Differential antitumor immunity mediated by NKT cell subsets in vivo

We reported that NKT cell–deficient TCR Jα18^−/− mice were more susceptible to the development and growth of MCA-induced sarcomas than WT mice (23). We also demonstrated that protection against sarcoma growth was restored to TCR Jα18^−/− mice upon adoptive transfer of liver-derived WT NKT cells (24) (Fig. 1 A). Liver-derived NKT cells were used in these studies because of their relative abundance, and it was assumed that these cells were functionally representative of all NKT cell populations in vivo. To test this hypothesis directly, NKT cells were isolated from the liver, spleen, and thymus of WT mice and adoptively transferred into MCA-1–injected TCR Jα18^−/− recipients. Consistent with our earlier report (24), transfer of 2.5 × 10⁵ liver-derived NKT cells into MCA-1–inoculated TCR Jα18^−/− mice induced complete inhibition of tumor growth (Fig. 1 A). In contrast, the equivalent number of spleen- or thymus-derived NKT cells from the same donor mice failed to inhibit tumor growth (Fig. 1, B and C). In fact, even twice as many (5 × 10⁵) thymus-derived NKT cells were unable to restore protection (Fig. 1 C), which suggests that liver NKT cells have unique functional characteristics that are necessary for efficient rejection of the MCA-1 sarcoma.

Given that liver-derived NKT cells are unique in their ability to mediate MCA-1 rejection, we sought to determine whether CD4⁺ and CD4⁻ subsets of liver-derived NKT cells possessed this capacity. Each subset was sorted and adoptively transferred into MCA-1–inoculated TCR Jα18^−/− recipients. As shown in Fig. 2 A, CD4⁻ NKT cells completely restored protection against MCA-1, whereas the equivalent number of CD4⁺ NKT cells did not. At best, CD4⁺ NKT cells slowed tumor growth when compared with PBS-injected controls. Considering that CD4⁻ cells make up ∼30% of the liver-derived NKT cell population, and that the antitumor effects resided primarily within this compartment, it was hypothesized that a lower dose of these cells than that established for unfractionated liver-derived NKT cells (2.5 × 10⁵) (24), would still restore complete protection. MCA-1–inoculated TCR Jα18^−/− recipients were injected with a range of CD4⁻ doses (2.5 × 10⁵, 10⁵, 5 × 10⁴ cells), and were monitored for tumor development (Fig. 2 B). Restoration of tumor rejection was dose dependent; 2.5 × 10⁵ and 10⁵ CD4⁻ cells completely inhibited tumor growth, while 5 × 10⁴ cells considerably delayed tumor onset and progression. Given that CD4⁻ NKT cells represent ∼30% of total NKT cells in liver (unpublished data), the number of CD4⁻ cells required for complete tumor inhibition (10⁵) was similar to the number of these cells (>0.8 × 10⁵) present in the minimum dose of unfractionated liver-derived NKT cells (2.5 × 10⁵) necessary for complete protection (24). This supports the concept that the CD4⁻ subset is mostly, if not completely, responsible for the antitumor effects of liver-derived NKT cells.

One explanation for the lack of protection mediated by thymus-derived NKT cells is that they might not survive after transfer into the periphery. To address this possibility, NKT cell–enriched thymocytes were carboxyfluorescein diacetate succinimidyl ester (CFSE)–labeled and transferred into WT recipients. 6 d later, organs were harvested and analyzed for the presence of CFSE⁺ CD1d/α-GalCer tetramer⁺ cells. These cells were clearly detected in liver (Fig. 3 A), and to a lesser extent in spleen, PBLs, and BM (not depicted) for at least a week after transfer. In other experiments, we could detect clear populations of thymus-derived NKT cells 8 weeks after adoptive transfer (unpublished data). Transferred liver- and spleen-derived NKT cells were also detectable in these tissues (albeit at a lower rate relative to the number of NKT cells injected, which was consistent with a previous report [31]) with a similar tissue distribution (unpublished data). Another explanation for the inability of CD4⁺ NKT cells to confer complete tumor inhibition is that CD4 labeling for FACS sorting may induce the preferential depletion of this subset in vivo. To address this possibility, mice received CFSE-labeled NKT cells that were, or were not, labeled with anti-CD4 before transfer. As shown in Fig. 3 B, prelabeled CD4⁺ NKT cells were detected at the same frequency as nonprelabeled CD4⁺ NKT cells 6 d after transfer, indicating that CD4 labeling did not affect survival of these cells.

