. 2B). This observation is in agreement with the notion that αGalCer loaded B cell targets accumulate in the liver in an antigen and iNKT cell dependent manner during at least the first hour in vivo after injection. This accumulation in the liver cannot account for the decrease in the spleen, as the liver has many fewer targets cells. We estimate that if all of the approximately 50% increase in αGalCer loaded target B cells in the liver were due to cells from spleen, this would contribute less than 5–10% of the total estimated cytotoxicity. Therefore the decrease in the spleen likely reflects mostly cytotoxicity there. Consequently, we concluded that the accumulation of αGalCer loaded B cell targets in the liver exceeds the capacity of the normal cytotoxic function of the liver resident iNKT cells at early time points. Nevertheless the cytotoxic capacity of spleen and liver resident iNKT cells appears similar.
We next analyzed the activation of the iNKT cells by αGalCer loaded B cell targets. As shown in figure 3, the αGalCer loaded B cell targets stimulated iNKT cells in spleen and liver to up-regulate the activation markers CD69 and CD25 after four hours. This up-regulation was seen as early as two hours after the injection of the B cell targets (data not shown). Furthermore, the iNKT cells produced the cytokines IFNγ, TNFα and IL-4 within four hours after the injection of the target cells (Fig. 3B). These data demonstrate that αGalCer loaded B cell targets activate iNKT cells in the spleen and liver. Like the iNKT effector cells, the B cell targets also were activated, as measured by the up-regulation of CD69 and CD25, both in spleen and liver (Fig. 3C). Together these data indicate that liver iNKT cells do not have an activation defect during the in vivo cytotoxicity assay.
To compare the intrinsic cytotoxicity of iNKT cells derived from different organs side by side, we used a flow cytometric assay similar in principle to the in vivo assay. A20 B lymphoma cells and A20 cells transfected with a construct driving the expression of mouse CD1d (A20-CD1d) were used as targets. Although A20 cells are H2d haplotype, we could not detect in vitro NK type killing during the four hours incubation time with H2b iNKT cells (data not shown). The A20 cell line does not express detectable levels of CD1d and is not killed by iNKT cells when loaded with αGalCer (data not shown). As A20 and A20-CD1d displayed identical growth within 48h, both cell lines could be used as targets without the need of prior CFSE labeling. Target cells were identified by size (FSC vs SSC) and CD19 expression and differentiated based on their CD1d expression levels. Using this protocol we assessed the antigen-specific cytotoxicity of iNKT cells from thymus, spleen and liver. In all experiments the cytotoxicity of iNKT cells from all three organs where comparable (Fig. 4), demonstrating that iNKT cells have similar cytotoxic capability irrespective of the organ of their origin.
We next analyzed the antigen-specific cytotoxicity of iNKT cell for CD1d-presented glycosphingolipid Ags with different antigenic potencies. We compared iNKT cell cytotoxicity against αGalCer loaded targets with two of its synthetic derivatives, OCH and C-Glycoside (C-Gly) (43). OCH is a weaker iNKT cell antigen than αGalCer and induces a systemic Th2 response (43–46). C-Gly is the weakest of these three antigens and induces a systemic Th1 response (43). As shown in figure 5 the cytotoxicity of iNKT cells in vivo directly correlated with the potency of the iNKT cell antigen (αGalCer > OCH > C-Gly). While we could not detect any cytotoxicity against B cell targets loaded with C-Gly, the cytotoxicity against OCH loaded targets reached 22% – 35% of the cytotoxicity observed with αGalCer (Fig. 5A). Furthermore, the in vivo cytotoxicity correlated with the intensity of CD25 up-regulation on the antigen loaded B cell targets (Fig. 5B) and the intensity of the observed iNKT cell stimulation, as measured by up-regulation of activation markers and cytokine production (data not shown). We also tested the in vivo iNKT cell cytotoxicity against B cells loaded with a synthetic version of the Sphingobium yanoikuyae glycolipid GalA-GSL (called GSL-1’ in earlier publications (47, 48)). The iNKT cell cytotoxicity against the GalA-GSL loaded targets was low, reaching only 3%–9% of the cytotoxicity observed with αGalCer loaded targets (Fig. 5C). Furthermore we did not observe up-regulation of CD25 on the GalA-GSL loaded B cell targets (data not shown).
