The length of lipids bound to human CD1d molecules modulates the affinity of NKT cell TCR and the threshold of NKT cell activation

We refolded hCD1d with a range of α-GalCer analogues with truncated acyl or phytosphingosine chains (Fig. 1). We then analyzed the rate of dissociation of each GSL from hCD1d molecules (Fig. 2 A) and the kinetics of iNKT TCR binding to each hCD1d–GSL monomer (Fig. 2 B). To measure the dissociation of each GSL from hCD1d molecules, we generated by phage display library a Fab antibody specific for hCD1d molecules loaded with α-GalCer (hereafter referred to as Fab 9B; unpublished data). Initial surface plasmon resonance (SPR) measurements and FACS staining of lipid-pulsed C1R-hCD1d cells demonstrated that Fab 9B specifically recognized hCD1d molecules loaded with the GSL shown in Fig. 1, whereas it failed to stain either unpulsed C1R-hCD1d cells or C1R-hCD1d cells pulsed with β-GalCer (unpublished data). Using Fab 9B, we measured the GSL dissociation from soluble hCD1d molecules over time using SPR (Fig. 2 A). We observed that shortening either the acyl chain or the phytosphingosine chain increased the rate of lipid dissociation from hCD1d molecules (Fig. 2 A). Interestingly, the rate of dissociation of compound C20:0 is 2.15-fold faster than that of compound C20:2 (half-life of 170 min and 367 min, respectively, at 25°C), demonstrating that unsaturation of bonds at carbons 11 and 14 of the acyl chain favors the formation of more stable hCD1d–GSL complexes.

We then assessed whether the length of the acyl and phytosphingosine chains could affect the affinity of iNKT TCR binding to the hCD1d–GSL complex. We refolded Vα24 and Vβ11 chains of iNKT TCR as previously described (10) and used the purified and refolded iNKT TCR in SPR studies against immobilized biotinylated hCD1d–GSL monomers. We demonstrated that the affinity of iNKT TCR for hCD1d–α-GalCer was substantially higher than that of hCD1d–OCH9 (Kd of 1.6 μM compared with 122 μM; Fig. 2 B).

To further understand the contribution of phytosphingosine chain length to the iNKT cell TCR affinity, we compared the affinity of iNKT TCR binding to hCD1d–OCH9 complexes with that of hCD1d molecules loaded either with OCH12 or OCH15 (Fig. 2 B, bottom). These experiments demonstrated that increasing the phytosphingosine chain length from 9 to 15 carbons increased the iNKT TCR binding affinity, suggesting that optimal iNKT TCR binding requires the lipid chain occupying the hCD1d F′ channel to range from 15 to 18 carbons. In contrast, the length and saturation of the fatty acid chain did not substantially alter the iNKT TCR binding affinity (Fig. 2 B, middle). Analysis of the binding of the anti-β2M antibody (BBM.1) confirmed that equivalent amounts of correctly refolded hCD1d–ligand complexes were immobilized to each sensor surface (unpublished data). As a further control, we showed that differences in the iNKT TCR binding affinity were not the result of differences in loading of hCD1d monomers with α-GalCer, OCH12, and OCH9 (Fig. S1). The rate constants (k_on and k_off) of iNKT TCR binding were measured directly for all hCD1d–GSL complexes with the exception of hCD1d–OCH9, in which the exceptionally fast k_off made direct determination of the k_on difficult. Nevertheless, it was possible to calculate k_on for iNKT TCR binding to hCD1d–OCH9 from the Kd and k_off (Table I). Comparison of the binding kinetics of iNKT TCR to the hCD1d–GSL monomers demonstrated that the decrease in affinity observed with shortening the phytosphingosine chain was as much the result of decreases in the k_on as increases in the k_off (Fig. 2 B and Table I). Agreement was observed between the affinities calculated from the kinetic constants and the affinities determined by equilibrium binding (Table I and unpublished data).

To ensure that differences in the affinity of iNKT TCR binding to hCD1d–GSL complexes were physiologically relevant, we used biotinylated iNKT TCR to stain C1R-hCD1d cells pulsed with a panel of GSL (Fig. 2 C). The results of these experiments confirmed that C1R-hCD1d cells pulsed with α-GalCer and OCH15 were stained more efficiently than C1R-hCD1d pulsed with either OCH12 or OCH9 (Fig. 2 C). As a control, GSL-pulsed C1R-hCD1d cells were stained with the Fab 9B antibody to ensure a similar loading of hCD1d molecules by the panel of GSL at a concentration of 10 μg/ml (Fig. S2).

