Consequences of ERp57 Deletion on Oxidative Folding of Obligate and Facultative Clients of the Calnexin Cycle*

Members of the protein-disulfide isomerase superfamily catalyze the formation of intra- and intermolecular disulfide bonds, a rate-limiting step of protein folding in the endoplasmic reticulum (ER). Here we compared maturation of one obligate and two facultative calnexin substrates in cells with and without ERp57, the calnexin-associated, glycoprotein-specific oxidoreductase. ERp57 deletion did not prevent the formation of disulfide bonds during co-translational translocation of nascent glycopolypeptides in the ER. It affected, however, the post-translational phases of oxidative influenza virus hemagglutinin (HA) folding, resulting in significant loss of folding efficiency for this obligate calnexin substrate. Without ERp57, HA also showed reduced capacity to recover from an artificially induced aberrant conformation, thus revealing a crucial role of ERp57 during post-translational reshuffling to the native set of HA disulfides. ERp57 deletion did not affect maturation of the model facultative calnexin substrates E1 and p62 (and of most cellular proteins, as shown by lack of induction of ER stress). ERp72 was identified as one of the ER-resident oxidoreductases associating with the orphan ERp57 substrates to maintain their folding competence.

Deletion of individual members of the calnexin chaperone system, namely calreticulin (6), UGT1 (7), and ERp57 (8), is embryonic lethal in mice. Deletion of calnexin does not result in embryonic lethality but causes a severe progressive pathology leading to premature death (9). The suboptimal glycoprotein folding efficiency upon chaperone deletion might be detrimental for organism viability, but all of these deletions are well tolerated at the cellular level. In fact, only a restricted number of endogenous, recombinant, or virus-encoded glycoproteins (e.g. the major histocompatibility complex class I peptide loading complex (8,10), the influenza virus HA (11), and the ADAM1/ADAM2 fertilin complex (12)) have so far been shown to strongly depend on the calnexin chaperone system for maturation. Consistently, and as previously shown for calnexin, calreticulin, and UGT1 deletions, cells lacking ERp57 show normal viability and proliferation rates and unperturbed maturation and transport of several surface glycoproteins containing disulfide bonds (8). Conditional deletion of ERp57 in B lymphocytes/ plasma cells revealed no defect in the production of immunoglobulin chains, the most abundant N-glycosylated/disulfide-bonded products of these cells (8). The thorough analysis by Garbi et al. showed that ERp57 deletion specifically affected assembly and stability of the major histocompatibility class I peptide-loading complex. These results are consistent with the important role of calnexin, calreticulin, and ERp57 in the biogenesis of this multimeric complex (reviewed in Refs. [13][14][15] and with the distinction we recently introduced between facultative versus obligate clients of the calnexin/calreticulin/ERp57 triad of chaperones predicting that most, but not all, cellular, viral, and recombinant glycopolypeptides may adopt surrogate chaperones to fulfill productive maturation (discussed in Ref. 16).
To better understand the differences between obligate and facultative clients of the calnexin chaperone system and to specifically assess the intervention of ERp57 in the co-and post-translational phases of oxidative folding for these two classes of glycoproteins, we compared in cells with (wt) or without ERp57 (57 Ϫ/Ϫ ) (Fig. 1A) the maturation of one obligate (influenza virus HA) and two facultative (Semliki forest virus (SFV) E1 and p62) clients of the calnexin chaperone system. HA was selected for its exquisite dependence on calnexin for efficient maturation (11,16,17). E1 and p62 were selected as representatives of the more numerous facultative clients of the calnexin cycle that can reach native conformation by exploiting undefined alternative folding pathways (11,16).
The ER hosts more than 20 members of the PDI superfamily. Their role in protein homeostasis is not fully understood. We present here the first analysis of oxidative protein folding in cells lacking one ER-resident oxidoreductase. We show that ERp57 activity is dispensable for cotranslational oxidation of cysteines in the nascent model glycoproteins. As a clear distinction between the facultative and the obligate N-glycosylated clients of the calnexin chaperone system, ERp57 was required only for the latter, for the post-translational reshuffling of intramolecular disulfides leading to acquisition of the native conformation.
