Tapasin is disulfide linked to ERp57 within the peptide loading complex. In cell-free assays, a soluble variant of the tapasin/ERp57 dimer recruits MHC class I molecules and promotes peptide binding to them, whereas soluble tapasin alone does not. Here we show that within cells, tapasin conjugation with ERp57 is as critical as its integration into the membrane for efficient MHC class I assembly, surface expression, and Ag presentation to CD8+ T cells. Elimination of both of these properties severely compromises tapasin function, in keeping with predictions from in vitro studies.
Stable peptide binding to MHC class I–β2m dimers is facilitated by the peptide loading complex (PLC), which is composed of TAP, tapasin, ERp57, calreticulin (CRT), MHC class I H chain (HC), and β2m (reviewed in 1). Association with peptide is essential for the stability and transport of MHC class I molecules, and tapasin plays a key role in their optimal assembly and peptide loading 2, 3. It physically links newly synthesized MHC class I–β2m dimers to TAP, favoring their exposure to translocated peptides 4, 5, and increases TAP steady-state levels, enhancing peptide transport into the ER 2. Within the PLC, tapasin forms a stable heterodimer with ERp57 6, 7, which is a member of the protein disulfide isomerase family with a four-domain structure (abb′a′) containing thioredoxin motifs (CXXC) in the a and a′ domains. These ensure correct disulfide bond formation in glycoproteins because ERp57 interacts with the ER chaperones calnexin and CRT, which bind to monoglucosylated N-linked glycans transiently present on newly synthesized glycoproteins 8, 9.
Tapasin and ERp57 are covalently connected by a disulfide bond between Cys 95 of tapasin and Cys 57 of ERp57, which is in the a domain active site. The recent structural characterization of the heterodimer indicates that it is stabilized by non-covalent tapasin interactions with both the a and a′ domains 10. Mutagenesis of the active site cysteine residues and structural considerations indicate that enzymatic activity of ERp57 plays no role in facilitating peptide binding 10, 11. The most likely function for ERp57 in the PLC is that it enhances MHC class I recruitment by providing a site of interaction for CRT associated with the MHC class I N-linked glycan 1, 10.
Certain alleles, for example, HLA-B27 and B*4405, are relatively well expressed on the cell surface in the absence of tapasin. However, the peptide repertoire is affected and class I turnover at the cell surface is increased 12. This is because tapasin not only facilitates peptide binding but also “edits” the repertoire to maximize peptide binding affinity 13. Recently these functions were analyzed in vitro, using recombinant soluble tapasin species lacking the transmembrane domain. Wearsch and Cresswell showed that a soluble tapasin/ERp57 dimer could bind to MHC class I molecules together with CRT, promote peptide loading, and function as a peptide editor 14. Chen and Bouvier 15 attached leucine zipper peptides to the C-termini of tapasin and MHC class I to stabilize their interaction. Forcing the MHC class I/tapasin interaction in this way facilitated peptide loading. In both cases, interactions between MHC class I and soluble ERp57-free tapasin were almost undetectable.
The in vitro analyses used soluble tapasin species, making comparison with cellular experiments difficult. Expression of soluble tapasin in .220.B8 cells, which associates with MHC class I but not with TAP, restored some surface expression of HLA-B8 2. Expression of tapasin mutated at Cys 95, which does not recruit ERp57 to the PLC, also allowed some MHC class I expression on the cell surface 6. To determine whether the importance of tapasin/ERp57 heterodimer formation suggested by the in vitro studies was true in the environment of the ER, we compared the properties of MHC class I molecules assembled in the presence of soluble and membrane-associated tapasin capable or incapable of recruiting ERp57.