To extend our findings on the molecular aspects of the iNKT cell immunological synapse and to investigate the possibility that differences in iNKT TCR affinity for the various hCD1d–GSL complexes may have an effect on the kinetics and/or pattern of hCD1d accumulation during synapse formation, we quantified the segregation and accumulation of hCD1d molecules and ICAM-1 in real time.

To this end, iNKT cells, from which the TCR for the kinetic studies was cloned, were settled onto bilayers containing AlexaFluor542-conjugated glycosyl-phosphatidylinositol (GPI)–linked ICAM-1 and biotinylated hCD1d monomers loaded with either α-GalCer or OCH9, which showed the greatest disparity in iNKT TCR affinity. We then followed the pattern of ICAM-1 (Fig. 3 A, red) and hCD1d (Fig. 3 A, green) distribution over time by confocal microscopy and correlated it with iNKT cell contact with the membrane as assessed by interference reflection microscopy (IRM), which allows the visualization of close membrane apposition as dark areas (Fig. 3 A).

We demonstrated that both hCD1d–OCH9 and hCD1d–α-GalCer monomers stimulated the formation of a classic immunological synapse, as characterized by the cessation of migration (Fig. 3 A and Video 1), the formation of a classic central supramolecular activation cluster (SMAC [cSMAC]), and aggregation of ICAM-1 to form a peripheral SMAC. This central zone of interaction was bordered by a ring of close apposition of the T cell membrane and planar bilayer (darkest area of IRM image). In the absence of ICAM-1 in the lipid bilayer, no clustering of hCD1d–α-GalCer molecules was observed (unpublished data). Quantification of the amount of hCD1d–α-GalCer–loaded molecules (Fig. 3 B) and ICAM-1 (Fig. 3 C) showed that a maximal clustering of both molecules was reached within 5 min and a plateau after 10 min. In contrast, quantification of the amounts of hCD1d–OCH9-loaded molecules (Fig. 3 B) showed slower kinetics and reduced maximal clustering of hCD1d–OCH9 monomers as compared with hCD1d–α-GalCer. Interestingly, consistent with the evidence that LFA-1 delivers a signal distinct from that of the TCR, accumulation of ICAM-1 to the peripheral SMAC was comparable between hCD1d–α-GalCer– and hCD1d–OCH9-stimulated synapses (Fig. 3 C). We also demonstrated that although hCD1d monomers loaded with β-GalCer did not induce any hCD1d clustering, some gathering of ICAM-1 was observed (Fig. 3 E). These results are consistent with previously published data analyzing the formation of the immunological synapse in T cells (15) and indicate that the affinity of the iNKT cells for hCD1d–GSL complexes may play an important role in determining ligand recognition.

To determine whether this difference in affinity translates into differential thresholds of hCD1d–GSL recognition, we compared the interaction of iNKT cells with lipid bilayers loaded with constant ICAM-1 concentration but decreasing densities of either hCD1d–α-GalCer or hCD1d–OCH9 monomers (Fig. 3, D and E). These results showed that a minimum density of 10 molecules/μm² was sufficient to trigger hCD1d–α-GalCer and hCD1d-C20:0 accumulation and immunological synapse formation (Fig. 3, D and E). In contrast, at this same density of hCD1d–OCH9, iNKT cells failed to aggregate antigen or show any sign of synapse formation. Indeed, a 10-fold increase in the density of hCD1d–OCH9 was required to observe similar frequencies of synapse formation. Because hCD1d–α-GalCer and hCD1d-C20:0 have a similar affinity of binding to iNKT TCR (Fig. 2 B) and OCH9 and C20:0 have a similar rate of dissociation from hCD1d molecules (Fig. 2 A), these results demonstrate that the affinity of iNKT TCR determines the capacity of iNKT cells to cluster hCD1d–lipid complexes, and this may play a role in defining the threshold of iNKT cell activation.

To establish whether the differential accumulation of hCD1d–GSL complexes to the synapses correlated with the strength of the subsequent iNKT cell activation, we monitored the changes in intracellular calcium levels of iNKT cells stimulated with planar bilayers loaded with either hCD1d–α-GalCer or hCD1d–OCH9. At the same densities, hCD1d–α-GalCer–stimulated synapses resulted in a greater and more rapid fluxing of calcium into the iNKT cells compared with hCD1d–OCH9-stimulated synapses (Fig. 4, A and B). These results further support the concept that the length of the phytosphingosine chain of the lipid antigen is important in determining the threshold of iNKT cell activation.

