Identification of novel glycolipid ligands activating a sulfatide-reactive, CD1d-restricted, type II natural killer T lymphocyte



Sulfatide-reactive CD1d-restricted natural killer T (NKT) lymphocytes belong to the type II NKT cell subset with diverse TCRs, and have been found to regulate experimental auto-immune encephalomyelitis, tumor immunity, and experimental hepatitis in murine models. NKT cells can be activated by self-lipids presented by CD1d, manifested as autoreactivity. The identity of most of these self-lipids remains unknown. By isolating lipids from a CD1d-expressing, highly stimulatory antigen presenting cell, we identified isoforms of β-glucosylceramide (GlcCer), with sphingosine and fatty acid chain lengths of C24:0 and C16:0, that activated a sulfatide-reactive type II NKT cell hybridoma. A screen of structurally related glycosphingolipids demonstrated β-galactosylceramide (GalCer) as another ligand, and further, that the lysoforms were the most potent isoform of the glycosphingo-lipid ligands, followed by isoforms with a long fatty acid chain of C24. Thus, the same type II NKT cell was activated by several ligands, namely sulfatide, GlcCer, and GalCer. However, CD1d-dependent reactivity to antigen presenting cells lacking all GlcCer-based glycosphingolipids, or all glycosphingolipids, was maintained. This suggests that other endogenous, nonglycosphingolipid, lipid ligands contribute to steady-state autoreactivity by type II NKT cells.


Natural killer T (NKT) lymphocytes are activated by lipids and glycolipids presented by the essentially nonpolymorphic

antigen presenting molecule CD1d, expressed predominantly by hematopoietic cells. Upon activation, NKT cells display innate-like features such as rapid production of large amounts of both inflammatory and anti-inflammatory cytokines, including IFN-γ and IL-4, and they enhance DC maturation. Through these features, they are important in the initiation of innate immunity and also in the direction and regulation of the adaptive immune response resulting in the modulation of tumor immunity, autoimmunity, and microbial infections [1-3].

NKT cells can be activated both by microbial and endogenous lipids. The latter may be upregulated in antigen presenting cells (APCs) upon their activation, or be derived from other cells of the body and presented by APCs [4-6]. Sulfatide was among the first self-glycosphingolipids demonstrated to stimulate T lymphocytes [7]. Sulfatide is found in large amounts in the central nervous system, mainly located in the myelin, a target of the autoimmune process during multiple sclerosis. Indeed, the frequencies of sulfatide-reactive T cells were increased in multiple sclerosis patients [7], and sulfatide-reactive CD1d-restricted NKT cells accumulated in the CNS of mice during experimental autoimmune encephalomyelitis [8], a model for the human disease, suggesting that the cells were engaged in the disease process. This notion was supported by studies showing that sulfatide injection protected CD1d-sufficient mice from the disease, consistent with the hypothesis that that sulfatide-reactive NKT cells could suppress the autoimmune process [8]. Subsequent studies in the mouse demonstrated that sulfatide-reactive NKT cells exhibit pivotal immunoregulatory properties in a variety of immune reactions [9-12]. Sulfatide-reactive NKT cells belong to the subset of type II NKT cells, with diverse TCRs. They are the best characterized NKT cell subset [3, 8, 13-16], next to the α-galactosyl ceramide (αGalCer) type I NKT cells expressing the invariant Vα14-Jα18 (mouse) or Vα24-Jα18 (human) TCR α-chain [1]. The size of the entire type II NKT cell subset is not possible to establish with high accuracy using current available methodologies. However, published data suggest that type II NKT cells are relatively abundant in humans, where type I NKT cell frequencies are highly variable but tend to be lower than in mice [3]. Previous studies suggest that murine sulfatide reactive type II NKT cells, like type I NKT cells, are preferentially located in the liver where they make up around 5% of lymphocytes, while their frequencies are 20-fold lower in spleen and thymus [8].

