Essential -dependent interactions optimize MHC class I peptide loading

Pamela A. Wearscha, David R. Peapera, and Peter Cresswella,b,1

aDepartment of Immunobiology and bDepartment of Cell Biology and Howard Hughes Medical Institute, Yale University School of Medicine, New Haven, CT 06520-8011

Contributed by Peter Cresswell, February 15, 2011 (sent for review January 5, 2011) In this study we sought to better understand the role of the (CNX). Both chaperones have a globular lectin-binding domain quality control machinery in the assembly of MHC and an extended arm known as the P-domain that binds specifi- class I molecules with high-affinity peptides. The lectin-like chap- cally to ERp57, a member of the disulfide isomerase (PDI) erone (CRT) and the thiol oxidoreductase ERp57 partic- family. ERp57 has a four-domain architecture of abb′a′ in which fi ipate in the nal step of this process as part of the peptide-loading the first and last contain a CXXC active site. CRT and CNX work complex (PLC). We provide evidence for an MHC class I/CRT in- in concert with ERp57 to promote proper folding and disulfide termediate before PLC engagement and examine the nature of that bond formation of newly synthesized . Upon release interaction in detail. To investigate the mechanism of of the glycoprotein from CRT or CNX, its glycan is deglucosylated peptide loading and roles of individual components, we reconsti- tuted a PLC subcomplex, excluding the Transporter Associated with by GlsII and is no longer a substrate for the chaperones. If the Processing, from purified, recombinant . ERp57 glycoprotein has not yet acquired its native structure, the enzyme disulfide linked to the class I-specific chaperone and CRT UDP-glucose:glycoprotein glucosyltransferase 1 (UGT1), a fold- were the minimal PLC components required for MHC class I as- ing sensor, reglucosylates it to reinitiate an interaction with CRT/ sociation and peptide loading. Mutations disrupting the interaction CNX/ERp57. However, a properly folded glycoprotein is not a of CRT with ERp57 or the class I glycan completely eliminated PLC substrate for UGT1 and can be exported to the Golgi. activity in vitro. By using the purified system, we also provide Important roles for CRT and ERp57 in MHC class I assembly direct evidence for a role for UDP-glucose:glycoprotein glucosyl- and PLC function have been established from studies of KO transferase 1 in MHC class I assembly. The recombinant Drosoph- mice and deficient cell lines (2, 3). In the absence of either ila enzyme reglucosylated MHC class I molecules associated with component, the cell surface expression and stability of MHC suboptimal ligands and allowed PLC reengagement and high- class I molecules are reduced. The cause of the defect has been fi af nity peptide exchange. Collectively, the data indicate that elucidated for ERp57-deficient cells. Tapasin forms a stable CRT in the PLC enhances weak tapasin/class I interactions in a man- disulfide-linked heterodimer with ERp57, which is required for ner that is glycan-dependent and regulated by UDP-glucose:glyco- the structural stability and optimal function of the PLC (4, 7, 8). protein glucosyltransferase 1. The disulfide linkage is between Cys95 of tapasin and Cys57 a protein folding | peptide editing of the ERp57 domain active site (9), and the dimer is further stabilized by noncovalent interactions between tapasin and the a′ domain active site (10). CRT interacts with both the HC glycan he assembly of MHC class I molecules is a critical step in the and the b′ domain of ERp57, but the nature and importance of Tgeneration of immune responses against viruses and tumors, and also a highly specialized example of glycoprotein folding in these interactions within the PLC are controversial, particularly – the (ER). MHC class I molecules display in regard to glycan-independent substrate binding (11 15). Fi- peptides representative of the cellular protein content to CD8+ nally, the potential roles of GlsII and UGT1 in regulating the T cells, and the stable association of the class I heavy chain (HC), CRT/class I interaction in the PLC have yet to be addressed. By using a variety of biochemical approaches, including re- β2-microglobulin (β2m), and a high-affinity 8- to 10-aa foreign peptide is essential for T-cell activation. As a result, a specialized constitution of a PLC subcomplex entirely from purified com- adaptation of the glycoprotein folding machinery has evolved to ponents, we have investigated the role of ER quality control ensure the loading of MHC class I molecules with optimal peptide components in MHC class I peptide loading. We demonstrate ligands. Following HC assembly with β2m, the empty heterodimer the CRT is recruited to and released from the PLC along with rapidly and stably associates with the peptide-loading complex class I molecules. By examination of the CRT/HC stoichiometry (PLC), which facilitates the final peptide-binding step (1). The and the HC state, we found no evidence for glycan- functions of the MHC class I-specific components of the PLC are independent interactions within the PLC. Consistent with this, in fi β well de ned. Tapasin interacts with both the HC/ 2m dimer as vitro reconstitution of the PLC required the lectin- and ERp57- well as Transporter Associated with (TAP), binding activities of CRT and could be enhanced by UGT1- thereby retaining the empty complexes in proximity to the peptide mediated glucosylation. Taken together, the data support impor- supply. More importantly, tapasin association stabilizes class I tant roles for CRT and the quality control machinery in regulating fi molecules and promotes loading with high-af nity peptides. class I peptide loading by the PLC. However, the optimal activity of tapasin requires the presence of calreticulin (CRT) and ERp57, two ER proteins involved in – general glycoprotein folding, in the PLC (2 4). Author contributions: P.A.W., D.R.P., and P.C. designed research; P.A.W. and D.R.P. per- The ER glycoprotein quality control machinery is a complex formed research; P.A.W., D.R.P., and P.C. analyzed data; and P.A.W. and P.C. wrote the system that uses the structural state of N-linked to dic- paper. tate the fate of newly synthesized proteins (1, 5, 6). Initially, a The authors declare no conflict of interest. Glc3Man9GlcNAc2 glycan is transferred to polypeptides during Freely available online through the PNAS open access option. translocation and subsequently trimmed by glucosidase I (GlsI) 1To whom correspondence should be addressed. E-mail: [email protected]. and glucosidase II (GlsII) to a monoglucosylated glycan, which is This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. the substrate for the lectin-like chaperones CRT and 1073/pnas.1102524108/-/DCSupplemental.

