Human ER Oxidoreductin-1α (Ero1α) Undergoes Dual Regulation through Complementary Redox Interactions with Protein-Disulfide Isomerase*

In the mammalian endoplasmic reticulum, oxidoreductin-1α (Ero1α) generates protein disulfide bonds and transfers them specifically to canonical protein-disulfide isomerase (PDI) to sustain oxidative protein folding. This oxidative process is coupled to the reduction of O2 to H2O2 on the bound flavin adenine dinucleotide cofactor. Because excessive thiol oxidation and H2O2 generation cause cell death, Ero1α activity must be properly regulated. In addition to the four catalytic cysteines (Cys94, Cys99, Cys104, and Cys131) that are located in the flexible active site region, the Cys208–Cys241 pair located at the base of another flexible loop is necessary for Ero1α regulation, although the mechanistic basis is not fully understood. The present study revealed that the Cys208–Cys241 disulfide was reduced by PDI and other PDI family members during PDI oxidation. Differential scanning calorimetry and small angle X-ray scattering showed that mutation of Cys208 and Cys241 did not grossly affect the thermal stability or overall shape of Ero1α, suggesting that redox regulation of this cysteine pair serves a functional role. Moreover, the flexible loop flanked by Cys208 and Cys241 provides a platform for functional interaction with PDI, which in turn enhances the oxidative activity of Ero1α through reduction of the Cys208–Cys241 disulfide. We propose a mechanism of dual Ero1α regulation by dynamic redox interactions between PDI and the two Ero1α flexible loops that harbor the regulatory cysteines.

Ero1 family members are highly conserved in eukaryotes, and two isoforms, Ero1␣ and Ero1␤, are found in vertebrates. Ero1␣ is expressed ubiquitously (11), whereas Ero1␤ is mostly expressed in the pancreas and stomach (12,13). Although yeast Ero1p is essential for cell viability (14), Ero1␣/␤ double-knockout mice do not display a significant phenotype, and oxidative folding of immunoglobulin proceeds normally in double knock-out cells (15). These results indicate the presence of backup systems that can complement and compensate for Ero1 in mammals.
When Ero1␣ oxidizes PDI, flavin adenine dinucleotide (FAD) bound to Ero1␣ accepts electrons from PDI and subsequently reduces O 2 to H 2 O 2 (4), which is further reduced to H 2 O by Ero1␣-associated GPx7/GPx8 (9,16,17). However, excessive formation of disulfide bonds and H 2 O 2 can cause cell death (18), and hence it is essential that Ero1␣ activity is strictly regulated. The disulfide bond pattern among four regulatory cysteines (Cys 94 , Cys 99 , Cys 104 , and Cys 131 ) in the so-called electron shuttle loop (residues Asp 90 -Cys 131 ) is a primary determinant of Ero1␣ activity; the catalytic Cys 94 -Cys 99 active site disulfide is present in the active form, whereas Cys 94 -Cys 131 and Cys 99 -Cys 104 disulfides are present in the inactive form (19 -22). More recently, we demonstrated that in addition to these four regulatory cysteines, the Cys 208 -Cys 241 pair that is conserved in vertebral Ero1 family enzymes has an auxiliary role in the regulation of Ero1␣. The Ero1␣ C104A/C131A/ C208S/C241S quadruple mutant displays higher PDI oxidation activity and H 2 O 2 production and is therefore more toxic than the Ero1␣ C104A/C131A double mutant (18). Furthermore, under high turnover conditions in which reduced PDI is constantly regenerated by reducing reagents such as GSH, the presence of Cys 208 and Cys 241 was required for maximal PDI oxidation activity (23). Additionally, formation of the Ero1␣-PDI complex involves Cys 208 and/or Cys 241 of Ero1␣ in human cells (18), and the Cys 208 -Cys 241 pair interacts with PDI to fine-tune Ero1␣ activity.
Importantly, all of the aforementioned regulatory cysteines are located in flexible loop regions of Ero1␣. It is assumed that the redox state of these cysteines influences the conformational dynamics of the flexible loops, thereby altering Ero1␣ activity. Consistent with this, the electron shuttle loop (loop I) that includes the Cys 94 -Cys 99 disulfide in the active form is more flexible and has a higher affinity for PDI than the loop containing Cys 94 -Cys 131 and Cys 99 -Cys 104 disulfides in the inactive form, which facilitates electron transfer from PDI to the bound FAD cofactor (21). In the present study, we focused on the role of another flexible loop flanked by Cys 208 and Cys 241 (loop II) in the regulation of Ero1␣ activity (see Fig. 1A). The results showed that the Cys 208 -Cys 241 disulfide was significantly reduced during Ero1␣-catalyzed oxidation of PDI, resulting in a more active Ero1␣. Deletion of loop II prevented the reduction of the Cys 208 -Cys 241 disulfide by PDI, but the overall structure of Ero1␣ was not altered. Loop II therefore serves as a platform for functional interplay with PDI, which in turn elevates Ero1␣ activity through the reduction of the Cys 208 -Cys 241 disulfide. The present findings shed new light on the mechanisms of PDImediated Ero1␣ regulation in the mammalian ER.

