Cell wall glycosphingolipids of Sphingomonas paucimobilis are CD1d-specific ligands for NKT cells


  • Venkataraman Sriram,

    1. Department of Microbiology and Immunology, Indiana University School of Medicine, Indianapolis, USA
    2. The Walther Oncology Center, Indianapolis, USA
    3. The Walther Cancer Institute, Indianapolis, USA
    Search for more papers by this author
  • Wenjun Du,

    1. Department of Chemistry, University of California, Davis, USA
    Search for more papers by this author
  • Jacquelyn Gervay-Hague,

    1. Department of Chemistry, University of California, Davis, USA
    Search for more papers by this author
  • Randy R. Brutkiewicz

    Corresponding author
    1. Department of Microbiology and Immunology, Indiana University School of Medicine, Indianapolis, USA
    2. The Walther Oncology Center, Indianapolis, USA
    3. The Walther Cancer Institute, Indianapolis, USA
    • Department of Microbiology and Immunology, Indiana University School of Medicine, Building R2, Room 302, 950 West Walnut Street, Indianapolis, IN 46202-5254, USA, Fax: +1-317-274-7592
    Search for more papers by this author


The current consensus on characterization of NKT cells is based on their reactivity to the synthetic glycolipid, α-galactosylceramide (α-GalCer) in a CD1d-dependent manner. Because of the limited availability of α-GalCer, there is a constant search for CD1d-presented ligands that activate NKT cells. The α-anomericity of the carbohydrate is considered to be an important requisite for the CD1d-specific activation of NKT cells. The gram-negative, lipopolysaccharide-free bacterium Sphingomonas paucimobilis is known to contain glycosphingolipids (GSL) with α-anomeric sugars attached to the lipid chain. Here, we report that GSL extracted from this bacterium are able to stimulate NKT cells in a CD1d-specific manner. In addition, soluble CD1d–Ig dimers loaded with this lipid extract specifically bind to NKT cells (but not conventional T cells). Further studies on the S. paucimobilis GSL could potentially lead to other natural sources of CD1d-specific ligands useful for NKT cell analyses and aimed at identifying novel therapies for a variety of disease states.





CD1KO mice:

CD1d1-knockout mice




Lipid antigens, including phospholipids and glycosphingolipids (GSL), are presented by CD1d – a subset of the CD1 family of MHC class I-like molecules – to specialized immune effector cells called NKT cells 1. Located on a different chromosome than the MHC, five different CD1 genes encode CD1 molecules – CD1a, CD1b, CD1c, CD1d and CD1e – in humans, whereas only two homologues of CD1d (CD1d1 and CD1d2) are present in mice 2. On the basis of the crystal structure of mouse CD1d1 3, it is predicted that the fatty-acyl chains are buried in the hydrophobic pocket of the molecule with the hydrophilic head-group of the lipid antigen available outside the molecule for its interaction with the NKT cell receptor. Even though the lipid-binding groove of CD1d is widely accommodative of many lipid groups, it is believed that only the sugar structures in α-anomeric orientation are able to stimulate NKT cells 4. Because of the lack of physiological glycolipids with terminal α-anomeric sugars, it is hypothesized that both altered self-glycolipids during a pathological process and exogenous antigenic glycolipids could be potential ligands presented by CD1d 4.

The presence of GSL in the cell wall of some bacterial strains indicates the possibility of these bacterial GSL being presented by CD1d. Although human CD1a, b and c molecules are known to present mycobacterial antigens to human NKT cells, there is a lack of substantial evidence to show that CD1d has the ability to present these microbial lipids 5. In a recent study, Fischer et al. 6 were able to show that mycobacterial phosphoinositolmannosides (PIM) could bind to soluble CD1d and activate human and mouse NKT cells. However, only a fraction of NKT cells detected by α-galactosylceramide (α-GalCer) could be stained by a CD1d tetramer loaded with mycobacterial PIM 6. A recent study 7 was able to show the binding of synthetic α-GalNAc containing GSL to cell surface CD1d, but no functional studies were done to demonstrate the T cell relevance of this antigen. To date, α-GalCer – originally isolated from marine sponges 8 – is the only available ligand for phenotypically and functionally characterizing NKT cells.

