The redox activity of ERp57 is not essential forits functions in MHC class I peptide loadingDavid R. Peaper and Peter Cresswell†

Department of Immunobiology, Howard Hughes Medical Institute, Yale University School of Medicine, 300 Cedar Street, New Haven, CT 06520-8011

Contributed by Peter Cresswell, May 23, 2008 (sent for review April 23, 2008)

ERp57 is an oxidoreductase that, in conjunction with calnexin andcalreticulin, assists disulfide bond formation in folding glycopro-teins. ERp57 also forms a mixed disulfide with the MHC classI-specific chaperone tapasin, and this dimeric conjugate edits thepeptide repertoire bound by MHC class I molecules. In cells unableto form the conjugate, because of tapasin mutation in humanstudies or ERp57 deletion in mouse studies, peptide loading isimpeded. Subtle differences between the mouse and human sys-tems have been observed. Here, we address these differences andexpand the analysis to investigate the role of ERp57 redox func-tions in MHC class I peptide loading. We show in human cells thatin the absence of conjugate formation MHC class I recruitmentand/or stabilization in the MHC class I peptide-loading complex isimpaired, similar to observations in mouse cells. However, wefound no role for the enzymatic activities of either the a or a�

domain redox sites of ERp57 in peptide loading. Our data arguethat the function of ERp57 in peptide loading is likely caused byother ERp57 functional domains or a combinatorial feature of thetapasin–ERp57 conjugate.

antigen presentation � antigen processing � human � protein folding �quality control

S table loading of peptides onto MHC class I/�2-microglobulin(�2m) dimers requires coordinated action within the pep-

tide-loading complex (PLC), which consists of TAP1, TAP2,tapasin, ERp57, calreticulin (CRT), MHC class I heavy chain(HC), and �2m (1). Tapasin is a critical component of the PLC,but the in vivo role of ERp57 in peptide loading and editing isnot entirely resolved. ERp57 is an oxidoreductase that promotesproper disulfide bond formation in folding glycoproteinsthrough its association(s) with calnexin (CNX) and/or CRT (2).Like protein disulfide isomerase, ERp57 is composed of fourdomains with the a and a� domains containing redox activeCXXC motifs. During the biosynthetic folding of MHC class IHC, it appears to act in a manner consistent with models ofglycoprotein quality control (3). However, within the PLC,Cys-57 of ERp57 forms a disulfide bond with tapasin Cys-95, andtapasin inactivates the substrate dissociation step, or ‘‘escapepathway,’’ of ERp57, making this interaction very stable (4).

We examined MHC class I assembly in human B lymphoblas-toid cells expressing HLA-B*4402 and a tapasin construct inwhich Cys-95 was mutated to Ala (C95A) to prevent conjugateformation (5). PLC formation was qualitatively normal exceptfor the absence of ERp57, but the stability of peptide–MHC classI complexes assembled in these cells was decreased, consistentwith association with a pool of lower-affinity peptides. In mouseB cells lacking ERp57, surface expression of H2-Kb moleculeswas reduced by �50%, and turnover was faster than in ERp57-expressing cells. H2-Kb recruitment into the PLC was affected,and its trafficking through the Golgi was accelerated. Further-more, presentation of an H2-Kb-restricted epitope derived fromovalbumin was reduced in ERp57-deficient mouse B lympho-cytes (6). The tapasin C95A mutation did not affect H2-Kb

binding of this ovalbumin-derived peptide, but mutation ofCys-95 prevented the association of H2-Ld with tapasin in human

cells (7, 8). Thus, the relative importance of conjugate formationfor PLC assembly in mouse and human systems is ambiguous.

A critical question is how ERp57 redox activity is involved inpeptide loading. The a domain active site is stably disulfide-linked to tapasin, and this bond would have to be reversiblyreduced for this site to have a functional role. The a� domain sitemight potentially play a role, and we observed an altered redoxstate of HLA-B*4402 associated with the PLC in C95A tapasin-expressing cells, which appeared to be consistent with thishypothesis (5). However, MHC class I redox changes were notobserved in ERp57-deficient mouse B cells (6). A recent studyidentified a disulfide-linked complex of MHC class I HC,tapasin, and ERp57 in the PLC, and the authors suggested thatthe a� domain cysteines of ERp57 may be required for ‘‘tripleconjugate’’ formation (9). However, this hypothesis was notdirectly demonstrated, and cysteines in the transmembrane orcytoplasmic domains of tapasin and MHC class I HC couldmediate these interactions (ref. 10 and D.R.P. unpublishedobservations). Kienast et al. (11) recently proposed that conju-gate formation inhibits ERp57 redox activity, suggesting thatredox-active ERp57 might negatively affect peptide loading.These discrepancies are clearly in need of resolution.

Here, we reinvestigate human cells expressing C95A tapasin toreconcile our data with those obtained in the mouse andsubsequently examine the role of the two redox domains ofERp57 in peptide loading. Some aspects of peptide loadingdiffer between mice and humans, but conjugate formation isrequired for the efficient association of MHC class I with thePLC in human cells. After conjugate formation, ERp57 isirreversibly sequestered in the PLC by tapasin, arguing that theERp57 a domain does not directly function in peptide loading.Additionally, elimination of the redox activities of both the a anda� domain CXXC motifs does not affect peptide loading ontoHLA-B*4402, suggesting that the positive functions of ERp57 inpeptide loading are not related to its role as an oxidoreductase.