4950–4955 | PNAS | March 22, 2011 | vol. 108 | no. 12 www.pnas.org/cgi/doi/10.1073/pnas.1102524108 Downloaded by guest on September 25, 2021 Results and Discussion to the conjugate with this assay. However, a significant portion of

CRT Associates with MHC Class I HC/β2m Heterodimers Before these complexes likely dissociate during the extract preparation Incorporation into the PLC. Work from our laboratory has dem- given the relatively low affinity of CRT for glycosylated class I K ∼ μ onstrated that the tapasin/ERp57 conjugate associates with TAP HC ( d of 1 M) (14) and the rapid rate of dissociation de- ∼ in cells lacking expression of MHC class I HC or β2m (8, 16), termined from kinetic experiments (t1/2 of 10 min at 4 °C; Fig. suggesting that this core serves as a scaffold onto which the S1). Consistent with our projections, the addition of excess remaining PLC components assemble. Furthermore, the GlsII recombinant CRT to the lysis buffer enhanced HC association inhibitor castanospermine (CST) prevents the interaction of MHC with the conjugate (Fig. 1C), presumably because of preservation β class I molecules with the PLC in intact cells (17) and a cell-free of CRT/HC/ 2m complexes in the assembly pathway. In agree- ment, Del Cid et al. (11) have proposed a recruiting role for assay (4). This suggests that CRT escorts HC/β2m dimers to tapasin, but this has not been demonstrated experimentally. Ini- CRT in the PLC based on reduced class I association with the fi tially, we sought to identify PLC-independent complexes of HLA- PLC in CRT-de cient cells. However, that interpretation is B8 with CRT by using tapasin-negative .220.B8 cells. We prepared complicated by the analyses being performed at steady state, i.e., extracts from radiolabeled cells in the absence or presence of the effect of CRT on the class I /tapasin interaction may be a DSP, a chemical cross-linker, and performed sequential immu- result of PLC stabilization. Although the two effects cannot be noprecipitations to detect CRT-associated proteins. As shown in discriminated biochemically, CRT may serve an important role β Fig. 1A, a CRT/HC interaction was observed in untreated extracts for both. The discovery that an HC/ 2m/CRT assembly in- and stabilized by DSP. β m, which is not glycosylated, also coim- termediate exists provides the most direct evidence for CRT- 2 A B munoprecipitated with CRT from cross-linked cell extracts, mediated recruitment (Fig. 1 and ), whereas the coupling of known interactions between PLC components is sufficient to demonstrating the presence of CRT-associated HC/β2m hetero- dimers (Fig. 1A). To determine whether this complex exists in the support CRT-mediated stabilization of existing class I/ tapasin presence of the PLC, we performed tapasin immunodepletions interactions (1). from radiolabeled .220.B8.Tpsn cells. As shown in Fig. 1B,a Analysis of the MHC Class I/CRT Interaction in the PLC. To better fraction of HLA-B8 molecules were associated with CRT after the understand the functions of CRT in class I assembly, we evalu- complete removal of tapasin. These results are consistent with ated the stoichiometry and mode of interaction between CRT the hypothesis that CRT/MHC class I complexes are an inter- and the class I HC in the PLC. The initial assessment of the mediate in the MHC class I assembly pathway. CRT/HC ratio (18) was complicated by the subsequent discovery To examine whether CRT binding to MHC class I molecules of ERp57 in the PLC, so we used a different approach that in- is required to initiate interactions with the remaining PLC volved quantitative immunoblotting. PLCs were isolated by components, we adapted a cell-free assay for PLC activity (4, 10). immunoprecipitation from .220.B8.Tpsn cells and analyzed in This system reconstitutes PLC subcomplexes lacking TAP by the parallel with purified CRT and class I HC as standards so the addition of recombinant soluble tapasin (sTpsn)/ERp57 conju- relative amounts could be determined. A representative blot is gate to .220.B8 cell extracts, which are a crude source of CRT shown in Fig. 2A and the average ratio calculated from five in- and empty class I molecules. We postulated that CRT/HC/β m 2 dependent experiments was 1.08 ± 0.19 CRT to 1.00 HC mole- complexes (Fig. 1 A and B) are the subset of molecules recruited cules. Thus, equimolar amounts of MHC class I HC and CRT are present in the PLC. In further support of this stoichiometry, we performed tapasin immunoprecipitations from CRT-depleted .220.B8.Tpsn extracts and found that there are essentially no CRT-deficient PLCs (Fig. 2B). Monoglucosylated glycans are required for CRT binding to the class I HC in vitro (14), which presents an apparent discrep-

