EN460

Regulation of Plant ER Oxidoreductin 1 (ERO1) Activity for Efficient Oxidative Protein Folding

ABSTRACT
In the endoplasmic reticulum (ER), ER oxidoreductin 1 (ERO1) catalyzes intramolecular disulfide-bond formation within its substrates in coordination with protein disulfide isomerase (PDI) and related enzymes. However, the molecular mechanisms that regulate the ERO1–PDI system in plants are unknown. Reduction of the regulatory disulfide bonds of the ERO1 from soybean, GmERO1a, is catalyzed by enzymes in five classes of PDI family proteins. Here, using recombinant proteins, vacuum ultraviolet circular dichroism (VUVCD) spectroscopy, biochemical and protein refolding assays, and quantitative immunoblotting, we found that GmERO1a activity is regulated by reduction of intramolecular disulfide bonds involving Cys121 and Cys146, which are located in a disordered region, similarly to their locations in human ERO1. Moreover, a GmERO1a variant in which Cys121 and Cys146 were replaced with Ala residues, exhibited hyperactive oxidation. Soybean PDI family proteins differed in their ability to regulate GmERO1a. Unlike yeast and human ERO1s, for which PDI is the preferred substrate, GmERO1a directly transferred disulfide bonds to the specific active center of members of five classes of PDI family proteins. Of these proteins, GmPDIS-1, GmPDIS-2, GmPDIM, and GmPDIL7 (which are group II PDI family proteins) failed to catalyze effective oxidative folding of substrate RNase A when there was an unregulated supply of disulfide bonds from the C121/146A hyperactive mutant GmERO1a, because of its low disulfide-bond isomerization activity. We conclude that regulation of plant ERO1 activity is particularly important for effective oxidative protein folding by group II PDI family proteins.

The folding of most proteins synthesized in the endoplasmic reticulum (ER) involves intramolecular disulfide-bond formation catalyzed by protein disulfide isomerase (PDI) (EC 5.3.4.1) and other PDI family proteins (1– 4). PDI, composed of four thioredoxin-folded domains (designated a, b, b’, and a’) catalyzes the formation, isomerization, and reduction of disulfide bonds via one or both of the CXXC active centers, which are located in domains a and a’ and assist in oxidative protein folding (Fig. 1). Disulfides in the active centers of PDI family proteins are reduced by catalyzing the oxidation of substrate dithiol groups. The reduced active center of PDI must then be re- oxidized prior to catalyzing the next oxidation reaction.Yeast ER oxidoreductin 1p (ERO1p) (5, 6), mammalian ERO1 (ERO1α and ERO1β) (7, 8), and plant ERO1 enzymes (9–11) reportedly oxidize PDI family proteins, and the respective enzymatic properties have been characterized. ERO1p, ERO1α, and ERO1β preferentially oxidize PDI but not other PDI family proteins (ERp57, ERp72, P5, ERp46) (12, 13). Bycontrast, plant ERO1 (soybean ERO1: GmERO1a) exhibits broad substrate specificity, oxidizing five PDI family proteins: GmPDIL-1, GmPDIM, GmPDIS-1, GmPDIS-2, and GmPDIL7 (10). The two active centers of PDI are known to undergo asymmetric oxidation catalyzed by ERO1s. In the yeast PDI ortholog (Pdi1p), the active center of domain a is preferentially oxidized by ERO1p(12).

By contrast, human ERO1 oxidizes the active center of PDI domain a’ (14-16).Such asymmetric active-center oxidation by PDIs provides a major clue as to how the enzymes catalyze both oxidation and isomerization/reduction reactions. It was suggested that yeast Pdi1p domain a catalyzes substrate oxidation, whereas domain a’ catalyzes isomerase or reductase reactions (14). Asymmetric oxidation of human PDI results from docking of ERO1 to PDI domain b’,which is also a substrate binding site and aligns the active centers of ERO1 and PDI a’ (17). Because the a active center of human PDI can be oxidized via intramolecular thiol-disulfide exchange with the a’ active center, PDI domain a can also oxidize substrates.Previously, we found that GmERO1a preferentially oxidizes the a’ active center of GmPDIM, a plant ortholog of mammalian P5(10). Because the GmPDIM a active center can be oxidized via intramolecular thiol-disulfide exchange with the a’ active center, both domains a and a’ catalyze substrate oxidation. GmPDIM oxidizes both the a and a’ active centers of GmPDIL-2, which facilitates oxidative folding of substrate proteins. However, the specificity of GmERO1a for active centers of other PDI family proteins including PDI ortholog has not been characterized to date.The active center of ERO1s, known as the “shuttle cysteine pair”, oxidizes PDI and transfers the resulting electrons to oxygen molecules in concert with flavin adenine dinucleotide (FAD) via another active center known as the “active cysteine pair”, generating one hydrogen peroxide (H2O2) molecule per disulfide bond (18–23).

