Identification of specific glycoforms of major histocompatibility complex class I heavy chains suggests that class I peptide loading is an adaptation of the quality control pathway involving calreticulin and ERp57.

Glycosylation analysis was used to probe the sequence of events accompanying the binding of antigenic peptides to the major histocompatibility complex class I heavy chains. Free heavy chains were isolated from the beta(2)-microglobulin-negative cell line Daudi and from the B-lymphoblastoid cell line Raji. Heavy chains were also isolated from Raji cells in multimolecular complexes (peptide loading complexes) containing the transporter associated with antigen processing, tapasin and ERp57 with and without the lectin-like folding chaperone, calreticulin. Calreticulin is a soluble protein that recognizes primarily the terminal glucose of Glc(1)Man(7-9)GlcNAc(2) glycans. This paper shows that monoglucosylated glycoforms of heavy chain, which exist transiently in the endoplasmic reticulum in the initial stages of the glycosylation processing pathway, are present in the peptide loading complex. The data are consistent with a model in which the release of peptide-loaded major histocompatibility complex class I molecules from calreticulin, induced by deglucosylation of the heavy chain N-linked glycan, signals the dissociation of the complex. This is consistent with the hypothesis that the class I loading process is an adaptation of the quality control mechanism involving calreticulin and ERp57.

In the endoplasmic reticulum (ER), 1 major histocompatibility complex (MHC) class I molecules bind short peptides that are generated in the cytosol by proteasomal degradation and translocated into the ER by the transporter associated with antigen processing (TAP). The class I-peptide complexes are then transported to the cell surface where they can be screened by CD8-positive T cells and can potentially trigger an immune response (1). The detailed mechanisms that regulate peptide binding to MHC class I molecules are not yet fully established.
Mature MHC class I molecules consist of two subunits: the heavy chain (HC), which is a transmembrane glycoprotein, and a small, soluble, nonglycosylated protein, ␤ 2 -microglobulin (␤ 2 m). A peptide must associate with the peptide-binding groove formed by the ␣1 and ␣2 domains of the HC for proper folding of the class I-␤ 2 m dimer and its subsequent transport to the cell surface. Class I assembly requires multiple coordinated intra-and inter-molecular events to ensure the continuous reporting of cellular contents to cytotoxic T-lymphocytes.
It has been proposed that in the ER of humans, unassembled heavy chains interact initially with the membrane-bound chaperone, calnexin (CNX) (2,3), or with the soluble chaperone BiP (4). Concomitant with ␤ 2 m association the heavy chain is released and binds to the soluble chaperone, calreticulin (CRT) (5). Either CNX or CRT recruits the thiol oxidoreductase ERp57 into the complex (6 -8). This appears to facilitate the formation and maintenance of an intrachain disulfide bond in the class I heavy chain, most likely that which anchors the ␣2 domain ␣-helix to the floor of the peptide-binding groove (9).
The assembled class I-␤ 2 m dimer, together with associated CRT and ERp57, is associated with a larger complex that also contains the TAP heterodimer and the transmembrane glycoprotein tapasin (10). Tapasin provides the bridge that links MHC class I to the TAP heterodimer (5), and a disulfide bond between tapasin and ERp57 is dependent upon the association of class I molecules with the complex (9).
Peptides destined for MHC class I binding are translocated in an ATP-dependent fashion into the ER by the TAP heterodimer. Here they may undergo further N-terminal trimming prior to binding to TAP-associated MHC class I molecules (11). Peptide binding induces dissociation of the class I-␤ 2 m dimer from the tapasin-TAP complex, as well as from CRT and ERp57, allowing glycan maturation and transport of the assembled MHC class I-peptide complex from the ER.
