Structural Features of the Acyl Chain Determine Self-phospholipid Antigen Recognition by a CD1d-restricted Invariant NKT (iNKT) Cell*

Little is known about the antigen specificity of CD1d-restricted T cells, except that they frequently recognize CD1d-expressing antigen-presenting cells in the absence of exogenous antigen. We previously demonstrated that the 24.8.A iNKT cell hybridoma was broadly reactive with CD1d-transfected cell lines and recognized the polar lipid fraction of a tumor cell extract. In the present study, the antigen recognized by the 24.8.A iNKT cell hybridoma was purified to homogeneity and identified as palmitoyl-oleoyl-sn-glycero-3-phosphoethanolamine (16:0–18:1 PE). The 24.8.A iNKT cell hybridoma recognized synthetic 16:0-18:1[cis] PE, confirming that this phospholipid is antigenic. Recognition correlated with the degree of unsaturation of the acyl chains. Using a panel of synthetic PEs, the 24.8.A iNKT cell hybridoma was shown to be activated by PEs that contained at least one unsaturated acyl chain. The configuration of the double bonds was important, as the 24.8.A iNKT cell hybridoma recognized unsaturated acyl chains in the cis, but not the trans, configuration. PEs with multiple double bonds were recognized better than those with a single double bond, and increasing acyl chain unsaturation correlated with increased binding of PE to CD1d. These data illustrate the potential importance of the acyl chain structure for phospholipid antigen binding to CD1d.

The CD1 proteins are ␤ 2 -microglobulin-associated antigenpresenting molecules that present lipid antigens to T cells. CD1d-restricted T cells can be divided into two subsets based on their T cell receptors (TCR) 1 : one that uses a diverse TCR repertoire and a second that uses an invariant TCR␣ chain (V␣14-J␣281) paired with a limited number of TCR␤ chains (V␤2, -7, -8). Both subsets can coexpress cell surface markers associated with NK cells, such as the NK1.1 antigen, and hence are referred to as NKT cells. The second subset of NKT cells will be referred to as invariant NKT (iNKT) cells. It is this subset that has been implicated in a variety of pathological processes, and activation of iNKT cells promotes host rejection of tumors, ameliorates autoimmune disease, and enhances immunity to certain infectious diseases (1)(2)(3)(4)(5)(6)(7)(8)(9). In some cases, the beneficial effects of iNKT cells appear to be mediated by other cell types, and this finding has led to the concept that CD1drestricted iNKT cells function as immunoregulatory T cells (10).
In many of these studies, iNKT cells were activated with the synthetic antigen ␣-galactosylceramide (␣GalCer), an antigen originally isolated from marine sponges and rarely detected in mammalian cells or bacteria (11). Most human and mouse iNKT cells recognize ␣GalCer when presented by CD1d, and undergo activation characterized by rapid cytokine production and enhanced cytotoxic activity (11)(12)(13). Although ␣GalCer has been extremely useful in elucidating the biology of iNKT cells, little is known about the naturally occurring antigens that are recognized by iNKT cells. To understand how iNKT cells participate in immunity, the physiological antigens recognized by the iNKT cell TCR require identification. Many of the first CD1d-restricted NKT cells studied were self-reactive, and recognized both human and murine CD1d in the absence of exogenously added antigen (14 -16). Several naturally occurring antigens potentially recognized by iNKT cells have since been identified. These include peptides, phosphatidylinositol (PI) and PI-containing compounds, and other phospholipids (17)(18)(19).
We have previously characterized a panel of CD1d-restricted T cells and T cell hybridomas (both invariant and diverse) that recognize CD1d-transfected cell lines in the absence of exogenously added antigen (14). However, these CD1d-restricted NKT cells are antigen-dependent and some have been shown to recognize endogenous cellular lipid antigens (17). Using an in vitro assay that was developed for the purpose of antigen discovery, the antigen recognized by the 24.8.A iNKT cell hybridoma was partially purified from RMA-S tumor cells and char-acterized as a polar lipid (14,17). We have now purified the antigen to homogeneity and identified the antigen as palmitoyl-oleoyl-sn-glycero-3-phosphoethanolamine (16:0 -18:1 phosphatidylethanolamine (PE)). We show that unsaturation of acyl chains is important for recognition and suggest that this is likely because of its effect on lipid packing and the ability of the lipid tails to load into the CD1d binding groove.