Results and discussion
Recruitment of PLC components by tapasin mutants lacking membrane and/or ERp57 association
We constructed a series of tapasin variants (Fig. 1A) encoding full-length tapasin (WT), C95A tapasin (cysteine to alanine substitution at position 95), or their soluble counterparts (sWT-K, sC95A-K). ER retention of the soluble constructs was assured by addition of a C-terminal KDEL motif. The constructs were stably expressed in the tapasin-negative cell line .220.B*4402 using a bicistronic EGFP-retroviral expression vector. We reasoned that the soluble conjugate used in vitro by Wearsch and Cresswell 14 is analogous to soluble, WT tapasin (sWT-K). The soluble leucine zipper-modified tapasin used by Chen and Bouvier 15 is analogous to full-length C95A tapasin (C95A) in that both provide the correct orientation relative to the MHC class I molecule, one by the leucine zipper interaction and the other by integration into the membrane, while both lack ERp57 association. The sC95A-K construct is analogous to soluble ERp57-free tapasin.
Tapasin was immunoprecipitated from MMTS-treated digitonin lysates of cells expressing the constructs, and TAP1, ERp57, MHC class I, and CRT association was evaluated by Western blotting (Fig. 1B and C). The amounts of both soluble versions of tapasin were higher than the transmembrane versions, probably because of the destabilizing effect of charged residues in the tapasin transmembrane domain normally masked by TAP association 16. As expected, the soluble species failed to bind TAP, while both C95A mutants did not bind ERp57 (Fig. 1B). Confirming previous data 17, the lack of ERp57 association with C95A tapasin led to a decrease in MHC class I and CRT association. The decrease was greater than that observed by eliminating the transmembrane domain (Fig. 1C). Loss of both ERp57 binding and membrane integration in the sC95A-K mutant reduced its association with MHC class I molecules to an undetectable level (Fig. 1C). Surprisingly, despite the absence of detectable MHC class I binding to sC95A-K, CRT association remained similar to that of full-length C95A tapasin, suggesting that CRT may interact with sC95A-K tapasin directly. The single N-linked glycan of WT tapasin is not substantially mono-glucosylated 18, but without ERp57 soluble tapasin may be improperly folded and a substrate for UDP-glucose glycoprotein transferase, which glucosylates unfolded glycoproteins 19, 20. Cys 95, which recruits ERp57 to mammalian tapasin, is not present in chicken tapasin, so significant misfolding of the full-length C95A mutant may be unlikely. Additionally, the steady-state levels of the two soluble tapasin species are similar, indicating that lack of ERp57 association is not sufficiently destabilizing to cause degradation by the ER-associated degradation pathway.
The tapasin variants and MHC class I expression
We next examined the ability of the tapasin mutants to support surface HLA-B*4402 expression (Fig. 2A). C95A and sWT-K tapasin restored expression to approximately half the level obtained with WT tapasin, measured by reactivity with the class I mAb w6/32 (left panel). Maturation of HLA-B*4402 in the presence of soluble human tapasin occurs at a similar rate to that of WT tapasin 13, 21, suggesting that in the presence of soluble tapasin, HLA complexes successfully exit the ER. Therefore, in our system, low level of expression of HLA-B*4402 in the presence of soluble tapasin may result from dissociation of HLA molecules in a post-ER compartment 17 or at the cell surface 13.
Expression of the HLA-Bw4 epitope of HLA-B*4402 assayed using the Tü109 mAb was more profoundly affected by the C95A mutation. Binding of Tü109 was previously reported to be influenced by the nature of the peptide cargo associated with HLA-B*5101 and HLA-B*4402 22, 23. Lower binding of Tü109 Ab to C95A expressing cells might therefore reflect the inability of the ERp57-free tapasin to effectively edit the peptide repertoire 14. The loss of the transmembrane domain and ERp57 association (sC95A-K) dramatically reduced MHC class I expression. HLA-Bw4 mAb binding was virtually the same as in cells lacking tapasin, i.e. less than 5% of that seen with WT tapasin (Fig. 2A, right panel). This is consistent with the in vitro studies of soluble tapasin and the soluble tapasin/ERp57 conjugate, where the latter was vastly superior in mediating peptide loading 14.
