CD1 proteins present various glycolipid antigens to T cells, but the cellular mechanisms that control which particular glycolipids generate T cell responses are not understood. We show here that T cell recognition of glucose monomycolate antigens with long (C80) alkyl chains involves the delivery of CD1b proteins and antigens to late endosomes in a process that takes several hours. In contrast, analogs of the same antigen with shorter (C32) alkyl chains are rapidly, but inefficiently, presented by cell surface CD1b proteins. Dendritic cells (DCs) preferentially present long-chain glycolipids, which results, in part, from their rapid internalization and selective delivery of antigens to endosomal compartments. Nonprofessional antigen-presenting cells, however, preferentially present short-chain glycolipids because of their lack of prominent endosomal presentation pathways. Because long alkyl chain length distinguishes certain microbial glycolipids from common mammalian glycolipids, these findings suggest that DCs use a specialized endosomal-loading pathway to promote preferential recognition of glycolipids with a more intrinsically foreign structure.
CD1 proteins are expressed on a variety of hematopoietic cells—including myeloid dendritic cells (DCs), B cells and cortical thymocytes—and bind and present glycolipids to antigen-specific T cells1,2. CD1 proteins are thought to accomplish their antigen-presenting function by binding the lipid portion of the antigen within a deep, hydrophobic groove on the membrane-distal surface of CD1. This allows the hydrophilic components of the antigen to protrude from the groove so that they may contact T cell antigen receptors (TCRs)3,4,5,6,7,8,9,10. CD1-mediated glycolipid presentation has been proposed to have two roles in the immune response. The constitutive expression of certain CD1 isoforms by myeloid lineage cells, B cells and cortical thymocytes may promote recognition of self- or altered self–glycolipids during immunosurveillance of transformed cells or immunoregulatory interactions between immune cells11,12. In addition, CD1a, CD1b and CD1c proteins can be up-regulated on myeloid DCs at sites of microbial infection, where they present bacterial antigens during host defense13,14,15.
Consistent with these proposed dual functions, the specific glycolipid targets of CD1-mediated T cell responses include self-glycolipids, such as mammalian phosphatidylinositols and gangliosides, as well as foreign microbial lipids, such as mycolyl glycolipids, lipoarabinomannan and mannophosphoisoprenoids15,16,17,18,19,20. T cell activation by a given antigen is mediated by clonally variable elements of TCRs and is not generally cross-reactive with structurally related glycolipids or with other known CD1-presented glycolipids9,11,20. Thus, CD1-restricted T cells do not simply respond to the presence or absence of glycolipid bound in the CD1 groove; instead they are highly specific for the structure of particular classes of glycolipids that vary in the composition of their lipid and carbohydrate moieties10. These observations suggest that antigen-presenting cells (APCs) could substantially influence immune responses by controlling which particular classes of self, altered self or foreign glycolipids are loaded onto CD1 proteins and presented to T cells.
Given the abundance of lipids as the chief constituents of cellular membranes, it is likely that mechanisms exist to selectively present lipid antigens involved in immunoregulation or host defense while simultaneously limiting the presentation of normal cellular lipids that could trigger deleterious autoimmunity. However, little information currently exists regarding the particular chemical features of glycolipid structures that control antigen trafficking within cells or loading onto CD1 proteins in APCs. In addition, there is no consensus regarding the cellular pathways that lead to the presentation of glycolipids by CD1b proteins. The CD1b-mediated antigen-presentation pathways in myeloid DCs are perhaps the most extensively studied to date. However, substantial controversy exists over whether glycolipid antigens can load directly onto CD1b proteins at the cell surface or instead require internalization of CD1b and antigens to late-endosomal compartments for loading at low pH.
Cellular studies of presentation of lipoarabinomannan, free mycolic acid and glucose-6-O-monomycolate (GMM) show that antigen presentation is inhibited by fixing APC membranes, treating APCs with inhibitors of endosomal acidification or by reducing CD1b trafficking to late endosomes and major histocompatibility complex (MHC) class II–enriched prelysosomal compartments (MIICs)14,18,21,22,23. These studies suggest that these microbial glycolipids require transport of CD1b and antigen to MIICs, where low pH may induce conformational changes of the CD1b protein or other effects that favor loading of lipids into the groove6,24. However, this intracellular pathway does not appear to be universally required for CD1b-mediated antigen presentation, as mammalian GM1 gangliosides can efficiently load onto recombinant CD1b proteins or onto CD1b at the surface of cells treated with fixatives17,25.