Hence, ERO1- catalyzed oxidation results in continual consumption of cellular glutathione (GSH) and accumulation of cytotoxic reactive oxygen species from H2O2 as a side product (19). As H2O2 is cytotoxic when overproduced, the oxidation mediated by ERO1s and reducing power (import of GSH into the ER lumen) are tightly regulated to link disulfide-bond formation with the overall ER redox status (24, 25). The activity of yeast ERO1p is regulated by reduction of the disulfide bond between the thiols of noncatalytic regulatory cysteine residues (Cys150 and Cys295) in the mechanism of oxidizing power control in the ER. Reduction of this disulfide bond induces confirmation changes that enable the transfer of electrons from the thiols of the shuttle cysteine residues to those of the active- cysteine residues. Human ERO1α is alsoregulated via reduction of the disulfide bond between thiols of shuttle Cys94 and noncatalytic Cys131 and shuttle Cys99 and noncatalytic Cys104, which directly blocks the pair of shuttle cysteines (26). A similar regulatory mechanism was reported for ERO1β (27). In ERO1α, another regulatory disulfide bond formed between Cys208 and Cys241 limits access to the FAD moiety by O2 molecules (28). Such regulatory reduction of disulfide bonds is catalyzed by reduced PDI and other PDI family proteins (29).

A third regulation system involving the formation of a dynamic mixed-disulfide complex with Cys166 and Cys165 in both the human ERO1 isoforms and PDI has been reported (30).Several systems in mammalian cells function to quench H2O2 produced by the PDI/ERO1 pathway and thus mitigate H2O2 cytotoxicity. Peroxiredoxin IV (PRDX4) (31– 34), PDI peroxidases (GPx7, GPx8) (35), and vitamin K epoxide reductase use H2O2 to oxidize PDI (36). These reaction systems are thus also useful as adjuncts to the PDI/ERO1 pathway. In addition, these enzymes exhibit a relatively broad specificity for PDI family proteins. Erp46 and P5 are oxidized by PRDX4, and ERp72, ERp57, and P5 are oxidized by GPx7/8 (i.e., oxidizing equivalents are supplied to multiple PDI family proteins by PRDX4 and Gpx7/8 using H2O2 generated by ERO1). A type II transmembrane GSH peroxidase-like protein located in the ER and Golgi apparatus of plant has also been described (37). However, whether this enzyme functions as an adjunct to the PDI/ERO1 system in the ER like GPx7/8 has not been determined. In addition, the mechanism regulating the ERO1/PDI system in plants remains to be elucidated.In the present study, we examined the mechanism by which plant ERO1 activity is regulated and the relationship of this mechanism to oxidative protein folding. We found that plant ERO1 activity is regulated by the reduction of intramolecular disulfide bonds and that this regulatory mechanism is essentialfor effective oxidative protein folding mediated by various PDI family proteins. Based on these data, we propose that the regulation of ERO1 activity functions as a safety mechanism that maintains the integrity of the oxidative protein folding process.

Results
Both human and yeast ERO1 are activated in vivo under reducing conditions created by the addition of dithiothreitol (DTT). The active form of ERO1 is detectable as a band known as “ox-1” by SDS-polyacrylamide gel electrophoresis(SDS-PAGE) under non- reducing conditions (38). Incubation of ERO1 with Escherichia coli thioredoxin 1 (Trx1), which is known to be a good substrate for ERO1s, reportedly generates ox-1 in vitro (24). Therefore, to determine whether plant ERO1 activity is regulated in a similar manner in yeast and human, we examined the redox status of GmERO1a in cultured soybean cells treated with DTT, and also in vitro during the reaction with Trx1. After treatment of soybean culture cells with DTT, a new GmERO1 band (ox-1) appeared on SDS-PAGE, with mobility on the gel lower than that of original GmERO1 (ox- 2) (Fig. 2A). Upon incubation of recombinant GmERO1a with Trx1, GmERO1a incubated for 2 min migrated slower through the gel than the un-incubated control, indicating the reduction of intramolecular disulfide bonds (Fig.2B). As Trx1 became fully oxidized by 60 min (Fig. 2C), GmERO1a migrated faster, indicating re-formation of the intramolecular disulfide bonds in GmERO1a. These results strongly suggested that activity of plant ERO1, similar to yeast and human ERO1, is regulated via reduction of intramolecular disulfide bonds. Next, we tried to identify the cysteine residues involved in the regulation of GmERO1a activity.

Alignment of the amino acid sequences of human ERO1α, yeast ERO1p and plant ERO1 (GmERO1a) revealed that active-center cysteines (shuttle cysteine pair and active cysteine pair) are conserved (Fig. 3A). In adding to the active-center cysteines, several regulatory cysteines in human ERO1α were also conserved in plants. In studies of the crystal structures, both the shuttle cysteines (Cys94, Cys99) and regulatory cysteines (Cys104, Cys131) of human ERO1α were found to be located in a single intrinsically flexible loop region (22). Unfortunately, no crystals of recombinant GmERO1a were obtained, despite repeated attempts. Therefore, vacuum-ultraviolet circular dichroism (VUVCD) spectra of recombinant GmERO1a were collected over the wavelength range 172-260 nm to compare the secondary structure of GmERO1a with that of human ERO1α (Fig. 3B). The spectrum exhibited two negative peaks around 222 and 208 nm and one positive peak around 190 nm, indicating that the recombinant GmERO1a had high α-helix content. The secondary structural contents of recombinant GmERO1a was estimated as 51.1% α-helix and 5.5% β-strand and the number of α-helix segments to be 20 (Table S1). These estimated contents by VUVCD analysis are similar to those of human ERO1α obtained from the crystal structure (48.9% α-helix and 6.4% β-strand) (Table S1).

