A Novel Glycolipid Antigen for NKT Cells That Preferentially Induces IFN-γ Production

Type 1 or invariant NKT (iNKT) cells are a lymphocyte population that is characterized by features of both the innate and adaptive immune responses. The multiple functions of these cells are remarkable in that they have been implicated in allergy, cancer, infection, autoimmunity, and a variety of other conditions (1). Similar to other T lymphocytes, iNKT cells arise from a CD4⁺, CD8⁺ double-positive thymocyte precursor (2). Unlike mainstream T cells, which recognize peptide moieties presented by MHC-encoded molecules, iNKT cells recognize lipid Ags that are often glycosphingolipids (GSLs). Lipid Ags are recognized when they are presented by CD1d, an MHC class I–like molecule (3). The CD1d binding groove is composed of two hydrophobic pockets labeled A′ and F′ (4). GSLs bind with the fatty acid localizing into the A′ groove and the sphingoid base into the F′ groove. This binding mode allows the carbohydrate head group to protrude out of the CD1d molecule, such that it is exposed to be recognized by the iNKT cell TCR (5).

The iNKT cell TCR contains a highly restricted, invariant TCR α-chain that is formed by a Vα14-Jα18 rearrangement in mice and a homologous Vα24-Jα18 (TRAV10-TRAJ18) in humans (6). Although the β-chain of these TCRs is not invariant, it is biased to Vβ8.2, Vβ7, or Vβ2 in mice and Vβ11 (TRBV25-1) in humans, with diverse CDR3 regions.

The prototypical GSL recognized by iNKT cells is α-galactosylceramide (αGalCer) (7, 8). When stimulated by a strong agonist, such as αGalCer, iNKT cells secrete both Th1 and Th2 cytokines (e.g., IFN-γ and IL-4) (9). Stimulation with αGalCer causes long-term changes in the iNKT cell population that originally were likened to anergy (10, 11). We recently showed that stimulation with αGalCer leads to an expansion of an iNKT cell population capable of secreting IL-10, referred to as NKT10 cells (12). Interestingly, subtle chemical/structural alterations in αGalCer were shown to alter the downstream cytokine response, skewing it toward either a Th1 or a Th2 phenotype (13). Generation of an Ag capable of stimulating a strong Th1 cytokine profile has been an area of great interest, because such an Ag would be beneficial for stimulating anticancer responses and for use as a vaccine adjuvant. Based on prior work, a heightened IFN-γ response is due, in large part, to the so-called “trans-activation” of NK cells occurring downstream of iNKT cell stimulation (14–17). Therefore, a strong Th1 response induced by this and some other GSL Ags represents not so much the tendency of an iNKT cell to produce more IFN-γ with reduced IL-4, but the output of a cellular network that involves dendritic cells (DCs) expressing CD1d, iNKT cells to activate the response, and NK cells that are stimulated downstream of iNKT cells that are crucial for continued IFN-γ release.

Although αGalCer was shown to suppress tumor metastases in mouse models (8), it has not been overwhelmingly successful in human trials; this may be due, in part, to the mixed Th1 and Th2 response or the anergy that it induces (18, 19) or to the potential induction of NKT10 (12). Because of this, many analogs of αGalCer have been generated in attempts to elicit a more pronounced Th1-skewed response. C-glycoside, which differs from αGalCer by the replacement of the carbon–oxygen glycosidic linkage with a carbon–carbon bond, was the first GSL Ag reported to have a Th1-polarizing potential (20, 21). C-glycoside cannot stimulate human iNKT cells and, therefore, it is not useful for therapeutic applications.

In this study, we characterized, in detail, the biochemical properties, immune responses, and mechanisms of action of a novel αGalCer analog, DB06-1. This compound is identical to αGalCer with the exception of the replacement of the C2 carbonyl oxygen on the acyl chain with a sulfur atom. DB06-1 was identified in a screen of lipids that associated with detergent-resistant membrane domains, which is characteristic of other Th1-skewing Ags tested (22). In this article, we provide evidence that DB06-1 promoted a Th1-skewed response that was more prominent than that of αGalCer. This correlated with increased loading of this Ag into CD1d, which may be related to a tighter fit in the CD1d groove. However, over the longer term, DB06-1 induced more IL-10–producing NKT10 cells than did αGalCer, suggesting that the Th1 effect of a strong antigenic stimulus may induce a counterregulatory pathway in vivo.

Unless otherwise noted, statistical comparisons were drawn using a two-tailed Student t test.

Stimulation of iNKT cell hybridomas on MicroWell plates coated with soluble mouse CD1d was carried out according to published protocols (23–25). The indicated amounts of compounds or vehicle were incubated for 24 h in MicroWells that had been coated with 1 μg CD1d. After washing, 5 × 10⁴ DN3A4-1.2 (1.2) Vα14 iNKT hybridoma cells were cultured in the plate for 24 h, and IL-2 in the supernatant was measured by sandwich ELISA (R&D Systems), following the manufacturer’s instructions. The 1.2 Vα14 iNKT cell hybridoma expresses a Vα14-Vβ8.2 TCR and was described previously (26).

