Different Interaction Modes for Protein-disulfide Isomerase (PDI) as an Efficient Regulator and a Specific Substrate of Endoplasmic Reticulum Oxidoreductin-1α (Ero1α)*

Background: Ero1α and PDI constitute the pivotal oxidative protein folding pathway in mammalian ER. Results: Both catalytic domains of PDI and PDI homologues rapidly regulate Ero1α activity while Ero1α asymmetrically oxidizes PDI. Conclusion: The modes for PDI as efficient regulator and specific substrate of Ero1α are different. Significance: This study reveals how Ero1α-PDI interplay ensures oxidative protein folding homeostatically. Protein-disulfide isomerase (PDI) and sulfhydryl oxidase endoplasmic reticulum oxidoreductin-1α (Ero1α) constitute the pivotal pathway for oxidative protein folding in the mammalian endoplasmic reticulum (ER). Ero1α oxidizes PDI to introduce disulfides into substrates, and PDI can feedback-regulate Ero1α activity. Here, we show the regulatory disulfide of Ero1α responds to the redox fluctuation in ER very sensitively, relying on the availability of redox active PDI. The regulation of Ero1α is rapidly facilitated by either a or a′ catalytic domain of PDI, independent of the substrate binding domain. On the other hand, activated Ero1α specifically binds to PDI via hydrophobic interactions and preferentially catalyzes the oxidation of domain a′. This asymmetry ensures PDI to function simultaneously as an oxidoreductase and an isomerase. In addition, several PDI family members are also characterized to be potent regulators of Ero1α. The novel modes for PDI as a competent regulator and a specific substrate of Ero1α govern efficient and faithful oxidative protein folding and maintain the ER redox homeostasis.

Disulfide bonds play important roles in the structure and function of many secretory and membrane proteins. The correct formation of disulfides during the folding of nascent peptides to native proteins, namely oxidative protein folding, takes place mainly in the endoplasmic reticulum (ER) 4 in eukaryotic cells (1). Protein-disulfide isomerase (PDI) and sulfhydryl oxidase ER oxidoreductin-1 (Ero1) constitute the pivotal pathway for oxidative protein folding from yeast to mammals. PDI contains four thioredoxin (Trx) domains arranged as a-b-bЈ-aЈ, with two -CGHC-active sites respectively located in domain a and aЈ. PDI can directly catalyze disulfide formation in reduced substrates, as well as the isomerization reaction to convert aberrant disulfides to correct ones (2). Ero1 flavoproteins can catalyze the re-oxidation of reduced PDI for continuous transfer of disulfides to substrate proteins. The -CXXXXC-outer active site located in an intrinsically flexible loop of Ero1 transfers electrons from the active site of PDI to the buried -CXXCinner active site, and the electrons are then used to reduce oxygen into hydrogen peroxide via flavin adenine dinucleotide cofactor (3,4).
There are two isoforms of Ero1 in mammalian cells: Ero1␣ is widely expressed (5) and Ero1␤ is abundantly expressed in select secretory tissues such as the pancreas (6). Both Ero1␣ and Ero1␤ activities are regulated by regulatory disulfides formed between catalytic and non-catalytic cysteines to avoid futile oxidation cycles with excess hydrogen peroxide production (7)(8)(9). For Ero1␣ particularly, the formation of two regulatory disulfides Cys94-Cys131 and Cys99-Cys104 in the inactive resting state blocks disulfide transferring from the inner active site (Cys394-Cys397) to PDI via the outer active site (Cys94-Cys99). These two regulatory disulfides need to be reduced to liberate the outer active site for activation of Ero1␣. A conserved longrange disulfide Cys85-Cys391 was suggested to also participate in the activity regulation of Ero1␣ (7,8), which was challenged later (10).
In cells, the poise of active and inactive Ero1␣ at steady state was demonstrated to be dependent on the level of PDI (7,11). Thus, PDI seems not only a substrate but also a physiological regulator of its oxidase Ero1␣. However, the dynamics of the transition between active and inactive Ero1␣ during the fluctuation of ER redox environment and the role of PDI in these processes remain largely unknown. For the interplay between Ero1␣ and PDI, it has been elucidated that the catalytic active Ero1␣ preferentially oxidizes the C-terminal active site in domain aЈ of PDI, rather than the N-terminal active site in domain a (8,12), although the reduction potentials of the two active sites are very similar (13). Also we and others have provided evidence that the primary substrate binding domain bЈ of PDI plays a critical role in binding with Ero1␣ for functional disulfide relay (10,12,14). On the other hand, the molecular mechanism of the reduction/oxidation of the regulatory disulfides of Ero1␣ by PDI is little understood. There are at least twenty PDI family members (PDIs) in mammalian ER (2), but Ero1␣ as well as its hyperactive isoform Ero1␤ poorly catalyzes the oxidation of other PDIs (9,15). Meanwhile other PDIs at steady state unsuccessfully modulate the redox states of Ero1␣ (15). Altogether, to reveal the molecular mechanism underlying the interplay between Ero1␣ and PDIs is central and crucial for understanding how efficient oxidative folding and redox balance in the ER are maintained in mammalian cells.
In this study, we report that (i) Cys85-Cys391 disulfide in Ero1␣ is stable and remains intact during the physiological activation of the enzyme; (ii) Cys94-Cys131 regulatory disulfide responds to the redox fluctuation in ER very sensitively, and its reduction/oxidation can be facilitated by not only PDI but also some other PDIs; (iii) either catalytic domain of PDI is able to facilitate the regulation of Ero1␣, and the substrate binding domain bЈ of PDI is not essential for activation/inactivation of Ero1␣; (iv) the functional oxidation of PDI catalyzed by Ero1␣ is asymmetric to make the aЈ domain act primarily as an oxidase and the a domain as an isomerase. The above findings shed great light on the mechanism underlying the interplay between Ero1␣ and PDI proteins, which ensures the efficiency and fidelity of oxidative protein folding and maintains the thiol-disulfide redox homeostasis in the ER.
