A semi-invariant V_α10⁺ T cell antigen receptor defines a population of natural killer T cells with distinct glycolipid antigen–recognition properties

Abstract

Type I natural killer T cells (NKT cells) are characterized by an invariant variable region 14–joining region 18 (V_α14-V_α18) T cell antigen receptor (TCR) α-chain and recognition of the glycolipid α-galactosylceramide (α-GalCer) restricted to the antigen-presenting molecule CD1d. Here we describe a population of α-GalCer-reactive NKT cells that expressed a canonical V_α10-J_α50 TCR α-chain, which showed a preference for α-glucosylceramide (α-GlcCer) and bacterial α-glucuronic acid–containing glycolipid antigens. Structurally, despite very limited TCRα sequence identity, the V_α10 TCR–CD1d–α-GlcCer complex had a docking mode similar to that of type I TCR–CD1d–α-GalCer complexes, although differences at the antigen-binding interface accounted for the altered antigen specificity. Our findings provide new insight into the structural basis and evolution of glycolipid antigen recognition and have notable implications for the scope and immunological role of glycolipid-specific T cell responses.

Natural killer T cells (NKT cells) arise in the thymus through random T cell antigen receptor (TCR) gene-recombination events that result in the expression of TCRs that preferentially interact with lipid-based antigens presented by CD1d on cortical thymocytes¹. This unusual form of positive selection diverts developing thymocytes into the NKT cell lineage, which is characterized by rapid and diverse cytokine production and multipotent immunoregulatory abilities. NKT cells are traditionally categorized into two distinct populations: type I and type II (ref. 1). Type I NKT cells express an invariant α-chain variable region 14–α-chain joining region 18 (V_α14-J_α18; TRAV11-TRAJ18) TCR α-chain in mice or V_α24-J_α18 (TRAV10-TRAJ18) in humans and can recognize the prototypical glycolipid antigen α-galactosylceramide (α-GalCer). Type II NKT cells are also CD1d restricted but do not express this invariant TCR α-chain or recognize α-GalCer. Instead, type II NKT cells express a range of TCRs², although they seem to be enriched for TCRs that incorporate V_α3.2-J_α9 and V_α8 (ref. 3) and recognize distinct glycolipid antigens such as sulfatide⁴. Mainly because CD1d–α-GalCer tetramers are the tool of choice for the study of NKT cells, most studies of glycolipid-reactive T cells have focused on type I NKT cells. The invariant V_α14-J_α18 TCR α-chain is a defining feature of type I NKT cells to the extent that these cells are not detected in mice with deletion of the gene encoding J_α18 (Tcra-J^tm1Tgi mice; called ‘J_α18^−/− mice’ here)⁵, now commonly used as a model of deficiency in type I NKT cells. Furthermore, differences between J_α18^−/− mice and mice with deletion of the gene encoding CD1d (Cd1d^−/− mice) are presumed to indicate a role for type II NKT cells, which are still present in the former strain but not the latter strain^6–9. Although human V_α24⁻ NKT cells reactive to the CD1d–α-GalCer tetramer have been described^10–12, these cells retain the J_α18 junctional region that characterizes type I NKT cells^10,12.

Type I NKT cells are stimulated by an array of microbial and self-derived lipid-based antigens¹³. Given the invariant nature of the NKT cell TCR α-chain, this suggests that the NKT cell TCR β-chain has a role in determining differences in antigen specificity. Indeed, mouse NKT cells, which have a more diverse V_β repertoire than do humans, frequently use three V_β genes (V_β8, V_β7 and V_β2; TRBV13, TRBV29 and TRBV1, respectively). The structures of various type I NKT cell TCR–CD1d–antigen complexes^14–18 have provided insight into the basis of type I NKT cell recognition and some clues about the role of differences in V_β use. In all NKT cell TCR–CD1d–antigen structures solved so far, a conserved, tilted and parallel docking mode relative to the CD1d antigen-binding cleft has been observed. In this common framework, the invariant NKT cell TCR α-chain, and in particular the J_α18 complementarity-determining region 3 (CDR3) α-loop, dominates the interaction with CD1d. Further, NKT cell TCR-mutagenesis experiments have demonstrated the importance of the V_α14 and V_α24 CDR1α and J_α18 CDR3α loops in interacting with CD1d when they are bound to a diverse array of lipid-based antigens^19–22. Although differences are observed in the way V_β8.2 and V_β7 NKT cell TCRs interact with CD1d, the overall docking mode is largely conserved. Collectively, these studies suggest that the invariant V_α14-J_α18 α-chain has the principal role in determining the conserved NKT cell TCR-CD1d docking topology.

Here we define a previously unknown population of V_α10-J_α50⁺ NKT cells (called ‘V_α10 NKT cells’ here) that fits neither the type I nor the type II NKT cell category. Analogous to classical type I NKT cells, the V_α10 NKT cells were defined by a previously unknown canonical TCR α-chain rearrangement and shared some functional attributes with type I cells, including CD1d-mediated recognition of α-GalCer, but recognized other glycolipids differently, including self and bacterial lipids. Here we provide an analysis of several key attributes of this NKT cell subset, including TCR use, antigen reactivity, functional responses and the structural basis of antigen recognition by V_α10 NKT cells.

RESULTS

Identification of T cells reactive to CD1d-α–GalCer in J_α18^−/− mice

J_α18^−/− mice are considered to completely lack type I NKT cells reactive to CD1d–α-GalCer¹; however, we observed a small population of cells reactive to the CD1d–α-GalCer tetramer in the thymus and liver of J_α18^−/− mice that was absent from Cd1d^−/− mice (Fig. 1a). This population was most obvious on the BALB/c background, although it was also detectable in C57BL/6 mice, and was most prominent after enrichment by depletion of CD8⁺ and CD24⁺ thymocytes. Similar to BALB/c type I NKT cells²³ (Fig. 1a), these cells could often be categorized into two subsets based on the amount of TCR expression; they also included CD4⁺ and CD4⁻ subsets, were CD44^hi and CD69^int (Fig. 1b) and expressed a TCR V_β repertoire enriched for V_β8 and V_β7 (Fig. 1c). In C57BL/6 mice, these cells expressed the activating NK cell receptor NK1.1, similar to C57BL/6 type I NKT cells (Fig. 1b). Accordingly, these cells, which lacked the invariant V_α14-J_α18 TCRα yet recognized CD1d–α-GalCer, were representative of neither classical type I NKT cells nor type II NKT cells and represent a previously unrecognized population of NKT cells.

Figure 1.

Identification of J_α18^−/− T cells reactive to CD1d–α-GalCer. (a) Flow cytometry of thymocytes and liver lymphocytes isolated from BALB/c and C57BL/6 wild-type (WT), J_α18^−/− and Cd1d^−/− mice (n = 7–9 mice per genotype) and stained with tetramer loaded with CD1d–α-GalCer and monoclonal antibody to αβTCR; thymocyte populations were also enriched for NKT cells by complement-mediated depletion of CD24⁺ and CD8⁺ thymocytes. Numbers above outlined areas indicate percent tetramer-positive αβTCR⁺ cells. (b) Expression of CD4 versus CD3, CD8, CD44, CD69 and CD49b (black lines) and isotype-matched control antibody staining (gray shading) on NKT cells reactive to CD1d–α-GalCer (BALB/c wild-type cells) or CD1d–α-GlcCer (BALB/c J_α18^−/− cells); right, expression of CD4 and NK1.1 in NKT cells from mice on the C57BL/6 background. Numbers above outlined areas, in quadrants or above bracketed lines, indicate percent positive cells in each area. (c) Expression of various TCRβ V_β regions by NKT cells from BALB/c wild-type and J_α18^−/− mice. Numbers in quadrants indicate percent positive cells in each. Data are representative of three (a) or two (b,c) separate experiments.