One possible explanation for the differential responsiveness of NKT cell subsets in these studies is that liver CD4⁻ NKT cells are activated selectively in response to the MCA-1 tumor. Therefore, we used a model in which all NKT cells are stimulated to test whether this distinction existed. We previously reported that adoptive transfer of liver-derived NKT cells into α-GalCer–treated, B16F10-inoculated TCR Jα18^−/− recipients inhibited B16F10 metastases in the lungs (25). To determine whether spleen- and thymus-derived NKT cells promoted similar antitumor immunity, unfractionated liver-derived lymphocytes, NKT cell–enriched thymocytes, and splenocytes were adoptively transferred into α-GalCer–treated, B16F10-inoculated, TCR Jα18^−/− mice. Each transferred population contained 10⁵ NKT cells, which we determined previously to be a protective dose (25). As expected, liver-derived NKT cells effectively inhibited B16F10 metastases in the lungs (Fig. 4 A). However, consistent with earlier findings from the MCA-1 model, spleen-derived NKT cells were less effective than were liver-derived NKT cells at reducing the number of metastases. Even more striking was the inability of NKT cell–enriched thymocytes or sorted thymus-derived NKT cells to inhibit tumor metastases, even when a fivefold increase in NKT cell dosage (5 × 10⁵) was used (Fig. 4 A). In some experiments, NKT cells were purified by FACS sorting based on NK1.1 versus αβTCR expression from liver, spleen, and thymus, with similar results.

To determine whether CD4⁺ and CD4⁻ subsets of liver-derived NKT cells were capable of mediating α-GalCer–induced antitumor immunity, each subset was sorted and adoptively transferred into α-GalCer–treated, B16F10-inoculated, TCR Jα18^−/− recipients (Fig. 4 B). Whereas both subsets promoted α-GalCer–induced antitumor immunity, this was mediated more potently by the CD4⁻ subset. Significant (P < 0.01) antitumor activity was observed after transfer of 2.5 × 10⁴ CD4⁻ NKT cells, whereas equivalent numbers of CD4⁺ NKT cells had little effect. For comparison, we also injected some mice with thymic NKT cells that had been subdivided into CD4⁺ and CD4⁻ subsets (Fig. 4 B). As for unfractionated thymic NKT cells, neither CD4⁺ nor CD4⁻ subsets showed enhanced protection.

We demonstrated previously that NKT cell–derived IFN-γ is critical for tumor immunosurveillance against MCA-1 and α-GalCer–induced antitumor immunity against B16F10 (24, 25). To investigate whether differences in IFN-γ production were responsible for the functional differences that were observed, WT mice were injected with α-GalCer, and CD4⁺ and CD4⁻ NKT cell subsets from the liver and spleen were analyzed for their production of intracellular cytokines. As shown in Fig. 5 A, no clear differences were observed; all populations produced similar levels of IFN-γ and IL-4 after α-GalCer stimulation. Because thymic NKT cells are not stimulated by α-GalCer in situ (16, 32), thymus- and liver-derived NKT cells were isolated from naive mice and stimulated in vitro with sorted spleen-derived DCs and 100 ng/ml α-GalCer for 12 h (Fig. 5 B). Intracellular cytokine staining of these cells demonstrated that IFN-γ and IL-4 production was similar between CD4⁺ and CD4⁻, subsets from the thymus and liver.