Recombinant Ero1p (18) and Ero1␣ (12) proteins were expressed and purified as described. PDI proteins and Pdi1p proteins were purified as for PDI (19). For reduced protein preparation, PDI proteins at 100 M or Ero1␣ at 10 M with 100 mM DTT and Pdi1p proteins at 100 M with 10 mM GSH were incubated in buffer A (50 mM Tris-HCl, 150 mM NaCl, 2 mM EDTA, pH 7.6) for 1 h at 25°C. Excess reductants were then removed using a HiTrap desalting column (GE Healthcare) pre-equilibrated with buffer A, and the reduced proteins were kept on ice for use only in the same day. For oxidized protein preparation, PDI proteins at 100 M or Ero1␣ at 50 M was incubated with 50 mM potassium ferricyanide in buffer A for 1 h at 25°C, and then chromatographed through a Superdex-200 10/300 GL column (GE Healthcare) pre-equilibrated with buffer A. Monomeric protein fraction was collected, concentrated and stored at Ϫ80°C in aliquots.
Cell Culture, Transfection, and Antibodies-HeLa cells were cultured in DMEM (Invitrogen) containing 5% fetal bovine serum, 100 units/ml penicillin and 100 g/ml streptomycin (Invitrogen) at 5% CO 2 . Plasmids were transfected using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. After 48 h, the transfected cells were harvested, or as needed treated with 50 M DTT and 50 M 16F16 (Sigma-Aldrich) dissolved in DMSO for 8 h before harvest.
RNA Interference-The pSUPER-retro-puro vector (Oligoengine) expressing the shRNA targeting PDI sequence 5Ј-GAG-TGTGTCTGACTATGAC-3Ј (20) was constructed according to the manufacturer's instructions. The pSUPER-shEro1␣ plasmid was used as described (17). pSUPER plasmids were transiently transfected into HeLa cells on day 1. Puromycin was added into the culture medium to a final concentration of 2 g/ml on day 2 to kill the negative cells. Then pSUPER plasmids were co-transfected with pcDNA3.1-Ero1␣ C99/104A on day 3. Cells were harvested on day 5.
Assay for the in Vivo Redox States of Ero1␣, Immunoprecipitation, and Western Blotting-For Ero1␣ activation, the harvested cells were re-suspended in DMEM containing 150 M DTT at 25°C. Aliquots were taken and immediately blocked by 20 mM N-ethylmaleimide (NEM, Sigma-Aldrich) at different times to trap disulfide bonds. For Ero1␣ inactivation, after incubation with 10 mM DTT in DMEM at 25°C for 10 min, cells were quickly washed twice by ice-cold phosphate-buffered saline to remove excess DTT and re-suspended in DMEM at 25°C. Aliquots were then taken and immediately blocked by 20 mM NEM at different times. Cells were lysed in radio immunoprecipitation assay buffer (Beyotime) containing 1 mM phenylmethanesulfonyl fluoride and 20 mM NEM. Post-nuclear super-natants were resolved by nonreducing SDS-PAGE to analyze the redox states of Ero1␣. Immunoprecipitation was carried out by incubation of cell lysates with ␣HA for 2 h at 4°C, followed by addition of protein A ϩ G Agarose (Beyotime) and rotation for another 2 h at 4°C. The beads were then washed for 3 times with RIPA buffer and the resuspended samples were analyzed by nonreducing SDS-PAGE. Proteins were transferred to polyvinylidene difluoride membranes (Millipore) using a semi-dry transfer apparatus, and the membranes were blocked in 5% milk and decorated by antibodies and enhanced chemiluminescence (Thermo Scientific), and visualized by using a Chemi-Scope mini imaging system (Clinx Science).
Redox State Determination of Activated and Inactivated Ero1␣ in Vitro-For the activation of Ero1␣, oxidized Ero1␣ was either incubated with 10-fold reduced PDI proteins or treated by 10-fold PDI proteins in the presence of GSH/GSSG at different ratios. For the inactivation of Ero1␣, reduced Ero1␣ was mixed with 10-fold oxidized PDI proteins. To study the self-oxidation of Ero1␣, reduced Ero1␣ proteins were incubated with equimolar GST-Ero1␣ proteins. All experiments were carried out in buffer A at 25°C, and aliquots were taken at different times and immediately quenched with 20 mM NEM. The samples were then analyzed by nonreducing SDS-PAGE followed by immunoblotting with ␣Ero1␣. Band intensity was quantified using Image J software (National Institutes of Health).
Oxygen Consumption Assay-Oxygen consumption was measured at 25°C using an Oxygraph Clark-type oxygen electrode (Hansatech Instruments) as described (9). Briefly, reactions were initiated by adding Ero1 proteins to a final concentration of 2 M into buffer B (100 mM Tris-HAc, 50 mM NaCl, 2 mM EDTA, pH 8.0) containing 20 M PDI proteins and various concentrations of GSH and GSSG.
Gel-based PDI Oxidation Assay-Oxidation of 20 M reduced PDI proteins by 3 M Ero1 proteins were carried out at 25°C in buffer B. At different time points, aliquots were taken for immediate mixing with equal volume of 2ϫ SDS-PAGE loading buffer containing 5 mM mPEG-5k (Sigma-Aldrich) followed by incubation at 33°C for 20 min. The samples were subjected to SDS-PAGE after consuming excess of mPEG-5k by 25 mM DTT.