As alkyl chain length is a major factor controlling the loading of lipids into the hydrophobic cavities of lipid-binding proteins other than CD1, we hypothesized that the size of the lipid moiety might affect loading of antigens into the CD1 groove10,26. Therefore, we measured CD1b-mediated T cell activation by GMM antigens that varied incrementally in the size of the lipid moiety, but shared the same carbohydrate and proximal α-branched, β-hydroxylated acyl chain structure that constitutes the TCR recognition epitope20. Our results show that antigens with short alkyl chains could rapidly load onto CD1b proteins at the cell surface or in other nonendosomal compartments, whereas efficient delivery of antigen and CD1b proteins to late endosomal compartments or MIICs favored the presentation of antigens with long alkyl chains. In addition, the endosomal pathway for presentation of foreign long-chain glycolipids was up-regulated in DCs. In contrast, the nonendosomal pathway that favors presentation of short-chain lipids similar to those found abundantly in normal host cell membranes was more active in cell types that are likely not involved in the stimulation of T cell responses during infection. These results suggest a strategy by which certain CD1b-expressing cells, such as cortical thymocytes, may be mainly programmed to present abundant classes of short chain self lipids, whereas DCs are skewed toward the presentation of long-chain lipid antigens that are more typical of bacterial pathogens.
DC presentation of glycolipid chain length variants
To investigate the role of lipid length on glycolipid antigen presentation, GMM antigens that varied in alkyl chain length were tested for activation of CD1b-restricted T cells when presented by monocyte-derived immature DCs. Like all known CD1b-presented antigens, these GMMs were composed of a hydrophilic head group that functions as the TCR epitope and two alkyl chains that probably mediate interactions with CD1b20,27. GMMs are naturally produced by mycobacteria and related actinomycetales, so we purified GMM from Mycobacterium phlei, Nocardia farcinica and Rhodococcus equi. Mass spectrometric detection of sodium adducts indicated that the natural GMMs from these three bacterial species had overall lipid lengths (that is, mycolate moieties) of C74–82, C50–58 and C30–36, respectively (Fig. 1a). To test the effects of lipid length on antigen presentation over the broadest possible size range, we also synthesized shorter synthetic GMMs with total alkyl chain lengths ranging from C32 to C12 (Fig. 1b)28.
Each of the seven GMM antigens activated the CD1b-restricted T cell line LDN5, which showed that DCs can present glycolipids that vary substantially in lipid moiety size (Fig. 1c,d). However, alkyl chain length strongly influenced antigenic potency. The synthetic GMM analog that contained a C12 mycolic acid only minimally activated T cells at high doses, which were just below those that caused toxicity to cells. This indicated that the C12 lipid provides the minimal lipid length that supports T cell activation in this system. Presentation of GMMs with longer alkyl chains by DCs activated T cells more strongly in all cases, and mycobacterial GMM with a C80 mycolic acid was 100-fold more potent than all other antigens tested.
Mycolyl lipid uptake by DCs
Because GMM recognition requires intracellular loading or processing by APCs, we studied GMM interactions with APCs to determine whether these marked differences in antigenic potency were due to the effects of lipid length on antigen delivery, uptake, intracellular transport or association with CD1b. Two purified natural GMMs—with predominant molecular species that corresponded to lipid lengths of either C80 or C32—were biosynthetically radiolabeled with 14C, which allowed direct measurement of their physical interactions with APCs. These two glycolipids were chosen from the larger panel because their lipid sizes were near the extremes found in natural glycolipids with two alkyl chains. C80 GMM is typical of mycolyl glycolipids that compose the outer cell walls of mycobacteria and C32 GMM more closely approximates the lipid length of glycolipids that comprise mammalian cell membranes.
Because lipid length affects glycolipid solubility in aqueous biological solutions, we first measured the delivery of exogenous [14C]C32 and [14C]C80 GMM to APCs. Both antigens were reproducibly and efficiently suspended in media by sonication, and scintillation counting of serially diluted antigen correlated well with predicted concentrations (R2>0.995, data not shown). The rate of antigen capture by DCs was measured by culturing radiolabeled antigens with DCs, extensively washing the cells, then scintillation counting of cell-associated antigen. In contrast to protein or particulate antigens, it is likely that glycolipids adhere to the DC surface or integrate into the plasma membrane. Therefore, assays were also done in the presence of metabolic inhibitors—2-deoxyglucose (2-DOG) and sodium azide (NaN3)—to distinguish the pool of antigen that adhered to the cell surface from that which was taken up into cells by energy-requiring mechanisms. Small amounts of GMM associated with DCs in the presence of metabolic inhibitors, which demonstrated a low but detectable rate of GMM adherence to metabolically inactive cells (Fig. 2a). Antigen association with DCs in the absence of metabolic inhibitors was dependent on the dose and time of antigen exposure, which indicated that energy-requiring mechanisms were used to capture exogenous glycolipid antigen (Fig. 2a and data not shown).