J_α18^−/− NKT cells express a V_α10-J_α50 TCR α-chain

We next determined whether NKT cells derived from J_α18^−/− mice expressed the gene encoding V_α14 or, if not, which genes encoding TCR V_α were used by these cells. RT-PCR-based analysis with a panel of primers for each gene encoding TCR V_α showed a strong band corresponding to V_α10 (TRAV13), but little expression of other V_α genes, apart from a faint band corresponding to V_α13 that sequencing confirmed was nonfunctional (Fig. 2a and data not shown). We then used single-cell RT-PCR analysis to determine the frequency of V_α10 NKT cells. As expected, most conventional T cells (negative for the CD1d–α-GalCer tetramer) did not express V_α10; however, most (14 of 16) of the NKT cells from J_α18^−/− mice reactive to the CD1d–α-GalCer tetramer were V_α10⁺ (Fig. 2b). Furthermore, sequence analysis of the PCR products showed that all V_α10⁺ NKT cells expressed the gene encoding J_α50 (Table 1). Of 33 V_α10⁺ sequences from four independent cell sorts, all included J_α50 and one of five similar CDR3α sequences, each with the same length. Moreover, V_α10 has nine subtypes (TRAV13-1 through TRAV13-5, and TRAV13D-1 through TRAV13D-4), and all sequences were TRAV13-3 or TRAV13D-3 (which differ by a single amino acid in the framework region), which suggested a potential role for both the CDR1α and CDR2α regions of the V_α10 NKT cell TCR in antigen recognition. Sequence analysis of the TCR β-chain with V_β8.1- and V_β8.2-specific primers showed diverse expression of TCR J_β and diverse length and composition of CDR3β (Table 1). Thus, similar to the V_α14 sequence of type I NKT cells, the TCR α-chain sequence of V_α10 NKT cells was nearly invariant, whereas the TCR β-chain was heavily biased toward V_β8 use but was diverse in the CDR3β region.

Figure 2.

J_α18^−/− CD1d–α-GalCer⁺ NKT cells express a semi-invariant V_α10-J_α50–V_β8⁺ TCR. (a) PCR analysis of cDNA isolated from CD1d–α-GalCer-reactive cells sorted from BALB/c J_α18^−/− thymuses and amplified with a panel of primers specific for each TCRα V-gene segment or the α-chain constant region (C_α). – (far right), C_α primers with no cDNA. (b) Single-cell PCR analysis for V_α10 on cDNA isolated from V_β8.1 and V_β8.2⁺ cells positive for CD1d–α-GalCer tetramer (right; J_α18^−/− mouse–derived) or V_β8.1 and V_β8.2⁺ CD4⁺ αβTCR⁺ cells negative for the CD1d–α-GalCer tetramer (left; conventional T cells) sorted from BALB/c J_α18^−/− thymuses; n = 16 cells per panel. (c) Staining of surface TCRβ (top row) and unloaded or α-GalCer-loaded CD1d tetramer (bottom row) on green fluorescent protein–gated human epithelial 293T cells transfected to express full-length rearranged V_α10-J_α50 or V_α14-J_α18 TCR α-chain, plus V_β8.1, V_β8.3 or V_β7 TCR β-chain, and CD3 complex. Isotype, isotype-matched control antibody. Numbers above bracketed lines indicate percent-positive cells. Data are from one experiment (a,b; one for each) or are representative of one (V_β8.1) or two (V_β8.3 and V_β7) experiments (c).

Table 1.

Sequence analysis of V_α10 NKT cells

TCRα	Gene segment
Sequence	Vα	V-N	Jα
1	Vα10	104CAIAS	SSFSKLVFGQGTSLSVVP (Jα50)
2	Vα10	104CAIRS	SSFSKLVFGQGTSLSVVP (Jα50)
3	Vα10	104CAMKA	SSFSKLVFGQGTSLSVVP (Jα50)
4	Vα10	104CAMRA	SSFSKLVFGQGTSLSVVP (Jα50)
5	Vα10	104CAMQF	SSFSKLVFGQGTSLSVVP (Jα50)
TCRβ
Sequence	Vβ	V-NDN	Jβ
1	Vβ8.1	104CASRLGG	YEQYFGPGTRLTVL (Jβ2.7)
2	Vβ8.2	104CASGDWGA	NTGQLYFGEGSKLTVL (Jβ2.2)
3	Vβ8.2	104CASGGTGGR	EQYFGPGTRLTVL (Jβ2.7)
4	Vβ8.2	104CASGDWGA	EQYFGPGTRLTVL (Jβ2.7)
5	Vβ8.1	104CASRGQG	TEVFFGKGTKLTVV (Jβ1.1)
6	Vβ8.2	104CASGAGLGGRD	NYAEQFFGPGTRLTVL (Jβ2.1)
7	Vβ8.2	104CASGGRLGG	YAEQFFGPGTRLTVL (Jβ2.1)
8	Vβ8.2	104CASSDIWGGH	EQYFGPGTRLTVL (Jβ2.7)
NKT cell subset	CDR1α	CDR2α	CDR3α
Vα10 (Vα10-Jα50)	TTLNS	SPSWA	CAIASSSFSKLV
Type I (Vα14-Jα18)	VTPDNH	LVHEND	CVVGDRGSALGR

PCR analysis of unique sequences for TCR α-chains (33 total sequences) and β-chains (13 total sequences) of V_α10 NKT cells sorted from thymus or liver, amplified with primers specific for V_α10, and V_β8.1 and V_β8.2 (top); and CDR1α, CDR2α and CDR3α sequences from V_α10 and type I NKT cell TCR α-chains (below). Superscripted numbers adjacent to sequences (top) indicate position of first amino acid in sequence; bolding (below) indicates type I NKT cell TCR residues critical for CD1d–α-GalCer interactions. N, nucleotide additions to TCR sequences; D, diversity region.

Data are representative of three (TCRα) or two (TCRβ) separate experiments.