Results
Reduction of the Cys 208 -Cys 241 Disulfide during Ero1␣-catalyzed PDI Oxidation-Our previous observation that the Ero1␣ C104A/C131A/C208S/C241S quadruple mutant oxidizes PDI more efficiently and produces more H 2 O 2 than the Ero1␣ C104A/C131A double mutant suggests that the Cys 208 -Cys 241 disulfide has a role in regulating Ero1␣ (18,23). To investigate the reduction of the Cys 208 -Cys 241 disulfide by PDI during Ero1␣ catalysis, which leads to further elevation of Ero1␣ activity, we first sought to measure the change in redox state of Ero1␣ during PDI oxidation. However, because of significant overlap of these two similar-sized proteins in SDS-PAGE, it was difficult to monitor PDI-induced redox state changes of Ero1␣ using the method (data not shown). We therefore fused maltose binding protein (MBP) to the N terminus of PDI and reacted the resulting fusion protein (MBP-PDI) with Ero1␣.
Ero1␣ in combination with MBP-PDI consumed molecular oxygen at almost the same rate as when in combination with PDI ( Fig. 1B). As previously observed with PDI (18), MBP-PDI was oxidized more efficiently by Ero1␣-AASS (C104A/C131A and C208S/C241S quadruple mutant) than by Ero1␣-AA (C104A/C131A double mutant) (Fig. 1C, upper left and right panels). Thus, the fusion of MBP had a minimal effect on the reaction between PDI and Ero1␣. A weak but significant band (marked by red) did appear in the upper part of the Ero1␣-AA band (marked by oxi) during the early stages (0.25-1.5 min) of MBP-PDI oxidation by Ero1␣-AA and disappeared as the reaction proceeded to completion (Fig. 1C, upper left panel). This band marked red was not observed with Ero1␣-AASS in which Cys 208 and Cys 241 were mutated to Ser (Fig. 1C, upper right panel). These results suggested that a portion of Ero1␣ underwent reduction of the Cys 208 -Cys 241 disulfide by reaction with reduced PDI. Ero1␣-Cysless (C94A/C99A/C104A/C131A quadruple mutant), which is incapable of oxidizing PDI via its active site, generated an even larger amount of the red species throughout the entire reaction (0.25-10 min; Fig. 1C, middle left panel). As expected, the red band was hardly detectable for Ero1␣-CyslessSS (C94A/C99A/C104A/C131A and C208S/ C241S sextuple mutant; middle right), as was the case for Ero1␣-AASS. The slight appearance of this band with the Ero1␣-CyslessSS mutant during the later stages of the reaction may indicate only partial reduction of the two solvent-exposed disulfide bonds in the N-terminal region (Cys 35 -Cys 48 and Cys 37 -Cys 46 ). It is conceivable that the Cys 208 -Cys 241 disulfide in Ero1␣ could be reduced to a significant extent by PDI during catalysis provided that sufficient amounts of reduced PDI are present.
The red species was also abundant with Ero1␣-WT (Fig. 1C, lower left panel), and despite the C208S/C241S double mutations, this was also the case with Ero1␣-WTSS, albeit to a lesser extent than with Ero1␣-WT (Fig. 1C, lower right panel). This observation suggests that a disulfide bond, possibly in the loop I of Ero1␣, can also be reduced (to some extent) by PDI.
To confirm that the red band actually corresponded to a species with a reduced Cys 208 and Cys 241 , we carried out peptide footprinting analysis. Gel fragments corresponding to the red and oxi bands of Ero1␣-Cysless were individually subjected to in-gel digestion, and the resulting short peptides were separated by HPLC and analyzed by MALDI-TOF MS (Fig. 1, D and  E). The peptide fragments identified are listed in supplemental  Table S1. The elution peak marked by # had a molecular mass equivalent to that of Ero1␣ residues Gly 193 -Lys 215 containing an AMS-modified Cys 208 (Fig. 1, D,   in air-saturated buffer at 30°C. At indicated time points, the reaction mixture was quenched with TCA, washed with acetone, and modified by maleimide-PEG-2k or AMS. The redox states of MBP-PDI and Ero1␣ mutants were separated by non-reducing SDS-PAGE and stained with CBB. Note that the red species of Ero1␣-WT are observed at 0 min because Ero1␣-WT includes OX1 and OX2 forms that differ in their electrophoretic mobility on non-reducing gels without modification, as reported previously (19,21,41).