PDI Isomerase Activity Assay-Scrambled bovine pancreatic RNase A (Sigma-Aldrich) was prepared as described (21). Ero1 and PDI proteins at equimolar were pre-incubated at 25°C for 30 min in buffer B, and the isomerase activities of PDI proteins at 3 M were then assayed by adding scrambled RNase A and cCMP (Sigma-Aldrich) to a final concentration of 8 M and 4.5 mM, respectively. The absorbance increase at 296 nm due to the hydrolysis of cCMP by refolded RNase A was monitored at 25°C, and the concentration of reactivated RNase A was calculated with details described elsewhere (22). The linear slope of RNase A reactivation was taken as the isomerase activity of PDI.
BPTI Oxidative Folding Assay-BPTI (Sigma-Aldrich) of 2.5 mg was denatured and reduced by incubation with 20 mM DTT and 6 M guanidine hydrochloride in 0.1 M Tris-HCl (pH 8.0) at 50°C for 5 h. Excess reductants and denaturants were removed using a HiTrap desalting column pre-equilibrated with 0.01 M HCl. The refolding assays were carried out at 25°C by adding denatured and reduced BPTI to a final concentration of 30 M into buffer B containing pre-incubated Ero1 and PDI proteins making a final concentration of 3 M of each. Aliquots were taken at different time points and quenched by adding 0.1 volumes of 5 M HCl. The samples were then loaded onto a Vydac C18 analytical HPLC column (250 ϫ 4.6 mm, GRACE) and eluted with a flow rate of 1 ml/min using a linear gradient of acetonitrile from 15% to 50% at a rate of 0.7%/min in 0.05% trifluoroacetic acid. The absorbance at 229 nm was monitored. The percentage of each folding intermediate during the refolding was quantified using Chromeleon software (Thermo Scientific).
We next examined the influence of PDI on the redox states of Ero1␣ in HeLa cells. Ero1␣ C99/104A mutant was used because this mutant retains both long-range disulfides of Cys85-Cys391 and Cys94-Cys131 for redox state examination but lacks intact outer active site (Cys94-Cys99) for catalyzing substrate oxidation, so that the interference from the re-oxidation of Ero1␣ regulatory disulfides by oxidized substrates can be excluded. As shown in Fig. 1E, Ero1␣ C99/104A migrated in both Ox1 and Ox2 forms, implying that in cells it exists in both activated and inactivated forms, similar to Ero1␣ WT. Co-expression of PDI at steady state led to moderate but significant increase of Ox1 form in C99/104A as well as pronounced increase of the disulfide-linked Ero1␣-PDI heterodimer, but no fully reduced form was observed, indicating that in cells the increase of PDI level promotes the reduction of Cys94-Cys131 but not Cys85-Cys391. Moreover, mutation of Cys85-Cys391 disulfide dramatically impaired the formation of Ero1␣-PDI heterodimer, underlining the importance of this bond for maintaining the Ero1␣-PDI functional complex. In summary, during the activation of Ero1␣ the Cys94-Cys131 regulatory disulfide is reduced resulting in the mobility shift from Ox2 to Ox1, whereas the Cys85-Cys391 disulfide remains intact, which is important for the catalytic activity of Ero1␣.
Dynamic Regulation of Ero1␣ Activity in Cells by PDI-To gain further insights into the dynamic regulation of Ero1␣ during the fluctuation of the ER redox environment in cells, the activation/inactivation processes of Ero1␣ were studied by monitoring the interconversion between Ox1 and Ox2 forms of the aforementioned Ero1␣ C99/104A mutant. Firstly, the activation kinetics of Ero1␣ was examined by stressing the cells with a low concentration of DTT to mimic the burst of free thiols during protein synthesis. Just in 1 min after DTT challenge, a large portion of Ero1␣ quickly shifted to Ox1 state, but after 10 min there was still a small fraction of Ox2. When PDI was overexpressed, all Ero1␣ shifted from Ox2 to Ox1 in 1 min ( Fig. 2A). When cells were pre-treated by a small molecule 16F16, which specifically inhibits the thiol-disulfide oxidoreductase activity of PDI (24), the reduction of Ero1␣ upon DTT addition was almost completely inhibited, strongly suggesting that PDI plays a critical role in mediating the thioldriven activation of Ero1␣ (Fig. 2B).
