|Home | About | Journals | Submit | Contact Us | Français|
CD1d is an MHC class I-like membrane glycoprotein that presents lipid antigens to NKT cells. Despite intensive biochemical, genetic and structural studies, the endogenous lipids associated with CD1d remain poorly defined because of the biochemical challenges posed by their hydrophobic nature. Here we report the generation of a protease-cleavable CD1d variant with a similar trafficking pattern to wild-type CD1d that can be purified in the absence of detergent and allows the characterization of the naturally associated lipids. In addition, we used soluble variants of CD1d that are secreted or retained in the endoplasmic reticulum (ER) to survey their acquired lipids. By using multiple mass spectrometry methods, we found that CD1d retained in the ER is predominantly loaded with the most abundant phospholipid in the cell, phosphatidyl choline, while the protease cleavable version of CD1d contains bound sphingomyelin and lysophospholipids in addition to phosphatidyl choline. The secreted soluble version of CD1d, on the other hand, lacks detectable phosphatidyl choline and the only detectable associated lipid is sphingomyelin. The data suggest that, in the absence of infection or stress, CD1d molecules survey the ER, the secretory pathway, and the endocytic pathway, and accumulate the most abundantly available lipids present in these compartments.
CD1 molecules are structurally similar to MHC class I molecules, consisting of a 43–49 kDa transmembrane glycoprotein associated with β2-microglobulin (β2m) ((reviewed in (1)). There are five genes encoding CD1 proteins in humans, called CD1a, b, c, d, and e. CD1a, b, c and e are more homologous to each other than they are to CD1d, and are referred to as Group 1 CD1 molecules. CD1d is the sole member of Group 2, and is the only CD1 species found in mice and rats.
Unlike MHC class I and class II glycoproteins, which bind to short peptides, CD1 molecules associate with lipids. The structural elements that comprise the peptide binding sites of MHC glycoproteins and the lipid binding site of CD1 molecules are similar, consisting of a pair of antiparallel α-helices overlaying an 8 strand β-sheet, but in the case of CD1 the site is highly hydrophobic and can accommodate the convoluted hydrocarbon chains of a wide variety of lipids (2–5). Similar to the way peptide antigens can be recognized by MHC class I and class II-restricted T cells, antigenic lipids associated with CD1 molecules can be recognized by effector T cells (6, 7). CD1d molecules also present autologous lipids to a subset of T lymphocytes, called Natural Killer T (NKT) cells, that play a major role in initiating adaptive immune responses (8). The element of the associated lipid that confers specificity to the T cell-CD1 interaction is generally, but not always, the hydrophilic head group.
There are two sites in the cell where CD1 molecules are believed to capture lipids. The first is the endoplasmic reticulum (ER), where lipid binding is facilitated by the Microsomal Triglyceride lipid transfer Protein (MTP). Data in support of this has been reported for CD1d in both mice and humans, and for CD1b in humans (8–10). The second is the endocytic pathway. CD1 molecules, with the exception of CD1a, are equipped with tyrosine-based targeting motifs in their cytoplasmic domains that induce their internalization and subsequent recycling through late endosomes or lysosomes, and here lipid binding is also catalyzed by lipid transfer proteins, predominantly the saposins (11–14). Four saposins (A, B, C and D) are generated from a precursor called prosaposin by lysosomal proteolysis, and they facilitate the degradation of a variety of glycolipids by removing or ‘lifting’ them from the membrane, making them susceptible to the action of lysosomal glycosidases (15). Saposin B has been shown to be the dominant saposin that mediates lipid binding to human CD1d molecules (12), and saposin C functions in CD1b loading (14). Other lysosomal lipid transfer proteins, such as GM2 activator protein (GM2A) (13) and Niemann-Pick Type C2 (NPC2) (16) have also been implicated in lipid binding to mouse CD1d. A role for CD1e has also been suggested in facilitating lipid binding by the other CD1 species in lysosomes (17).
