Distinct CD1d docking strategies exhibited by diverse Type II NKT cell receptors

Type I and type II natural killer T (NKT) cells are restricted to the lipid antigen-presenting molecule CD1d. While we have an understanding of the antigen reactivity and function of type I NKT cells, our knowledge of type II NKT cells in health and disease remains unclear. Here we describe a population of type II NKT cells that recognise and respond to the microbial antigen, α-glucuronosyl-diacylglycerol (α-GlcADAG) presented by CD1d, but not the prototypical type I NKT cell agonist, α-galactosylceramide. Surprisingly, the crystal structure of a type II NKT TCR-CD1d-α-GlcADAG complex reveals a CD1d F’-pocket-docking mode that contrasts sharply with the previously determined A’-roof positioning of a sulfatide-reactive type II NKT TCR. Our data also suggest that diverse type II NKT TCRs directed against distinct microbial or mammalian lipid antigens adopt multiple recognition strategies on CD1d, thereby maximising the potential for type II NKT cells to detect different lipid antigens.

N atural killer T (NKT) cells recognise lipid-based antigens (Ag) presented by the major histocompatibility class I (MHC I)-like molecule CD1d (reviewed in ref. 1 ). There are two broad classes of NKT cells, defined as type I and type II NKT cells. The most extensively studied are the type I NKT cells, which express an invariant TCR α-chain (Vα14-J18 in mice, Vα24-Jα18 in humans and strongly respond to the glycolipid αgalactosylceramide (α-GalCer)) (reviewed in ref. 1 ). Upon activation, type I NKT cells rapidly produce an array of cytokines and are implicated in a broad range of diseases through their ability to modulate Ag-presenting cells and effector T, B and natural killer (NK) lymphocytes (reviewed in ref. 1 ). Type II NKT cells are also CD1d restricted, but they are distinguished from type I NKT cells because they neither express the invariant TCR α-chain that characterises type I NKT cells nor do they recognise α-GalCer (reviewed in refs. 1,2 ). The specificity and function of type II NKT cells is poorly understood although some evidence suggests that they play an immunosuppressive role that may oppose the function of type I NKT cells 3,4 (and reviewed in refs. 1,2 ). Given that type II NKT cells appear to be more abundant in humans than their type I NKT cell counterparts [5][6][7] (and reviewed in ref. 8 ), it is important to understand the molecular mechanisms underpinning their antigen reactivity and specificity.
The availability of α-GalCer and CD1d-α-GalCer tetramers 9 has facilitated the study of type I NKT cells and the development, functional potential, and molecular basis for Ag-recognition by these cells is now well understood (reviewed in ref. 1 ). Conversely, a limited understanding of the Ags recognised by type II NKT cells has hampered the study of these cells, although some studies have identified Ags recognised by some type II NKT cell lines, including sulfatide, β-glucosylceramide (β-GlcCer), phosphatidylglycerol (PG), diphosphatidylglycerol (DPG), lysophosphatidylcholine (LPC) and lysophosphatidylethanolamine (LPE) [10][11][12][13][14][15][16][17][18][19][20] . Structural studies have determined how a type II NKT TCR can bind to sulfatide or lysosulfatide presented by mouse CD1d 21,22 , and revealed a docking mode that was distinct to the type I NKT TCR-CD1d-α-GalCer interaction (reviewed in ref. 23 ). Namely, type I NKT TCRs bind with a parallel docking mode over the F′-pocket of CD1d, with the TCR α-chain dominating the interaction with the galactose headgroup, while the TCR β-chain binds solely to CD1d (reviewed in ref. 23 ). In contrast, the sulfatide-reactive type II NKT TCR docked orthogonally over the A′-pocket of CD1d with only the CDR3β loop interacting with the sulfatide headgroup 21,22 . These studies suggests that the structural and molecular basis for CD1d-Ag recognition between type I and type II NKT cells is fundamentally distinct (reviewed in refs. 1,24 ). However, it remains to be determined whether this represents a general dichotomy that distinguishes the mode of Ag recognition by type I and type II NKT cells. We previously identified an atypical population of CD1d-restricted NKT cells that utilise a Vα10Jα50 TCR α-chain that recognise α-GalCer and α-GlcCer. Some of these atypical NKT cells were activated by a mycobacterial Ag, α-glucuronosyl-diacylglycerol (α-GlcADAG) 25 presented by CD1d.
