Natural variations at position 93 of the invariant Vα24-Jα18 α chain of human iNKT-cell TCRs strongly impact on CD1d binding

Authors

  • Joseph P. Sanderson,

    Corresponding author
    1. Academic Unit of Clinical & Experimental Sciences, Faculty of Medicine, University of Southampton, Southampton, UK
    • Joseph P. Sanderson, Max Planck Institute of Biophysics, Theoretical Molecular Biophysics Group, Frankfurt am Main, Germany Fax: +49-6963-031502

      Kathrin Waldburger-Hauri, Mailpoint 811, Level E, Sir Henry Wellcome Laboratories, South Block, Southampton General Hospital, Southampton SO16 6YD, UK Fax: +44-2380796603

    Search for more papers by this author
    • These authors contributed equally to this work.

  • Kathrin Waldburger-Hauri,

    Corresponding author
    1. Department of Rheumatology, Faculty of Medicine, University of Bern, Bern, Switzerland
    • Joseph P. Sanderson, Max Planck Institute of Biophysics, Theoretical Molecular Biophysics Group, Frankfurt am Main, Germany Fax: +49-6963-031502

      Kathrin Waldburger-Hauri, Mailpoint 811, Level E, Sir Henry Wellcome Laboratories, South Block, Southampton General Hospital, Southampton SO16 6YD, UK Fax: +44-2380796603

    Search for more papers by this author
    • These authors contributed equally to this work.

  • Diana Garzón,

    1. Max Planck Institute of Biophysics, Theoretical Molecular Biophysics Group, Frankfurt am Main, Germany
    Search for more papers by this author
  • Gediminas Matulis,

    1. Department of Rheumatology, Faculty of Medicine, University of Bern, Bern, Switzerland
    Search for more papers by this author
  • Salah Mansour,

    1. Academic Unit of Clinical & Experimental Sciences, Faculty of Medicine, University of Southampton, Southampton, UK
    Search for more papers by this author
  • Nicholas J. Pumphrey,

    1. ImmunoCore Limited, Milton Park, Abingdon, Oxfordshire, UK
    Search for more papers by this author
  • Nikolai Lissin,

    1. ImmunoCore Limited, Milton Park, Abingdon, Oxfordshire, UK
    Search for more papers by this author
  • Peter M. Villiger,

    1. Department of Rheumatology, Faculty of Medicine, University of Bern, Bern, Switzerland
    Search for more papers by this author
  • Bent Jakobsen,

    1. ImmunoCore Limited, Milton Park, Abingdon, Oxfordshire, UK
    Search for more papers by this author
  • José D. Faraldo-Gómez,

    1. Max Planck Institute of Biophysics, Theoretical Molecular Biophysics Group, Frankfurt am Main, Germany
    Search for more papers by this author
  • Stephan D. Gadola

    1. Academic Unit of Clinical & Experimental Sciences, Faculty of Medicine, University of Southampton, Southampton, UK
    Search for more papers by this author

Abstract

Human invariant natural killer T (NKT) cell TCRs bind to CD1d via an “invariant” Vα24-Jα18 chain (iNKTα) paired to semi-invariant Vβ11 chains (iNKTβ). Single-amino acid variations at position 93 (p93) of iNKTα, immediately upstream of the “invariant” CDR3α region, have been reported in a substantial proportion of human iNKT-cell clones (4–30%). Although p93, a serine in most human iNKT-cell TCRs, makes no contact with CD1d, it could affect CD1d binding by altering the conformation of the crucial CDR3α loop. By generating recombinant refolded iNKT-cell TCRs, we show that natural single-nucleotide variations in iNKTα, translating to serine, threonine, asparagine or isoleucine at p93, exert a powerful effect on CD1d binding, with up to 28-fold differences in affinity between these variants. This effect was observed with CD1d loaded with either the artificial α-galactosylceramide antigens KRN7000 or OCH, or the endogenous glycolipid β-galactosylceramide, and its importance for autoreactive recognition of endogenous lipids was demonstrated by the binding of variant iNKT-cell TCR tetramers to cell surface expressed CD1d. The serine-containing variant showed the strongest CD1d binding, offering an explanation for its predominance in vivo. Complementary molecular dynamics modeling studies were consistent with an impact of p93 on the conformation of the CDR3α loop.