ResultsImpaired PLC Formation in Cells Expressing C95A Tapasin. To examinethe ability of C95A tapasin to recruit and/or stabilize MHC classI/�2m dimers, known PLC-associated proteins were detected byimmunoprecipitation and blotting of extracts of .220 cells express-ing HLA-B*4402 and either wild type (WT) or C95A tapasin (Fig.1A). TAP1 and tapasin levels were identical in cells expressing WTand C95A tapasin, but no ERp57 was associated with C95A-containing PLCs. Both CRT and MHC class I HC were recruitedto TAP1 by C95A tapasin, but PLCs from C95A-expressing cellscontained only 25% of the levels seen in WT-expressing cells. Thislevel could result from decreased recruitment, accelerated release,and/or loss during detergent solubilization. To determine the cause,

Author contributions: D.R.P. and P.C. designed research; D.R.P. performed research; D.R.P.and P.C. analyzed data; and D.R.P. and P.C. wrote the paper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.

†To whom correspondence should be addressed. E-mail: peter.cresswell@yale.edu.

© 2008 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0805044105 PNAS � July 29, 2008 � vol. 105 � no. 30 � 10477–10482

IMM

UN

OLO

GY

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

18, 2

020

we examined the kinetics of MHC class I interaction with the PLCby pulse–chase analysis (Fig. 1B). Throughout the chase, less MHCclass I HC was associated with PLCs containing C95A tapasindespite similar tapasin labeling in WT- and C95A-expressing cells.MHC class I associated with WT- and C95A-containing PLCs withsimilar kinetics, but the duration of the interaction was significantlyreduced in the absence of ERp57. Thus, the differences in steady-state MHC class I association with C95A-containing PLCs are atleast partially attributable to decreased retention. We cannotdistinguish between defects in MHC class I recruitment and post-solubilization loss of MHC class I, but the presence of ERp57 isrequired for stable formation and maintenance of the completePLC in human cells, consistent with results obtained in ERp57-deficient mouse B cells and for H2-Ld (6, 7).

Delayed Maturation of HLA-B*4402 in Cells Expressing C95A Tapasin.H2-Kb molecules assembled in ERp57-deficient B cells weremore rapidly eliminated from the cell surface, a marker ofinstability, than those assembled in ERp57-expressing cells (6).Similarly, the thermostability and half-life of HLA-B*4402–�2mcomplexes assembled in C95A-expressing cells were reducedcompared with WT-expressing cells (5). In ERp57-negativecells, H2-Kb complexes exited the PLC and traversed the Golgimore rapidly, but our original analysis of C95A-expressing cellsdid not assess rates of trafficking. Thus, we performed a pulse–chase analysis and harvested cells at 15-min intervals followed byimmunoprecipitation with the anti-MHC class I mAb W6/32,endoglycosidase H (EndoH) digestion, and SDS/PAGE (Fig. 2Upper). The acquisition of EndoH resistance by MHC class I/�2mdimers was significantly delayed in C95A-expressing cells (Fig. 2Lower Left), and fewer W6/32 complexes survived for the 4 h ofthe experiment (Fig. 2 Lower Right; WT t1�2

� 4 h, C95A t1�2�2

h). Those complexes that acquired EndoH resistance wererelatively long-lived, however (data not shown), suggesting thatpassing ER quality control checkpoints may correlate with theloading of higher-quality peptides.

Tapasin-Mediated Inhibition of a Domain Redox Activity SequestersERp57 in the PLC. The absence of ERp57 from the PLC affects theassociation of MHC class I with the PLC and the stability ofgenerated MHC class I/�2m dimers. ERp57 has several func-tional domains that could mediate these effects. To determinewhether the redox active sites of ERp57 are involved, we firstasked whether the a domain CXXC motif is active when ERp57is associated with tapasin. If it is active, exchange should occurbetween the free and tapasin-conjugated ERp57 pools in theER. Hence, in cells coexpressing FLAG-tagged C60A ERp57and endogenous ERp57, the elimination of the a domain escapepathway in the Cys-60 mutant should lead to the accumulationof C60A-FLAG ERp57 in the PLC over time. To test thishypothesis, we labeled and chased WT- or C60A-FLAG ERp57-expressing cells. The cells, not treated with methyl methaneth-iosulfonate (MMTS) unless indicated, were solubilized in digi-tonin and immunoprecipitated with PaSta1 before boiling inSDS sample buffer. Eluted material was then resolved by SDS/PAGE under reducing or nonreducing conditions (Fig. 3). As

MH

C c

lass

I H

Cto

Tap

asin

Rat

io

0

5

10

15

20

2400 60 120 18030 90Time (min.)

Tapasin NormalizedMHC class I HC

0 60 120 180 24030 90Time (min)

% Maximum TAP-associatedMHC class I HC

025

5075

100 WTC95A

A

0

25

50

75

100

% WT PLCComponents

TAP1

Tapa

sinCRT

MHC HC

ERp57

Rb.CRT

64

Rb.Tapasin48

64

Rt.MHC class I HC37

48

Rb.ERp5748

Rb.TAP1

80 kDa

IP TAP1 Ctrl 15 4.8 1.5 4.8 Cell Line WT 95 WT 95 WT 95 WT 95

Cell# (x106)