The positions of secondary structure of GmERO1 on the amino-acid sequence are predicted by combining a neural network (NN) analysis treating the amino-acid sequence with the secondary structure data obtained from the VUVCD analysis (VUVCD-NN method) (Fig. 3C). Conserved putative shuttle cysteines (Cys113, Cys118) and adjacent putative regulatory cysteines (Cys121, Cys123, Cys146) were located in a disordered region of GmERO1a, analogous to the context of shuttle and regulatory cysteines in human ERO1α. To assess the function of the putative shuttle, active and regulatory cysteines in GmERO1a, we prepared recombinant GmERO1a mutants in which alanine residues were substituted for Cys113, Cys118 (putative shuttle cysteine residues), Cys370 (putative active-cysteine residue) and Cys121, Cys123, and Cys146 (putative regulatory cysteine residues). All mutants were expressed as soluble proteins in E. coli cells. However, the C370A mutant was not used in the experiments because this protein became insoluble and lost during the purification process. The other mutants were purified as stable, soluble proteins. These proteins were analyzed by SDS-PAGE under reducing and nonreducing conditions (Fig. 4A). The non-reducing SDS- PAGE gel indicates the oxidation state of the protein as purified, which could include a number of different disulfide bond states. Under non-reducing conditions, all of the GmERO1a mutants migrated slower on SDS- PAGE than wild-type GmERO1a, suggesting that Cys113, Cys118, Cys121, Cys123 and Cys146 form intramolecular disulfide bonds. Although there are slight differences in the VUVCD spectra among the mutants, the secondary structures of all mutants were basically similar to those of wild GmERO1a (Fig. 3B, Fig. S1 and Table S1), indicating that the mutations did not affect proper folding of the polypeptides.

To assess the function of Cys113, Cys118, Cys121, Cys123 and Cys146, we assayed the ability of the mutant GmERO1a enzymes to oxidize GmPDIL7, a soybean PDI family protein with a single redox active site CGHC in domain a that is readily oxidized by recombinant wild-type GmERO1a in vitro (39). Oxidation of GmPDIL7 by GmERO1a was assayed by experiments coupled with glutathione disulfide reductase and nicotinamide adenine dinucleotide phosphate reduced (NADPH) consumption (Fig. 4B). The reaction was performed in the presence of GSH, as a substrate for oxidation by GmPDIL7 oxidized by GmERO1a. As expected, the GmERO1a C113A and C118A mutants did not consume NADPH because these cysteines constitute the shuttle cysteine pair essential for oxidation of the substrate (Fig. 4C). A pronounced lag phase was observed in NADPH consumption by wild-type GmERO1a, which we interpreted to reflect activation of the GmERO1a catalytic activity. The rate of NADPH consumption reached a maximum 12 min after initiation of the reaction (Fig. 4C and D). The C121A and C146A GmERO1a mutants exhibited enhanced catalytic activity, with 1.2- and 1.5- fold increases in the maximal rate of NADPH consumption relative to the wild-type enzyme (Fig. 4D).

The C121/146A double mutant exhibited a higher maximal rate of NADPH consumption (1.7-fold higher than that of wild- type GmERO1a). Additionally, the time required to reach the maximum NADPH consumption rate was shortened to 8.5, 6.5, and 7 min in the C121A, C146A, and C121/146A mutants (Fig. 4D). These results suggest that both Cys121 and Cys146 normally form regulatory disulfide bonds with other cysteines (perhaps with shuttle cysteines Cys113 and Cys118), thus suppressing GmERO1a activity (Fig. 4E). Interestingly, the C123A mutant exhibited a 1.6-fold decrease in the maximal rate of NADPH consumption relative to the wild-type enzyme, and the time required to reach the maximum NADPH consumption rate was prolonged to 14 min. However, mutation of Cys123 had no other effect on either the maximum NADPH consumption rate or the lag phase observed with the C146A and C121/146A mutants. Taken together, these results suggest that the activity of GmERO1a is fine-tuned by the formation or reduction of disulfide bonds between multiple combinations of cysteines. PDI regulates the activities of yeast and human ERO1 in a feedback manner. In the reduced state, yeast and human PDI can reduce the regulatory disulfides, and thereby activate yeast ERO1p and human ERO1α/β (40, 41). In the oxidized state, human PDI can re-oxidize the regulatory disulfides, and thereby inhibit the activity of ERO1α (42). Based on those observations and our own experimental results, we hypothesized that the activation of wild- type GmERO1a, which involves Cys121 and Cys146, is dependent upon the presence of PDI. We therefore examined the ability of various PDI family proteins to reduce the regulatory disulfides of GmERO1a.