Remarkably, all species of the gram-negative bacteria Sphingomonas contain GSL in their cell envelope and are completely devoid of LPS 9. Sphingomonas GSL characteristically contain 2-hydroxy-tetradecanoic acid with an α-anomeric sugar attached directly to the lipid chain 9. Specifically, the GSL mixture from S. paucimobilis has been found to contain two major GSL, namely a tetraglycosylated GSL (GSL-4A) and a monoglycosylated α-glucuronosylceramide (α-GlcUCer) 9, 10. These GSL of S. paucimobilis have already been shown to induce TNF-α production in human mononuclear cells 11, stimulate phagocytosis and phagosome–lysosome fusion 12, and activate the human complement system 13. Because of the close structural similarities of the α-GlcUCer from S. paucimobilis and the NKT cell activating ligand, α-GalCer, we investigated the activating potential of this bacterial glycolipid on CD1d-mediated antigen presentation. We describe here a unique species of Sphingomonas GSL capable of stimulating NKT cells in a CD1d-dependent manner.


CD1d-restricted stimulation of NKT cells by S. paucimobilis GSL

To determine the potential of GSL from S. paucimobilis to act as CD1d-specific ligands for NKT cell activation, crude lipid extracts from bacterial cells were used in the preliminary experiments. The bacterial lipid extract was able to stimulate all tested NKT cell hybridomas in a dose-dependent manner (Fig. 1). Although the degree of stimulation varied between the different Vα14+ (DN32.D3, N38-2C12, N38-3C3) hybridomas, we consistently found that the two α-GalCer-dependent hybridomas (i.e., N38-2C12 and N38-3C3) were stimulated to a maximum extent when the CD1d1+ cells were treated with the S. paucimobilis lipid extract (Fig. 1 and data not shown). The results shown were verified by an ELISA analysis of IL-13 and GM-CSF secretion by the NKT cells and the trends were similar to those observed for IL-2 secretion (data not shown). Our laboratory has been able to show the secretion of IL-13 and GM-CSF by the NKT cell hybridomas in a CD1d-dependent manner (unpublished observations). No significant increase in the NKT cell stimulation was observed with the Vα5+ hybridoma, N37-1A12, even at the highest lipid concentrations tested, demonstrating the Vα14+ TCR-restricted specificity of the bacterial lipids

Figure 1.

Cell wall GSL of S. paucimobilis activate NKT cells in a dose-dependent manner. LMTK-CD1d1WT cells (2.5×104 cells per well) were pulsed with different concentrations of a crude lipid mixture in 100 μl of assay medium. After a 4-h pulse at 37°C, the cells were extensively washed in PBS to remove lipids bound non-specifically and were cocultured with NKT cell hybridomas (5×104 cells/well). After 20 h, the supernatants were harvested and analyzed for IL-2 production by ELISA. As N38-2C12 cells do not secrete any detectable IL-2 on stimulation with vehicle-pulsed APC, the net IL-2 production is shown in the top panel. The data for the remaining NKT hybridomas (in the lower panel) are presented as the percent increase in IL-2 production compared to stimulation with vehicle-treated cells (set at 100%). The experiment is representative of more than seven performed.

Notably, the OVA-specific DO11.10 CD4+ T cell hybridoma – used as a negative control – did not produce any detectable IL-2 either when stimulated with the lipid alone or with lipid-pulsed fibroblasts. The specificity of the S. paucimobilis GSL in the CD1d1-NKT system was thus established by these experiments.

Mouse and human CD1d can present S. paucimobilis GSL to NKT cells

Total splenocytes have both classical antigen-presenting cells and NKT cells. Therefore, using splenocytes as an antigen-presentation system for the study of NKT cell stimulation by exogenous lipids has been well recognized 14. We used freshly isolated splenocytes from both C57BL/6 WT and CD1d1 knockout (CD1KO) mice to test the ex vivo splenocyte stimulation capacity of S. paucimobilis GSL. Splenocytes from WT (but not CD1KO) mice responded to S. paucimobilis GSL in a dose-dependent manner (Fig. 2A). The lack of a significant increase in cytokine secretion in α-GlcUCer-pulsed splenocytes from CD1KO mice is a direct demonstration of the CD1d-dependence for this stimulation.

Figure 2.