To gain insight into the antigen-specific cytotoxic mechanisms employed by iNKT cells in vivo, we analyzed perforin and CD95 deficient hosts. As shown in figure 6, the lack of perforin had no significant effect on the observed in vivo iNKT cell cytotoxicity, either alone or in conjunction with the lack of CD95 on the targets cells (Fig. 6A and B). By contrast, deficiency for either CD178 in the host (Fig. 6A) or of CD95 on the B cell targets (Fig. 6B), resulted in a significant decrease in the observed antigen-specific in vivo iNKT cell cytotoxicity. The remaining cytotoxicity in the absence of the CD95/CD178 pathway was between 17.6% and 28.5%, with a mean 23.4% of the C57BL/6 value, when normalized to 100%. In accordance with the results of the in vivo cytotoxicity assay, we observed up-regulation of CD178 on iNKT cells stimulated with αGalCer loaded B cells (Fig. 6C) and the up-regulation of CD95 expression by the B cell targets from the same mice (Fig. 6D). Furthermore, we could not detect granzyme B or perforin by either stimulated or control iNKT cells (data not shown). Altogether, these data demonstrate that the in vivo cytotoxicity of iNKT cells depends mainly on the CD95/CD178 interaction.
Given the strong cytotoxicity we observed against CD1d expressing targets, we determined if this direct antigen-specific cytotoxicity could be utilized for anti-tumor responses. The B16 melanoma aggressively grows in vivo, but protection can be conferred by simultaneous stimulation of iNKT cells via IL-12 (10, 12, 49) or αGalCer (17, 19, 26, 27). However the B16 melanoma does not express CD1d ((50, 51) and data not shown), excluding a direct presentation of αGalCer. We stably transfected B16 melanoma cells with mouse CD1d (B16-CD1d) and used them as targets in an in vivo cytotoxicity assay. B16 or B16-CD1d cells were loaded with αGalCer, mixed 1:1 with RMA cells as internal control, labeled with CFSE and incubated for four hours with iNKT cell enriched splenocytes in vitro. Whereas no cytotoxicity was detected against the control B16 melanoma (data not shown), the αGalCer loaded B16-CD1d cells were eliminated (Fig. 7A). We also observed an up-regulation of CD95 on the B16-CD1d melanomas (Fig. 7B), but not on the B16 (data not shown), during the four hours incubation with the splenocytes. To determine the CD95/CD178 dependency of the observed cytotoxicity we inhibited this death pathway using a blocking αCD178Ab. This resulted in a decrease of the cytotoxicity against the B16-CD1d cells of 41% – 59% (Fig. 7A), indicating that the CD95/CD178 pathway is important for the observed cytotoxicity. We next determined, whether expression of CD1d on the B16 melanoma could also confer protection in vivo. B16 and B16-CD1d melanomas were pulsed shortly (90 min) with αGalCer before i.v. injection and the number of formed tumor metastases in the lung were recorded 14 days later. It has previously been shown that loading of B16 cells in vitro with αGalCer for two days protects mice from tumor metastases (51, 52). In contrast αGalCer loading for only 1.5 hours resulted in a minor protective effect compared to unloaded B16 tumor cells (Fig. 7C). By contrast αGalCer loading of B16-CD1d melanomas protected the animals from metastases (Fig. 7C). These data indicate that direct antigen presentation by the B16-CD1d melanoma leads to an efficient eradication of the tumor in vivo.