Ero1␣ Regulation via Redox Interplay with PDI
HPLC/MALDI-TOF MS analysis (Fig. 1F). The synthetic peptide was eluted at the same elution time as the peak marked # and had an identical molecular mass. The corresponding elution peak was significantly weaker in the elution profile of digested peptides derived from the oxi band ( Fig. 1, D, lower panel, and E, lower panel). An Ero1␣ species with reduced Cys 208 and Cys 241 was therefore generated during Ero1␣-catalyzed PDI oxidation. . Note that Ero1␣-CyslessSS migrates more slowly on a non-reducing gel than does Ero1␣-Cysless, because of the absence of the Cys 208 -Cys 241 long range disulfide. n.s., not significant; *, p Ͻ 0.05; **, p Ͻ 0.01. B, time course of the redox state changes of Ero1␣ mutants during incubation of Ero1␣-Cysless or Ero1␣-CyslessSS (4 M each) with reduced PDIs (10 M each). All experiments were performed as described for Fig. 1C. The redox states of Ero1␣ mutants were detected as described in A. C, quantification and statistical analysis of the red fraction of Ero1␣-Cysless shown in B (n ϭ 3, means Ϯ S.D.). D, redox states of Ero1␣-Cysless during reaction with the reduced form of PDI mutants in which a CXXC sequence in either the a or a domain is replaced with AXXA. All experiments were performed as described for A (left panel). The right panel indicates the quantification of red Ero1␣-Cysless species shown in the left panel (n ϭ 3, means Ϯ S.D.). **, p Ͻ 0.01; ***, p Ͻ 0.001. E, time course of the redox state changes of Eo1␣-Cysless during incubation with the reduced form of PDI active-site mutants. All experiments were performed as described in Fig. 1C. F, Quantification and statistical analysis of the red fraction of Ero1␣-Cysless shown in E (n ϭ 3, means Ϯ S.D.). WB, Western blotting. Efficient Reduction of the Ero1␣ Cys 208 -Cys 241 Disulfide by PDI and ERp46 -We next explored whether the Cys 208 -Cys 241 disulfide in Ero1␣ could be reduced selectively by PDI or nonselectively by PDI family members. To this end, we assessed how efficiently Ero1␣-Cysless was converted into the red species upon reaction with separate PDI family enzymes. Whereas Ero1␣-Cysless could be reduced to some extent by all tested PDIs, PDI and ERp46 were the particularly efficient producers of the red species of Ero1␣ within 0.25 min ( Fig. 2A). Meanwhile, Ero1␣-CyslessSS did not exhibit this upward band shift upon addition of any PDI family enzyme, as expected because of the lack of Cys 208 and Cys 241 . ERp57, which has a similar overall domain arrangement to PDI, catalyzed a slower reduction of the Cys 208 -Cys 241 disulfide than did PDI. After a 10-min incubation, however, ERp57 generated an equivalent amount of the red Ero1␣ species as did PDI (Fig. 2B). Statistical analysis of the time course of the change in redox state of Ero1␣ indicated that PDI and ERp46 produced the red Ero1␣ species at a faster rate than did the other tested PDIs (Fig. 2C).
We next explored whether the Ero1␣ Cys 208 -Cys 241 pair has a preference for thiol-disulfide exchange with the PDI a domain active site or the a domain active site. For this purpose, Ero1␣-Cysless was incubated with two different PDI mutants in which the redox active site cysteines in either the a or a domain were replaced by alanines. The resulting constructs were named PDI AXXA-CXXC and PDI CXXC-AXXA, respectively. PDI AXXA-CXXC reduced Ero1␣-Cysless more efficiently than PDI CXXC-AXXA (Fig. 2D), indicating that the C-terminal a domain serves as a more efficient reductant of the Ero1␣ Cys 208 -Cys 241 disulfide than the N-terminal a domain. The time course of the change in the Ero1␣ redox state confirmed the higher reactivity of PDI AXXA-CXXC with the Ero1␣ Cys 208 -Cys 241 disulfide compared with PDI CXXC-AXXA (Fig.  2, E and F). These results were consistent with our previous observation of mixed disulfide complexes trapped in the ER of living cells (18).