Cells and Cell Lines-The CD1d-restricted 24.8 iNKT cell clone and the 24.8.A iNKT hybridoma, both of which express a TCR encoded by the invariant TCR␣ (V␣14-J␣281) and V␤8.2 genes, have been previously described (14,17). The RMA-S cell line was derived from the RBL-5 T cell lymphoma and has a defective transporter associated with antigen processing-2 (TAP2) gene, which disrupts its MHC class I processing pathway (21). It also expresses low levels of endogenous CD1d and does not express cell surface MHC class II protein.
Fractionation of Lipids-As previously described, total cellular lipid was isolated from RMA-S cells and fractionated into neutral lipids, glycolipids, and phospholipids, by sequential elution of the lipid from a silica column using solvents of increasing polarity (17). The strategy used to purify and identify the antigen was selected based on our initial observation that the relevant antigen was enriched in the phospholipidcontaining methanol fraction (17). The methanol-eluted fraction was separated by two-dimensional thin layer chromatography (TLC) using 500 m silica gel-coated plates (Scientific Adsorbants, Inc, Atlanta, GA). The first dimension was resolved with 60:35:6 chloroform/methanol/water, and the second dimension was run in 40:25:3:6 chloroform/ acetic acid/methanol/water. Areas containing lipids were detected using water mist, scraped, and eluted with 2:1 CHCl 3 /methanol to extract lipids. Certain low abundance lipids were detected with iodine vapor and then extracted in the same way. Eluted lipids were assayed for stimulation of T cells and highly bioactive fractions were used to determine the structure of the antigen by mass spectrometry.
Mass Spectrometric Analysis of the Antigens-Structural analysis was performed on lipid fractions capable of stimulating the 24.8.A iNKT hybridoma. Electrospray ionization (ESI) mass spectrometry experiments were performed on a Quattro II triple quadrupole (Micromass) mass spectrometer to acquire both single stage (MS) and tandem (MS/ MS) data at unit mass resolution. Accurate mass determinations were performed using an IonSpec Ultima Fourier transform ion cyclotron resonance mass spectrometer with a 7T actively shielded superconducting magnet, also using electrospray ionization.
Plate-bound mCD1d Hybridoma Stimulation Assay for T Cell Recognition of Phospholipids-Purification of the mCD1d fusion protein was as previously described (17). The plate-bound mCD1d fusion protein hybridoma stimulation assay was performed as previously described (17), with the exception that protein G-coated plates (Pierce) were used in place of protein A-coated plates. This assay will be referred to hereafter as the plate-bound CD1 assay. Briefly, phospholipid antigens diluted in PBS were added to the wells of a protein G-coated plate, followed by mCD1d fusion protein (0.5-1.0 g/well in PBS). Following incubation of the plates for 24 h at 37°C, the plates were washed with PBS and medium. 24.8.A iNKT hybridoma cells were added to the coated wells at a concentration of 1 ϫ 10 5 cells/well in a total volume of 150 l/well. Assays were performed using 3 replicate wells and repeated a minimum of twice. Controls included wells coated with anti-CD3 (positive) or PBS (negative). In some experiments, anti-CD1d (19G11), anti-MHC I (M1/42.3.9.8.HLK), or anti-MHC II (M5/114.15.2) blocking monoclonal antibody was included at a concentration of 20 g/ml. The coefficient of variation between replicate wells was generally Ͻ10%.