The V_α10-J_α50 TCRα facilitates recognition of CD1d–α-GalCer

Alignment of the sequences of V_α10-J_α50 and V_α14-J_α18 showed little homology; moreover, none of the residues with a key role in CD1d-antigen recognition in the V_α14-J_α18 NKT cell TCR^19–21 were conserved in the V_α10-J_α50 TCR (Table 1). Therefore, we directly determined whether this TCRα facilitated recognition of CD1d–α-GalCer. We transfected human epithelial 293T cells with rearranged V_α10-J_α50 or V_α14-J_α18 TCR α-chains, plus V_β8.1 TCR β-chain with a sequence derived from V_α10 NKT cells (Table 1, sequence 1) or with V_β8.3 or V_β7 TCR β-chains derived from irrelevant T cell lines (H-2D^b restricted and influenza A specific), and CD3 complex. The V_α10-J_α50⁺ TCR paired with each TCRβ to support surface expression of αβTCR, and when paired with the V_β8.1⁺ TCRβ, it conferred recognition of CD1d–α-GalCer (Fig. 2c). The reactivity of the V_α14-J_α18 TCR α-chain to CD1d–α-GalCer was more permissive, with TCR β-chains V_β8.1, V_β8.3 and V_β7 all conferring binding. Thus, we have identified a previously unknown canonical NKT cell TCR α-chain distinct from V_α14-J_α18 yet still able to confer recognition of CD1d–α-GalCer when paired with a permissive TCR β-chain. The finding that the irrelevant V_β8.3 and V_β7 TCR β-chains did not support binding of the CD1d-α-GalCer tetramer suggested that CDR3β is also important for the reactivity of the V_α10 NKT cell TCR to the CD1d-α-GalCer tetramer.

Glycolipid recognition by V_α10 NKT cells

In addition to recognizing α-GalCer, type I NKT cells can recognize a range of glycosphingolipid antigens. Given the very different amino acid sequences of the V_α10-J_α50 TCR and the V_α14-J_α18 TCR of type I NKT cells, we labeled thymocyte samples enriched for V_α10 and type I NKT cells with the fluorescent dye CFSE and assessed their proliferative responses in vitro to a panel of lipid-based CD1d ligands that included bacterial and self lipid antigens (Fig. 3, Supplementary Fig. 1 and data not shown). Although both types of NKT cells proliferated considerably in response to α-GalCer with a saturated 26-carbon (C_26:0) acyl chain (KRN7000) and a α-GalCer analog with a di-unsaturated 20-carbon (C_20:2) acyl chain²⁴, the strongest proliferative response of V_α10 NKT cells was in response to C_20:2 α-glucosylceramide (α-GlcCer)²⁵. The mammalian glycolipids iGb3 and β-GalCer stimulated both types of NKT cells, although the response of V_α10 NKT cells to β-GalCer was more variable among experiments (Fig. 3a and data not shown).

Figure 3.

V_α10 NKT cells have a unique hierarchy of antigen recognition. (a) Frequency of NKT cells among thymocyte populations obtained from BALB/c wild-type and J_α18^−/− mice, depleted of CD8⁺ and CD24⁺ cells and cultured for 72 h with glycolipid-pulsed antigen-presenting cells (J_α18^−/− splenocytes), and proliferation of CFSE-labeled type I and V_α10 NKT cells (gated on CD1d–α-GalCer tetramer, with an additional gate to exclude CFSE spectral overlap (not shown)). Numbers above outlined areas or brackets indicate percent positive cells in each; numbers above bracketed lines indicate percent divided cells. Data are from one of two similar experiments. (b) Proliferation (top) and cytokines in supernatants (below) of NKT cells positive for the CD1d–α-GalCer tetramer sorted from BALB/c wild-type and J_α18^−/− mice, then labeled with CFSE and cultured for 72 h (4 × 10³ cells per well) in the presence of no glycolipid, α-GalCer (C_26:0; 500 ng/ml), α-GlcCer (C_20:2; 500 ng/ml), α-GlcA–DAG(mixture of variants; 10 µg/ml), GSL-1 (1 µg/ml) or iGb3 (10 µg/ml), plus 20 × 10³ sorted splenic CD11c⁺ dendritic cells. Each symbol shape represents a different experiment. IFN-γ, interferon-γ. Data are from up to four independent experiments (mean and s.e.m. of three to twelve replicates). (c) Proliferation of gated V_α10 NKT cells positive for the CD1d–α-GalCer tetramer, sorted from BALB/c J_α18^−/− thymus, labeled with CFSE and cultured for 72 h in the presence (1 × 10³ cells per well) or absence (5 × 10³ cells per well) of α-GlcA-DAG (10 µg/ml), plus sorted CD11c⁺ dendritic cells (20 × 10³ per well). Numbers above bracketed lines indicate percent divided cells. Data are from one experiment with two replicates. (d) Proliferation and cytokine concentrations in supernatants of NKT cells positive for the CD1d–α-GalCer tetramer, sorted from BALB/c wild-type thymus (type I) or J_α18^−/− thymus (V_α10), labeled with CFSE and cultured for 72 h (2 × 10³ cells per well) with 20 × 10³ sorted CD11c⁺ dendritic cells and doubling dilutions of C_19:0-C_16:0 α-GlcA-DAG. Data are from one of two similar experiments (mean of duplicate cultures).

To determine if V_α10 NKT cells recognize glucose-containing microbial glycolipid ligands, we assessed the proliferation and cytokine production (interferon-γ, interleukin 4 (IL-4), IL-13 and IL-17A) in additional experiments involving coculture of purified CFSE-labeled V_α10 and type I NKT cells in the presence of purified CD11c⁺ dendritic cells plus the following antigens: α-glucuronosyl diacylglycerol (α-GlcA-DAG) derived from Mycobacterium smegmatis (a mixture of three variants (C_19:0-C_16:0, C_18:0-C_16:0 and C_16:0-C_19:0) at a ratio of 1:1:1 (wt/wt/wt), with one being the naturally occurring glycolipid; Supplementary Fig. 1), α-glucuronosyl ceramide (GSL-1) derived from Sphingomonas species, α-GalCer, α-GlcCer and iGb3 (Fig. 3b). Although most V_α10 and type I NKT cells proliferated in response to α-GalCer and α-GlcCer, α-GlcA-DAG elicited a stronger proliferative response from V_α10 NKT cells than from type I NKT cells. Furthermore, V_α10 NKT cells produced 10- to 100-fold more IL-4, IL-13 and IL-17A than did type I NKT cells in response to α-GlcA–DAG. Compared with the response of the control cells without glycolipid, the response of type I NKT cells to this antigen ranged from weak to undetectable. The proliferative response to GSL-1 was similar for V_α10 and type I NKT cells, although we also observed more production of IL-4 and IL-13 by V_α10 NKT cells in response to this glycolipid (albeit to a lesser extent than in response to α-GlcA–DAG). We also detected low concentrations of interferon-γ (<50 pg/ml) in cultures of V_α10 NKT cells in response to these microbial glycolipids. Conversely, type I NKT cells seemed to produce stronger responses to α-GalCer and iGb3.

To determine which of the three α-GlcA-DAG variants stimulated the V_α10 NKT cells, we tested each ligand separately (Fig. 3c). The naturally occurring compound C_19:0-C_16:0 α-GlcA-DAG²⁶ elicited the strongest proliferative response. Dose-response experiments with C_19:0-C_16:0 α-GlcA-DAG showed greater stimulation of V_α10 NKT cells at a range of doses from 10 µg/ml down to 0.6 µg/ml (Fig. 3d). Type I NKT cells showed little response at any dose but responded well to the positive control (α-GlcCer; data not shown). These data suggest that although there is some overlap in specificity, V_α10 NKT cells are functionally distinct from type I NKT cells and tend to recognize glucose- and glucuronic acid–containing glycolipids.