ancy between the stoichiometry (Fig. 2A) and the mixed glycan IMMUNOLOGY structures of PLC-associated HC previously reported (19). We therefore reevaluated the glycosylation state of class I in the PLC with the use of two assays. The first used jack-bean mannosidase (JBM), an exomannosidase that trims nonglucosylated glycans more extensively than those with a terminal glucose residue (Fig. 2C), which can then be detected as subtle shifts by SDS/PAGE (20). We performed a pulse-chase experiment with .220.B8.Tpsn cells followed by JBM or EndoH digestion of PLC-associated tapasin or class I HC. As shown in Fig. 2D, the tapasin glycan contains terminal glucose residues shortly after synthesis (0 min chase), which are subsequently trimmed by the 90-min chase point, consistent with its maturation to the native state. In con- trast, the JBM-treated class I HC band did not shift during the Fig. 1. HC/β2m heterodimers associate with CRT independently of the PLC. (A) .220.B8 cells were labeled for 30 min and lysed with or without 0.3 mM chase; thus, its glycan is not trimmed while it is associated with the DSP. Immunoprecipitations were performed with tapasin (R.gp48C), CRT, or PLC. To more accurately quantify the fraction of PLC-associated

β2m Abs, followed by reimmunoprecipitation with 3B10.7 (HC) or β2m Abs. (B) class I HC with monoglucosylated glycans, our second assay used Radiolabeled .220.B8.Tpsn cells (60-min pulse) were lysed and subjected to pull-downs with CRT-GST fusion protein immobilized on gluta- four immunodepletion steps with PaSta1-coupled beads, followed by im- thione beads. As shown in Fig. S2, the binding of class I HC to the munoprecipitation with tapasin (R.gp48C), CRT, or ERp57 Abs. The associated CRT-GST beads was completely eliminated by treatment with HCs were reimmunoprecipitated with 3B10.7 and analyzed by SDS/PAGE. (C) GlsII and is therefore glycan-dependent. Compared with the total Cell extracts were prepared from .220.B8 cells in lysis buffer containing the indicated concentrations of recombinant CRT. Samples were then incubated amount of PLC-associated class I HC, determined by 3B10.7 with or without 0.35 μM C60A conjugate for 15 min at RT followed by im- immunoprecipitations, we found that 97% of the total could be munoprecipitation with PaSta1-coupled beads. The bound proteins were recovered from sequential CRT-GST pull-downs (Fig. 2E). We eluted and analyzed by immunoblotting with CRT and 3B10.7 Abs. therefore conclude that virtually all MHC class I molecules in

Wearsch et al. PNAS | March 22, 2011 | vol. 108 | no. 12 | 4951 Downloaded by guest on September 25, 2021 Fig. 2. Analysis of the CRT/HC interaction in the PLC. (A) TAP immunoprecipitations were performed from the indicated number of .220.B8.Tpsn cells and analyzed by SDS/PAGE along with recombinant class I HC and CRT standards. Quantitative immunoblotting was performed with 3B10.7 or CRT Abs, and the standard curves are shown in Fig. S2. Two sample sets were processed and averaged per experiment, but only one is shown. (B) Extracts from radiolabeled .220.B8.Tpsn cells (60-min pulse) were subjected to four CRT immunodepletion steps. Immunoprecipitations were then performed with control, PaSta1, or CRT Abs and analyzed by SDS/PAGE. (C) Schematic of the JBM assay. (D) .220.B8.Tpsn cells were pulse-labeled for 30 min and chased for as long as 90 min. Primary immunoprecipitations were performed with 148.3 Ab-coupled beads and secondary immunoprecipitations were performed with 3B10.7 or R.gp48C. Digests were then performed with JBM or EndoH and samples were analyzed by SDS/PAGE. Note that, after JBM digestion, the migration of glucosylated species is more similar to that of the undigested band (red asterisk), whereas the migration of deglucosylated species is more similar to that of the EndoH-treated band (blue asterisk). (E) .220.B8.tpsn cells were labeled for 60 min and sequential immunoprecipitation/pull-down assays were performed with 3B10.7 Ab, GST, or CRT-GST immobilized on beads followed by 3B.10.7 (Upper). The total amount of HC recovered was calculated from two sequential 3B10.7 immunopreci- pitations or from three CRT-GST pull-down steps (Lower).