Introduction

Invariant natural killer T (iNKT) lymphocytes are a versatile and potent subset of T cells which play important roles in both host defense and tolerance. In contrast to conventional T cells, they recognize the non-polymorphic lipid-antigen-presenting molecule CD1d via highly conserved TCRs. In fact, all iNKT-cell TCRs use an “invariant” TCRα chain (iNKTα), Vα24-Jα18 in humans, Vα14-Jα18 in mice, paired to “semi-invariant” TCRβ chains (iNKTβ), which in humans solely utilize Vβ11 rearranged with diverse TCR Jβ segments. For all iNKT-cell TCRs, binding to CD1d is primarily mediated by the Vα-Jα rearranged “invariant” CDR3α loop as well as the two germline-encoded CDR1α and CDR2β loops 1, 2. In addition, we have recently shown that the only variable region of human iNKT-cell TCRs, i.e. the CDR3β loop, strongly modulates the affinity to CD1d 3.

Human iNKT-cell clones with single-amino acid variations in the “invariant” Vα24-Jα18 iNKTα chain have been previously identified 4, 5, and a recent study found that 4-30% of human iNKT-cell clones carried such variations 4. However, the effect of these variations on iNKT-cell TCR binding to CD1d has not been examined. All of the described variations to date are located immediately upstream of the CDR3α loop region and most of these involve residue 93 (p93). The most commonly used amino acid at this position is serine (Ser93), which is encoded for by the germline sequence of human Vα24S1 (AE00521.1), but other identified natural variations at this position include asparagine (Asn93), threonine (Thr93) and isoleucine (Ile93) 5.

The recent crystal structure of the human and mouse iNKT-cell TCRs complexed with CD1d shows that the amino acid at p93 of iNKTα, Ser93 in the human iNKT-cell TCR:CD1d structure, makes no direct contact with either CD1d or the CD1d-bound lipid antigen 1, 6. Interestingly though, the human Ser93 makes polar contacts (H-bonds) with other residues within CDR3α, namely Arg95 and Arg103, both of which are absolutely essential for CD1d antigen recognition 1. Hence, consistent with a “lock-and-key” mechanism of iNKT-cell TCR binding to CD1d, Ser93 could strengthen the conformational stability of the “invariant” CDR3α loop, whereas substitutions at this position might fail to do so and thus impact on iNKT-cell TCR–CD1d binding. In order to test this hypothesis, we generated a panel of human iNKT-cell clones, identified natural iNKTα variants at p93, and compared the binding of recombinant replica of these natural iNKT-cell TCR variants with recombinant and endogenous CD1d–ligand complexes. This strategy enabled us to determine the impact of p93 variations on CD1d binding independent of the effect of CDR3β variation. Our results show that natural variations at p93 of the human “invariant” iNKTα chain exert a powerful impact on iNKT-cell TCR binding to human CD1d. Complementary molecular dynamics modeling studies, based on the human iNKT-cell TCR:CD1d complex crystal structure (PDB 3HJU), were consistent with an effect of p93 on the conformational stability of the invariant CDR3α loop.

Results

Ex vivo identification of natural iNKTα variants in CD1d-restricted “invariant” Vα24-Ja18/Vβ11 TCRs

Exley et al. have recently published that a significant fraction, up to 30%, of human iNKT-cell clones expressed p93 variants of the “invariant” TCR Vα24-Jα18 chain 4. Consistent with this, we identified four different iNKTα variants with single-nucleotide variations at the middle position of codon 93 within the Vα24-Jα18 junction from a small panel of 12 ex vivo FACS-sorted human TCR Vα24+/Vβ11+ T-cell clones. In these iNKT-cell clones, codon 93 of the iNKT TCRα chain encoded for a serine (Ser93) in 9 of the 12 clones, whereas three clones used a different residue, namely threonine (Thr93), asparagine (Asn93) or isoleucine (Ile93), at this position. All 12 clones bound strongly to CD1d-tetramers loaded with the α-galactosylceramide (αGC) KRN7000 (K7) verifying them as CD1d-restricted iNKT-cell clones.