Next, we studied the inactivation process of Ero1␣ C99/104A by using a DTT pulse-chase assay. In DTT-flooded cells, Ero1␣ was in Ox1 state with the Cys94-Cys131 regulatory disulfide reduced. At the end of the pulse, Ox2 species quickly re-emerged and increased during the 'chase', but not all Ox1 were re-oxidized to Ox2 after 10 min. Knockdown of PDI significantly delayed the re-emergence of Ero1␣ Ox2 following the DTT-pulse (Fig. 2C). Moreover, when cells were pre-treated with 16F16, the delay of Ero1␣ inactivation became more conspicuous (Fig. 2D), emphasizing the role of redox active PDI in FIGURE 1. Identification of the regulatory disulfide in Ero1␣. A, schematic representation of the disulfide pattern in active (Ox1) and inactive (Ox2) states of Ero1␣ (7,14,23). The cysteine residues are shown as white, gray (outer active site), and black (inner active site) circles with numbering, and disulfides are indicated as lines. Regulatory disulfides (Cys94-Cys131 and Cys99-Cys104) formed in Ox2 state are indicated by curves (7,8). Note that the long-range disulfide Cys85-Cys391 is explicitly identified to be a structural but not regulatory disulfide in this work. The flexible loop region is represented by an open bar. B, Ero1␣ WT (lanes 1-9) of 1 M was incubated with or without 10 M PDI in 10 mM glutathione redox buffer composed of various concentrations of GSH and GSSG as indicated at 25°C for 5 min, then analyzed by nonreducing SDS-9% PAGE and Western blotting (WB) with ␣Ero1␣. The distinct gel mobilities of three Ero1␣ Cys-to-Ala mutants (lanes 10 -12) and fully reduced Ero1␣ treated with 2-mercaptoethanol (2-ME) (lane 13) indicated the reduction of different disulfides. Reduced Ero1␣ (Red), two oxidized Ero1␣ species (Ox1 without Cys94-Cys131 disulfide and Ox2 with both Cys85-Cys391 and Cys94-Cys131 intact) were indicated. Asterisk indicates a species with only Cys85-Cys391 absent (23). C, oxygen consumption catalyzed by 2 M Ero1␣ WT was monitored in the presence of 20 M PDI and various concentrations of GSH/GSSG as indicated by numbering the same as in B. D, 2 M Ero1␣ WT and mutants were incubated in the absence (Ϫ) or presence of 20 M PDI and 10 mM GSH for indicated time, analyzed as in B. E, Myc-tagged Ero1␣ WT and mutants were co-expressed with (ϩ) or without (Ϫ) HA-tagged PDI in HeLa cells. After NEM blocking, cells were lysed and analyzed by nonreducing (NR) or reducing (R) SDS-8% PAGE, followed by ␣myc or ␣HA WB as indicated. Molecular size markers were shown on the left margin and redox states of Ero1␣ were indicated as in B. Double asterisks indicate the Ero1␣-PDI complex. The ratios of Ox1/Ox2 of Ero1␣ C99/104A from lanes 4 and 5 are shown on the right panel (mean Ϯ S.D., n ϭ 3).
Ero1␣ inactivation. As the recovery of ER redox homeostasis (25) and PDI redox balance (26) from a reductive challenge both largely depend on Ero1␣ oxidase activity, we next looked into whether modulating Ero1␣ level can affect the inactivation of Ero1␣ itself. Silencing endogenous Ero1␣ dramatically inhibited the recovery of Ox2 in Ero1␣ C99/104A (Fig. 2E). Conversely, co-expression of the hyperactive Ero1␣ C104/131A markedly accelerated the transition from Ox1 to Ox2 in Ero1␣ C99/104A and reset the redox poise eventually at 30 min (Fig.  2F). Taken together, in cells Ero1␣ is very sensitive to the fluctuation of thiol-disulfide redox states in the ER, and the regu-latory disulfides of Ero1␣ can be modulated by the redox active PDI.
The Two Catalytic Domains of PDI Can Activate Ero1␣ Independently-PDI contains two -CGHC-active sites respectively located in domain a and aЈ. To understand the contribution of the two catalytic domains to the activation of Ero1␣, five PDI mutants were prepared (Fig. 3A). We tested the in vitro ability of these PDI proteins at reduced form to reduce the Cys94-Cys131 regulatory disulfide in Ero1␣ C99/104/166A mutant. The mutation of the unpaired Cys166 to Ala on the background of C99/104A was to avoid aberrant formation of homodimer during the preparation of homogenous oxidized Ero1␣ monomer (14). Addition of 10-fold reduced PDI WT to oxidized Ero1␣ resulted in rapid appearance of Ox1 form within 15 s, but up to 5 min there was still Ox2 remained ( Fig.  3B), as the reduction potential of the Cys94-Cys131 regulatory disulfide is much lower than that of PDI as mentioned above. PDI C1 or PDI C2, in which both cysteines were replaced by serines in the active site of domain a or aЈ, reduced the Cys94-Cys131 regulatory disulfide as efficiently as PDI WT, indicating that either active site of PDI is sufficient to activate Ero1␣. PDI C1/2 with both active sites mutated as a negatively control, had little effect on the reduction of Ero1␣ even though excess GSH was supplied (Fig. 3, B and C). The ability of catalytic domains of PDI to activate Ero1␣ was further assessed by isolated PDI a and aЈc domains. Similar to PDI WT both isolated domains reduced ϳ40% of the Cys94-Cys131 regulatory disulfide in 5 min, although Ox1 form appeared somewhat slower during the initial stage (Fig. 3, B and C). The above results clearly demonstrated that PDI can directly modulate the fast transition of Ero1␣ from Ox2 to Ox1 state, and the two catalytic domains of PDI can independently perform the task of activating Ero1␣.
Autonomous and PDI-mediated Inactivation of Ero1␣-Next, the inactivation process of Ero1␣ was dissected to address whether the re-oxidation of the regulatory disulfides in Ero1␣  can occur autonomously. Purified Ero1␣ protein was firstly treated with excess amounts of DTT for full reduction of the regulatory disulfides. After DTT was washed out, most of Ero1␣ WT immediately converted to the Ox2 form, with only a small portion remained in fully reduced and/or Ox1 form, and a complete conversion to Ox2 form was achieved at 30 min (Fig.  4A, upper panel). The re-oxidation of Ero1␣ mutants lacking the outer active site (Ero1␣ C99/104/166A) or the inner active site (Ero1␣ C394A) was greatly impaired (Fig. 4A, middle panel) or completely inhibited (Fig. 4A, lower panel). Thus, the regulatory disulfides in Ero1␣ can be autonomously re-oxidized aerobically, and the oxidizing power of oxygen is likely expended by the inner active site (Cys394-Cys397) and transferred via the outer active site (Cys94-Cys99) to re-oxidize the regulatory disulfides. We then examined whether the autonomous re-oxidation of Ero1␣ is via intramolecular or intermolecular disulfide transfer. To test the latter possibility, reduced Ero1␣ C394A was incubated with GST-fused Ero1␣ WT or GST-fused hyperactive Ero1␣ C104/131A, because they could be clearly separated by distinct molecular weights on SDS-PAGE. As shown in Fig. 4B, little oxidized Ero1␣ C394A was observed up to 60 min after the addition of equimolar active GST-Ero1␣ proteins, suggesting that the intermolecular disulfide exchange between two Ero1␣ molecules is not favorable, and the autonomous re-oxidation of Ero1␣ occurs predominantly via intramolecular disulfide transfer.