the PLC are monoglucosylated and bound to CRT via the N- of HC/β2m complexes, extraction and purification were carried linked glycan. out in the presence of an intermediate-affinity peptide, the pre- viously described RAL variant of the antigenic peptide EBNA3 CRT and ERp57 Provide Structural Roles in Soluble PLCs Reconstituted (339-447) (4). The RAL ligand binds and stabilizes HLA-B8 from Purified Components. Conflicting data regarding the mecha- molecules (4), but can be displaced by the PLC in vitro (Fig. S4). nism of CRT and ERp57 activity in the PLC have been reported Our strategy was to assemble the PLC subcomplex in vitro from in studies in intact cells (11, 12, 15). We sought to address these purified CRT, sTpsn/ERp57 conjugate, and HC/β2m/RAL com- discrepancies using an in vitro assay for PLC activity. Since the plexes (Fig. S3C). First, we incubated the recombinant proteins discovery of tapasin in 1996, two experimental systems have been in various combinations and assessed PLC reconstitution by the devised to address its function and both have limitations. Chen coimmunoprecipitation of CRT and class I HC with tapasin (Fig. and Bouvier used recombinant soluble MHC class I complexes 3B). As expected, CRT bound to the conjugate at these concen- and tapasin purified from Escherichia coli (21), but excluding trations as a result of its specific interaction with the ERp57 b′ CRT and ERp57 from the system necessitated the addition of domain (Kd of 0.55 μM) (11). MHC class I complexes did not leucine zippers to tapasin and class I HCs to induce dimerization. interact with sTpsn/ERp57 alone, but did so efficiently in the In contrast, we developed a cell-free assay that successfully presence of CRT, demonstrating that PLC subcomplexes were reconstituted a subcomplex of the PLC (4) but used cell extracts, generated (Fig. 3B). These observations explain the need for meaning that a role for unidentified components could not be leucine zippers to force the dimerization of class I with tapasin in eliminated. For these reasons, we continued our endeavor to the absence of an HC glycan, CRT, and ERp57 (21). We next reconstitute a soluble subcomplex of the PLC from purified tested the reconstituted system for peptide loading activity by components (Fig. 3A). This has been a considerable challenge using the high-affinity NP(380-387L) ligand specific for HLA-B8 because of the multiple components and specialized interactions (4). We incubated HC/β2m/RAL complexes with purified PLC within the PLC (1). components and [125I]-NP and then determined radiolabeled The production of the appropriate PLC substrate (empty peptide loading by immunoprecipitations with the mAb w6/32, N MHC class I complexes with monoglucosylated -linked glycans) which binds HC/β2m/peptide complexes (Fig. 3C). HC/β2m/RAL has been the most problematic. We previously described the complexes were capable of a low level of [125I]-NP exchange that expression of soluble HLA-B8 bearing such a glycan in a Sac- was not affected by the presence of CRT or the sTpsn/ERp57 charomyces cerevisiae glycosylation mutant (14), but the refolding conjugate alone. However, when both CRT and the conjugate yields without high-affinity peptides were insufficient for exten- were present, a 10-fold increase in high-affinity peptide binding sive structural or functional analyses. Consequently, we tested was observed, which directly demonstrates catalysis of peptide a different strategy that involved baculovirus expression, as the exchange by the PLC. Some laboratories have reported that PDI class I HC assembles with β2m in insect cells but the complexes is a member of the PLC and affects peptide loading (24), but these are largely devoid of peptides because of the absence of tapasin findings have been contentious (1). Recombinant human PDI was and TAP (22, 23). Coexpression of the luminal domain of HLA- not required for PLC activity, nor did it enhance peptide loading B8 with human β2m in Sf21 cells resulted in the formation of in this assay. These experiments demonstrate successful recon- heterodimers that were completely retained in the ER as de- stitution of the PLC from purified proteins and establish MHC A termined by EndoH digestion (Fig. S3 ). We therefore postu- class I HC, β2m, CRT, tapasin, and ERp57 as the only compo- lated that a significant fraction of the HC molecules might possess nents essential for peptide exchange in vitro. the correct oligosaccharide structure for CRT binding. Analysis To directly test the roles of ERp57 and CRT in the PLC, we of samples from multiple infections by CRT-GST pull-downs generated several mutants and tested their ability to support pep- indicated that approximately 20% of the class I molecules pro- tide loading. The 3X mutant of ERp57 (C60A/C406A/C409A) duced in Sf21 cells were monoglucosylated (Fig. S3B). HLA-B8 allows covalent trapping of tapasin, but is redox-inactive (7). The molecules expressed in Sf21 cells were unstable, however, as a ΔPDB ERp57 construct (K214A/K274A/K284A) is a combina- result of the absence of bound peptides. To prevent dissociation tion of the individual point mutations that most profoundly