CNX and CRT are both lectin-like molecules that can facilitate the folding of newly synthesized glycoproteins in the ER, in part by binding to their N-linked glycans (12,13). In the early stages of oligosaccharide processing, HCs, in common with all nascent glycoproteins, transiently carry the Glc 1 Man 9 GlcNAc 2 oligosaccharide (Fig. 1). The ␣1,3 arm of this glycan, in particular the terminal Glc residue (G3), provides the ligand that enables the partially folded protein to interact with the lectinbinding sites on both CNX (12) and CRT (13). A quality control cycle has been proposed in which CNX and CRT in turn recruit ERp57, which facilitates the proper formation of disulfide bonds in the nascent glycoprotein (14). Removal of the glucose residue by glucosidase II eliminates CNX or CRT binding. Only * This work was supported in part by the Howard Hughes Medical Institute. The MALDI mass spectrometer was purchased with a grant from the Biotechnology and Biological Sciences Research Council. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ To whom correspondence may be addressed. 1 The abbreviations used are: ER, endoplasmic reticulum; ␤ 2 m, ␤ 2microglobulin; CNX, calnexin; CRT, calreticulin; GU, glucose unit; HC, heavy chain; MALDI, matrix-assisted laser desorption/ionization; MHC, major histocompatibility complex; NP-HPLC, normal phase high performance liquid chromatography; TAP, transporter associated with antigen processing. misfolded, non-native glycoproteins are substrates for the enzyme, UDP-glucosyl transferase, which adds back the terminal glucose residue, allowing the glycoprotein to rebind to CNX or CRT. Correctly folded glycoproteins, which are not reglucosylated, escape further interactions with the lectin-like chaperones and can generally leave the ER after further trimming of the mannose residues (15).
There is no direct evidence for the involvement of the MHC class I N-linked glycan in CNX-mediated folding, and it has been proposed that this interaction may be mediated by the MHC class I polypeptide chain (16 -19). There is also suggestive evidence that the interaction of MHC class I with CNX may be important for the CRT-class I interaction and even for the incorporation of class I molecules into the class I loading complex. However, MHC class I association with CRT and with the peptide loading complex is inhibited in the presence of the glucosidase I inhibitor castanospermine (5). This observation suggests that the class I peptide loading process may be an adaptation of the general quality control mechanism involving CRT and ERp57 that regulates folding of glycoproteins in the ER. In this paper we investigated this hypothesis by determining the glycosylation status of MHC class I heavy chains, both free in the ER and associated with the loading complex, and we also analyzed the glycosylation of TAP-associated tapasin. The results are consistent with a role for the N-linked glycanmediated, CRT-dependent, quality control pathway in modulating class I peptide loading.

EXPERIMENTAL PROCEDURES
Materials-The following monoclonal antibodies were used for the affinity columns: MaP.ERp57 (anti-human ERp57) (20), 148.3 (anti-TAP1) (21), and HC10 (anti-class I HC) (21). Tapasin C-terminal specific rabbit antiserum R.gp48c and anti-HC-terminal peptide specific rabbit antiserum R.A3e7 used in Western blotting were prepared in the Cresswell laboratory (9). Peptide N-glycanase F (EC 3.5.1.52) was obtained from Roche Diagnostics (Lewes, UK), Jack bean ␣-mannosidase (EC 3.2.1.24) was from Glyko Inc. (Upper Heyford, UK), and glucosidase II was prepared in the Glycobiology Institute. The activity of this enzyme was measured by hydrolysis of [ 14 C]glucose-labeled Glc 2 Man 9 GlcNAc 2 substrate and was found to release 5.7% min Ϫ1 of labeled glucose (22). All of the other reagents used throughout these experiments were of the highest quality available.
Sample Preparation-Raji wild type B cells or Daudi ␤ 2 m-negative B cells (10 10 ) were lysed in 200 ml of 1% digitonin in 0.15 M NaCl, 0.01 M Tris, pH 7.4 (Tris-buffered saline) containing protease inhibitors as previously described (10) and passed through one of three affinity columns consisting of MaP.ERp57 (␣-ERp57), 148.3 (␣-TAP), or HC10 (␣-HC) coupled to BioGel A15m beads (Bio-Rad) to produce ERp57associated proteins, TAP-associated proteins, or free class I HC, respectively. After extensive washing with 0.1% digitonin in Tris-buffered saline, the proteins were eluted in 1% octylglucoside in 3.5 M MgCl 2 and dialyzed against 0.1% octylglucoside in phosphate-buffered saline. During dialysis some protein precipitated that was removed by centrifugation. These preparations produced the following samples: (i) from the ␣-ERp57 affinity column: ERp57/CRT, ERp57-associated tapasin and ERp57-associated class I HC in solution (gel A), and ERp57-associated class I HC as a precipitate (gel B); (ii) from the ␣-TAP affinity column: ERp57/CRT, TAP-associated tapasin and TAP-associated class I HC in solution (gel C), and TAP-associated class I HC as a precipitate (gel D); (iii) from the ␣-HC affinity column: free class l HC from Raji cells (gel E) and Daudi cells (gel F); and (iv) from the ␣-TAP affinity column: tapasin from Daudi cells (gel G) (Fig. 2). Precipitated samples were dissolved with sonication in 2-10% SDS.