Competitive mCD1d Phospholipid Binding Assay-A competitive phospholipid binding assay was developed to evaluate the ability of phospholipids to compete with biotinylated di-18:0 PE bound to streptavidin-coated plates for binding to mCD1d. The concentrations of biotinylated di-18:0 PE and mCD1d used in this assay were selected to fall on the linear portion of their binding curve. Immulon-2 (Dynatech, Chantilly, VA) plates were coated for ϳ16 h at 4°C with 5 g/ml streptavidin (ICN, Irvine, CA) in 0.015 M sodium carbonate buffer, pH 9.6. The coated plates were washed 3 times with PBS and then blocked with PBS containing 10% heat-inactivated fetal bovine serum (Invitrogen, Burlington, ON) and 0.05% Tween 20 for 1 h at 25°C. Following three washes with PBS, 30 l of biotinylated di-18:0 PE (Northern Lipids, Vancouver, BC), diluted to 50 g/ml in 0.01 M Tris-buffered saline, pH 7.4 (TBS), was added to the blocked steptavidin-coated wells and incubated for 1 h at 25°C. Competitive binding samples for the assay were prepared by incubating 30 l of phospholipid competitors, diluted from the Me 2 SO stock solution into TBS containing 0.1% ovalbumin (assay buffer), with 20 l of purified mCD1d (approximate final concentration of 4 g/ml). The samples were incubated in borosilicate glass tubes for 16 h at 37°C. 50 l of the competitor/CD1 solution were added in duplicate to the blocked streptavidin-coated plates, and incubated for 4 h at 25°C. The plates were washed four times with PBS. Bound mCD1d was detected by the addition of alkaline phosphataseconjugated human-absorbed goat anti-mouse IgG (Southern Biotechnology Associates, Birmingham, AL), diluted at 1:1000 in PBS containing 0.4% bovine serum albumin, and incubation for 16 h at 4°C. A colored reaction was developed by adding p-nitrophenol phosphate substrate (Sigma) and developing for ϳ20 min at 37°C. The OD 405 was read in an ELISA reader (Titertek Multiskan MCC/340, Flow Systems, Huntsville, AL). The coefficient of variation between replicate wells was generally Ͻ10%.

Purification and Identification of Palmitoyl-oleoyl PE as the RMA-S Antigen Recognized by the 24.8.A iNKT Cell Hybridoma-Previous
studies had shown that the iNKT cell hybridoma 24.8.A was most potently stimulated by the phospholipidenriched methanol fraction of RMA-S cellular lipids (17). To determine the identity of the natural antigens, the methanol fraction was further fractionated by two-dimensional TLC, which yielded 20 distinct lipid species (data not shown). Ten spots were visualized based on the exclusion of a fine water mist (spots 1-10), and a further ten spots were visualized using iodine vapor (spots [11][12][13][14][15][16][17][18][19][20]. The silica corresponding to each spot was scraped and the lipids eluted with CHCl 3 /methanol. The ability of each of these fractions to stimulate the 24.8.A NKT cell hybridoma was tested using the plate-bound CD1 assay. One of the fractions (fraction 7) strongly stimulated the 24.8.A iNKT cell hybridoma to produce IL-2 in the plate-bound CD1d assay, similar to the original methanol fraction (Fig. 1). Plate-bound CD1d alone, in the presence of PBS, showed little or no reactivity.
To determine the identity of the reactive lipid, fraction 7 was analyzed by negative ion mass spectrometry. It contained a single isotopic cluster with the nominal mass per unit charge of m/z 716. The accurate mass m/z 716.5233 was determined (calc. m/z 716.5236 for the C 39 H 75 NO 8 P Ϫ anion). Deuterium exchange experiments indicated that two exchangeable protons were present (data not shown). Both of these findings are consistent with the structure of palmitoyl-oleoyl PE. Further  (22). Comparison of the MS/MS data to data in the literature for the two possible isomers, where carboxylate anion were preferentially generated from the sn-2 position during low energy CID of PEs, indicates that the 18:1 acyl group is located at the sn-2 position, although a mixture with lower levels of the other isomer cannot be ruled out (22). 1-Palmitoyl-2-oleoyl-snglycero-3-phosphoethanolamine (16:0 -18:1[cis] PE) is postulated as the most probable structure (Fig. 2c).