B IP Ctrl TAP1 Ctrl Time (min) 0 0 15 30 45 60 90 120 180 240 240

46 kDa Tapasin

46 kDa TapasinMHC class I HC{

MHC class I HC{

220.B4402.WT-Tapasin

220.B4402.C95A-Tapasin

Fig. 1. PLC formation is impaired in C95A-expressing cells. (A) (Left) Thesteady-state association of MHC class I HC and CRT is reduced in PLCs contain-ing C95A tapasin. WT and C95A-expressing .220.B*4402 cells were treatedwith MMTS, and digitonin lysates were immunoprecipitated (IP) with 148.3(TAP1) or B7/21 (Ctrl) coupled to beads. Precipitated material was resolved byreducing SDS/PAGE, and PLC components were detected by immunoblottingwith R.RING4C (TAP1), R.CRT (CRT), R.ERp57-C (ERp57), R.gp48N (tapasin), or3B10.7 (MHC class I HC) and secondary antibodies coupled to alkaline phos-phatase. A background band of lower mobility than MHC class I HC consis-tently present in 3B10.7 blots was excluded from the analysis. All bands werequantitated by fluorometry and were within the linear range of detection.(Right) The relative amount of PLC-associated proteins in C95A-expressingcells was calculated, and the mean � SEM of three dilutions from twoindependent experiments is shown. (B) MHC class I PLC association is alteredin C95A-expressing cells. Cells from A were pulse-labeled for 15 min andchased for the indicated periods of time before MMTS treatment. Digitoninlysates were immunoprecipitated with R.RING4C (TAP1) or NRS (Ctrl) andprotein A–Sepharose, and 0.1% Triton X-100 was used to release tapasin andassociated proteins. The eluted material was resolved by reducing SDS/PAGE,and the tapasin and MHC class I HC bands were quantitated. The mean � SEMof two independent experiments for the MHC class I HC/tapasin ratio (Left)and percentage of maximum TAP-associated MHC class I HC (Right) for WT (■ )and C95A (Œ) tapasin are shown. In both graphs the curves for WT and C95Atapasin are significantly different, as assessed by two-way ANOVA.

46 kDa EndoH Res.EndoH Sens.

46 kDa EndoH Res.EndoH Sens.

% EndoH Resistant

0 60 120 180 24030 900

25

50

75

100

Time (min)

WTC95A

% Max Total

0 60 120 180 24030 900

25

50

75

100

Time (min)

IP Ctrl W6/32 Ctrl Time (min) 0 0 15 30 45 60 90 120 180 240 240

220.B4402.WT-Tapasin

220.B4402.C95A-Tapasin

Fig. 2. MHC class I/�2m dimers mature more slowly and are less stable inC95A-expressing cells. (Upper) Cells were pulse-labeled and solubilized as inFig. 1B and MHC class I/�2m dimers were immunoprecipitated with W6/32 or51.1.3 (Ctrl) and protein A–Sepharose. Samples were digested with EndoHbefore reducing SDS/PAGE, and EndoH-resistant and -sensitive MHC class I HCwere quantitated. (Lower Left) The percentage of EndoH-resistant MHC classI HC was calculated, and the best-fit curves for WT (■ ) and C95A (Œ) tapasin areshown. The curves are significantly different (P � 0.0001). (Lower Right)Signals were further expressed as the percentage maximum MHC class I HC,and these curves were significantly different as assessed by two-way ANOVA.The dotted horizontal line corresponds to 50% of the maximum value.

10478 � www.pnas.org�cgi�doi�10.1073�pnas.0805044105 Peaper and Cresswell

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

18, 2

020

expected, in the cells expressing WT-FLAG ERp57, no conju-gate was seen without MMTS treatment after boiling in SDSwithout reduction (lanes 2, 4, 6, 8, and 10), but in cells where theescape pathway was blocked by MMTS treatment, the conjugatewas preserved (lane 13). In contrast, cells expressing bothendogenous ERp57 and C60A-FLAG ERp57 had substantialamounts of detectable conjugate in the absence of MMTStreatment (lanes 2, 4, 6, 8, and 10). Conjugates containingC60A-FLAG ERp57 are resistant to escape pathway-mediatedconjugate reduction after denaturation in SDS (D.R.P., unpub-lished observation), and the constant ratio of free to conjugatedtapasin argues that C60A-ERp57 does not accumulate in theloading complex with time (Lower). Thus, regulated conjugatereduction through the a domain CXXC motif and ERp57exchange do not appear to occur during peptide loading. Onceincorporated into the MHC class I loading complex, ERp57 ispermanently sequestered through inactivation of its a domainredox activity.

Generation of Cells Expressing ERp57 Redox Mutant Conjugates. Toprobe further the function of the a and a� domain CXXC motifsin peptide loading, we knocked down endogenous ERp57 byusing shRNA constructs and then reexpressed FLAG-taggedWT and mutant ERp57. Cells expressing HLA-B*4402 and WTtapasin were transduced with an ERp57-specific shRNA retro-virus, and, after sorting, �90% knockdown was achieved com-pared with cells transduced with nontargeting shRNA constructs(Fig. 4A). Despite this knockdown, only subtle differences wereseen in PLC composition and MHC class I trafficking in thesecells (data not shown). These differences are likely because allof the residual ERp57 is conjugated to tapasin, and some tapasinremains in the conjugate (Fig. 4 B and C). These data reinforce

our previous finding that ERp57 is recruited preferentially bytapasin in IFN-�-stimulated cells (4). We next wished to reex-press WT or mutant ERp57 to examine their effects on MHCclass I loading. ERp57-suppressed cells were transduced withretroviruses encoding FLAG-tagged WT, C60A mutant, ortriple cysteine mutant (C60A/C406A/C409A) ERp57 biscistroni-cally with EGFP and sorted for high EGFP expression. TheC60A/C406A/C409A triple redox mutant (3x) combined thetrapping feature that inactivates the a domain active site andmutations that inactivate the a� domain active site. Mutation ofCys-406 and Cys-409 prevents conjugate formation, but the

C60A-associated Tapasindoes not accumulate in the PLC

0 2.5 5 7.5 1010

25

50

75

100

Time (hr)

% N

orm

aliz

edF

ree

Tapa

sin

MMTS (-/+) - + IP Ctrl Tapasin Ctrl Tpsn Time (hr) 0 1 2.5 5 10 1 DTT (-/+) - - + - + - + - + - + + - +