We show that members of the PDI superfamily may have overlapping and interchangeable activities in the living cell. Accordingly, for glycoproteins that successfully completed the oxidative folding process in cells lacking the dedicated oxidoreductase ERp57, at least one surrogate PDI family member, namely ERp72, proficiently replaced ERp57 upon deletion and upon complete inactivation of the calnexin cycle in cells exposed to the glucosidase inhibitor castanospermine.

EXPERIMENTAL PROCEDURES
Cell Lines-ERp57 deletion is embryonic lethal for mice (8). To generate an ERp57-deficient cell line, fibroblasts were first prepared from ERp57 flox/flox mice, which express wild-type levels of ERp57 protein. A cloned line immortalized with SV40 large T antigen (referred to as wt) was then transfected with a plasmid coding for Cre recombinase to obtain stable ERp57 knock-out fibroblasts (referred to as 57 Ϫ/Ϫ ). Levels of ERp57 protein were reduced by stably co-transfecting cells with plasmids coding for small interfering RNA specific for three different ERp57 mRNA regions. A clone showing stable ϳ80 -90% reduction in ERp57 protein expression was used for these studies (referred to as si57).
Infections and Sample Analysis-Cell infection with the X-31 strain of influenza virus or with SFV, pulse-labeling, immunoprecipitation, and endoglycosidase H digestion are described in Refs. 11 and 18. N-Ethylmaleimide (20 mM) or methyl methanethiosulfonate were used as cell-permeable alkylating agents to prevent postlysis disulfide bond formation and isomerization. Quantifications were by ImageQuant software (Amersham Biosciences) as in Ref. 19. For figures, gels were exposed to Biomax films (Eastman Kodak Co.) and scanned with an Agfa scanner. In Figs. 3 and 6, selected experiments are shown in duplicate or triplicate. Analysis of co-translational formation of intra-and intermolecular disulfide bonds by diagonal gel electrophoresis is described in Ref. 20.

ERp57 Deletion Does Not Elicit Compensatory Unfolded Protein
Responses-As previously shown for cells lacking calnexin or calreticulin (11), also in cells deleted of ERp57 there was no sign of ER stress ( Fig.  1, B and C) (8). Accordingly, the ER stress marker sXbp1 was not expressed unless cells were exposed to tunicamycin, a compound that elicits ER stress by blocking N-glycosylation of newly synthesized secretory proteins (Fig. 1B). Moreover, the luminal level of several stressresponsive chaperones was comparable in 57 Ϫ/Ϫ and wt MF (Fig. 1C). The lack of UPR-induction shows that ERp57 is dispensable for maturation of most cellular glycoproteins and that its deletion must be proficiently compensated by the redox activity of surrogate ER oxidoreductases.
ERp57 Deletion Does Not Affect Co-translational Formation of HA Disulfides-We first determined consequences of ERp57 deletion on the oxidative folding of a glycoprotein strictly dependent on calnexin association for efficient maturation, the influenza virus HA. 4 h after infection, cells were pulse-labeled with [ 35 S]methionine/cysteine. To analyze co-translational events, cells were flooded on ice immediately after the 2-min pulse with a membrane-permeable alkylating agent to prevent postlysis oxidation of free cysteines (18). Labeled influenza virus proteins were immunoisolated from cell lysates with a specific antibody and were analyzed using a two-dimensional gel electrophoresis technique (diagonal gel electrophoresis (21)), in which the first dimension is run under nonreducing conditions in a glass capillary tube. For the second dimension, the gel is extruded from the tube, boiled in reducing sample buffer, and placed on a conventional slab gel. In diagonal gels, nascent polypeptides are seen as a stripe of radioactivity from the lower left corner (shorter chains) to the upper right corner (longer chains) (Fig.  2B). Polypeptides without disulfide bonds have the same mobility under nonreducing and reducing conditions; they therefore migrate on the gel diagonal. Polypeptides with intramolecular disulfides migrate faster in the nonreducing gel and move below the diagonal.