SDS-PAGE Gels-Samples of each of the proteins were run on 10% SDS-PAGE gels in a vertical mini-gel system, 80 ϫ 80 ϫ 0.75 mm at 500 V and 25 mA/gel. The gels were prepared according to Kuster et al. (23), and 20 l of sample and 5 l of 5ϫ Laemmli sample buffer were used per well. Reduction and alkylation of the proteins were carried out prior to running the gels. Dithiothreitol (10 mM, 0.5 l) was added, and each sample was incubated at 70°C for 10 min. Iodoacetamide (100 mM) was added to a final concentration of 10 mM, and the samples were incubated for 30 min at room temperature in the dark. Up to 8 wells of each sample were run to provide sufficient glycans for NP-HPLC analysis, enzyme digestions, and mass spectrometry. The protein was visualized in the gel by staining with Coomassie Blue for at least 2 h, destaining with 50% methanol/7% acetic acid for a few minutes and then overnight with 5% methanol/7% acetic acid. The bands were excised, cut into pieces of ϳ1 mm 3 , and frozen at Ϫ20°C for at least 2 h. The gel pieces were washed with 2ϫ 300 l of 20 mM NaHCO 3 , pH 7.0, for 30 min and then with 300 l of 1:1 acetonitrile:20 mM NaHCO 3 for 60 min to remove residual SDS (23). The gel pieces were dried in a vacuum centrifuge.
In Situ Digestion of N-Glycans with Peptide N-Glycanase F and Glycan Extraction-The volumes given are for 10 -15 mm 3 of SDS-PAGE gel band containing the protein of interest. 30 l of peptide N-glycanase F (100 units/ml) were added to the gel pieces and left for about 5 min for the gel to re-swell, and then the gel was covered with 20 mM NaHCO 3 . The sample was incubated at 37°C for 16 h. After incubation, the sample was centrifuged at 1,100 ϫ g for 5 min. The supernatant was retained, and the glycans were extracted from the gel using three changes of 200 l of sub-boiling point distilled water and sonicated for 30 min, followed by extractions with 200 l of acetonitrile, 200 l of water, and 200 l of acetonitrile. The extracts and the above supernatant were combined. 30 l of activated anion exchange resin, AG-50 X12 (H ϩ ) slurry were added to the extracts and incubated for 5 min at room temperature to desalt. It was centrifuged for 5 min at 550 ϫ g, and the supernatant was filtered with a 0.45-m Millex-LH/ hydrophilic polytetrafluoroethylene filter attached to a 2.5-ml syringe. The supernatant was partially dried in a vacuum centrifuge, transferred to a 0.5-ml microcentrifuge tube, and dried completely.
Fluorescent Labeling and NP-HPLC-The samples were labeled with 2-aminobenzamide by reductive amination according to Bigge et al. (24) and processed through NP-HPLC using the low salt buffer system as described by Guile et al. (25). This uses a 4.6 ϫ 250-mm GlycoSep-N column (Glyko) with a gradient of 20 -58% solvent A (solvent A, 50 mM formic acid adjusted to pH 4.4 with ammonia; solvent B, acetonitrile). Each 2-aminobenzamide-labeled sample was dissolved in 100 l of sub-boiling point distilled water, and 20 l was run on the NP-HPLC with 80 l of acetonitrile.