We have recently shown that CD1d-tetramers loaded with the weak iNKT-cell agonist ligand OCH (OCH-tetramers) allow for discrimination of human iNKT cells according to their TCRs' CD1d-binding affinity 3. Here, the iNKT-cell clones using Ile93 and Asn93 showed no detectable binding of OCH-tetramers, while the clone containing Thr93 was brightly stained by OCH-tetramers (data not shown). OCH-tetramer binding of human iNKT-cell clones is strongly dependent on CDR3β loop variability and, consistent with this, different iNKT-cell clones using the major iNKTα variant Ser93 exhibited up to 200-fold differences in their OCH-tetramer staining intensity 3.

Natural amino acid variations at p93 of iNKTα exert a strong effect on iNKT TCR binding to CD1d

Because of this confounding effect of CDR3β variability, the influence of individual iNKTα chain variations on CD1d binding cannot be determined using natural human iNKT-cell clones. To overcome these hurdles, we generated recombinant replica of the four “WT” iNKT-cell TCRs that were present in the original iNKT-cell clones, as well as 12 novel hybrid iNKT-cell TCRs by pairing each of the four iNKTβ chains (1–4β, see Table 1) with each of the four iNKTα variants (i.e. Ser93, Thr93, Asn93 and Ile93). This enabled us to determine the impact of the four TCRα p93 variants on CD1d binding independently of the effect of CDR3β variations. In a series of surface plasmon resonance experiments, we directly determined the binding affinities of the four WT and the 12 hybrid iNKT-cell TCRs to three different CD1d/ligand complexes, i.e. CD1d/K7, CD1d/OCH and CD1d/β-galactosylceramide (βGC), at binding equilibrium. These experiments revealed differences of up to 28-fold in the CD1d-binding affinity at equilibrium between iNKT-cell TCRs made from different iNKTα variants but identical TCR Vβ11 chains (Fig. 1, Table 2). Of note, when paired with identical iNKTβ chains, TCRs using the major iNKTα variant, Ser93, always exhibited stronger binding affinities towards the three different CD1d/lipid complexes than TCRs which used any of the three minor variants.

Figure 1.

Natural iNKTα chain variations at position 93 impact on CD1d-binding affinity (KD). Each graph displays the dissociation constants of a panel of four iNKT-cell TCRs binding to three different CD1d/ligand complexes (K7-CD1d, OCH-CD1d, βGC-CD1d). Each iNKT-cell TCR consists of a specific iNKTβ chain (β1–4, see Table 1) and one of the four iNKTα chains (Ser93, Thr93, Asn93, Ile93), as determined by surface plasmon resonance (SPR; BiaCore3000) at equilibrium binding. Data points represent the mean KD (μM) from three independent experiments+SEM.

Table 1. TCR usage of four iNKTα variant containing human iNKT cell clones
iNKT cell cloneCodon 93 iNKTαVα24-Jα18 sequenceINKTβ (Vβ11)TCRβJ usageVβ – N/D/N – Jβ region predicted sequence
1AGCVVS-DRG“β1”2–5CASSE-FGGTERT-QETQYFGPGTRLLVL
2ACCVVT-DRG“β2”1–3CASSE-PS-SGNTIYFGEGSWLTVV
3AACVVN-DRG“β3”2–5CASSE-PPTGF-QETQYFGPGTRLLVL
4ATCVVI-DRG“β4”2–5CASSE-AHPTG-TQYFGPGTRLLVL
Table 2. Binding affinity of recombinant iNKT cell TCRs to CD1d loaded with glycolipids (KD; μM±SEM)
 iNKTα:Ser93iNKTα:Thr93
 K7OCHβGCK7OCHβGC
β10.4±0.052.1±0.12.0±0.21.7±0.111.7±0.934.2±5.2
β21.3±0.18.1±0.624.6±3.14.7±0.519.0±1.342.4±13.5
β30.9±0.17.8±0.48.5±1.02.7±0.320.3±2.472.2±8.4
β41.7±0.18.5±0.311.5±0.79.8±0.451.8±2.454.4±12.8
 iNKTα: Asn93iNKTα: Ile93
 K7OCHβGCK7OCHβGC
β10.6±0.04.5±0.22.6±0.12.9±0.23.3±0.85.6±1.1
β22.4±0.113.0±0.723.3±3.44.7±0.224.3±1.732.2±8.0
β35.0±0.754.2±13.2245.8±72.01.5±0.125.8±3.978.1±16.5
β418±1.083.2±6.445.8±11.613.1±0.556.4±4.160.6±30.8

