CD1d-restricted Recognition of Synthetic Glycolipid Antigens by Human Natural Killer T Cells

Human NK T cell clones DN2.B9, DN2.C7, and DN2.D5 were derived, maintained, and phenotypically analyzed as previously described (4, 8). The cervical carcinoma cell line HeLa was transfected with the expression vector pSRα-Neo containing cDNA inserts encoding CD1a, CD1b, CD1c, or CD1d according to previously described methods (20). The HeLa CD1d/a transfectant expressed a chimeric form of CD1d with the extracellular domains of CD1d fused to the transmembrane plus cytoplasmic domains of CD1a. This stable transfectant was produced by the introduction into HeLa cells of a previously described expression construct encoding this chimeric protein (4).

5 × 10⁴ T cells were plated in triplicate in flat-bottomed 96-well plate wells with 10⁴ mitomycin C–treated APCs (0.1 mg/ml for 1 h at 37°C; Sigma Chemical Co., St. Louis, MO). Media for all T cell cultures was RPMI 1640 (GIBCO BRL, Gaithersburg, MD) plus 10% fetal bovine serum (HyClone Labs., Logan, UT) and other additives as previously detailed (21). The synthetic glycolipid antigens used in this study were provided by the Pharmaceutical Research Laboratory, Kirin Brewery Co., Ltd. (Gunma, Japan), and have been described in detail previously (18). Although these were previously designated as glycosyl-ceramides, here we have referred to them as glycosyl-APS derivatives because of the hydroxylations of the third and fourth carbons in their sphingosine base that are characteristic of phytosphingosines (19). Synthetic glycolipids were dissolved at 100 μg/ml in DMSO, and diluted into culture medium to the indicated final concentrations. For mAb blocking experiments, the following mAbs were added as purified IgG to a final concentration of 20 μg/ml: W6/32 (anti-HLA-A, -B, and -C; reference 22); CD1d51 (anti-CD1d; reference 4); and P3 (nonbinding IgG1 control; reference 23). For light fixation of the APC surface, CD1d-transfected HeLa cells were incubated for 1 h on ice in 2 ml of 0.9% NaCl containing 75 mM 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide-HCl (ECDI; Pierce Chemical Co., Rockford, IL), followed by extensive washing in RPMI 1640 medium. Cultures were incubated at 37°C in a 5% CO₂ incubator, pulsed with 1 μCi of [³H]thymidine (2 Ci/mmol) on day 2 and harvested 16 h later for β scintillation counting using a Tomtec harvester and a Betaplate scintillation counter (Wallac, Gaithersburg, MD).

5 × 10⁴ T cells were cultured in triplicate wells with 10⁴ mitomycin C–treated CD1d-transfected HeLa cells in the absence or presence of 100 ng/ml of α-galactosyl-APS (αGalAPS). Phytohemagglutinin (PHA-P; Difco, Detroit, MI) was added to some cultures (1:2,000 final dilution) as a positive control. Supernatants were harvested after 48 h, and cytokine levels determined by capture ELISA as described (4). T cell cytotoxic activity was measured using a standard 4-h ⁵¹Cr-release assay (21). Mock-transfected or CD1d-transfected HeLa cells were labeled with 200 μCi of ⁵¹Cr for 2 h at 37°C, followed by incubation for 12 h at 37°C in complete medium with or without 200 ng/ml of αGalAPS, and then washed and used as target cells. Assays were performed in triplicate and results were expressed as percentage of maximum specific ⁵¹Cr release.

Three previously characterized human NK T clones (DN2.B9, DN2.C7, and DN2.D5; reference 4) bearing the invariant Vα24-JαQ TCR-α chain paired with different Vβ11⁺ TCR-β chains were tested for their ability to respond to synthetic APS-containing glycolipid antigens presented by CD1d-transfected HeLa cells. Because human NK T cells generally have intrinsic autoreactivity to CD1d-expressing cells (4), modifications of the culture conditions were required for optimal demonstration of the response to synthetic glycolipids. Preliminary experiments (data not shown) revealed that CD1d-transfected HeLa cells were particularly suited as APCs for these studies, since they did not require additional exogenous signals such as PMA to support the proliferation of human NK T cells. Using a relatively low APC/T cell ratio (1:5), we observed low to moderate autoreactivity and consistent augmentation of the proliferation of human NK T cells clones when the synthetic glycolipid αGalAPS was added (Fig. 1 a).

The response to αGalAPS appeared to be specific for NK T cell clones, since this compound did not stimulate proliferation of a panel of other human T cell clones that did not express the invariant Vα24-JαQ TCR-α chain (data not shown). The specificity of the response to synthetic glycolipids was also assessed by examining the responses of NK T cell clones to a panel of structurally related glycolipids. These included two previously demonstrated strong stimulators of murine NK T cells (αGalAPS and αGluAPS), as well as four related analogues that were nonstimulatory or weakly stimulatory in the murine system (APS, αManAPS, βGalAPS, and 3,4-deoxy αGalAPS) (18). NK T cell clones responded strongly to αGalAPS and αGluAPS but not to the other analogues (Fig. 1 b, and additional data not shown), demonstrating a pattern very similar to that observed for murine NK T cells. Thus, the specific recognition of synthetic glycolipids by a human NK T cell clone required an α-anomeric sugar (either galactose or glucose) and the presence of hydroxylations at positions 3 and 4 of the phytosphingosine base, exactly as described for murine NK T cells.