The majority of recombinant GmERO1a molecules purified from E. coli exist in the ox-2 (oxidized) state. No transition of GmERO1a from the ox-2 to the reduced and activated ox-1 form was observed in the presence of GSH but absence of PDI family protein (Fig. 5A). As expected, incubation of GmERO1a with GmPDIL-1 in the presence of GSH resulted in the appearance of the ox-1 form by 0.5 min. The ox-1 level increased with continued incubation as the ox- 2 level decreased. These results indicated that GmPDIL-1, reduced by GSH, converted GmERO1a from the ox-2 to ox-1 form. In contrast to GmPDIL-1, GmPDIL-2 did not convert GmERO1a from the ox-2 to ox-1 form. GmPDIM, GmPDIS-1, GmPDIS-2, and
GmPDIL7 did however convert GmERO1a from the ox-2 to ox-1 form. The rate of ox-2 to ox-1 conversion differed between PDI family proteins. GmPDIL-1 exhibited the highest ox-2 to ox-1 GmERO1a conversion rate, reaching a plateau within 10 min (Fig. 5B). GmPDIL-1 converted approximately 80% of GmERO1a to the ox-1 form within 5 min. To determine the contribution of the active centers in domains a and a’ of GmPDIL-1, we examined the conversion activity of the active-center mutants. Both the C418/421A and C73/76A active-center mutants converted GmERO1a from the ox-2 to ox-1 form, indicating that both active centers of can convert GmERO1a (Fig. 5B and Fig. S2A). However, the active centers differed in terms of conversion activity; the C418/421A mutant converted GmERO1a to the ox-1 form more rapidly and to a greater percentage than either wild-type GmPDIL-1 or the C73/76A mutant.

Compared with GmPDIL-1, GmPDIM catalyzed conversion of the ox-2 form of GmERO1a to the ox-1 form at a lower rate (Fig. 5C). The reaction plateaued with approximately 80% of GmERO1a converted to the ox-1 form. The C192/195A GmPDIM mutant converted GmERO1a to the ox-1 form more rapidly and to a greater percentage than the C64/67A mutant (Fig. 5C and Fig. S2B). Compared with GmPDIL-1 and GmPDIM, GmPDIS-1 and GmPDIS-2 exhibited lower activity in converting GmERO1a from the ox- 2 to ox-1 form (Fig. 5D, and E). The rates of conversion of GmPDIS-1 and GmPDIS-2 were lower than that of GmPDIL-1, with only 50% of GmERO1a converted to the ox-1 form even after 20 min. The C176/179A GmPDIS-1 mutant exhibited higher activity than the C57/60A GmPDIS-1 mutant, suggesting that domain a has higher conversion activity than domain a’ (Fig. 5D and Fig. S2C). The conversion activity of the C175/178A GmPDIS-2 mutant is comparable to the wild type and slightly higher than that of the C56/59A (Fig. 5E and Fig. S2D). GmPDIL-7, which has one active site, also converted GmERO1a from the ox-2 to ox-1 form in a manner comparable to that of GmPDIL-1 (Fig. 5A and F), converting >80% of GmERO1a to the ox-1 form within 10 min. Next, using semi-quantitative Western blot analysis, we determined the content and molar ratios of each of the PDI family proteins and GmERO1 in immature soybean cotyledons (Table 1). The GmPDIL-1 was markedly higher than that of other PDI family proteins, with a molar ratio of GmPDIL-1 to the total amount of other PDI family proteins of 3.4. Previously, we reported that a wheat ortholog of GmPDIL-1 (TaPDIL1) exhibited the highest content among PDI family proteins in the wheat caryopses throughout the grain filling period (43). Furthermore, the content of GmERO1 was extremely low compared with that of other PDI family proteins (Table 1). The molar ratio of total PDI family protein content to that of GmERO1 was 178. Collectively, these data suggest that the redox status of GmPDIL-1 plays a crucial role in the
regulation of GmERO1a activity in the ER.

Previously, we have found that plant ERO1 has a broader substrate specificity than human and yeast ERO1, which preferentially oxidize PDI (10, 39). GmERO1a oxidizes five soybeanPDI family proteins (at different rates): GmPDIL-1, GmPDIM, GmPDIS-1, GmPDIS-2, and GmPDIL7 (Fig. 6A, black bars), but not GmPDIL-2, which has a PDI-like domain organization identical to GmPDIL-1 and exhibits high oxidative refolding activity in the presence of GSH redox buffer (44). Such differences in rates of oxidation of PDI family proteins by GmERO1a likely reflect differences in the abilities of PDI family proteins to activate GmERO1a. To confirm this, we determined the oxidation rates of PDI family proteins by the C121/146A hyperactive GmERO1a. The C121/146A hyperactive GmERO1a oxidized GmPDIL-1, GmPDIM, GmPDIS-1, GmPDIS-2 and GmPDIL7 athigher reaction rates than wild-type GmERO1a (Fig. 6A, white bars), but the relative rates for the different PDI proteins was similar to the pattern observed with wild-type GmERO1a. And like wild-type GmERO1a, C121/146A hyperactive GmERO1a did not oxidize GmPDIL-2. For the human proteins, crystal structure and domain-swapping analyses demonstrated that replacement of PDI domain b’ with domain b’ of ERp57, which has a PDI- like domain organization but low reactivity with ERO1α, substantially altered the affinity and reactivity for ERO1α (17, 22, 45). Therefore, it was expected that the differences in rates of oxidation of PDI family proteins by activated GmERO1a were associated with differences in the affinity of GmERO1a for the various PDI family proteins.