The system was calibrated using an external standard of hydrolyzed and 2-aminobenzamide-labeled glucose oligomers to create a dextran ladder (25). The number of glucose residues in each dextran peak was plotted against the retention times of the peaks to obtain a standard curve. The retention times for the individual glycans were converted to glucose units (GU) using this curve. The retention times were then compared with a data base of experimental values. The higher the GU value the larger the glycan it represents. The incremental GU values are measures of the affinity of each glycan for the column matrix and are related to the hydrophilicity of each monosaccharide residue in the chain. It also depends on how much of the hydrophilic surface is exposed to the column matrix. In the case of GlcNAc, the N-acetyl side chain hinders some of the interactions which are available to glucose; therefore the incremental value for GlcNAc is about 50% of that of glucose (25). It is this feature that gives the fine specificity of the columns, allowing the distinction between arm-specific isomers. The area of each ERp57 (A and B) and the ␣-TAP (C and D) affinity columns. The bands containing ERp57/CRT, HC, and tapasin are indicated. The NP-HPLC chromatograms (A HC , B HC , C HC , and D HC ) are of the glycans released from the CRT associated HCs in gels A, B, C, and D, respectively. Each of these glycan profiles contains predominantly Man 9 GlcNAc 2 (peak 4) and Glc 1 Man 9 GlcNAc 2 (peak 5).

FIG. 4. SDS-PAGE gels of complexes eluted from the ␣-HC affinity column from Raji cells (E) and Daudi cells (F).
The NP-HPLC chromatograms (E HC and F HC ) are of the glycans released from free HC, which were not associated with CRT, in gels E and F, respectively. The glycan profiles of E and F contain no Glc 1 Man 9 GlcNAc 2 but more of the smaller glycans, Man 6 GlcNAc 2 and Man 7 GlcNAc 2 .
FIG. 5. SDS-PAGE gels of Raji cell complexes eluted from the ␣-ERp57 affinity column (A) and from the ␣-TAP affinity column (C) and of a Daudi cell complex also eluted from the ␣-TAP affinity column (G). The NP-HPLC chromatograms (A TPN , C TPN , and G TPN ) are of the glycans released from the tapasin bands in gels A, C, and G, respectively. The tapasin bands contain mainly Man 9 GlcNAc 2 and a small peak of Glc 1 Man 9 GlcNAc 2 . of the glycan peaks was measured and expressed as a percentage of the total glycans of each sample. Comparing the percentage areas of intact and digested glycan pools helps to identify glycans that co-elute.
Enzyme Digestions-Approximately 100 -300 fmol of each glycan pool were dried, and 10 l of 50 units/ml Jack bean ␣-mannosidase in 100 mM sodium acetate with 2 mM Zn 2ϩ , pH 5, were added. The reaction mixture was incubated at 37°C overnight, after which a further 10 l of 100 units/ml enzyme were added, and the sample was incubated for a further 24 h. The sample was then heated at 100°C for 5 min to denature the enzyme and dried in a vacuum centrifuge. It was redissolved in 20 l of sub-boiling point distilled water and analyzed by NP-HPLC. Similarly two additions of 10 l of ␣1,3-glucosidase II (in 80 mM triethylamine, pH 7, 0.1 M sodium chloride, 10% glycerol and sodium azide) were made to 100 -300 fmol of the released glycans. The samples were processed as for the mannosidase digests. The NP-HPLC profiles were analyzed according to Guile et al. (25).
Matrix-assisted Laser Desorption Ionization Time-of-flight Mass Spectrometry-Positive ion Reflectron MALDI time-of-flight mass spectra were recorded with a Micromass TofSpec 2E Reflectron mass spectrometer (Micromass Ltd., Manchester, UK) fitted with delayed extraction and a nitrogen laser (337 nm). The acceleration voltage was 20 kV, the pulse voltage was 3200 V, and the delay for the delayed extraction ion source was 500 ns. The samples were prepared by adding 0.5 l of an aqueous solution of the sample to the matrix solution (0.3 l of a saturated solution of 2,5-dihydroxybenzoic acid in acetonitrile) on the stainless steel target plate, allowing it to dry at room temperature and then recrystallizing it from ethanol.