To directly visualize the delivery of mycolyl lipids to intracellular compartments, C80 mycolate was conjugated to the fluorescent label bodipy (bodipy-mycolate), incubated overnight with DCs, then analyzed by confocal microscopy. DCs labeled in this way showed a vesicular staining pattern, including staining in compartments that colabeled with HLA-DM, a marker of MHC class II compartments (MIIC, Web Fig. 1 online). In addition, an acid-mediated cleavage product of [14C]C80 GMM could be recovered from DCs cultured in media, but this cleavage product was not found when DCs were treated with concanamycin B21 (data not shown). This provided functional evidence for the delivery of GMM to compartments with vacuolar proton-ATPases, including late endosomes and lysosomes. Taken together, these data provide evidence for the delivery of long-chain mycolyl lipids into the endosomal network of DCs.
Selective sorting of C80 GMM to MIICs
To determine whether differential cellular uptake of C80 and C32 GMM could account for their differing potency for T cell activation, we measured the metabolism-dependent uptake rates at 0.5 μM. At this concentration, the T cell response to C80 GMM was maximal, but the T cell response to C32 GMM was not detected (Fig. 1c). There were no significant differences in the metabolism-dependent association of antigens with DCs, which showed that the greater antigenic potency of C80 GMM was not accounted for by differences in its overall uptake by DCs (Fig. 2b). Studies have shown that diacylglycerols and dialkylindocarbocyanine derivatives enter into early-sorting endosomes, where analogs with longer alkyl chains are selectively sorted to late endosomal or lysosomal compartments29. Therefore, we considered the possibility that cellular sorting of C80 GMM to these compartments might underlie its higher potency for T cell activation, as published data have indicated that C80 GMM likely loads onto CD1b in late endosomal or lysosomal compartments21,22.
To test this, DCs were incubated for 24 h with [14C]C32 or [14C]C80 GMM; cells were then washed, disrupted and their membranes fractionated by density gradient centrifugation over Percoll30. Fractions prepared with either antigen yielded virtually identical profiles of β-hexosaminidase, a marker of lysosomes, indicating that separation of organelles was comparable for cells pulsed with either antigen (Fig. 2c). However, a significant fraction of C80 GMM reproducibly accumulated in fractions 1 to 3 that contained CD1b and were identified as lysosomal fractions based on their high density, enrichment of β-hexosaminidase and lack of detectable transferrin receptor, MHC class I and Rab5 (Fig. 2c,d). In each of four separate experiments, C80 GMM accumulated in higher amounts than C32 GMM in lysosomal fractions; on average, C80 GMM was present at 2.9-fold higher amounts than C32 GMM in the peak lysosomal fraction (Fig. 2d, fraction 2, P=0.025). This was not due to the intrinsic physical properties of the antigens that could have differentially affected separation in Percoll, as neither antigen accumulated in dense Percoll fractions when they were separated as free lipids that were not associated with cellular membranes (data not shown). These results suggest that DCs sorted the antigens based on alkyl chain length and preferentially concentrated the long-chain antigen in dense lysosomal or prelysosomal MIICs, where efficient loading onto CD1b probably occurs21,22,23,30.
GMM presentation in B cells versus DCs
To further investigate the role of cellular pathways on presentation of antigens with different alkyl chain lengths, we measured the presentation of GMM chain–length analogs by immature DCs and CD1b-transfected C1R B lymphoblastoid cells (referred to hereafter as C1R.CD1b cells). These two cell types differ in their modes of exogenous antigen acquisition and delivery to endosomal pathways31. Whereas alkyl chain length affected the efficiency of antigen presentation by several hundred–fold in DCs, C1R.CD1b cells presented GMMs with C32, C54 and C80 alkyl chains with similar efficiency (Fig. 3a). Therefore, the effects of alkyl chain length could not be solely explained by differing interactions of the GMM chain–length analogs with wild-type CD1b proteins, but were most likely due to inherent differences in uptake or processing of the glycolipids between these cell types.
Although DCs and B lymphoblastoid cells differ in many ways that could potentially explain the differing efficiency of antigen presentation based on chain length, known differences in antigen internalization by these two cell types provided a potential explanation for the preferential presentation of longer chain GMM by DCs. Immature DCs are capable of phagocytosis and rapid macropinocytosis, which potentially allows them to internalize exogenous glycolipid antigens more rapidly than B lymphoblastoid cells31. Therefore, these results might be explained if the long chain form of the antigen required internalization for optimal loading onto CD1b but the short chain form did not. To test this possibility, we first measured the uptake of [14C]GMM by both cell types. As expected, metabolism-dependent association of GMM with immature DCs was much greater than with C1R.CD1b cells (Fig. 3b). In fact, metabolism-dependent and metabolism-independent rates of GMM association with C1R.CD1b cells were similar, which suggested that GMM could adhere to metabolically inactive C1R.CD1b cells, but that energy-requiring uptake mechanisms—if present at all—were inefficient.
Cell surface loading of C32 GMM
To determine the requirements for antigen internalization into APCs before T cell recognition, DCs were fixed with glutaraldehyde before or after exposure to GMM and their ability to activate T cells was tested. Unfixed DCs or DCs that were fixed after exposure to antigen preferentially presented C80 GMM to T cells (Fig. 3a,c). In contrast, fixation before antigen exposure completely and reproducibly blocked the ability of DCs to present C80 GMM but not C32 GMM (Fig. 3c). This suggested that C32 GMM, unlike C80 GMM, could directly load onto CD1b at the surface of fixed cells.