Here, using α-GlcADAG-loaded CD1d tetramers, we characterise a population of type II NKT cells that are selectively reactive to the α-GlcADAG Ag presented by CD1d and do not recognise α-GalCer. Using a panel of synthetic α-GlcADAG analogues, we demonstrate the key moieties of this molecule that facilitate type II NKT TCR recognition. Structural analysis of a type II Vβ8.2 + NKT TCR-CD1d-α-GlcADAG complex revealed a parallel docking mode over the F′-pocket of CD1d, which contrasted the type II NKT TCR-CD1d-sulfatide ternary complexes 21,22 . Extensive CD1d mutational analysis against a panel of diverse self and microbial-antigen-reactive type II NKT TCRs reveal that type II NKT cell TCRs can adopt differing docking strategies to engage specific CD1d-Ag complexes. Thus, this study defines how a microbial-Ag is recognised by a type II NKT TCR, and demonstrates that type II NKT TCRs can bind with diverse docking modes to CD1d-Ag. This highlights the diverse lipid-based Ag recognition by type II NKT TCRs.
Diverse CD1d-α-GlcADAG tetramer + NKT TCRs. We next determined the TCR sequences used by the CD1d-α-GlcADAG tetramer + cells that were sorted as single cells from both wt and Jα18 −/− BALB/c thymi, following tetramer-associated magnetic enrichment (TAME) based on gates depicted in Fig. 1d and Supplementary Fig. 1b. CD1d-α-GalCer + CD1d-α-GlcADAG tetramer − type I NKT cells from wt mice were also sorted as controls. Single cell TCR αand TCR β-chain paired analysis was performed using multiplex PCR, as previously described 26 (Supplementary Table 1). CD1d-α-GalCer tetramer + cells are known to express the canonical Vα14Jα18 + type I NKT TCR α-chain rearrangement 27 . In contrast, approximately half (12 out of 25 paired TCR sequences) of the CD1d-α-GlcADAG tetramer + sorted cells expressed Vα10Jα50 TCR α-chain rearrangements, similar to the Vα10 + NKT cells present in Jα18 −/− mice that we previously described 25 . Interestingly, four CD1d-α-GlcADAG tetramer + clones from wt BALB/c mice expressed a TCR α-chain in which the Jα50 gene was rearranged with Vα15, Vα4, Vα11, or Vα4.4, instead of the Vα10 gene. These TCR α-chains displayed little or no homology in their CDR1α and CDR2α regions, yet possessed highly similar CDR3α regions suggesting that the Jα50encoded region confers CD1d-α-GlcADAG recognition in the context of different Vα gene usage. This may be due to the conservation of the CDR3α residues Ser109, Ser110 and Phe113 in each of these TCRs, three residues that were involved in the recognition of CD1d-α-GlcCer complexes by Vα10Jα50 + NKT TCRs 25 . The different CDR1α and CDR2α loops may also facilitate CD1d binding in different ways. Indeed, in a previous study we demonstrated that a Vα10Jα50 + NKT TCR utilised residues within the CDR1 and CDR2 loops to establish contact with CD1d, whilst conserved CDR3α residues contacted both CD1d and the antigen 25 . Interestingly two unique TCR α-chain sequences that did not express Vα14, Jα18 or Jα50 gene Pre TAME Post TAME segments (Vα4Jα17, Vα8Jα49) were also identified amongst wt thymocytes sorted with CD1d-α-GlcADAG tetramers (Supplementary Table 1). TCRβ sequence analysis indicated diversity within the composition and length of the CDR3β regions amongst all sorted populations, and they almost exclusively used Vβ8 (24 out of 25 paired TCR sequences), consistent with predominant Vβ8 gene usage within type I NKT cells 9 , Vα10Jα50 + NKT cells 25 , and type II NKT cells 28 . Some TCR sequences with identical TCRα and TCRβ nucleotide homology were identified in independently sorted cells (Supplementary Table 1), suggesting clonal expansion in vivo. Similar results were obtained when NKT cells were isolated from Jα18 −/− BALB/c mice (Supplementary Table 1). In the absence of Jα18, Vα10Jα50 TCRs were prominent within the population that stained with CD1d-α-GlcADAG tetramers (5 out of 11 paired TCR sequences), and all used the Vβ8 gene. Five unique TCR α-chain sequences that did not express Vα14, Vα10 or Jα50 gene segments (Vα11Jα9, Vα5Jα12, Vα3Jα4, Vα17Jα2, Vα8Jα16) were observed amongst Jα18 −/− thymocytes sorted with CD1d-α-GlcADAG tetramers (Supplementary Table 1). Interestingly, the Vα11Jα9-Vβ8.2 TCR (designated clone name: A11B8.2) was detected in 12 out of 16 sorted clones. As these NKT TCRs displayed non-germline-encoded residues in both the CDR3α and CDR3β regions, this suggests that these T cells had expanded in vivo prior to their isolation.