Role of the Cys 208 -Cys 241 Disulfide in the Overall Structure of Ero1␣-In general, there are three functions of protein disulfide bonds: structural disulfides that stabilize the conformation of proteins (24 -27), catalytic disulfides that directly engage in redox reactions, and regulatory disulfides that modulate enzymatic activities (3,4). To investigate the effect of the Cys 208 -Cys 241 disulfide on the conformational stability of Ero1␣, we performed differential scanning calorimetry (DSC) measurements for Ero1␣-AA, Ero1␣-AASS, Ero1␣-WT, and Ero1␣-WTSS at scan rates of 60, 120, and 200°C h Ϫ1 (Fig. 3A and supplemental Fig. S1, A and C). Although Ero1␣ appears to assume a single-domain globular fold, all constructs displayed two or more melting points during heat denaturation, indicating a non-2-state denaturation process for Ero1␣. Thermodynamic parameters of this denaturation were calculated by fitting a non-2-state model using ORIGIN software (Origin Lab) (Fig. 3B, supplemental Fig. S1, B and D, and Table 1). The denaturation of Ero1␣ involved two melting temperatures, T m 1 and T m 2, a three-state model that included native, partially denatured and fully denatured states was the best fit. Ero1␣-AASS showed a 3.8°C higher T m 1 value than Ero1␣-AA (Table 1), and a significant increase in T m 1 was observed with the double mutant C208S/C241S in the presence of Ero1␣-WT-based constructs (i.e. Ero1␣-WTSS versus Ero1␣-WT). Additionally, Ero1␣-WT-based constructs tended to have higher T m 1 values than Ero1␣-AA-based constructs, indicating that the presence or absence of disulfide bonds in loop I and loop II influences the first thermal denaturation step of Ero1␣. By contrast, T m 2 was affected to a much lesser extent by disulfide bonds in loops I and II ( Table 1), suggesting that the second denaturation step could be ascribed primarily to the denaturation of the main ␣-helical domain of Ero1␣. In general, the T m 2 value was comparable among all Ero1␣ mutants tested, regardless of the scan rate (supplemental Fig. S1, A and C, and Table 1), indicating that Ero1␣ retained virtually all thermodynamic stability upon deletion of the Cys 208 -Cys 241 disulfide.
To gain further insight into the role of the Cys 208 -Cys 241 disulfide in the overall fold of Ero1␣, we carried out small angle X-ray scattering (SAXS) measurements. Fig. 4A shows the SAXS profiles of Ero1␣-AA and Ero1␣-AASS extrapolated to zero concentration. Guinier plots are linear without any upward curvature at low Q 2 (Fig. 4A, inset), indicative of no protein aggregation. The apparent radius of gyration, R g , and the normalized forward intensity, I(0)/c, were determined from the slope and intercept of the linear fits (supplemental Fig. S2 and Table 2). The R g values of Ero1␣-AA and Ero1␣-AASS were estimated to be 26.6 Ϯ 0.1 and 26.2 Ϯ 0.1 Å, respectively. The molecular mass estimated from the normalized forward intensity I(0) value using BSA (66.4 kDa) as the standard reference was 54.3 kDa for both Ero1␣-AA and Ero1␣-AASS, suggesting that both were monomeric in solution. Furthermore, the pair distribution function, P(r), was calculated from the SAXS curves using GNOM (28), and P(r) for Ero1␣-AASS was almost superimposable to that of Ero1␣-AA (Fig. 4B), although the largest r value (D max ) of Ero1␣-AA and Ero1␣-AASS was 96 and 92 Å, respectively ( Table 2). These results suggest that the overall molecular shape of Ero1␣ was not altered by the C208S/ C241S double mutation.
In our previous crystallographic study on human Ero1␣, the electron density was completely missing for both loop I and loop II (Fig. 1A), indicating considerable flexibility in these two segments (21). To further analyze the conformation of Ero1␣ in these flexible regions, the ensemble optimization method (EOM) was performed using the SAXS data (29) and the Ero1␣ crystal structure (Protein Data Bank code 3AHQ) as a rigid body scaffold. Consequently, whereas a SAXS curve estimated from any certain conformation of Ero1␣ did not coincide with the observed curve, ensembles of multiple conformations with distinct loop I and loop II structures produced a better fit (Fig.  4C). However, it was impossible to reach an optimal ensemble of different R g and D max values for both Ero1␣-AA and Ero1␣-AASS, because the conformational distribution did not converge into a unique pattern (Fig. 4D). Such diversity in allowed conformation ensembles suggests that Ero1␣ in solution is likely to be highly dynamic, particularly in the loop I and loop II regions.