Our understanding of the mechanisms that generate peptides bound by MHC class I and class II molecules has been furthered considerably by analysis of the pools of associated peptides in normal or cultured cell lines. For example, MHC class I-associated peptides were shown to derive mostly from cytosolic proteins, while those associated with MHC class II were predominantly from the extracellular domains of plasma membrane proteins, exogenous soluble proteins, and even lysosomal proteins (18, 19). This information generalized the principles governing antigen processing pathways that arose from studies of the processing of specific protein antigens for T cell recognition. In the case of CD1 molecules, while considerable effort has been expended using a variety of antigenic lipids to investigate the mechanisms of lipid loading, there is little information on the normal profile of endogenous lipids associated with CD1 molecules that might similarly enhance our understanding of lipid loading principles. There is a technical difficulty in identifying such lipids in that detergents must be used to solubilize transmembrane proteins for purification, and detergents can induce lipid dissociation from CD1 molecules. To overcome this problem, a soluble, secreted version of mouse CD1d was produced that could be purified in the absence of detergents, and glycosylphosphatidylinositol (GPI) was found to be the major bound lipid (20). Using the same approach, soluble versions of human CD1d and CD1b proteins were made and the lipids associated with them were analyzed. Surprisingly, a single cellular phospholipid, phosphatidyl inositol (PI), was found to be associated with both CD1d and CD1b proteins (17). The same lipid was found to be associated with soluble CD1d molecules retained in the ER by the addition of a KDEL sequence to its C-terminus (17). However, an intrinsic problem with using soluble CD1 molecules is that they do not have access to the endocytic pathway.
To allow analysis of the lipids associated with CD1 molecules that recycle through the endocytic pathway we generated a version of human CD1d that could be proteolytically cleaved to produce a soluble molecule. Here we present mass spectrometric (MS) analyses of the lipids bound to the recycling form of CD1d as well as a re-analysis of lipids purified from soluble secreted CD1d and soluble ER-retained CD1d. The data show that all three versions are predominantly associated with the dominant lipids present or produced in the compartments to which they have access.
The .221.ERCD1d and .221.secCD1d cell lines have been described previously (17). To construct a cDNA encoding protease-cleavable CD1d (pclCD1d), an oligonucleotide encoding 12 amino acids (QLYQGKKPDVSP) derived from the proGILT sequence was inserted between the luminal and transmembrane domains of CD1d by PCR, and the cDNA was cloned into the pIRESneo vector (Clontech). The.221.pclCD1d cell line was generated by electroporation as previously described (21). The mouse mAbs to human CD1d, CD1d51 (51.1.3) (22) and D5 (22), were gifts from Dr. Steven Porcelli (Albert Einstein College of Medicine) and Dr. Steven Balk (Harvard Medical School), respectively. The rabbit anti-calnexin antibody was a gift from Dr. Ari Helenius (Zurich). Alexa 488-conjugated goat anti-rabbit IgG secondary antibody and Alexa 594-conjugated goat anti-mouse IgG secondary antibody were from Invitrogen, and FITC conjugated mouse anti-Lamp1 mAb was from BD Biosciences. Rabbit anti-ERp57 serum and the anti-DM mAb K589 were described previously (21, 23).
Purification of ERCD1d and secCD1d was performed as described previously (17). Briefly, for secCD1d, cell culture supernatants were filtered before affinity purification. For ERCD1d, .221.ERCD1d cells (1010) were frozen and thawed on ice in 10 mM Tris.HCl, pH7.4 containing phenyl methyl sulfonyl fluoride (PMSF, 0.5 mM) and iodoacetamide (5mM). After centrifuging at 1000g for 10 minutes to remove nuclei and debris, the supernatants were sonicated briefly and centrifuged at 100,000g for 1 hr at 4°C to pellet the membranes. The supernatants were then used for affinity purification using the anti-CD1d mAb CD1d51 coupled to Biogel A15m (BioRad), eluting with 3.5M MgCl2. For purifying papain-cleaved pclCD1d and HLA.A2 proteins, .221.pclCD1d or .221.A2 cells (1010) were frozen and thawed as described above and pelleted at 1000g for 10 min to remove nuclei. After resuspending in 0.15M NaCl containing 2 mM cysteine, the membranes were digested with 8 units/ml papain (ICN) at 37°C for 5 minutes before adding iodoacetamide to 10 mM to stop the reaction. Insoluble material was removed by centrifugation at 100,000g for 90 minutes. The supernatant was loaded onto a CD1d51 affinity column or w6/32 affinity column for CD1d and HLA.A2 purification, respectively. The specific columns were washed with TBS (10 mM Tris, pH 7.4, 150 mM NaCl) and eluted with 3.5 M MgCl2. The fractions containing protein were pooled and dialyzed against 50 mM ammonium acetate. Protein purity was ascertained by SDS-PAGE and Coomassie Blue and silver staining.