Accordingly, we examined the affinity of GmERO1a for soybean PDI family proteins using far-Western blot analysis with recombinant wild-type GmERO1a as the bait and each of the PDI family proteins as prey. GmERO1a bound to all of the PDI family proteins, including GmPDIL-2, but with different affinities (Fig. 6B). The rates of oxidation by GmERO1a and affinities for PDI family proteins showed poor correlation, with exception of GmPDIL7, which was found to be a superior substrate of GmERO1a, though exhibiting very weak affinity for GmERO1a.Human ERO1α/β oxidizes domain a’ of human PDI preferentially than it oxidizes domain a (13, 15). This asymmetric oxidation of PDI domains is thought to be related to the opposing functional roles of domains a and a’ as a disulfide isomerase and disulfide oxidase, respectively. We determined the rates of oxidation of the active-cysteine mutants of PDI family proteins by wild type and C121/146A hyperactive GmERO1a. Both GmPDIL-1 (C418/421A) and GmPDIL-1 (C73/76A) wereoxidized at almost the same rate (Fig. 6C), suggesting that GmERO1a oxidizes both domains a and a’ of GmPDIL-1. However, domain oxidation of GmPDIM, GmPDIS-1, and GmPDIS-2 was asymmetric. GmERO1a oxidized GmPDIM (C64/67A) much more effectively than it oxidized GmPDIM (C192/195A) (Fig. 6D), indicating that domain a’ is more readily oxidized than domain a. In contrast to GmPDIM, domain a of GmPDIS-1 and GmPDIS-2 was preferred by GmERO1a because GmERO1a oxidized GmPDIS-1 (C176/179A) and GmPDIS-2 (C175/178A)much more effectively than GmPDIS-1 (C57/60A) and GmPDIS-2 (C56/59A) (Fig. 6Eand F).

GmPDIS-1 (C176/179A) and wild type GmPDIS-1, and GmPDIS-2 (C175/178A) andwild type GmPDIS-2 were oxidized by GmERO1a at almost the same rate. GmPDIM (C64/67A) was oxidized more slowly than wild type GmPDIM, suggesting that the active center of domain a of GmPDIM plays an important role for oxidation of domain a’.In human PDI, the redox reaction is affected by differences in the reduction potential E0 of the active-center disulfides (46, 15). However, there was no significant difference in the E0 values of soybean PDI family protein active centers (Table 2). The domain specificity of GmERO1a is therefore probably associated with the interactions between the various PDI family proteins and GmERO1a.Maintenance of the appropriate redox status in the active centers of PDI family proteins is critical for effective oxidative folding of unfolded proteins. Feedback regulation of ERO1 activity is thought to play a critical role the maintenance of proper redox status (24). We therefore examined the PDI- catalyzed oxidative folding of RNase A as a model substrate in the presence of C121/146A hyperactive GmERO1a lacking feedback regulation activity.

The maximum oxidative refolding rate of RNase A by GmPDIL-1 in the presence of C121/146A hyperactive GmERO1a increased to 1.7-times the rate observed in the presence of wild-type GmERO1a (Fig. 7A). GmPDIL-2, which is not oxidized by GmERO1a, could not mediate the refolding of RNase A. The refolding activities of GmPDIM, GmPDIS-1, GmPDIS-2, andGmPDIL-7 were significantly lower than that of GmPDIL-1 in the presence of wild-type GmERO1a. As the rates at which GmERO1a oxidized GmPDIS-1, GmPDIS-2, GmPDIM, and GmPDIL-7 were equal to or higher than that of GmPDIL-1 (Fig 5A), the very low refolding activities of these PDI family proteins in the presence of GmERO1 are likely associated with their low activities in terms of isomerizing disulfide bonds into the native configuration. Other noteworthy results were also obtained in the presence of C121/146A hyperactive GmERO1a. Minimal oxidative refolding of RNase A catalyzed by GmPDIS-1 and GmPDIS-2 was detected in the presence of C121/146A hyperactive GmERO1a (Fig. 7A).We then sought the cause of the lower activities of some PDI family proteins, such as GmPDIS-1, in the presence of C121/146A hyperactive GmERO1a. Oxidative refolding of unfolded proteins is generally thought to proceed via a two-step process, consisting of the introduction of non-native disulfide bonds in the substrate followed by isomerization of the disulfide bonds to restore the native conformation.