The protein bands on the SDS-PAGE gels (Figs. [3][4][5] were identified according to their migration positions and by West-ern blotting. The approximate apparent molecular masses were as follows: ERp57/calreticulin, 55 kDa; tapasin, 50 kDa; and HC, 45 kDa. The tapasin bands were confirmed by N-terminal sequencing, which indicated that the Raji cell tapasin ran as a doublet (Figs. 3 and 5). Fig. 6 shows Western blotting performed on a selection of samples. The upper panel shows the reactivity of the samples with the tapasin C-terminal specific rabbit antiserum R.gp48c, and the lower panel shows the reactivity of the samples with the anti-HC-terminal peptidespecific rabbit antiserum R.A3e7. The specificity of both the tapasin-reactive and the HC-reactive antibodies is demonstrated by samples G (tapasin from Daudi cells) and E (free HC purified from Raji cells). Samples A-C all contain tapasin and HC. Samples A and C contain tapasin and HC in solution, and sample B has some tapasin co-precipitating with HC during dialysis.
Glycans were released by incubating the excised gel bands with peptide N-glycanase F, which cleaves the N-glycosidic linkage to asparagine, and the glycan pools labeled with the fluorophore 2-aminobenzamide. The chromatograms of both the HC and the tapasin samples were all found to contain the same sugars in the peaks, labeled from 1 to 5, although in different proportions (Figs. [3][4][5]. Peak a (Figs. 3-5) was not identified. It is present in the samples eluted from the ␣-ERp57 and ␣-HC columns but not in those eluted from the ␣-TAP column (gels C and D), and it is not digested by either mannosidase or glucosidase II and was not seen in the MALDI mass spectrometric analysis (see Table III). As expected, the ERp57/ CRT contained no glycans that could be detected by NP-HPLC because ERp57 and human CRT are not glycosylated (data not shown). In every case, glycan structures were assigned from the GU values of the peaks in the NP-HPLC trace of the intact glycan pools (Table I) combined with data from glucosidase II and Jack bean ␣-mannosidase digests including consideration of peak areas (Table II). The assignments are consistent with MALDI mass spectrometric data (Table III).
Glucosidase II Digestions-The identity of the peaks in the chromatograms in Figs. 3-5 was confirmed with glucosidase II and Jack bean ␣-mannosidase digestions. Glucosidase II cleaves the nonreducing terminal glucose ␣1-3-mannose linkage. Fig. 7 shows the glucosidase II digest of ERp57 associated class I HC (B HC ), in which peak 5 (GU 10.15) is completely digested to Man 9 GlcNAc 2 , peak 4 (GU 9.49) ( Table I). The incremental GU value for one glucose monomer is 0.7, so the digestion of peak 5 to peak 4 with a difference of ϳ0.7 and an increase in peak area of peak 4 from 37.4 to 61.8% confirms the presence of Glc 1 Man 9 GlcNAc 2 in peak 5. Peak 5 was collected and digested with glucosidase II producing a single peak of Man 9 GlcNAc 2 (Fig. 7, inset a), confirming that peak 5 contains only Glc 1 Man 9 GlcNAc 2 . A comparison of the percentage peak areas of the intact glycan pool of the ERp57 associated class I HC (B HC ) with that of its glucosidase II digest shows the  presence of Glc 1 Man 8 GlcNAc 2 (ϳ9% of total glycans), which co-elutes with Man 9 GlcNAc 2 (Table II). Similarly, peak 4 of ERp57-associated class l HC (A HC ) contains ϳ9% Glc 1 Man 8 GlcNAc 2 as well as Man 9 GlcNAc 2 (data not shown). Peak 4 was collected and digested with glucosidase II producing two peaks, Man 8 GlcNAc 2 and Man 9 GlcNAc 2 , derived from 42 and 58% of peak 4, respectively (Fig. 7, inset b), confirming that peak 4 contains both Man 9 GlcNAc 2 and Glc 1 Man 8 GlcNAc 2 . These data show that Glc 1 Man 8 GlcNAc 2 is ϳ16% of the total glycans, which is a more accurate assessment than the data from the total glycan pool where there is overlap of peaks 3a and 3b, which contain two isoforms of Man 8 GlcNAc 2 . None of the tapasin samples show evidence of Glc 1 Man 8 GlcNAc 2 when comparing the peak areas (Table II).