The Ero1␣ C99/104/166A mutant with slow autonomous reoxidation was used to further study the contribution of PDI to the inactivation dynamics of Ero1␣. Consistent with the data in cells that rapid inactivation of Ero1␣ depends on active PDI (Fig. 2), oxidized PDI WT markedly accelerated the reappearance of Ox2, with ϳ70% of Ero1␣ being oxidized to Ox2 state at 15 s and complete oxidation at 5 min (Fig. 4, C and D). PDI C1 and PDI C2 promoted the re-oxidation of Ero1␣ as efficiently as PDI WT, suggesting that the two active sites functions also independently in the inactivation of Ero1␣. The catalytic inactive mutant PDI C1/2 did not show any effect as expected. Again, the isolated PDI a and PDI aЈc domains accelerated the re-oxidation of Ero1␣, albeit PDI aЈc was less efficient (Fig. 4, C  and D). Taken together, the inactivation process of Ero1␣ as monitored by the re-oxidation of Cys94-Cys131 regulatory disulfide can be directly promoted by either active site of PDI.
Asymmetric Oxidation of PDI by Ero1␣ Ensures Efficient Oxidative Folding-The above results showed that the two catalytic domains of PDI contribute to the activation/inactivation of Ero1␣ equally, whereas we and others previously observed that Ero1␣ drives substrate oxidation preferentially through the aЈ domain of PDI (8,12). So we explored the biological significance of asymmetric oxidation of PDI by Ero1␣ catalytic activity. The gel-based methoxy polyethyleneglycol 5000 maleimide (mPEG-5k) modification assay can monitor the redox state change of PDI (7,27), and PDI oxidized by Ero1␣ would have less free cysteines to be modified by mPEG-5k and migrate faster. As shown in Fig. 5A after reacting with Ero1␣ for 30 min, PDI WT was not completely oxidized with significant fully/ semi-reduced species remained, PDI C1 was more oxidized, and PDI C2 was barely oxidized. As reduced PDI is required to rearrange the non-native disulfides (2), we speculated that the asymmetric and incomplete oxidation of PDI by Ero1␣ may be important for PDI to catalyze disulfide isomerization in addition to disulfide formation. To test this hypothesis, reactivation of scrambled RNase A with promiscuous disulfides in the Ero1␣/PDI system was used for isomerase activity assay. As shown in Fig. 5C, after oxidation by Ero1␣ the isomerase activities of PDI WT and PDI C1 were compromised, but not completely depleted. Of note, the isomerase activity of PDI C2 was little affected, supporting the idea that the a domain of PDI is resistant to Ero1␣ oxidation and can function as an efficient isomerase. To further examine this rationale, yeast Ero1p/ Pdi1p system was studied because the oxidation pattern of the two active sites in Pdi1p catalyzed by Ero1p is different from that in mammalian system (28). Under our experimental conditions Pdi1p was very efficiently oxidized by Ero1p with little fully/semi-reduced Pdi1p left at 30 min, and the oxidation rate of Pdi1p C2 by Ero1p was slightly faster than that of Pdi1p C1 (Fig. 5B), which is quite different from the Ero1␣/PDI system indeed (Fig. 5A). Consequently, the isomerase activities of Pdi1p WT and its active site mutants were fully suppressed upon Ero1p oxidation (Fig. 5D).
To gain further insights into the merits of asymmetric oxidation of PDI by Ero1␣, we monitored the de novo oxidative folding of another classical folding substrate, bovine pancreatic trypsin inhibitor (BPTI). Native BPTI contains three disulfide bonds, and its successful folding depends on the rearrangement of the kinetically trapped disulfide intermediates (29). The quantitative dissection of BPTI refolding steps showed that in the first 10 min Ero1␣/PDI catalyzed the oxidation of reduced BPTI to various intermediates, which were further oxidized to native BPTI at 60 min with the yield over 90% (Fig. 5E). In Ero1p/Pdi1p system, although the oxidation of reduced BPTI in the first 10 min was slightly faster than that in Ero1␣/PDI system, the one-disulfide and two-disulfide species remained in later steps and only ϳ50% BPTI was refolded to native state after 60 min (Fig. 5F). Thus, human Ero1␣/PDI system is more efficient than yeast Ero1p/Pdi1p system in proofreading nonnative disulfides. In conclusion, the asymmetry in Ero1␣ catalyzed oxidation of PDI is functionally significant, which makes the aЈ domain act primarily as an oxidase and the a domain act as an isomerase so as to ensure efficient oxidative folding of client proteins.