In contrast, no consistent pattern of effect could be observed between the Thr93, Asn93 and Ile93 iNKTα variants. For example, TCRs containing the Asn93α variant supported strong binding to CD1d, which was comparable to Ser93α, when paired to either iNKTβ chain 1β or 2β (Fig. 1, upper two panels; Table 2). In contrast, iNKT-cell TCRs containing the same iNKTα variant, but in pairing with either 3β or 4β chains exhibited lower CD1d binding than any of the other iNKTα variants (Fig. 1, lower two panels; Table 2). These results indicated that complex interactions between the CDR3α and CDR3β loops in iNKT-cell TCRs affect CD1d binding.

Differential binding of natural iNKTα chain variants to CD1d-presenting endogenous ligands

In order to examine the impact of the four iNKTα variants on iNKT-cell TCR binding to cell surface CD1d-presenting endogenous ligands, we generated a panel of four fluorescent-conjugated iNKT-cell TCR tetramers made from each of the four iNKTα variants in complex with the iNKTβ chain of one clone (iNKT clone 3β). The four p93 variant iNKT-cell TCR tetramers stained OCH- and K7-pulsed CD1d-T2 cells with different mean fluorescence intensities; Ser93-containing iNKT-cell TCR tetramers showed the brightest staining and Ile93-containing iNKT-cell TCR tetramers the weakest staining for both OCH and K7 pulsed CD1d-T2, while Asn93- and Thr93-containing iNKT-cell TCR tetramers showed highly similar staining patterns. (Fig. 2A and B). In contrast, detectable staining of endogenous ligand presenting CD1d-T2 cells could only be achieved with the Ser93-containing iNKT-cell TCR tetramer and none of the other three variant iNKT-cell TCR tetramers (Fig. 2C). Neither CD1d-negative T2 cells (Fig. 2) nor anti-CD1d antibody (mAb cd1d42) pre-treated cells (data not shown) were stained by iNKT-cell TCR tetramers, thus confirming the specificity of the interaction.

Figure 2.

Impact of residue 93 of iNKTα TCR chains on human iNKT cell TCR binding to cell-bound CD1d. The binding of four recombinant PE-conjugated iNKT cell TCR tetramers made from the TCR Vβ11 chain of clone iNKT3 (3β) and the four natural iNKTα variant TCR chains (Ser93, Thr93, Asn93, Ile93) towards cell-bound CD1d loaded with synthetic ligands (A) OCH or (B) K7, or (C) endogenous ligands was examined by flow cytometry. Filled histograms represent staining of CD1d-negative T2 cells (T2). The data presented are representative of three independent experiments.

Molecular dynamics simulations reveal structural variations in the CDR3α loop

In order to further explore the hypothesis that substitutions at p93 of iNKTα might impact on the conformational stability of CDR3α, we carried out molecular dynamics simulations of the iNKT-cell TCR:CD1d complex with all four iNKTα variants (Fig. 3A). These simulations, in which both the protein and its environment were represented in atomic detail, revealed a subtle structural adaptation in the CDR3α loop upon substitution of Ser93 with other residues, becoming either more extended (in the Asn93 and Thr93 variants) or more compact (in the Ile93 variant) (Fig. 3B). These changes seem to arise from distinct patterns of hydrogen-bond interactions within the CDR3α loop, both directly with the side chain at p93 as well as mediated by water molecules (Fig. 3C). For example, in the Thr93 variant, the additional methyl group causes a slight rotation of this side chain relative to the Ser93 variant, distancing its hydroxyl group from the backbone of Arg95. This H-bond cannot be supported by the Ile93 variant, but it is largely preserved in the Asn93 substitution. In this case, however, the loop structure is perturbed as the water-mediated interaction with the backbone of Ser96 (Fig. 3C) pushes this residue away, on account of the larger size of the Asn93 substitution. Conversely, the absence of this water-bridge in the Ile93 variant allows the loop to be slightly more compact. These simulations therefore indicate that natural substitutions of the serine at p93 in the iNKTα chain could induce structural variations in the CDR3α loop with the potential to impact on binding properties of the CD1d interaction.

Figure 3.