C1R.A2.WT

97 kDa

66

46

97 kDa

66

46

37C1R.A2.C60A

Lane 1 2 3 4 5 6 7 8 9 10 11 12 13 14

Conjugate

Tapasin

Conjugate

Tapasin

Fig. 3. ERp57 is sequestered in the conjugate. The conjugate does notundergo escape pathway mediated reduction. C1R cells expressing HLA-A*0201 and FLAG-tagged WT or C60A ERp57 were pulse-labeled for 45 minand chased for the indicated times before treatment with or without MMTSas indicated. Digitonin-solubilized cells were immunoprecipitated withPaSta1 (tapasin)- or Ox-68 (Ctrl)-coupled beads, and precipitated material wasresolved by SDS/PAGE under reducing or nonreducing conditions. Bandscorresponding to tapasin were quantitated, and the amount of free tapasinunder nonreducing conditions was normalized to the total amount of tapasinpresent after reduction. The chart depicts the mean � SEM of two indepen-dent experiments.

97 kDa6646

30

Conjugate

TapasinERp57

MHC class I HC

1st IP TAP1 Ctrl 2nd IP FLAG DTT (+/-) - + + ERp57-FLAG WT 60 3x WT 60 3x WT 60 3x

E

B RVH1 UT ERp57 Ctrl DTT (+/-) - + - + - +

113 kDa80

63

48

M.GAPDH

37

24

Rb.ERp57

C

Rb.Tapasin

113 kDa80

63

48

24

37

RVH1 UT ERp57 Ctrl DTT (+/-) - + - + - +

M.GAPDH

D

M.ERp57 M.FLAG

M.GAPDH M.GAPDH

64 kDa

4837

26

ERp57 RNAi - + + + + - + + + + ERp57-FLAG - - WT 60 3x - - WT 60 3x

66 kDa48

109

80Rt.GRP94

Rb.ERp57

Protein (µg) 15 4.75 1.5 ERp57 RNAi - + - + - +

A

Fig. 4. Successful stable knockdown of ERp57. (A) Generation of ERp57-suppressed cell lines. 721.220 cells expressing HLA-B*4402 and WT tapasintransduced with ERp57-specific or control shRNA sequences were solubilizedin 1% Triton X-100. Serial dilutions of ERp57 suppressed (�RNAi) or control cellextracts (�RNAi) were resolved by reducing SDS/PAGE. Proteins were detectedby immunoblotting with rabbit anti-ERp57 (R.ERp57-C) and monoclonal ratanti-GRP94 antibody as a loading control. (B) All remaining ERp57 is conju-gated to tapasin. Cells from A or untransduced cells (UT) were treated with 10mM MMTS and solubilized in 1% Triton X-100. Postnuclear supernatants wereresolved by reducing/nonreducing SDS/PAGE. Proteins were detected by im-munoblotting with rabbit anti-ERp57-C or mouse anti-GAPDH. (C) Conjugatelevels are slightly decreased in ERp57-suppressed cells. Cell lysates were pre-pared as in B, but blots were probed with rabbit anti-tapasin serum (R.SinE) ormouse anti-GAPDH. (D) Expression of FLAG-tagged ERp57. 721.220 cells ex-pressing HLA-B*4402 and WT tapasin with or without ERp57 RNAi and ex-pressing FLAG-tagged WT, C60A, or C60A/C406A/C409A (3x) ERp57 weresolubilized in 1% Triton X-100, and postnuclear supernatants were resolved byreducing SDS/PAGE. Total and FLAG-tagged ERp57 were detected by immu-noblotting with MaP.ERp57 and M2 (FLAG). Mouse anti-GAPDH was used asa loading control. (E) FLAG-tagged ERp57 is incorporated into the PLC andforms the conjugate. Cells from D were pulse-labeled for 30 min and chasedfor 30 min. Digitonin lysates of MMTS-treated cells were immunoprecipitatedwith RING4c (TAP1) or NRS (Ctrl). Triton X-100 (0.1%) was used to releasetapasin from TAP, and subcomplexes incorporating FLAG-tagged ERp57 wereimmunoprecipitated with M2-agarose (FLAG). Samples were resolved by re-ducing/nonreducing SDS/PAGE as indicated. FLAG-associated PLC componentsare indicated.

Peaper and Cresswell PNAS � July 29, 2008 � vol. 105 � no. 30 � 10479

IMM

UN

OLO

GY

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

18, 2

020

additional mutation of Cys-60 allows conjugation formation tooccur (5). As shown in Fig. 4D, ERp57 expression in thetransduced cells was comparable to endogenous ERp57. Allthree ERp57 constructs, including the triple mutant, wereincorporated into the PLC and formed the conjugate (Fig. 4E),and virtually all detectable conjugated ERp57 in these cells wasFLAG-tagged (Fig. 5A and data not shown). Thus, this systemallowed us to specifically address the role of the redox activitiesof the ERp57 a and a� domains in MHC class I peptide loading.