Binding affinity of the V_α10 NKT cell TCR

We used a soluble V_α10-J_α50 NKT cell TCR composed of paired TCR α- and β-chains (Table 1, sequence 1) from a V_α10 NKT cell clone carrying a TCR α-chain sequence present in both thymus and liver and measured affinity by surface plasmon resonance (Fig. 4a). As expected, control type I TCRs (V_α14-J_α18–V_β8.2 and V_α14-J_α18–V_β7; derived from published work¹⁴) bound to CD1d–α-GalCer with very high affinity (dissociation constants of 41 nM and 170 nM, respectively) and to CD1d–α-GlcCer with approximately 10% as much affinity (dissociation constants of 530 nM and 1.9 µM, respectively). Consistent with the results of the proliferation experiments (Fig. 3a), the V_α10 NKT cell TCR bound to CD1d–α-GlcCer with approximately fivefold higher affinity than it showed for CD1d–α-GalCer (dissociation constants of 4.4 µM and 23 µM, respectively). Thus, V_α10 NKT cell TCRs seemed to preferentially recognize CD1d–α-GlcCer better than CD1d–α-GalCer.

Figure 4.

V_α10 NKT cells have a higher affinity for α-GlcCer and are present in wild-type mice. (a) Binding of graded concentrations of V_α10 soluble TCR (V_α10–V_β8.1 (175–0.05 µM)) or type I soluble TCR (V_α14–V_β8.2 (150–0.04 µM) or V_α14–V_β7 (200–0.05 µM)) to CD1d–α-GalCer or CD1d–α-GlcCer, after subtraction of results from those of a control flow cell (unloaded CD1d). Far right, saturation plots showing equilibrium binding to immobilized CD1d–α-GalCer or CD1d–α-GlcCer and the equilibrium dissociation constant (K_d) derived by equilibrium analysis. RU, response units. Data are representative of two independent experiments. (b) Staining profiles of α-GalCer tetramer (left) and α-GlcCer tetramer (middle) and dual tetramer labeling (right) in BALB/c wild-type or J_α18^−/− thymocyte populations depleted of CD8⁺ and CD24⁺ cells, then simultaneously costained with CD1d tetramers loaded with α-GalCer and α-GlcCer. Numbers adjacent to outlined areas indicate percent cells in each. Far right, single-cell PCR analysis of V_α10 and V_α14 on wild-type NKT cells with high expression of the α-GlcCer tetramer (α-GlcCer tet^hi) or α-GalCer tetramer (α-GalCer tet^hi). Data are representative of three independent experiments.

On the basis of those differences in recognition, we determined whether costaining with CD1d tetramers loaded with either α-GlcCer or α-GalCer would allow us to distinguish V_α10 NKT cells from type I NKT cells in wild-type mice. Costaining of thymocytes showed a minor population that stained more brightly with the α-GlcCer tetramer (Fig. 4b), and a major population was stained more brightly with the α-GalCer tetramer. We assessed the TCR expression of these populations by single-cell PCR. All nine cells with high expression of the CD1d–α-GlcCer tetramer were V_α10⁺ (Fig. 4b), whereas none were V_α14⁺ and we successfully sequenced eight of nine as V_α10-J_α50⁺ (data not shown). In contrast, only one of nine cells with high expression of the CD1d–α-GalCer tetramer was V_α10⁺, and sequencing determined that this was a nonproductive J_α50⁻ rearrangement (data not shown), whereas six of nine of these cells were V_α14⁺. Thus, V_α10⁺ NKT cells were present in wild-type mice and could be distinguished from type I NKT cells by costaining with CD1d tetramer loaded with either α-GalCer or α-GlcCer.

The V_α10 TCR–CD1d–α-GlcCer ternary complex

We determined the structure of the V_α10 TCR–CD1d–α-GlcCer complex at a resolution of 2.2Å (Fig. 5 and Supplementary Table 1). The V_α10 TCR adopted a parallel docking mode above the F′ pocket of the CD1d antigen-binding cleft (Fig. 5a,b) in a manner very similar to that of other type I NKT cell TCR–CD1d–α-GalCer complexes (Fig. 5c). However, there was a 14° difference in the docking angles, with the V_α10 TCR leaning more toward the F′ pocket than did the type I NKT cell TCR–CD1d–α-GalCer complex (Fig. 5d).

Figure 5.

Structural comparison of V_α10 NKT cell TCR–CD1d–α-GlcCer and type I NKT cell TCR–CD1d–α-GalCer. (a) NKT cell V_α10–V_β8.1 TCR in complex with CD1d–α-GlcCer: magenta, α-GlcCer; gray, CD1d; salmon, V_α10; light green, V_β8.1; purple, CDR1α; dark green, CDR2α; yellow, CDR3α; teal, CDR1β; ruby, CDR2β; orange, CDR3β. β₂m, β₂-microglobulin. (b) Footprint of the NKT cell V_α10–V_β8.1 TCR on the surface of CD1d–α-GlcCer: spheres indicate α-GlcCer; colors of CD1d and CDR loops as in a. (c) Type I NKT cell V_α14–V_β8.2 TCR in complex with CD1d–α-GalCer¹⁴: blue, α-GalCer; cyan, V_α14; dark green, V_β8.2; colors of CD1d and CDR loops as in a. (d) Superposition of NKT cell V_α10–V_β8.1 TCR–CD1d–α-GlcCer and type I NKT cell V_α14–V_β8.2 TCR–CD1d–α-GalCer (colors as in a,c). (e) Footprint of the type I NKT cell V_α14–V_β8.2 TCR on the surface of CD1d–α-GalCer: spheres indicate α-GalCer; colors of CD1d, α-GalCer and CDR loops as in a,c.

The interactions of the V_α10 TCR with CD1d spanned residues 72–87 and 145–154 of the α1 helix and α2 helix, respectively (Supplementary Table 2), which buried ~910Ų after ligation; this was slightly greater than the buried surface area at the interface of type I NKT cell TCR–CD1d–α-GalCer (buried surface area, 760–860Ų for V_β8.2 and V_β7 NKT cell TCRs¹⁴; Fig. 5b,e). At this interface, the TCR α- and β-chains contributed 59% and 41% of the buried surface area, respectively, with all CDRs contacting CD1d. CDR1α, CDR2α and CDR3α contributed 14%, 13% and 32%, respectively, whereas the CDR1β, CDR2β and CDR3β loops contributed 8%, 21% and 12%, respectively. Thus, whereas the CDR3α and CDR2β loops dominated interactions at the V_α10 TCR–CD1d–α-GlcCer interface, the involvement of CDR2α was unique to V_α10 NKT cells, in contrast to the type I NKT V_β8⁺ TCR, in which only CDR1α, CDR3α and CDR2β mediated contact with CD1d–α-GalCer. Thus, the V_α10 and V_α14 NKT cell TCR–CD1d–antigen complexes showed similar docking modes, despite substantial differences in their V_α and J_α sequences.