Total lipids were extracted by a modified Bligh and Dyer method (24). Briefly, purified CD1d protein was mixed with 3.75 volumes of chloroform-methanol (1:2) and vortexed three times for 1 minute each. The extract was kept at room temperature for 1 hour before 1.25 volumes each of chloroform and aqueous 0.88% KCl were added. The mixture was again vortexed three times for 1 minute each time and then centrifuged at 2500 rpm for 5 min to separate the organic and aqueous phases. The lower organic phase was retrieved and dried under a stream of nitrogen gas and the dried lipids were redissolved in chloroform-methanol (1:1). Linked-scan tandem quadrupole mass spectrometry was performed for detection of phospholipids by the methods described by (25), using a Waters Quattro-II triple quadrupole mass spectrometer (Waters) in the nanospray mode with fused silica nanospray capillaries (New Objectives). Linked scan monitoring of parents of the m/z 241 ion was used in the negative mode to specifically detect PI species. Phosphatidylcholine (PC) and sphingomyelin were detected in the positive ion mode by monitoring parents of the m/z 184 ion. Phosphatidylethanolamine (PE) and phosphatidylserine (PS) were detected in positive mode by monitoring neutral losses of 141D and 185D, respectively. PS and PE were also detected in the negative ion mode by monitoring the neutral losses of 196D and 87D, respectively. The lipid samples were also analyzed by MALDI-TOF MS using the method of Schiller et al (2001) (26) using a Waters MALDI L/R-TOF instrument (Waters Co.) and a LCQ ion trap mass spectrometer (Thermo Finnigan) to non-selectively detect a wide range of lipids that might be present.
Immunofluorescence was performed as previously described (27). CD1d-expressing HeLa cells and HeLa transfectants were overlayed on coverslips and fixed in 3.7% formaldehyde. After permeablization in 0.1% saponin, the cells were stained by indicated antibodies and mounted on slides in Mowiol 4–88 (Calbiochem). The cells were examined using a Zeiss Axioplan2 microscope or a Leica S2 confocal microscope and images were collected by Openlab (Improvision, England) and Leica software.
Percoll gradient separation was performed as previously described (28). Cells were harvested and washed in homogenization buffer (10 mM triethanolamine, 10 mM acetic acid, 1 mM EDTA, 250 mM sucrose, pH 7.4). They were then disrupted with a ball bearing homogenizer and nuclei were removed by centrifugation at 1000 g for 5 min. The postnuclear supernatant was mixed with 1.1 volumes of homogenization buffer and 0.6 volume of Percoll stock (90% Percoll in homogenization buffer, Sigma) and underlaid with 27.6% Nycodenz (Sigma). The gradient mix was centrifuged at 17,500 rpm (Rotor SW41Ti at 4C) for 1 hour and fractions were collected from the top of the gradient.
To identify lipids associated with CD1d that recycles through endocytic pathway, we constructed a CD1d variant (pclCD1d) which has a short peptide of 12 amino acids (QLYQGKKPDVSP) inserted between the extracellular, membrane proximal, α3 domain and the transmembrane domain (Figure 1A). This peptide corresponds to a site in the pro-form of Gamma Interferon-inducible Lysosomal Thioreductase (GILT) that is proteolytically cleaved during the maturation of GILT in lysosomes (29), and was expected to allow the isolation of recycling CD1d without the use of detergents (17). Constructs encoding soluble human CD1d molecules (Figure 1A) have been previously described (17). Secreted secCD1d consists of only the ectodomain associated with β2m. ER-retained ERCD1d contains the ectodomain of human CD1d attached to an ER-retrieval signal, KDEL. Pulse-chase experiments showed that ERCD1d remained sensitive to digestion with endoglycosidase H (Endo H), consistent with ER retention, whereas secCD1d was secreted into the culture supernatant ((17), data not shown). pclCD1d, on the other hand, acquired Endo H-resistance in a similar manner to wild-type CD1d (Figure S4, compare to data in ref. 27). Transient transfection of HeLa cells with the three CD1d constructs followed by immunoflourescence analysis confirmed that they localized to the desired subcellular compartments. While secCD1d had a weak and diffuse intracellular pattern (data not shown), ERCD1d was exclusively in the ER based on its co-localization with the ER resident protein, calnexin (Figure 1B). Like the wild-type CD1d protein (21), pclCD1d was found both on the cell surface and within lysosomes based on co-staining with the lysosomal marker Lamp1 (Figure 1C). This was confirmed by subcellular fractionation using Percoll density gradients of the human MHC class I-negative B cell line .221 transfected with the same construct, detecting the pclCD1d by Western blotting (Figure 1D). The lighter fractions contain broad, heavily glycosylated, CD1d heavy chain bands, consistent with the CD1d molecules having traversed the Golgi apparatus. This indicates that they are associated with the plasma membrane, which co-sediments with ER in Percoll gradiernts. Our original hope was that natural cleavage in lysosomes would yield a soluble CD1d molecule that we could purify. However, though some probable cleaved product was seen in the lysosomal fraction, at the bottom of the gradient, this was not a significant fraction of the total expressed protein.