To confirm that 16:0 -18:1[cis] PE was an antigen recognized by the 24.8.A iNKT cell hybridoma, synthetic 16:0 -18:1[cis] PE was obtained and evaluated in the plate-bound CD1 assay. Synthetic 16:0 -18:1[cis] PE was recognized by the 24.8.A iNKT cell hybridoma and its recognition was completely blocked by antibodies specific for CD1d (Fig. 3). Antibodies specific for MHC I or MHC II did not block antigen recognition. These data indicate that PE is an endogenous self-lipid antigen that can be presented by CD1d to the self-reactive 24.8.A iNKT cell hybridoma.
The 24.8.A iNKT Cell Hybridoma Recognizes Microbial, Mammalian, and Plant PEs-Once it was confirmed that PE was an endogenous antigen recognized by the 24.8.A iNKT cell hybridoma, we systematically examined reactivity to naturally occurring PEs. These included PE from the bacterium, Escherichia coli. We were particularly interested in bacterial PE, as many species of bacteria modify the acyl chains of their phospholipids through cyclopropanation and cis 3 trans isomerization. These modifications are not commonly reported in eukaryotic cells and can affect recognition of lipid antigens by CD1-restricted iNKT cells (23,24). Plant PE was recognized most strongly, followed by bovine brain PE and egg PE, heart PE, and liver PE (Fig. 4). PE from E. coli was recognized by the 24.8.A iNKT cell hybridoma, but not better than PEs from plant or animal sources. The degree of recognition of E. coli PE was similar to that of liver PE. In general, we found that most sources of natural PEs were well recognized, and the degree of recognition correlated with the percentage of 18:1 ϩ 18:2 ϩ 18:3 acyl chains for the PEs for which this information is known (Fig. 4). For example, plant PE was the most potent activator of the 24.8.A iNKT cell hybridoma and has 73.2% of its acyl chains (18:1, 18:2, and 18:3) unsaturated. In contrast, liver PE was the least potent activator and contained only 14.7% unsaturated 18-carbon acyl chains. This observation led us to more carefully examine the role of the acyl chains in recognition of antigen by the 24.8.A hybridoma using a panel of synthetic PE.
Fatty Acyl Chain Unsaturation Is Critical for Recognition of PE-The paradigm for binding of lipid antigens to CD1 is that the acyl chains are buried deep in the hydrophobic cleft formed by the CD1 ␣1 and ␣3 domains, leaving the polar head group exposed on the surface of the molecule, a model confirmed by the crystal structure of murine CD1d, human CD1a, and human CD1b (25)(26)(27)(28)(29). Furthermore, alteration of the polar head group of the lipid antigen or the polar amino acid residues in the TCR␤ CDR3 region both have a dramatic effect on TCRdependent recognition of the antigen (24,30). However, changes in the structure of acyl chains can also affect T cell recognition and lipid antigen processing (31,32). The platebound CD1d assay detects the recognition of PE by the 24.8.A T iNKT cell hybridoma independently of antigen processing. As the lipid antigen is not processed and recognition depends only on the direct binding of the phospholipid to the CD1d fusion protein, we were able to determine the effect of different acyl chains on recognition of PE by the 24.8.A iNKT cell hybridoma.
To determine the effect of length and saturation of the fatty acyl chains on recognition of PE by the 24.8.A iNKT cell hybridoma, we first examined a series of synthetic saturated 1,2diacyl-sn-glycero-3-phosphoethanolamines (Fig. 5b). Although the chain length varied between 6 and 18 carbons, none of the saturated diacyl PEs was recognized, with the exception of di-12:0 PE. As the PE purified from RMA-S cells (i.e. 16:0 -18:1 PE) contained one monounsaturated acyl chain, we next examined whether PEs containing unsaturated acyl chains were better recognized than those containing saturated acyl chains. We first evaluated PEs containing acyl chains with a single double bond. Both di-16:1[cis] PE and di-18:1[cis] PE were recognized by the 24.8.A iNKT cell hybridoma, while PEs with saturated acyl chains of the same length (di-16:0 and di-18:0, respectively) were not (Fig. 5c). Furthermore, the introduction of a second double bond conferred greater recognition upon the PE, as shown by di-18:2[cis] PE (Fig. 5d) (Fig. 5e).