To assess kinetically the impact of the tapasin variants on peptide loading, cells were acid washed to denature surface MHC class I–β2m dimers and the rate of arrival of new molecules at the cell surface was monitored by flow cytometry (Fig. 2B). The total MHC class I recovery, and the rate of recovery, was highest in cells expressing WT tapasin. In the absence of tapasin, the rate of recovery was almost 80% less, whereas in cells expressing sWT-K tapasin, it was reduced by approximately 20%. Introduction of the C95A mutation into full-length tapasin also reduced the recovery rate by 20%, whereas for soluble tapasin the C95A mutation decreased the recovery rate by 40%. Accumulation of HLA-B*4402 molecules at the cell surface over time appeared to be lower after loss of TAP/membrane association and/or ERp57 interaction. Expression levels are likely to reflect the assembly efficiency of HLA-B*4402 molecules in combination with their stability at the cell surface, which generally correlates with the quality of the associated peptide repertoire.
The tapasin variants and MHC class I–β2m dimer stability
As a surrogate measure of the affinity of their peptide repertoire 6, 12, 13, we examined the thermostability of HLA molecules in .220.B*4402 cells expressing the tapasin variants (Fig. 3A). After heating the cells at temperatures ranging from 37 to 57°C, the w6/32 mAb was used to quantitate residual surface MHC class I–β2m dimers 24. We observed that lack of tapasin membrane association (sWT-K) or of recruitment of ERp57 into the PLC (C95A) led to similar reductions in thermostability. This suggests that the quality of the peptide repertoire is affected similarly by the loss of ERp57/tapasin interaction and by the lack of TAP/membrane association. A more drastic reduction in thermostability, interpreted as a severe reduction in the affinity of the peptide repertoire, was found after loss of both these interactions in cells expressing sC95A-K tapasin.
Roles of ERp57 recruitment and membrane association of tapasin in CTL recognition
The effectiveness of the Ag processing machinery is ideally assessed by sensitivity to CTL. We used two HLA-B*4402-restricted T-cell clones recognizing peptides derived from the proteins MUM-1 25 and MAGE-A1 (in preparation). Cells expressing the tapasin mutants were transiently transfected with constructs encoding MUM-1 or MAGE-A1 and tested for recognition by the relevant CTL (Fig. 3B). For both CTL, recognition was reduced by 10–20% in cells expressing either C95A-FL or sWT-K. However, in line with the biochemical observations, loss of ERp57 recruitment and membrane association reduced CTL activation by 70–80%.
Tapasin is necessary for efficient assembly and high-affinity peptide binding to MHC class I molecules. It achieves this by directly associating with MHC class I, predominantly through a likely interaction with the α2 helix 10, and by increasing TAP levels 2. Physical coupling of tapasin and ERp57 also plays a major role in tapasin function. This may result from the interaction of ERp57 with CRT, which in turn recruits MHC class I molecules bearing monoglucosylated N-linked glycans to the PLC. However, recently it has been suggested that the ERp57-CRT interaction is not essential for MHC class I assembly 26, indicating that this issue is not entirely resolved. Nevertheless, the experiments reported here agree well with in vitro studies using recombinant proteins 14, 15: in the environment of the ER both the ERp57 interaction as well as the correct orientation and TAP association provided by the transmembrane domain are required for optimal tapasin function, whereas loss of both renders tapasin ineffective in facilitating peptide loading and T-cell recognition.
Materials and methods
Cell lines and Ab
The 721.220 cells transfected with HLA-B*4402 6 expressing full-length or soluble WT and C95A tapasin bicistronically with EGFP were generated by retroviral spinfection 27. Transduced cells were isolated by sorting for EGFP+ cells. CTL clones LB33-CTL 159/5 and LB1801-CTL 461/G4.2 recognizing peptides presented by HLA-B*4402 and derived respectively from MUM-1 and MAGE-A1 tumor Ag (25, in preparation) were maintained in IMDM (Invitrogen, Carlsbad, CA, USA) supplemented with 10% human serum, 50 U/mL rIL-2 (Chiron, Emeryville, CA, USA), 200 μM 1-methyl tryptophan (Sigma, St. Louis, MO, USA), L-arginine (116 mg/mL), L-asparagine (36 mg/mL), L-glutamine (216 mg/mL), penicillin (100U/mL), and streptomycin (100 μg/mL). mAbs used were 3B10.7 (anti-MHC class I H chain 6), 148.3 (anti-TAP128), w6/32 (anti-HLA-A,B,C 6), Tü109 (anti-HLA-Bw4 23), and PaSta-1 (anti-tapasin 6). Rabbit antisera against TAP-1 (R.RING4C), CRT (PA3-900, ABR-affinity Bioreagents), ERp57 (R.ERp57-C), and tapasin (R.gp48N, R.SinE) were used as in 11.