In wt MF, HA acquires intra-molecular disulfide bonds co-translationally (17,22). Consistently, labeled HA growing nascent chains ran below the gel diagonal (Fig. 2, B and C, panel 1). Some nascent chains acquired small loop disulfides and diverged below the gel diagonal (arrow 1); they were completed into the IT1 form of full-length HA. Others, starting from a length of about 50 kDa (arrow 2), acquired one big loop disulfide and terminated in the IT2 form characterized by a larger distance from the gel diagonal. Formation of the second big loop disulfide originating a fully oxidized NT form with an even faster mobility in the second dimension occurred just before or after chain termination (arrow 3 in Fig. 2C, panel 1; for nomenclature of the HA folding intermediates expressed in wt cells, see Ref. 18).
Co-translational formation of intramolecular disulfides did not require ERp57. Comparison of co-translational HA oxidation in cells with and without ERp57 showed that a comparable amount of nascent HA was below the gel diagonal and thus had acquired intramolecular disulfides in the two cell lines. Oxidation actually started on shorter HA chains in cells lacking ERp57 (compare the position of arrow 2 in Fig. 2C, panel 1 versus panel 2). Thus, although we cannot exclude the possibility that the intramolecular HA disulfides formed with and without ERp57 are different (for that reason, the latter were named po (for partially oxidized) and eo (for extensively oxidized) in Fig. 2C, panel 2), the data clearly show that ERp57 is dispensable for co-translational oxidation of HA cysteines.
ERp57 Deletion Affects Post-translational Reshuffling of HA Disulfides-After a 3-min chase (Fig. 2C, panels 3 and 4), most HA nascent chains had been completed, and radioactivity was mainly condensed in the cytosolic influenza virus nuclear protein (NP) and FIGURE 1. Analysis of ER chaperone content. A, 3 g of wt and of si57 and 10 g of 57 Ϫ/Ϫ cell lysates were subjected to SDS-PAGE and blotted on polyvinylidene difluoride membrane. The membranes were decorated with ␣-ERp57-specific (upper panel) or ␣-tubulin-specific (lower panel as loading marker) antibodies. B, spliced sXbp1 is an active transcription factor generated in cells upon ER stress (e.g. treatment with the N-glycosylation inhibitor tunicamycin (Tun)). In untreated wt and 57 Ϫ/Ϫ cells (Ϫ), semiquantitative reverse transcription-PCR (RT-PCR) shows absence of sXbp1 transcripts. sXbp1 is only detected upon overnight incubation with 2.5 g/ml tunicamycin (ϩ). C, comparison of calnexin (Cnx), ERp72, ERp57, BiP, calreticulin (Crt), and PDI content in wt and 57 Ϫ/Ϫ cells; tubulin served as a loading marker (immunoblot).
in full-length HA in the IT1, IT2, or NT forms. In 57 Ϫ/Ϫ MF, about 20% of the labeled HA was extensively oxidized after a 3-min chase ( Fig. 2C, panel 4) versus the 40% in wt MF (Fig. 2C, panel 3). In 57 Ϫ/Ϫ MF, part of the newly synthesized HA had entered intermolecular disulfide-bonded complexes after chain termination. These are shown above the diagonal of the gel (disulfide-bonded HA (DB HA) in Fig. 2C, panel 4). DB HA only contained full-length disulfidebonded HA as labeled component (Fig. 2C, panel 4) and did not co-precipitate with the oxidoreductases PDI and ERp72. A fraction of DB HA did co-precipitate with BiP, confirming the involvement of this abundant ER chaperone in ER retention of aggregated HA (data not shown) (11). Formation of DB HA is a symptom of inefficient HA folding upon deletion of ERp57, because these complexes were not formed in wt cells (panel 3). DB HA gradually increased in amount during the chase (Fig. 3A, panel 57 Ϫ/Ϫ ). Thus, deletion of the glycoprotein-specific oxidoreductase ERp57 resulted in formation of nonnative intermolecular complexes containing HA after chain termination.