The culture of bone marrow–derived DCs in GM-CSF was described previously (27). Briefly, cells were isolated from mouse femurs and cultured in media containing GM-CSF for 7 d. The cells were then pulsed with GSL Ags overnight (o/n) and incubated with 5 × 10⁴ 1.2 Vα14 iNKT hybridoma cells for 20–24 h. Similarly, the Ag-presentation assay using A20 B lymphoma cells was described previously (28). Briefly, A20-CD1d transfectants expressing wild-type (WT) CD1d or tail-deleted (TD) CD1d were pulsed with a GSL Ag overnight. APCs (1 × 10⁵/well) were incubated with 5 × 10⁴ iNKT hybridomas for 20–24 h. IL-2 in the supernatant of hybridoma cultures was measured by sandwich ELISA (R&D Systems). Human Vα24⁺ iNKT cells were purified by magnetic enrichment and expanded according to a previously published protocol (29). Briefly, PBMCs were isolated by Percoll (Sigma-Aldrich) density-gradient centrifugation. Human donor PBMCs (1–1.5 × 10⁶/ml) were cultured in RPMI 1640 (Invitrogen) supplemented with 10% (v/v) FBS and 1% (v/v) Pen-Strep-Glutamine (10,000 U/ml penicillin, 10,000 g/ml streptomycin, 29.2 mg/ml l-glutamine; Invitrogen). Human iNKT cell cultures were expanded by weekly restimulation with αGalCer-pulsed irradiated PBMCs and recombinant human IL-2. Ag-pulsed PBMCs (1 × 10⁵/ well) were seeded in 96-well plates and cultured in the presence of 5 × 10⁴ Vα24⁺ human iNKT cells for 20–24 h. GM-CSF release, as a marker of iNKT cell activation, was measured by sandwich ELISA (R&D Systems).

C57BL/6 mice were purchased from The Jackson Laboratory. CD1-TD mice were generously provided by the laboratory of Dr. Albert Bendelac (University of Chicago, Chicago, IL) (30). Cd1d^f/f mice were generated in the laboratory using conventional strategies and crossed with a CD11c- Cre–transgenic mouse obtained from The Jackson Laboratory (31). IL-12^−/− (strain B6.129S1-Il12a^tm1Jm/J) mice also were purchased from The Jackson Laboratory. All mice were housed in specific pathogen–free conditions, and the experiments were approved by the Institutional Animal Care and Use Committee of the La Jolla Institute for Allergy and Immunology. Humanized mice (hCD1d-Vα24 Tg) (X. Wen, S. Kim, R. Xiong, M. Li, A. Lawrenczyk, X. Huang, S.-Y. Chen, P. Rao, G.S. Besra, P. Dellabona, G. Casorati, S.A. Porcelli, O. Akbari, M.A. Exley, and W. Yuan, submitted for publication) were generated by crossing a human CD1d knockin mouse line (32) and a human Vα24 TCR mouse line (33) and were tested in the laboratory of W.Y. following the Animal Care and Use Guidelines of the University of Southern California. Mice were injected i.v. with 1–4 μg DB06-1 or αGalCer (positive control), and blood serum or spleens were harvested 2, 6, 22, or 24 h later or, in the case of NKT10 staining, 4 wk later. Standard sandwich ELISAs for mouse IFN-γ, IL-12p70, and IL-4 (R&D Systems) were performed to measure cytokines in the sera.

Single-cell suspensions of splenocytes were generated as described previously (34). For DC isolation, the tissue was diced into 1-mm pieces and digested using spleen dissociation media (Stem Cell Technologies), and DCs were enriched by positive selection using a CD11c⁺ isolation kit (Stem Cell Technologies) or MACS, according to the manufacturer’s protocols (Miltenyi Biotec). Isolated DCs were cocultured at various concentrations with 5 × 10⁴ 1.2 Vα14 iNKT hybridoma cells o/n, and activation was measured by sandwich ELISA of culture supernatants for IL-2.