S. paucimobilis GSL stimulate both mouse and human NKT cells in a CD1d-restricted manner. (A) Freshly isolated splenocytes (5×105 cells per well) from C57BL/6 WT and CD1KO mice were stimulated with different doses of alkali-resistant lipids for 24 h at 37°C. IFN-γ (left panel) and IL-4 (right panel) secretion was determined by ELISA. The results are expressed as the net lipid-specific cytokine release. (B) Human C1R and C1R-CD1d cells (5×105) were pulsed with either vehicle or 250 μg/ml of alkali-resistant lipids for 4 h at 37°C. The cells were extensively washed and cocultured with 5×104 human NKT cells for 24 h. IL-4 secretion was measured by ELISA.

Human C1R B-lymphoblastoid cells or C1R transfected with the human CD1d cDNA (C1R-CD1d cells 15) preincubated with S. paucimobilis GSL were co-cultured with a human PBMC-derived NKT cell line to determine whether these lipids can stimulate human NKT cells in a CD1d-dependent manner. Significant amounts of IL-4 were produced by human NKT cells stimulated by C1R-CD1d cells (but not the parental C1R cell line) pulsed with the S. paucimobilis GSL (Fig. 2B). This bacterial-lipid-dependent stimulation of human NKT cells by GSL-pulsed CD1d transfectants resulted in a ∼30–60% increase in IL-4 production compared with vehicle-treated C1R cells.

Requirement of processing for the presentation of S. paucimobilis GSL

Because of the presence of a tetraglycosidic moiety attached to the ceramide in GSL-4A, we wanted to determine if processing was required for the presentation of GSL-4A and α-GlcUCer mixture. LMTK-CD1d1WT cells fixed in 0.05% paraformaldehyde prior to the lipid pulse induced no detectable antigen presentation to the ligand-specific hybridomas, N38-2C12 (Fig. 3B) and N38-3C3 (data not shown), or the canonical NKT cell hybridoma, DN32.D3 (Fig. 3A). On the other hand, fixing the LMTK-CD1d1WT cells after a 4-h period of lipid pulse resulted in low but detectable levels of IL-2 production by all the hybridomas (Fig. 3), showing that some amount of processing is required for the presentation of these bacterial lipids. These results are in accordance with those shown by Prigozy et al. 16, and imply that GSL with more than one carbohydrate moiety attached to the ceramide bind to cell surface CD1d, but are endocytosed and processed to a simple GSL before being presented to NKT cells. The anti-mouse-CD1d monoclonal antibody 1H6 17 (but not the negative control Ab) completely blocked the bacterial GSL presentation to NKT cells, confirming the CD1d-dependence of GSL recognition (Fig. 3).

Figure 3.

GSL from S. paucimobilis require endosomal (acidic pH) processing. (A, B) LMTK-CD1d1WT cells were pulsed with either vehicle or 250 μg/ml of alkali-digested lipids for a total of 4 h, washed and fixed either before or after the lipid pulse with 0.05% paraformaldehyde, and cocultured with the DN32.D3 (A) and N38-2C12 (B) NKT cell hybridomas. Antigen- and CD1d-dependent IL-2 production was measured by ELISA after a 20-h culture period. The anti-mouse-CD1d mAb 1H6 (or negative control Ab) was included during the lipid pulse as well as during the coculture. (C, D) Chloroquine (100 μM final concentration), primaquine (100 μM final concentration), and bafilomycin A1 (100 nM final concentration) were added to the LMTK-CD1d1WT cells from a 100× DMSO stock. The "inhibitor→lipid" bars denote pretreatment of indicated inhibitors for a 30-min period before the lipid pulse. The "lipid→inhibitor" bars indicate treatment of cells with the inhibitors for an additional hour in the presence of lipid after a 3-h pulse with the lipid (i.e., GSL) mixture. All lipid pulse times were maintained at 4 h. The "inhibitor alone" controls were treated with the drugs for the total 4-h pulse period. After treatment, LMTK-CD1d1WT cells were extensively washed to remove inhibitors and lipids, and fixed in 0.05% paraformaldehyde before coculture with the DN32.D3 (C) and N38-2C12 (D) NKT cell hybridomas. IL-2 production was measured by ELISA after a 20-h coculture.