To investigate whether IL-4 and/or IL-10 production antagonizes the potential of some NKT cell subsets to confer protection, NKT cells were enriched or purified from the thymus and liver of WT and IL-4^−/− mice and adoptively transferred into α-GalCer–treated, B16F10-inoculated, TCR Jα18^−/− mice, some of which also were treated with a neutralizing anti–IL-10 mAb. The activity of this batch of mAb was confirmed by using an inhibition ELISA (unpublished data) at two to four times the amount (400 μg/mouse on days 0, 4, and 8) known to effectively block IL-10 signaling in other in vivo models (33–35). As shown in Fig. 6 A, IL-4^−/− liver-derived NKT cells were significantly (P < 0.05) more protective than were the same number of WT liver-derived NKT cells. Furthermore, IL-4^−/− thymus-derived NKT cells also were able to protect in this model, in sharp contrast to WT thymus-derived NKT cells. Neutralizing anti–IL-10 had little additional effect. One possible explanation for the enhanced tumor immunity after transfer of IL-4^−/− cells was that in the absence of IL-4 production, NKT cell–derived IFN-γ production was enhanced. We tested this by comparing IFN-γ production by NKT cells in IL-4^−/− and WT mice after α-GalCer administration in vivo. As shown in Fig. 6 B, IFN-γ production was comparable between CD4⁺ and CD4⁻ NKT cell subsets from each strain 2 and 48 h after stimulation. Taken together, this suggests that thymus-derived NKT cells are capable of responding to the tumor, but are unable to do so because IL-4 production antagonizes their ability to mediate IFN-γ–dependent tumor rejection. How liver-derived NKT cells inhibit tumor growth—despite similar levels of IL-4 production to thymus-derived NKT cells—is unclear but might relate to other (undefined) cytokines that are expressed differentially between these NKT cell subsets, or by differences in their ability to transactivate NK cells and DCs. This latter possibility was investigated using in vitro cultures in which NKT cells from thymus or liver of WT mice were cocultured for 12 h, in the presence of α-GalCer, with liver mononuclear cells from Jα18^−/− mice (Fig. S1, available). This induced IFN-γ production by NK cells, enhanced CD69 expression by DCs and NK cells, and increased the percentage of CD86⁺ DCs. However, we found no obvious difference in the ability of thymus- and liver-derived NKT cells in this assay that would explain the differential antitumor immunity observed.

We previously reported that adoptive transfer of liver-derived NKT cells restores antitumor immunity to TCR Jα18^−/− mice (24, 25). These and other studies presumed that liver-derived NKT cells were a functionally homogenous population that was representative of all NKT cells. However, our current study demonstrates that liver NKT cells, and primarily the CD4⁻ subset, are unique in their ability to confer protective antitumor responses. It is not clear why spleen- and thymus-derived NKT cells do not confer similar protective responses, but there are several plausible explanations. We considered the possibility that thymus-derived NKT cells are functionally less mature than NKT cell populations that are found in the periphery. Yet our data, and that of other investigators, clearly demonstrate that thymus-derived NKT cells are potent cytokine producers after in vitro stimulation (15, 17, 36–39), and that these cells survive in the periphery after adoptive transfer (31). Furthermore, adoptive transfer of thymus- (40) or spleen-derived (41) NKT cells into NOD mice conferred IL-4– and/or IL-10–dependent suppression of type 1 diabetes, which demonstrated that these cells can mediate effector functions in vivo. Spleen-derived NKT cells are also clearly functional because they can mediate anterior chamber-associated immune deviation in an IL-10–dependent manner (42), and can promote airway hyperresponsiveness by way of IL-4 and -13 production (43); this further supports the idea that they are not functionally immature. Clearly, thymus- and spleen-derived NKT cells can influence immune responses upon adoptive transfer in vivo. This raises the possibility that these cells are specifically less efficient at mediating tumor rejection compared with liver-derived NKT cells, and by definition, should be regarded as a functionally distinct subset. This has interesting implications for other studies that involve the adoptive transfer of liver-derived NKT cells (e.g., references 44, 45), and raises several important questions, such as whether the effects that are mediated by NKT cells in these studies are also influenced by the NKT cell tissue of origin. Considering that these responses appeared to be dependent on NKT cell–derived IL-4 (i.e., the opposite of our current study), this raises the question of whether these functional roles also are unique to liver NKT cells, and whether CD4⁺ and CD4⁻ subsets of liver NKT cells contribute equally to the various diseases in which NKT cells are known to play a role.

The ability of IL-4^−/− thymus-derived NKT cells to restore α-GalCer–induced antitumor immunity supports the idea that these cells are functionally active upon adoptive transfer, and that they can recognize and respond to the tumor, but that the production of IL-4 antagonizes their ability to promote effective tumor rejection. These findings are reminiscent of observations in a model of experimental autoimmune encephalomyelitis, in which the Th1-type disease was heightened in IL-4^−/− mice and was inhibited in IFN-γ^−/− mice after α-GalCer treatment (46).