Molecular simulations of an iNKT cell TCR-CD1d subcomplex. (A) All-atom simulation system comprising the protein–antigen subcomplex, consisting of the variable domain of the TCR and the binding domain of CD1d, in a 0.1-M NaCl solution. (B) Probability distribution of the distance between Cα atoms at positions 93 and 97 in the CDR3α loop of the iNKT cell TCR, for all four variants of the side chain at position 93. (C) Representative close-up snapshots of the conformation of the CDR3α loop in all four variants of the side chain at position 93, highlighting direct H-bonds as well as water bridges. The snapshots approximately correspond to the peaks in the probability histograms in panel (B). The pairwise variability (RMSD) of the loop conformations sampled within each simulation is ∼0.5 Å.

Discussion

The key importance of the “invariant” TCR Vα24-Jα18 rearrangement for human and mouse iNKT-cell TCR binding to CD1d has been previously analyzed by site-directed mutagenesis studies replacing residues within the invariant CDR3α loop with alanine 2, 7, and by crystal structure analysis 1, 6. Considerable similarities are apparent between mouse and human iNKT-cell receptor interactions with CD1d, including in a highly similar orientation and energetic footprint of TCR binding to CD1d 1, 2, 6, 8. Here, we showed that natural single-amino acid variants immediately upstream of the human invariant CDR3α loop, where a canonical serine at p93 (Ser93) of the iNKTα chain is replaced by threonine, asparagine or isoleucine, can exert a strong impact on iNKT-cell TCR binding to CD1d. The CD1d-binding affinities between these human iNKT-cell TCR p93 variants differed up to 28-fold, and clear-cut qualitative differences in their binding to both synthetic and endogenously bound CD1d ligands were also observed in cell-based assays using iNKT-cell TCR-tetramers. We and others have previously shown that quantitative differences in the strength of iNKT-cell TCR-binding affinity are directly linked to significant functional differences between T-cell clones 3, 9. However, direct functional comparison of human iNKT-cell clones bearing the different iNKTα p93 variants would not be expected to yield useful information on the impact of the p93 variants because of the confounding effect of CDR3β loop sequence variability on iNKT-cell TCR binding to CD1d 3. Notwithstanding, our finding that single-amino acid variations at p93 of the iNKTα chain can, independently of iNKTβ, translate into strong differences in CD1d binding, indicates a crucial role for the canonical serine (Ser93) for iNKT-cell function. Ser93 is directly adjacent to Asp94 and interacts via hydrogen bonds with both Arg95 and Arg105, all of which are critical for iNKT-cell TCR binding to CD1d 2. Consistent with this hypothesis, in silico molecular dynamics modeling for the four iNKTα variants was consistent with an effect of p93 residue substitutions on CDR3α loop conformation. Interestingly, for any given iNKTβ chain, the major iNKTα variant Ser93 conferred stronger binding than any of the minor variants to all tested CD1d/ligand complexes, including endogenously presented CD1d ligands on the cell surface. These results may therefore offer a plausible explanation for the predominance of the Ser93 variant in vivo, as immature T cells expressing the Ser93 iNKT-cell TCR would stand a better chance of being selected by cortical thymocytes to enter the iNKT-cell differentiation pathway 10. On the other hand, the germline sequence of the Vα24 gene encodes for a serine at this position, and so the greater prevalence of the Ser93 variant may simply be a function of the stochastic recombination process.