Interestingly, the stereochemistry of the double bond had a great impact upon recognition by the 24.8.A iNKT cell hybridoma. We evaluated two different isomers of di  7-12°C, respectively, while the T m and T h of di-18:1[trans] PE are 38°C and 60 -63°C (33). Consequently, at 37°C, the cis isomer of di-18:1 PE will be in the hexagonal phase, while the trans isomer will likely be in the gel phase.
In contrast with unsaturated PEs, PEs containing long saturated acyl chains, such as di-14:0 PE, di-16:0 PE, di-18:0 PE, and di-20:0 PE, have relatively high transition temperatures (T m ϭ 50 -80°C and T h Ն 100°C, in water), making it probable that they will all be in the gel phase in our assay buffer at 37°C. However, di-12:0 PE, with shorter saturated acyl chains, has a T m of 17°C and T h of 43-44°C in aqueous solution. Therefore, it would certainly be in the liquid crystalline phase and may even adopt the hexagonal (H II ) phase in our system, as the presence of salt or Me 2 SO can under certain circumstances decrease transition temperatures. This may explain why di-12:0 PE was reactive, while the other saturated PEs (with acyl chains Ͼ12 carbons long) were not recognized by the 24.8.A iNKT cell hybridoma. The lack of recognition of PEs with saturated acyl chains Ͻ12 carbons long (e.g. di-6:0, di-8:0, and di-10:0) (Fig. 5), which would be expected to be in the liquid crystalline phase at 37°C, suggests that acyl chains of Ͼ10 carbons are required for recognition of saturated diacyl PEs.
Fatty Acyl Chain Unsaturation Is Critical for Binding of the PE to mCD1d-To address whether phospholipid recognition by the 24.8.A iNKT cell hybridoma reflected differences in the binding of lipids to mCD1d, di-18:n PEs of increasing unsaturation were evaluated in a competitive binding assay. We assessed the ability of each phospholipid to competitively inhibit the binding of mCD1d to biotinylated di-18:0 PE bound to streptavidin-coated wells. As shown in Fig. 6

DISCUSSION
Little if anything is known about the natural antigens recognized by CD1d-restricted iNKT cells. We have previously characterized the 24.8.A iNKT cell hybridoma and shown that it recognizes a polar lipid of RMA-S cells (17). In the present study, we have purified the antigen to homogeneity and identified it to be palmitoyl-oleoyl PE (16:0 -18:1 PE) by mass spectrometry. Synthetic 1-palmitoyl-2-oleoyl-sn-glycero-3phosphoethanolamine (16:0 -18:1[cis] PE) was recognized by the 24.8.A iNKT cell hybridoma in a plate-bound CD1d assay, and this recognition was specifically blocked by anti-CD1d monoclonal antibodies. The 24.8.A iNKT cell hybridoma also recognized natural PEs of other cellular origins, including E. coli, mammalian tissues, and plant (soybean). The strength of recognition correlated with the degree of unsaturation of the PEs, with plant PE being the most potent activator, but all of the natural PEs were recognized by the 24.8.A iNKT cell hybridoma.