For IFN-γ intracellular cytokine staining (ICCS), the cells were cultured in media consisting of RPMI 1640 (Invitrogen) supplemented with 10% (v/v) FBS and 1% (v/v) Pen-Strep-Glutamine (Invitrogen) for 4 h at 37°C in the presence of GolgiPlug (BD Biosciences). To measure NKT10 cell IL-10, splenocytes were isolated from mice and restimulated with PMA (0.1 μg/ml) and ionomycin (0.9 μg/ml) (both from Sigma-Aldrich) for 4 h at 37°C in the presence of GolgiPlug and GolgiStop in media consisting of RPMI 1640 (Invitrogen) supplemented with 10% (v/v) FBS and 1% (v/v) Pen-Strep-Glutamine (Invitrogen). Single-cell lymphocyte suspensions were purified using Lymphoprep (Axis-Shield, Oslo, Norway) density gradient centrifugation. Anti-mouse CD16/32 Ab (2.4 G2), purified in the Kronenberg laboratory, was used for FcR blocking. The following Abs were used in experiments: CD1d-αGalCer tetramers labeled with the fluorochrome BV421 (BD Biosciences) (generated in our laboratory) (35); CD1d, CD45R/B220, CD8, CD40, CD44, CD80, CD86, IL-10, and CD11b (all from BD Biosciences); NK1.1, IFN-γ, and CD11c (all from eBioscience); CD4 (Life Technologies); and TCRβ, CD3ε, CD19, and L363 (all from BioLegend). Dead cells were labeled with LIVE/DEAD yellow or aqua (Life Technologies). The cells were then fixed and permeabilized using Cytofix/Cytoperm buffer (BD Biosciences). The data were collected on an LSR II or Fortessa flow cytometer (BD Biosciences) and analyzed using FlowJo software (TreeStar).

C57BL/6 mice were depleted of NK cells by injecting 50 μl anti-asialo GM1 rabbit polyclonal Ab (Wako) 24 h prior to Ag challenge. NK cell depletion (NK1.1⁺ TCRβ⁻ cells) was verified by flow cytometry.

Mouse CD1d–β2-microglobulin heterodimeric protein was expressed in a baculovirus-expression system, as reported previously (36). Human CD1d–β2-microglobulin was prepared similarly to the mouse protein. The Vα14-Vβ8.2 TCR construct design, refolding, and purification processes were identical to the ones reported previously (37), whereas the autoreactive human Vα24-Vβ11 TCR (auto-Vα24) 4C1369 (38, 39) was generously provided by Dr. Jamie Rossjohn (Monash University, Clayton, VIC, Australia) and prepared as reported (40).

The DB06-1 lipid, synthesized in the laboratory of G.S.B., was dissolved in DMSO at 1 mg/ml. Before loading, 25 μl was diluted to 0.25 mg/ml with 25 μl vehicle solution (50 mM Tris-HCl [pH 7], 4.8 mg/ml sucrose, 0.5 mg/ml sodium deoxycholate, and 0.022% Tween 20) and 50 μl 1% Tween 20 and incubated at 80°C for 20 min. DB06-1 was loaded onto CD1d o/n (molar ratio of protein/lipid of 1:3) in the presence of 10 mM Tris-HCl (pH 7). Refolded TCR was incubated at room temperature for 1 h with lipid-loaded CD1d at a 1:2 molar ratio, and the ternary CD1d–lipid–TCR complex was isolated from uncomplexed CD1d and TCR by size-exclusion chromatography using a Superdex S200 10/300 GL (GE Healthcare).

Surface plasmon resonance (SPR) binding studies were conducted using a Biacore 3000, as reported previously (37). Briefly, ∼300 response units of biotinylated CD1d (either human or mouse) loaded with DB06-1 were immobilized onto a streptavidin sensor chip (GE Healthcare) surface by injecting the CD1d–DB06-1 mixture at 3 μl/min in HEPES buffered saline running buffer. A reference surface was generated in another flow channel with unloaded CD1d. Mouse or autoreactive human TCRs were flowed over at a constant rate. Experiments were carried out at 25°C with a flow rate of 30 μl/min and were performed at least twice. Kinetic parameters for the mouse molecule interactions were calculated after subtracting the response to CD1d molecules in the reference channel, using a simple Langmuir 1:1 model, in BIAevaluation software version 4.1. Human kinetic parameters were obtained using steady-state solution graphs plotting T_eq versus concentration and were fitted with binding response at equilibrium using BIAevaluation software version 4.1.

The mouse CD1d–DB06-1–TCR complex was isolated by Superdex S200 10/300 GL (GE Healthcare) column chromatography in 50 mM HEPES (pH 7.4), 150 mM NaCl and concentrated to 0.86 mg/ml. Crystals were grown at 22.3°C by sitting drop vapor diffusion while mixing 2 μl protein with 2 μl precipitate (0.2 M ammonium citrate dibasic [pH 4.98] 20% PEG 4000). Crystals were flash-cooled at 100°K in mother liquor containing 20% glycerol. Diffraction data were collected at the Stanford Synchrotron Radiation Laboratory beamline 11-1 and processed with Mosflm software (41). The CD1d–DB06-1–TCR crystallized in space group C222₁. The structure was solved by molecular replacement in Collaborative Computational Project, Number 4 (1994) using the protein coordinates from the CD1d-iGb3 structure as the search model [PDB code 2Q7Y] (42), followed by the iNKT cell TCR [PDB code 3QUZ] (43). The model was rebuilt into σA-weighted 2Fo – Fc and Fo – Fc difference electron-density maps using the COOT program (44). The lipid was built into a 2Fo – Fc map and refined using REFMAC (1994). The final refinement steps were performed using the TLS procedure in REFMAC with five domains (α1–α2 domain including carbohydrates and glycolipid, α3-domain, β2-microglobulin, variable domain, and constant domain of TCR). The CD1d–DB06-1–TCR structure was refined to 2.83 Å to an Rcryst and Rfree of 20.9 and 25.6%, respectively. The quality of the model was excellent, as assessed with MolProbity software (45) (Supplemental Table I).