While it is well known that iNKT cells can augment the cytotoxic activity of NK cells, here we have analyzed the cellular and molecular parameters involved in the in vivo and in vitro antigen-dependent cytotoxicity of targets cells by iNKT cells. In particular, we provide evidence that the in vivo cytotoxicity of iNKT cells correlates directly with CD1d expression levels on target cells and with the potency of the iNKT cell antigen. The interaction of iNKT cells with their targets led to a mutual activation of the both cell types and the iNKT cells from spleen and liver showed a similar degree of cytotoxicity. Most surprising, we show that the antigen-specific cytotoxicity of iNKT cells in vivo relies almost entirely on the interaction between CD95 (Fas) and CD178 (FasL). It has been reported for αGalCer and its derivatives OCH and C-Gly that the intensity of cytokine produced by the iNKT cells following i.v. injection correlated directly with the antigenic potency, i.e. αGalCer > OCH > C-Gly (43). Utilizing αGalCer, OCH, C-Gly, and a synthetic version of the Sphingobium yanoikuyae derived Ag GalA-GSL, we demonstrate that the antigenic strength directly correlated in a similar fashion with the iNKT cell in vivo cytotoxicity (Fig. 5). Surprisingly, limited cytotoxicity was observed with the weaker Ags, including C-glycoside and the Sphingomonas yanoikuyae Ag GalA-GSL. These data suggest that the described in vivo potency of C-glycoside in tumor rejection (53) or the anti-malarial response (54) may not be due principally to antigenic specific killing by iNKT cells. Similarly, while mice that have iNKT cells clear S. yanoikuyae more rapidly when compared to iNKT cell deficient animals, this also may occur independently of an antigen-specific cytotoxic function (48, 55). Our experiments focused on testing the immediate or short term cytotoxic function of iNKT cells, and therefore we cannot rule out the possibility that iNKT cell mediated Ag specific killing is important in immune responses that take place over several days like during tumor or pathogen challenges.
Furthermore, under different circumstances, some of the less potent compounds may stimulate cytotoxic activity more efficiently. For example, DC may be more effective at taking up C-glycoside and loading it into CD1d than B lymphocytes, and similarly, uptake of whole Sphingomonas bacteria or membrane fragments may promote more effective loading of GalA-GSL into the groove of mouse CD1d. In addition the overall avidity of an antigen depends not only on its strength of interaction with the TCR when bound to CD1d, i.e. the TCR affinity, but also on its density, i.e. on the amount of available epitopes presented and recognized by the T cell. In line with this notion, we show that the iNKT cell cytotoxicity in vivo also correlated with the amount of CD1d expressed on the target cells (Fig. 1). Although NK cells expressed lower levels of CD1d than any other cell type analyzed (Fig. 1C), it was still surprising that we did not observe αGalCer induced cytotoxicity against splenic NK cell targets (Fig. 1B). The expression of CD95 induced on the NK cells was lower than on B cells (data not shown), which together with the low amount of CD1d, may explain the reduced NK cell susceptibility to killing. Furthermore, several mechanisms have been reported that protect cytotoxic lymphocytes against cytotoxicity (56–58), which could be involved in our experimental system as well.
The two major mechanisms to induce cell death by cytotoxic lymphocytes involve on the one hand different surface receptors, most prominently CD95/CD178, but also TNFα/TNF-R and TRAIL/TRAIL-R, or alternatively, soluble mediators, mainly perforin and granzymes (59, 60). iNKT cell express CD178 following activation with Concanavalin A (Con A) (61), and αGalCer (Fig. 6 and (17, 62)). Furthermore, human iNKT cell lines have been reported to express TRAIL (CD253) following restimulation with αGalCer loaded DCs and they exerted TRAIL-dependent cytotoxic activity against some leukemia cells in vitro (62, 63). Despite this diversity of mechanisms, most studies have implicated the perforin/granzyme B pathway in the cytotoxic activity of iNKT cells (19, 64–66). However, these studies relied on chemical inhibitors of granule release, and most likely addressed NK-type cytotoxicity rather than antigen-specific CD1d dependent cytotoxicity. A typical experimental set-up for these studies involved the injection of αGalCer i.v., and the determination of the cytotoxicity of purified splenocytes one day later against NK cell sensitive targets in vitro. Such an approach does not directly address the mechanisms or even the role of iNKT cells in the observed cytotoxicity. Indeed, in several studies it was shown that the perforin required for the observed cytotoxicity against tumors resided not within the iNKT cells, but rather in NK cells that had been activated down-stream of the iNKT cell activation (25, 67, 68). In fact, this trans-activation of NK cells appears to be the general mechanism for the αGalCer induced anti-tumor activity. Following αGalCer exposure, NK cells are activated by several mechanisms, including iNKT cell derived IFNγ (in mice (27, 67, 69–71)) or IL-2 (in humans (24)), leading to NK cell cytotoxicity and cytokine production. The importance of this trans-activation of NK cells by iNKT cell derived IFNγ has been demonstrated for pathogen infections (39, 72) and in tumor models (17, 27, 67, 73). Apparently, NK cells act generally in amplifying the iNKT cell signal in a feed forward loop (14, 49, 74).