4952 | www.pnas.org/cgi/doi/10.1073/pnas.1102524108 Wearsch et al. Downloaded by guest on September 25, 2021 Fig. 3. In vitro reconstitution of a soluble subcomplex of the PLC. (A) Schematic of the soluble PLC subcomplex. (B) Recombinant proteins (0.5 μM each) were incubated at RT for 1 h and immunoprecipitations were performed with PaSta1-coupled beads. The tapasin-associated CRT and HC were detected by im- munoblotting. (C and D) The indicated recombinant proteins—CRT or Y92A, a CRT glycan-binding mutant; disulfide-linked conjugates of sTpn with C60A, an ERp57 mutant that traps tapasin, with 3X, a redox-inactive ERp57 mutant that traps tapasin, or with ΔPDB, an ERp57 mutant that traps tapasin but does not interact with CRT—were incubated at a final concentration of 0.4 μM each for 15 min at RT with [125I]-NP. Peptide loading was measured by w6/32 im- munoprecipitation and γ-counting.

disrupt CNX/CRT P-domain binding (25). Both ERp57 mutants that CRT does not associate with TAP in the absence of MHC efficiently formed disulfide-linked heterodimers when coex- class I heterodimers (16). pressed with sTpsn and were purified. We also produced a Y92A Because CRT and MHC class I molecules are released si- mutant of human CRT based on the analogous mutations of multaneously from TAP (Fig. 4A) in a manner that is apparently rabbit and mouse CRT that disrupt glycan binding (11, 26). affected by GlsII inhibition (27), we considered the possibility Consistent with the crystal structure of the sTpsn/ERp57 conju- that trimming of the HC glycan triggers dissociation of the PLC. gate (10) and the characterization of the 3X mutant in ERp57- To address this question in greater detail, we labeled .220.B8. deficient cells (7, 15), inactivation of ERp57 redox activity did not Tpsn cells and followed the kinetics of HC release from TAP in impair the function of the PLC (Fig. 4D). However, when either the absence or presence of CST. We found that the interaction Y92A CRT or the sTpsn/ΔPDB ERp57 conjugate was used, no of class I molecules with the PLC was dramatically prolonged B tapasin-mediated peptide loading was observed. Thus, the CRT/ when GlsII activity was inhibited during the chase (Fig. 4 ). This glycan and CRT/ERp57 interactions are required for PLC func- suggested that glycan trimming is required for release of class tion in vitro. Contradictory results have been reported on this is- I from the PLC. To further investigate this, we compared the IMMUNOLOGY ability of purified GlsII to trim the monoglucosylated glycan of sue with the use of mutants expressed in CRT- or ERp57-deficient PLC-associated and free HC. TAP immunoprecipitations were mouse cells (11, 12, 15). With an independent experimental sys- performed from radiolabeled .220.B8.Tpsn cells, samples were tem involving soluble components, our results agree with the incubated with GlsII before or after reimmunoprecipitation of findings of Del Cid et al. (11). the class I HC, and then JBM digestions were performed as C CRT Exits the PLC with MHC Class I, but Release Is Not Induced by a readout. As shown in Fig. 4 , the glycans of free HCs, but not PLC-associated HCs, were accessible to GlsII (compare lanes 3 GlsII. The events that govern MHC class I peptide loading and and 4 in Fig. 4C). Glycan trimming must therefore occur after dissociation of the PLC are poorly understood. Based on our release of CRT or MHC class I from the PLC. This finding would finding that empty MHC class I heterodimers and CRT are appear to disagree with the conclusion implied by the data in Fig. corecruited to the core complex of tapasin/ERp57/TAP (Fig. 1), 4B and ref. 27, i.e., that glucose removal is required for disso- we postulated that they are also simultaneously released from ciation. To reconcile these findings, we suggest that the pro- the PLC upon peptide binding. To test this hypothesis, we used β longed interaction of MHC class I molecules with the PLC is siRNA to knock down 2m, preventing the association of newly maintained by a dynamic equilibrium that results from multiple synthesized MHC class I molecules with tapasin, and then fol- rounds of release and reengagement of the class I glycan pro- lowed the fate of the preexisting PLC components. Transient vided it is monoglucosylated. transfections of .220.B8.Tpsn cells were performed with control This hypothesis presents interesting implications for the mech- β fi or 2m-speci c siRNA, and samples were harvested for as long a anism of peptide optimization, including a potential role for 12 h (Fig. 4A). Analysis by coimmunoprecipitation and blotting UGT in promoting repeated peptide loading events until the showed that the amount of TAP-associated tapasin/ERp57 con- highest-affinity ligand is acquired. In support of this conclusion, jugate remained constant over the 12-h period after transfection, we have found and reported in a companion paper in PNAS that whereas the CRT and class I HC levels decreased in parallel. the assembly of MHC class I molecules is impaired in UGT1- This suggests that CRT is released from the core PLC compo- deficient murine fibroblasts (28). To further these observations, nents following peptide loading, and agrees with the observation we adapted our in vitro assay to directly demonstrate that MHC