Despite the identification of several lipid ligands that can activate NKT cells, an unresolved conundrum of NKT cell biology is the identity of the lipid ligands presented on CD1d that drive the selection of NKT cells during their thymic development. The same, or different, endogenous ligands may also be expressed by APCs and contribute to activation of NKT cells during microbial infections, and regulation of tumor responses and autoimmunity [4, 5]. Revealing the regulation of the synthesis and structure of CD1d-restricted NKT self-lipids is central for our understanding of how these cells function to modulate immunity. The sulfatide isoforms that activate type II NKT cells are well characterized [8, 14, 17]. Still, sulfatide-reactive NKT cells remain in mice lacking sulfatide [8], and the sulfatide-reactive type II NKT cell hybridoma XV19 is autoreactive to CD1d-expressing APCs from sulfatide-deficient mice [14]. Thus, other CD1d-restricted ligands are involved in the thymic selection and peripheral activation of sulfatide-reactive type II NKT cells. To identify these ligands, here we have performed a screen for cellular glycosphingolipids that activate sulfatide-reactive type II NKT cells in a CD1d-dependent manner, and investigated the role of the identified glycosphingolipids in the CD1d-dependent autoreactivity of these NKT cells.


Glycosphingolipid fractionation of A20CD1d cells identified β-GlcCer as a possible NKT cell ligand

Like other NKT cell hybridomas, the sulfatide reactive type II NKT cell hybridoma XV19 demonstrates variable levels of CD1d-dependent autoreactivity to different cell lines [14, 18, 19] (Fig. 1). It is presumed that autoreactivity requires the presentation of endogenous lipid ligands by CD1d. The diverse stimulatory ability of the APCs could reflect the presentation of distinct endogenous ligands, but is likely to also be influenced by different CD1d levels and the expression of costimulatory and coinhibitory molecules on the APCs (Fig. 1 and Supporting Information Fig. 1). Regardless, the strong reactivity of XV19 cells toward A20 cells transfected with tail-deleted CD1d (here called A20CD1d cells) indicates the presentation of potent self-lipid ligands on CD1d by these cells. It further confirms that CD1d auroreactivity of type II NKT cells, including XV19, is not dependent on recycling of CD1d to late endosomes [18], suggesting that the endogenous ligands underlying natural autoreactivity of type II NKT cells are loaded in the ER–Golgi pathway.

Figure 1.

XV19 cells demonstrate high autoreactivity to A20CD1d cells but not to JAWSII cells. (A) Overlay showing CD1d expression of A20 (white peak), A20CD1d (black peak), and JawsII cells (gray peak). (B) Autoreactivity of XV19 cells to A20, A20CD1d (left), and JawsII cells (right). Activation was measured as IL-2 secretion determined in a CTLL-2 bioassay. Data are shown as mean and SD of triplicate cultures. Data are representative of more than three experiments.

In an attempt to identify endogenous ligands presented by A20CD1d cells, we isolated and fractionated the total cellular glycosphingolipids (Fig. 2A), and tested the fractions obtained for recognition by XV19 cells (Fig. 2B–D). XV19 cells reproducibly showed activity toward the neutral but not the acidic lipid fraction (Fig. 2B). Further fractionation of the neutral lipids was performed according to hydrophobicity. Two neutral lipid fractions (eluates II and III) stimulated XV19 cells (Fig. 2C), while subfractions of acidic lipids were not stimulatory (data not shown). As a final step, both active neutral eluates were separated by high-performance thin-layer chromatography (HPTLC). The entire HPTLC lane of each eluate was divided into sections, the solid matrix scraped off, and lipids were extracted and tested for activation of XV19 cells. One fraction derived from eluate II reproducibly activated XV19 cells in a dose-dependent manner (Fig. 2D). This fraction ran in the HPTLC at a position identical to that of GlcCer (Glcβ1–1′Cer) and GalCer (Galβ1–1′Cer). Analysis on a borate-impregnated HPTLC-orcinol plate, where GlcCer and GalCer are separated, demonstrated the presence of GlcCer but not GalCer (data not shown). Further, electrospray mass spectometry (ESI-MS) of the active fraction revealed two major peaks determined to be monoglycosylceramide with a saturated long (C24:0) and short (C16:0) fatty acid chain (Fig. 2E). Taken together, analysis of the active glyco-sphingolipid fraction identified the two major components as GlcCer with C24:0 and C16:0 fatty acid chains.

Figure 2.