Normal MHC Class I Assembly in the Absence of a and/or a� DomainRedox Activity. The three most pronounced phenotypes in cellsexpressing HLA-B*4402 and C95A tapasin are: impaired incor-poration of MHC class I in the PLC, decreased stability ofassembled MHC class I complexes, and an altered redox state ofMHC class I HC in the PLC (5). Because poor incorporation/stabilization of MHC class I into the PLC likely underlies thedefects seen in peptide loading in conjugate-deficient cells, wefirst examined whether elimination of the redox activity ofERp57 adversely affected PLC stability. Interactions betweenPLC components are preserved in the detergent digitonin, butincubation of TAP1 immunoprecipitates with Triton X-100releases subcomplexes containing different combinations oftapasin, ERp57, CRT, MHC class I HC, and/or �2m. Reprecipi-tation of these eluates from cells expressing WT tapasin withPaSta1 isolates tapasin, ERp57, and associated MHC class Icomplexes (Fig. 5A Left). When subcomplexes were isolatedfrom Triton X-100 eluates of PLCs from C95A-expressing cells,tapasin is precipitated without ERp57, and MHC class I com-plexes are lost during biochemical isolation (Fig. 5A Left). Thus,the interaction of MHC class I with unconjugated tapasin isdestabilized in Triton X-100, and this correlates with the differ-

ences seen at steady-state and in pulse–chase shown in Fig. 1.When the same technique was applied to cells expressing WT,C60A, or C60A/C406A/C409A ERp57, MHC class I remainedassociated with tapasin after incubation in Triton X-100 (Fig. 5ARight). We further confirmed that MHC class I was associatedwith the exogenously expressed FLAG-tagged ERp57 by pre-cipitating with an anti-FLAG antibody. Comparable amounts oftapasin, ERp57, and MHC class I HC were precipitated whetherWT, C60A, or the triple mutant versions of ERp57 were present(Fig. 5A Right). Thus, elimination of the redox activity of ERp57does not affect the stabilization of MHC class I with the PLC,and this function of ERp57 is likely attributable to other ERp57domains or to a combinatorial property of tapasin and ERp57.

Although the interaction of MHC class I with the mutantconjugates is the same as with WT conjugates, there could beeffects of the redox domains on the generation of stable MHCclass I/peptide complexes. Surface staining of HLA-B*4402 incells expressing exogenous WT and redox mutant ERp57 wassimilar (data not shown), but elimination of conjugation bymutation of Cys-95 only decreased HLA-B*4402 surface expres-sion by 50% (5). The most pronounced defects in MHC class Istability in C95A-expressing cells are seen by pulse–chase anal-ysis in both short-term (Fig. 2) and long-term assays (5).However, as seen in Fig. 5B, the long-term (24-h) stability ofHLA-B*4402 complexes assembled in cells predominantly ex-pressing redox mutant ERp57 was not significantly differentfrom those assembled in cells expressing WT-FLAG ERp57.Thus, elimination of the a� and/or a domain CXXC motifs ofERp57 does not adversely affect the generation of stable MHCclass I/peptide dimers.

We identified an alteration in the redox state of HLA-B*4402associated with C95A tapasin, and this finding has recently been

1st IP TAP1 2nd IP Tapasin Ctrl Cell Line WT 95 WT 95

66 kDa

48

ERp57

TapasinMHC HC

TAP1 Tapasin FLAG Ctrl WT 60 3x WT 60 3x WT 60 3x

Mutant Tapasin ERp57-FLAG

D TAP1 X TAP1 Ctrl WT 60 3x X WT 60 3x WT 60 3x - DTT X + DTT + DTT

Rt.MHC class I HC

IP Cell Line DTT

48 kDa

37

BC60AC60,406,409A

WT

% M

axim

um W

6/32

Rea

ctiv

e M

HC

HC

Chase Time (hr)

MHC Class I/β2m Stability

0 8 16 2412 40

25

50

75

100

Lane # 1 2 3 4 5 6 7 8 9

IP TAP1 X TAP1 Ctrl TAP1 Cell Line WT 95 X WT 95 WT 95 WT 95 DTT - DTT X + DTT + DTT - DTT

48 kDa

37

∆Ox

Rt.MHC class I HC MHC HC

A

C

C60,406,409A-FLAG

Fig. 5. Absence of ERp57 redox activity does not affect conjugate function. (A) MHC class I dissociates from unconjugated but not WT or mutantERp57-conjugated tapasin. Cells as in Figs. 1 and 2 (Left) or Fig. 4E were pulse-labeled as in Fig. 4E. PLCs were precipitated from digitonin lysates of MMTS-treatedcells with RING4c (TAP1), and Triton X-100-released material was immunoprecipitated with PaSta1 (tapasin), M2 (FLAG), or Ox-68 (Ctrl) coupled to agarose beadsand subjected to SDS/PAGE. Associated PLC components are indicated. (B) MHC class I stability is not affected by the loss of ERp57 redox activity. Cells in A werepulse-labeled for 15 min and chased for the indicated periods of time. MHC class I HC/�2m dimers were immunoprecipitated from digitonin lysates with W6/32or 51.1.3 (Ctrl) and digested overnight with EndoH. (Left) Samples were resolved by reducing SDS/PAGE and quantitated. (Right) The percentage maximumW6/32-reactive MHC class I HC was calculated, and the mean � SEM of two independent experiments is shown. (C) A small population of PLC-associated MHCclass I HC is partially oxidized in C95A-expressing cells. Cells expressing WT or C95A tapasin from A were treated with MMTS, and PLCs were immunoprecipitatedfrom digitonin lysates by using 148.3 (anti-TAP1) coupled to beads. B7/21 (anti-HLA-DP mAb) coupled to beads was used as a control (Ctrl). Triton X-100-elutedmaterial was resolved by reducing/nonreducing SDS/PAGE, and MHC class I HC was detected by immunoblotting with 3B10.7. Lanes 8 and 9 were overexposedto emphasize the partially oxidized MHC class I HC band (�OX) associated with C95A-containing PLCs. (D) PLC-associated MHC class I HC is fully oxidized in thepresence of redox-inactive ERp57. Cells expressing WT or mutant ERp57 from A were treated as described in C.