To further assess the functional consequences of the differences in CD1d–GSL and TCR binding affinities, we analyzed the efficiency of polarization of iNKT cell lytic granules on the stimulation of iNKT cells with C1R- hCD1d target cells pulsed with either α-GalCer or OCH9, which are the strongest and weakest iNKT cell agonists, respectively. Titration of the lipid concentrations for both α-GalCer and OCH9 revealed that the number of polarized iNKT cells conjugated to targets was greatly reduced at lower concentrations of OCH9 when compared with α-GalCer (Fig. 5, B and C). When C1R-hCD1d cells were pulsed with OCH9, polarization was poor (i.e., 10 ng/ml OCH was required to see 50% of iNKT cell conjugates with tightly polarized granules at the synapse). In contrast, 0.001 ng/ml α-GalCer was sufficient to see 60% of iNKT cell conjugates with tightly polarized granules at the synapse. Because the affinity of Fab 9B for α-GalCer–CD1d and OCH9–CD1d monomers is very similar (Kd of 408 nM and 595 nM, respectively; unpublished data), FACS staining of C1R-hCD1d cells pulsed with 1 μM α-GalCer and OCH9 with Fab 9B confirmed that within the time frame of the experiment shown in Fig. 5, the amount of α-GalCer and OCH9 presented by C1R-hCD1d cells was comparable (not depicted). Consistent with the observed slower and less efficient polarization of cytotoxic granules of iNKT cells activated by OCH9, we demonstrated that human DCs pulsed with OCH9 were killed less efficiently by iNKT cells than human DCs pulsed with α-GalCer (Fig. 5 D).

The density of clustered hCD1d–α-GalCer and the proportion of cells forming immunological synapses with α-GalCer as compared with OCH9 correlated with the in vitro activation of human iNKT cells as defined by IL-4 and IFN-γ secretion (Fig. 6 A). Furthermore, unlike results obtained with mouse iNKT cells (13, 16), the cytokine profile of human iNKT cells stimulated by OCH9 and OCH12 was not biased toward a Th2 response; rather, less IL-4 and IFN-γ were secreted (Fig. 6 A). Comparable cytokine profiles were obtained using iNKT cells stimulated with α-GalCer and with compounds with truncated acyl chains and OCH15. Thus, the pattern of cytokine production and target cell lysis by iNKT cells in response to lipid-pulsed target cells was consistent with the differential activation of iNKT cells.

Finally, we demonstrated the ability of both α-GalCer and OCH9 to stimulate the expansion of iNKT cells from peripheral blood lymphocytes from healthy donors (Fig. 6 B). Although α-GalCer was capable of sustaining iNKT cell proliferation at concentrations < 10⁻¹¹ M, OCH9 failed to induce iNKT cell proliferation at concentrations <10⁻¹⁰ M, and, at all concentrations, α-GalCer stimulated greater numbers of iNKT cells.

Our results provide important insights into the mechanisms controlling lipid-specific T cell responses by demonstrating that modifications of the lipid chain buried in the F′ channel of hCD1d molecules can modulate the TCR affinity, formation of stable synapses, and, ultimately, lymphocyte activation. Previous studies have established that modifications of the alkyl chains may alter the activation of lipid-specific T cells (3, 13, 17–21). However, a combined kinetic and functional analysis has not been previously performed.

Because α-GalCer has the maximum lipid chain lengths that are able to fit into the antigen-binding groove of hCD1d and there are conformational differences between lipid-bound and nonlipid-bound CD1d molecules (4), we reasoned that binding of hCD1d molecules with analogues of α-GalCer containing shorter lipid chains may result in the partial collapse of unfilled portions of the A′ and F′ channels, resulting in surface-exposed structural changes. This reasoning led us to study whether the binding affinity of soluble iNKT TCR to hCD1d molecules loaded with GSL containing an identical polar head can be fine tuned by modifying the length of the lipid chains.