In order to identify the rate- determining step in the oxidative refolding by PDI family proteins, we used non-reducing SDS-PAGE to analyze the folding intermediates. These experiments focused on GmPDIL-1 and GmPDIS-1, because GmPDIL-1 is the plant ortholog of PDI, and immunoprecipitation experiments revealed that GmPDIL-1 associates with GmPDIS-1 in the ER and in vitro (Fig. S3). Incubation of denatured and reduced RNase A and GmPDIL- 1 for various times in the presence of wild-type GmERO1a resulted in the oxidation of most RNase A molecules to form an intermediate with non-native disulfide bonds, leading to the appearance of a weakly staining band at the same position as native RNase A within 5 min (Fig. 7B). With incubation between 10 and 20 min, this band increased in intensity, with an accompanying decrease in that of the intermediates. As 80% of the RNase activity was recovered within 10 min (Fig. S4A), the band migrating at the same position as native RNase A was identified as refolded RNase A. In the presence of C121/146A hyperactive GmERO1a, a clearly distinguishable native RNase A band appeared within 5 min (Fig. S4B), and most RNase A molecules were folded by 10 min (Fig. S4A), indicating that the more rapid oxidation of GmPDIL-1 by C121/146A hyperactive GmERO1a accelerated the overall oxidative folding of RNase A. The C418/421A and C73/76A active-center cysteine mutants also refolded RNase A in the presence of GmERO1a (Fig. S4A and B), suggesting that the active centers in domains a and a’ are capable of catalyzing the refolding of RNase A using oxidation equivalents supplied by both wild-type and C121/146A hyperactive GmERO1a. These results suggest that the first-step reaction (introduction of non-native disulfide bonds) is the rate-limiting step in the overall oxidative folding process catalyzed by GmPDIL-1 (Fig. 7C).

The ability of GmPDIL-1 to isomerize disulfide bonds into their native conformation is very strong. Even in the presence of 10 molar equivalents of C121/146A hyperactiveGmERO1a against GmPDIL-1, the rate of RNase A refolding catalyzed by GmPDIL-1 remained constant, although the lag time before the commencement of RNase A refolding declined (Fig. S5B).GmPDIS-1 introduced non-native disulfide bonds in all RNase A molecules within 5 min; by 5 to 10 min, oxidized intermediates and a few native RNase A molecules were generated (Fig. 7D). In the presence of C121/146A hyperactive GmERO1a, non-native disulfide bonds were introduced in all RNase A molecules within 5 min. After 10 min, the RNase A migrated as a single band immediately above that of native RNase A. As no recovery of RNase A activity was detected in the reaction consisting of GmPDIS-1 in the presence of C121/146A hyperactive GmERO1a (Fig. 7A and Fig. S4C), the band migrating immediately above native RNase A was determined to represent dead- end products containing non-native disulfide bonds. In contrast to the case of GmPDIL-1, these results suggest that the second-step reaction (isomerization of disulfide bonds into the native conformation) is the rate-limiting step in the oxidative folding process catalyzed by GmPDIS-1 (Fig. 7E).

Continuous oxidation of GmPDIS-1 by non-regulated GmERO1a results in the introduction of non-native disulfide bonds in the substrate that cannot be isomerized into the native conformation by GmPDIS-1, resulting in the accumulation of dead-end products. These data indicate that the feedback regulation of GmERO1a plays an important role in oxidative folding catalyzed by PDI family proteins such as GmPDIS-1.Because GmPDIL-1 and GmPDIS-1 associate in the ER, we examined the impact of dysregulation of GmERO1a-associated cooperative refolding with GmPDIL-1 and GmPDIS-1. At low molar ratios of C121/146A hyperactive GmERO1a to GmPDIL-1 and RNase A (0.02 μM to 1 μM and 8 μM), the maximum oxidative folding rate of RNase A increased with the addition of GmPDIS-1 in a GmPDIS-1 concentration-dependent manner (Fig. 7F). The effect of GmPDIS-1 additionwas synergistic, suggesting that GmPDIS-1 compensates for the insufficient formation of non-native disulfide bonds in RNase A by GmPDIL-1. However, at high molar ratios of C121/146A hyperactive GmERO1a to GmPDIL-1 and RNase A (3 μM to 1 μM and 8 μM), adding GmPDIS-1 inhibited oxidative refolding (Fig. 7G). In the presence of wild- type GmERO1a, by contrast, the addition of GmPDIS-1 accelerated oxidative refolding at both low and high molar ratios of wild-type GmERO1a to GmPDIL-1 and RNase A (Fig. 7H and I).