More detailed analysis of post-translational HA folding was performed by chasing cells for up to 40 min before lysis (Fig. 3). In wt MF, oxidative folding of HA was as rapid as in other cell lines (18). After a 2-min chase, 50% of the labeled HA was in the partially oxidized IT1 or IT2 forms, and 50% had completed the oxidation process and was in the NT form (Fig. 3A, wt). Consistent with efficient maturation, HA was released from calnexin (t1 ⁄ 2 of less than 10 min) (Fig. 3, C and D, wt) and from the ER. ER exit was confirmed by the acquisition of Endo H-resistant N-glycans (t1 ⁄ 2 of about 20 min, Fig. 3, E and F, wt) that occurs when native glycopolypeptides transit the Golgi compartment (23).
Oxidative folding of HA was significantly less efficient without ERp57, because 40 -50% of the labeled HA was trapped in DB HA after a 40-min chase (DB HA in Fig. 3, A (57 Ϫ/Ϫ ) and B). Moreover, the oxidation of the HA protected from covalent aggregation was slower without ERp57 (Fig. 2C, compare panel 3 with 4; Fig. 3A, compare wt with 57 Ϫ/Ϫ ). An inverse correlation between the intraluminal content of ERp57 and the gravity of the HA folding defect was confirmed in a stable cell line subjected to specific RNA interference and expressing only 20% of the wt level of ERp57 (si57; Fig. 1A). These cells had an intermediate phenotype characterized by an HA oxidation rate closer to that of wt MF (Fig. 3A, si57) and a partial recovery of HA folding efficiency, as shown by the substantial reduction of DB HA compared with 57 Ϫ/Ϫ MF (Fig.  3B). After a 40-min chase, virtually no HA was aggregated in wt cells (Fig. 3B, lane 1), versus the 10 -20% in si57 MF and the 40 -50% in 57 Ϫ/Ϫ MF (lanes 3 and 2, respectively).
The extensively oxidized HA (eo in Fig. 3A, 57 Ϫ/Ϫ ) generated without ERp57 and the extensively oxidized HA generated in wt MF (NT in Fig.  3A, wt) had similar mobility in the nonreducing gels, but at least part of the former contained nonnative intramolecular disulfides. Consistent with the persistence of a folding defect despite extensive oxidation, the release of HA from calnexin ( Fig. 3, C, D, and G) and from calreticulin ( Fig. 3G) was delayed in cells lacking ERp57, and, notably, less labeled HA acquired the native epitopes exposed by trimeric HA in 57 Ϫ/Ϫ MF (Figs. 3H (lower panel) and 4B). Thus, fine structural defects caused by inappropriate cysteine pairing in cells lacking ERp57 delayed HA release from calnexin/calreticulin and inhibited the formation of HA homotrimers, which is required for forward transport of HA (24). As a consequence, only about 30% of the labeled HA acquired Endo H-resistant N-glycans after a 40-min chase in cells lacking ERp57 versus the 70 -80% in wt MF (Fig. 3, E, F, and H, upper panel). For HA, we were unable to identify surrogate oxidoreductases assisting the slower and less efficient HA maturation or catalyzing the formation of the aberrant intermolecular disulfides causing HA aggregation in 57 Ϫ/Ϫ MF (see below for identification of surrogate oxidoreductases for other model glycoproteins).