Overall, the population of thymocytes identified by CD1d-α-GlcADAG tetramers was characterised by heterogeneous TCR usage in both Jα18 −/− and wt BALB/c mice, with a range of TCR α-chain sequences and an almost exclusive use of the Vβ8 chain. Moreover, these results demonstrate a means by which microbial Ag-specific type II NKT cells can be isolated.
As expected, the XV19 type II NKT TCR + cell line showed clear staining with sulfatide-loaded CD1d tetramer whereas the other CD1d tetramers did not stain these cells at higher levels than CD1d-endo (Fig. 2a). The VB8-STD type I NKT TCR + cell line was stained by CD1d-α-GlcADAG tetramer with an intensity that was between CD1d-endo tetramer and CD1d-α-GalCer tetramers (Fig. 2a). This shows that some type I NKT cell clones can recognise the microbial Ag α-GlcADAG in the context of CD1d, consistent with the detection of a subset of type I NKT cells by CD1d-α-GlcADAG tetramer (Fig. 1c, d) and the presence of Vα14Jα18 rearrangements amongst cells that costained with CD1d-α-GlcADAG and CD1d-α-GalCer tetramers (Supplementary Table 1).
Collectively, these results highlight the existence of diverse NKT cells expressing TCRs that can interact with CD1d-α-GlcADAG including subsets of type I, type II, and atypical NKT cells.
CD1d-α-GlcADAG complexes can activate NKT TCR + cells. Plate bound CD1d-lipid Ag activation assays were carried out to investigate if the TCR + cell lines could respond to CD1d-α-GlcADAG recognition (Fig. 2b). The A11B8.2 type II, the A10B8.2 atypical, the A10B8.3 TCR + atypical, and the VB8-STD type I NKT TCR + cell lines, were activated by α-GlcADAGloaded CD1d, leading to CD69 up-regulation and IL-2 production at levels above those elicited by CD1d-endo. Consistent with the tetramer staining ( Fig. 2a) the A10B8.2, A10B8.3 atypical NKT TCR + and the controls A10B8.1 atypical NKT TCR + and VB8-STD type I NKT TCR + cell lines responded more strongly to stimulation with CD1d-α-GalCer, while this induced only a weak response from A11B8.2 and A17B8.2 type II NKT TCR + cells. Notably, and in contrast to A11B8.2, A17B8.2, or VB8-STD NKT TCR lines, no response was detected by the A10B8.1, A10B8.2, and A10B8.3 atypical NKT TCR + cell line following stimulation with CD1d-endo, implying that these TCRs are highly Ag-dependent (Fig. 2b). Thus, CD1d-α-GlcADAG is capable of stimulating cells expressing both classical type I, type II, and atypical NKT TCRs, in an Ag-dependent manner. This provides the first example of a microbial Ag that can stimulate all three classes of NKT cells.