Together, the thermodynamic and structural analyses indicated that the overall fold of Ero1␣ was not altered by mutation of the Cys 208 -Cys 241 pair. In other words, the Cys 208 -Cys 241 disulfide functions mainly as a regulatory disulfide that fine-

Ero1␣ Regulation via Redox Interplay with PDI
tunes the activity of Ero1␣. However, the influence of its structural role that determines the thermodynamic stability of Ero1␣ cannot be ignored (see the next section).
Deletion of Loop II Mimics the Phenotypes of Ero1␣-AASS-The fact that loop II is not conserved in yeast Ero1p suggests that Ero1 family enzymes in higher eukaryotes acquired a new mechanism of regulation through the insertion of this loop region. To explore the enzymatic role of this loop, we constructed Ero1␣ mutants Ero1␣-⌬AA and Ero1␣-⌬AASS, in which the Arg 216 -Gly 239 segment corresponding to loop II was deleted in Ero1␣-AA and Ero1␣-AASS, respectively. We first monitored the time course of PDI oxidation by these Ero1␣ mutants in the absence of GSH (Fig. 5A, upper panel). Ero1␣-AASS oxidized PDI with higher efficiency than did Ero1␣-AA ( Fig. 5B and supplemental Fig. S3). Noticeably, both Ero1␣-⌬AA and Ero1␣-⌬AASS oxidized PDI faster than Ero1␣-AA, indicating that deletion of loop II enhanced the PDI oxidation activity of Ero1␣. Thus, these mutants
As previously reported (23), the presence of GSH dramatically changes the ranking of the oxidative activity of Ero1␣ mutants; under conditions where reduced PDI is constantly regenerated by GSH, NADPH consumption driven by Ero1␣-AASS was not faster but rather slower than that by Ero1␣-AA (Fig. 5, A, lower panel, and C). Again, both Ero1␣-⌬AA and Ero1␣-⌬AASS displayed similar kinetics to Ero1␣-AASS; mutants lacking loop II oxidized PDI more slowly than Ero1␣-AA in the presence of GSH.
We also measured O 2 consumption and H 2 O 2 production in the absence and presence of GSH. These assays also demonstrated that although the Ero1␣ mutants lacking loop II oxidized PDI faster and hence generated higher levels of H 2 O 2 than Ero1␣-AA in the absence of GSH (Fig. 5, D, left panel, and  E, left panel), the opposite was true in the presence of GSH; Ero1␣-AA consumed O 2 fastest and hence produced the largest amount of H 2 O 2 among the Ero1␣ derivatives tested (Fig. 5, D,  right panel, and E, right panel).
To address the mechanistic influence of loop II on Ero1␣ activity in the presence of GSH, we analyzed the redox state of Ero1␣ loop II deletion mutants during catalytic PDI oxidation. As expected, a significant portion of Ero1␣-AA was reduced by PDI in the presence of GSH (Fig. 6A, left panel), but this reduction was significantly delayed with Ero1␣-⌬AA (Fig. 6A, right  panel). Thus, deletion of loop II inhibited PDI-mediated reduction of the Cys 208 -Cys 241 disulfide. Given that the redox interaction between PDI and the Cys 208 -Cys 241 pair is rate-limiting under such reductive conditions (23), this finding likely explains the compromised catalytic activity of Ero1␣-⌬AA in the presence of GSH (Fig. 5, C-E; see also "Discussion"). Furthermore, in the absence of GSH, Ero1␣-⌬AA was reduced by PDI to a much lesser extent compared with Ero1␣-AA (Fig. 6B,  left panel), as was Ero1␣-⌬Cysless compared with Ero1␣-Cysless (Figs. 2B and 6B, right panel). These results suggest that loop II plays a significant role in the efficient reduction of the Ero1␣ Cys 208 -Cys 241 disulfide by PDI. However, in the absence of GSH, loop II deletion mutants appear to adopt an active, Ero1␣-AASS-like conformation that is independent of the redox state of the Cys 208 -Cys 241 pair (see "Discussion").

Discussion
Our previous study demonstrated that the human Ero1␣ C104A/C131A/C208S/C241S quadruple mutant oxidizes PDI more rapidly and therefore produces more H 2 O 2 than does Ero1␣ C104A/C131A with the Cys 208 -Cys 241 disulfide intact. As a result, cell proliferation is greatly inhibited, and cell viability is significantly diminished (18), indicating a physiologically significant role in regulation for Ero1␣ Cys 208 and Cys 241 . Consistent with this, the Cys 208 -Cys 241 pair is highly conserved in Ero1 family enzymes in higher eukaryotes, although not in the yeast enzyme Ero1p (4).