The single-amino acid variation in the four identified p93 variants all resulted from single-nucleotide variations at the precise point of Vα24-Jα18 recombination. As TdT is not required for iNKT-cell development 11, the change of a single nucleotide at the Vα24-Jα18 junction is likely to be due to a P nucleotide added during somatic recombination. Exley et al. recently reported that variations around p93 in the “invariant” iNKTα chain occur in a substantial proportion, i.e. up to 30% of human iNKT-cell clones 4, and our identification of three variants among a small panel of only 12 human iNKT-cell clones is consistent with this. In addition to the variants characterized in this report, others have previously identified iNKT-cell clones with non-canonical Vα24-Jα18 region sequences at positions 92, 93 and 94 of iNKTα (S92A, S93G, S93V, D94Y) 4, 5, but their CD1d-restriction has not been determined. In contrast to the single-nucleotide variants studied here, these other variants resulted from more complex nucleotide changes from the germline sequence, and may therefore occur less commonly in vivo. The mouse iNKTα chain uses a glycine at position 93, with rare variants containing alanine, valine or isoleucine 11 but the impact of these variants on iNKT-cell binding or function has not been assessed, and this specific site has not been included in any alanine scanning studies of the mouse iNKT-cell TCR:CD1d interaction. Interestingly, natural iNKTα variants at p93 also occur in the rat, where they have been shown to impact on iNKT-cell function 12. Furthermore, subtle variations at the TCR Vα7.2-Jα33 junction that are regularly found in mucosa-associated invariant T (MAIT) cells, which are another “invariant” T lymphocyte subset in humans restricted by the non-polymorphic MHC class I-related MR1 molecule 13. For both iNKT-cells and MAIT cells a role in antibacterial host defense has been recently postulated, and microbial glycolipids can be presented by CD1d to human iNKT-cell TCRs. Interestingly, the “invariant” iNKT-cell TCRα chain provides the main binding energy to the TCR's interaction with microbial CD1d presented ligands. Future studies may examine the role of the above characterized p93 variants in iNKT-cell TCR-mediated recognition of different microbial CD1d presented ligands. In conclusion, natural variations at position 93 of the Vα24-Jα18 junction of human iNKT-cell TCRs exert a strong impact on CD1d-binding affinity via an indirect effect on the “invariant” CDR3α loop. The role of these variants for iNKT-cell biology in health and disease remains to be further explored.

Materials and methods

Cell lines and clones

The study was approved by the local ethics committee. All donors gave informed consent. Human iNKT-cell clones were generated by FACSVantage sorting from healthy donors' PBMCs as previously described 3. Stable human CD1d-expressing T2-lymphoblast lines and clones were maintained in complete medium (RPMI, 10% FBS, 1% glutamine, 1% penicillin/streptomycin) 3.

Recombinant CD1d and iNKT TCR proteins

Human CD1d/β2-microglobulin (β2m)/lipid complexes were generated by in vitro refolding as described previously 3, 14, 15. In brief, both the extracellular region of CD1d containing an engineered specific BirA-tag biotinylation site at the C-terminus and β2m were expressed separately as inclusion bodies from E. coli BL21-DE3. Inclusion body proteins were then thoroughly washed, fully denatured and reduced in 5 M guanidinium-HCL and 20 mM DTT, and used for in vitro refolding. The proteins were refolded in the presence of defined lipid species (K7, OCH, βGC) using an oxidative refolding matrix containing agarose bead-immobilized recombinant GroEL minichaperone and recombinant DsbA enzyme. The correctly refolded CD1d/β2m/lipid complex of the expected size was purified from aggregates by repeated preparative grade and analytical size-exclusion chromatography using FPLC (Pharmacia, Sweden) with Hi Load 26/600 Superdex pg and Superdex GL 10/300 gel filtration columns (GE Healthcare, UK). Specific biotinylation of the complex at the BirA-tag site was performed using a BirA biotinylation kit (Avidity, USA). Stability of the refolded biotinylated CD1d proteins was assessed during repurification by FPLC gel filtration following overnight biotinylation. For production of fluorescent-conjugated CD1d-tetramers, biotinylated CD1d/β2m/lipid complexes were bound to PE-streptavidin (Molecular Probes, USA) at a 4:1 molar ratio.

Soluble TCR heterodimers were generated as described previously 16. Briefly, the extracellular region of each TCR chain was produced as inclusion bodies from E. coli BL21-DE3 (pLysS) following cloning into the bacterial expression vector pGMT7. To produce stable, disulfide-linked heterodimers, cysteines were incorporated into the TCR α- and β-chain constant regions, replacing residues Thr48 and Ser57 respectively. Expression, refolding and purification of the resultant disulfide-linked iNKT-cell TCR αβ heterodimers were carried out as previously described 17. The stability of the purified refolded TCR proteins was assessed by further gel filtration chromatography. Fluorescent-conjugated iNKT-cell TCR tetramers were generated using modified TCRβ chains, containing a C-terminus BirA-tag motif, which was specifically biotinylated using a BirA biotinylation kit (Avidity, USA). Biotinylated iNKT-cell TCRs were conjugated to PE-streptavidin (Molecular Probes, USA) at a 4:1 molar ratio, purified by FPLC (Pharmacia, Sweden) on an SD200 column (Pharmacia, Sweden) and concentrated to 1 mg/mL using Vivaspin20 concentrators (Vivascience, UK). Conformational stability of the purified TCRs was determined by further gel-filtration chromatography.