In mice, two CD1 genes encode extremely homologous proteins, both of which are most closely related to human CD1d. In contrast, humans have a more diversified family of CD1 proteins, with genes encoding five distinct isoforms designated CD1a, -b, -c, -d, and -e (24). Previous work has shown that the human CD1d protein is a target for direct recognition by human NK T cell clones (4), whereas CD1a, -b, and -c present lipid and glycolipid antigens to other populations of T cells (15–17, 21). To determine if the human NK T cell response to α-glycosyl-APS was restricted only by the CD1d protein, or if other human CD1 isoforms could also support this response, we tested the ability of HeLa cells stably transfected with each of the human CD1 proteins to present αGalAPS. For all three human NK T cell clones tested, the response to αGalAPS was seen only when CD1d⁺ HeLa cells were used as APCs (Fig. 2,a and data not shown). Confirmation of this CD1d-restriction was obtained by mAb blocking experiments, which demonstrated that an mAb against the cell surface expressed conformation of CD1d protein (CD1d51) significantly inhibited the response to αGalAPS (Fig. 2 b). Thus, responses of human NK T cells to glycosyl-APS derivatives were dependent on APC expression of CD1, and CD1d was the only human CD1 isoform that presented these glycolipids to NK T cells.

To determine if antigen internalization by CD1d⁺ APCs was required for presentation, we examined the effects of blocking plasma membrane trafficking by light chemical fixation of APCs. Because aldehyde fixatives caused strong augmentation of the autoreactivity of human NK T cells to CD1d⁺ APCs (4), we instead used the chemical cross-linker ECDI as a fixative (25). CD1d⁺ HeLa cells were surface cross-linked by ECDI either before or after a 12-h pulse with αGalAPS (100 ng/ ml), and then tested for their ability to stimulate proliferation of clone DN2.C7. Whereas APCs treated with ECDI after the antigen pulse stimulated strong proliferation of DN2.C7, fixation with ECDI before antigen pulsing completely abrogated the response to αGalAPS (Fig. 3 a). This strongly suggested a requirement for antigen uptake by APCs in the response to this synthetic glycolipid.

Consistent with this requirement for antigen uptake, studies of αGalAPS presentation to murine NK T cells have demonstrated that this process is inhibited by chloroquine, and thus may require endosomal acidification (18). To further examine this possibility, we assessed the ability of a modified form of human CD1d that is deficient in endosomal localization to present αGalAPS. Human CD1d, like CD1b and CD1c but not CD1a, has a short cytoplasmic tail containing a tyrosine-based four-amino acid motif (24). For CD1b, this motif has been shown to be critical for endosomal localization and efficient microbial lipid antigen presentation by this molecule (20). Previously, we generated a chimeric form of the human CD1d protein in which the CD1d ectodomains are fused to transmembrane and cytoplasmic domains of CD1a, eliminating the endosomal targeting signal normally present in CD1d. Previous studies with this construct expressed in B lymphoblastoid cells showed that the chimeric CD1d/a protein was defective in endosomal localization, but was well expressed on the cell surface (4). To assess the importance of endosomal delivery of CD1d, we examined the dose response of clone DN2.C7 to αGalAPS presented by HeLa cells expressing either wild-type CD1d or the CD1d/a chimera. This revealed a striking effect of the deletion of the targeting motif, essentially abolishing the ability of CD1d/a to present αGalAPS (Fig. 3 b). This finding, along with the requirement for antigen internalization by the APCs, strongly suggested that synthetic glycolipid antigens colocalize with CD1d in the endocytic system, and that this is required for the generation of a complex that is recognized by NK T cells on the APC surface.

A hallmark of both murine and human NK T cells has been their ability to rapidly produce cytokines associated with both Th1 (IFN-γ) and Th2 (IL-4) responses upon TCR engagement (1, 4, 11). Interestingly, however, it has been noted that several conditions exist in which this apparent Th0 profile of cytokine production may be skewed either toward a predominance of IFN-γ (11, 26) or of IL-4 (14, 27). We found that stimulation with αGalAPS generated substantial levels of both IFN-γ and IL-4 by the three human NK T cell clones tested (Fig. 4,a, and data not shown). Some variation between clones in the relative amounts of IFN-γ and IL-4 production was observed, but in all cases the cytokine production mirrored that observed with activation by PHA. Human NK T cell clones also showed enhanced cytotoxic activity against CD1d⁺ HeLa cells pulsed with αGalAPS (Fig. 4 b). Thus, CD1d-restricted recognition of αGalAPS augmented the known effector functions of human NK T cell clones, and preserved their Th0 pattern of cytokine secretion.

Our results demonstrated CD1d-restricted recognition of synthetic glycosyl-APS by human NK T cells, and confirmed the striking conservation of this T cell subset between humans and mice in both phenotype and function. The presentation of synthetic glycolipids by CD1d also supports the growing body of data demonstrating that the main function of CD1 proteins may be the presentation of foreign microbial lipid and glycolipid antigens (15–17, 21). In spite of this substantial progress, the main immunological functions of NK T cells remain unclear. The synthetic glycolipids used in this study are not yet known to have precise counterparts in relevant infectious agents, and the prototype natural compounds in this family were originally isolated not from microbial pathogens but from marine sponges (18). Nevertheless, given the impressive antitumor effects of NK T cells and a variety of studies indicating that they may play a role in regulating autoimmune inflammatory diseases, the discovery of exogenous ligands that can modulate the activity of this unique T cell subset is of great potential importance. Our results demonstrating that synthetic glycosyl-APS derivatives are active modulators of the human NK T cell system suggests that these compounds may be promising candidates for immunotherapy of human diseases.

The authors thank Drs. Branch Moody and Brian Wilson for thoughtful suggestions, and Drs. Steven Balk and Mark Exley (Beth Israel-Deaconess Medical Center, Boston, MA) for many helpful discussions and for providing the CD1d/a plasmid DNA construct.

This work was supported by National Institutes of Health grant R29 AI-40135, a grant from the American Cancer Society, and an Arthritis Foundation Investigator Award to S. Porcelli.