DB06-1 is identical to αGalCer, with the exception of the replacement of the C2 carbonyl oxygen on the acyl chain by a sulfur atom (Fig. 1A). We used several assays to measure the antigenic potency of this compound. Initially, we tested DB06-1 in a cell-free Ag-presentation assay, whereby a soluble CD1d molecule was coated on a plate, GSL Ags were added, and IL-2 release from an iNKT cell hybridoma was used to determine whether the lipid could activate the iNKT cell TCR. DB06-1 only weakly stimulated the iNKT cell hybridoma compared with αGalCer (Fig. 1B). We also used a cell-based Ag-presentation assay, with bone marrow–derived DCs as the APCs. This a more physiologically relevant experimental set-up because it allows for the uptake by APCs and endolysosomal loading of GSL Ags into CD1d. In this experimental set-up, DB06-1 was a more effective Ag, although it remained weaker in comparison with αGalCer (Fig. 1C).

As noted in the 1Introduction, some lipids that activate mouse iNKT cells do not stimulate their human counterparts; therefore, such compounds in this category are not relevant for the development of immune therapies. To address the possible usefulness of DB06-1, we performed two studies. In the first, we determined whether DB06-1 could efficiently activate human iNKT cells derived from human blood. We showed that DB06-1 activated iNKT cells from two donors using human PBMCs as CD1d-expressing APCs (Fig. 2A). We also used a mouse strain in which the mouse Cd1d1 gene was replaced with its human CD1d counterpart. These mice also contained a human iNKT cell Vα24 TCRα transgene and were crossed onto the Cα^−/− background. We refer to this strain as iNKT cell humanized mice. To assess their immune response, we injected DB06-1 into these humanized mice and measured IFN-γ at 12 and 26 h postinjection (Fig. 2B), as well as IL-4 at 2 h postinjection (Fig. 2C). IFN-γ in the sera of humanized mice was higher when mice were injected with DB06-1 compared with αGalCer, whereas at the 2-h time point, IL-4 was decreased in the mice injected with DB06-1. In summary, these two studies indicate that DB06-1 is able to activate human iNKT cells and generate a Th1 response in a humanized iNKT cell mouse model.

Th1 cytokine skewing following GSL stimulation is believed to be the product of a cellular network that responds within the first 24 h (16). Therefore, we analyzed the in vivo response to DB06-1 by measuring the concentration of cytokines (IFN-γ and IL-4) in the sera of mice 2 and 22 h after injection (Fig. 3A). Previous results (22) showed that DB06-1 can induce a robust serum IFN-γ in vivo. The initial IFN-γ response induced by DB06-1, measured at 2 h, was similar to the response induced by αGalCer (Fig. 3A, Supplemental Fig. 1A) and is due to the rapid IFN-γ secretion from iNKT cells. Although ICCS IFN-γ at 2 h was higher in mice injected with αGalCer compared with DB06-1, (Supplemental Fig. 1C, 1D), this was not reflected in the sera data. The production of IFN-γ at 22–24 h after GSL injection has been attributed to the trans-activation of NK cells, due in part to IL-12 production from APCs (46–49). Serum IL-12 levels from mice injected with DB06-1 at 6 h postinjection were higher than in mice injected with αGalCer (Fig. 3B). To measure NK cell trans-activation, we analyzed IFN-γ production of splenic NK cells from mice injected with DB06-1 24 h earlier. After a 4-h culture with GolgiPlug, splenic NK cells (NK1.1⁺ TCRβ⁻) were identified by flow cytometry, and intracellular IFN-γ was measured. NK cells of mice injected with DB06-1 produced more IFN-γ than NK cells from αGalCer-injected mice (Fig. 3C), with 11–24% of NK cells from mice injected with DB06-1 producing IFN-γ compared with 1.5–6.5% of NK cells from mice injected with αGalCer (Supplemental Fig. 2A, 2B). To further determine that the sera IFN-γ production at this 24-h time point was indeed due to NK cell trans-activation, we repeated the experiment in mice depleted specifically of NK cells using anti-asialo GM1 Abs (50) This treatment effectively depleted NK cells (Supplemental Fig. 2C, 2D) but did not affect iNKT cells (Supplemental Fig. 2E). The serum IFN-γ levels 24 h after DB06-1 injection were significantly lower in these NK cell–depleted mice than in control mice (Fig. 3D). We also injected Il12^−/− mice and measured serum IFN-γ at 24 h by ELISA. In the absence of IL-12, the amount of IFN-γ in the serum from mice injected with DB06-1 was reduced ∼10-fold (Supplemental Fig. 2F). ICCS demonstrated that NK cells from DB06-1–injected Il12^−/− mice did not produce IFN-γ (Supplemental Fig. 2G). Based on these data, we conclude that DB06-1 causes a strongly Th1-skewed response in vivo, in both iNKT cell humanized and WT mice. Furthermore, it acts by causing increased IL-12, which stimulates NK cells to produce IFN-γ, with NK cells providing the majority of IFN-γ in the serum.