A specific fraction of the neutral lipid extract derived from A20CD1d cells stimulates XV19 cells. (A) Scheme of the fractionation process of A20CD1d cells (described in the Materials and Methods). Step 1: total lipids from A20CD1d cells were fractionated into uncharged neutral (Neu) and charged acidic (Acidic) lipids. Step 2: neutral fractions were further separated into neutral lipid eluate II and eluate III. Step 3: glycolipids in eluates II and III were further separated by HPTLC. Fractions of lipids were extracted from the HPTLC plate for testing in the T hybridoma assay. (B–D) JawsII cells (5 × 103 cells/well) were pulsed with (B) neutral or acidic lipids, or (C) neutral eluates II or II, or (D) eluted lipid fractions from the HPTLC separation (for clarity fraction 1 and DMSO vehicle control are plotted separately in the lower graph), or (B–D) vehicle (DMSO), for 4–5 h before addition of XV19 NKT hybridoma cells (40 × 103 cells/well). Activation was measured as IL-2 secretion determined in a CTLL-2 bioassay. The data show the average and SD of lipid fractions prepared in parallel triplicates (each tested in triplicate cultures) and are representative of at least three independent glycolipid fractionation experiments. (E) ESI-MS histogram of the active fraction 1. The two major peaks in the ESI-MS histogram (circled) were determined as monoglucosylceramide with a saturated long-chain (C24:0) and short-chain (C16:0) fatty acid composition. Data show one of the two experiments performed with similar results.

Semisynthetic GlcCer isoforms activated XV19 cells in a CD1d-dependent manner

To verify that the identified glycolipids from A20CD1d cells were stimulatory, the GlcCer isoforms with sphingosine and fatty acid chains of C16:0 and C24:0 were produced semisynthetically. Using JawsII APCs, saturated GlcCer with a long fatty acid chain, C24:0 was superior in activating XV19 cells compared with the shorter fatty acid chain isoform, C16:0, while C24:1 and native GlcCer only marginally stimulated XV19 cells (Fig. 3). In contrast, GM1 as well as sphingomyelin (Fig. 3A and data not shown) failed to induce activation of XV19 cells. The activation of XV19 cells by GlcCer isoforms was dependent on CD1d (Fig. 3B and C). These data demonstrate that primary APCs and JawsII cells can present GlcCer isoforms to XV19 NKT hybridoma cells in a CD1d-dependent manner.

Figure 3.

Activation of XV19 cells by semisynthetic isoforms of GlcCer. (A) XV19 cells were stimulated with JawsII cells prepulsed with semisynthetic C24:0 and C16:0 GlcCer, GM1, or vehicle (DMSO). (B) JawsII cells were pulsed with 100 nmol/mL C24:0 or vehicle (DMSO), then 15 μg/mL anti-CD1d mAb 20H2 for 15 min, prior to addition of XV19 cells. (C) XV19 cells were stimulated with BMDCs derived from CD1d+/− or CD1d−/− mice prepulsed with 100 nmol/mL lyso-, C24:0, C16:0, C24:1, and native GlcCer or vehicle. Activation was measured as IL-2 secretion determined in a CTLL-2 bioassay, and all data are shown as mean and SD of triplicate cultures and are representative of at least two experiments.

The lysoforms of sulfatide, GalCer, and GlcCer were superior in activating XV19 cells

The lysoform (lacking the fatty acyl chain) of sulfatide potently activates XV19 cells ([14, 17], and Fig. 4A). In the same manner, lyso-GlcCer was a superior ligand compared with the other tested isoforms of GlcCer (Fig. 2C) suggesting that lyso-glycosphingolipids may provide an optimal configuration together with CD1d for stimulation of XV19 cells. To explore this further, additional lysoforms of glycosphingolipids were tested for their ability to activate XV19 cells. In addition to lysosulfatide and lyso-GlcCer, we found that lyso-GalCer activated XV19 cells to a similar level as lyso-GlcCer, while lyso-LacCer, lyso-Gb3, lyso-GM1, and lyso-sphingomyelin all failed to activate XV19 cells (Fig. 4A). This demonstrated that, in this assay, the high level of activation of XV19 cells was specific for the lysoforms of sulfatide, GalCer, and GlcCer and further that sulfated GalCer (i.e. sulfatide) was superior to nonsulfated GalCer and GlcCer (for structures, see Fig. 4F). We investigated whether other type II NKT cell hybridomas shared reactivity to lyso-GlcCer and lyso-GalCer with XV19 cells. The sulfatide-reactive 14S.15.5D type II NKT cell hybridoma demonstrated a reproducible stimulation by lyso-GalCer (Supporting Information Fig. 2), while other type II hybridomas tested so far did not respond to these ligands. The activation by the lysoforms could be blocked by anti-CD1d mAb demonstrating a CD1d-dependent activation (Figs. 3C and 4B, and data not shown).

Figure 4.