This study is the first in vivo demonstration of functionally distinct NKT cell subsets in two disease models. In vitro studies have reported differences in cytokine production between CD4⁺ and CD4⁻ NKT cell subsets from mice (15) and humans (28, 30, 47, 48); however, the physiological significance of these results is unclear and does not easily explain the differences in our study. We were unable to determine any clear differences in cytokine production that could explain the functional disparity between CD4⁺ and CD4⁻ liver NKT subsets. Both subsets produced similar levels of IFN-γ and IL-4 by cytokine staining, and were detected in similar locations after adoptive transfer, including at the tumor site (albeit in very low numbers) (unpublished data). Further study of factors that are expressed differentially between NKT cell subpopulations is warranted.

Our results demonstrate the importance of studying distinct NKT cell subsets in the various models in which these cells are involved. Specifically, the tissue of origin (e.g., liver, thymus, or spleen) and the subset within that tissue (e.g., CD4⁺ or CD4⁻) needs to be considered because the use of different subsets may influence results dramatically. It should be emphasized that our study may not prove to be the most complete breakdown of NKT cell subsets, and that additional ways to subdivide NKT cell populations, based on their functional distinction, may become apparent. An attractive hypothesis is that different NKT cell subsets have different functional roles in vivo. Thus, CD4⁻ liver-derived NKT cells may be very efficient at promoting Th1-type cell-mediated immunity, whereas other NKT cell subpopulations, such as thymus-, spleen-, and liver-derived CD4⁺ NKT cells, might be more efficient at conferring immunosuppressive effects. This needs to be tested by comparing the activity of these subsets in other models of NKT cell–dependent immune responses that require a cell-mediated Th1-like or immunosuppressive Th2-like effect. The ability to separate potentially suppressive and/or aggressive NKT cell subsets will be an important step forward for NKT cell research, especially now that these cells are being targeted in α-GalCer–based clinical trials. Global activation of NKT cells by administration of soluble α-GalCer and/or α-GalCer–pulsed DCs could be detrimental because they may induce opposing effects that reduce the efficiency of treatment and/or cause undesirable effects. These could be eliminated by targeting individual NKT cell subsets, which might be achieved by in vitro isolation and stimulation of the desired subset before transfer, or potentially by the targeted suppression of particular subsets in vivo. Future studies should also aim to determine whether NKT cell subsets are preconditioned in their relative environments before transfer, or whether distinct functional capabilities are intrinsic to each subset upon export from the thymus. Thus, the novel findings that are reported in this study are likely to have major implications for future studies that investigate the potential of NKT cells in immunotherapy.

Inbred C57BL/6 WT mice were bred in-house at the Department of Microbiology and Immunology Animal House, Melbourne University, or the Peter MacCallum Cancer Institute Animal House. TCR Jα18^−/− mice (provided by M. Taniguchi, Chiba University Graduate School of Medicine, Chiba, Japan and backcrossed to C57BL/6 for 12 generations) and C57BL/6 IL-4^−/− mice (The Jackson Laboratory) were bred at the Peter MacCallum Cancer Institute. Mice of 6–14 wk of age were used in all experiments that were performed according to animal experimental committee guidelines.

Lymphocytes were isolated from the liver by cutting it into small pieces and gently pressing it through 200-μm mesh sieves into PBS containing 2% FCS (Commonwealth Serum Laboratories). Lymphocytes were separated from hepatocytes and cellular debris by way of a 33% isotonic Percoll density gradient (GE Healthcare) performed at room temperature. Exposed DNA was removed using DNase (35 μg/ml final concentration) (Roche Diagnostics). Lymphocytes were isolated from the thymus, spleen, and LNs by cutting the organs and gently grinding them between two frosted glass slides into PBS containing 2% FCS. BM was harvested by flushing femurs with PBS containing 2% FCS. Blood was collected, by way of cardiac puncture, into 10 ml of PBS to prevent clotting. Liver, spleen, BM, and PBLs were depleted of RBCs using red cell lysis buffer (Sigma-Aldrich).