The present study revealed that mutation of Cys 208 and Cys 241 had little effect on the thermodynamic stability and overall shape of Ero1␣ (Figs. 3 and 4B, supplemental Fig. S1, and Tables 1 and 2), leading us to conclude that the Cys 208 -Cys 241 disulfide functions mainly as a regulatory element that fine-tunes the Ero1␣ activity. The Cys 208 -Cys 241 disulfide was preferentially reduced by two members of the PDI family; PDI and ERp46 (Fig. 2, A-C). In this context, our previous studies indicated that Ero1␣ is capable of selectively oxidizing PDI and, to a lesser extent, ERp46 (5), although Ero1␣ is incapable of efficiently oxidizing other members of the PDI family (30). The present study also showed that the Cys 208 -Cys 241 disulfide was reduced by the PDI a domain more efficiently than by the a domain (Fig. 2, D-F). In line with this, Ero1␣ preferentially oxidizes the PDI a domain and oxidizes the a domain to a lesser extent (30,31). Thus, there seems to be a close correlation between the reactivity of PDIs with the Ero1␣ Cys 208 -Cys 241 disulfide and the selective oxidation of PDIs by the Ero1␣ active site Cys 94 -Cys 99 .  In this study, we demonstrated that reduction of the Cys 208 -Cys 241 disulfide increased the oxidative activity of Ero1␣ (Figs.  5 and 6, A and B). As shown in supplemental Fig. S4 (B and D), the proportion of Ero1␣ with a reduced Cys 208 -Cys 241 pair increased with absolute levels of reduced PDI. We also observed that the Cys 208 -Cys 241 disulfide could be reduced more effectively by a combination of PDI and GSH than by GSH alone (supplemental Fig. S4, A and B). Together, these results show that the differences in the reactivity of Ero1␣ in the presence and absence of GSH are dependent on the availability of reduced PDI or the rate of recycling of reduced PDI. In this respect, GSH may serve as an indirect enhancer of Ero1␣ activity.
We previously demonstrated that a non-covalent Ero1␣-PDI complex forms through van der Waals contacts between a pro-

Ero1␣ Regulation via Redox Interplay with PDI
truding ␤-hairpin loop in Ero1␣ and the substrate binding pocket in the PDI b domain and via a sustained thiol-disulfide exchange between the loop I cysteines of Ero1␣ and the a domain active site of PDI during catalysis (32). More recently, however, we found that a non-catalytic mixed disulfide complex involving Ero1␣ Cys 208 or Cys 241 (Ero1␣-PDI fast ) is the predominant species detectable in the ER at steady state (18). However, whereas the Ero1␣-PDI fast complex appears to be kinetically stabilized in cells, this was hardly detectable in in vitro assays, suggesting that there might be additional players that stabilize the mixed disulfide complex in cells. Similar mixed disulfide complexes were also observed with ERp57 (18), which has a very similar overall domain arrangement to PDI (33). ERp46, another preferred substrate of Ero1␣, is composed of three Trx-like domains linked by unusually long flexible loops, where the three solvent-exposed redox active sites are separately located (34). These structural features of PDI and ERp46 could be advantageous for their close access to the regulatory cysteines of Ero1␣, leading to the facilitated activation of Ero1␣.
EOM analysis suggested that loop I and loop II are highly flexible in solution (Fig. 4D), and this flexibility is likely key to the functional redox interplay between Ero1␣ and PDI. Accordingly, deletion of loop II compromised the efficiency of the  Fig. 1C. B, redox state changes of Ero1␣ loop II deletion mutants during incubation with reduced MBP-PDI. All experiments were performed in air-saturated buffer at 30°C by mixing each Ero1␣ mutant (4 M) with reduced MBP-PDI (10 M). At 0.25 min into the reaction (upper panels) or at other indicated time points (lower panels), the reaction mixture was quenched with TCA, washed with acetone, and modified by AMS. The redox states of Ero1␣ mutants were assessed as described for Fig. 1C. C, newly proposed model of the dual regulation of Ero1␣ via redox interplay between PDI and flexible loops I and II of Ero1␣. Under conditions where reduced PDI is continually provided through reduction by GSH, the Cys 208 -Cys 241 pair is reduced in a large proportion of Ero1␣ molecules, leading to a further elevation of Ero1␣ activity. Close and Open indicate Cys 208 /Cys 241 disulfide bonded and Cys 208 /Cys 241 reduced Ero1␣, respectively. This model is based mainly on the results of in vitro experiments performed in the present study. Its applicability to Ero1␣ regulation in living cells needs to be further verified. NOVEMBER 11, 2016 • VOLUME 291 • NUMBER 46

JOURNAL OF BIOLOGICAL CHEMISTRY 23961
Cys 208 -Cys 241 disulfide reduction by PDI. At this stage, the PDI b domain that houses the substrate binding pocket may perform an auxiliary role in another mode of the Ero1␣-PDI interaction. Specifically, Ero1␣ loop II might be recognized by the substrate binding pocket in the PDI b domain, resulting in the facilitated reduction of the Cys 208 -Cys 241 disulfide by the PDI a domain.