We have previously shown that the pharmacological inhibitors including chloroquine 18, 19, primaquine, and bafilomycin A1 20, that inhibit the endosomal recycling pathway and lysosomal enzyme activities by different modes of action, inhibit antigen-presentation by CD1d without altering its cell surface levels 1, 17. We thus assessed the requirement for intracellular processing of Sphingomonas GSL. Blocking the vesicular acidification of intracellular vacuoles resulted in a substantial inhibition of antigen presentation, indicating a need for GSL antigen-processing before presentation to NKT cells (Fig. 3C and D). The drugs used alone (without the lipid pulse) served as internal experimental controls (Fig. 3C and D). Treatment of the cells with lipids for 3 h prior to a 1-h incubation of these cells with chloroquine, primaquine, or bafilomycin did not completely eliminate antigen presentation. A lipid pulse before chloroquine treatment completely rescued the enhanced antigen presentation by S.paucimobilis lipids to DN32.D3 (Fig. 3C). However, even a brief pretreatment of cells with these pharmacological inhibitors before the lipid pulse substantially blocked antigen presentation compared with lipid alone (Fig. 3C and D). These results suggest that there is a requirement for GSL-antigen-processing or endosomal uptake and loading during the 3-h pulse period.

It should be noted that pure α-GalCer has been shown to activate NKT cells in a CD1d-restricted manner without the need for any processing 4, 14, 16. However, the lipid mixtures used in our experiments also contain a tetraglycosylated lipid (GSL-4A) along with the simple α-GlcUCer. It is our belief that GSL4-A undergoes endo-lysosomal trimming to the core α-GlcUCer structure for efficient presentation.

The NKT cell stimulating ligand is alkali-stable

When the alkali-labile lipids (glycerolipids and some phospholipids) were removed from the extract by mild alkaline hydrolysis, the alkali-resistant GSL were able to efficiently stimulate NKT cells (data not shown). It has been shown that following alkaline hydrolysis, two major GSL, namely α-GlcUCer and GSL-4A, are recovered from the S. paucimobilis lipid extract 10. This was further confirmed by TLC analysis (Fig. 4). An orcinol+ lipid migrating just below α-GlcUCer and the orcinol lipids in the spotting origin observed in the crude lipid mixture (Lane 1 of TLC in Fig. 4) disappeared after alkali-treatment (Lane 2 of TLC). However, an alkali-resistant glycolipid was observed with a relative migration between α-GlcUCer and GSL-4A. This could be a degradative product of GSL-4A with three sugar residues 9, 21.

Figure 4.

Both α-GlcUCer and GSL-4A are alkali-stable. An aluminium TLC plate with silica-gel matrix (Merck) was pre-run with the chromatographic solvent system and warmed to dry before sample application. Lipids (50 μg) were applied on 5-mm lanes using a microcapillary pipette. The samples were chromatographed using chloroform:methanol:acetic-acid:water (25:15:4:2 by volume) to a height of 6 cm and air-dried free of solvent. The lipids were detected colorimetrically by orcinol spray. Lane 1 was loaded with crude lipid extract and Lane 2 with mild-alkali-digested lipid derived from 50 μg of the crude lipid mixture. The white arrow is identified as GSL-4A and the black arrow indicates α-GlcUCer as determined by their relative mobility. The cartoon depiction of a general glycolipid structure modeled on the published α-GlcUCer structure 9, 10 is accompanied by a table comparing α-GalCer with bacterial lipids.

The alkali-resistant GSL – α-GlcUCer (α-glucuronosyl ceramide) and GSL-4A – both contain carbohydrate units in α-anomeric orientation. In a separate set of experiments, GSL from S. paucimobilis were partitioned into neutral and acidic fractions by the solvent-phase partitioning method 22, 23. The negative charge imparted by glucuronic acid (6′-COOH group instead of 6′-CH2OH group as in α-GalCer) drives partitioning of α-GlcUCer into the acidic fraction, but a majority of the α-GlcUCer was also recovered in the neutral fraction as confirmed by TLC analysis (data not shown). Although LPS (which fractionates into the interphase during solvent partitioning) has been shown to activate NKT cell clones 24, the fact that Sphingomonas cell walls lack LPS and the absence of any stimulatory activity in the interphase provides definite identity for the activating ligand being an alkali-resistant glycolipid (which fractionates into the organic/aqueous phase in the partitioning).