10480 � www.pnas.org�cgi�doi�10.1073�pnas.0805044105 Peaper and Cresswell

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

18, 2

020

expanded (5, 11). There are several possible explanations forthese observations, one being that the a� domain CXXC motifmay be required to reoxidize MHC class I HC that becomesreduced during peptide loading. Thus, we wished to examine theredox state of HLA-B*4402 HCs associated with ERp57 mutant-containing PLCs. First, we confirmed that redox changes wereapparent in MHC class I HCs in cells expressing C95A tapasin.In contrast to the previous results, only a small fraction ofPLC-associated MHC class I HC in C95A-expressing cells waspartially reduced (Fig. 5C, lanes 2 and 9). We have no explana-tion for this difference, but our present results are more similarto those of Kienast et al. (11) than our original observation.However, the absence of the a� domain active site did not affectthe redox status of PLC-associated HLA-B*4402 HCs (Fig. 5D).Taken together, these results indicate that the function(s) ofERp57 within the PLC are largely if not entirely independent ofits role as a redox enzyme. Instead, ERp57 likely plays anundefined structural role in recruiting MHC class I complexesinto the PLC through its conjugation with tapasin and/or non-redox functions such as interactions with CRT and CNX.

DiscussionConsistent with observations in ERp57-negative mouse B cells,our data argue that tapasin-associated ERp57 is required for theefficient incorporation of MHC class I/�2m dimers into the PLC(6). The use of a sensitive quantitative approach likely explainsour ability to demonstrate a decrease in PLC association of MHCclass I HC in C95A-expressing cells not revealed in previousstudies (5, 11). How conjugate formation leads to MHC class Irecruitment and stabilization is unclear. ERp57 has three knownfunctional elements: two redox active sites, in the a and a�domains, and a CNX/CRT-interacting site in the b and b�domains (12). However, the a domain redox site is inactivated bytapasin association, and we showed here that there is no ex-change of ERp57 molecules conjugated through that site. Wefurther showed that complete inactivation of the thiol exchangecapacity of the a domain site by mutation of the C-terminalcysteine residue combined with the elimination of the a� domainactive site does not affect MHC class I incorporation into thePLC. Thus, the redox activity of PLC-associated ERp57 plays norole in MHC class I recruitment. Furthermore, the expressionand stability of the MHC class I molecules assembled in thepresence of the ERp57 triple mutant are indistinguishable fromthose assembled in the presence of WT ERp57. Thus, the redoxactivity also appears to play no role in selecting high-affinitypeptides or peptide editing. CNX/CRT binding is the only knownfunctional property of ERp57 that might mediate a direct effecton MHC class I recruitment and peptide loading. Interaction oftapasin-conjugated ERp57 with CRT that is simultaneouslybound to the monoglucosylated N-linked glycan of the class I HCmay explain why recombinant conjugate can recruit MHC classI/�2m dimers and function as a peptide editor whereas freetapasin cannot (13). Consistent with this, free tapasin acquiresthe ability to mediate peptide exchange when artificially tetheredto MHC class I HC (14).

A substantial fraction of MHC class I-�2m dimers assembledin C95A-expressing cells do not exit the ER, likely because ofinefficient recruitment and peptide loading. Those complexeswith bound peptide exit the ER and pass through the Golgi, butthis population is less stable than the pool that exit the ER in WTcells (5). In contrast, most H2-Kb complexes assembled inERp57-deficient mouse B cells successfully exit the ER butdissociate in post-ER compartments (6). The reasons for thedifferences in trafficking between HLA-B*4402 and H2-Kb areunknown, but to exit the ER proteins must pass ER qualitycontrol checkpoints. These alleles may differ in their intrinsicability to pass these checkpoints. Consistent with this possibility,peptide-free H2-Kb molecules expressed in human TAP-

deficient cells accumulate at the cell surface whereas TAP-dependent human class I alleles in the same cell do not (15). Inaddition, surface-expressed H2-Kb molecules in mouse TAP-deficient cells can be stabilized by exogenous �2m or tempera-ture reduction, but neither �2m nor lower temperatures stabilizeTAP-dependent human alleles in either mouse or humancells (16).

Kienast et al. (11) recently suggested that the function of theconjugate is to inhibit ERp57 reductase activity toward MHCclass I HC. They observed the tapasin-dependent allele HLA-B*4402 in a partially oxidized state in cells lacking tapasin orthose expressing C95A tapasin, whereas the tapasin-indepen-dent allele HLA-B*4405 was fully oxidized under these condi-tions. Most of their experiments did not distinguish betweenpools of MHC class I molecules (e.g., newly synthesized, PLC-associated, ER-retained and targeted for degradation, surface-expressed, etc.) We observed a small population of partiallyoxidized HLA-B*4402 HCs associated with the PLC in C95A-expressing cells, but much more profound oxidation differenceswere seen in non-PLC-associated MHC class I HC in the ER(D.R.P., unpublished observations). MHC class I HCs arepartially reduced shortly after synthesis and immediately beforeER-associated degradation (ERAD), and ERp57 contributes toreduction under both circumstances (3, 17). Mutation of eitherCys-101 or Cys-164 in the MHC class I HC peptide-bindinggroove substantially inhibits peptide loading (11, 18). Con-versely, poor peptide loading leads to reduction of that disulfidebond and ERAD. Because C95A tapasin is less efficient atcatalyzing peptide loading (5, 19), it is difficult to determinewhether nonsequestered ERp57 might inhibit peptide loading byreducing MHC class I molecules or whether reduction occursbecause C95A tapasin is inefficient in mediating peptide loading.Generation of tapasin mutants that do not catalyze peptideloading but form the conjugate could resolve this question.Finally, the model is at best incomplete because purified con-jugate can promote peptide loading and act as a peptide editorin a cell free assay to a much greater extent than free tapasin(13). This would not be so if the sole purpose of conjugation wasto inhibit ERp57 activity toward MHC class I HC.