Bicistronic plasmids encoding WT or C95A tapasin and EGFP were prepared by ligating tapasin DNA fragments into PBMN-IRES-EGFP 6. DNA fragments encoding the tapasin C-terminus truncated at Asp-392 and supplemented with a KDEL ER retrieval motif were ligated into digested WT and C95A constructs. pDsRed-N1 was generated by replacing the EGFP coding sequence of pEGFP-N1 with the DsRed monomer coding sequence of the vector pDsRed (Clontech, Mountain View, CA, USA). DNA encoding 35 –amino acids including the MUM-1 peptide recognized by LB33-CTL 159/5 was fused to the C-terminus of DsRed in pDsRed-N1 following a short linker. The MAGE-A1 cDNA encoding the peptide recognized by LB1801-CTL 461/G4.2 was cloned in pcDNA1. All constructs were confirmed by sequencing.
Immunoprecipitation and Western blotting
Cells were incubated with 10 mM methyl methanethiosulfonate (MMTS, Pierce, Rockford, IL, USA) in PBS and extracted in 1% digitonin as described 11. Post-nuclear supernatants were incubated for 1 h at 4°C with the appropriate mAb coupled to Biogel A15M beads. After washing, precipitated material was eluted in 2× reducing SDS-PAGE sample buffer at 95°C for 5 min and resolved by SDS-PAGE. Later transfer membranes were blocked, probed, washed and proteins detected as described 11.
Cells were incubated at 4°C with w6/32 tissue culture supernatant or a 1:100 dilution of Tü109 ascites for 30 min, washed, and stained with goat anti-mouse Ig serum coupled to allophycocyanin (Invitrogen) for 30 min. For the thermostability experiments, cells were incubated for 10 min in PBS containing 0.1% sodium azide and then heated at various temperatures for 10 min. Cells were transferred to 4°C and stained with w6/32 directly conjugated to Alexa-647 (Invitrogen). To assess recovery after acid treatment, surface HLA complexes were denatured for 90 s in 50 mM sodium citrate buffer, pH 3.0, and neutralized using three volumes of 150 mM Na2HPO4, pH 10.5. After washing and incubation at 37°C, cells were collected and kept on ice until staining with w6/32. Data were acquired with a FACScalibur (Becton Dickinson, Franklin Lakes, NJ, USA).
2 × 106 721.220B*4402 cells expressing the tapasin mutants were nucleofected using Amaxa kit R, program V01 (Lonza, Walkersville, MD, USA), with 4 μg of pDsRed-N1 with or without 1 μg pDsRed-MUM-1. Transfection efficiency was assessed by flow cytometry using the dsRed signal. For the MAGE-A1 CTL recognition, transfection was with 0.5 μg pcDNA1-MAGE-A1. After 4 h, 104 nucleofected cells were incubated with 5×103 CTL in 150 μL of culture medium supplemented with rIL2 25 U/mL. After 20 h co-culture, supernatants were collected and IFN-γ was measured by ELISA (Invitrogen).
We thank Nancy Dometios for aiding in the preparation of this manuscript, and Dr. Pamela Wearsch for providing reagents and helpful discussions. We are grateful to Professors P. Coulie and P. van der Bruggen for providing CTL lines, T. Aerts and S. Ottaviani for valuable technical assistance, and Professor B. Van den Eynde for a careful review of the manuscript. Support for this work was provided by the Howard Hughes Medical Institute (P. C.), the NIH/National Institute of General Medical Sciences, Medical Scientists Training Grant-GM07205 (D. P.), Marie Curie Outgoing International Fellowship (OIF) from the European Union and the Fonds National de la Recherche Scientifique (F.N.R.S), Belgium (N. V.).
Conflict of interest: The authors declare no financial or commercial conflict of interest.