Lipidomic scanning of self-lipids identifies headless antigens for natural killer T cells

Significance Natural killer T (NKT) cells are an abundant T cell subset that recognizes CD1d and lipid antigens. Whereas most antigens are glycolipids from bacteria, NKT cells are selected, reside in tissues, and are activated in sterile conditions. Therefore, we comprehensively profiled human and mouse cells for self-antigens, identifying ceramides as an abundant and highly regulated class of self-antigens. These data conclusively identify a molecular target of NKT cells in the sphingomyelin cycle, which is regulated in cell stress and transformation. Further structural data disrupt the corecognition model, where NKT cells normally recognize protruding glycolipid headgroups that function as epitopes. Here, ceramide is a headless molecule, and a ternary crystal structure shows contact of the NKT cell receptor with CD1d itself.

Protein production.The truncated mouse CD1d ectodomain harboring a C-terminus His-6 tag (GSGLNDIFEAQKIEWHEHHHHHH) and a C-terminus biotinylation tag, along with β2M was cloned into the pBacp10pH vector and the recombinant protein was produced in the baculovirus expression system (2).In brief, 1% of P3 viral stock transfected 1 litre of High Five cells.The secreted protein was harvested after 72 hours and dialysed against 10 mM Tris pH 8.0 and 150 mM NaCl buffer for 16 hrs.Alternatively, mouse CD1d was produced in 293S Human Embryonic Kidney cells (293S) with or without the N-acetylglucosaminyltransferase-I gene (HEK-293S.GnTI−) by co-transfection with pHLsec vectors encoding truncated mouse CD1d ectodomain with a C-terminal biotinylation motif and His6-tag, or β2-microglobulin, using polyethylenimine essentially as described (3).The protein was purified using a two-step purification protocol including conventional Nickel Transfer Agarose (Ni-NTA) and size exclusion chromatography (4,5).
The genes encoding TCRα and TCRβ chains of the 2C12 TCR were codon optimised, synthesized (Genscript) and cloned into pET30 (Novagen) for expression in Escherichia coli strain BL21 and purified as inclusion bodies (IBs) (6).Inclusion bodies (120 mg and144 mg of α:β chain refold), in the presence of 1 mM DTT, were resuspended in 8 M urea 1000 mM Tris-HCl-pH 8.0, 0.2 mM Na-EDTA, 0.4 M arginine, 0.5 mM oxidised glutathione, 5 mM reduced glutathione and 0.2 M phenylmehtylsufonyl fluoride and X pepstatin, stirred at 4C for 24 hrs, dialysed in 10 mM Tris-HCl buffer with three changes over 16 hrs.A four-step purification process involving DEAE anion exchange, size exclusion, anion exchange (Hitrap Q) and Hydrophobic Interaction chromatography (HIC) was carried out, followed by SDS-PAGE gel electrophoresis under reducing and non-reducing conditions (7).
CD1d-TCR complex formation.Fraction 1 of lipids extracted from thapsigargin treated J774.2 macrophages was solubilized in 0.5% v/v tyloxapol in (TBS, pH8.0).Solubilized fraction 1 was incubated with insect cell or mammalian derived soluble mCD1d in a 6:1 lipid:protein molar ratio for 16 hours at room temperature and passed over a Superdex 200 16/60 column (GE Healthcare) using TBS, pH8.0, to remove excess detergent.The TCR trap assay (8,9) was performed by mixing a 1:1 molar ratio of mCD1d containing lipids from fraction 1 with soluble 2C12 TCR, and then passage over a Superdex 200 16/60 column in TBS, at a flow rate of 0.5mls/min.Fractions obtained from the gel filtration column were analysed by SDS-PAGE gel.
The generation of 2C12 TCR-CD1d-ceramide ternary complex for crystallization was generated in a similar manner.In brief, C42:2 ceramide was solubilized in 0.5% v/v tyloxapol and mixed with mCD1d in a 5:1 molar ratio, prior to incubation at room temperature overnight.Excess tyloxapol detergent was removed from mCD1d using a Superdex 300 10/300 column (GE Healthcare), and mCD1d loaded containing ceramide was mixed with the 2C12 TCR in a 1:1 molar ratio and incubated overnight at 4C.Ternary complex was isolated by performing a size exclusion chromatography (S200 16/60) (GE Healthcare).
Lipid extractions and analysis.CD1 protein fractions were extracted of lipids by the Bligh and Dyer method.For shotgun nano-ESI-MS, extracted lipids were analysed on a Thermo Fisher LXQ Ion Trap in the negative ion mode.Eluents were normalized to 10 µM based on protein input and 10 µL were loaded on a reversed-phase HPLC column Agilent Poroshell EC-C18 column (1.9-micron, 3 x 50 mm) with a guard column (3 x 5 mm, 2.7 µm; 1260 HPLC system) and monitored with Agilent 6530 Accurate-Mass ESI-QToF-MS using previously published gradient (10).Protein normalized eluents were loaded on a reversed-phase HPLC column and monitored with ESI-QToF-MS.Lipids from early eluting co-complexes enriched for CD1d-TCR were compared late fractions containing CD1d and TCR monomers, to identify targets whose intensity was >10-fold higher in the co-complex fractions.We removed the redundantly detected isotopes and alternate adducts, as well as salt clusters identified based on low mass defects, yielding 15 molecular events in 6 lipid classes as the TCR-trap antigens.Similar results were obtained using High Five insect cells or human 293S cells.
Mice.C57BL/6, Traj18 -/-and CD1d1/2 -/-mice, all on C57BL/6 background, were purchased from Jackson Laboratory (USA).All mice were maintained under specific pathogenfree conditions in the Animal Care Facility of Ghent University as approved by the Institutional Animal Care and Ethics Committee.Experiments were performed with age-and sex-matched mice at 10-12 weeks of age.Cell suspensions from the thymus were resuspended in PBS containing 1 mM ethylenediaminetetraacetic acid and 0.5% BSA.
Flow cytometry and antibodies.Murine thymocytes were isolated as described (11).
X-ray crystallography and structure determination.The isolated ternary complex fractions were pooled and concentrated to 5 mg/ml.Using reported methods (13), crystals were grown in conditions containing 12-16% PEG 3350, 8% Tacsimate, pH 5.0 using the hanging drop vapour diffusion method at 4C.Crystals with good morphology were flash frozen in the mother liquor containing 10% glycerol as a cryoprotectant.The crystal diffraction data were collected at the MX2 beamline at the Australian Synchrotron.The data were processed with the program XDS (14) and scaled using SCALA (15) in the CCP4 program suite.Using 2C12 TCR and mCD1d without the lipid in the 2C12 TCR-mCD1d-KRN7000 (GalCer) complex (PDB 6BNK) as two independent search ensembles, the structure solution was obtained by molecular replacement with the help of PHASER-MR program (16).An initial rigid body refinement followed by iterative rounds of refinement was performed with the Phenix program.Model building of amino acid residues was carried out in COOT prior to running the refinement cycles (PyMOL Molecular Graphics System, version 2.0).