Soluble conjugate, but not tapasin alone, acts as a peptideeditor and exerts a positive function on MHC class I peptideloading (13). Here, we showed that those effects are not theresult of a� domain redox activity. Additionally, the steady-stateinhibition of the ERp57 a domain escape pathway we describedtranslates into the permanent sequestration of ERp57 in theconjugate. This strongly suggests that redox regulation by ERp57does not have a role in MHC class I peptide loading independentof its general role in glycoprotein folding, but the positivefunction of ERp57 in the PLC remains elusive. H2-Kb traffickingin CRT-deficient fibroblasts is similar to that seen in the absenceof ERp57 (20), consistent with the idea that interaction of the bb�domains of ERp57 with CRT could stabilize MHC class I in thePLC. However, more experiments are needed to evaluate theimportance of the ERp57–lectin interactions in MHC class Iloading.

Materials and MethodsPlasmids. Retroviral vectors encoding shRNA constructs targeting ERp57 (RVH1-ERp57) or a control (RVH1-Ctrl) were constructed by using the RVH1 vector (21).The following oligonucleotides synthesized with 5�-phosphates were annealedand ligated as described: ERp57-F, 5�-gat ccc cGG ACT CTT CCA TCA GAG ATt tcaaga gaA TCT CTG ATG GAA GAG TCC ttt ttg gaa c-3�; ERp57-R, 5�-tcg agt tcc aaaaaG GAC TCT TCC ATC AGA GAT tct ctt gaa ATC TCT GAT GGA AGA GTC Cgg g-3�;Ctrl-F,5�-gatccccGCTTCAACAGCAGGCACTCttcaagagaGAGTGCCTGCTGTTGAAG Ctt ttt gga ac-3�; Ctrl-R, 5�-tcg agt tcc aaa aaG CTT CAA CAG CAG GCA CTCtct ctt gaa GAG TGC CTG CTG TTG AAG Cgg g-3�. Correct clones were identifiedby sequencing. Retroviral vectors encoding WT, C60A, and C60A/C406A/C409AFLAG-tagged ERp57 were generated by ligating fragments from ERp57-FLAG in

Peaper and Cresswell PNAS � July 29, 2008 � vol. 105 � no. 30 � 10481

IMM

UN

OLO

GY

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

18, 2

020

pCDNA3.1-Puro (5) into the retroviral vector pBMN-IRES-EGFP (a gift of A. Both-well, Yale University, New Haven, CT).

Cell Lines and Antibodies. The cell line 721.220 and its transfectants have beendescribed in ref. 5. Cells were first transduced with the RVH1 retroviral shRNAconstructs by spinfection and sorted for high CD4 expression with a FACSVantage SE (22). Cells with stably suppressed ERp57 were then transducedwith pBMN-ERp57-IRES-EGFP, and EGFP-positive cells were collected as above.All derivatives of 721.220 were maintained as described in ref. 5. The mAbsused were: 3B10.7 [anti-MHC class I HC (5)], 148.3 [anti-TAP1 (5)], W6/32[anti-HLA-A, B, C (5)], 51.1.3 [anti-CD1d (23)], BB7.2 [anti-HLA-A2 (24)], PaSta1[anti-tapasin (5)], M2 (anti-FLAG; Sigma), 6C5 (anti-GAPDH; Research Diag-nostics), SPA-850 (anti-GRP94; Stressgen), and B7/21 [anti-HLA-DP (25)]. Rabbitantiserum against TAP1 [R.RING4C (5)], CRT (PA3-900; ABR-Affinity Biore-agents), ERp57 [R.ERp57-C (4)], and tapasin [R.gp48N (26)] were also used. Therabbit antiserum R.SinE was raised against soluble recombinant tapasin.

Pulse–Chase Analyses and Immunoprecipitation. Pulse–chase analyses wereperformed as described in ref. 4. For all experiments, starved cells werepulse-labeled with 1 mCi of [35S]Met/Cys labeling mix (ICN or PerkinElmer) per20 106 cells for the indicated times. Harvested, labeled cells were washedonce with PBS with or without 10 mM MMTS (Pierce) as indicated and frozen.Cell solubilization in 1% digitonin and preclearing were performed as de-scribed in ref. 4. Proteins were immunoprecipitated by incubating with mAbscoupled to Bio-Gel A15m beads or mAbs or antiserum and protein A–Sepha-rose for 1 h at 4°C followed by washing in 0.1% digitonin. For EndoH diges-

tions, beads were heated to 95°C in 2 EndoH buffer (0.05 M sodium phos-phate, 0.25% SDS, pH 6.5) for 5 min. Eluted material was digested with 1milliunit of EndoH (Roche) overnight at 37°C before reducing SDS/PAGE. TAP1immunoprecipitates were eluted in 0.1% Triton X-100 (American Bioanalyti-cal) for 5 min on ice, and SDS/PAGE sample buffer was added to elutedmaterial. All other immunoprecipitates were eluted directly into SDS/PAGEsample buffer. Samples were resolved by reducing or nonreducing SDS/PAGEas described in ref. 4. Quantitation was performed with ImageQuant software(GE Healthcare).

Immunoprecipitation and Western Blotting. Cells were treated with 10 mMMMTS in PBS and extracted as described above. Postnuclear supernatantswere precleared with protein A–Sepharose and precipitated with mAb cou-pled to Bio-Gel A15m beads for 1 h at 4°C. After washing, precipitated materialwas eluted in 2 reducing SDS/PAGE sample buffer at 95°C for 5 min or with0.1% Triton X-100 for reducing/nonreducing SDS/PAGE as described above.After transfer, membranes were blocked, probed, and washed, and proteinswere detected as described in ref. 4 for quantitative and nonquantitativeblots.