A previous study indicated that CD8α⁺ CD11c⁺ DCs are the dominant APC type essential for activation of iNKT cells by injected GSL Ags (51), although, in some circumstances, macrophages were shown to be important too, especially for Ags that are in particulate form (52). To determine whether DCs were essential for the presentation of DB06-1, we generated a mouse strain with floxed CD1d alleles (Cd1d^f/f mice) and crossed this line with a CD11c-Cre–transgenic mouse strain (Cd1d^f/f Cre⁺ mice), thereby deleting CD1d expression on CD11c⁺ cells, including most DCs (Fig. 4A). When Cd1d^f/f Cre⁺ mice were injected with DB06-1, we observed a significant decrease in the amount of IFN-γ in mouse sera at 24 h (Fig. 4B). However, because IFN-γ production was not completely absent, these data suggest that CD11c⁺ DCs may not be the sole population capable of presenting DB06-1 to iNKT cells in vivo. IL-4 in the sera at 2 h also was determined and was decreased in Cd1d^f/f Cre⁺ mice as well (Supplemental Fig. 3A). Therefore, we conclude that CD11c⁺ DCs likely were important for DB06-1 presentation in vivo, although the participation of other cell types was not excluded. To further investigate how αGalCer and DB06-1 might affect DCs, we analyzed surface markers on CD8α⁺ DCs, comparing the 2- and 24-h time points. Surface levels of CD1d did not change in mice injected with either GSL Ag; however, CD80 and CD86 were increased at 24 h postinjection of αGalCer or DB06-1. There was a trend toward higher expression of both CD80 and CD86 with injection of DB06-1 at 24 h, but the difference was not significant (Supplemental Fig. 3B)

We previously found that a common feature of several Th1 cytokine–skewing αGalCer analogs is that they persist longer as complexes with CD1d on the surface of APCs in vivo (53). To address this, we injected lipid Ags and we used an Ab that binds specifically to αGalCer–CD1d complexes (L363) to measure surface GSL–CD1d complexes on DCs using flow cytometry. After injection of either αGalCer or DB06-1, complexes with CD1d were barely detectable on the surface of DCs by flow cytometry at 2 h postinjection compared with control uninjected mice. However, at 24 h, DB06-1–CD1d complex staining was higher and increased compared with the αGalCer–CD1d complex (Supplemental Fig. 3C).

We analyzed the presence of these complexes using a T cell functional assay, which is more sensitive than flow cytometry, because it is likely that very few Ag–CD1d complexes are required to activate an iNKT cell. We injected mice with one of the GSL Ags, isolated a splenic fraction enriched for DCs at 2 or 24 h, and used these APCs to stimulate iNKT cell hybridomas in vitro (13, 53). The Th1-skewing lipids that had been analyzed in this way previously showed an increased ability to activate iNKT cell hybridomas at 24 h compared with αGalCer (53). In accordance with these studies, we observed that APCs purified 24 h after Ag injection were better able to activate iNKT cell hybridomas in vitro when they had been exposed in vivo to DB06-1 than when exposed to αGalCer (Fig. 4D). However, unlike the previous studies, the presentation of DB06-1 by APCs loaded in vivo induced a clearly stronger iNKT cell response ex vivo compared with αGalCer, even at 2 h after Ag injection (Fig. 4C). Although we did not detect surface Ag–CD1d complexes by flow cytometry on DCs of mice injected 2 h earlier, it is likely that an amount of complexes below the detection limit of flow cytometry was able to give an optimal stimulation of iNKT cell hybridomas in vitro. However, it is also possible that some lipid Ag taken up by DCs in vivo was able to load into CD1d during the culture period with the iNKT cell hybridoma cells. Regardless, our data demonstrate that DB06-1 is taken up and presented more effectively by DCs in vivo.