For purification of pclCD1d, membranes were isolated from .221.pclCD1d cells. pclCD1d was released by papain cleavage and affinity purified using the CD1d51 mAb as described in the Materials and Methods. As a control, papain-cleaved HLA-A2 was purified from .221.A2 cells. Affinity chromatography using the anti-HLA class I mAb w6/32 was used to purify the HLA-A2 molecules. ERCD1d was released from .221.ERCD1d cells by freeze-thaw followed by sonication of post-nuclear supernatants and purified by affinity chromatography (17). secCD1d was purified from the tissue culture supernatants of .221.secCD1d cells, as previously described (17). The purified proteins were subjected to SDS-PAGE to determine their purity, and a Coomassie Blue-stained gel is shown in Figure 2. As expected, secCD1d had a slightly higher Mr than the ERCD1d, consistent with late stage glycosylation steps occurring along the secretory pathway, while the pclCD1d band was smeared in addition to the Mr shift, perhaps consistent with progressive deglycosylation occurring during repeated recycling between the cell surface and late endosomes and lysosomes. Associated β2m was allowed to run off the gels to increase the resolution of the heavy chain bands. A similar SDS-PAGE analysis of purified HLA-A2, shown in Figure S1, revealed a single band corresponding to the cleaved A2 heavy chain.
Bound lipids were extracted from 50 μg of the purified proteins using chloroform/methanol as described in Material and Methods, and subsequently analyzed by MS and linked scan MS/MS. Figure 3 shows the results for pclCD1d. Panel A shows the total profile obtained by MALDI-TOF, while panel B shows the result of linked scan analyses for parental ions of phosphocholine and sphingomyelin with choline daughter ions using a Quattro-II triple quadrupole mass spectrometer. PC with a range of hydrocarbon chain lengths was the dominant lipid detected, although sphingomyelin was also detected at lower levels. Analysis of the choline-containing lipids present in total lipid extracts from .221 cells and the Fetal Bovine Serum (FBS) in the tissue culture medium used to grow the .221.ERCD1d cells indicated that the lipids isolated from pclCD1d were derived from the cells and not the tissue culture medium (Figure S2). Further analysis showed that lysophospholipids were also associated with pclCD1d (Figure 3C). These were lyso PC, lyso PS, and lyso PE. The molecular species resembled the lysophospholipid species from .221 cells rather than those found in FBS, for example in the ratio of 18:1a to 18:2a ions (highlighted in Figure 3C), indicating that they were derived from the .221.pclCD1d cells (compare the highlighted peaks in Figure 3C to the indicated peaks in Figure S2A and S2B). Using specific parental ion scans and the neutral loss scans in both negative and positive modes of electrospray ionization, we did not detect significant amounts of other phospholipids. No significant PC, PE, PS, lysophospholipids or sphingomyelin was detectable when the same amount of papain-cleaved, affinity-purified HLA-A2 was subjected to identical analyses (Figure S1C).