Surface Plasmon resonance (SPR):
Affinity measurements were carried out at 25C using a buffer containing 10 mM Hepes-HCl, pH 8.0 and 150 mM NaCl on a Biacore 3000 instrument.The biotinylated mouse CD1d monomers loaded with ceramide were immobilised on the streptavidin (SA) sensor chip to ~2500 response units.Then, 2C12 TCR was passed over the chip at a flow rate of 5l/min with increasing concentrations from 0-40 M.Using a one-site binding model, the equilibrium dissociation constant (KD) was calculated using Graph pad prism software 6.0d.

Statistics:
Statistical analyses were performed using Genstat V.19 and GraphPad Prism V.9.For the plate binding assay, an exponential curve of the form y = A + B*R X was fitted to the data, where A (asymptote), A-B (starting point when X = 0), and R (speed at which the asymptote is approached), and X (concentration) generate an accumulated analysis of variance for each pair of compounds, using the F-test.

Fig. S2 .
Fig. S2.Protein analysis of fractionsCD1d from High Five cells and 2C12 TCR heterodimers were mixed in equimolar ratios and loaded onto a size exclusion column, collected as fractions and subjected to polyacrylimide gel electrophoresis to estimate molar ratios and protein yields needed to normalize lipid fractions.

(Fig. S4 .
Fig. S3.Nano-electrospray ionization mass spectrometry (nano-ESI) analysis of CD1dbound lipids.Total mouse CD1d (mCD1d) expressed in insect High Five cell line was captured and treated with chloroform and methanol to detect eluted lipids by negative mode nano-ESI MS, finding numerous bright ions corresponding to endogenous CD1d lipids (mCD1d-endo) loaded in cells.mCD1d-endo was treated with macrophage neutral lipids (fraction 1 purified from thapsigargin treatment) and then mixed with recombinant 2C12 TCR and subjected to size exclusion chromatography to detect CD1d-TCR complexes in early (31-33) and CD1d and TCR monomers in late (37-38) fractions, which were simultaneously monitored by nano-ESI-MS.

Fig. S6 .
Fig. S6.Deoxyceramide recognition by murine NKT TCRs.a) Summary of NKT clones and variable (V) or joining (J) gene usage for TCR or TCR chains, as per (1).The complementary determining region amino acid sequences are shown with residues partly or wholly encoded by non-germline insertions highlighted in bold.b) Mock or antigen treated mouse CD1d-tetramers were assessed for staining of BW58 lines expressing the indicated TCRs with cells gated for similar surface TCR expression +/-SEM from two experiments performed with mouse CD1d expressed in High Five cells and three experiments using proteins from 293S GNtI-/-.Plots from one representative experiment is shown in Figure 3f.c) Mouse CD1d-tetramers in 293S GNtI-/cells were assessed for staining of type I NKT TCR-transduced BW58 lines expressing the TCRs indicated in (a).Graphs show MFI fold-increase compared to vehicle treated mCD1d-tetramers from three experiments performed, one representative shown in (d) Numbers in d represent mean fluorescence intensity (MFI) on cells with similar surface TCR expression.Each dot shown in graphs for (b) and (c) represents an independent experiment .

dFig. S7 .
Fig. S7.Creation and tetramer staining of a human NKT cell-enriched cell line.a) CD1d-PBS-57 tetramer high cells were sorted from PBMC from a healthy donor and expanded.After the first expansion, the sorted cells were analyzed by flow cytometry, expanded again, yielding a line that is 96.5 percent Type I NKT cells, which was used for further experiments.b) Single tetramer stainings with a series of tetramers that were antigen or mock treated in-house in parallel.