To test this further, the time required to generate antigenic CD1b-GMM complexes in APCs was determined by exposing antigen to APCs for various time-periods and then immediately measuring T cell activation by intracellular calcium-flux. This was accomplished by flow cytometric analysis of T cells loaded with Fluo-4, a dye with emission in the range of 516 nm (FL1) which is enhanced upon calcium binding, and Fura red, a dye with emission in the range of 670 nm (FL3) which is suppressed by calcium binding32. The percentage of activated T cells was measured by detecting the number of cells in an activation gate defined by high FL1 and low FL3. Initial experiments measuring LDN5 activation by calcium-induced alterations in fluorescence demonstrated that 1% of resting T cells or T cells coincubated with C1R.CD1b cells without antigen were detected within the activation gate (Fig. 4a,b). Treatment of T cells with the calcium ionophore ionomycin or coincubation with C1R.CD1b cells pulsed with GMM resulted in a substantial percentage of cells moving into the activation gate. Further experiments showed that calcium flux was dependent upon antigen dose, expression of CD1b on APCs and cell-cell contact that resulted from brief centrifugation (Fig. 4b and data not shown).
Comparison of T cell activation by cells that had been pulsed with GMM for various time-periods demonstrated that treatment of C1R.CD1b cells with C32 GMM for less than 1 min resulted in substantial activation of GMM-specific T cells; this rose to a maximum within 10 min of preincubation. In contrast, C80 GMM required >4 h of preincubation with C1R.CD1b cells to generate maximal T cell responses (Fig. 4c). Similar results were obtained for the presentation of these GMMs with immature DCs (data not shown). The rapid sensitization of APCs to recognition of C32 GMM was most consistent with the loading of this antigen at the cell surface, and the much longer time for presentation of GMM with the C80 lipid was consistent with the use of more complex cellular processing or loading pathways.
To assess the loss of stimulation by C32 and C80 GMM–treated APCs, we measured presentation of these two antigens after the APCs were antigen-pulsed. C1R.CD1b cells were pulsed overnight with either C32 or C80 GMM at concentrations that led to equivalent, but suboptimal, T cell activation under steady-state antigen–loading conditions. Subsequent chase of antigen-loaded C1R.CD1b cells in antigen-free media showed that stimulation by C32 GMM declined by ∼50% over 5 h, whereas T cell stimulation by C80 GMM was unchanged (Fig. 4d). The decline in T cell activation by C32 GMM was even more rapid and more complete after a brief 30 min antigen pulse, whereas the stimulation by C80 GMM–pulsed cells actually increased under these experimental conditions. The marked differences in the kinetics of processing of these two antigens in the presence and absence of extracellular antigen demonstrated that although C80 GMM is initially processed more slowly (Fig. 4c), it is presented in a form that is recognized by T cells on the surface of APCs with a substantially longer half-life than C32 GMM (Fig. 4d).
Disruption of endosomal trafficking of CD1b
Because only C80 GMM required cellular uptake for efficient presentation, the greater antigenic potency when presented by immature DCs may have resulted from its more rapid internalization and delivery to intracellular compartments by this cell type (Figs. 3 and 4c). However, DCs and C1R cells also differed in cellular functions other than antigen internalization that could affect the efficiency of GMM presentation. To investigate the role of CD1b trafficking to endosomes in the presentation of chain length analogs, we measured GMM presentation by C1R cells transfected with CD1b proteins in which sequences near the transmembrane region that target CD1b to late endosomal compartments had been mutated. Replacement of the transmembrane and cytoplasmic domains of CD1b with a sequence from decay-accelerating factor (referred to hereafter CD1b.DAF), which encodes a signal for a glycosyl phosphatidylinositol (GPI) modification, allowed expression of CD1b as a GPI-anchored protein33,34. Anchoring with GPI leads to the accumulation of proteins in the cholesterol-rich microdomains of the plasma membrane, and GPI modification of CD1b leads to decreased presentation of C80 GMM35.
Due to the intrinsic limitations of total cellular production of GPI-CD1b, the cell surface amounts of CD1b in C1R.CD1b.DAF cells were lower than those of DCs or C1R cells transfected with wild-type CD1b. Thus, the absolute efficiency of presentation could not be directly compared among the three cell types. Nevertheless, C1R.CD1b.DAF cells presented the C32 GMM more efficiently than the C80 GMM, in contrast to wild-type CD1b proteins expressed in either C1R cells or DCs (Figs. 3a and 5a). The preferential presentation of C32 GMM by C1R.CD1b.DAF was consistent with the view that redirecting CD1b trafficking from late endosomes to other compartments favors presentation of the short-chain form of the antigen.