To better understand how the structure of the acyl chains can influence recognition of PE, a panel of synthetic PEs was analyzed. We focused primarily on synthetic PEs that contain 16-and 18-carbon acyl chains, since these were the chains found in the native antigen and are more commonly represented in mammalian phospholipids. We determined that di To demonstrate definitively that the 24.8.A iNKT cell hybridoma shows preferential recognition of unsaturated, but not saturated, acyl chains, we then evaluated two series of synthetic PEs: 1) PEs with saturated acyl chains of different lengths (from 6 to 18 carbons) and 2) PEs with increasing degrees of unsaturation in acyl chains of one length (18 carbons). The results were striking. Of the seven saturated PEs tested, only one (di-12:0 PE) was recognized by the 24.   6. Binding specificity of mCD1d for unsaturated PE. Phospholipids (PL), or the CHCl 3 control, in Me 2 SO were diluted serially in assay buffer and incubated with mCD1d. The samples were then added to biotinylated di-18:0 PE bound to streptavidin-coated plates and mCD1d binding was detected with an alkaline phosphatase-conjugated goat anti-mouse IgG. All phospholipids shown are synthetic forms of PE. The data represent the mean OD 405 values of duplicates and are representative of two independent experiments. the phospholipid. It is noteworthy that there is an association between transition temperature (T m ) and the ability of PEs to activate the 24.8.A iNKT cell hybridoma. Our findings indicate that PEs likely to be in the liquid crystalline phase at 37°C are recognized by the 24.8.A iNKT cell hybridoma, while those in the gel phase at this temperature are not recognized. The adoption of the liquid crystalline phase (either lamellar or hexagonal [H II ]) and greater acyl chain fluidity may favor the interaction of the acyl chains with the CD1d protein, facilitating the loading of the lipid into the antigen-binding pocket. Similarly, unsaturation in the cis configuration results in looser packing of acyl chains, while the trans configuration exhibits tighter packing of acyl chains. Most eukaryotic membrane phospholipids bear unsaturations in the cis configuration, and the membranes are in the liquid crystalline (as opposed to the gel) phase at physiological temperatures. However, some bacteria produce trans isomers when subjected to certain environmental stress (e.g. heat); these isomers exhibit tighter packing of the acyl chains and higher transition temperatures (23).
To ascertain whether T cell recognition of unsaturated PE was due to better binding of unsaturated forms of PE to mCD1d, we determined the ability of unsaturated and saturated di-18:n PEs to competitively inhibit the binding of biotinylated di-18:0 PE to mCD1d. Fatty acid unsaturation was found to have an important effect on this interaction, and the ability of the PE to inhibit the binding increased with increasing fatty acyl chain unsaturation ( PE, and suggests that binding of the phospholipid to mCD1d may not be the only factor involved in its recognition by a T cell. Second, while di-18:1[cis] was recognized by the 24.8.A hybridoma, this phospholipid did not inhibit the binding of biotinylated di-18:0 PE to mCD1d. The latter can be explained by the differences between biotinylated and nonbiotinylated PEs. The biotinylated headgroup on the phospholipid is very large and bulky, resulting in different packing of the lipid and interdigitation of the fatty acyl chains, which is not seen in nonbiotinylated phospholipids of similar fatty acyl chain composition (LIPIDAT databases, www.lipidat.chemistry.ohiostate.edu). For this reason, the structure and phase that biotinylated di-18:0 PE assumes is very different from that of nonbiotinylated di-18:0 PE. It appears likely that biotinylated di-18:0 PE binds with greater affinity to mCD1d than nonbiotinylated di-18:0 PE and di-18:1[cis] PE, thereby resulting in no competition by either of these phospholipids in our binding assay. It is noteworthy that Naidenko et al. demonstrated that biotinylated dipalmitoyl (di-16:0) PE binds to CD1d protein, using surface plasmon resonance (35). Although the biotinylated di-16:0 PE had a slower rate of association and dissociation than biotinylated ␣GalCer, its K D (0.2 M) did not differ greatly from that of ␣GalCer (0.34 M). The effect of the biotin moiety on the affinity of di-16:0 PE for CD1d was not addressed in that study. We were also able to demonstrate binding of biotinylated di-16:0 PE to mCD1d (data not shown), despite the inability of nonbiotinylated di-16:0 PE to activate the 24.8.A iNKT cell hybridoma.