Jack Bean ␣-Mannosidase Digestions-Jack bean ␣-mannosidase is an exoglycosidase that acts on nonreducing terminal mannose ␣1-2,3 and 6 linkages. Fig. 8 shows the mannosidase digest of ERp57-associated class I HC (B HC ). All of the oligomannose peaks were digested to Man 1 GlcNAc 2 , 65.5% of the total digested glycans (peak 6, GU 2.63). Glc 1 Man 4 GlcNAc 2 (29.6%, peak 7, GU 5.97) and Glc 1 Man 5 GlcNAc 2 (4.8%, peak 8, GU 6.88) were also identified as digestion products of Glc 1 Man 9 GlcNAc 2 and Glc 1 Man 8 GlcNAc 2 . Jack bean ␣-mannosidase is sometimes unable to cleave a single Man␣1-6 linked to the core ␤ Man, so peak 8 may be the result of incomplete digestion. Figs. 7 and 8 also show representative cartoons of some of the glycan structures. Fig. 3 shows the chromatograms of all the glycans of the calreticulin-associated heavy chain samples. The heavy chains of ERp57-associated class I MHC in solution (A HC ) and the precipitated ERp57-associated class I MHC (B HC ), the TAP-associated class I MHC in solution (C HC ), and the precipitated TAP-associated class I MHC (D HC ) all show a predominance of Man 9 GlcNAc 2 and Glc 1 Man 8 GlcNAc 2 (peak 4) and Glc 1 Man 9 GlcNAc 2 (peak 5). Man 6 -8 GlcNAc 2 (peaks 1-3) were also identified (Table I). Fig. 4 shows the chromatograms of the glycans of the free HC from Raji and Daudi cells (E HC and F HC , respectively). E HC and F HC show peaks corresponding to glycans 1-4 but do not show peak 5 (Glc 1 Man 9 GlcNAc 2 ). The additional minor peaks on these chromatograms have not been identified; they do not correspond to known elution positions of N-linked glycans, were not seen in the MALDI mass spectrometric analysis, and may be artifacts. Fig. 5 shows the chromatograms of the glycans from the tapasin samples from the ERp57-associated tapasin (A TPN ) and TAP-associated tapasin (B TPN ) together with the Daudi tapasin (G TPN ). They all show a similar pattern of glycosylation with a predominant peak of Man 9 GlcNAc 2 (peak 4) and only a small peak (Ͻ10%) of Glc 1 Man 9 GlcNAc 2 (peak 5). The peak areas for all of the samples are shown in Table IV. DISCUSSION

MHC Class I Assembly May Be an Adaptation of the CNX/ CRT Quality Control Mechanism-The involvement of CNX
and CRT in the assembly of class I molecules suggests that this process may be an adaptation of the glycan-regulated folding and quality control pathway postulated by Helenius and Aebi (13). This pathway governs the folding and assembly of many glycoproteins, retaining them in the ER either until they are correctly folded and transit to the Golgi or until they are targeted for ER-associated degradation.
CRT has been shown to bind free Glc 1 Man 9 GlcNAc 2 oligosaccharides directly (26), and we have determined that the N-linked glycans of class I HCs bound to CRT, purified by virtue of their association with TAP or ERp57, contain a high proportion (ϳ50%) of monoglucosylated glycans. These are Glc 1 Man 9 GlcNAc 2 (ϳ32%; Table IV) and Glc 1 Man 8 GlcNAc 2 (ϳ16%), both of which are potential ligands for CRT. These data, which suggest that class I HCs bind CRT through the monoglucosylated glycans, are consistent with experiments showing that the glucosidase I inhibitor castanospermine reduces MHC class I association with the peptide loading complex (5) and also with mutagenesis data suggesting that CRT might interact directly with the class I HC N-linked glycans (27). Thus, in addition to the previously defined interaction between CRT and ERp57 (14), at least two types of interactions exist within the class I loading complex: a noncovalent, lectinlike interaction between CRT and the class I HC and a covalent disulfide bond between ERp57 (Cys 57 ) and tapasin (Cys 95 ) (9). The combined data suggest that the loading complex, excluding the TAP heterodimer, which probably interacts with the tapasin transmembrane domain (28), is held together by the set of co-operative interactions shown schematically in Fig. 9.