V_α10 TCR–CD1d interactions

As V_α10 shares only 40% sequence identity with V_α14, we observed many differences in the interactions mediated by the respective TCR α-chains. In the V_α10 complex, Thr28α and Asn30α of CDR1α made van der Waals interactions with Val72 and Asp153 of CD1d, respectively (Fig. 6a), whereas in the type I complex, the CDR1α loop exclusively contacted α-GalCer¹⁴. The CDR2α loop of V_α10 TCR contained a bulky Trp59α residue that made extensive van der Waals interactions with residues spanning Ala152–Gly155 of the α2 helix of CD1d (Fig. 6a and Supplementary Table 2), whereas the corresponding loop in the type I TCR does not mediate any contacts with CD1d–α-GalCer¹⁴ or CD1d–α-GlcCer¹⁷. J_α50-encoded CDR3α interacted with residues from both the α1 and α2 helices of CD1d, including one hydrogen bond (between Ser109α and Asp80). Most notably, the bulky side chain of Phe113α protruded into the binding cleft between the α1 helix and α2 helix and made van der Waals interactions with Asp80 and Glu83 of the α1 helix and Pro146 and Val149 of the α2 helix (Fig. 6b, Supplementary Table 2). Accordingly, the CDR3α-CD1d interactions in the V_α10 complex were dominated by van der Waals contacts, whereas the type I NKT cell TCR CDR3α-CD1d interface was characterized by a large network of polar interactions¹⁴ (Fig. 6c).

Figure 6.

CD1d-mediated interactions with V_α10–V_β8.1 NKT cell TCR. (a) Contacts of V_α10 NKT cell TCR CDR1α and CDR2α with CD1d. (b) Contacts of V_α10 NKT cell TCR CDR3α with CD1d. (c) Contacts of type I NKT cell TCR CDR3α with CD1d¹⁴. (d) Contacts of V_α10 NKT cell TCR CDR1β and CDR3β with CD1d. (e) Contacts of V_α10 NKT cell TCR CDR2β with CD1d. Purple, CDR1α; dark green, CDR2α; yellow, CDR3α; teal, CDR1β; ruby, CDR2β; orange, CDR3β; gray, CD1d; black dashed lines, hydrogen bonds and salt-bridge interactions.

V_β8.1 shares 87% and 52% sequence identity with V_β8.2 and V_β7, respectively, and thus many contacts were conserved between the TCR β-chain of V_α10 and type I complexes. As noted before for type I NKT cell TCRs¹⁴, the TCR β-chain of the V_α10 NKT cell TCR contacted only CD1d (Supplementary Table 2). The interactions between V_α10 NKT cell TCR CDR1β and CD1d were mediated by Asp30β and Tyr31β, with Asp30β forming a salt bridge with Lys148, whereas the aromatic side chain of Tyr31β interacted with Val149 (Fig. 6d). These CDR1β-CD1d interactions were absent from the V_β8.2 NKT cell TCR–CD1d–α-GalCer complex, whereas the V_β7 complex included an equivalent salt bridge (Glu30β-Lys148)¹⁴.

In the CDR2β loop, Tyr55β, Tyr57β and Glu63β contacted CD1d, with Tyr55β and Tyr57β forming hydrogen bonds and van der Waals contacts with Glu83 and Lys86, which also formed a salt bridge with Glu63β. In addition, Tyr57β formed van der Waals interactions with Met87 and Leu145 of the CD1d α1 helix and α2 helix, respectively (Fig. 6e). These interactions were analogous to those mediated by the CDR2β loop of the V_β8.2 type I NKT cell TCR¹⁴.

The hypervariable CDR3β loop ‘sat’ above the α2 helix, which allowed Leu108β, Gly109β, Gly113β and Tyr114β to make van der Waals interactions with CD1d (Fig. 6d). This suggests that the CDR3β loop has an important role in the V_α10 TCR–CD1d interaction and is consistent with the observation that V_β8.3 and V_β7 β-chains containing irrelevant CDR3β sequences failed to confer reactivity to the CD1d–α-GalCer tetramer when paired with V_α10 (Fig. 2c). Accordingly, despite the similar footprints of the V_α10 and type I NKT cell TCRs with CD1d, specific differences at the binding interface were readily apparent.

Plasticity of the V_α10 TCR interaction

We also determined the structure of the V_α10 NKT cell TCR not bound to a ligand at a resolution of 2.9Å (Supplementary Table 1) and assessed the conformational changes that took place after ligation. In addition to small differences between the TCR alone and in complex in the V_α-V_β domain positioning, some of the CDR loops had also moved to allow optimal binding to CD1d–α-GlcCer. In addition to the small change (root mean square deviation (r.m.s.d.), 1.1Å) in the CDR1α loop, we observed notable conformational changes in the CDR3β loop (r.m.s.d., 2.4Å, with maximal shift at Gly109β; data not shown) and CDR3α loop (r.m.s.d., 1.5Å, maximal shift of 2.8Å at Phe113α; Fig. 7a). In the V_α10 NKT cell TCR not bound to a ligand, the CDR3β interacted with the tip of CDR3α; however, in the complex, CDR3β moved away from CDR3α, making contacts with the α2 helix of CD1d instead (Fig. 6d). The movement in the CDR3α loop was particularly notable, as reorientation of side-chain residues enabled enhanced contacts with CD1d–α-GlcCer: Ser109α moved toward and formed a hydrogen bond with Asp80 of CD1d, and the Phe113α side chain was orientated to allow contacts with residues from both the α1 helix and α2 helix of CD1d (Fig. 7a). This Phe113α was in a position equivalent to that of Leu99α of the type I TCR (Fig. 6b,c), in which Leu99α was in a small hydrophobic niche formed by Leu84, Leu150 and Val149 of CD1d. Similarly, Phe113α was also in a hydrophobic niche formed by equivalent residues of CD1d. However, the longer side chain of Phe113α protruded further into the CD1d cleft, causing Leu84 and Leu150 to reorientate. The reorientation of Leu84 in turn pushed the tip of the sphingosine tail deeper into the F′ pocket of the binding groove compared with the CD1d-lipid binary complex (Fig. 7a). Thus, in contrast to the rigidity of the type I NKT cell TCR–CD1d–α-GalCer interaction, we observed a greater degree of plasticity for the V_α10 NKT cell TCR after interaction with CD1d–α-GlcCer.

Figure 7.

Lipid antigen specificity. (a) Overlay of V_α10 TCR not bound to a ligand and the binary complex of CD1d–PBS-25 (Protein Data Bank accession code, 1Z5L) on the V_α10 TCR–CD1d–α-GlcCer ternary complex: gray, CD1d in the V_α10 complex; magenta, α-GlcCer; salmon, V_α10 TCR in complex; blue, V_α10 TCR not bound to a ligand; light green, CD1d in binary complex; yellow, PBS-25 (α-GalCer analog with a shorter acyl chain). (b) Interactions with the V_α10 NKT cell TCR mediated by α-GlcCer: purple, CDR1α; dark green, CDR2α; yellow, CDR3α; magenta, α-GlcCer; gray, CD1d; blue, H₂O molecules; −OH^S, −OH on the sphingosine chain. (c) Interactions with the type I NKT cell TCR mediated by α-GalCer¹⁴: blue, α-GalCer; colors of CD1d and CDR loops as in b. Black dashed lines, hydrogen bonds; red dashed lines, van der Waals interactions.