For the A11B8.2 type II NKT TCR + cell line, of the α-GlcADAG variants, only the bona fide mycobacterial species (R-C19:0/C16:0) was clearly bound by the TCR at levels greater than CD1d-endo (Fig. 3a). This type II NKT cell line showed no binding to any of the other CD1d tetramer-Ag complexes, which highlights the fine specificity of this type II NKT cell-derived TCR, and also the importance of both the polar head group and the acyl chain composition and regioselectivity for recognition of this Ag. In contrast, the A10B8.2 atypical NKT TCR + cell line bound to CD1d tetramers loaded with α-GlcADAG R-C19:0/ C16:0 and R-C16:0/C19:0 analogues, but not to the C18:0/C16:0 analogue (Fig. 3a). As expected 25 , this atypical NKT cell line stained strongly with both α-GalCer and α-GlcCer-loaded CD1d tetramers and also showed moderate binding to α-GlcDAG from S. pneumoniae and both the Sphingomonas-derived lipids α-GlcACer and α-GalACer (Fig. 3a). This cell line showed no appreciable binding to tetramers loaded with sulfatide, GD3 or 'endo'.
We investigated more analogues of α-GlcADAG including a C18:1/C16:0 variant that has been isolated from M. smegmatis and Corynebacterium glutamicum, in which the TBSA is replaced by oleic acid bearing a double bond between C9 and C10 [40][41][42][43] ( Supplementary Fig. 2 and Fig. 3b). This is representative of related compounds that have been isolated from the pathogenic fungus Aspergillus fumigatus 44 . We also synthesised and investigated analogues with R-or S-versions of TBSA and oleic acid-containing α-GlcADAG or α-GlcDAGs, which lack the glucuronic acid on their glycosidic headgroup (Supplementary Fig. 2 and Fig. 3b). The strongest staining of the type II A11B8.2 NKT TCR + cell line was achieved with CD1d tetramers loaded with the α-GlcADAG variant bearing the R-TBSA chain (R-C19:0/C16:0). Moderate staining of this line was achieved by the oleic-acid (C18:1/C16:0) followed by the S-TBSA variant (S-C19:0/C16:0), and to a much lower degree by the 'isooleoyl' regioisomer α-GlcADAG (C16:0/C18:1) (Fig. 3b). Furthermore, IL-2 was produced in response to plate-bound CD1d loaded with R-and S-TBSA variants of C19:0/C16:0, C18:1/C16:0 or C16:0/ C18:1 α-GlcADAGs (Fig. 3b), which demonstrates that these variants can also elicit functional responses of type II NKT cells. Notably, none of the variants lacking the uronic acid on the glycosidic group was able to elicit responses by the A11B8.2 NKT TCR + cell line nor allow CD1d-tetramer staining, highlighting a high degree of specificity of the A11B8.2 TCR towards the glucuronic acid. These experiments show that the A11B8.2 TCR has the ability to recognise and respond to α-GlcADAG Ags not only from M. smegmatis but also other α-glucuronosyl-producing species, such as those in C. glutamicum and A. fumigatus.
Type II NKT TCR/CD1d-α-GlcADAG interactions. The centrally positioned CDR1α loop mediated contacts with both α1-and α2-helices of CD1d, whereby Lys31α intersected both helices by forming a salt bridge with Asp80 and van der Waals (vdw) contacts with Asp153 (Fig. 5a), while Thr29α interacted with Val72 of CD1d. The CDR2α loop was positioned directly above the α2-helix of CD1d, in which Ala51α and Gln52α contacted solely Gln154 via vdw and hydrogen bond interactions, respectively (Fig. 5a)  hydrophobic pocket lined by Leu84, Val149, and Leu150 (Fig. 5a). Further, Met93α and Gly94α from the CDR3α loop contributed to the interface by contacting Val75, Ser76, Arg79, Asp80, and Glu83 via vdw contacts (Fig. 5a). Within the TCR β-chain, the CDR2β and CDR3β loops dominated the interactions at the interface. Here, the CDR2β loop sat peripherally above the α1-helix of CD1d, whereby Tyr50β contacted the side chains of Met87 and Glu83, while the adjacent framework residues Tyr48β and Glu56β also made contacts with Lys86 and Glu83 or Lys86 and Arg21 (respectively) (Fig. 5b, Supplementary Table 3).