The results of the carbohydrate analysis undertaken here are consistent with the hypothesis that the glucosylated glyco-  forms (Glc 1 Man 7-9 GlcNAc 2 ) of MHC class I bind to CRT. In an adaptation of the CNX/CRT quality control pathway, we propose that the association of a peptide with the binding groove of MHC class I indicates that the folding/assembly process is complete. At this stage in the CRT quality control pathway, the final glucose residue is removed permanently by glucosidase II because the UDP-glucosyl transferase does not reglucosylate fully folded glycoproteins. Therefore the removal of the glucose residue allows the class I-peptide complex to leave the quality control pathway and to dissociate from the other components of the loading complex (Fig. 10). The TAP-and ERp57-associated HCs that are not monoglucosylated (ϳ50% of the total) may correspond to class I-␤ 2 m dimers that have been deglucosylated but have not yet dissociated from the loading complex. At first sight these data appear to be inconsistent with previous studies indicating that the ratio of CRT to HCs in the loading complex was ϳ0.9 to 1 (10). This ratio might be used to argue that all of the HC in the loading complex should contain a monoglucosylated glycan. However, the ratio of CRT to HC was determined prior to the identification of ERp57 in the loading complex. Because CRT and ERp57 are difficult to separate by SDS-PAGE, it is likely that the amount of CRT in the loading complex was overestimated, perhaps by 2-fold if ERp57 and CRT are present in equimolar amounts. Thus, the amounts of monoglucosylated HC and CRT may in fact be quite similar. Whether or not the deglucosylated class I molecules in the loading complex contain bound peptides is an interesting question for future studies. In addition, these data could also be consistent with a glycan-independent interaction between CRT and HC. Other interpretations of these data include the possibility that the Man 9 GlcNAc 2 sugar attached to the HC still retains sufficient binding affinity for CRT for the glycoforms to remain associated with the loading complex during purification. The data would also be consistent with the proposal that unfolded glycoproteins interact with both a lectin site and a nonspecific polypeptide-binding site in CNX and CRT (29 -31).
The sugars attached to HLA class I expressed in Epstein- FIG. 7. NP-HPLC chromatogram of the glycans released from the SDS gel band containing HC (B HC ) eluted in a complex from Raji cells that bound to the ␣-ERp57 affinity column (see Fig.  3). The lower chromatogram shows the glucosidase II digest of the same sample. Glc 1 Man 8 GlcNAc 2 (of peak 4) and Glc 1 Man 9 GlcNAc 2 (peak 5) digest to Man 8 GlcNAc 2 and Man 9 GlcNAc 2 , respectively. Inset a is of an individual fraction of Glc 1 Man 9 GlcNAc 2 (peak 5) and the product of glucosidase II digestion, a single peak of Man 9 GlcNAc 2 . Inset b is of an individual fraction of Man 9 GlcNAc 2 and Glc 1 Man 8 GlcNAc 2 (peak 4) and the product of glucosidase II digestion, Man 8 3). The lower chromatogram shows the Jack bean ␣-mannosidase digest of the same sample. Glc 1 Man 8 GlcNAc 2 (peak 4) and Glc 1 Man 9 GlcNAc 2 (peak 5) were digested to Glc 1 Man 4 GlcNAc 2 (peak 7) and Glc 1 Man 5 GlcNAc 2 (peak 8). Man 6 -9 GlcNAc 2 (peaks 1-4) were digested to Man 1 GlcNAc 2 (peak 6). Molecular representations of some of the glycoforms are included. For key see Fig. 7.