It has recently been reported that diacylglycerol lipids from Borrelia burgdorferi are recognized by iNKT cells and that their antigenic potency is dependent on acyl chain length and saturation (3). In particular, it was shown that mouse iNKT cells more efficiently recognize mouse CD1d molecules loaded with diacylglycerol analogues with 16-carbon-long R₂ fatty acids. In contrast, human iNKT cells more efficiently recognize hCD1d molecules loaded with diacylglycerol analogues with 18-carbon-long R₂ fatty acids (3). It is likely that these differences between human and mouse iNKT cells can be accounted for by the shorter length of the mouse CD1d F′ channel as compared with the hCD1d F′ channel (4–7). However, these results did not address whether the enhanced potency of diacylglycerol analogues is caused by their more stable binding to CD1d molecules or rather to a higher TCR affinity. To distinguish between these possibilities, we independently analyzed the role of lipid chains occupying the A′ and F′ channels in hCD1d molecules in modulating the TCR affinity and/or stability of hCD1d–GSL complexes.

The central claim of this study is that the length of the lipid chain occupying the F′ channel modulates the binding affinity of iNKT TCR. This claim is supported by two complementary experimental systems: (a) SPR, analyzing the binding of soluble iNKT TCR to in vitro refolded CD1d molecules coated to BIAcore chips (Fig. 2, A and B), and (b) FACS staining, analyzing the binding of biotinylated tetramerized iNKT TCR to C1R-hCD1d cells pulsed with a panel of GSLs (Fig. 2 C).

The difference in TCR binding affinity between α-GalCer, OCH12, and OCH9 could be accounted for by the hypothesis that shortening of the phytosphingosine chain could cause a repositioning of this lipid chain in the hCD1d F′ channel and, thus, directly alter the position of the α-linked galactose or cause a shift in the α helices of hCD1d as a result of a collapse of the partially unfilled F′ channel.

If it is assumed that OCH9 retains the same orientation as α-GalCer, a comparison of the crystal structures of α-GalCer–loaded and empty hCD1d molecules (4) suggests that loading hCD1d molecules with OCH9 will only drive part of the F′ channel into the filled conformation because of the shorter phytosphingosine chain (Fig. 7 A). This modification results in predicted differences between hCD1d–OCH9 and hCD1d–α-GalCer structures of ∼2 Å in the main chain of the α2 helix between residues Leu139 and Thr154. This region is at the center of the putative TCR binding footprint as mapped by mutagenesis studies (22, 23) and includes two residues (Asp151 and Thr154) that hydrogen bond to the α-GalCer headgroup. Any substantial alterations in the positions of these residues may be expected to impact on the hydrogen bond network, which dictates the position of the headgroup, and, thus, to affect the recognition of α-GalCer by iNKT TCR.

These models and our finding that truncating the phytosphingosine chain decreases the k_on support a conformational change mechanism whereby hCD1d can exist in more than one conformation, a subset of which can bind iNKT TCR, and filling the F′ channel favors the TCR-binding conformations. A similar modeling exercise to assess the effect of shortening the acyl chain (i.e., the structural differences associated with incomplete occupation of the A′ channel; Fig. 7 B) predicts changes in the position of the α1 helix from residues Phe58 to Phe70, but, in contrast to the differences resulting from a partially unfilled F′ channel, these changes are in a region that is peripheral to the iNKT cell TCR docking area and have no influence on the orientation of the α-GalCer headgroup. Collectively, these structural considerations concur with our data, which indicate that changes in the length of the phytosphingosine but not the acyl chain can modulate TCR recognition.

The observation that fatty acid spacers have been identified in the A′ channel seems to suggest that the presence of these short-chain lipids could be a conserved element, ensuring full occupancy of the A′ channel (7, 21). In contrast, because the F′ channel can accommodate an alkyl chain up to 18 carbon atoms long rather than 26 carbon atoms long, F′ channel spacer lipids are unlikely to exist when the F′ channel is occupied by lipid antigens shorter than 12 carbon atoms (such as OCH12 and OCH9) because fatty acids ranging from 9 to 6 carbon atoms required to fill the remaining unoccupied space of the F′ channel comprise only a minor percentage of the total fatty acid composition of mammalian cells (24).

We confirmed the results obtained using hCD1d molecules coated to the BIAcore chips by developing a FACS-based assay in which hCD1d molecules were expressed on live cells rather than coated to a solid phase (Fig. 2 C). The finding that fluorescently labeled recombinant iNKT TCR tetramers can be used as a staining reagent for GSL-pulsed C1R-CD1d cells is of interest because in a previous study, the staining of peptide-pulsed target cells by MHC class I–restricted TCR could only be obtained using an affinity-matured TCR with Kd values of ∼10⁻¹² M (i.e., 10⁶-fold higher than the affinity values of iNKT TCR; reference 25). Because iNKT TCR and wild-type MHC class I–restricted TCRs have similar affinities, it is tempting to speculate that the ability of iNKT TCR to stain GSL-pulsed cells may be caused by the clustering of hCD1d–GSL complexes on the surface of C1R-hCD1d cells.