We also measured antigen presentation by C1R cells that expressed CD1b proteins with mutations in a tyrosine-containing targeting motif that controls CD1b delivery to late endosomes. The short cytoplasmic tail of CD1b contains the sequence YQNI, which conforms to a motif that allows interaction with clathrin adaptor protein (AP) complexes and promotes trafficking of CD1b proteins to late endosomes36. Consistent with published data, deletion of the entire cytoplasmic tail (CD1b.TD) or alanine substitution of the critical tyrosine at position 311 in this motif (CD1b.Y311A) caused redistribution of CD1b from MHC class II+LAMP 1+ endosomes to the cell surface and reduced efficiency of presentation of C80 GMM22,23 (Fig. 5b,c). Alteration of the endosomal-targeting sequence had the opposite effect on C32 GMM presentation. C1R cells that expressed CD1b.TD or CD1b.Y311A presented C32 GMM more efficiently than C1R cells that expressed equivalent amounts of wild-type CD1b (Fig. 5b,c). The increased C32 GMM antigen-presentation efficiency by CD1b.TD proteins was even apparent in C1R cells expressing low mean fluorescence intensity (MFI=423) of CD1b.TD compared to C1R cells expressing high amounts of wild-type CD1b (MFI=887) (Fig. 5d).
Thus, three mutations that disrupt the normal trafficking of CD1b to late endosomes augmented presentation of C32 GMM and inhibited presentation of C80 GMM. This suggests that the effects of alkyl chain length on antigen recognition by T cells were dependent upon whether or not the cellular pool of CD1b proteins efficiently trafficks through endosomal compartments. The more efficient presentation of C32 GMM by cells expressing mutant CD1b proteins that lack endosomal-targeting motifs suggested that delivery of CD1b to endosomes was not only unnecessary for C32 GMM presentation, but that passage of CD1b through endosomes inhibits presentation of this form of the antigen in nonendosomal compartments.
GMM presentation by thymocytes
B lymphoblastoid cells provided an experimental system that lacked wild-type CD1b proteins; thus, the effects of CD1b mutations on trafficking or presentation of antigens with varied chain lengths could be assessed. However, the two principal cell populations that naturally express CD1b in vivo are DCs and cortical thymocytes, rather than B cells. Thymocytes constitutively express CD1b when present in the thymic cortex, but lose expression upon maturation, which is consistent with a role for CD1b presentation of endogenous glycolipids during thymic selection events37,38. Such a role is supported by in vivo studies of CD1d-restricted T cell development in mice in which the production of peripheral CD1d-restricted T cells requires CD1d expression39,40. Therefore, we measured GMM presentation by a thymic leukemia cell line (HPB-ALL) that retains the characteristics of immature cortical thymocytes, including constitutive expression of CD1b41. In contrast to DCs, which presented the C80 GMM more efficiently, HPB-ALL cells preferentially presented the C32 GMM (Fig. 6). Thus, two cell types, which represent the two major populations of human CD1b+ APCs, showed marked differences in their ability to present GMM antigens of varying chain length.
CD1b proteins can bind and present antigens containing lipid moieties composed of diacylglycerols, sphingolipids and mycolic acids, which—with their variously glycosylated derivatives—represent a diverse range of potential targets of CD1b-mediated T cell responses17,18,19,20. The pool of glycolipids that is available for recognition by T cells after loading onto CD1 proteins is small compared to the total pool of lipids present in the membranes of APCs. Thus, events that control which particular glycolipids are loaded onto CD1 proteins at the cell surface, or during their trafficking through secretory and endocytic pathways, may play a major role in determining the outcome of CD1-mediated immune responses. Certain proposed functions of CD1-expressing APCs—such as immunosurveillance of altered self–lipids in transformed cells or the induction of tolerance to self-glycolipids—would likely require that antigen-presentation pathways randomly sample the entire complement of lipids that compose the membranes of APCs. On the other hand, T cell activation by trace amounts of microbial or altered self–glycolipids that are intermixed with a much larger pool of self-antigens of normal structure would be facilitated by mechanisms for sorting and selectively presenting lipids with distinct structural features, such as long alkyl chain length.
We show here that glycolipid alkyl chain length controls the kinetics, subcellular localization and, ultimately, the efficiency of antigen presentation by different types of CD1b-expressing APCs. Because C32 GMM was recognized by GMM-specific T cells within seconds and was presented by fixed and unfixed APCs, this antigen is likely to load directly onto CD1b proteins at the cell surface without cellular processing requirements. In addition, C1R cells with mutant CD1b proteins that cannot efficiently internalize CD1b from the cell surface to endosomes presented C32 GMM more efficiently than longer chain GMMs. In contrast, antigens that contained the same TCR epitope, but with longer alkyl chains, required several hours of incubation with APCs, could not be loaded onto fixed cells and were presented less efficiently by C1R cells expressing CD1b proteins that lacked endosomal targeting motifs. These results indicate that long alkyl chain length prevents mycobacterial C80 GMM from loading onto cell-surface CD1b proteins; instead, they must be presented with the use of intracellular antigen-processing pathways that involve late endosomes or lysosomes. Deletion of endosomal-targeting motifs had pronounced and opposite effects on the presentation of C32 and C80 GMMs; this implicated endosomal processing events, rather than other aspects of APC-antigen interactions, as the basis for the differing presentation efficiencies of chain length analogs. Taken together these results indicate that lipid length controls the entry of glycolipids into separate, but coexistent, pathways of endosomal and nonendosomal antigen presentation by DCs.