Role of Hydrophobic Interaction in Ero1␣-PDI Interplay-We further explored the mechanisms underlying the intriguing phenomenon that both catalytic domains of PDI can react with the regulatory disulfides of Ero1␣ without discrimination but only the aЈ domain can mediate efficient electron transfer to the outer active site of Ero1␣. We and others previously found that the principle substrate binding domain bЈ in PDI is critical for the Ero1␣/PDI disulfide relay (12,14), therefore we studied whether the substrate binding ability of PDI is also required for the activation/inactivation of Ero1␣. Here we took the advantage of the mutation of two residues Phe-275 and Ile-289 in the bЈ domain of PDI, which dramatically eliminates binding with peptide or Ero1␣ (10,30). Surprisingly, PDI binding mutant F275W/I289A at reduced form was able to reduce the Cys94-Cys131 regulatory disulfide in Ero1␣ as efficiently as PDI WT (Fig. 6A), and the oxidized binding mutant also facilitated the re-oxidation of Ero1␣ (Fig. 6B). Thus, the peptide binding ability of PDI bЈ domain is not required in the reduction/oxidation of the regulatory disulfides in Ero1␣.
To explore which form of Ero1␣ binds with PDI, HA-tagged PDI WT was further used to co-immunoprecipitate Ero1␣ in HeLa cells. Only Ox1 but not Ox2 species of Ero1␣ was detected with PDI under nonreducing conditions (Fig. 6C), indicating that PDI specifically recognizes the active form of Ero1␣ through non-covalent interaction. A surfactant Triton X-100 and a hydrophobic probe 1-anilinonaphthalene-8-sulfonate (ANS) both markedly inhibited the oxygen consumption by active Ero1␣ during the oxidation of PDI (Fig. 6, D and E), emphasizing the functional role of hydrophobic binding in Ero1␣ and PDI disulfide relay. Interestingly, Triton X-100 and ANS only slightly affected Ero1p/Pdi1p activity (Fig. 6, D and  E), implying that the hydrophobic interaction between Ero1␣ and PDI is a critical requirement for mammals. Gel filtration chromatography of mixed PDI and Ero1␣ showed a new peak fraction with the molecular weight larger than that of the separated proteins (Fig. 6F), confirming the presence of a stable complex between PDI and Ero1␣. The majority of this complex was formed via non-covalent interaction and the minority was linked by intermolecular disulfide bridges (Fig. 6F). In line with activity assays, negligible complex of Ero1p and Pdi1p was FIGURE 5. Asymmetric oxidation of PDI by Ero1␣ endows PDI with both oxidoreductase and isomerase activities. A and B, loss of reactivity toward mPEG-5k upon disulfides formation in PDI (A) or Pdi1p (B) proteins catalyzed by Ero1␣ or Ero1p, respectively, was monitored by SDS-10% PAGE and visualized by Coomassie staining. a red , a ox , aЈ red , and aЈ ox indicate the PDI proteins with a domain reduced, a domain oxidized, aЈ domain reduced and aЈ domain oxidized, respectively. The doublet bands of each PDI species are resulted from one or both cysteines in domain bЈ being alkylated by mPEG-5k. Asterisks in B indicate the Pdi1p species with a structural disulfide (Cys90-Cys97) in domain a being reduced and modified by mPEG-5k. C and D, isomerase activities of PDI (C) and Pdi1p (D) proteins for the reactivation of scrambled RNase A were determined after incubation with or without Ero1 oxidases as indicated (mean Ϯ S.D., n ϭ 3). E and F, oxidative folding of denatured and reduced BPTI was catalyzed by Ero1␣/PDI (E) or Ero1p/Pdi1p (F) and analyzed by reverse-phase HPLC. The percentage of reduced BPTI, one disulfide containing species, two disulfides containing species and native BPTI at 0, 1, 5, 10, 30, and 60 min were quantified. Values were the mean of two independent experiments with very similar profiles. detected on gel filtration (Fig. 6G). Collectively, all the above data support our model that modulation of the regulatory disulfides in Ero1␣ by both PDI active sites is independent from the bЈ domain, while the hydrophobic interaction is necessary for the catalytic oxidation of PDI aЈ domain by active Ero1␣. In contrast, there is no stable hydrophobic binding between yeast Ero1p and Pdi1p, and both active sites of Pdi1p can freely react with the regulatory disulfides (31) as well as the catalytic disulfide ((28) and Fig. 5B) in Ero1p due to less conformational restriction.
Interplay between Ero1␣ and Other PDIs-It is known that Ero1␣ and its hyperactive homologue Ero1␤ poorly catalyze the oxidation of other PDI family members (9, 15), but transient mixed-disulfides between Ero1␣ and several PDIs were trapped in cells (32,33). We speculated these intermediates might be formed during the reduction/oxidation of the regulatory disulfides in Ero1␣ by PDIs. Therefore PDIp, P5, ERp46, ERp57, and ERp72 were overexpressed in HeLa cells to check whether they were able to function as regulators of Ero1␣. Overexpression of these PDIs to a similar level had only moderate effects on the Ox1/Ox2 ratio of Ero1␣ at steady state compared with PDI (Fig.  7A, first lane in upper panel). However, when the cells were treated with a low concentration of DTT, these oxidoreductases except ERp72 rapidly promoted the reduction of Ero1␣ (Fig.  7A), underscoring their roles in facilitating Ero1␣ activation against reductive challenge in the ER. Thus, the inefficient functional oxidation of other PDIs by Ero1␣ may not be attributed to poor Ero1␣ activation. As the substrate binding domain bЈ in PDI plays a critical role in enzymatic disulfide relay between PDI and Ero1␣, we realized that lack of the unique bЈ domain of PDI in other PDIs could be a reason for their inefficient oxidation by Ero1␣. To test this possibility, chimeric PDI-PDIs proteins were constructed by fusing one catalytic domain from each of the five PDIs to the C terminus of the rigid bbЈ base of PDI. The rates of Ero1␣ catalyzed oxygen consumption were dramatically increased in the presence of the chimeras PDI-PDIp, PDI-ERp57 and PDI-ERp72 (Fig. 7B). PDI-ERp46 only modestly increased the reaction rate and PDI-P5 was not effective (Fig. 7C). As the catalytic domains from PDIp, ERp57, and ERp72 used to generate the chimeras are aЈ-type while those from ERp46 and P5 are a-type, these results explained why other PDIs are poor substrates of Ero1␣, and strengthened the molecular mechanism we proposed that both the bЈ and aЈ domains are necessary for the functional disulfide relay between Ero1␣ and PDI (12). As expected, all the PDIs and PDI-PDIs chimeras efficiently promoted the transition of Ero1␣ from Ox2 to Ox1 state during Ero1␣ catalyzed reactions (Fig. 7D), indicating that all PDIs tested are capable to activate Ero1␣ in vitro.