Although the role of peptide MHC clustering in TCR signaling remains an area of investigation (26), a cooperative interaction between mouse CD1d–α-GalCer complexes and iNKT TCRs has previously been shown by fluorescence resonance energy transfer measurements (27). Our finding that the FACS-based assays confirmed the hierarchy of responses to the different GSLs observed in the SPR results strongly suggests that monomeric hCD1d/GSL–TCR interactions (i.e., SPR data) are in line with the situation in which the physiological multimeric hCD1d/GSL–TCR interactions may occur between hCD1d molecules expressed on the surface of APCs and TCRs expressed on iNKT cells. Consistent with our data, differences in the ability of mouse iNKT cells to be stained by mouse CD1d tetramers loaded with either α-GalCer or OCH9 have also been shown (27).

The use of a Fab antibody specific to the hCD1d–α-galactose–linked sphingolipid complex permitted us to compare the efficiency of binding to hCD1d molecules for a broad range of α-GalCer analogues. It is of interest that unlike iNKT TCR, Fab 9B has a more similar affinity for hCD1d monomers loaded with either α-GalCer or OCH9 (unpublished data), suggesting that the footprint of Fab 9B on the hCD1d–GSL complex may be different from the footprint of iNKT TCR. We used Fab 9B to study the role of each alkyl chain in controlling the rate of dissociation from hCD1d molecules. Consistent with previously published data with mouse CD1d molecules (28), we demonstrated that the shortening of either alkyl chain reduces the stability of the hCD1d–lipid complex. We further extended these results by demonstrating that the presence of two cis double bonds in the acyl chain, at positions 11 and 14 (C20:2; reference 6), reduced the rate of dissociation from hCD1d molecules as compared with the compound C20:0, which has a saturated acyl chain. It is of interest that the compound C20:2, unlike the compound C20:0, was shown to be efficiently loaded onto mouse CD1d molecules lacking an endosomal targeting motif, suggesting that the presence of the two double bonds at positions 11 and 14 could facilitate its binding to CD1d molecules (28). Because the acyl chain of α-GalCer is anchored in place in the A′ channel by circumnavigating a central pole (4), the observed differences in the rate of dissociation of the C20:0 and C20:2 compounds suggest that a preformed kink in the acyl chain caused by the presence of the unsaturated bonds may stabilize binding of the lipid by favoring the tightly curved conformation required for binding in the A′ channel.

Our results describe for the first time the immunological synapse of iNKT cells and polarization of iNKT cell cytotoxic granules. To determine the effect of the disparity in TCR affinities and TCR/hCD1d half-lives on the earliest events in human iNKT cell synapse formation, we used a planar lipid bilayer (15, 29) with constant ICAM-1 density and with equal numbers of hCD1d–OCH9 or hCD1d–α-GalCer complexes. The pattern of segregation of hCD1d molecules and ICAM-1 is similar to the pattern described for MHC class I– and II–restricted T cells (15). Interestingly, not only did the rate of accumulation of hCD1d–GSL complexes to cSMAC vary considerably between α-GalCer and OCH9, but the final density of aggregated antigen also differed.

The amount of signal that T cells receive by interacting with APCs is determined by several parameters, including the concentration of peptide MHC complex (class I peptide binding affinity) and the duration of the interaction between T cells and APCs (TCR binding affinity). Previous studies have compared the structures of peptide MHC complexes with those of altered peptide ligands, for example (30–35). Consistent with our findings describing the role of CD1d-bound lipid chains in modulating TCR binding affinity, these comparisons have demonstrated that substitutions of anchor residues can cause slight structural alterations that may indirectly impact on TCR recognition, synaptic strength, fluxing of calcium, and cytokine secretion (35–38).

T cell activation has recently been shown to be initiated and sustained in TCR-containing microclusters generated at the initial contact sites and at the periphery of the mature immunological synapse (39, 40). TCR signaling, which is sustained by stabilized microclusters, is terminated in the cSMAC, a structure from which TCRs are sorted for degradation (41). After the cSMAC forms, the microclusters continue to form and are sites of TCR signaling based on the recruitment of phosphorylated Lck, ZAP-70, and LAT (39, 40). Although analysis of microcluster formation in the immunological synapse of iNKT cells was beyond the scope of this study, our results indicate that the differential rate of accumulation of hCD1d–GSL complexes to the synapses substantially correlated with the strength of the subsequent iNKT cell stimulation.