ACKNOWLEDGMENTS. We thank Nancy Dometios for aiding in the prepara-tion of this manuscript and our colleague Dr. Pamela Wearsch for reagents andhelpful discussions. This work was supported by the Howard Hughes MedicalInstitute (to P.C.) and the National Institutes of Health/National Institute ofGeneral Medical Sciences Medical Scientist Training Grant GM07205 (toD.R.P.).

1. Garbi N, Tanaka S, van den Broek M, Momburg F, Hammerling GJ (2005) Accessorymolecules in the assembly of major histocompatibility complex class I/peptide com-plexes: How essential are they for CD8(�) T cell immune responses? Immunol Rev207:77–88.

2. Helenius A, Aebi M (2004) Roles of N-linked glycans in the endoplasmic reticulum.Annu Rev Biochem 73:1019–1049.

3. Zhang Y, Baig E, Williams DB (2006) Functions of ERp57 in the folding and assembly ofmajor histocompatibility complex class I molecules. J Biol Chem 281:14622–14631.

4. Peaper DR, Wearsch PA, Cresswell P (2005) Tapasin and ERp57 form a stable disulfide-linked dimer within the MHC class I peptide-loading complex. EMBO J 24:3613–3623.

5. Dick TP, Bangia N, Peaper DR, Cresswell P (2002) Disulfide bond isomerization and theassembly of MHC class I–peptide complexes. Immunity 16:87–98.

6. Garbi N, Tanaka S, Momburg F, Hammerling GJ (2006) Impaired assembly of the majorhistocompatibility complex class I peptide-loading complex in mice deficient in theoxidoreductase ERp57. Nat Immunol 7:93–102.

7. Turnquist HR, et al. (2004) The Ig-like domain of tapasin influences intermolecularinteractions. J Immunol 172:2976–2984.

8. Howarth M, Williams A, Tolstrup AB, Elliott T (2004) Tapasin enhances MHC class Ipeptide presentation according to peptide half-life. Proc Natl Acad Sci USA 101:11737–11742.

9. Santos SG, et al. (2007) Major histocompatibility complex class I-ERp57–tapasin inter-actions within the peptide-loading complex. J Biol Chem 282:17587–17593.

10. Chambers JE, Jessop CE, Bulleid NJ (2008) Formation of a major histocompatibilitycomplex class I tapasin disulfide indicates a change in spatial organization of thepeptide-loading complex during assembly. J Biol Chem 283:1862–1869.

11. Kienast A, Preuss M, Winkler M, Dick TP (2007) Redox regulation of peptide receptivityof major histocompatibility complex class I molecules by ERp57 and tapasin. NatImmunol 8:864–872.

12. Kozlov G, et al. (2006) Crystal structure of the bb� domains of the protein disulfideisomerase ERp57. Structure 14:1331–1339.

13. Wearsch PA, Cresswell P (2007) Selective loading of high-affinity peptides onto majorhistocompatibility complex class I molecules by the tapasin-ERp57 heterodimer. NatImmunol 8:873–881.

14. Chen M, Bouvier M (2007) Analysis of interactions in a tapasin/class I complex providesa mechanism for peptide selection. EMBO J 26:1681–1690.

15. Wei ML, Cresswell P (1992) HLA-A2 molecules in an antigen-processing mutant cellcontain signal sequence-derived peptides. Nature 356:443–446.

16. Anderson KS, Alexander J, Wei M, Cresswell P (1993) Intracellular transport of class IMHC molecules in antigen-processing mutant cell lines. J Immunol 151:3407–3419.

17. Antoniou AN, et al. (2002) The oxidoreductase ERp57 efficiently reduces partiallyfolded in preference to fully folded MHC class I molecules. EMBO J 21:2655–2663.

18. Warburton RJ, et al. (1994) Mutation of the �2 domain disulfide bridge of the class Imolecule HLA-A*0201: Effect on maturation and peptide presentation. Hum Immunol39:261–271.

19. Hughes EA, Hammond C, Cresswell P (1997) Misfolded major histocompatibility com-plex class I heavy chains are translocated into the cytoplasm and degraded by theproteasome. Proc Natl Acad Sci USA 94:1896–1901.

20. Gao B, et al. (2002) Assembly and antigen-presenting function of MHC class I moleculesin cells lacking the ER chaperone calreticulin. Immunity 16:99–109.

21. Barton GM, Medzhitov R (2002) Retroviral delivery of small interfering RNA intoprimary cells. Proc Natl Acad Sci USA 99:14943–14945.

22. Lackman RL, Cresswell P (2006) Exposure of the promonocytic cell line THP-1 toEscherichia coli induces IFN-�-inducible lysosomal thiol reductase expression by inflam-matory cytokines. J Immunol 177:4833–4840.

23. Exley M, Garcia J, Balk SP, Porcelli S (1997) Requirements for CD1d recognition byhuman invariant V�24�CD4�CD8� T cells. J Exp Med 186:109–120.

24. Parham P, Brodsky FM (1981) Partial purification and some properties of BB7.2: Acytotoxic monoclonal antibody with specificity for HLA-A2 and a variant of HLA-A28.Hum Immunol 3:277–299.

25. Royston I, Omary MB, Trowbridge IS (1981) Monoclonal antibodies to a human T cellantigen and Ia-like antigen in the characterization of lymphoid leukemia. TransplantProc 13:761–766.

26. Lehner PJ, Surman MJ, Cresswell P (1998) Soluble tapasin restores MHC class I expres-sion and function in the tapasin-negative cell line .220. Immunity 8:221–231.

10482 � www.pnas.org�cgi�doi�10.1073�pnas.0805044105 Peaper and Cresswell

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

18, 2

020