The presence of loop II was intrinsically inhibitory to the activity of Ero1␣. Accordingly, both Ero1␣-⌬AA and Ero1␣-⌬AASS displayed higher PDI oxidation activity than Ero1␣-AA in the absence of GSH (Fig. 5, A, upper panel, and B). Importantly, the effects on Ero1␣ activity caused by deletion of this loop and the substitution of Cys 208 and Cys 241 were neither additional nor synergistic to each other. In this context, a previous molecular dynamics simulation suggested that the removal of the Cys 208 -Cys 241 disulfide induces a slight but significant rearrangement of the four helices embracing the FAD moiety, resulting in the enhanced access of O 2 to the FAD isoalloxazine ring (18). Presumably, deletion of loop II perturbs the local conformation in the vicinity of the engineered site and possibly the orientation of the two helices connected by loop II, resulting in similar effects on the structure and functionality of Ero1␣ as occur with the C208S/C241S double mutations. As demonstrated previously (18), a loop II deletion mutant that is directly comparable with Ero1␣-⌬AA was not properly expressed in human cells, possibly because of compartmentspecific ionic strength, redox balance, or degradation factors. Thus, loop II appears to be essential for the stability of Ero1␣ in the native environment of the ER.
The loop II deletion mutants presented fundamentally different phenotypes in the presence of GSH. Under reducing conditions in which Ero1␣ works at maximal turnover rate (23), these mutants exhibit a defect with regard to PDI oxidation and H 2 O 2 production. Our results indicate that this defect may be due to a decreased propensity for the deletion mutants to become reduced at the Cys 208 -Cys 241 disulfide by PDI. This defect resembles the compromised activity of Ero1␣-AASS under reducing conditions, which cannot be activated by reduced PDI because of the absence of the Cys 208 -Cys 241 pair (23).
Collectively, our results showed that loop II provides a platform for functional interplay with reduced PDI. Consequently, our in vitro experiments establish a novel regulation mechanism of human Ero1␣ (Fig. 6C). During Ero1␣ activation, reduced PDI acts on the regulatory disulfides Cys 94 -Cys 131 and Cys 99 -Cys 104 that are located in loop I to generate active Ero1␣ with the Cys 94 -Cys 99 disulfide (35). The active Ero1␣ then undergoes further activation through reduction of the Cys 208 -Cys 241 disulfide by the reduced PDI a domain, possibly leading to more efficient entry of O 2 into the FAD-binding pocket. We previously reported that the four regulatory cysteines in loop I communicate with the Cys 208 -Cys 241 pair in loop II, either intramolecularly or intermolecularly. Furthermore, cleavage of the Cys 208 -Cys 241 disulfide appears to involve a thermodynamic preference for the Cys 94 -Cys 131 disulfide over the Cys 94 -Cys 99 disulfide, which stabilizes the inactive form of Ero1␣ (23). Thus, the Ero1␣ hyperactive state with the Cys 94 -Cys 99 disulfide and reduced Cys 208 and Cys 241 is presumably short-lived unless reduced PDI is continually provided. We propose that the Cys 208 -Cys 241 pair acts as the second PDImediated switch that controls Ero1␣ activity by somehow communicating with the four regulatory cysteines in loop I. At this stage, the availability of reduced PDI, which is representative of the redox environment in the ER, essentially regulates the disulfide bond pattern in loop I, as well as the redox state of the Cys 208 -Cys 241 disulfide in loop II. These findings highlight a dual regulation mechanism of mammalian Ero1␣ that proceeds via redox interplay between PDI and both loop I and loop II of Ero1␣.

Experimental Procedures
Recombinant Protein Expression and Purification-cDNAs encoding human Ero1␣ and human PDIs (PDI, ERp72, ERp57, ERp46, and P5) and their mutants were subcloned into the NdeI and BamHI sites of the pET15b vector (Novagen). Proteins were overexpressed in Escherichia coli strain BL21 (DE3) and purified as described previously (18). Plasmids for overexpression of a set of Ero1␣ mutants with the deletion of the Arg 216 -Gly 239 segment were constructed using a PrimeSTAR mutagenesis basal kit (Takara Bio). The resultant Ero1␣ mutants are single-polypeptide chains in which residues 215 and 240 are covalently linked. The expression and purification of the Ero1␣ mutants were performed as described previously (21).