α-GlcUCer is the NKT cell activating ligand

GSL from some other species of Sphingomonas may vary not only in the length of the carbohydrate chain, but also in the sugar sequence 9. Given that both GSL-4A and α-GlcUCer have the structural features necessary to activate NKT cells, it is necessary to use purified glycolipids to test their individual stimulation capacity. We purified α-GlcUCer and GSL-4A to homogeneity by solid-phase C18 chromatography followed by preparative TLC and used these lipids individually to assess their CD1d-specific stimulation capacity. Being similar to α-GalCer in structure, α-GlcUCer was able to stimulate NKT cells 4-fold better than GSL-4A alone was (Fig. 5), suggesting that α-GlcUCer is the major stimulating ligand in the S.paucimobilis GSL mixture. Notably, the stimulation capacity of GSL-4A was almost comparable to α-GlcUCer when the cells were not fixed after the lipid pulse (Fig. 5), suggesting that the tetraglycosylated lipid may be processed for its efficient presentation to NKT cells in metabolically active cells.

Figure 5.

α-Glucuronosylceramide is a potent NKT cell stimulant as compared with GSL-4A. Individual lipids purified by preparative TLC were used at equimolar concentrations from a 500-μg alkali-resistant lipid mixture. LMTK-CD1d1WT cells were pulsed with vehicle, the alkali-stable GSL mixture, or purified α-GlcUCer or GSL-4A for 4 h. The cells were extensively washed and fixed in 0.05% paraformaldehyde before coculture with the indicated NKT cell hybridomas for 20 h at 37°C. IL-2 secretion was measured as above. The experiment was repeated three times.

Although it was intriguing to note the increase in antigen-presentation capacity of pure α-GlcUCer when the APC were unfixed (Fig. 5), it could be accounted for by the fact that unfixed, metabolically active cells may have relatively more cell surface CD1d and the molecules are unaffected by paraformaldehyde fixation. This fixation could possibly affect the protein conformation, leading to a decreased recognition by the TCR in fixed cells. Compared with GSL-4A, the antigen-presentation capacity of α-GlcUCer was 4-fold better in stimulating N38-2C12 cells (Fig. 5). Thus, the fact that α-GlcUCer (but not GSL-4A) could be efficiently presented when the cells were fixed leads us to conclude that α-GlcUCer (unlike GSL-4A) does not require extensive intracellular processing before presentation.

During the preparation of this manuscript, Wu et al. 25 reported that glycolipids from a related species (S. wittichii) can activate human and mouse NKT cells. Although Wu et al. 25 had used synthetic α-GlcUCer to show that it is a CD1d-restricted antigen for NKT cells, this is in support of our results using natural α-GlcUCer extracted from S. paucimobilis. GSL-4A by itself did not stimulate NKT cell cytokine production any more than vehicle treatment and therefore served as a proper control for the bacterial-derived α-GlcUCer. We also used β-GalCer at a 500 μg/ml concentration in these assays as a negative control. Differing only by the anomeric attachment of the sugar compared to α-GalCer, as expected, β-GalCer was not presented to NKT cells 14, 26, 27 (data not shown).

CD1d-restricted identification of mouse NKT cells using S. paucimobilis GSL

Recombinant CD1d–Ig dimers loaded with the alkali-resistant S. paucimobilis GSL stained all NKT cell hybridomas tested but not the OVA-specific CD4+ T cell hybridoma, DO11.10 (Fig. 6A and data not shown). The dose-dependent staining of NKT cells but not mainstream T cells demonstrates the CD1d-restricted T cell specificity of this lipid staining. MHC class I dimers loaded with this lipid did not show any significant staining at the highest lipid concentrations used as compared to empty MHC class I dimer staining (data not shown). In addition, there was no change in the TCR levels of the hybridomas stained with lipid-loaded dimers as assessed by flow cytometry (data not shown).

Figure 6.