When we compared hCD1d–α-GalCer with hCD1d–OCH9-stimulated synapses, hCD1d–α-GalCer–stimulated synapses triggered a stronger and more rapid fluxing of calcium into the iNKT cells. Thus, the strength of signaling through the iNKT cell TCR is far greater for α-GalCer than for OCH9, and this determines downstream responses by the human iNKT cells. To strengthen these conclusions, we also compared the pattern of aggregation in lipid bilayers of hCD1d molecules loaded with OCH9 and C20:0 and demonstrated that the pattern and kinetics of hCD1d aggregation is very similar to those obtained with hCD1d–α-GalCer monomers. OCH9 and C20:0 have a similar rate of dissociation from hCD1d molecules, strongly suggesting that differences in the iNKT cell TCR affinity are controlling the pattern of hCD1d clustering in the lipid bilayers. Because the comparison of the binding kinetics of iNKT TCR to the hCD1d monomers loaded with different lengths of phytosphingosine chains was as much the result of decreases in the k_on as increases in the k_off, our results are consistent with previously published data, demonstrating that the differential accumulation of peptide MHC to cSMAC is ascribed to the TCR half-life for the peptide MHC complex (15). A recent study by Qi et al. (42) explores this concept further by providing an algorithm for relating solution half-life to 2D half-life by incorporating the heat capacity.

Differences in lymphokine profiles and in the proliferation of iNKT cells from PBL stimulated with α-GalCer and OCH9 can be accounted for by a combined effect of the higher affinity of α-GalCer to hCD1d molecules and higher affinity of the iNKT cell TCR for the hCD1d−α-GalCer complex. The reduced killing of DCs pulsed with OCH9 as compared with the killing of DCs pulsed with α-GalCer is consistent with the reduced polarization of cytotoxic granules in iNKT cells stimulated with weaker agonists. Because it has been shown that activation in vivo of iNKT cells can assist priming of antigen-specific immune responses (43, 44) reduced lysis of DCs pulsed with OCH9 suggests that more attention should be focused on the use of weaker iNKT cell agonists that, unlike α-GalCer, may ensure a longer life span of APCs. This is particularly important because α-GalCer–pulsed DCs have already been used in clinical trials (45).

In conclusion, our results highlight the effects of variation in the length of lipid chains of exogenously loaded lipids occupying the CD1d F′ channel in fine tuning the affinity of TCR of hCD1d-restricted lymphocytes. Although further experiments are required to extend these findings to the recognition of endogenous loaded lipids, our results demonstrate that the sensitivity of TCR recognition of hCD1d-restricted T lymphocytes can extend to detect differences in lipid chains buried in the F′ channel, resulting in altered TCR binding kinetics and different levels of activation signals.

GSLs were synthesized by a previously described strategy (46), and their structures were confirmed by mass spectrometry. Glucocerebrosides (Gaucher's spleen) were purchased from Sigma-Aldrich. Human Vα24 iNKT cell lines were generated and maintained as previously described (47) in RPMI medium (Sigma-Aldrich).

hCD1d and β₂M protein were refolded with GSL shown in Fig. 1 by oxidative refolding chromatography using a protocol that was previously described (48).

The generation of soluble TCR heterodimers was based on the procedure described by Boulter et al. (49). In brief, the Vα24 and Vβ11 iNKT cell TCR sequences were amplified by PCR from the cDNA of an established iNKT cell clone. The α and β chains were individually cloned into pGMT7 bacterial expression vectors encoding the c-Jun and v-Fos fragments, respectively. TCR chains were expressed, refolded, and purified as previously described (49).

Soluble biotinylated TCR was made by inserting a 15–amino acid AviTag recognition peptide (Avidity) after the c-Jun fragment of the α chain. In brief, the α chain Jun fragment was amplified by PCR and subcloned into a pGMT7 vector encoding the AviTag peptide. This new α protein was expressed, refolded with the β chain, and purified as before. The TCR heterodimer was biotinylated with BirA enzyme before the final SD75 gel filtration (50). The biotinylated TCR heterodimers were tetramerized by mixing at a 4:1 molar ratio with Extravidin-PE (Sigma-Aldrich).