Previous data indicated that presentation of Th1 cytokine–skewing lipid Ags was increased when CD1d cycled through endosomal compartments. Furthermore, CD1d molecules presenting Th1-skewing Ags were preferentially associated with lipid rafts (22). CD1d contains a tyrosine motif in its cytoplasmic tail that allows trafficking to endosomal compartments. To study the role of endosome trafficking in the presentation of DB06-1, we used transfected cell lines that either expressed WT CD1d or TD CD1d lacking the cytoplasmic tail, including the tyrosine motif important for endosomal location. As a reference Ag, we used galactosyl (α1-2) galactosyl ceramide (GGC), which is known to require lysosomal processing to cleave the terminal sugar to yield the active compound, αGalCer. Without recycling through endolysosomal compartments, GGC is not recognized by iNKT cell hybridomas (54). Like GGC, DB06-1 presentation to an iNKT cell hybridoma also was greatly reduced when CD1d lacked the ability to recycle through endosomal compartments compared with the WT CD1d cells (Fig. 5A). To confirm that this recycling was also important in vivo, we used a mouse strain that lacked the tyrosine motif of the Cd1d gene (CD1-TD). Although surface expression of CD1d is significantly higher on APCs from CD1-TD mice compared with control mice, iNKT cells do not develop to WT numbers in CD1d-TD mice, likely due to the inability of CD1d to acquire the ligands appropriate for iNKT cell positive selection (30). To study the capability of APCs from these mice, we injected CD1-TD mice with DB06-1 or αGalCer, isolated splenic CD11c⁺ cells at 24 h, and used different concentrations of the DCs to stimulate an iNKT cell hybridoma ex vivo as a readout of Ag presentation. αGalCer was used as the control Ag because, although it shows a dependence on CD1d recycling in some experiments, in our experience it can load effectively into CD1d molecules on the cell surface (54). Although an iNKT cell response could be elicited by DCs derived from CD1-TD mice that had been injected with DB06-1, this response was reduced compared with the one stimulated by DCs from mice injected with αGalCer (Fig. 5B). This profile is opposite from the one obtained when these two Ags were injected into WT mice (Fig. 4D). Together, these data demonstrate that DB06-1 is presented when CD1d traffics normally through late endosomal compartments, and it is more dependent on CD1d recycling than is αGalCer.

One hypothesis for the ability of Ags to induce a Th1 cytokine–skewed response in vivo is that those Ags have an increased affinity for the iNKT cell TCR when bound to CD1d (55). Equilibrium-binding analysis using SPR demonstrated a binding affinity (K_D) of the mouse Vα14-Vβ8.2 iNKT cell TCR for mouse CD1d–DB06-1 complexes of 56 ± 6 nM (Fig. 6A), ∼2-fold weaker than the binding affinity observed with αGalCer-loaded CD1d (24 nM; data not shown). K_a (5.2 ± 0.7 10⁴ M⁻¹ s⁻¹) and K_d (2.9 ± 0.7 10⁻³ s⁻¹) were comparable to the rates for αGalCer (K_a = 7.84 10⁴ M⁻¹ s⁻¹ and K_d = 1.61 10⁻³ s⁻¹; data not shown).

We also analyzed the affinity of a human autoreactive iNKT cell TCR for the human CD1d–DB06-1 complex (Fig. 6B). Using steady-state kinetic modeling, whereby we plotted the residual units of the sensorgram against the concentration of the TCR and calculated a line of best fit (Fig. 6C), we determined that the affinity of the autoreactive human TCR for the CD1d–DB06-1 complex (0.485 ± 0.265 μM) was again ∼2-fold weaker than for the αGalCer–CD1d complex (0.22 ± 0.11 μM). Therefore, we conclude that, although the iNKT cell TCR has a high affinity for the CD1d–DB06-1 complex, the interaction with the αGalCer–CD1d complexes is even stronger. These results are consistent with other data demonstrating that increased TCR affinity cannot explain increased Th1 cytokine release resulting from iNKT cell stimulation.

To characterize the biochemical features of the binding of the DB06-1 GSL to CD1d and the iNKT cell TCR, we determined the structure of the ternary complex by x-ray crystallography (Fig. 7A, Supplemental Table I). The complex crystallized in the space group C222₁, with one complex in the asymmetric unit. The binding orientation of the iNKT cell TCR is consistent with the conserved parallel docking mode described previously (56). The TCR docks over the CD1d Ag binding groove in an orientation that is markedly different from the diagonal binding footprint seen in MHC–peptide–TCR complexes. As shown previously with models using the same TCR construct (53), the TCR interacts with CD1d using amino acids in the TCR CDR3α and CDR2β loops, as well as a CDR3β-dependent contact (Fig. 7B). The same previously seen polar contacts are also formed between the α-chain (N30, R95, G96) of the iNKT cell TCR and the DB06-1 Ag (Fig. 7B). There is well-defined density for the DB06-1 ligand (Fig. 7C), with the galactose head group exposed to recognition by the iNKT cell TCR. Polar contacts between the GSL Ag and amino acids E80, E153, and T156 of CD1d are virtually identical to the interactions of CD1d with αGalCer. The binding orientation of DB06-1 within the hydrophobic A′ and F′ grooves of CD1d is also conserved, with the acyl chain localizing to the A′ groove and the sphingoid base to the F′ groove, with minimal to no differences when superimposed on the αGalCer–CD1d structure (data not shown). Because the two molecules are chemically similar in these regions, this similarity was expected. DB06-1 differs from αGalCer by the replacement of an oxygen with a sulfur. The sulfur atom, which is 50% bigger than the oxygen atom present in αGalCer (Fig. 7D), took up more space in CD1d (Fig. 7E), possibly forming more intimate contacts with the surrounding CD1d residues.