Lysoforms of stimulatory glycosphingolipids are superior in activating XV19 cells.XV19 cells were stimulated with JawsII cells prepulsed with (A, B, D, E) 30 nmol/mL or (C) a titration of the indicated glycosphingolipids or vehicle (DMSO). (B, D) The anti-CD1d mAb 20H2 was added as indicated 15 min before XV19 cells. Activation was measured as IL-2 secretion determined in a CTLL-2 bioassay, data are shown as mean and SD of triplicate cultures, and are representative of three experiments. (F) Structures of sulfatide, GlcCer, and GalCer.

Identification of an additional glycosphingolipid, GalCer, activating XV19 cells

We next tested whether other isoforms of GalCer also stimulated XV19 cells, similar to sulfatide and GlcCer isoforms. Native GalCer as well as the semisynthetic C24:1 isoform with sphingosine base stimulated XV19 cells in a dose- and CD1d-dependent manner, while C16 and C24:0 were not stimulatory (Fig. 4C–E).

An APC free asssay system verified reactivity to stimulatory glycosphingolipid isoforms

To demonstrate that the identified glycosphingolipid ligands, when loaded on CD1d, were directly stimulating the XV19 TCR, we investigated the activation of XV19 cells by the ligands when presented on plate-bound CD1d. We could demonstrate that C24:1 sulfatide, C24:1 GalCer, and C24:0 GlcCer efficiently stimulated XV19 cells in this assay (Fig. 5A). In the CD1d–plate-bound assay, the three glycosphingolipids displayed similar level of stimulation of XV19 cells in three experiments performed. The lysoforms potently activated XV19 cells also in this assay, similar to the level achieved with anti-CD3 stimulation (Fig. 5A); here, the lysoforms induced around twofold higher IL-2 secretion compared with the other isoforms tested (Fig. 5A compared with Fig. 5B). Thus, when loaded on plate-bound CD1d, C24 isoforms, and lysoforms of sulfatide, GlcCer and GalCer could directly stimulate the XV19 TCR resulting in activation and IL-2 secretion.

Figure 5.

Glycosphingolipids activate XV19 cells in an APC-free assay system. CD1d-coated culture wells were incubated with (A) the indicated glycosphingolipid C24-based isoforms or (B) lysoforms or vehicle (Empty) overnight, followed by three washes. A total of 40 × 103 XV19 cells were added per well and cultured over night, and IL-2 in supernatants was measured by ELISA. XV19 + SA – indicates control without CD1d. Data are shown as mean and SD of triplicate cultures and are representative of three experiments.

The autoreactivity of XV19 cells was not dependent on glycosphingolipids

In our search for APC-derived endogenous ligand/s responsible for the CD1d-dependent autoreactivity of XV19 cells, we have identified two additional glycosphingolipids that can stimulate the cells, GlcCer and GalCer. Thus, three structurally similar endogenous glycosphingolipids, including sulfatide (Fig. 4D and F), have been demonstrated to activate XV19 cells, suggesting that glycosphingolipids may play a role in the autoreactivity of XV19 cells. Sulfatide was previously shown not to be required for the auto-reactivity of XV19 cells to spleen cells [14]. To investigate the role of GlcCer for the autoreactivity, we used a CD1d-transduced B16 melanoma-derived cell line deficient in ceramide glucosyltransferase (GM95mCD1d), and thereby deficient in GlcCer [20]. The autoreactiviy of XV19 cells to CD1d+ GM95mCD1d cells was not altered compared with that of the CD1d+ restored control cells expressing ceramide glucosyltransferase (GluCerTmCD1d) [20], despite the lower CD1d expression on the mutant GM95mCD1d cells (Supporting Information Fig. 3), indicating that the autoreactivity of XV19 hybridoma cells was independent of GlcCer (Fig. 6A). Consistent with this, the selective ceramide glucosyltransferase inhibitor NB-DGJ did not reduce the XV19 stimulatory capacity of A20CD1d cells (data not shown).

Figure 6.

Glycosphingolipids are not required for the autoreactivity of the type II NKT cell hybridoma XV19. Autoreactivity of XV19 cells to (A) a titration of CD1d-transduced GluCerTmCD1d or GM95mCD1d cells, or (B) CD1d-transfected LY-B and LY-B/cLCB1 cells (left), or to 40 × 103 CD1d-transfected or nontransfected LY-B and LY-B/cLCB1 cells per well (right). Activation was measured as IL-2 secretion determined in a CTLL-2 bioassay, data are shown as mean and SD of triplicate cultures, and are representative of two experiments.