The differential use of endosomal versus nonendosomal pathways may account for preferential presentation by DCs of GMMs with longer alkyl chains; this contrasts with transformed B cells or thymocytes, which present antigens with short alkyl chains as well as or better than C80 GMM. Entry into the endosomal pathway first requires that antigens are internalized and delivered to late endosomes or MIICs. DCs internalized GMM more rapidly than other cell types and sorted glycolipids based on alkyl chain length, preferentially delivering a larger percentage of the long-chain GMM to MIICs. Thus, the rapid delivery of exogenously supplied long-chain antigens to late endosomes or MIICs of DCs can selectively promote the presentation of long-chain antigens by this cell type. In contrast, the inability of transformed B cells or thymocytes to efficiently deliver antigens to this compartment probably underlies the preferential presentation of short-chain GMM by these nonphagocytic APCs. Other aspects of endosomal-loading mechanisms may also promote the preferential presentation of long-chain antigens observed in DCs. In particular, our studies of transfected C1R cells, which showed that reduced endosomal trafficking enhanced the ability of CD1b to present C32 GMM, suggested that trafficking of CD1b proteins through endosomal compartments inhibits the loading of short-chain GMM in nonendosomal compartments.
Identification of these pathways through studies of GMM chain–length analogs may account for what was previously viewed as differing cellular requirements for processing of other CD1b-presented antigens, including free mycolate and GM1 gangliosides. There is evidence that each of the known CD1b-presented antigens with long alkyl chains (C80 GMM, C54 GMM and C80 mycolate) require cellular uptake and delivery to endosomes for presentation by CD1b, and each of these long chain antigens are efficiently presented by DCs when present at low (nanomolar) concentrations18. In contrast, antigens with relatively short alkyl chains (C32 GMM and C34–44 GM1 gangliosides) can load directly onto CD1b proteins in the absence of endosomal processing, and both of these antigens require high (micromolar) concentrations to sensitize DCs to recognition by CD1b-restricted T cells17,25. Thus, the apparent differences in the cellular requirements for presentation of several CD1b-presented antigens can be explained by alkyl chain length controlling the entry of long chain lipids into efficient endosomal antigen-presentation pathways and short-chain lipids into less efficient nonendosomal pathways. In fact, the only known CD1b-presented antigen that does not conform to this model is LAM, a lipoglycan with an unusually large (20 kD) glycan moiety that is delivered to late endosomes by the mannose receptor24.
Although the molecular basis for the selective ability of short-chain antigens to directly load onto cell-surface CD1b proteins has not been experimentally determined, quantitative molecular models provide insights into this process. The x-ray crystal structure of murine CD1d1 indicates that the groove is large enough to optimally accommodate 31 or 32 methylene units, a value which corresponds to the size of the lipid moieties of phoshatidylinositols eluted directly from cellular CD1d proteins3,7,10. Our results showed that CD1b-expressing DCs can present glycolipids ranging in alkyl chain length from C12 to C80. Antigens composed of lipids that approximate the predicted volume of the CD1 groove can load directly onto CD1b, whereas antigens with larger lipid moieties must undergo endosomal delivery before T cell recognition. Therefore, targeting of glycolipids and CD1b proteins to late endosomes may be required for molecular modification of antigens or of CD1b proteins to enable glycolipids with large lipid moieties that exceed the size of the CD1b groove to be loaded. This may involve enzymatic cleavage of the lipid chains or pH-induced relaxation of the α-helices that form the groove. The latter mechanism could facilitate insertion of long-chain lipids so that part of the acyl chain protrudes from the groove, analogous to MHC class II–loading of peptides of varied length6,10.
MHC class I loads self and viral antigens in the endoplasmic reticulum, and MHC class II proteins predominantly load exogenous antigens in endosomal compartments42. Our results suggest that CD1b also optimizes antigen loading in cellular compartments that are enriched in those antigens that serve as the physiological targets for T cells. The relative inefficiency of nonendosomal-loading mechanisms may allow the sampling of abundant classes of short-chain self-diacylglycerols (C30–38) or sphingolipids (C34–44) during CD1b trafficking through the endoplasmic reticulum, golgi, cell surface and other compartments. The preferential use of this pathway by thymocytes is consistent with their proposed function of providing ligands for the TCRs of CD1-restricted T cells during thymic selection39. In addition, we speculate that the use of the nonendosomal pathway for presentation of self–lipid antigens with relatively short alkyl chains by APCs in the periphery could be involved in the maintenance of tolerance in vivo against antigens such as gangliosides. A CD1d-presented α-galactostyl ceramide with truncated alkyl chains promotes interleukin 4 (IL-4) secretion by T cells and protection against experimental allergic encephalomyelitis43. This indicates that alkyl chain length can effect whether antigen recognition results in tolerance versus an aggressive inflammatory response.