Defective Recovery of Native HA Conformation in Cells Lacking ERp57-To confirm that ERp57 deletion impaired disulfide reshuffling occurring during HA maturation, we artificially induced formation of nonnative disulfides in newly synthesized HA. To this end, infected cells were flooded, after the pulse with radioactivity, with diazenedicarboxylic acid bis(N,N-dimethylamide) (diamide), a powerful cell-permeable thiol oxidant (25). 5 min later, diamide was washed out. Conversion from the nonnative to the native set of HA disulfides in wt and in 57 Ϫ/Ϫ MF was monitored by determining acquisition of trimer-specific epitopes and of Endo H-resistant N-glycans.
In both cell lines, incubation with diamide caused rapid formation of nonnative intra-and intermolecular disulfides in newly synthesized polypeptides. In cells lysed immediately after the chase with diamide, immunoisolation of the labeled nuclear protein and HA was very inefficient (compare the amount of labeled protein in lanes 1 and 2 and lanes  13 and 14 versus lanes 3-12 and lanes 15-24 in Fig. 4A). Evidently, the epitopes recognized by the antibodies were inaccessible in the diamideinduced aggregates. Moreover, the solubilization of the diamide-induced HA-containing aggregates was suboptimal. At longer times after diamide wash-out (5-60 min in Fig. 4A), more HA and nuclear protein  MARCH 10, 2006 • VOLUME 281 • NUMBER 10 were immunoisolated from the detergent extracts, showing reversibility of the aggregation process and reappearance of the epitopes of the antibody in both cell lines.

ERp57 and Glycoprotein Folding
In wt MF chased for 5 min after the diamide wash-out, most of the immunoisolated HA entered the nonreducing gel (Fig. 4A, lane 3). With progression of the chase, HA gradually acquired faster mobility and therefore more compact conformation (Fig. 4A, lanes 3, 5, and 7, respectively). This showed a slow progression of disulfide isomerization that eventually converted part of the HA from large aggregates into a partially oxidized (po in Fig. 4A) and then into a more extensively oxidized form (eo in Fig. 4A). 40 and 60 min after diamide wash-out, a significant fraction of labeled HA was native and was, accordingly, immunoisolated from cell lysates with a trimer-specific monoclonal antibody (Fig. 4B). Confirming acquisition of the native conformation, 30 and 50% of the labeled HA had acquired Endo H-resistant N-glycans after 40 and 60 min of chase, respectively (Fig. 4C). Thus, in wt cells artificially misfolded, disulfide-bonded HA was slowly converted into native, trimeric, and transport-competent HA upon diamide wash-out.
Deletion of ERp57 had three consequences. First, a significant fraction of detergent-soluble, labeled HA remained in disulfide-bonded complexes after diamide wash-out (DA in the nonreducing gel; Fig. 4A,  lanes 4, 6, 8, 10, and 12). This showed that disassembly of covalent aggregates formed upon diamide treatment was defective in cells lacking ERp57. Second, the disulfide reshuffling to the extensively oxidized HA (eo in Fig. 4A) after diamide wash-out was significantly delayed without ERp57 (t1 ⁄ 2 of 10 min versus 5 min in wt MF). Third, a lower fraction of HA acquired a native, trimerization-competent structure, confirming the incorrect pairing of HA cysteines without ERp57. Consistently, significantly less labeled HA reacted with trimer-specific antibodies in 57 Ϫ/Ϫ MF (Fig. 4B), and acquisition of Endo H-resistant N-glycans was not observed (Fig. 4C).
To summarize, ERp57 deletion did not prevent co-and post-translational oxidation of the cysteines of HA. However, the several oxidoreductases in the ER lumen did not replace with due efficiency the deleted ERp57 in assisting the post-translational phases of HA maturation. In fact, shortly after chain termination, half of the newly synthesized HA formed covalent aggregates, and even the fraction of HA protected from aggregation only hardly acquired native architecture despite extensive oxidation. Moreover, aberrant diamide-induced disulfides were slowly reshuffled into the native set of disulfides of intramolecular HA in the wt ER. Reshuffling was substantially delayed upon ERp57 deletion. Taken together, the data showing accumulation of HA aggregates in 57 Ϫ/Ϫ MF (Figs. 2 and 3) and the defect in disassembly of disulfide-bonded HA upon diamide wash-out in 57 Ϫ/Ϫ MF (Fig. 4) hint at ERp57 chaperone and/or redox activity involved in reducing the kinetic partitioning of the off pathway during HA folding.