Thus, neither sulfatide nor GlcCer alone were required for the autoreactivity of XV19 cells. It remained a possibility that both glycosphingolipids could serve as ligands, and in the lack of one the other was sufficient to induce activation. In addition, GalCer was also a candidate ligand for the autoreactivity. We, therefore, investigated whether the autoreactivity of XV19 cells required glycosphingolipids of any kind by using LY-B cells, mutant for sphingosine long-chain base subunit 1 (LCB1), a subunit of serine palmitoyltransferase, resulting in dramatically reduced levels of sphingomyelin and glycosphingolipids [20]. These were compared with control cells with rescued enzyme activity (LY-B/cLCB1 cells). Both lines had also been transfected to express CD1d [20] (Supporting Information Fig. 3). The autoreactivity induced by the glycosphingolipid-deficient LY-B/CD1d cells and the control LY-B/cLCB1/CD1d cells was comparable (Fig. 6B). This suggested the involvement of other lipids than glycosphingolipids in the natural autoreactivity of XV19 cells.


In this study, we aimed to identify novel endogenous ligands that stimulate the sulfatide reactive type II NKT cell hybridoma XV19, in search for the endogenous ligand/s responsible for the CD1d-dependent autoreactivity of these cells. To this end, we fractionated glycosphingolipids from a CD1d-expressing cell line (A20-CD1d) that induces high autoreactivity by XV19 cells. By further analysis of a positive glycolipid fraction, and the use of semisynthetic glycosphingolipid candidates, we identified GlcCer and GalCer as stimulatory ligands for the XV19 cells.

We used two different assay systems to demonstrate the activ-ation of XV19 cells by sulfatide, GlcCer, and GalCer. First, we employed APCs (the DC cell line JawsII and bone marrow-derived dendritic cells (BMDCs) that after pulsing with glycosphingolipids efficiently activate XV19 cells. Using the APC-based assay, we demonstrate that GlcCer C24:0 and GalCer C24:1 were both able to stimulate XV19 cells, although not to the same extent as the previously identified ligand sulfatide C24:1. With all three glycoshingolipids, the lysoforms were superior ligands compared with the other isoforms tested. In APC-based assays for stimulation of NKT cells with exogenous lipid ligands, activation of NKT cells will result from the TCR affinity to the specific CD1d-ligand complex, as well as the cell surface density of CD1d loaded with the specific ligand. The latter will in turn depend on lipid solubility, uptake, intracellular localization, and loading efficiency on CD1d, which may differ substantially between different glyco-sphingolipids and between isoforms. In addition, it was possible that lipid pulsing of APCs could result in APC activation and increased presentation of endogenous stimulatory lipids on CD1d. Therefore, to demonstrate a direct activation of XV19 cells with defined CD1d-ligand complexes, we also used plate-bound CD1d loaded with different ligands to stimulate hybridoma cells. Using this assay, we found that sulfatide C24:1, GlcCer C24:0, and GalCer C24:1 stimulated the XV19 cells to a similar level. The respective lysoforms more potently activated XV19 cells compared with the other isoforms also in this assay, but the difference was only twofold. Further, lysosulfatide, lyso-GlcCer, and lyso-GalCer all activated the hybridoma to the same level as anti-CD3. We, therefore, concluded that the three glycosphingolipids and their lysoforms loaded on CD1d could all directly stimulate the XV19 TCR. The difference in results seen between the two assays may be a consequence of differential uptake and intracellular loading of the different ligands, resulting in more divergent stimulation by the ligands in the APC assay.

The crystal structure of CD1d loaded with sulfatide suggests that the β-linked sulfated galactose head group would point upwards and be exposed toward the TCR [13]. The three ligands recognized by the XV19 TCR, GlcCer, GalCer, and sulfatide, are structurally similar; sulfatide differs from GalCer in the sulfate group added to the 3′ position of the galactose head group of GalCer, while GalCer and GlcCer only differ from each other in the planar positioning of an OH-group of the hexose head group (Fig. 3F). It therefore seems likely that the same TCR could accommodate the three ligands when bound to CD1d. The invariant type I NKT cell TCR binds to CD1d with relatively high affinity in an interaction where germline-encoded residues in the complementarity determining regions of the TCR-α and β-chains make major contributions, referred to as an innate-like pattern recognition [21]. This is thought to allow the recognition of structurally very different CD1d-presented ligands that are in some cases forced to adopt a shared flattened position between the TCR and CD1d [22, 23]. We show here that type II NKT TCR can demonstrate a relatively broad reactivity and recognize different lipid ligands presented on CD1d, although the array of nonstimulatory glyco-sphingolipids, or lyso-compounds indicated that there is a limit to the flexibility of ligand recognition by the XV19 TCR. Defining nonglycosphingolipid endogenous ligand/s that underpin CD1d autoreactivity by XV19 cells will add further complexity to the picture. The physiological consequences of this cross-reactive nature of NKT cell TCR are not known. Resolution of the trimolecular interaction of type II NKT TCR with CD1d-ligand will settle whether these TCR share with type I TCR an interaction with CD1d determined by germline-encoded elements, or if type II TCR, being diverse, will have more unique nongermline-encoded interactions with CD1d-ligand reminiscent of conventional MHC-restricted TCR.