Discussion
In the present study, we demonstrated that the activity of plant ERO1 (GmERO1a) is regulated by the reduction and/or exchange of disulfide bonds involving residues Cys121 and Cys146. The location of these cysteine residues in GmERO1a is similar to that of the regulatory cysteines (Cys104 and Cys131) in human ERO1α, near the shuttle cysteine pair. Analyses of the crystal structures of constitutively active and inactive human ERO1α mutants revealed that the regulatory cysteines are located within an intrinsically flexible loop exhibiting electron shuttle activity that is finely tuned by intra-loop disulfide-bond rearrangement (22). As VUVCD-NN method predicted that Cys146 of GmERO1a is located in the intrinsically flexible loop, GmERO1a activity could be regulated in a manner similar to human ERO1α. The additional Cys208-Cys241 disulfide bond in human ERO1α functions to “clamp” the two helices that seal the flavin cofactor FAD. The a’ active center of PDI unclamps this seal by forming a Cys208- and Cys241-dependent mixed-disulfide complex with ERO1α to facilitate entry of O2 molecules into the flavoprotein active center (28, 47).

In GmERO1a, Cys217and Cys226 are located in positions similar to Cys208 and Cys241 of human ERO1α. Indeed, a brief but non- negligible lag phase in the GmPDIL-1– catalyzed oxidative folding of RNase A was observed in the presence of the C121/146A hyperactive GmERO1a, suggesting the presence of a regulation mechanism that does not involve Cys121/Cys146. We also found that substitution of Cys123 with an Ala residue diminished the oxidation activity of GmERO1a and prolonged the lag phase until the oxidation rate reached a maximum. Moilanen et al. identified a third regulatory disulfide bond involving human ERO1α Cys166 and ERO1β Cys165, which form a mixed-disulfide-linked complex with the a’ active-center cysteine of PDI in the inactive state (30). Substitution of ERO1α Cys166 or ERO1β Cys165 with an Ala residue reportedly causes activity of human ERO1s to decline over the course of the assay, with a concomitant shortening of the lag phase (i.e., enhancement of the activation rate). As the Cys123 mutation in C121/146A hyperactive GmERO1a had no effect on either activity or activation rate, this suggests that Cys123 stabilizes the active form of wild-type GmERO1a. Analyses of the interactions between GmERO1a and PDI family proteins revealed that PDI family proteins also regulate GmERO1a. However, GmPDIL-2 (which is not a substrate of GmERO1a) did not regulate GmERO1a, in contrast to the observation that human PDI family proteins, which are not good substrates of ERO1α, function as potent regulators of ERO1α/β (41). Furthermore, the activity of GmPDIL-1, GmPDIM, GmPDIS-1, GmPDIS-2, and GmPDIL7 as regulators of GmERO1a differed.

The specific characteristics of PDI family proteins that lead to differential activation of GmERO1a may be necessary to elicit the oxidative folding enzymatic activity of each PDI family protein. GmERO1a oxidizes both the a and a’ domains of GmPDIL-1. GmERO1a also preferentially oxidizes the C-terminal a’ domain of GmPDIM and the N-terminal a domains of GmPDIS-1 and GmPDIS-2. This broad specificity and the differences in domain preference of GmERO1a for PDI family proteins appears to be important in the intra- and inter-molecular relay of electrons between GmERO1a, the PDI family protein, and the substrate during oxidative folding, although the underlying structural details remain unclear. We previously found that GmPDIL-2 and GmPDIM associate in the ER and that GmPDIL-2 synergistically accelerates oxidative refolding in vitro using oxidizing equivalents obtained by GmPDIM from GmERO1a (10). In this process, a disulfide bond introduced in the active center of domain a’ of GmPDIM by GmERO1a is transferred to the active center of GmPDIM domain a and both domains a and a’ of GmPDIL-2. In the present study, we demonstrated that GmPDIS- 1 associates with GmPDIL-1 in the ER and accelerates oxidative refolding catalyzed by GmPDIL-1 in such a way as to compensate for the lack of disulfide bonds.

In this study, we showed that plant PDI family proteins are categorized into two groups based on the rate-limiting enzymatic activity (e.g., formation of non-native disulfide bonds in the substrate protein or isomerization of disulfide bonds into the native conformation). Formation of non-native disulfide bonds in the substrate protein is the rate-limiting activity of group I PDI family proteins (GmPDIL-1 and GmPDIL-2), whereas isomerization of disulfide bonds into the native conformation is the rate-limiting activity of group II PDI family proteins (GmPDIM, GmPDIS-1, GmPDIS-2, and GmPDIL7). Our folding experiments involving the C121/146A hyperactive mutant GmERO1a indicated that dead-end products containing non-native disulfide bonds that cannot be isomerized to the native conformation are generated by group II PDI family proteins. The generation of these dead- end products is prevented by feedback regulation of GmERO1a activity. The rapid generation of a supply of oxidizing equivalents by C121/146A hyperactive GmERO1a was found to accelerate the rate of oxidative folding by the group I PDI family protein GmPDIL-1. In addition, the rate of oxidative folding catalyzed by GmPDIL-1 was increased by interaction with GmPDIS-1, which associates with GmPDIL-1 in the presence of wild-type GmERO1a or C121/146A hyperactive GmERO1a at a molar ratio similar to that occurring in vivo. However, levels of C121/146A hyperactive GmERO1a above this threshold inhibit synergistic folding catalyzed by GmPDIL-1 and GmPDIS-1, demonstrating the importance of feedback regulation of GmERO1a for group II PDI family proteins. These data also importantly suggest that GmPDIL-1 can activate GmERO1a and increase supply of oxidizing equivalents via GmPDIS-1 as required in vivo by feedback. Decreased activation of GmERO1a by GmPDIS-1 and GmPDIS-2 probably prevents excessive oxidation of their active-center cysteines by GmERO1a.