Specificity of V_α10 TCR for α-GlcCer

The V_α10 NKT cell TCR interacted with α-GlcCer solely via the TCR α-chain (Supplementary Table 2), with the sugar directly under the CDR1α loop (Fig. 7b), which allowed the 2′ OH to form a hydrogen bond with Asn30α. In addition, Thr28α and Leu29α of the CDR1α loop made van der Waals interactions with α-GlcCer. Unexpectedly, CDR2α also made contact with the sugar group; Trp59α made van der Waals interactions with the 3′ OH, and there was an H₂O-mediated hydrogen bond between the OH group of Ser58α and the 4′ OH (Fig. 7b). In the CDR3α loop, Ser108α, Ser109α and Ser110α made van der Waals contacts with the glucose as well as the OH groups of the sphingosine tail. Further, the OH group of Ser110α made an H₂O-mediated hydrogen bond with 4′ OH of the sphingosine tail (Fig. 7b). These interactions, characterized by many van der Waals interactions and H₂O-mediated hydrogen bonds were in contrast to the type I NKT cell TCR–α-GalCer interactions, in which many polar OH groups of the galactosyl ring were sequestered by hydrogen bonds with residues from the CDR1α and CDR3α loops (Fig. 7c).

The main difference in the glycosyl head groups of α-GlcCer and α-GalCer was in the orientation of the 4′ OH group. The ‘downward-pointing’ glucosyl 4′ OH group contacted the main chains of Gly155 and Thr156 of CD1d, with Gly155 and its neighboring residues interacting with the CDR2α loop. Accordingly, replacement of α-GlcCer with α-GalCer (Fig. 7b) may have resulted in a destabilizing loss of interactions with CD1d. Moreover, an upward-pointing galactosyl 4′ OH group may have sterically clashed with Ser58α of the CDR2α loop of V_α10, a residue that formed an H₂O-mediated hydrogen bond with the glucosyl 4′ OH. Together these data provide a molecular basis for the differences between the type I NKT cell TCR and V_α10 NKT cell TCR in their fine specificity for glycolipid antigens.

DISCUSSION

In this study we have identified a unique population of NKT cells that fit neither the type I category nor the type II category; we have called these ‘V_α10 NKT cells’. These cells recognized α-GalCer presented by CD1d and were characterized by use of a semi-invariant α-chain (V_α10-J_α50) and a bias toward V_β8 expression. Although the development of V_α10 NKT cells remains to be investigated, the very limited junctional diversity and the absence of these cells in Cd1d^−/− mice suggest that they arise through TCR-mediated intrathymic selection events in a manner akin to that of type I NKT cells. In further support of that idea, we found that both CD4⁺ and CD4⁻ subsets of V_α10 NKT cells existed; they were CD44^hi, CD69^lo−int and NK1.1⁺, they produced large amounts of a range of cytokines, and they were present at greater frequency in liver than spleen (data not shown), which suggests a developmental pathway that makes them more like type I NKT cells than conventional T cells.

The V_α10 NKT cells had a hierarchy of antigen reactivity different from that of type I NKT cells²⁷, as demonstrated by their preferential recognition of α-GlcCer and other glucose-based glyco-lipids, including the microbial glucuronic acid–containing ligand α-GlcA-DAG from M. smegmatis²⁶ and, to a lesser extent, GSL-1 from Sphingomonas species^28–30. The finding that V_α10 NKT cells produced 10–100 times more interferon-γ, IL-4, IL-13 and IL-17 in response to α-GlcA-DAG than did type I NKT cells suggested that these cells are functionally distinct. Furthermore, the type of antigen may determine which population dominates the response and may also influence, through differences in cytokine production, the outcome of the response. Further studies are warranted comparing the response of V_α10 NKT cells with that of type I NKT cells in mouse-infection studies, especially with bacteria containing glucose-based α-linked glycolipids^26,28–30. Notably, human V_α24⁻J_α18⁺ type I NKT cells respond less well to α-GlcCer than to α-GalCer¹², which further discriminates between the V_α10 NKT cells defined here and human V_α24⁻J_α18⁺ NKT cells. Although CD1d-mediated intrathymic selection seems to impose limitations on TCR α-chain diversity, variations in αβTCR use nonetheless imbue CD1d-restricted T cells with a range of antigen specificities and cytokine-producing potential. The finding that these cells recognized and responded to the mammalian glycolipid antigen iGb3 also suggests that they have the potential for self-reactivity in a manner similar to that of iGb3-reactive type I NKT cells^29,31,32.

Although type I and V_α10 NKT cells expressed highly disparate TCR α-chains, they docked onto CD1d-glycolipid antigen in a very similar manner. Nevertheless, we observed key differences in the type I and V_α10 NKT cell TCR-CD1d-antigen interfaces, which included the CDR2α loop of the V_α10 NKT cell TCR in contact with the lipid antigen. Moreover, we observed differences between the type I and V_α10 NKT cell TCRs in the degree of structural plasticity after they engaged their respective CD1d-antigen complex. Furthermore, the CDR3α loop of the V_α10 NKT cell TCR interacted with CD1d-antigen mainly through van der Waals interactions, in contrast to the polar bond–mediated network of the CDR3α-mediated interactions of the type I NKT cell TCR^14,15. Collectively, the differences at the antigen-binding interface explain the differences between type I and V_α10 NKT cell TCRs in their antigenic specificity. The CDR3α loop is crucial for docking of type I NKT cell TCR–CD1d–antigen^{14,15,19–21,33}. However, the V_α10 NKT cell TCR docked in a similar manner, which suggests that it may be the NKT cell TCR β-chain that dictates such a conserved NKT cell TCR-CD1d docking topology. In particular, the CDR2β loop of mouse V_β8⁺ and human V_β11⁺ NKT cell TCRs has two tyrosine residues that form a conserved set of interactions with a defined region on CD1d. These observations have resonance with the V_β8.2-mediated interaction ‘codons’ observed in the major histocompatibility complex–restricted response, in which two conserved tyrosine residues are considered to dictate TCR–major histocompatibility complex bias^34,35. However, the V_β7 NKT cell TCR has only one tyrosine residue in the corresponding CDR2β loop yet docks onto CD1d in a conserved manner¹⁴. Similarly, mutagenesis studies of the V_β2 NKT cell TCR, which has no tyrosine residues in the CDR2β loop, is also predicted to dock onto CD1d in a conserved manner²¹. Thus, the conserved docking topology may be a consequence of the fact that an innate-like TCR must recognize a relatively monomorphic antigen-presenting molecule.

In summary, we have identified a previously unknown population of CD1d-restricted NKT cells with a unique canonical V_α10-J_α50 TCRα chain that recognize a partially but not completely overlapping array of glycolipids, including self and bacterial antigens. Our findings have also identified a feature of the interaction of TCRs with antigen-presenting molecules: although the diverse repertoire of αβTCRs allows them to bind peptide–major histocompatibility complex molecules in a range of docking modes³⁶, the diversity of the NKT cell TCR repertoire is associated with a conserved CD1d docking mode. Accordingly, our findings highlight a fundamental difference between peptide- and lipid-mediated recognition by the TCR.