For thymus-derived NKT cell enrichment, thymocytes were labeled with anti-CD8 (clone 3.155) and anti-CD24 (J11D) (both grown in-house), and bound cells were depleted using rabbit complement (C-six Diagnostics). Clumping of dead cells was avoided by using DNase, and viable cells were collected using a histopaque (1.083 g/ml) gradient (Sigma-Aldrich) performed at room temperature. Cells were washed twice before being surface labeled for flow cytometric sorting. Spleen-derived NKT cells were enriched by B cell depletion. This was achieved in one of two ways in different experiments, either by incubating spleen cells on 90-mm petri dishes that were coated with purified goat anti–mouse Ig (15 μg/ml; Caltag), or anti-B220 labeling followed by magnetic depletion with anti-rat Ig–coated Dynabeads (Dynal). In an alternative enrichment procedure, splenocytes were labeled with α-GalCer–loaded CD1d tetramers (CD1d/α-GalCer) followed by anti-PE microbeads (Miltenyi Biotec). Cells were passed through an automated magnetic cell sorter (autoMACS) enrichment column (Miltenyi Biotec), and collected into PBS containing 2% FCS and 2 mM EDTA (Sigma-Aldrich).

For sorting of NKT cell populations, unfractionated or NKT cell–enriched populations were labeled with combinations of antibodies, including FITC-conjugated anti-αβTCR (clone H57-597), FITC-conjugated anti-CD4 (clone RM4-5), PE-conjugated anti-NK1.1 (clone PK-136), biotinylated anti-CD4 (clone RM4-5), or allophycocyanin-conjugated anti-αβTCR (clone H57-597). For DCs, splenocytes were labeled with PE-conjugated anti-MHC class II (clone M5/114.15.2) and biotinylated anti-CD11c (clone HL3). Biotinylated antibodies were revealed using streptavidin-CyChrome. To avoid nonspecific binding of antibodies to FcR-γ, cells were incubated with anti–mouse CD16/32 (clone 2.4G2) (grown in-house). All flow cytometry reagents were purchased from BD Biosciences unless otherwise indicated. After cell surface labeling, cells were sorted using a FACStar^PLUS (Becton Dickinson) or MoFlow (DakoCytomation). Desired populations were collected into 2% normal mouse serum (NMS) in PBS. A sample of sorted cells was always analyzed to assess the purity of these populations. Sorted NK1.1⁺αβTCR⁺ (NKT cell) populations were always ≥95% pure, and CD4^high (CD4⁺) and CD4^−/low (CD4⁻) NKT cell subsets were always ≥96% pure, unless otherwise indicated.

Culture media contained RPMI 1640 (GIBCO BRL) supplemented with 5% FCS (JRH), 100 U/ml penicillin (GIBCO BRL), 100 μg/ml streptomycin (GIBCO BRL), and 2 mM Glutamax (GIBCO BRL).

For in vivo stimulation, WT or IL-4^−/− mice received 2 μg α-GalCer or vehicle by way of i.p. injection, and were killed 2 or 48 h later. Lymphocytes were isolated from the liver and spleen and cultured for 2 h, without further stimulation, in 5 μg/ml Brefeldin A (Sigma-Aldrich). For in vitro stimulation, NKT cell–enriched thymocytes or unfractionated liver-derived lymphocytes were cultured for 12 h with sorted spleen-derived DCs and 100 ng/ml α-GalCer in culture media (Life Technologies). Brefeldin A was added for the last 2 h of culture. After culture, in vivo and in vitro stimulated cells were labeled with combinations of cell surface antibodies, including FITC-conjugated anti-CD4 (clone RM4-5), CyChrome-conjugated CD1d/α-GalCer tetramer (produced in-house), PerCP-conjugated anti-CD4 (clone RM4-5), and allophycocyanin-conjugated anti-αβTCR (clone H57-597). Following cell surface labeling, cells were washed once in PBS, and fixed in 0.5% paraformaldehyde in PBS (BDH Chemicals) for 30–45 min in the dark at room temperature. Cells were washed twice in PBS and incubated for 1 h in the dark at room temperature with PE-conjugated anti–IFN-γ (clone XMG1.2), anti-IL-4 (clone 11B11), or rat IgG1 (clone R3-34 as an isotype control) in 0.05% saponin (Sigma-Aldrich) in 5% FCS in PBS.