Unlike thymocytes, which express CD1b but probably do not play a role in presentation of foreign antigens, DCs migrate to sites of microbial infection where they up-regulate CD1b expression and take-up exogenous microbes or microbial glycolipids into the endosomal network13,44. Certain classes of microbial glycolipids, such as mycolyl glycolipids (C70–80), have lipid moieties that greatly exceed the length of self glycolipids typically found in mammalian APCs45. Because microbial glycolipids are more likely to have long alkyl chains and to be concentrated in the endosomal network after internalization, preferential use of the endosomal pathway by DCs may favor the presentation of foreign antigens, consistent with the function of this cell type in host defense. Therefore, the endosomal and nonendosomal pathways of glycolipid antigen presentation by CD1b may perform important, but separate, functions in shaping immune responses to this class of T cell antigens.
M. phlei, N. farcinica and R. equi (American Type Culture Collection, Manassas, VA) were grown in 7H9 medium (Gibco, Grand Island, NY) supplemented with TWEEN 80 and 10 g/l of d-glucose (Sigma, St. Louis, MO), then subjected to extraction with chloroform:methanol (2:1) and GMM purification on silica columns as described27. Radiolabeled GMM was produced in the same manner except that 1 mCi/l of [14C]acetate (Perkin Elmer, Boston, MA) was added in early log phase for 24 h. Synthetic GMMs were prepared by condensation of Cx fatty acids to yield C2x mycolates that were 3-tert-butyldimethylsilylated and coupled to 1,2,3,4 tetra-O-dimethylsilylated glucose as described28. Each of the synthetic GMMs contained similar amounts of threo and erthro mycolates, as determined by thin layer chromatography (TLC) analysis of saponified methyl esters20,27. Structures were confirmed by positive mode electrospray ionization mass spectrometry of sodium adducts as described20.
The efficiency of antigen presentation was measured by activation of the CD1b-restricted human αβ TCR+ CD4−CD8− T cell line LDN5 or the Jurkat T cell leukemia line J.RT-3 cells transfected with LDN5 TCRα and β chains (J.RT-3/LDN5αβ), both of which have the same antigen specificity)27,20. Monocyte-derived DCs were prepared from the peripheral blood of healthy donors by centrifugation over Ficoll-Hypaque (Amersham, Piscataway, NJ), adherence of mononuclear cells to plastic tissue-culture flasks (Falcon, Franklin Lakes, NJ), culture of adherent cells with 300 U/ml of granulocyte-macrophage colony-stimulating factor and 200 U/ml of IL-4 for 72–96 h, followed by γ-irradiation (5000 rad) as described27. C1R lymphoblastoid cells were transfected with pREP7β vector containing DAF-derived GPI signal sequence fused with the extracellular domain of CD1b truncated at Trp277 as described35. C1R cells transfected with the vector pSRα-NEO vector alone (CD1b.mock) or the vector that contained the cDNA encoding full-length human CD1b (C1R.CD1b), CD1b with an alanine substitution at position 311 (CD1b.Y311A) or CD1b lacking the 7–amino acid cytoplasmic tail distal to Arg308 (CD1b.TD) were also used22. The HPB-ALL thymic leukemia cell line was a gift of H. Band41. CD1b cell surface expression was evaluated by flow cytometry (BD FACSort, BD Pharmingen, San Diego, CA) with the use of a monoclonal antibody (mAb) to CD1b (4A7 or BCD1b3.1) stained with fluorescein isothiocyanate–conjugated goat anti–mouse Ig F(ab′)2 (Biosource, Camarillo, CA).
Immature DCs were incubated with 7.5 μg/ml of either [14C]C32 or [14C]C80 GMM for 24 h, washed four times with PBS, homogenized and fractionated on Percoll gradients as described30. Fractions were assayed for β-hexosaminidase activity as a marker for enrichment of lysosomes30. To analyze the distribution of lipids, duplicate samples of the fractions were counted in a liquid scintillation counter. For immunoblotting, 300 μl of fractions were centrifuged for 45 min in a TLA100.1 rotor at 70,000 rpm. The membrane pellets were isolated and taken up in sample buffer and loaded onto a 12% SDS–polyacrylamide gel, transferred to PVDF membranes and blotted for CD1b, TfR, MHC class I and Rab5. Antibodies and methods for immunoblotting were as described30, except for the rabbit anti–human Rab5, which was used at a dilution of 1:500. Statistical comparison of GMM accumulation on β-hexosaminidase positive fractions was done with the Student's t-test for paired values.