Notably, the Vβ8.2 in the type I NKT TCR contains equivalent tyrosine residues in the CDR2β loop (Tyr48β and Tyr50β) that contacted similar residues in CD1d 31 . Further, the CDR3β loop reached across the external side of the CD1d cleft whereby residues Pro96β, Gln97β, Val99β, Ser100β, Tyr101β interacted exclusively with the α2-helix (Leu145, Lys148, Val149, Ala152, and Asp153) (Fig. 5b). The aromatic moiety of Tyr101β pointed towards the CD1d cleft to form hydrogen bonds with the main chain of Val149 and Ala152 (Fig. 5b). Accordingly, the nongermline-encoded CDR3 loops dominated the interactions at the A11B8.2 TCR/CD1d-α-GlcADAG interface (Figs. 4c, 5a, b). This CDR3 loop-mediated dominance at the CD1d-Ag interface was reminiscent of the type II NKT TCR-CD1d-sulfatide complex 21,22 and thus in clear contrast to the type I NKT TCR-CD1d-Ag complex, which is dominated by the germline encoded CDR1α and CDR2β loops and the invariant CDR3α loop (Fig. 4c) 31 . Collectively, the structure of the type II A11B8.2 NKT TCR-CD1d-α-GlcADAG ternary possessed features that were markedly distinct from the type I and type II NKT TCR-CD1d-Ag complexes that were previously characterised structurally.
Here, Lys31α hydrogen bonded to the carbonyl of the stearoyl chain, whilst Thr29α interacted via vdw contacts with the α-GlcADAG carbohydrate moiety. The positioning of the A11B8.2 TCR α-chain indicated that the CDR1α loop would sterically clash with the headgroups of a number of CD1d lipid Ags, including α-GalCer, sulfatide and microbial-derived ligands, consistent with the lack of staining of the A11B8.2 TCR + cell line for CD1d tetramers loaded with such Ags (Figs. 2a and 3a).
Overall, α-GlcADAG was involved in a coordinated series of interactions, with both CD1d and the A11B8.2 TCR, which allowed it to acquire a unique position where the headgroup of this ligand was tilted towards the A′-pocket of CD1d, which is in contrast to binary and ternary complexes with other CD1dbinding Ags. Thus, the A11B8.2 type II NKT TCR-CD1d-α-GlcADAG ternary complex reveals a distinct mechanism of lipid recognition and provides the first structural insight to diverse mechanisms of CD1d-Ag recognition amongst the type II NKT TCRs.
Type II NKT TCRs adopt multiple CD1d-binding modes. To more broadly investigate the docking mode of diverse type II NKT TCRs with CD1d, we generated soluble CD1d molecules with 18 single residue mutations across the length of the Agbinding cleft and tested their ability to stimulate a panel of NKT cell lines (Fig. 6). These mutants were chosen based on the available CD1d crystal structures, where residues selected were solvent exposed and most did not contact the lipid antigens. These six type II NKT cell lines exhibited diverse TCR usage and reactivity towards self-lipids or microbial lipids, thereby providing a broad perspective on the range of CD1d-docking modes that type II NKT TCRs could potentially employ. As a control, we also examined impact of these mutants on a type I NKT cell line (VB8-STD). Consistent with previous studies, the type I NKT cell line (VB8-STD) was impacted by the CD1d mutations: Arg79Ala, Glu83Ala, Lys86Ala, Met87Ala, Leu145Ala and Val149Ala. These residues are localised around the F′-pocket and mediate contacts with the type I NKT cell TCRs (Fig. 6) 31 . Further, and consistent with the crystal structure, mutational analyses indicated that Met87, Glu83, Lys86 and Val149 were critical for CD1d-Ag recognition by the A11B8.2 TCR (Fig. 6), and that Leu145, Thr159 and Ala152 also played a moderate role. Thus, similar to type I NKT TCRs, CD1d recognition by the A11B8.2 NKT TCR appears to be highly dependent on CDR2β interactions with the CD1d residues that comprise the "energetic hot spot" 31,49,50 . Similarly, the A17B8.2 