The identification of the endogenous ligands of natural killer T cells reveals the presence of mammalian α-linked glycosylceramides

Abstract

Glycosylceramides in mammalian species are thought to be present in the form of β-anomers. This conclusion was reinforced by the identification of only one glucosylceramide and one galactosylceramide synthase, both β-transferases, in mammalian genomes. Thus, the possibility that small amounts of α-anomers could be produced by an alternative enzymatic pathway, by an unfaithful enzyme, or spontaneously in unusual cellular compartments has not been examined in detail. We approached the question by taking advantage of the exquisite specificity of T and B lymphocytes and combined it with the specificity of catabolic enzymes of the sphingolipid pathway. Here, we demonstrate that mammalian immune cells produce constitutively very small quantities of α-glycosylceramides, which are the major endogenous ligands of natural killer T cells. Catabolic enzymes of the ceramide and glycolipid pathway tightly control the amount of these α-glycosylceramides. The exploitation of this pathway to manipulate the immune response will create new therapeutic opportunities.

Figure 1. β-GluCer Is Not the Endogenous Ligand of NKT Cells

(A) Comparison of the stimulatory capacity of commercial (closed circles) and in-house(open circles) C24:1 β-GluCer presented by DCs to DN32.D3 NKT cells. (B) CD1d tetramers loaded with control lipid, C24:1 β-GluCer, or C24:1 α-GalCer were used to stain DN43.D3 NKT hybridoma cells and splenocytes and were examined by flow cytometry. (C) β-GluCer was digested with recombinant GBA for 2 hr at 37°C, analyzed by TLC for confirmation of complete digestion (left panel), and tested functionally for its ability to stimulate DN32.D3 NKT cells when presented by WT splenocytes before (squares) and after (circles) digestion (right panel). (D) The stimulatory activity of commercial β-GluCer was blocked by L363 (10 μg/ml; diamonds) and 20H2 (5 μg/ml; triangles), but not control (squares) antibodies (left panel), and could be recovered with Pisum Sativum lectin (right panel), an agglutinin specific to α-glucose (closed squares). Unbound fractions were used as a negative control (open circles). Seven two-fold dilutions of the purified material were used. All experiments were repeated a minimum of three times with similar results. See also Figure S1.

Figure 2. The Anti-CD1d-α-GalCer Antibody L363 Blocks the Autoreactivity of CD1-Expressing Cells toward NKT Cells

(A) IL-2 production by Vα14-expressing DN32.D3 NKT cells was tested after a 24 hr exposure to increasing numbers of RBL-CD1d cells in the presence of L363 (red dots) or control (black dots) antibody (10 μg/ml; left panel). (B) Tested under similar conditions, the non-Vα14 NKT cell hybridoma, TBA7, was not affected by antibody treatment. (C) The stimulatory activity of WT thymocytes toward DN32.D3 cells was tested in the presence of control (black dots) or L363 (red dots) antibody (20 μg/ml). (D) A total of 2 × 10⁴ DC3.2 cells treated for 16 hr with increasing concentrations of LPS were used to stimulate DN32.D3 cells in the presence of control (black dots) or L363 (red dots) antibody (10 μg/ml). (E and F) Stimulation of DN32.D3 cells (E) and TBA7 cells (F) was tested against RBL-CD1d cells (black dots) or RBL-CD1d saposin-knockdown cells, in which saposin expression was reduced to undetectable levels with interfering RNAs. IL-2 production was measured with the NK reporter cell line from triplicate wells. Experiments shown are representative of at least five independent experiments. See also Figure S2.

Figure 3. The L363 Antibody Binds with Different Affinity to Both CD1-α-GalCer and CD1-α-GluCer Complexes but Does Not Recognize β-Glycosylceramide-CD1d Association

(A) Predicted L363 binding to glycosylceramides. In the crystal structure, L363 contacts α-GalCer with two H-bonds, G50 interacts with the axial 4″-OH, and R32 is specific to the sphingosine chain (Protein Data Bank [PDB] ID 3UBX, left panel). Modeling the interaction with α-GluCer illustrates the loss of the H-bond with G50, which as a result of equatorial rather than axial position of 4″-OH, leads to weaker L363 binding affinity (middle panel). However, N31 and R32 together form a cap over the sugar and bind through van der Waals interactions, predominantly through N31. The upright positioning of β-GalCer (modeled with the crystal structure of mCD1d-sulfatide [PDB ID 2AKR]) would prevent L363 binding as a result of steric clashes (right panel). (B) The binding of the various α- and β-anomers of glycosylceramides mentioned and depicted in Table 1 was measured by SPR on a Biacore T200 instrument. Single-cycle analysis was performed on CM5 chips with 500–1,000 response units of immobilized antibody and increasing concentrations of the various CD1-lipid complexes. Analysis was performed with the Biacore T200 Biaevaluation global analysis software with subtracted sensorgrams (L363 antibody was used as a control antibody). Similar results were obtained in five independent experiments.