T cell activation assays.
Glycolipid antigens were prepared by aliquoting antigens in organic solvents into glass tubes. Solvent was removed by drying under nitrogen, then T cell media was added and antigen was suspended by sonication in a water bath for 2 min. T cell proliferation was measured by incubating 5×104 T cells with antigen and 5×104 irradiated (5000 rad) monocyte-derived DCs in 96-well microtiter plates for 3–4 days; this was followed by the addition of 1 μCi [3H]thymidine (Perkin Elmer) and an additional 6 h of culture before samples were collected and β-emissions counted as described49. IL-2 release was measured by culture of 105 JRT-3/LDN5αβ cells with 10 ng/ml of PMA and 5×104 DCs or C1R cells plus antigen in 200 μl/well in 96-well microtiter plates. After 24 h, 50 μl of supernatant was transferred to wells containing 100 μl of media and 5×103 IL-2–dependent HT-2 cells. Cells were cultured for 24 h before 1 μCi of [3H]thymidine was added for an additional 6–24 h of culture; samples were then collected and β-emissions counted.
Kinetic analysis of T cell activation of calcium flux was completed by incubating T cells with 4 μM Fura red and 2 μM Fluo-4 (Molecular Probes, Eugene, OR) at room temperature for 45 min. Cells were then washed three times with media and resuspended at 106 cells/ml in Hank's Buffered Salt Solution (HBSS, Gibco-BRL) with 5% bovine serum albumin (BSA, Sigma). APCs that had been pulsed with antigen were washed and suspended in HBSS at 106 cells/ml. They were then added in equal volume and subjected to centrifugation in a table-top centrifuge (∼1000g) for 60 s, vortexed and analyzed by flow cytometry (FACSort, Becton Dickinson, San Jose, CA, FL1 filter=530 nm and FL3 filter=650 nm). Activation was expressed as the percentage of Fura red–and Fluo-4–stained cells in a high FL1, low FL3 gate. Kinetic analysis of peak alterations in fluorescence occurred within 10–60 s of cell contact, so calcium flux was measured for 20 s or 5000 events and expressed as the percentage of cells within the high FL1, low FL3 gate or MFI in the FL1 channel.
The efficiency of antigen suspension in media was measured by scintillation counting of [14C]C32 GMM and [14C]C80 GMM suspended in media at 0.2–2 μM, as compared to the direct addition of antigen in organic solvents to all-purpose scintillant (Beckman ReadySafe, Miami, FL). Both antigens were suspended with equivalent efficiency. Typically, 90–110% of expected counts were recovered from aqueous sonicates. Cellular uptake was measured by coincubation of radiolabeled GMMs with 105 APCs in triplicate in 96-well plates in the presence or absence of NaN3 (0.02%) and 2-DOG (50 mM). This was followed by the addition of wash buffer (PBS with 5% fetal bovine serum (Gemini, Woodland, CA), NaN3 and 2-DOG), centrifugation (2000g for 90 s) and two additional washes. Pilot experiments showed that no significant radioactivity was detected in the supernatant of the third wash. The cell pellet was resuspended in 20 μl of wash buffer and added to all-purpose scintillant. Assays were done in triplicate and reported as mean±s.d. for total or metabolism-dependent uptake. The latter was determined by subtracting the mass of antigen taken up in the presence of inhibitors from the amount adherent to cells in the absence of inhibitors.
C80 mycolate was coupled to bodipy (Molecular Probes) and yielded monosubstituted mycolates, as confirmed by mass spectrometric detection of ions that corresponded to the predicted structure of bodipy-mycolate. Human monocyte–derived DCs were incubated with bodipy-mycolate (2 μM/ml, Molecular Probes), then stained with anti–human HLA-DM (MaP. DM1, 0.4 μM/ml, Pharmingen) and analyzed with Zeiss axiovert S100 microscope equipped with a Bio-rad Mrc 1024 confocal imaging system (Zeiss, Thornwood, NY).
Note: Supplementary information is available on the Nature Immunlogy website.
We thank J. Gumperz, J. Hartt, M. Brenner and M. Sugita for advice and reagents; D. Olive for the 4A7 hybridoma cell line; P. Chavrier for the Rab5 antiserum; and I. Wilson and B. Segelke for molecular modeling studies of CD1 proteins. Supported by grants from the American College of Rheumatology Research and Education Foundation, the Human Frontiers Science Program, the Irene Diamond Foundation, NIAMS (ARO1988 to D. B. M.), the NIAID (AI49313, AI45889, AI48933, AI 31044 and AI 38960 to D. B. M., S. A. P. and M. L. T.), the NCI (CA74958 to M. L. T) and the Medical Research Council (49343 and G0000895 to G. S. B.).