The MCA-1 sarcoma cell line was derived from a TCR Jα18^−/− mouse injected with 100 μg MCA and maintained as described previously (23). For growth of MCA-1, WT and TCR Jα18^−/− mice were injected s.c. in the right hind flank with 10⁵ sarcoma cells. From 1–2 h later, subgroups of TCR Jα18^−/− recipient mice received one of the following populations of cells from WT donors by way of i.v. injection: unfractionated liver-derived lymphocytes; sorted liver-, spleen-, or thymus-derived NK1.1⁺αβTCR⁺ cells; sorted liver-derived CD4⁺ or CD4⁻ NKT cell subsets; or 2% NMS in PBS. Tumor growth was measured every second day with a caliper square as the product of two diameters. Results were recorded as the mean tumor size (cm²) ± SEM.

B16F10 melanoma cells were maintained as described previously (25). WT or TCR Jα18^−/− mice received 5 × 10⁵ B16F10 cells by way of i.v. injection. 3 h later, some TCR Jα18^−/− recipient mice received one of the following populations from WT donors by way of i.v. injection, unless otherwise indicated: unfractionated liver-derived lymphocytes, NKT cell–enriched thymocytes or splenocytes, sorted thymus-derived NK1.1⁺αβTCR⁺ cells, sorted liver- and thymus-derived CD4⁺ or CD4⁻ NKT cell subsets, or 2% NMS in PBS. 1–2 h after lymphocyte transfer, and again on days 4 and 8, mice also received 2 μg α-GalCer (provided by Kirin Breweries Pharmaceutical Research Laboratories, Gunma, Japan, that was prepared in saline supplemented with 0.5% polysorbate-20). For IL-10 neutralization, some mice also were treated with 400 μg anti–IL-10 (clone JES-2A5) or control IgG on days 0, 4, and 8. Mice were killed on day 14, and surface lung metastases were counted with the aid of a dissecting microscope as described previously (49). Data were recorded as the mean number of metastases ± SEM.

NKT cell–enriched thymocytes were washed once in 0.1% BSA in PBS, before being labeled in 1 ml 0.1% BSA in PBS with 4 μl of 1 mM 5-(and-6)-CFSE (Invitrogen) for 10 min at 37°C in the dark. The reaction was quenched with PBS containing 20% FCS before cells were washed twice in RPMI 1640 and once in PBS, and transferred into recipient mice by way of i.v. injection. In some experiments, cells were labeled with anti-CD4 before transfer. Organs (liver, spleen, PBLs, BM, LNs, and thymus) were harvested 6 d after transfer, and screened by flow cytometry for the presence of CFSE⁺ cells. CFSE⁺ cells were analyzed for CD1d/α-GalCer tetramer-binding, αβTCR, and CD4 expression.

Significant differences in tumor growth rate between PBS/vehicle-injected TCR Jα18^−/− control groups and multiple test groups were determined using a Kruskal-Wallis test, followed by a Dunn's post test. Significant differences between vehicle-injected TCR Jα18^−/− control groups and a single test group were determined using a Mann-Whitney U test.

Fig. S1 shows that thymus, spleen, and liver NKT cells induce similar amounts of bystander cell activation.

The authors wish to thank R. Cameron and S. Griffiths for the care and maintenance of mice at the Peter MacCallum Cancer Centre, and D. Taylor and staff at University of Melbourne, Department of Microbiology and Immunology, for their care and maintenance of the mice used in these studies. We also thank K. Field for assistance with flow cytometry and M. Dowling for technical assistance.

D.I. Godfrey and M.J. Smyth were supported by Research Fellowships and a Program grant from the National Health and Medical Research Council of Australia. S.P. Berzins was supported by a Human Frontiers Science Program fellowship. Y. Hayakawa was supported by a Cancer Research Institute Postdoctoral Fellowship. N.Y. Crowe was supported by a Cancer Council of Victoria Postdoctoral Fellowship. J.M. Coquet was supported by a Cancer Research Institute Postgraduate Fellowship. R. Keating was supported by an Australian Postgraduate Award.

The authors have no conflicting financial interests.