Next, the abilities of PDIs to inactivate Ero1␣ were studied by using the reconstituted system. PDIp, P5, ERp46 and ERp57 re-oxidized reduced Ero1␣ as efficient as PDI, and ERp72 showed somewhat weaker effects in terms of the disappearance of fully reduced Ero1␣ (Fig. 7E). We were then interested in whether the -CGHC-active site containing motif, which is conserved among these PDIs, is sufficient to reduce/oxidize the regulatory disulfides in Ero1␣. A synthesized octapeptide (PWCGHCKA) derived from PDI in its reduced form indeed promoted the activation of Ero1␣ in a dose dependent manner, and the oxidized octapeptide facilitated the inactivation of Ero1␣ in a similar way (Fig. 7F), suggesting that the -CGHCactive site is the minimal element for regulation of Ero1␣. The high-dose of octapeptide used here implied that the intact catalytic Trx domain is optimized for efficient reduction/oxidation of Ero1␣. Altogether, the above data strongly suggested that PDIs with catalytic domain containing the -CGHC-active sites, if not all, are potent regulators of Ero1␣. However, active Ero1␣ can only efficiently catalyze disulfide production via PDI due to the specific recognition of PDI bЈ-aЈ domains.

DISCUSSION
Ero1␣ activity is tightly regulated in mammalian ER by a feedback mechanism. At resting state of cells, formation of the two regulatory disulfides (Cys94-Cys131 and Cys99-Cys104) in Ero1␣ restricts the availability of the Cys94-Cys99 outer active site (7,23). The inactive Ero1␣ must be promptly activated once robust protein oxidative folding capacity is required, and the activated Ero1␣ must be adequately inactivated to prevent over-oxidation within the ER. Here, we demonstrate that both the activation and inactivation of Ero1␣ by PDI in vitro occur very fast (Ͻ15 s). Similarly, when cells suffer or recover from a reductive challenge the reduction/oxidation of the regulatory disulfides in Ero1␣ happens rapidly within 1 min. Therefore, the interconversion between inactivated and activated Ero1␣ responding to the ER redox environment is prompt. Interestingly, we also find that after reductive challenge substantial oxi- dase activity of Ero1␣ is required to re-establish the thiol-disulfide balance in the ER on a relative longer time-scale. Previous studies have shown that after reductive challenge the resetting of the steady-state ratio of GSSG to total glutathione is very fast on a time-scale of seconds (25), but the re-oxidation of disulfides in protein substrates takes longer time to be completed (34,35). Thus, we propose when cells encounter a reductive challenge Ero1␣ is quickly activated by reduced PDI to catalyze the oxidation of GSH by cooperating with PDI, which results in fast reset of GSH/GSSG balance within 1 min, and Ero1␣ is then partly inactivated. The remaining active Ero1␣ drives the oxidation of protein substrates until the thiol-disulfide status in the ER reaches rebalance. After that, Ero1␣ is inactivated by oxidized PDI to prevent futile consumption of GSH and excessive peroxide production.
Previously, it was reported that several other PDIs (including ERp46, ERp57, ERp72, and P5) can hardly modulate the redox states of Ero1␣ at steady state, although their redox equilibrium constants with glutathione are close to that of PDI (15). By real-time monitoring the redox states of Ero1␣, we clearly show that the activation of Ero1␣ upon reductive challenge is indeed significantly promoted when PDI, PDIp, P5, ERp46, or ERp57 was individually overexpressed in cells, and these PDIs at oxidized form can accelerate the inactivation of Ero1␣ in vitro (Fig.  7). One exception with unknown reasons is ERp72, which seems less efficient than the other PDIs to regulate Ero1␣. Our new finding that many PDIs are potent regulators of Ero1␣ leads us to propose that the redox state of Ero1␣ is controlled by the redox balance of PDIs ensemble in the ER (Fig. 8). Since PDIs have substrate specificity in different tissue or cell types, the oxidase activity of Ero1␣ can thus be precisely regulated in cells with distinct proteomes. Recently, studies have reported the redox regulation of the different unfolded protein response (UPR) sensors was mediated by specific PDIs (36). Therefore, a redox-based feedback regulation loop for controlling the strength of UPR signaling likely exists in the ER, that fine-tuning Ero1␣ activity by PDIs maintains ER redox homeostasis, which will in turn keep balance of the redox states of PDIs and promote the adaptive UPR as well as attenuate the fatal UPR to avoid cell death.