Barr virus transformed human B-lymphoblastoid cells have been shown to be remarkably homogeneous. The site at Asn 86 contained a complex type bi-antennary core fucosylated glycan with and without a bisecting GlcNAc (32), and no oligomannose sugars were identified. Based on the well established N-glycosylation pathway (33), these data indicate that before leaving the ER, the oligomannose structures attached to MHC class I HCs are modified to Man 6 GlcNAc 2 . The analysis of HCs in the loading complex eluted from the ␣-ERp57 column indicates that trimming to Glc 1 Man 8 GlcNAc 2 can take place while CRT is still associated with the loading complex and that trimming to Man 8 -6 GlcNAc 2 can take place on the intact loading complex after dissociation of CRT but while ERp57 is still present. The low levels of Man 8 -6 GlcNAc 2 suggest that class I molecules dissociate from ERp57 and leave the ER soon after dissociation of CRT. The HCs analyzed here may represent residual MHC class I-peptide complexes that have yet to dissociate from ERp57. Fig. 10 is a "snapshot" that gives a qualitative picture of the dissociation of the loading complex from ERp57. In contrast, glycans attached to empty MHC class I molecules are expected to be reglucosylated by UDP-glucosyl transferase, and, if peptide is not loaded in subsequent cycles, the empty class I molecules will eventually be targeted for degradation. Trimming to Glc 1 Man 8 GlcNAc 2 has been proposed to be a signal that initiates degradation through the ER-associated degradation pathway. This involves retrotranslocation from the ER to the cytosol followed by proteasomal degradation (34,35). A small fraction of empty molecules may escape the ER and be transiently expressed on the cell surface, a situation more readily detectable in murine cells (36,37).
Unassembled MHC Class I HCs Do Not Contain Glc 1 Man 9 -GlcNAc 2 -Free HCs do not leave the ER for the Golgi but are degraded (34,35). In contrast to the HCs recovered from ␣-ERp57 and ␣-TAP affinity columns, unassembled HCs expressed in Raji cells and isolated from ␣-HC affinity chromatography (Fig. 4) did not contain glucosylated glycans. Neither did free HCs expressed in Daudi cells, which do not contain ␤ 2 m. (Table I). Free HCs have been shown to associate with CNX rather than CRT, but we do not know the fraction of the total that are CNX-associated. We did not specifically address the issue of whether CNX-associated HCs have monoglucosylated N-linked glycans. The percentage of the total that carry monoglucosylated N-linked glycans may be too low for detection, or the interaction between CNX and HCs may have a glycan-independent component, as recently suggested by Danilczyk and Williams (38).
Glycosylation of Tapasin-Tapasin glycosylation is significantly different from the class I HCs (Figs. 3-5 and Table IV). Tapasin co-purifying with ERp57 or TAP contained no detectable Glc 1 Man 8 GlcNAc 2 , when the peak areas of the intact glycan pool and the glucosidase II digest were compared (Table  II), and less than 10% of the glycan has the Glc 1 Man 9 GlcNAc 2 structure. 40 -60% of the tapasin glycans are Man 9 GlcNAc 2 ( Table IV). The small fraction of the TAP-and ERp57-associated tapasin molecules containing Glc 1 Man 9 GlcNAc 2 may be folding intermediates. Previously it was established that a subset of tapasin-TAP complexes are associated with CNX and ERp57 and that the interaction of CNX and MHC class I molecules with the TAP complex are mutually exclusive (20). It was suggested that the CNX-associated species might represent tapasin in the process of folding. The CNX-tapasin interaction can be seen even with truncated tapasin molecules that lack the N-linked glycan, but nevertheless the presence of the monoglucosylated glycoform in a subset of the TAP-associated tapasin species may indicate that the subset is associated with CNX and ERp57 via the lectin site on CNX. The predominance of Man 9 GlcNAc 2 glycans in the tapasin pool probably reflects the fact that tapasin, which is properly folded and retained in  9. Proposed model of human MHC class I in association with TAP, tapasin, CRT, and ERp57. Tapasin is glycosylated with Man 9 GlcNAc 2 and interacts covalently with ERp57 via a disulfide bond. 50% of the HC glycoforms contain Glc 1 Man 9 GlcNAc 2 and interact noncovalently with tapasin and with CRT through the oligosaccharide substrate for CRT. The remaining 50% contain Man 6 -9 GlcNAc 2 glycans that may provide sufficient affinity for CRT allowing the HCs to remain associated with the complex. Alternatively these HCs may associate with CRT independently of the glycosylation through protein-protein interactions (31). the ER by a specific retention signal in its cytoplasmic domain (10), contains glycans that are somewhat inaccessible to the ER mannosidases.
Conclusion-In conclusion, this paper demonstrates that ϳ50% of the sugars N-linked to class I heavy chains in association with CRT and ERp57 are monoglucosylated. This is consistent with the class I peptide loading process being an adaptation of the general quality control mechanism involving CRT and ERp57.