ONLINE METHODS

Mice

C57BL/6 and BALB/c mice were derived from the Peter MacCallum Cancer Centre or were from the Walter and Eliza Hall Institute of Medical Research. The following mice were bred at Peter MacCallum Cancer Centre: BALB/c CD1d1.CD1d2-deficient mice (BALB/c Cd1d^−/−; 11 backcrosses; from Jackson Laboratories)³⁷; BALB/c J_α18-deficient mice (BALB/c J_α18^−/−; 10 backcrosses; originally provided by M. Taniguchi)⁵; C57BL/6 Jα18-deficient mice (C57BL/6 J_α18^−/−; 12 backcrosses; originally provided by M. Taniguchi)⁵; C57BL/6 CD1d-deficient mice (C57BL/6 Cd1d^−/−; 11 backcrosses; originally provided by Jackson Laboratories)³⁸. Mice over 6 weeks of age were used in all experiments, which were done in accordance with the Animal Ethics and Experimentation Committee of Peter MacCallum Cancer Centre.

Flow cytometry

Single-cell suspensions of thymus, spleen and liver were made as described³⁹. Where indicated, thymocyte samples were enriched for NKT cells by depletion of CD8⁺ and CD24⁺ thymocytes as described³⁹. Monoclonal antibody to CD16-CD32 (FcR block; 2.4G2; grown in-house) was included in cell-labeling experiments. Cells were stained with the following antibodies (all from BD Pharmingen): antibody to CD3 (anti-CD3; 145-2C11) anti-CD4 (RM4-5), anti-CD8 (53-6.7), anti-CD44 (IM7), anti-CD49b (HMα2), anti-CD69 (H1.2F3), anti-NK1.1 (PK136), anti-TCRβ (H57-597), anti-V_β2 (B20.6), anti-V_β7 (TR310), anti-V_β8.1-V_β8.2 (MR5-2), anti-V_β8.3 (1B3.3) or rat IgG2b isotype-matched control antibody (A95-1). Mouse CD1d tetramers (produced in-house with a baculovirus-based CD1d expression system originally derived from M. Kronenberg) were loaded with α-GalCer (from Alexis Biochemicals or P. Savage (C_24:1 PBS-44 analog)). In some cases, CD1d tetramer was loaded with α-GlcCer (C_20:2 DB06-15 analog^24,25). In some experiments, 7-amino-actinomycin D (Molecular Probes) and/or unloaded mouse CD1d tetramers were used for the exclusion of dead and nonspecifically stained cells. Cells were analyzed with an LSR II or FACSCanto (BD Biosciences) and data were processed with FlowJo (Tree Star), with bi-exponential scaling.

In vitro stimulation

For in vitro proliferation assays, two approaches were used. In one approach (Fig. 3a), thymocyte samples were depleted of CD8⁺ and CD24⁺ cells³⁹ and labeled for 10 min at 37 °C with 2 µM CFSE (carboxyfluorescein diacetate succinimidyl ester; Molecular Probes), then the thymocytes (1.5 × 10⁵) were cultured for 72 h with 3 × 10⁵ BALB/c J_α18^−/− splenocytes that had been pulsed the night before with α-GalCer (C_26:0 or C_20:2; 100 ng/ml), α-GlcCer (C_20:2; 100 ng/ml)^24,25, β-GalCer (C₁₂; 10 µg/ml; Avanti Polar Lipids) or iGb3 (C_26:0; 10 µg/ml; Alexis Biochemicals). Glycolipid stock solutions were dissolved in either dimethyl sulfoxide or a solution of 0.5% (vol/vol) Tween-20, sucrose (56 mg/ml) and l-histidine (7.5 mg/ml) and were sonicated for ~15 min immediately before use. For the second approach (Fig. 3b–d), purified αβTCR⁺ thymic NKT cells positive for CD1d–α-GalCer tetramer (1 × 10³ to 4 × 10³ cells) were labeled with 1 µM CFSE and were cultured for 72 h in 50 µl culture medium in the presence of 2 × 10⁴ splenic CD11c⁺ dendritic cells (from BALB/c J_α18^−/− mice) with or without the following synthetic glycolipids: C_26:0 α-GalCer (500 ng/ml), α-GlcCer (500 ng/ml), M. smegmatis α-GlcA-DAG (0.6–10 µg/ml; either separate or a mixture of three analogs (C_18:0-C_16:0, C_19:0-C_16:0 and C_16:0-C_19:0) at a ratio of 1:1:1 (wt/wt/wt); structures, Supplementary Fig. 1), Sphingomonas species GSL-1 (C_14:0 α-glucuronosyl ceramide⁴⁰; 1 µg/ml; National Institutes of Health Tetramer Core Facility) or iGb3 (10 µg/ml; Alexis Biochemicals). Cells were cultured in RPMI-1640 medium (Invitrogen Life Technologies) supplemented with 10% (vol/vol) FCS (JRH Biosciences), penicillin (100 U/ml), streptomycin (100 µg/ml), Glutamax (2 mM), sodium pyruvate (1 mM) and nonessential amino acids (0.1 mM) and HEPES buffer (15 mM), pH 7.2–7.5 (all from Invitrogen Life Technologies), plus 50 µM 2-mercaptoethanol (Sigma-Aldrich). Culture supernatants were analyzed for cytokines with a Cytometric Bead Array Mouse Flex Set (BD Biosciences).

Synthesis of α-GlcA-DAG (C_19:0-C_16:0)

Three glycolipid variants derived from M. smegmatis²⁶ were synthesized: 1-O-((R)-10-tuberculostearyl)-2-O-palmityl-sn-glyceryl α-D-glucopyranosiduronic acid (C_19:0-C_16:0;) was prepared from 1-O-tert-butyldiphenylsilyl-2-O-p-methoxybenzyl-sn-glycerol and 2,3,4-tri-O-benzyl-6-O-acetyl-α-d-glucopyranosyl iodide, followed by C6-oxidation and esterification (B.C. et al., personal communication). Hydrogenolysis of 1-O-((R)-10-tuberculostearyl)-2-O-palmityl-sn-glyceryl (benzyl 2,3,4-tri-O-benzyl-α-d-glucuronate) gave the title compound [α]_D²³ +39.2 (c 0.25 in dimethyl sulfoxide), δ_H (500 MHz; dimethyl sulfoxide) 0.81 (d, 3H), 0.85 (t, 6H), 1.03–1.23 (m, 51H), 1.50 (m, 4 H), 2.24–2.31 (m, 4 H), 3.22 (dd, 1H), 3.55 (dd, 1H), 3.69 (dd, 1H), 3.77 (d, 1H), 4.16 (dd, 1H), 4.32 (dd, 1H), 4.69 (d, 1H), 5.11 (m, 1H); δ_C (125 MHz; dimethyl sulfoxide) 13.91, 19.55, 22.09, 24.41, 24.44, 26.40, 26.44, 28.43, 28.45, 28.69, 28.71, 28.76, 28.96, 29.02, 29.07, 29.36, 29.37, 31.29, 32.06, 33.42, 33.53, 36.42, 36.44, 62.19, 65.48, 69.58, 71.38, 71.76, 72.61, 99.37, 171.06, 172.25, 172.53; high-resolution mass spectrometry–electrospray ionization [M + Na]⁺ calculated for C₄₄H₈₂O₁₁, 809.5749; found a mass/charge ratio of 809.5739.