In the APC assay, the preferred acyl chains of the GalCer-based glycosphingolipids (GalCer and sulfatide) was C24:1, while C24:0 was the most stimulatory GlcCer isoform. Both acyl chain length and saturation clearly play a role for activity in the APC assay, and hydroxylation of the sulfatide acyl chain is not tolerated [14]. Unsaturation of the acyl chain may tilt the head group altering the surface exposed toward the TCR [24]. The ligand glycosphingolipid lysoforms were superior in stimulating XV19 cells in both the APC-based and plate-bound CD1d assays, and lysosulfatide was also a preferred isoform for another sulfatide-specific type II hybridoma [14]. It is not clear how the lysoforms would bind in CD1d. Although with ceramide lipids the sphingoid base is bound in the F′ pocket, in the absence of the fatty acid, the long chain base potentially could anchor in either of the A’ or F’ pockets. The lack of the fatty acid may also influence the positioning and/or flexibility of the head group, contributing to a superior activation of the XV19 TCR. It is interesting to note that the recognition of GlcCer by type I NKT cells demonstrated a preference for GlcCer C24:1, while lyso-GlcCer was not stimulatory [25], a pattern clearly distinct from that of XV19 cells.

Having defined the novel ligands GlcCer and Galcer for a type II NKT cell hybridoma, the question arises whether this reactivity is common among type II NKT cells. While the type II NKT-cell population is defined as being diverse, data imply that sulfatide reactive type II NKT cells make up a substantial fraction of the type II population [8], suggesting that the type II NKT cell population may be oligoclonal. Stimulation of spleen cells from Jα18-deficient mice (lacking type I NKT cells) with GlcCer and GalCer, also after injecting BMDCs pulsed with the glycolipid ligands, has not revealed a glycolipid-specific response (data not shown). We could demonstrate reactivity to lyso-GalCer by the sulfatide-reactive 14S.15.5D type II NKT hybridoma, but other type II NKT cell hybridomas tested so far were not stimulated by GlcCer or GalCer (data not shown). Different TCR can be expressed by distinct sulfatide reactive type II NKT cells arguing against a shared canonical/invariant TCR by all sulfatide-reactive cells [14, 15], and thus, not all sulfatide reactive type II NKT cells may share TCR fine specificity and cross-reactivity to GlcCer and GalCer with XV19 cells. These initial results suggest that GlcCer/GalCer reactivity is present among type II NKT cells and not unique to XV19 cells, however, further studies are required to shed light on this issue.

CD1d-dependent reactivity was maintained to GlcCer- and glycosphingolipid-deficient cell lines, suggesting that the ligands responsible for the homeostatic/basal autoreactivity of XV19 cells, as recently suggested for iNKT cells [20], may not exclusively be glycosphingolipids. On the other hand, type I NKT cells were recently found to depend on GlcCer for their CD1d-dependent autoreactivity to BMDCs [25], emphasizing that distinct cell types are likely to differ significantly in the lipid antigens presented on CD1d. It should be noted that type I NKT cells, but not XV19 and other type II NKT cells, require recycling of CD1d to the lyso-somal compartments for their autoreactivity [18, 26], demonstrating that type I and II NKT-cell autoreactivity is likely induced by distinct ligands. In addition to glycosphingolipids, other classes of lipids such as phospholipids, diacylglycerol-based microbial ligands, and ether-bonded monoalkyl glycerophosphates and precursor and degradation products of plasmalogens are ligands for type I NKT cells [24, 27], underscoring the diversity of lipid ligands presented on CD1d. Moreover, cellular glycosphingolipid synthesis is highly dynamic [28]. The glycosphingolipid repertoire loaded on CD1d in TLR-stimulated APCs is altered through increased glyco-sphingolipid synthesis and reduced degradation [6, 25, 29, 30], leading to augmented stimulation of type I NKT cells. Further, increased levels of GlcCer, GalCer, or sulfatide, are associated with TLR stimulation, carcinogenesis, and other pathological states [28, 31]. Taken together, this supports a scenario where CD1d-associated lipids other than glycosphingolipids may provide a basal homeostatic autoreactivity of XV19 cells, while upregulation of sulfatide/GlcCer/GalCer may contribute to an increased level of activation of these cells in situations of tumor immunity, autoimmunity, and infection. A similar scenario has been proposed to apply to type I NKT cells as well, and the results from a recent study suggest that the autoantigens permissive for recognition and CD1d autoreactivity by the semi-invariant TCR could be diverse [32]. Further studies of how self-lipids control autoreactivity of NKT cells will be of great importance to increase our understanding for the regulation of NKT-cell activation, and their engagement in diverse immune reactions.