CD1d-restricted NKT cells can be stained with S. paucimobilis GSL-loaded CD1d1–Ig dimers. (A) NKT and mainstream T cell hybridomas (5×105) were stained with either empty or lipid-loaded CD1d–Ig dimers for 3 h at 4°C. The dimers were then detected using a biotinylated anti-mouse-IgG1 and streptavidin–allophycocyanin and analyzed by FACS. (B) Liver mononuclear cells from C57BL/6, CD1KO, and Jα18KO mice were blocked with 2.4G2 and stained with 4 μg α-GlcUCer-loaded CD1d–Ig dimers. The plots shown were gated for live lymphocytes and are representative of at least two individual experiments.

About 20–30% of liver T cells have been reported to be of the NKT cell phenotype 28, 29. This liver NKT cell population was recently confirmed to be invariant-TCR+ by staining with CD1d dimers or tetramers loaded with α-GalCer 30, 31. We detected α-GlcUCer-reactive T cells in WT mouse liver, but observed only background staining with CD1KO (which lack both invariant and non-classical CD1d-dependent NKT cells) and Jα18KO (which lack only invariant NKT cells) mice (Fig. 6B). Even though only ∼50% of α-GalCer-specific invariant NKT cells were detected by using α-GlcUCer, this proportion is significantly higher than that reported using PIM 6, and could indicate a subpopulation of NKT cells 31. Unlike α-GalCer-loaded tetramers 30, 32, 33, the lipid-loaded CD1d–Ig dimers exhibit a Vβ-dependent staining of invariant NKT cells 31. Therefore, the subset of NKT cells detected by α-GlcUCer-loaded dimers have TCR with α- or β-chains that are not affected by the negatively charged carboxylic acid group of α-GlcUCer that distinguishes it from α-GalCer. The liver mononuclear cells detected by α-GlcUCer-loaded CD1d dimers were not present in CD1KO mice or Jα18KO mice, demonstrating the usefulness of this reagent in detecting NKT cells.


NKT cells with a diverse TCRβ repertoire are able to effectively recognize α-GalCer presented by CD1d as long as they express the invariant TCR α-chain rearrangement Vα14Jα18 14, 32, 34, 35. It has also been suggested that these canonical NKT cells are heterogeneous in their reactivity to cellular antigens because of their promiscuous recognition capacity of multiple ligands 34, 35. Along these lines, our data confirm the ability of these canonical NKT cells to respond strongly to S. paucimobilis GSL as compared with a minimal or lack of an effect on the stimulation of non-canonical, CD1d-restricted NKT cells.

The structural features of α-GalCer essential for efficient binding to CD1d and subsequent NKT cell stimulation have been the topic of interest in several studies 4, 26, 3638. On the basis of these reports, it is clear that lipid chain length, although possibly contributing to CD1d binding, is not critical for efficient NKT cell stimulation. Further, the hydroxyl position at C4 in the sphingosine base is also dispensable 26. The carbohydrate anomericity has been found to be the most crucial determinant for effective NKT cell stimulation and α-GlcUCer satisfies this requirement. In relation to the substitutions in the carbohydrate, the equatorial orientation of the 2-hydroxyl is the only other requirement (α-mannosylceramide with axial orientation does not stimulate NKT cells) for being a CD1d-specific ligand 4. Tagging bulky groups like biotin or an amino functional group on the C6 position of the galactose in α-GalCer has been shown to be inconsequential to the NKT cell stimulation capacity of the glycolipid 39 and therefore the carboxyl group at C6 position in α-GlcUCer should not theoretically affect its ability to stimulate NKT cells. Our results support this observation. The fact that NKT cell stimulation is not dependent upon the anomericity of the outermost sugar residues, as long as the inner core sugar attachment to ceramide is in an α-orientation, makes GSL-4A a candidate ligand for CD1d that may require further processing before its presentation 4, 16.

The ability of S. paucimobilis GSL to stimulate both human and mouse NKT cells is a promising indicator for the possible use of this GSL in the identification, enumeration, and functional characterization of CD1d-dependent NKT cells. Being a natural ligand, the bacterial α-GlcUCer is an ideal tool to study the characteristics of NKT cell subsets that recognize this glycolipid compared with the total invariant NKT cell pool recognized by α-GalCer. It will be interesting to define how modifications in the fatty-acyl chain and sphingosine base of α-GlcUCer might affect CD1d binding and NKT cell function 38 by the chemical synthesis of these variants, similar to that described recently for other Sphingomonas species 25.