Recent studies identified an iNKT cell subset, which we called NKT10 cells, that produces IL-10 and may have regulatory function (12, 57). Previous experiments demonstrated that iNKT cells exposed to αGalCer in vivo were more capable of producing IL-10 when restimulated weeks to months later. To compare a strongly Th-1–biasing GSL Ag with αGalCer for the induction of NKT10 cells, we injected mice with DB06-1 or αGalCer; 4 wk later, we measured the capacity for splenic iNKT cells to produce IL-10 following a brief stimulation in vitro with PMA and ionomycin, followed by ICCS. Remarkably, the frequency of IL-10⁺ iNKT cells 1 mo after DB06-1 immunization was significantly greater than after αGalCer immunization (Fig. 8). A similar enhancement of IL-10 production was observed after immunization with DB06-1 when the iNKT cells were restimulated with Ag 1 mo later (data not shown). These data suggest a possible link between a strong Th1 cytokine response and the induction of IL-10–producing iNKT cells.

Understanding how structural changes in GSL Ags can differentially modulate the immune response is important for understanding how iNKT cells influence immunity, as well as for developing therapeutic GSLs. For example, Ags that preferentially skew the immune response toward Th1 cytokine production could be useful as vaccine adjuvants and anticancer therapeutics (58, 59). In this study, we characterized DB06-1, a GSL Ag that differs from the well-studied αGalCer by only a single atom. Despite this subtle change, we show that DB06-1 leads to changes in the immune system of mice that are more pronounced than those induced by αGalCer. Within 1 d after immunization, DB06-1 caused an increased Th1-cytokine response, and in the long-term it induced more NKT10 cells than αGalCer. Moreover, because DB06-1 can activate human iNKT cells, it, or related Ags with thioamide groups, could be therapeutically relevant. It is noteworthy that DB06-1 was not better or markedly different from αGalCer using in vitro assays, considering TCR binding to the GSL–CD1d complex, as well as activation in vitro using CD1d-coated plates or APCs. However, in almost all of the in vivo assays it was superior, including increased loading onto DCs in vivo, increased IFN-γ in the serum, augmented IL-12 in the serum, and increased NK cell IFN-γ production. These findings emphasize the complexity of assaying GSL Ag efficacy and the need to study these compounds in vivo.

A number of theories have been proposed for Th1 cytokine skewing following GSL Ag immunization; however, despite much effort and investigation, this process remains incompletely understood. Selective uptake and presentation by an APC type, for example DCs versus B lymphocytes, could be important (60), although recent evidence indicates that CD8α⁺ DCs are the most essential APCs for the presentation of injected lipid Ags, regardless of whether they are Th1 or Th2 skewing (51). The effect of a GSL Ag on the APC could be critical, perhaps as a result of its trafficking in the cell and the site where it is loaded into CD1d. In fact, Th1 cytokine skewing has been associated with a requirement for Ag internalization for loading into CD1d, as opposed to loading into CD1d on the cell surface (61). This type of immune response is also correlated with the appearance of GSL–CD1d complexes on the cell surface in detergent-resistant domains (22). Consistent with this, DB06-1 originally was identified in a screen of lipid Ags that were associated with detergent-resistant domains when bound to CD1d. Furthermore, this Ag was shown previously to stimulate robust IFN-γ production in vivo; however, cytokine production in response to it was not compared with αGalCer, which could be classified as a Th0 Ag, considering the ratio of IL-4/IFN-γ that it induces (22). For unknown reasons, the different trafficking of GSLs in APCs is linked to changes in the DCs, such as increased expression of CD86, which could help to stimulate Th1 responses.

Another theory for Th1 cytokine skewing proposes that IFN-γ production is a result of prolonged stimulation by GSL–CD1d complexes, which could be due to increased TCR affinity, increased stability of GSL–CD1d complexes, or pharmacokinetic properties. For example, a decreased rate of compound degradation, as originally proposed for C-glycoside, could contribute to prolonged iNKT cell stimulation. In fact, our data indicate that several Th1-skewing GSL Ags have an increased half-life in vivo as Ag–CD1d complexes on the surface of APCs (13, 53). Furthermore, the results from structural studies indicate that some Th1-skewing glycolipids may have increased contacts with CD1d, and this enhanced interaction with CD1d may contribute to the prolonged antigenic stimulation (53, 62). Although these different theories have merit, one class of theory alone probably does not account for all compounds that cause Th1-biasing cytokine release, and these mechanisms are not mutually exclusive.