The affinity and kinetic properties of Vα24 NKT TCR binding to hCD1d–lipid complexes were measured with a spectrometer (model 3000; Biacore) as previously described (51). The purified biotinylated hCD1d–lipid complexes were immobilized onto streptavidin-coated CM5 sensor chips (Biacore) at a level of ∼1,000 response U. In all experiments, hCD1d–β-GalCer molecules were used as negative controls. Equilibrium binding was performed at a flow rate of 10 μl/min from the lowest TCR concentration. To exclude any effect of lipid dissociation, which would result in a gradual decrease in the level of immobilized CD1d–lipid during the course of the experiment, the affinity was determined with both increasing (from low to highest) and decreasing (from highest to lowest) concentrations of TCR. The affinity values obtained were not significantly different, indicating that lipid dissociation during the course of the experiments had a negligible effect on the affinity measurements. The data was plotted using Origin software (OriginLab), and Kd values were calculated using the standard hyperbolic model. Kinetics of the TCR–hCD1d/lipid interactions was measured at 50 μl/min. Equilibrium dissociation constants (Kd) and k_on and k_off were obtained by nonlinear curve fitting of the binding curves obtained after subtracting the response in the reference flow cells with equations derived from the simple 1:1 Langmuir binding model using the BIA evaluation program (version 3.02.2; Biacore). SPR was also used to estimate the dissociation rate of GSL from hCD1d molecules. 1,000 response U of hCD1d monomers loaded with either α-GalCer, OCH9, OCH12, OCH15, C20:0, C20:2, or C11:1 were immobilized on a streptavidin-coated sensor chip. As a negative control, hCD1d–β-GalCer biotinylated monomers were used. The amount of hCD1d–GSL remaining at a particular time was measured by injecting saturating amounts (30 μl at 500 nM) of Fab 9B antibody.

C1R-hCD1d cells were pulsed with 10 μM, 5 μM, 1 μM, 500 nM, and 100 nM of lipid for 16 h at 37°C and washed twice in RPMI. Cells were incubated with iNKT cell TCR tetramers at 37°C for 30 min and washed again twice with ice-cold PBS/1% FCS. Samples were analyzed on a flow cytometer (FACSCalibur; BD Biosciences), and the data were processed using CellQuest software (BD Biosciences).

The GPI-linked ICAM-1 protein was purified from transfected L cells as previously described (52). In brief, the protein was solubilized in 1% Triton X-100 buffer and captured on an anti–ICAM-1 YN1/1.7 antibody-coupled Sepharose column (N-hydroxysuccinimide–activated HiTrap; 1 ml; GE Healthcare). The protein was labeled with AlexaFluor532 (Invitrogen) and eluted at low pH. The purified, labeled GPI-linked ICAM-1 was reconstituted into liposomes by detergent analysis with 0.4 mM egg phosphatidylcholine (Avanti Polar Lipids, Inc.). Planar lipid bilayers were formed in FCS2 closed chambers as previously described (15). AlexaFluor532-labeled GPI-linked ICAM-1 liposomes and biotinylated lipids were mixed with 1,2-dioleocyl-phosphatidylcholine lipids (Avanti Polar Lipids, Inc.) at different ratios to obtain the required molecular densities. The chambers were blocked with PBS and 2% FCS. Monobiotinylated CD1d monomers were loaded on the bilayers using AlexaFluor633-streptavadin (Invitrogen) as a bridging with the biotin lipids. The iNKT cells were injected into the 37°C chambers in chamber buffer (PBS, 0.5% FCS, 2 mM Mg²⁺, 0.5 mM Ca²⁺, and 1 g/liter d-glucose). Confocal fluorescence and differential interference contrast images were simultaneously obtained at specified times, and iNKT cell contact with the planar bilayer was visualized by IRM. Images were acquired on an inverted microscope (Axiovert LSM 510-META; Carl Zeiss MicroImaging, Inc.) with a 63× oil immersion objective (Carl Zeiss MicroImaging, Inc.) and analyzed by LSM 510 software (Carl Zeiss MicroImaging, Inc.).

Intracellular Ca²⁺ influx was measured by confocal microscopy with a fluorometric assay. Cells were labeled with fluo-4FF calcium indicator (1 μM in culture medium; Invitrogen) for 30 min at room temperature, washed, and injected into the FCS2 closed chamber in chamber buffer. The Ca²⁺ flux was monitored by the increase of fluorescence at ∼520 nm with an open pinhole to acquire the total signal. The total mean fluorescence per cell was quantified as the intensity in the selected cell – intensity in the neighboring area at each time point. Data represent the mean of 20 cells.