Single-cell PCR and RT-PCR

For analysis of genes encoding V_α regions, RNA was extracted from sorted NKT cells with an RNeasy kit (Qiagen) and cDNA was synthesized with oligo(dT₁₅) primers (Promega) and AMV reverse transcriptase (Promega) in accordance with the manufacturers’ instructions. NKT cell cDNA was amplified by PCR with a panel of V_α-specific primers⁴¹ and GoTaq Master Mix (Promega). For single-cell PCR, cDNA from sorted αβTCR⁺ V_β8⁺ thymic NKT cells positive for CD1d–α-GalCer tetramer was generated⁴² and then was amplified by two rounds of semi-nested PCR⁴³ with sense primers for V_α10 (external, 5′-CCTTGGTTCTGCAGGAGGGGGAG-3′, and internal, 5′-AACGTCGCAGCTCTTTGCAC-3′), V_α14 (external: 5′-TTGTCCGTCAGGGAGAGAACTGC-3′, and internal: 5′-GACACAGGCAAAGGTCTTGTGTCC-3′) or C_α (5′-GAACCTGCTGTGTAC-3′), each used with the C_α antisense primer (both rounds; 5′-TGGCGTTGGTCTCTTTGAAG-3′). PCR products were separated by electrophoresis through a 1.5% agarose gel and were sequenced as described⁴³. TCR residue numbering is in accordance with the International ImMunoGeneTics database⁴⁴.

Cloning and expression of genes encoding mouse NKT cell TCRs

Paired TCRα and TCRβ sequences from a single V_α10 NKT cell (Table 1, sequence 1 for both TCRα and TCRβ) were used to design a chimeric transcript that also contained a modified human C_α domain to aid in refolding¹⁴. Sequences codon-optimized for Escherichia coli were synthesized (Genscript) and then transferred into a pET30 expression vector (Novagen). Soluble TRAV13D-3-TRAJ50–TRBV13-3-TRBJ2-7 (V_α10-J_α50–V_β8.1-J_β2.7) and TRAV11-TRAJ18–TRBV13-2 (V_α14-J_α18–V_β8.2-J_β2.1) NKT cell TCRs were expressed in E. coli strain BL21 and inclusion body proteins were prepared, refolded and purified by gel filtration as described⁴⁵.

Transfection of 293T cells with TCR

Full-length TCRα transcripts TRAV13D-3 (V_α10; CDR3α sequence 1, Table 1) and TRAV11 (V_α14) were cloned from V_α10 and type I NKT cell cDNA, respectively, by RT-PCR. Full-length TCRβ (TRBV13-3-TRBJ2-7; V_β8.1-J_β2.7; CDR3β sequence 1, Table 1) was synthesized (Genscript), and TRBV13-1 (V_β8.3) and TRBV29 (V_β7) were generated from H-2D^b-restricted influenza A–specific clones (E.B.D. et al., personal communication). TCR sequences were verified and inserted into the plasmid pMIGII, and 293T cells were transfected with hydrolase element P2A–linked genes encoding CD3, TCRα and TCRβ as described⁴⁶. After 48 h of culture to allow expression of proteins, cells positive for the expression of green fluorescent protein were analyzed for surface expression of TCR by flow cytometry.

Surface plasmon resonance

A ProteOn XPR36 protein-interaction array system (Bio-Rad) was used for surface plasmon resonance as described¹⁴. Streptavidin was coupled to a GLC Sensor Chip (Bio-Rad) by amine coupling (~500 response units), and biotinylated mouse CD1d–α-GalCer, CD1d–α-GlcCer or unloaded CD1d was captured on a separate flow cell for each (~600 response units each). Soluble TCRs were serially diluted and simultaneously injected over test and control surfaces at a rate of 30 µl/min. After subtraction of data from the control flow cell (unloaded CD1d), interactions were analyzed with ProteOn Manager software version 2.1 (Bio-Rad) and the Scrubber 2.0a program (Prot version; BioLogic Software), and steady-state dissociation constants were derived at equilibrium.

Crystallization, structure determination and refinement

The V_α10 TCR–CD1d–α-GlcCer ternary complex and the V_α10 TCR not bound to ligand (both at a concentration of 10 mg/ml in 10 mM Tris, pH 8.0, and 150 mM NaCl) were crystallized at 18 °C in 20% (vol/vol) PEG 1500 and 0.1 M 2-(N-morpholino)ethanesulfonic acid, pH 6.5, and in 20% (vol/vol) PEG 3350, 0.2 M NaI and 0.1 M bis-Tris-propane, pH 6.5 (Hampton Research), respectively. Crystals were flash-frozen before data collection in crystallization solution containing 25% (vol/vol) glycerol as a cryoprotectant. Data for the ternary complex and the TCR were collected at 100 K on the MX1 beamline and MX2 beamline, respectively, at the Australian Synchrotron (Melbourne). The ternary complex crystals diffracted to 2.2Å and belonged to space group P2₁ with two ternary complexes in each asymmetric unit, with an r.m.s.d. of 0.57Å over 784 Cα atoms. Thus, structural analysis was restricted to one complex in the asymmetric unit. The crystals of the TCR not bound to a ligand diffracted to 2.9Å and belong to the space group C2, with two TCR molecules in each asymmetric unit (Supplementary Table 1). Data for the ternary complex were processed with the Mosflm program for integrating single-crystal-diffraction data from area detectors (version 7.0.5)⁴⁷ and were scaled with the SCALA scaling and data-merging program from the CCP4 Suite (Collaborative Computational Project, number 4)⁴⁸. The complex was solved by the molecular-replacement method with the MOLREP program in CCP4, with V_α14–Vα8.2 TCR minus the CDR loops (Protein Data Bank accession code, 3HE6) and mouse CD1d minus the lipid (Protein Data Bank accession code, 1Z5L) as the search models. The electron density at the TCR–CD1d–α-GlcCer interface was unambiguous and the initial experimental phases showed an unbiased density for α-GlcCer in the antigen-binding cleft of CD1d. The data for the TCR not bound to a ligand were processed and scaled with the HKL-2000 program package⁴⁹. The structure of the TCR was solved by the molecular-replacement method with the Phaser crystallographic software in CCP4, with V_α14-V_α8.2 NKT cell TCR minus the CDR loops (Protein Data Bank accession code, 3HE6) as a search model. COOT (Crystallographic Object-Oriented Toolkit)⁵⁰ was used for model building (Supplementary Table 1). The quality of both structures was confirmed at the Research Collaboratory for Structural Bioinformatics Protein Data Bank Data Validation and Deposition Services website. All presentations of molecular graphics were created with the PyMOL molecular visualization system. For the NKT cell TCR not bound to a ligand, all CDR loops were modeled except CDR2α, which was also involved in crystal contacts and was thus excluded from analysis. For structural analysis, although the structure of the binary complex of CD1d–α-GlcCer was not available, the degree of plasticity in CD1d after engagement of the V_α10 TCR was evaluated by comparison of the V_α10 TCR–CD1d–α-GlcCer in complex with CD1d linked to PBS-25, an analog of α-GalCer with a shorter (C₈) acyl chain⁵¹.