The novel compound characterized in this study fits with several of the theories and mechanisms described above for the activity of Ags that cause Th1-biased cytokine responses. For example, DB06-1 is preferentially presented by CD11c⁺ DCs to iNKT cells. As previously mentioned, it was known that DB06-1 bound to CD1d localized to detergent-resistant domains (22). Moreover, DB06-1 also shows a strong preference for internalization by APCs for effective presentation. Although DB06-1 can be loaded into CD1d in vitro in a cell-free assay, optimal presentation of this lipid is achieved by recycling through endosomal compartments. Although DB06-1 has a slightly weaker TCR affinity when presented by CD1d than αGalCer, a weaker TCR affinity also has been observed with other Th1-biasing GSL Ags, including the most well-studied Ag, C-glycoside (13). Similar to other Th1-biasing iNKT cell GSL Ags (53), DB06-1 persists as GSL–CD1d complexes on DCs in vivo, as measured by the ability of ex vivo DCs to activate an iNKT cell hybridoma. When DCs from mice injected with DB06-1 were analyzed ex vivo, they were more effective at stimulating the iNKT cell hybridoma at both the early (2 h) and late (24 h) time points, demonstrating increased loading into the CD1d groove in vivo. Consistent with this, we also observed increased Ag–CD1d complexes by flow cytometry but only at 24 h, likely because the increased epitope density at the earlier times was not sufficient to be detected by flow cytometry. Therefore, we propose that, although the larger sulfur atom may make CD1d loading more difficult in vitro in a cell-free assay, this may be overcome in the presence of lipid-transfer proteins in the lysosome. Conversely, the sulfur atom also may allow for better locking within the CD1d groove, inhibiting the GSL Ag from replacement. Furthermore, sulfur is also less electronegative than oxygen, and this may allow DB06-1 to be maintained in the CD1d hydrophobic pocket longer. Therefore, we speculate that, as for other Th1-biasing iNKT cell Ags, the enhanced molecular interaction of DB06-1 with CD1d permits an increased and prolonged iNKT cell stimulation that leads to increased IFN-γ production by trans-activated NK cells.

A population of iNKT cells that produces IL-10, called NKT10 cells, was described recently (12, 57). These cells are induced or expand greatly after a strong antigenic stimulation, for example after αGalCer immunization, they are long-lived, and they preferentially localize to adipose tissue. An intriguing property of DB06-1 described in this article is the robust IL-10 production by iNKT cells from mice injected with this compound 1 mo earlier compared with those injected with αGalCer. We made a similar observation of increased IL-10 production when iNKT cells were restimulated weeks after injection with several other Th1-biasing GSL Ags (G. Wingender, A. Birkholz, D. Sag, A.R. Howell, and M. Kronenberg, manuscript in preparation), suggesting that this property could be a more general one. This effect, extending far beyond the initial IFN-γ burst, could have profound implications regarding the development of iNKT cell GSL Ags as therapeutics. Previous studies showed that mice pretreated with αGalCer have a decreased ability to reject tumors due to the IL-10 production by NKT10 cells (12). If we extrapolate, it is possible that a GSL that leads to a more pronounced IL-10 response would theoretically lead to less tumor rejection and, thus, would be a poor target for a cancer therapeutic. However, other diseases may benefit from activation of NKT10 cells. Repetitive αGalCer injections were shown to lead to a reduced disease score in a mouse model of multiple sclerosis. This protection was suggested to be correlated with the ability of αGalCer to induce NKT10 cell IL-10 production, because IL-10^−/− mice are not protected (63). Single injections of αGalCer are unable to lead to this protection (64); however, a strong IL-10–inducing GSL like DB06-1 may prove to be more effective, and IFN-γ is known to be necessary to promote this response (65). In addition to this, αGalCer was shown to be protective in a mouse rheumatoid arthritis model through an IL-10–mediated response (66). Although humans have NKT10 cells, it is unknown whether the profound response seen in mice will be found in humans; more studies are needed to address the implications of IL-10–producing NKT10 cells on human therapeutics.

In summary, DB06-1 is a powerful activating GSL Ag capable of impacting the mouse immune system days and weeks after immunization. Its chemical properties allow for the stable formation of complexes with CD1d when it can be internalized within DCs in vivo. The characterization of this GSL, along with previous iNKT cell GSL Ags, contributes to our understanding of the mechanisms for diverse iNKT cell influences on the immune response and will aid in the logical design of potential future iNKT cell GSL Ag therapeutics.

We thank Dr. Jamie Rossjohn for the autoreactive human TCR plasmid, Dr. Albert Bendelac for the CD1-TD mice, and Kyowa Hakko Kirin for αGalCer. We also thank the Stanford Synchrotron Radiation Laboratory, the Flow Cytometry Core Facility, and the Department of Laboratory Animal Care at the La Jolla Institute for Allergy and Immunology for excellent technical assistance.

The authors have no financial conflicts of interest.