TCR + cell line also showed an interaction map that correlates with a docking position over the F′-pocket of Fig. 3 Fine Ag-specificity of BW58 cell lines expressing CD1d-α-GlcADAG-reactive NKT TCRs. a BW58 cell lines expressing the A11B8.2 and A10B8.2 TCRs were tested by flow cytometry for their ability to bind a panel of CD1d tetramers loaded with: Mycobacteria smegmatis natural occurring α-GlcADAG species (R-C19:0/C16:0) and two synthetic analogue variants (C16:0/C19:0) and (C18:0/C16:0); Sphingomonas spp. α-GlcACer C14:0 (GSL-1); Sphingomonas spp. α-GalAcer C14:0 (GSL-1′); Streptococcus pneumoniae α-GlcDAG C16:0/C18:1; Borrelia burgdorferi α-GalDAG C17:1/C16:0; α-GlcCer C20:2; α-GalCer C:20:2; sulfatide C24:1 and GD3. CD1d-endogenous and vehicle-loaded CD1d tetramers were included as controls. The VB8-STD type I NKT TCR + and the XV19 type II NKT TCR + cell lines were also included as a control. Graphs represent the mean fluorescence intensity (MFI) of CD1d tetramer staining within cells with similar TCR expression from duplicate values of n = 2 (n = 3 for A11B8.2) independent experiments and SEM. b Synthetic versions of the naturally occurring forms of α-GlcADAG bearing a methyl group with R-enantiomer or S-enantiomer (R-C19:0/C16:0 and S-C19:0/16:0) or the oleic version containing a double bond between C9 and C10 (C18:1/C16:0) or the isooleic variant (C16:0/C18:1), were loaded into CD1d tetramers and assessed for their ability to stain the A11B8.2 type II NKT TCR + cell line (left), or to induce IL-2 following 16 h culture on platebound CD1dlipid (right). Variants in which the glucoronic headgroup (-GlcA) was substituted for a glycosydic headgroup (-Glc) were also tested. S. pneumoniae α-GlcDAG (SPN α-GlcDAG), α-GalCer, vehicle-loaded and endogenous-loaded CD1d tetramers were included as controls. Graph on left represents the MFI of CD1d tetramer fold increase over CD1d-endogenous staining within cells with similar TCR expression from four independent experiments ± SEM, and graph on right depicts the mean IL-2 production of three independent experiments ± SEM CD1d, similar to the A11B8.2 type II NKT TCR, with mutations Arg79Ala, Glu83Ala, Lys86Ala, Val149Ala, Thr159Ala causing a major reduction in activation. For the A11B8.2 and A17B8.2 TCRs, it is possible that the Thr159Ala mutation was indirectly effecting binding by impacting on lipid loading as this residue formed H-bonds with the α-GlcADAG headgroup (Fig. 5).
In contrast, for the type II XV19 TCR + cell line, the CD1d mutations that impacted on cellular activation were localised to the A′-pocket (His68Ala, Val72Ala, Thr159Ala, Asp166Ala and Lys65Ala, Gln62Ala, Leu170Ala, and Ser76Ala, Met162Ala) (Fig. 6) which correlates with the contact residues identified in the crystal complex structure of the XV19 TCR-CD1d-sulfatide 21,22 . The CD1d autoreactive 14S.6, TBA7 and VII68 type II NKT TCR + cell lines were assessed using CD1d mutants loaded with 'endo' Ags ( Fig. 6) 10,20,51,52 . For the 14S.6 TCR + cell line, the interaction map suggested that this TCR spanned across both the A′-and F′-pockets, ranging from Gln62 present in the distal A′-pocket α1-helix, to Leu145, present in the distal F′-pocket. The interaction map generated for the TBA7 TCR (Fig. 6) revealed binding to CD1d over the F′-pocket and closer to the α1-helix. Lastly, the interaction map of the VII68 TCR (Fig. 6)  with residues (Glu83, Lys86, Met87 and Val149) located in both α1and α2-helices. Taken together, these data suggest that, unlike type I NKT cells, the population of type II NKT cells can adopt diverse docking modes of that span the length of the CD1d-binding cleft. This is likely to reflect the high degree of variability in TCR usage and lipid Ag specificity that is observed for Type II NKT cells.