Figure 4. Endogenous NKT Ligands Recognized by L363 and L317 Antibodies Are Monoglycosylceramides and α-Linked

(A) Liquid chromatography-MRM-MS identified an 810 Da monoglycosylceramide in DC3.2 and RBL-CD1d cells. The lipid content of L363 and L317 antibody immunoprecipitations from DC3.2 and RBL-CD1d cells (2 × 10⁹ cells) was analyzed by MRM-MS. Ionization-transition profiles were defined for α-GalCer, C24:1 α-GalCer (shown in the figure), psychosines, and phytopsychosines. Untransfected RBL cells (5 × 10⁹ cells) were used as a negative control. (B) The α-linkage of the anomer was demonstrated by analysis of control antibody (ctl, normal rabbit IgG), anti-α-GalCer (āα, left), and anti-L363 (right) immunoprecipitations of C¹⁴-labeled RBL cells by TLC. C¹⁴UDP-galactose was used to label 5 × 10⁷ RBL cells for 48 hr. In separate experiments, protein extracts were immunoprecipitated with L363, whereas lipid extracts were immunoprecipitated with a purified anti-α-GalCer rabbit antibody. Immunoprecipitates were extracted with a 2:1 chloroform-methanol mixture and analyzed by high-performance TLC next to standards (std.) for α-GalCer and β-GalCer on plates that had been pretreated with borate. Standards were cut and stained with cerium-ammonium-molybdate stain and charring. The experimental part of the TLCs was fluorographed on Biomax MS films for 4–8 weeks. Each experiment was repeated a minimum of three times. See also Figures S3 and S4.

Figure 5. Catabolic Enzymes Control the Availability of α-Glycosylceramides

(A) High-performance TLC was used to separate glycosylceramides (left panel) and lyso-glycosylceramides (right panel) before and after digestion with recombinant GLA. Lipids were visualized with a cerium ammonium molybdate stain. Abbreviations are as follows: Gal, GalCer; Glu, GluCer. (B) Samples from α-GalCer (left) and α-psychosine (right) were tested for their stimulatory ability toward DN32.D3 NKT cells before (empty circles) and after (filled circles) digestion with recombinant GLA. DC3.2 cells (20,000 cells/well) were used as antigen-presenting cells. A representative experiment out of three independent experiments is shown. (C) DC3.2 cells were differentiated with LPS, treated with inhibitors of α-glycosidases (GLAi and/or GAAi), 1-deoxygalactonojirimycin (0.5 μM), and 1-deoxy-gluconojirimycin (2.0 μM) or with ceramidase inhibitors ASAHLi (20 μM;Li et al., 2012) or ASAH1i (carmofur, 1.0 μM) for 24 hr, and used to stimulate DN32.D3 NKT cells. A control (empty circles) is included in each panel for comparison with inhibitors (filled circles). (D) Confocal microscopy of α-GalCer in RBL-CD1d cells by indirect immunofluorescence. Flattened Z series are presented. Anti-α-GalCer antibody is in red, whereas anti-CD1d antibody is in green. Control stains include normal rabbit Ig, 14.4.4 antibody, and secondary antibodies (a and h). In (b), (c), and (d), cells were incubated in the presence of 10 μg/ml β-GalCer at the time of staining, whereas in (e), (f), and (g), cells were incubated in the presence of 10 μg/ml α-GalCer. An overlay of each group is presented in (d) and (g). In (h)–(k), cells were either untreated (i) or treated with 0.5 μM 1-deoxygalactonojirimycin and 1.0 μM carmofur (j) or 2.0 μg/ml α-GalCer (k) for 24 hr in 0.05% serum-free media containing BSA. All cells were permeabilized with 0.05% saponin. All scale bars represent 20 μm. See also Figures S4–S6.

Figure 6. Thymocytes Present More Than One α-Glycosylceramide

(A) Titration of DN32.D3-cell-stimulation inhibition by L363 antibody with the use of thymocytes (left panel) or RBL-CD1 d cells as antigen-presenting cells. Seven two-fold dilutions were tested from a 20 μg/ml stock. In the lower panel, the percentage of inhibition was plotted as the percentage of maximal response (100%) for RBL-CD1d cells (black symbols) and thymocytes (empty circles). (B) Day 14.5 thymic lobes were cultured for 18 days in the presence of 40–60 μg of antibody. 14.4.4 s was the negative control antibody, and 20H2 was the positive control antibody. In this particular experiment, percentages of CD1-PBS57-positive cells were 0.27%, 8.02%, 1.14%, and 0.27% for adult thymus, 14.4.4 s, L363, and 20H2, respectively. Similar results were obtained on four other thymic lobes. The experiment was repeated three times. FTOC, fetal thymic organ culture. (C) Mice were treated with either 14.4.4 control or L363 antibody for 4 weeks and examined for thymic NKT cell numbers. The lowering of NKT cell numbers in the L363-treated group was significant (p = 0.0147). (D) Staining of WT and Gla^−/− thymic sections with purified anti-α-GalCer rabbit antibody (green) and anti-CD3 antibody (red). Control staining was with secondary antibodies only. Sections (a), (b), and (c) were magnified 20×. Section (d), from a WT thymus, was magnified 40×. Similar images were obtained in three independent experiments on WT and Gla^−/− thymi. Scale bars represent 100 μm in (a)–(c) and 50 μm in (d). See also Figure S6.