Wearsch et al. PNAS | March 22, 2011 | vol. 108 | no. 12 | 4953 Downloaded by guest on September 25, 2021 Fig. 4. Reglucosylation of MHC class I molecules by UGT promotes reengagement with the PLC. (A) Control or β2m-specific siRNA oligos were introduced into .220.B8.Tpsn cells and samples were harvested at 0, 12, and 24 h after nucleofection. PLC components were detected by immunoprecipitation with 148.3 Ab and quantitative immunoblotting (Left). Amounts of PLC-associated CRT, tapasin, and class I HC were calculated and normalized to TAP (Right). (B) .220.B8. Tpsn cells were labeled with [35S]-methionine for 30 min and chased for 0 to 120 min with or without 2.5 mM CST. Immunoprecipitations were performed with control or 148.3 Abs followed by elution and reimmunoprecipitation with 3B10.7 mAb (Left). Right: PhosphorImager quantification. (C) .220.B8.Tpsn cells were labeled for 60 min and then subjected to immunoprecipitations and enzymatic digestions as shown in the flowchart (Upper). (D) Purified recombinant MHC class I complexes depleted of those with monoglucosylated glycans were incubated with or without UGT and UDP-glucose. Subsequently, [125I]-NP loading was measured in the absence or presence of the remaining PLC components (0.4 μM C60A conjugate and CRT).

class I complexes with intermediate-affinity peptides are sub- action and for UGT1 in peptide exchange and optimization. The strates for reglucosylation by using the Drosophila homologue reconstituted system finally opens the door for the detailed (i.e., UGT) of human UGT1. Purified HC/β2m/RAL preparations mechanistic analysis of the core PLC components, as well as were depleted of monoglucosylated species with CRT-GST beads, possible accessory roles of other ER factors such as peptidases in incubated with or without recombinant UGT and UDP-glucose, MHC class I peptide loading. and then subsequently tested for PLC-mediated loading of [125I]- NP. As shown in Fig. 4D, no peptide exchange was observed fol- Materials and Methods lowing the CRT-GST depletion, indicating that all HC-bearing Plasmids. A baculovirus construct was generated for the coexpression of – β fi monoglucosylated glycans had been removed. However, peptide soluble HLA-B*0801 (AA-24 273) with human 2m. Inserts were ampli ed by exchange was observed when the depleted HC/β m/RAL com- using PCR and cloned into pFastBac Dual (Invitrogen). The mature domain of 2 human CRT with a C-terminal 6× His tag was cloned into pET15b (Novagen). plexes were enzymatically reglucosylated by using UGT and UDP- The QuikChange mutagenesis kit (Agilent) was used to generate Y92A CRT in glucose. Thus, as demonstrated directly in the companion paper pET15b and the two ERp57 mutations (ΔPDB and 3X) in pFastBac Dual (28), UGT recognizes MHC class I molecules loaded with sub- with sTpsn. optimal ligands, promoting peptide optimization by the PLC. Cell Lines and Reagents. The human .220.B8 cell line and its WT tapasin Summary. Taken together, the results from the present study transfectant (.220.B8.Tpsn) have been described previously (18). The anti- support a critical role for the lectin-like chaperone activity of bodies used were 3B10.7 (free class I HC), w6/32 (class I conformation- CRT in the PLC. Although it has been speculated that CRT specific), PaSta1 (tapasin), R.gp48C (tapasin), R.ERp57 (full length ERp57), escorts class I molecules to tapasin and TAP (11, 17), we provide and CRT (4, 8). The peptides NP(380-387L) and EBNA3(339-447) RAL were experimental evidence for the existence of this intermediate. In synthesized by GenScript and iodinated as described (4). addition to a recruiting function, CRT stabilizes class I molecules Immunoprecipitations and Immunoblotting. Cells (5 × 106 per sample unless within the PLC via an equimolar, glycan-dependent interaction. indicated otherwise) were lysed in 1% digitonin (EMD Biosciences) in 150 mM By coupling interactions with the HC glycan and ERp57, the NaCl, 25 mM Tris-Cl, pH 7.4, 1 mM CaCl2 (TBS-C) with complete EDTA-free presence of CRT greatly enhances the otherwise weak affinity protease inhibitor tablets (Roche) and 10 mM methyl methanethiosulfonate between MHC class I molecules and the tapasin/ERp57 hetero- (Pierce). Following a 30-min preclear, immunoprecipitations were performed dimer. The glucosylation state of the class I molecule therefore for 1 h at 4 °C with the indicated Abs and protein G Sepharose or directly dictates its engagement with the PLC, permitting a role for UGT1 conjugated beads. Normal rabbit serum was used as a control Ab. After three in the optimization of the peptide repertoire as described in an- washes with 0.1% digitonin in TBS-C, proteins were eluted in sample buffer other study in PNAS (28). or in 0.5% SDS for subsequent reimmunoprecipitation after dilution into 1% Triton in TBS-C. The following steps were used when indicated. Pulse/chase For more than a decade, the analysis of the PLC and its experiments were performed as described (14). Immunodepletions were ex- components has been largely restricted to the biochemical ecuted with four 45-min incubations with Ab-coupled beads. Enzymatic characterization of cell lines. Although such studies have been digests were performed overnight at 37 °C with EndoH (NEB) or JBM (Sigma). highly informative, they can provide only an indirect analysis of The standard blotting procedures and detection with ECL (Pierce) were per- PLC function. The development of an in vitro assay for PLC formed as described (4, 7). For quantitative immunoblots, CRT and soluble function now permits precise analysis of the mechanisms of HLA-B8 purified from E. coli (14) were used as standards, and proteins were peptide loading. This objective has been hampered for many detected with alkaline phosphatase-conjugated secondary Ab (Jackson Labs) years by the inability to produce the appropriate class I substrate and ECF substrate (GE Healthcare). PhosphorImager and FluorImager quan- β tifications were performed by using the Storm860 system and ImageQuant (4, 14, 21). By using soluble HLA-B8 expressed with 2m in in- software (GE Healthcare). sect cells, we were able to reconstitute a soluble subcomplex of the PLC with purified CRT and sTpsn/ERp57. In addition β2m siRNA. Cells (.220.B8.Tpsn) were transfected with β2m-specific or control to establishing the minimal components required for function, siRNA oligos by using the Amaxa Nucleofector (Lonza) with program V01 in we unambiguously established a role for the CRT/ERp57 inter- five 100-μL aliquots. Briefly, 20 × 106 cells were resuspended in 500 μLof