Discussion
We have identified a polyclonal NKT cell population that can recognise the microbial glycolipid α-GlcADAG presented by CD1d. To comprehensively confirm the reactivity of this NKT cell population, we have employed a combination of techniques, including single-cell TCR sequencing, TCR-transduced cell line generation and stimulation, and soluble TCR generation for biophysical and structural analysis. The α-GlcADAG used in this study is a synthetic version of one of the most abundant cell membrane lipids in M. smegmatis 40 . We also show that these cells can respond to similar α-linked glycolipids including α-GlcAdag from A. fumigatus and C. glutamicum, α-GlcDAG from S. pneumoniae and α-GlcACer from Sphingomonas species. The α-GlcADAG-responsive cells include a small subset of classical Vα14Jα18 type I NKT cells and a population of atypical Vα10Jα50 NKT cells. Moreover, we identify here a novel population of α-GlcADAG-specific type II NKT cells. This highlights the potential for overlap in the Ag specificity between these subclasses of cells and shows that distinct NKT TCRs can adopt diverse molecular mechanisms to interact with CD1d to recognise a common Ag.
While several Ag-specific type II NKT cell populations have been identified [10][11][12][13][14][15][16][17][18][19][20] , structural investigations of these TCRs interacting with their CD1d-lipid Ag targets are limited to two studies, both using the same, sulfatide-reactive XV19 TCRs 21,22 . A striking observation from these studies was that the XV19 TCR bound over the A′-roof, in sharp contrast to type I NKT TCRs that invariably bind over the F′-pocket (reviewed in ref. 24 ). This raised the question of whether other type II NKT TCRs would adopt this A′-roof-binding mode with CD1d-Ag. We show that type II NKT TCRs can also bind over the F′-pocket of CD1d. Indeed, the structure of the type II A11B8.2 TCR-CD1d-α-GlcADAG complex revealed that the interactions between the two Tyr residues within the CDR2β loop with a defined region over the F′-pocket of CD1d were similar to that observed for mouse Vβ8 + NKT TCRs 23,31,53 . However, analysis of the binding footprint of a range of type II NKT TCRs with diverse TCR αand β-chains, indicates that some type II NKT TCRs can bind in multiple ways to CD1d. This likely reflects the high TCR diversity of type II NKT cells and can engender differing reactivities to various CD1d-restricted Ags, including non-lipid molecules 54 or even peptides and lipo-peptides 55 . There are some parallels with that in MR1-mediated immunity, whereby diversity in the     Control  AR21A  Q62A  H68A  V72A  S76A  R79A  E83A  K86A  M87A  L145A  K148A  V149A  A152G  Q154A  T159A  M162A D166A L170A Fig. 6 Binding modes of a panel of type II NKT cell lines. CD69 up-regulation on NKT TCR-expressing BW58 cell lines, following 16 h culture on plates coated with alanine (Ala) mutants of CD1d loaded with either α-GalCer, sulfatide, or α-GlcADAG. The level of activation elicited by each mutant is normalised to the response elicited by a control CD1d mutant (Asp226Ala). Graphs show the average of duplicate wells ± SEM from one experiment, which is representative of three independent assays. Corresponding surfaces of CD1d (PDB code: 1Z5L) are shown to the right of each graph, depicting residues that when mutated had no effect (dark grey); a 25-50% decrease (orange); a >50% decrease (red); and a >150% increase (green) MR1-restricted T cell repertoire can also lead to differing reactivities towards metabolite-based ligands and differing docking modes atop MR1 56 . In summary, we identify a new, microbial Ag-specific, population of type II NKT cells. Our findings also suggests that a universal potent Ag for type II NKT cells is unlikely to exist. Rather, our understanding of type II NKT cells will be built upon studies of individual Ag-specific subsets, such as those identified here. Nonetheless, given that type II NKT cells may be more abundant than type I NKT cells in humans [5][6][7] (and reviewed in ref. 8 ), and considering the unique functional roles that these cells appear to play in the immune system (reviewed in ref. 2 ) these studies are important if we are to understand the full scope of Ags recognised by type II NKT cells and to harness the therapeutic potential of this arm of the immune system.