De Silva et al. (36) have proposed that cellular phospholipids, such as PI, play a chaperone-like role during the assembly of CD1d molecules in vivo (36). They demonstrated that cellular lipids associate with CD1d early in the secretory pathway and do not require glycosylation, and suggest that other lipids may be exchanged with these natural ligands later within the recycling compartment for presentation on the cell membrane. Interestingly, cellular lipids labeled with [ 3 H]ethanolamine have been found to be selectively bound by soluble murine CD1d1 molecules (18). The acyl chains of the phospholipid portion of glycosylphosphatidylinositol (GPI) isolated from mouse CD1d1 were stearic (18:0) and arachidonic (20:4) acid, respectively (18). Similar to the natural PE ligand isolated in the present study, one acyl chain was saturated (stearic acid) and the other was unsaturated. It is possible that unsaturated and saturated acyl chains differ in their ability to load into the CD1d antigenbinding site. The endogenous lipids of the endoplasmic reticulum (ER) membrane are relatively unsaturated and the isolated PE component (ϳ30% of microsomal phospholipid) prefers the hexagonal (H II ) phase at physiological temperatures (37). Relatively minor perturbations of the ER membrane could result in local nonbilayer (e.g. hexagonal [H II ] phase) structures (34). As the assembly of CD1d1 with cellular phospholipids has recently been shown to occur in the ER (36), unsaturated forms of PE or other lipids may play an, as yet, undefined role in this assembly.
The recognition of ubiquitous phospholipids is consistent with the broad specificity of the 24.8.A iNKT cell hybridoma, which recognizes CD1d-transfected cell lines of diverse origins, as well as murine splenocytes and thymocytes (17). Thus, the 24.8.A hybridoma represents one pole of a spectrum of antigen specificities reported for CD1d-restricted iNKT cells, as it fails to recognize ␣GalCer and is instead specific for phospholipids. iNKT cells that recognize ␣GalCer, but are not autoreactive, represent the other end of the spectrum (17). In between are iNKT cells that recognize ␣GalCer and are autoreactive, and some of these recognize phospholipids, albeit weakly (13, 17, 38 -40). These various patterns of recognition highlight a potential role for the TCR␤ CDR3 region in determining the antigen specificity of iNKT cells (14).
Although the in vivo relevance of the endogenous antigen recognized by the 24.8.A iNKT cell hybridoma remains unclear, it is interesting to speculate that ubiquitous phospholipid antigens may play a role in selecting the CD1d-restricted T cell repertoire. In addition, iNKT cell recognition of weakly crossreactive self-lipid antigens presented by CD1d may be responsible for the activated/memory phenotype observed for iNKT cells (41,42). Although NKT cell activation is likely initiated by a high affinity interaction between the TCR and CD1d/lipid complex, preferential binding of certain acyl chains to CD1d may contribute to the selective displacement of endogenous self-lipids by antigenic lipids as they are loaded in the endosome (43).
Taken together, our data indicate that the effect of unsaturation on T cell recognition of PE is due to differences in the ability of PEs of differing degrees of unsaturation to bind to mCD1d. Increasing fatty acid unsaturation appears to directly correlate with the ability of PE to bind to mCD1d. It is not yet clear whether fatty acid unsaturation plays a similar role in the binding of the other phospholipids to mCD1d, but data from our laboratories suggest that this is likely a feature of other phospholipids. 2 Our data are consistent with the possibility that PEs containing unsaturated acyl chains load more efficiently than PEs with saturated acyl chains, resulting in a higher affinity complex with CD1d. It is not yet clear whether this is related to a particular phase requirement, but it seems virtually certain that the antigenic lipids will be in the fluid phase in vivo. An alternative hypothesis is that unsaturated acyl chains may induce a conformational change in the CD1d protein that is recognized by the 24.8.A TCR. Finally, the third possibility is that phospholipids with unsaturated acyl chains bind in a different manner than saturated phospholipids, (e.g. a different portion of the phospholipid is presented to the TCR). These considerations are important and lack of attention to acyl chain saturation may explain discrepant findings regarding CD1d-restricted T cell recognition of phospholipids in the current literature (17,36,44).