ERp57 Deletion Does Not Affect Co-translational Formation and Post-translational Reshuffling of E1 and p62 Disulfides-We next determined the consequences of ERp57 deletion on the maturation of calnexin substrates that remain folding-competent upon inactivation of the calnexin chaperone system. We selected the E1 and p62 glycoproteins of SFV because of their propensity to abundantly engage ER oxidoreductases in transient mixed disulfides (MD) during maturation (26). E1/p62-containing MD can be immunoisolated with oxidoreductase-specific antibodies for analysis in conventional and in diagonal gels. We reasoned that these model glycoproteins could in principle be exploited to assess consequences of ERp57 deletion and, more importantly, to identify the ER-resident oxidoreductases intervening to replace the missing, inactive, or busy ERp57.
We first compared co-translational oxidation of E1 and p62 cysteines in cells with and without ERp57. Infected cells were pulsed for 2 min with radioactivity (as described above). At the end of the pulse, they were flooded with ice-cold NEM to alkylate free sulfhydryl groups of nascent chains. The virus-encoded glycoproteins were immunoisolated with specific antibodies and separated in diagonal gels (Fig. 5, A and B). The data showed that a similar amount of nascent E1/p62 chains acquired intramolecular disulfides and diverged below the gel diagonal in the two cell lines (arrow 1 in Fig. 5, A and B). Thus, as previously reported for HA, ERp57 deletion did not inhibit oxidation of E1 and p62 cysteines emerging in the ER lumen during the synthesis of the glycopolypeptides.
The post-translational phases of E1/p62 maturation progressed normally in cells lacking ERp57. This was shown by the equivalent generation of E2 (a p62 cleavage product only generated when native E1/p62 heterotrimers leave the ER and transit in the Golgi compartment) in cells with and without ERp57 (Fig. 5C). Consistent with unperturbed maturation of SFV glycoproteins, the secretion of labeled viral particles from SFV-infected 57 Ϫ/Ϫ MF was normal (Fig. 5D). Thus, analysis of E1/p62 maturation confirmed that ERp57 was dispensable during cotranslational phases of glycoprotein oxidation and revealed that, unlike HA, these facultative clients of calnexin did proficiently complete the folding process in cells lacking ERp57. The next challenge was the identification of the surrogate oxidoreductase(s) involved in the ERp57-independent glycoprotein maturation. . Conversion from diamide-induced covalent aggregates into native, transportcompetent HA in wt and in 57 ؊/؊ MF. A, at the end of a 3-min pulse with radioactivity, cells were flooded with diamide (5 mM in PBS) for 5 min to promote formation of nonnative disulfides. Diamide was washed out, and chase was continued for the indicated times to allow correction of the nonnative disulfides. Detergent-soluble, disulfidebonded HA is shown as DA (for diamide-induced aggregates) at the top of the nonreducing gel. B, native HA homotrimers formed after the diamide wash-out were immunoisolated from cell lysates with a specific monoclonal antibody. C, acquisition of Endo H-resistant N-glycans allows quantification of the recovery of the native, transport-competent HA structure.

Compensatory Folding Pathways Involving the Oxidoreductase ERp72
Are Activated in Cells Lacking ERp57-To identify alternative chaperone-assisted folding pathways used by newly synthesized glycopolypeptides when the calnexin assistance based on ERp57 function is inactive or missing, we next compared chaperone intervention during E1/p62 maturation in wt and in 57 Ϫ/Ϫ MF. To this end, the labeled viral glycoproteins were immunoisolated from detergent extracts of cells after a 6-min chase, when association with ER-resident factors was maximal. We made use of viral protein-specific antibodies and of antibodies to several ER-resident molecular chaperones (calnexin, calreticulin, and BiP (Fig. 6A)) or oxidoreductases (ERp57, PDI, ERp72 (Fig. 6B)). The immunoprecipitates were analyzed in nonreducing and reducing SDS-PAGE.