By contrast with these earlier studies, our experiments were designed to investigate the direct, short-term, CD1d-dependent, antigen-specific cytotoxicity of iNKT cells. CD1d dependent cytotoxicity of iNKT cells has been reported previously (2, 24, 29, 30, 65, 75–77), but the underlying mechanism was not addressed. Here we demonstrate that the CD1d and antigen-dependent cytotoxicity of iNKT cells requires the CD95/CD178 pathway and is completely independent of the perforin/granzyme pathway (Fig. 6). iNKT cells expressed CD178 following interaction with αGalCer loaded B cell targets and induced CD95 expression on these target B cells (Fig. 6). The remaining, relatively minor cytotoxicity observed in the absence of CD95 and perforin expression could be due to TRAIL (CD253) and/or TNFα mediated apoptotic pathways. However, we could not detect any TRAIL staining (clone N2B2) on iNKT cells by flow cytometry in our experiments (data not shown). Our data are in agreement with earlier studies showing an iNKT cell-mediated and CD95/CD178 pathway dependent elimination of B cells during nickel tolerance (78) and that of hepatotcytes following ConA-induced hepatitis (61) depends on both. However, the requirements for expression of CD95 by iNKT cells, and the role of CD1d and TCR recognition, were not addressed in these studies. Numerous studies have established the strong anti-tumor activity following the activation of iNKT cells with IL-12 (7–14) or αGalCer (1, 2, 15–27). Importantly, NKT cells were also shown to be involved in the un-manipulated tumor surveillance (25, 68, 79, 80). It has been demonstrated previously that loading of αGalCer onto B16 and other tumor lines can induce a strong immune response and protection against tumor growth in vivo (51, 52). However, in these studies the loading of αGalCer was carried out for two days, which may allow the tumor cells to take up high amounts of this glycolipid Ag (51).
Consequently, in this experimental format, the responses of iNKT cells (51) and DCs (52), after injection of either αGalCer-loaded B16 or αGalCer-loaded B16-CD1d cells, were indistinguishable. This indicates that direct presentation of αGalCer by the tumor cells was not required, suggesting that potent cross-presentation of αGalCer can occur in vivo. As mentioned above the anti-tumor response following αGalCer treatment in vivo depends on IFNγ mediated trans-activation of NK cells (17, 67, 81). Furthermore, it has been shown that IFNγ induces up-regulation of CD95 on B16 tumor cells (82, 83). Therefore the αGalCer induced anti-tumor response for most of these studies can be explained by the trans-activation of NK cells, which then attack the tumor. In line with this interpretation is the observation that protocols that augment the IFNγ response by iNKT cells, thereby intensifying the NK cell trans-activation, augment the anti-tumor response. This could be achieved either by using DCs loaded with αGalCer instead of free αGalCer (22, 84), or by utilizing the Th1 cytokine-inducing iNKT cell antigen C-Gly (85).
By contrast, we show here that expression of CD1d by B16 melanoma cells makes them susceptible to cognate antigen-dependent cytotoxicity by iNKT cells, mainly via the CD95/CD178 pathway. Furthermore, we demonstrate that following a short pulsing of B16 tumor cells with αGalCer, the efficiency of the anti-tumor response depends on the expression of CD1d and most probably on direct presentation of the antigen by the tumor cells (Fig. 7). Several studies have shown that tumor cells that express CD1d can be lysed by iNKT cells in vitro when the tumor cells were loaded with αGalCer (24, 30, 65, 76, 77). Here we extended these reports by demonstrating the CD1d dependency of the antigen-specific iNKT cell cytotoxicity against tumors in vivo (Fig. 7). Therefore, our data suggest that activation of iNKT cells for an anti-tumor therapy, a strategy currently applied in clinical trials (86), might be most effective against those tumors that express CD1d. In such cases, the activation of iNKT cells might facilitate cognate-Ag killing as well as trans activation of other cell types.
The authors wish to thank Archana Khurana for excellent technical assistance. We are grateful to the scientific contributions of Barbara Sullivan, Bo Pei, Aaron Tyznik and Jennifer Matsuda.
Funding: This work was funded by NIH grants RO1 AI45053 and R37 AI71922 (M.K.), an Outgoing International Fellowship by the Marie Curie Actions (G.W.) and by a long-term EMBO fellowship and a fellowship from the Swiss National Science Foundation (P.K.).
The authors have no competing interests regarding this work.