Materials and methods


C57BL/6 mice were from Taconic (Ry, Denmark). CD1d−/− mice on a C57BL/6 genetic background were bred in a specific pathogen-free animal facility at the University of Gothenburg. Female mice were used at the age of 8–20 weeks.

Cells and flow cytometry

The type II NKT hybridoma XV19, A20 cells expressing a tail-deleted murine CD1d (A20CD1d), and JawsII cells were cultured as described before [18, 33, 34]. BMDCs were obtained by culturing bone marrow cells with 20 ng/mL GM-CSF for 6 days, and matured with 1 μg/mL LPS for 24 h, before use as APCs. CD1d-transfected and nontransfected LY-B and LY-B/cLCB1 cells (see Supporting Information Fig. 2 for CD1d expression levels) were cultured in sphingolipid- and serum-free medium before hybridoma stimulation as described [20]. The CD1d-transfected GM95mCD1d and GluCerTmCD1d cell lines were cultured as described [20]. Cells were incubated with 2.4G2 Fc-block followed by PE-conjugated anti-mouse CD1d mAb (clone 1B1, BD Pharmingen), PE-CD80 (clone1610A1, BD Pharmingen), PE-CD86 (clone GL1, BD Pharmingen), biotinylated CD40 (homemade, clone FGK) followed by allophycocyanin-conjugated streptavidin (clone MAR18.5, BD Pharmingen), and PE-conjugated PDL-1 (clone M1H5, eBioscience) or control mAb and analyzed with a FACSCalibur using CellQuest software (BD Biosciences).

APC-based T-cell hybridoma assay

Before addition to APCs, A20-derived lipid fractions were dissolved in DMSO, heated 15 min at 37°C, sonicated 5 min, further diluted in culture medium preheated to 37°C, incubated 15 min at 37°C, and sonicated 5 min. Sulfatide isoforms were diluted in culture medium, sonicated 15 min, heated at 37°C 15 min, and diluted in preheated medium (37°C) before addition to APCs. Lyso-GM1 ganglioside (lyso-GM1) was dissolved in DMSO by incubation at 37°C for 15 min, and diluted in 37°C preheated medium before addition to APCs. Other glycosphingolipids were dissolved in DMSO, sonicated 15 min, and incubated at 80°C for 2 min in cycles until dissolved, then diluted in preheated (80°C) culture medium, sonicated 15 min, and heated at 80°C for 2 min before addition to APCs. A total of 5 × 103 JawsII cells or 5 × 103 BMDCs/well were pulsed with lipids 4–5 h at 37°C, washed in medium, and added to T hybridoma cells (40 × 103 cells/well). The anti-CD1d mAb 20H2 was kindly provided by Albert Bendelac, University of Chicago [14, 35]. IL-2 secretion was determined in culture supernatants in a bioassay using CTLL-2 cells [14].

Antigen presentation assay with plate-bound CD1d

The lipid presentation assay with plate-bound CD1d was performed essentially as described before [25, 36]. ELISA plates (high binding, Greiner) were coated with 10 μg/mL recombinant streptavidin (Invitrogen) over night at 4°C. After washing with PBS three times, 0.25 μg/well biotinylated mouse CD1d monomers (NIH Tetramer Facility) were added and incubated for 3 h at room temperature. Parallel wells were incubated with 5 μg/mL anti-CD3 (homemade, KT-3, used as positive control). Wells were washed three times with PBS. A total of 10 nmol/well of indicated isoforms of sulfatide, GalCer, and GlcCer were diluted as described above and incubated in CD1d-coated wells over night at 37°C, followed by two washes with PBS and one with culture medium. A total of 40 × 103 XV19 hybridoma cells were added per well and cultured over night. IL-2 in supernatants was analyzed with ELISA (R&D systems, mouse IL-2 DuoSet).