Although not widely studied as a pathogenic bacterium, S. paucimobilis has been associated with several water-borne hospital infections that can cause severe and recurrent bacteremia in patients with debilitating conditions such as hematological malignancies 40, 41. In light of our studies, the role of NKT cells in septicemia caused by bacterial infections and the causative or protective role of bacterial cell wall glycolipids needs to be evaluated. The pathophysiological function of CD1d-restricted NKT cells in microbial immunity and the involvement of non-protein ligands in the process have been recently reviewed 1, 5. Several species of bacteria contain unusual glycolipids capable of binding to CD1d.

During the review of this manuscript, two groups have published their observations on the ability of related bacterial glycolipids to stimulate NKT cells 42, 43. Whereas the studies of Kinjo et al. 42 lend support to our results, they used synthetic glycolipids as compared with the natural α-GlcUCer extracted from S. paucimobilis used in this report. Mattner et al. 43 have directly demonstrated the physiological relevance of our studies by showing that the bacterial glycolipids from heat-killed S. capsulata can be presented by bone-marrow-derived DC and splenocytes to NKT cells in a CD1d-dependent manner. It is our hypothesis that similar to pattern-recognition molecules responsible for diverse microbial product recognition, CD1d-mediated presentation of bacterial glycolipids can specifically "fine tune" NKT cells in an immunoregulatory fashion based on ligand structure and binding strength as compared with endogenous ligands.

Materials and methods


WT C57BL/6 mice were purchased from the Jackson Laboratory (Bar Harbor, ME, USA). CD1KO and Jα18KO mice on the C57BL/6 mice were kindly provided by L. Van Kaer (Vanderbilt University, Nashville, TN, USA); Jα18KO were obtained by permission of M. Taniguchi (Chiba University). Mice were age- and sex-matched and all experiments were approved by the Indiana University School of Medicine's Animal Care and Use Committee.

Cell lines and other reagents

Murine LMTK fibroblasts transfected with either the empty vector pCDNA3.1 (–) (LMTK-vector) or vector containing cd1d1 cDNA (LMTK-CD1d1WT) were cultured and used as described previously 44. The CD1d1-specific murine NKT cell hybridomas DN32.D3 (Vα14+) 45, N37-1A12 (Vα5+), and the α-GalCer-dependent N38-2C12 (Vα14+), N38-3C3 (Vα14+) NKT cell hybridomas have been described 14, 27. The murine T cell hybridoma DO11.10 specific for ovalbumin was kindly provided by Dr J. Blum. All NKT and T cell hybridomas were cultured in Iscove's modified Dulbecco's medium supplemented with 5% FBS (HyClone, Logan, UT, USA) and 2 mM L-glutamine (assay medium; Cambrex, Walkersville, MD, USA). The human B lymphoblastoid cell line C1R and the human CD1d transfectant C1R-CD1d (kindly provided by Dr S Balk) were grown in RPMI-1640 medium containing 2 mM L-glutamine and 5% bovine growth serum, with C1R-CD1d also cultured in the presence of 0.5 mg/ml G418 [Mediatech (Cellgro), Herndon, VA, USA]. A human NKT cell line (kindly provided by Dr M. Exley) was propagated and maintained as described previously 15.

Capture and detection antibodies used in the ELISA for the quantitative analysis of mouse IL-2, GM-CSF and IL-13 were purchased from Pharmingen. Recombinant cytokines used as standards in the ELISA were obtained from PeproTech (Rocky Hill, NJ, USA). The α-GalCer and 4-deoxy α-GalCer were synthesized as will be described elsewhere (Du et al., submitted).