C1R-hCD1d target cells were pulsed with 1 μg/ml, 100 ng/ml, 10 ng/ml, 1 ng/ml, 100 pg/ml, 10 pg/ml, or 1 pg/ml of either α-GalCer or OCH9 lipid for 16 h at 37°C and washed twice in RPMI. NKT cells and C1R-hCD1d cells were washed in RPMI, and each cell pellet was resuspended to a final concentration of 10⁶ cells/ml in RPMI. NKT cells and targets were mixed 1:1, left for 5 min in suspension, plated on glass multiwell slides, and incubated for 20 min at 37°C. Cells were fixed in 100% methanol precooled to –20°C, washed in PBS, and blocked in PBS/2% BSA (Sigma-Aldrich). The cells were stained, and the slides were mounted in PBS containing 90% glycerol and 2.5% 1,4-diazabicyclo[2.2.2]octane. Samples were examined using a laser-scanning microscope (Radiance 2000 MP; Bio-Rad Laboratories), and the conjugation rate and granule polarization were quantified using an epifluorescent microscope (Axioplan 2; Carl Zeiss MicroImaging, Inc.).

Monocyte-derived DCs were differentiated as described previously (53). Immature DCs were collected at day 4, and 30,000 cells were plated in 96-well flat-bottom plates with equal numbers of iNKT cells and different concentrations of lipids. Cells were incubated for 24 h before FACS analysis on a flow cytometer (FACSCalibur; BD Biosciences) with CellQuest software (BD Biosciences).

C1R-hCD1d cells were pulsed for 2 h with GSLs and diluted to 10⁶cells/ml. 10⁵ cell-pulsed targets were incubated at 37°C with 10⁴ iNKT cells in a final volume of 200 μl. After 18 h, the supernatants were harvested, and the concentrations of IFN-γ and IL-4 were determined by ELISA according to the manufacturer's instructions (Mabtech).

PBMCs were isolated from healthy donors' buffy coats by density gradient centrifugation over Lymphoprep (Nycomed). Monocytes were positively selected using anti-CD14 mAb–coated magnetic beads (MACS; Miltenyi Biotec), and monocyte-depleted lymphocyte fractions (CD14 negative) were frozen until needed. Monocytes were cultured as previously described with 50 ng/ml GM-CSF (Novartis) and 1,000 U/ml IL-4 in six-well plates at 4 × 10⁵ cells/ml (3 ml/well). After 5 d, maturation was induced using bacterial LPS (1 μg/ml LPS of Salmonella abortus equi; Sigma-Aldrich). Immature and mature monocyte-derived DCs were phenotypically analyzed for maturation markers. 2 × 10⁵ monocyte-derived DCs were pulsed with GSL for 2 h in 24-well plates in 200 μl RPMI 1640. 2 × 10⁶ autologous lymphocytes were added in 1.8 ml of medium (5% human serum). After 3 d, 25 IU/ml IL-2 was added to cultures. Thereafter, cultures were fed every 3–4 d with fresh medium containing 1,000 U/ml IL-2. iNKT cell frequencies were determined using APC CD1d tetramer and Vα24 antibody (Serotec).

The structural effects of binding lipid chains of various lengths to hCD1d were assessed by comparison of the hCD1d–α-GalCer and empty hCD1d crystal structures (Protein Data Bank code 1ZT4). Structural superposition used the Structure Homology Program (54), and analysis was performed using the interactive graphics display program O (55).

Fig. S1 shows iNKT TCR and Fab 9B antibody binding to immobilized hCD1d–GSL monomers. hCD1d complexed with α-GalCer, OCH12, OCH9, and β-GalCer were loaded onto a streptavidin-coated CM5 sensor chip. Fig. S2 shows equivalent loading of the C1R-hCD1d cells with different GSLs. Video 1 shows the time-lapse reconstruction of the kinetics of hCD1d–α-GalCer aggregation (density = 200 molecules/μm²).

This work was funded by Cancer Research UK grants C399/A2291 and C399/A3213, a UK Medical Research Council grant, and a European Community European Autoimmune Polyendocrine Syndrome Consortium grant to V. Cerundolo as well as Cancer Research UK grants C375/A3964 and C375/A5261 and a UK Medical Research Council grant to E.Y. Jones. P.A. van der Merwe acknowledges support from the UK Medical Research Council. G.S. Besra acknowledges support from James Bardrick in the form of a Personal Research Chair, is a former Lister Institute-Jenner Research Fellow, and is supported by the UK Medical Research Council and the Wellcome Trust.

The authors have no conflicting financial interests.