The previous studies with the soluble ER-retained and secreted versions of CD1d indicated that PI was the dominant associated lipid (17). Because we did not detect any PI or PI derivatives associated with the recycling pclCD1d version of CD1d we performed an independent re-analysis of the lipids associated with ERCD1d and secCD1d. Figure 4 shows the results for the lipids associated with ERCD1d. In the new analysis we failed to detect PI. In fact, PC was the dominant phospholipid found, confirmed by a choline daughter ion scan (precursors of m/z 184). In contrast to the pclCD1d-associated lipids no sphingomyelin was detected. As we observed for pclCD1d, the profile of the PC molecular species was more similar to that of .221 cells than that isolated from FBS (Figure S2).
Figure 5 shows an analysis of the lipids associated with secCD1d. In this case MALDI/TOF analysis indicated that the dominant lipid was sphingomyelin (panel A), which was confirmed by a choline daughter ion search (panel B). The dominant sphingomyelin species identified had the same carbohydrogen composition as that from pclCD1d sample (d18:1/24:1). Notably, this is the most abundant sphingomyelin species synthesized by mammalian cells (30). PE was detectable at low levels in association with both ERCD1d and pclCD1d in some analyses (data not shown), but the lysophospholipids were only detectable in pclCD1d preparations.
Given the previous results with ERCD1d and secCD1d (17) we were surprised that in these studies no PI or its derivatives were found to be associated with any of the three preparations of CD1d. This was not due to an inability to detect PI, for example because of low sensitivity of detection or because of the extraction or ionization conditions. We used a variety of approaches, including varying lipid extraction methods, different chromatographic techniques for lipid purification, and multiple instruments for MS, including MALDI/TOF, Quattro II Tandem quadrupole and LCQ ion trap instruments (data not shown). Furthermore, trace amounts (3–5 ng or 4–6 pmol) of plant derived PI (34:3, m/z 833.5) were readily detectable when added as an internal standard to the lipid samples (Figure S3). Our calculations suggest that even if the loading efficiency of CD1d is low (10%), and the extraction efficiency suboptimal (50% recovery), PI should be readily detectable. The identification of PI as a CD1d-associated lipid in the previous work remains unexplained.
The three derivatives of CD1d studied here have differential access to intracellular environments. ERCD1d is restricted to the early secretory pathway by the presence of an ER retention signal at its C-terminus. The KDEL sequence is actually a retrieval motif that binds to a receptor that recovers proteins bearing the motif from the cis-Golgi. secCD1d assembles with β2m in the ER, passes through the Golgi apparatus and Trans Golgi Network (TGN) and is secreted from the cell. Finally, the pclCD1d construct was designed to have access to the endocytic pathway and to behave as much like the native CD1d molecule as possible. It was expected to assemble in the ER, travel through the Golgi apparatus to the plasma membrane, and then to constitutively recycle through the endocytic pathway. The presence of pclCD1d in the endocytic pathway was confirmed by immunofluorescence and subcellular fractionation (Figure 1), and a variety of immunofluorescence and biosynthetic studies confirmed that all three species behaved as expected (Figure 1, (17), and data not shown). Based on this, the lipids bound by the three species should reflect the environments to which they have access. Because of the relative insensitivity of the MS approaches used in these studies, and because the hydrocarbon chains are the dominant feature of lipids governing their binding to CD1 molecules, one might simplistically expect the lipids observed to be the dominant ones present unless specific loading mechanisms dictate otherwise. In the earlier studies (17) the surprising finding was that PI was the dominant lipid found in both secreted and ER-retained species. However, here we have failed to confirm those findings and the simplistic expectations appear to have been met for ERCD1d and pclCD1d. However, the finding that sphingomyelin was the dominant lipid associated with secCD1d was unexpected.