Bacterial lipid extraction and purification

S. paucimobilis (originally Flavobacterium devorans, ATCC 10829) grown in 2 liters of nutrient agar (Becton Dickinson, Franklin Lakes, NJ, USA) at 30°C for 24–48 h was collected by centrifugation, washed in PBS and lyophilized. The lyophilized bacterial pellet was homogenized with methanol before adding chloroform to obtain a solvent ratio of chloroform:methanol of 2:1 by volume. Extraction was continued by brief sonication (20 s × 3) followed by refluxing at 80°C for 1 h. The lipid extract was separated from the non-lipid residue by centrifugation. The non-lipid residue was re-extracted with chloroform:methanol (1:1, vol:vol) as above and the lipid extracts were combined, dried by flash evaporation under vacuum, and the chloroform:methanol (2:1, vol:vol)-soluble fraction was used as a crude lipid extract. The crude lipid extract was reconstituted in warm methanol containing 0.5 N KOH and saponified at 37°C for 4 h. The alkali-stable lipids were purified after neutralization with glacial acetic acid using C18 reverse-phase cartridges (Accubond, J&W Scientific, Folsom, CA, USA) 46. The α-GlcUCer and GSL-4A were purified to homogeneity using preparative TLC plates (EMerck, Germany). The alkali-resistant lipid mixture (500 μg) was applied to the plate in a narrow concentrating zone, and chromatographed in chloroform:methanol:acetic-acid:water (100:20:12:5, by volume) as the solvent system for 20 cm height, and air-dried to remove solvents. A parallel lane was cut from the plate and GSL were detected using orcinol spray 47. Silica gel zones identified as α-GlcUCer and GSL-4A on the basis of mobility and parallel orcinol detection were scraped and eluted with chloroform:methanol (2:1 by volume). The α-GlcUCer and GSL-4A derived from this 500 μg lipid mixture were compared side by side with 500 μg of alkali-resistant lipid equivalents in a functional assay.

NKT cell stimulation assay

The lipids were dried and sonicated in normal saline containing 0.05% Tween-20 (vehicle) as a stock solution and then diluted in assay medium with sonication before use. LMTK-vector or LMTK-CD1d1WT cells were treated with either vehicle or lipids for 4 h at 37°C, and washed extensively before culture with NKT cell hybridomas 44. IL-2 release after 20 h of co-culture was measured by ELISA as described previously 17. Where necessary, the target cells were also treated with the indicated concentrations of lysosomotropic agents (chloroquine and bafilomycin A1; Sigma-Aldrich, St. Louis, MO, USA), or an inhibitor of recycling (primaquine; Sigma-Aldrich), but the total lipid pulse time was maintained at 4 h. In some assays, the lipid-pulsed and drug-treated cells were fixed after the 4 h treatment period in 0.05% paraformaldehyde before being cocultured with NKT cells 17.

CD1d–Ig dimer staining of NKT cells

Recombinant soluble mouse dimeric CD1d–Ig (Pharmingen) was loaded with different molar excesses of purified GSL from S.paucimobilis according to the manufacturer's instructions. In brief, lipids resuspended in PBS (pH 7.4) were mixed with the dimers and incubated at 37°C overnight. Mouse liver mononuclear cells, isolated as described previously 48, were treated with 2.4G2 hybridoma supernatant before staining for cytofluorography with empty or lipid-loaded CD1d dimers, FITC-conjugated anti-mouse-TCRβ and biotinylated anti-NK1.1 (Pharmingen). The dimers were directly labeled with phycoerythrin using a mouse IgG1-specific Zenon labeling kit following the instructions of the manufacturer (Molecular Probes, Eugene, OR, USA). Biotinylated anti-NK1.1 was detected using streptavidin–allophycocyanin (Pharmingen). NKT and T cell hybridomas were also stained with lipid-loaded or empty dimers followed by detection with a PE-conjugated anti-mouse-IgG1 monoclonal antibody (Pharmingen) 31.


This work was supported by grants from the National Institutes of Health to R. R. B. (R01 AI46455 and CA89026) and funds from the Walther Cancer Institute and NSF CHE-0194682 from the National Science Foundation to J.G.H. The authors would like to express their gratitude for the kind donation of the lyophilizer by the late Dr Raoul Rosenthal, cell lines from Drs J. Yewdell, J. Bennink, S. Balk, and M. Exley, and CD1KO, and Jα18KO mice from L. Van Kaer and M. Taniguchi, respectively. V. S. is thankful to Sungyoo Cho for propagating the human NKT cells, and members of the Brutkiewicz laboratory for their helpful comments and suggestions. R. R. B. is a Scholar of the Leukemia and Lymphoma Society.


  1. 1


  2. 2


  3. 3


  4. 4


  5. 5


  6. 6