Methods
Mice. BALB/c wt, BALB/c Jα18 −/− , BALB/c CD1d −/− were bred and maintained according to ethics approval from The University of Melbourne Animal Ethics Committee, under specific pathogen-free conditions at the Biological Research Facility at the Department of Microbiology and Immunology, The University of Melbourne, Australia. BALB/c wt colonies were sourced from either the same facilities, or the Animal Resources Centre (ARC; Canning Vale, WA, Australia). CD1d −/− (CD1d1 and CD1d2 double knock-out, described in ref. 57 ) mice were sourced from the Peter MacCallum Cancer Centre (Australia) and backcrossed to BALB/c for at least 10 generations. Jα18 −/− mice were originally provided by M. Taniguchi (Chiba University Graduate School of Medicine, Japan) and backcrossed to BALB/c for at least 10 generations.
Lymphocyte isolation. Thymocytes were isolated by grinding thymi through a 50 μM mesh or cell strainer in the presence of 2% FBS PBS. For complementmediated enrichment of mature thymocytes, the cell suspension was then incubated with anti-CD24 (I) (clone J11d-in house production) on ice for 30 min. Rabbit complement (GTI Diagnostics, Wisconsin, USA) and 70 μg/mL deoxyribonuclease I (DNase; Roche) were then added and allowed to incubate for another 30 min at 37°C. Thymocytes were washed with 10% FBS/RPMI, then overlayed onto Histopaque-1083 (Sigma), and centrifuged at 1000 × g for 10 min, with no brake, at room temperature. Cells at the interphase were recovered, centrifuged at 350 × g for 4 min at 4°C and washed with 2% FBS/PBS, or 10% FBS/RPMI for cell culture. All cells were quantified in a Neubauer counting chamber (Merck), with dead cell exclusion by trypan blue solution 0.4% (Sigma).
Crystallisation, structure determination and refinement. The A11B8.2 TCR (8-10 mg/mL) and A11B8.2-CD1d-α-GlcADAG ternary complex (5-7.5 mg/mL) were crystallised in 10-20% PEG 3350/4% Tascimate, pH 4 and 15-25% PEG 1500/ 10% succinate phosphate glycine, pH 6.8, respectively, using the hanging drop vapour diffusion method. The crystals were flash frozen and data were collected at the MX1 and MX2 beamlines (Wavelength 0.9537 Å and 100 K) of the Australian Synchrotron, respectively. The data for the A11B8.2 monomer and ternary complex were processed with the XDS software 64 and iMOSFLM 7.0.5/SCALA from the CCP4 suite of programmes 65 , respectively. The TCR and the ternary complex X-ray structures were determined by molecular replacement using the software PHASER 66 . For the A11B8.2 TCR, the deposited structures (pdb code: 3QIB for the variable domain Vα11 and pdb code: 3HE6 for the variable domain Vβ8.2) were used as separate search ensembles while for the A11B8.2-CD1d-α-GlcADAG structure, the structures of mouse CD1d minus the antigen (pdb code: 3HE6) and the previously determined A11B8.2 structure were used. The initial experimental phases showed an unbiased density for the headgroup of α-GlcADAG. The programmes BUSTER 67 and COOT 68 were used for refinement and model building, respectively. Subsequent refinement cycles and model building resulted in final R work /R free values of 18.1%/21.9% (For A11B8.2) and 18.7%/24.6% (for A11B8.2-CD1d-α-GlcADAG complex). The quality of both structures was validated at the Research Collaboratory for Structural Bioinformatics Protein Data Bank (RCSB) Validation and Deposition Services. All presentations of molecular graphics were created with the programme PyMOL 69 .
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.