4954 | www.pnas.org/cgi/doi/10.1073/pnas.1102524108 Wearsch et al. Downloaded by guest on September 25, 2021 Solution R with 250 pmol control (GCUUCAACAGCAGGCACUC) or 125 pmol rified as described (6). Drosophila UGT produced in Sf9 cells was a gift of

each of β2m-specific (GAGUAUGCCUGCCGUGUGAUU and GCAAGGACUGG- Karin Reinisch (Yale University, New Haven, CT). UCUUUCUAUU) siRNA oligos (Dharmacon). Samples were harvested at the indicated times and frozen for subsequent analysis. Immunoprecipitations Peptide Loading Assay. Purified components (0.4 μM) were incubated in 50 μL and blotting were performed as described earlier, and the data were nor- TBS-C with 0.8 μM[125I]-NP(380-387L) for 15 min at room temperature (RT). malized to TAP levels before calculating the percentage of control siRNA- (The residual RAL concentration from the purified MHC was 5 μM.) Next, 750 treated value. μL of 0.1% Triton in TBS-C with 50 μM cold NP peptide was added and w6/32 immunoprecipitations were performed. After three washes with 0.1% Triton Protein Purification. All CRT and sTpsn/ERp57 recombinant proteins were in TBS-C, the amount of peptide bound on the beads was measured by using purified by using the established protocols (4, 14). Unless indicated other- a Wallac 1420 counter (Perkin-Elmer). Reactions were performed in duplicate wise, the C60A conjugate (WT sTpsn disulfide-linked to C60A ERp57), which and the baseline values (i.e., no recombinant protein) were subtracted as

is fully active (4), was used for experiments. Free class I HCs (soluble HLA-B8) background from all data sets. For reglucosylation experiments, the HC/β2m/ fi β were puri ed from E. coli inclusion bodies (14), and HC/ 2m/RAL complexes RAL preparations were depleted of monoglucosylated complexes by using were isolated from Sf21 cells by using the following protocol. Frozen cell CRT-GST beads and then incubated with 1 μM UGT and 100 μM UDP-glucose pellets were lysed in 1% Triton in 50 mM sodium phosphate, 300 mM NaCl, (Sigma) in 50 mM NaCl, 10 mM CaCl2, 25 mM Tris-Cl, pH 8, for 1 h at 30 °C 10 mM methyl methanethiosulfonate, with Complete EDTA-free tablets before the peptide loading assay. (Roche). The complexes were purified by using Talon beads (Clontech) ac- ’ cording to the manufacturer s instructions and then by MonoQ chroma- ACKNOWLEDGMENTS. The authors thank Wei Zhang for helpful discus- tography with a linear gradient of 0 to 500 mM NaCl in 25 mM Tris-Cl, pH 8. sions, Karin Reinisch for valuable reagents, and Nancy Dometios for β μ To limit dissociation of HC/ 2m dimers, 25 to 50 M RAL peptide was in- manuscript preparation. This work was supported by The Howard Hughes cluded in the lysis, purification, and storage buffers. Pig liver GlsII was pu- Medical Institute.