After a 6-min chase, E1 and p62 are still undergoing folding in the wt ER (26,27). Accordingly, the E2 band formed in the Golgi complex at later chase times (Fig. 5C) was still not visible (Fig. 6A, lane 1). Instead, a heterogeneous population of mixed disulfides containing E1, p62, and several oxidoreductases is formed in the ER during the oxidative folding of the viral glycoproteins (26). MD are transient (Fig. 5C), short lived intermediates of the folding process (26) shown in the nonreducing gels (Figs. 5C and 6A, lane 1). They are disassembled upon sample reduction with 100 M dithiothreitol (Fig. 6A, lane 9). Without ERp57, E1, p62, and p97 (an uncleaved E1-p62 precursor) are unchanged, but E1/p62containing MD had a clearly different pattern (Fig. 6A, compare MD in lanes 1 and 2).
Abundant MD were also co-precipitated with BiP (Fig. 6A, lanes 7  and 8), showing that BiP was also associated with an oxidoreductase engaged in MD with viral glycoproteins. The BiP-associated oxidoreductase was not ERp57, because in cells without ERp57, the BiPassociated MD increased rather than disappeared (compare lanes 7 and 8 in Fig. 6A) and because in contrast to the ERp57-containing MD, the  BiP-associated MD had E1 as the major viral component (compare Fig.  6A, lane 15, with Fig. 6B, lane 7).
Two ER-resident oxidoreductases were found to have E1 as a major covalent partner during E1/p62 oxidation in wt MF. The first was PDI (Fig. 6B, lane 3 (nonreducing) and lane 9 (reducing)). The second was ERp72 (Fig. 6B, lanes 5 and 11). Neither PDI nor ERp72 were the BiPassociated oxidoreductase, because the pattern of PDI-and ERp72-containing MD differed from that of BiP-associated MD (Fig. 6, A-C). A possible candidate for the BiP-associated oxidoreductase would be ERdj5. This oxidoreductase contains, in fact, a BiP-specific J domain (28 -30). Unfortunately, our attempts to prove the association of ERdj5 with E1 and/or p62 failed.
Deletion of ERp57 slightly increased the engagement of PDI in MD, but the pattern of PDI-containing MD did not change (Fig. 6B, compare  lanes 3 and 4). MD disassembly in the reducing gel showed that in both cell lines PDI was mostly associated with E1. The engagement of the oxidoreductase with the viral glycoproteins was quantified by determining the ratio of E1/p62 upon disassembly in the reducing gels of the viral glycoproteins/oxidoreductase MD. This ratio remained 10:1 for PDI even upon ERp57 deletion (Fig. 6B, compare lanes 9 and 10). Thus, PDI did not replace ERp57 and did not associate with orphan p62 in 57 Ϫ/Ϫ MF.
This was in contrast to BiP and ERp72. Although the BiP-associated MD only slightly changed their pattern upon ERp57 deletion (Fig. 6A,  compare lanes 7 and 8), their disassembly revealed a more consistent presence of p62 in 57 Ϫ/Ϫ MF (Fig. 6A, compare lanes 15 (ratio of E1/p62 is 10:1) and 16 (ratio of E1/p62 is 10:3)). For ERp72, both the MD pattern (Fig. 6B, compare lanes 5 and 6) and the ratio of viral polypeptides in the MD changed clearly upon ERp57 deletion. ERp72 engaged significantly more p62 during the oxidative phases of folding upon deletion of the p62-dedicated oxidoreductase ERp57 (compare p62 co-precipitated with the ERp72-specific antibody in Fig. 6B, lanes 11 and 12).