Flow cytometry

Human iNKT-cells were stained with PE-conjugated human CD1d tetramers or PE-anti-TCR Vα24 along with a panel of fluorescently conjugated antibodies: FITC-anti-human TCR Vβ11 (Serotec, UK); PerCP-anti-CD3, FITC-anti-CD3, allophycocyanin-anti-CD4 and allophycocyanin-anti-CD8 (BD Pharmingen). Staining was performed for 45 min at 4°C, and stained and washed cells were acquired on an FACSCalibur flow cytometer (Becton Dickinson). CD1d-expressing T2-lymphoblasts were stained with PE-conjugated iNKT-cell TCR tetramers for 45 min at 4°C, washed and analyzed on an FACSCaliber flow cytometer (Becton Dickinson). Propidium iodide was used to exclude dead cells. Data were processed using the CellQuest Pro software (BD Biosciences, USA).

Surface plasmon resonance

Streptavidin (∼5000RU) was amino-coupled to a Biacore CM-5 chip (BIAcore AB, UK) and ∼50 μg/mL biotinylated lipid-CD1d complexes or control proteins (βGC-CD1b and HLA-A2*01-NY-Eso-1(157–165) complex) were loaded on individual flow cells until the response measured ∼1000RU. Recombinant iNKT-cell TCRs were serially diluted and flowed over the protein-loaded flow cells at a rate of 5 or 50 μL/min for determination of equilibrium binding or kinetics. Responses were recorded in real time on a Biacore 3000 machine at 25°C, and data were analyzed using the BIAevaluation software (Biacore, Sweden) as described previously 3.

Molecular dynamics simulations of iNKT-cell TCR:CD1d complexes

All atom molecular dynamics simulations of the iNKT-cell TCR:CD1d complex were carried out in all variants of the amino acid at position 93 using NAMD2.7 18 and the CHARMM27/CMAP force field 19, 20. Lipids and sugar parameters were adapted by comparison with pre-existing parameters 21. Each molecular system contains ∼16 000 explicit TIP3P water molecules, and a neutralizing NaCl electrolyte buffer at a concentration of 100 mM. In total, each simulation system comprised ∼54 000 atoms. Atomic coordinates for the iNKT-cell TCR:CD1d complex at 2.5 Å 6 were obtained from the Protein Data Bank, entry 3HUJ. The simulations include only the variable domain of the TCR, and the binding domain of CD1d.

The preparation of each molecular system entailed a series of energy minimizations, to remove steric clashes between the protein-antigen complex and the added solvent, and upon substitution of Ser93 for Asn, Thr or Ile. This was followed by an equilibration phase, i.e. a series of simulations during which the structure of the protein-antigen complex is softly constrained, and gradually released, with respect to the initial crystal conformation, while the dynamics of the solvent environment is unrestricted. Finally, fully unconstrained simulations of each variant were carried out for 10 ns.

All simulations were carried out at constant temperature of 298 K and a constant pressure of 1 atm. Periodic boundaries were used. Electrostatic interactions were calculated using the Particle–Mesh Ewald algorithm with a real space cut-off of 12 Å. A Lennard–Jones potential cut-off at 12 Å was employed to calculate van der Waals interactions. Non-bonded pair-lists were re-generated every 10 steps using a distance cut-off of 14 Å. A time-step of 2 fs was used; bonds to hydrogen atoms were constrained via the SHAKE (protein, lipid-antigen) or SETTLE (water) algorithms.

Acknowledgements

J. P. S., S. M. and S. D. G. were supported by the Higher Education Funding Council for England (HEFCE). G. M. was the recipient of a Novartis Research Foundation student grant. K. W. H. was supported by a grant from the Swiss National Science Foundation. D. G. and J. F. G. are funded by the Max Planck Society and the DFG Cluster of Excellence “Macromolecular Complexes”.

Conflicts of interest: The authors declare no financial or commercial conflict of interest.

Ancillary