1. Wearsch PA, Cresswell P (2008) The quality control of MHC class I peptide loading. 16. Diedrich G, Bangia N, Pan M, Cresswell P (2001) A role for calnexin in the assembly of Curr Opin Cell Biol 20:624–631. the MHC class I loading complex in the endoplasmic reticulum. J Immunol 166: 2. Gao B, et al. (2002) Assembly and a ntigen-presenting function of MHC class I 1703–1709. molecules in cells lacking the ER chaperone calreticulin. Immunity 16:99–109. 17. Sadasivan B, Lehner PJ, Ortmann B, Spies T, Cresswell P (1996) Roles for calreticulin 3. Garbi N, Tanaka S, Momburg F, Hämmerling GJ (2006) Impaired assembly of the major and a novel glycoprotein, tapasin, in the interaction of MHC class I molecules with histocompatibility complex class I peptide-loading complex in mice deficient in the TAP. Immunity 5:103–114. oxidoreductase ERp57. Nat Immunol 7:93–102. 18. Ortmann B, et al. (1997) A critical role for tapasin in the assembly and function of 4. Wearsch PA, Cresswell P (2007) Selective loading of high-affinity peptides onto major multimeric MHC class I-TAP complexes. Science 277:1306–1309. histocompatibility complex class I molecules by the tapasin-ERp57 heterodimer. Nat 19. Radcliffe CM, et al. (2002) Identification of specific glycoforms of MHC class I heavy Immunol 8:873–881. chains suggests that class I peptide loading is an adaptation of the quality control 5. Helenius A, Aebi M (2004) Roles of N-linked glycans in the endoplasmic reticulum. pathway involving calreticulin and ERp57. J Biol Chem 277:46415–46423. Annu Rev Biochem 73:1019–1049. 20. Cannon KS, Helenius A (1999) Trimming and readdition of glucose to N-linked 6. D’Alessio C, Caramelo JJ, Parodi AJ (2010) UDP-GlC:glycoprotein glucosyltransferase- oligosaccharides determines calnexin association of a substrate glycoprotein in living glucosidase II, the ying-yang of the ER quality control. Semin Cell Dev Biol 21:491–499. cells. J Biol Chem 274:7537–7544. 7. Peaper DR, Cresswell P (2008) The redox activity of ERp57 is not essential for its 21. Chen M, Bouvier M (2007) Analysis of interactions in a tapasin/class I complex provides functions in MHC class I peptide loading. Proc Natl Acad Sci USA 105:10477–10482. a mechanism for peptide selection. EMBO J 26:1681–1690. 8. Peaper DR, Wearsch PA, Cresswell P (2005) Tapasin and ERp57 form a stable disulfide- 22. Jackson MR, Song ES, Yang Y, Peterson PA (1992) Empty and peptide-containing linked dimer within the MHC class I peptide-loading complex. EMBO J 24:3613–3623. conformers of class I major histocompatibility complex molecules expressed in 9. Dick TP, Bangia N, Peaper DR, Cresswell P (2002) Disulfide bond isomerization and the Drosophila melanogaster cells. Proc Natl Acad Sci USA 89:12117–12121. assembly of MHC class I-peptide complexes. Immunity 16:87–98. 23. Lauvau G, et al. (1999) Tapasin enhances assembly of TAP-dependent and inde- 10. Dong G, Wearsch PA, Peaper DR, Cresswell P, Reinisch KM (2009) Insights into MHC pendent peptides with HLA-A2 and HLA-B27 expressed in insect cells. J Biol Chem 274: class I peptide loading from the structure of the tapasin-ERp57 thiol oxidoreductase 31349–31358. heterodimer. Immunity 30:21–32. 24. Park B, et al. (2006) Redox regulation facilitates optimal peptide selection by MHC 11. Del Cid N, et al. (2010) Modes of calreticulin recruitment to the major histo- class I during antigen processing. Cell 127:369–382. compatibility complex class I assembly pathway. J Biol Chem 285:4520–4535. 25. Kozlov G, et al. (2006) Crystal structure of the bb’ domains of the protein disulfide 12. Ireland BS, Brockmeier U, Howe CM, Elliott T, Williams DB (2008) Lectin-deficient isomerase ERp57. Structure 14:1331–1339. calreticulin retains full functionality as a chaperone for class I histocompatibility 26. Thomson SP, Williams DB (2005) Delineation of the lectin site of the molecular molecules. Mol Biol Cell 19:2413–2423. chaperone calreticulin. Cell Stress Chaperones 10:242–251. 13. Rizvi SM, Mancino L, Thammavongsa V, Cantley RL, Raghavan M (2004) A polypeptide 27. van Leeuwen JEM, Kearse KP (1996) Deglucosylation of N-linked glycans is an IMMUNOLOGY binding conformation of calreticulin is induced by heat shock, calcium depletion, or important step in the dissociation of calreticulin-class I-TAP complexes. Proc Natl Acad by deletion of the C-terminal acidic region. Mol Cell 15:913–923. Sci USA 93:13997–14001. 14. Wearsch PA, et al. (2004) MHC class I molecules expressed with monoglucosylated N- 28. Zhang W, Wearsch PA, Yajuan Z, Leonhardt RM, Cresswell P (2011) A role for UDP- linked glycans bind calreticulin independently of their assembly status. J Biol Chem glucose glycoprotein glucosyltransferase in expression and quality control of MHC 279:25112–25121. class I molecules. Proc Natl Acad Sci USA:4956–4961. 15. Zhang Y, et al. (2009) ERp57 does not require interactions with calnexin and calreticulin 29. Trombetta ES, Simons JF, Helenius A (1996) Endoplasmic reticulum glucosidase II is to promote assembly of class I histocompatibility molecules, and it enhances peptide composed of a catalytic subunit, conserved from yeast to mammals, and a tightly loading independently of its redox activity. J Biol Chem 284:10160–10173. bound noncatalytic HDEL-containing subunit. J Biol Chem 271:27509–27516.

Wearsch et al. PNAS | March 22, 2011 | vol. 108 | no. 12 | 4955 Downloaded by guest on September 25, 2021