ERp72 is the only member of the PDI superfamily that conserves the ERp57 residues mediating association with calnexin (30). We therefore hypothesized that ERp72 could replace ERp57 in a functional complex with the ER lectins in 57 Ϫ/Ϫ MF. Our prediction was wrong. Unlike ERp57-containing MD, those containing ERp72 did not disappear in cells treated with castanospermine, a sugar analogue that inhibits substrate association with calnexin by inactivating ER ␣-glucosidases (Fig.  6, C and D). Thus, ERp72 acts as surrogate oxidoreductase to replace ERp57, and it does so independently of substrate association with calnexin or calreticulin.

DISCUSSION
Studies performed in test tubes in the early 1960s revealed that only the information contained in the amino acid sequence is required for proper folding of polypeptides (31,32). The discrepancy between in vitro and in vivo folding rates led to the discovery of an enzyme system that catalyzes the rate-limiting step of protein folding in the ER, consisting in formation of native disulfides between the cysteines of the protein (33,34). 43 years of further study showed that the mammalian ER hosts about 20 different PDI-like proteins. For some of them, the redox activity in vitro has been established, but their functions in vivo are less well studied (30,(35)(36)(37)(38).
ERp57 is one of the best characterized PDI family members. It has been shown that, providing an appropriate redox buffer, ERp57 shows multifunctional oxidoreductase activity in vitro (39,40) and in semipermeabilized cells (41), which is enhanced by the interaction with calnexin or calreticulin (42). In cells, ERp57 is noncovalently associated with calnexin and calreticulin and is therefore appropriately placed to assist the oxidative maturation of newly synthesized N-glycosylated polypeptides (43)(44)(45)(46)(47).
Oxidative protein folding in the ER progresses in two distinct phases. The co-translational phase consists of oxidation of nascent chain cysteines to form intramolecular disulfides. It is followed by a post-translational phase, where disulfides are rearranged into the native set that fulfills quality control requirements. The first phase is completed in about 2 min, the time required to synthesize and translocate in the ER lumen the few hundred residues of the nascent chains (514 for HA, 438 for E1, and 482 for p62, respectively, for the model proteins used in this study) and was investigated in this work with a sophisticated technical approach based on metabolic labeling of nascent chains and their analysis by diagonal gel electrophoresis (20,21). The second phase lasts for about 15-20 min in wt cells for HA and a bit longer for E1 and p62. It is concluded with the release of the native polypeptides from the ER into the Golgi complex.
For the substrates analyzed in this work, deletion of ERp57 did not inhibit cysteine oxidation occurring during the entry of newly synthesized glycopolypeptides in the ER lumen. Two options remain open to explain this result: 1) ERp57 is not involved in co-translational oxidation of substrate cysteine (this would be in keeping with published data reporting that luminal ERp57 is found in the reduced state and should therefore not be involved in cysteine oxidation (48 -50)); 2) an unidentified oxidoreductase(s) proficiently replace the missing ERp57 in oxidizing free cysteines of growing nascent chains.
Our data show that ERp57 plays an important role during the posttranslational phases of glycoprotein folding, when the primary requirement for the promotion of glycoprotein maturation is an isomerase/ reductase activity that catalyzes the reshuffling of disulfides formed as intermediates of the folding process into the definitive set that characterizes the architecture of the native polypeptide. In fact, for both obligate and facultative clients of the calnexin chaperone system, deletion of ERp57 had consequences on protein oxidation after chain termination. For obligate clients, maturation stalled, because, evidently, ERp57 could not be proficiently replaced by any of the several other oxidoreductases of the ER lumen. For facultative users, folding proceeded to completion under the assistance of at least one surrogate PDI family member, ERp72.
The in vivo function and catalytic specificity of most cellular oxidoreductases are unknown. Our analysis of oxidative glycoprotein folding in the first mammalian cell line deleted of one member of the PDI superfamily proves that functional redundancy exists in vivo among the several ER oxidoreductases.