The only readily detectable lipid associated with ER-retained CD1d molecules was PC, although occasional experiments revealed a small amount of PE. PC is the most abundant phospholipid found in eukaryotic cells, and given that sphingomyelin is absent from the ER, it is perhaps to be expected that this is the species associated with ERCD1d. The terminal stages of sphingomyelin synthesis occur in the Golgi apparatus (31), and therefore there are no topological constraints that could prevent the acquisition of sphingomyelin by secCD1d. However, if PC is acquired by all CD1d molecules in the ER prior to their release into the secretory pathway, it would seem highly unlikely that it would be replaced so efficiently during secretion that none would be detected. One possible explanation is that most of the CD1d molecules escaping the ER are devoid of associated lipids, and sphingomyelin is readily bound during transport or from FBS following secretion. Unfortunately, the experimental systems used did not allow us to quantify the various lipid species isolated, and we cannot estimate the fractional occupancy of either the secreted or ER-retained CD1d. It is conceivable that secCD1d assembles and leaves the ER in a predominantly lipid-free form, but that the ER-retained version accumulates PC during the prolonged period it spends recycling between the ER and cis-Golgi: the half-life of sec CD1d is > 20 hours ((17), data not shown). There are no data to indicate how efficiently MTP catalyzes lipid loading in the ER, and in fact one view is that it catalyzes the binding of the antigenic lipid α-galactosylceramide in the endocytic pathway and not in the ER (32), which is entirely consistent with empty CD1d molecules being assembled and secreted. Alternatively, PC binding in the ER may be efficient but lipid exchange factors present in the secretory pathway after the cis-Golgi preferentially exchange sphingomyelin for PC. These two hypotheses are not mutually exclusive; there may be a factor within the Golgi or TGN that preferentially loads empty CD1d molecules with sphingomyelin. We know of no candidates for such a factor, but this possibility certainly needs to be explored. Determining the true CD1d lipid loading efficiency and establishing whether CD1d can leave the ER without a bound endogenous lipid will shed light on whether lipids bound in the ER play a chaperone-like role in CD1d assembly and transport, a role clearly played by endogenous peptides or the invariant chain in the MHC class I and class II systems, respectively. Temporary retention of secCD1d in the ER by a trafficking inhibitor, such as brefeldin A, combined with quantitative MS analysis, might be used to address this question.
Recycling, membrane-associated, pclCD1d molecules are associated with a combination of the species bound to the two soluble variants. Both sphingomyelin and PC were detected, although PC is dominant. Given the lack of PC in the secreted version, it seems unlikely that the PC present in pclCD1d was acquired in the ER. It is perhaps more likely that the lipids associated with this species were accumulated during continuous recycling through the endocytic pathway, although differences in the susceptibility of soluble and membrane-bound CD1d molecules to exchange cannot be excluded. Our earlier in vitro data describing the role of saposins in lipid loading indicated that the binding of a phospholipid, in that case a biotinylated derivative of PE, was catalyzed as efficiently by saposin B as was the binding of a similar derivative of α-galactosylceramide (12). In addition, the published structures of CD1-lipid complexes do not support a major role for lipid head groups in the regulating the specificity of lipid binding (2–5).
The finding most consistent with accumulation of lipids in the endocytic pathway by pclCD1d molecules is that lysophospholipids were found to be associated with them. Lysophospholipids are generated in the lysosome during lipid degradation by the action of phospholipases (33). How stable the interaction might be between CD1d and lysophospholipids, which bear a single fatty acyl chain, is unclear. However, recognition of CD1d-associated lysophospholipids by non-invariant NKT cells has been described, arguing for a potential in vivo function for such an association (34, 35). Notably, no detectable lysolipids were found in association with either the ER-retained or secreted forms of CD1d. The absence of lysophospholipids associated with secCD1d, which was isolated from tissue culture supernatants, is perhaps the strongest evidence that the lysophospholipids were indeed acquired by pclCD1d in the endocytic pathway.
The detection of PC and sphingomyelin in pclCD1d is consistent with their particularly high abundance in mammalian cells. In fact, PC, PE and sphingomyelin are the three most abundant lipid species in mammalian membranes. However, only PC and sphingomyelin were detectable at significant levels. Interestingly, most PC and sphingomyelin species are asymmetrically located in the luminal/extracellular leaflet of the bilayer (31, 36). Although it is not known if a lipid transfer protein might be functioning in the Golgi complex that is involved in loading sphingomyelin, all of the identified lipid transfer proteins, MTP, saposin, GM2 activator and Niemann-Pick C proteins are luminal soluble protein that are likely to preferentially extract lipids from the luminal leaflet.