In this report, we find that the reduction/oxidation of the regulatory disulfides in Ero1␣ relies on the catalytic activity of PDI but is independent from the peptide binding activity of PDI, and either active site of PDI can facilitate the regulation of Ero1␣ independently. This novel mode for PDI-mediated regulation of Ero1␣ is supported by several evidences: 1) single catalytic domain a or aЈ of PDI functions well to promote the reduction/oxidation of the regulatory disulfides of Ero1␣; 2) an octapeptide containing the -CGHC-active site of PDI alone is capable of regulating Ero1␣; 3) the activation and inactivation of Ero1␣ in cells are dramatically abolished by a PDI inhibitor 16F16, which modifies the active sites of PDI; 4) the substrate binding deficient mutant PDI F275W/I289A regulates Ero1␣ as efficiently as PDI WT. This mode also well explains why other PDIs containing the -CGHC-active site are also potent regulators of Ero1␣, even though they lack the unique bЈ domain of PDI. During preparation of this paper we noticed a very recent study claiming that the substrate binding ability of PDI is cru-cial for the inactivation of Ero1␣, according to the observation that a binding mutant PDI I272A/D346A/D348A (residue numbered without the 17-residue signal sequence) was not able to inactivate Ero1␣ under anaerobic conditions (37). This triple mutant was originally screened out for stabilization of the bЈx fragment of PDI in a capped conformation (38), but in fulllength PDI the existence of the neighboring aЈ domain limits the x-linker to be in a fully capped conformation (39). Thus, the inefficiency of the I272A/D346A/D348A mutant to inactivate Ero1␣ in their experiments may not be properly attributed to the loss of peptide binding activity.
Once being activated, the reduction of the regulatory disulfides probably increases the conformational flexibility of the regulatory loop between Cys-94 and Cys-131 (14), so that Ero1␣ can specifically bind to PDI via hydrophobic interactions (Fig. 6), like an unfolded nascent peptide being captured by the substrate binding site of PDI. By recognizing the bЈ-aЈ fragment of PDI, Ero1␣ specifically catalyzes the oxidation of the active FIGURE 8. Model for efficient oxidative protein folding in mammalian ER functioned by interplay between Ero1␣ and PDIs. At steady states, Ero1␣ in the ER lumen is predominantly in the inactive states, and PDIs oxidoreductases are present in a balanced reduced (blue) and oxidized (red) distribution (only the catalytic domain is shown for simplicity). Once the loading of reduced nascent substrates explodes, the redox homeostasis in the ER will be disturbed, and reduced substrates as well as increased GSH generate more reduced PDIs. These reduced PDIs quickly reduce the regulatory disulfides (Cys94-Cys131 and Cys99-Cys104) of Ero1␣, and liberate the outer active site located in the loop region (green arrow, right). The activated Ero1␣ specifically recognizes the bЈ-aЈ domains of PDI and preferentially oxidizes the active site in the aЈ domain, which further introduces disulfides into reduced substrates and GSH (red arrows). The asymmetric oxidation of PDI by Ero1␣ keeps the a domain in reduced state, which functions to catalyze efficient disulfide isomerization for the production of correctly folded substrates with complicated disulfides (blue arrow). Once the thiol-disulfide equilibrium in the ER is reestablished, the regulatory disulfides of Ero1␣ are easily re-formed either by self-oxidation or facilitated by oxidized PDIs to decrease the flow of oxidizing power into ER and avoid the futile oxidation cycles (green arrow, left). site in the aЈ domain of PDI. Other PDIs are poor substrates of Ero1␣ and cannot increase the ratio of activated/inactivated Ero1␣ at steady state because they cannot bind and stabilize the active form of Ero1␣ due to the lack of the peptide binding domain bЈ in PDI. Intriguingly, if the aЈ-type domain of PDIp, ERp57 or ERp72 is fused to the bbЈ base of PDI, the chimera becomes competent substrate for Ero1␣ oxidase activity. The two different interaction modes between Ero1␣ and PDIs described above ensures that the activity of Ero1␣ can be elegantly regulated by sensing the redox states of the PDIs ensemble, while Ero1␣ specifically produces disulfides through the aЈ domain of PDI, so that the ER is protected from over-oxidation caused by promiscuous oxidation of PDI a domain and other PDIs.
The biological implication of the strong preference on the aЈ domain by Ero1␣ oxidase activity has been revealed in this study. In the Ero1␣-driven oxidative folding, PDI aЈ domain acts primarily as an oxidase to transfer disulfides into folding substrate, while PDI a domain acts as an efficient isomerase to proofread incorrect disulfides. Under physiological conditions the semioxidized state of PDI is important for the efficiency and fidelity of oxidative protein folding (Fig. 8). Distinct from the mammalian system, there is little stable hydrophobic binding between yeast Ero1p and Pdi1p, which makes less conformational restriction for active Ero1p to transfer disulfides into either catalytic domain of Pdi1p. The strong oxidation of Pdi1p by Ero1p favors fast disulfide generation but compromises the isomerase activity required for catalyzing native disulfide formation (Fig. 5). Our results are in line with the observation that Pdi1p oxidase activity is critical to yeast growth and viability and less than 6% of its isomerase activity is needed (40). Actually, Pdi1p exists predominantly in oxidized state in yeast cells (27), whereas a majority of the active sites of PDI in human cells is in reduced state (41). It is known that only ϳ1% proteins (78 in 6,623 total proteins) in Saccharomyces cerevisiae are predicted to contain disulfides, much smaller than the percentage of ϳ16% (3,297 in 20,258 total proteins) in human (www.uniprot.org). Therefore, the less demand of isomerase activity in yeast than in mammals appears reasonable, and the Ero1␣/PDI system in mammals has evolved to adapt to the folding of larger and more complicated disulfide proteome.