Glycosphingolipids and other lipids

Preparation of native GlcCer (Glcβ1-1′Cer) and GalCer (Galβ1-1′Cer), and production of their semisynthetic isoforms have been described before [37, 38]. Semisynthetic isoforms were free of detectable contamination in analysis by thin-layer chromatography and mass spectrometry. Isoforms produced from pig brain-derived sulfatide have been described previously [14]. Semisynthetic lyso-lactosylcermide (lyso- LacCer) (Galβ4Glcβ1Cer), lyso-sphingomyelin, lyso-GM1 (Galβ3GalNAcβ4(NeuAcα3)Galβ4Glcβ1Cer), and lyso-globotriaosylceramide (lyso-Gb3) (Galα4Galβ4Glcβ1Cer) were purchased from Matreya (PA, USA).

Cellular lipid extracts and fractionation

A total of 20×106 A20CD1d cells were harvested, suspended in ultrapure water, homogenized, and total protein amount was determined with BCA Protein Assay Reagent (Pierce, Rockford, USA). Lipids were extracted using chloroform/methanol/water (C/M/W, 4:8:3 by volume) [39], purified from salt and low molecular contaminants, and separated into acidic and neutral lipids by hydrophobic chromatography using BondElut LRC-18 columns [40]. Duplicate preparations of neutral and acidic fractions were evaporated and analyzed in the T hybridoma assay. Further purification of the neutral fraction was routinely performed in triplicates. The neutral fraction was evaporated, dissolved in 0.2 M NaOH in methanol and incubated at 37°C for 2 h, neutralized with 0.1 M HCl, evaporated, and resolved in chloroform. The fractions were further purified on 1 g silica gel-60 column (230–400 mesh, Merck, Darmstadt, Germany) eluted with 20 bed volumes of chloroform, 20 bed volumes of C/M (9:1) (neutral eluate II), and 15 bed volumes of C/M/W (60:35:8) (neutral eluate III). The neutral lipid eluates II and III were evaporated and analyzed in the hybridoma assay.

Preparative HPTLC

The neutral fractions were further purified using preparative HPTLC with C/M/W 80:20:2 (fraction II) and C/M/W 60:35:8 (fraction III) (Merck). The entire migration area, subdivided into fractions, was scraped off and each fraction was mixed with 1–2 mL C/M/W (30:60:20). Lipids were extracted by sonication and purified on Sephadex G-25 columns (Pharmacia & Upjohn, Uppsala, Sweden) [41], and analyzed in the T hybridoma assay.


The active sample was dissolved in C/M/W (30:60:20, by vol.) before ceramide characterization by mass spectrometry. ESI-MS in the positive mode was performed on a quadrupole-time-of-flight mass spectrometer (Micromass) equipped with a z-spray nonospray ion source. A capillary voltage of 1.2 kV, a source block temperature of 80°C, and a cone voltage of 30 eV were used. Samples were manually loaded into nanoflow probe tips (type F thin wall, Micromass). For MS-MS, the gas cell was pressurized with argon gas, and fragments of precursor ions were produced by collision energy of 30–40 eV.


We thank Vincenzo Cerundolo for providing the LY-B and LY-B/cLCB1 cell lines, and Mitch Kronenberg for carefully reviewing the manuscript. The NIH Tetramer Facility is gratefully acknowledged for provision of biotinylated CD1d monomers. This work was supported by the Strategic Research Center for Mucosal Immunobiology and Vaccines (MIVAC), and grants from the Swedish Cancer Foundation, the Swedish Research Council, LUA-ALF Göteborg to S.C., and the foundations of Adlebert and W. Lundgren and M. Lundgren to S.R. B.P. was supported by the NIH grant AI45053, and S.R. was financed by a PhD project grant to S.C. from the Sahlgrenska Academy, University of Gothenburg.

Conflict of interest

The authors declare no financial or commercial conflict of interest.




bone marrow-derived dendritic cell








GM1 ganglioside


high-performance thin-layer chromatography




long-chain base subunit 1