Analysis of the Redox State of PDIs and Ero1␣-Purified PDIs were reduced with 10 mM DTT for 10 min at 4°C, and thereafter, DTT was removed by passing the sample through a PD-10 column (GE Healthcare) pre-equilibrated with 50 mM Tris/HCl (pH 7.5) buffer containing 300 mM NaCl. The catalytic oxidation of PDIs by Ero1␣ was initiated by mixing 10 or 100 M of reduced PDIs with 4 M of each Ero1␣ derivative in air-saturated buffer containing 50 mM Tris/HCl (pH 7.5) and 300 mM NaCl at 30°C. At selected time points, the reaction mixture was quenched with 5% TCA, washed with acetone, and dissolved in buffer containing 100 mM Tris/HCl (pH 7.0), 1% SDS, and 1 mM maleimide-PEG-2k or AMS. All samples were separated by non-reducing SDS-PAGE followed by staining with Coomassie Brilliant Blue (CBB) or immunoblotting with antibody. The band intensity was measured for each redox state of PDIs and Ero1␣ using a LAS-3000 image reader (Fujifilm).
Statistical Analysis-The statistical significance of differences was examined by one-way analysis of variance with Tukey honestly significant difference (HSD) post hoc testing. All statistical tests were performed using KaleidaGraph statistical software (Synergy Software) at a significance level of ␣ ϭ 0.05.
Identification of Digested Peptides from Reduced/Oxidized Ero1␣-CBB-stained SDS-PAGE gels were washed with 50% methanol, 10% acetic acid, and 100 mM ammonium bicarbonate, subjected to in-gel digestion with trypsin in 25 mM ammonium bicarbonate for 16 h at 37°C, and extracted with 33 and 70% acetonitrile containing 0.1% TFA. The digested peptide fragments were analyzed by HPLC (GL Science) equipped with a COSMOSIL 5C 18 -AR-II column (Nacalai Tesque) pre-equilibrated with 5% acetonitrile in 0.05% TFA and eluted with a linear acetonitrile gradient ranging from 5% to 65% at a rate of 1%/min, with detection at an absorbance of 220 nm. The molecular masses of separated peptides were determined by MALDI-TOF MS (Bruker Daltonics) (36,37). Synthetic peptides were prepared by the Fmoc (N-(9-fluorenyl)methoxycarbonyl) solid phase method, modified with AMS, and analyzed by HPLC and MALDI-TOF MS as described previously (5,34,37).
Differential Scanning Calorimetry-DSC measurements in 20 mM phosphate buffer (pH 7.0) were performed for Ero1␣-AA and Ero1␣-AASS using a MicroCal VP-capillary DSC (Malvern Instruments) at a scan rate of 60, 120, and 200°C/h (38). The results were cubic baseline adjusted, analyzed by integration of the total area under the curve, and fitted to a non-2-state model using ORIGIN DSC analysis software (Origin Lab).
Small Angle X-ray Scattering-SAXS measurements were performed for Ero1␣-AA, Ero1␣-AASS, and BSA (Sigma-Aldrich) as a reference for determination of the molecular mass (39) in 20 mM phosphate buffer (pH 7.0) containing 150 mM NaCl and 5% glycerol at the RIKEN SPring-8 Beamline BL45XU (Hyogo, Japan) (40). For each sample, 20 SAXS images were collected using a PILATUS 3X 2 M detector (DECTRIS) at an X-ray wavelength of 1.0 Å with a camera distance of 2.0 m and an exposure time of 1 s at 20.2°C. Analysis of SAXS data were carried out as described previously (34,38).
Measurement of O 2 and H 2 O 2 Levels during Ero1␣-catalyzed PDI Oxidation-Measurement of O 2 consumption were performed with a FireStingO 2 fiber optic oxygen meter (Pyro Science) by mixing 100 M reduced PDI with 4 M of Ero1␣ mutants in air-saturated buffer or by mixing 10 M of reduced PDI with 4 M of Ero1␣ in the presence of 1 mM GSH, 1 unit of GR, and 200 M NADPH at 30°C. The level of Ero1␣-generated H 2 O 2 was measured using a Pierce quantitative peroxide assay kit (Thermo Scientific) and Multiskan FC microplate reader (Thermo Scientific) as described previously (18).