Several recent studies on the purification and identification of CD1 natural ligands, as well as in vitro lipid loading studies and structural studies have showed that PC is naturally associated with purified mouse CD1d and human CD1b secreted from mammalian or insect cells (3, 5). Different species of PC can be loaded into human CD1d and other group I CD1 molecules and recognized by T cells (37). Sphingomyelin was also isolated from secreted mouse CD1d (3) and can be loaded onto human CD1d and other human group I CD1 molecules in vitro (38). Although we can readily detect various PC species in the ER-retained version of CD1d, it is surprising that sphingomyelin was almost exclusively found in secreted CD1d. Although CD1b and CD1d could acquire different endogenous lipids along secretory pathway, there are substantial difference between our protein purification and mass spectrometry methodologies and that used for human CD1b (5). Furthermore, during the MS/MS analysis of CD1b, which relied on MS of the native protein with the bound ligand, sphingomyelin could be present but less efficiently dissociated from the protein/lipid complex by collisional activation and therefore less readily detected than PC. There was still significant complete protein/lipid complex present after collisional activation of the native protein MS analysis (5) and it would be interesting to investigate whether higher energy collisional activation MS/MS of extracted lipids could identify remaining complexed sphingomyelin. Interestingly, in addition to PC some sphingomyelin could be detected in association with soluble mouse CD1d secreted from insect cells, which lack endogenous sphingomyelin. The sphingomyelin in this situation was almost certainly bound post-secretion from FBS-derived lipids in the tissue culture medium.
In the data reported here the dominance of sphingomyelin and the absence of PC associated with seccD1d makes it unlikely that loading is exclusively occurring in the culture supernatants. The level of PC in FBS is high whereas high levels of sphingomyelin were not readily detectable (Supplemental Figure 2). Unless CD1d molecules preferentially bind sphingomyelin, or specific exchange factors are secreted into the tissue culture supernatant, one would expect that PC would be the dominant lipid associated with secCD1d. However, at the moment is must be regarded as unproven that sphingomyelin actually binds to secCD1d within the secretory pathway.
Earlier studies of the endogenous peptides associated with MHC class I and class II molecules helped define the mechanisms of peptide generation, and the data we have generated here helps to define the cellular compartments where lipid binding to CD1d can occur. The dominance of PC associated with ER-retained CD1d is consistent with it being a major lipid in ER membranes, and the absence of sphingomyelin is consistent with the lack of this lipid in the ER. The association of sphingomyelin with secCD1d supports the idea that binding might occur in the secretory pathway, with the caveat that binding in the tissue culture medium may be at least in part responsible. While the physiological significance of the observations using the soluble CD1d molecules is unclear, the detection of PC, sphingomyelin, and lysophospholipids associated with pclCD1d supports preconceptions derived from earlier studies examining the binding of antigenic lipids by CD1 molecules (1); lipid loading and/or exchange predominantly occurs during recycling through the endocytic pathway.
Figure S1: PC is specifically associated with pclCD1d. .221.A2 membranes were digested with papain, soluble HLA-A2 was affinity purified and subjected to the identical lipid extraction procedures used for pclCD1d. (A). Western blots of .221.A2 membranes before and after papain digestion, the soluble papain digest, the residual membranes following papain digestion, and the affinity purified HLA-A2 molecules. (B). Purified HLA-A2 protein separated by SDS-PAGE and stained by Coomassie blue. (C). Lipids extracted from A2 protein were subjected to ESI linked scan MS/MS analysis for major phospholipids.
Figure S2: PC isolated from ERCD1d and pclCD1d is most likely derived from .221 cells. Total lipids from .221 cells (A) and fetal bovine serum (B) were extracted and analyzed by ESI linked scan MS/MS analysis. The peaks of m/z 522.4 (18:2a lyso PC) and m/z 524.3 (18:1a lyso PC) were highlighted in bold and used for the comparison with CD1d-associated lyso-lipids detected in Figure 3.
Figure S3: PI is readily detectable by the MS procedures used. Purified soybean PI was added to the lipids extracted from purified secCD1d and the lipid mixture was analyzed by ESI linked-scan MS/MS to search for PI species (parental ions of m/z 241 in negative mode).
Figure S4: Biogenesis of pclCD1d molecules. .221.pcl.CD1d cells were pulsed with 35S-methionine for 15 min and chased for various times. Detergent extracts were immunoprecipitated with CD1d51 mAb, stripped in SDS, and re-immunoprecipitated with the D5 mAb followed by elution, Endo H digestion and SDS-PAGE. Acquisition of Endo H resistance is indicated by the arrows.
We thank Nancy Dometios for assistance in preparing the manuscript.
This work was supported by the Howard Hughes Medical Institute and NIH Grant RO1 AI059167 (PC). W.Y. was supported by a postdoctoral fellowship provided by the Cancer Research Institute.