In eukaryotes, the PDI/ERO1 system is a major supplier of disulfides necessary for the oxidative folding of nascent proteins in the ER. In human, electron transfer between PDI family proteins occurs in a hierarchical cascade from PDI, the main substrate of Ero1α
(13). Here, we demonstrated that plant ERO1 (GmERO1a) has broad substrate specificity, and can oxidize not only PDI ortholog (GmPDIL-1), but also other PDI family proteins (GmPDIM, GmPDIS-1, GmPDIS-2, and GmPDIL7). These findings suggest that, in the plant system, the electron transfer network of the ERO1/PDI family proteins is distinct from the human system. The oxidation power of this system is taken from O2 by ERO1, resulting in the generation of H2O2, which is highly toxic to the cells. Regulation of ERO1 activity is thus thought to be necessary to protect cells from excessive H2O2 generation (48–50). In the present study, we demonstrated that the presence of an unregulated supply of disulfide bonds provided by hyperactive plant ERO1 led to inhibition of folding. This implies that feedback regulation of ERO1 plays important roles not only in regulating the redox status of the active center of PDI family proteins, especially group II PDI proteins, but also in protection from H2O2 generation. In addition, we showed that the concentration ratio between PDI family members and ERO1 is important for efficient oxidative protein folding. Fig. 8 provides a framework for understanding oxidative protein folding in the ER of plants, at different abundance ratios of ERO1 to PDI family proteins, and for either wild-type or hyperactive ERO1. At a proper molar ratio of wild-type ERO1 to PDI family proteins, oxidized PDI family proteins introduce disulfide bonds in unfolded protein (Dred) producing protein with nonnative disulfide bonds (Dox), and reduced PDI family proteins rearrange the disulfide bonds to the native ones (N) (Fig. 8A). However, when a molar ratio of wild-type ERO1 to PDI family proteins is improperly high, too many disulfide bonds beyond the isomerization capacity of PDI family proteins are supplied by ERO1, resulting in production of denatured proteins with dead-end disulfides (Dend) simultaneously with production of native proteins (Fig. 8B). At a proper molar ratio of hyperactive ERO1 to PDI family proteins, disulfide bonds within the limit of isomerization capacity of PDI family proteins are supplied to Dred by hyperactive ERO1 via PDI family proteins and the disulfide bonds are rearranged to the native ones efficiently (Fig. 8C). However, at an improperly high molar ratio of hyperactive ERO1 to PDI family proteins, too many disulfide bonds are supplied by hyperactive ERO1, resulting in most of Dred becoming Dend (Fig. 8D). The properties of plant PDI family proteins and the regulation of plant ERO1 revealed in this work should be useful for understanding oxidative protein folding in other eukaryotes as well.

Recombinant glutathione S-transferase (GST)-GmERO1a, GST-GmPDIL-1, GST- GmPDIL-2, GST-GmPDIM, GST-GmPDIS-1, GmPDIS-2, and GST-GmPDIL-7 were expressed as described previously (10, 39, 44, 51, 52). Expression plasmids encoding the cysteine mutants of these proteins were constructed as follows. The appropriate DNA fragments were amplified by PCR using primers listed in Supplementary Table S2 and then sub-cloned into the pGEX6p-2 vector (GE Healthcare Life Science). BL21 (DE3) cells (Takara Bio, Inc.) transformed with (GST- GmERO1a and its cysteine-mutants) or without (GST-PDI family proteins and respective cysteine-mutants) pTf16 (Takara Bio, Inc.) were used for the expression of recombinant proteins. Expression of recombinant proteins was induced in the presence of 0.5 mM isopropyl-β-D- thiogalactopyranoside, extracted by sonication, and purified by GSH Sepharose 4B column chromatography (GE Healthcare Life Science) followed by digestion with PreScission protease (GE Healthcare Life Science) and further purification by TSK gel G3000SW column chromatography (Tosoh Co., Ltd.). Purified GmERO1a and respective mutants were incubated in Laemmli’s SDS-loading buffer (38) with and without reducing reagent containing 25 mM N-ethylmeleimide at 37°C for 2 h and then subjected to non-reducing SDS-PAGE. Proteins were stained with Coomassie Brilliant Blue G-250. For expression and purification of Trx1, an expression vector for Trx1 was constructed as follows. The cDNA was amplified from total RNA extracted from E. coli (TOP-10F) (Invitrogen) using the RT-PCR primers shown in Supplementary Table S2. The vector was then sub-cloned into the pET-46 Ek/LIC expression vector (Novagene). Expression of recombinant Trx1 was induced in LB medium containing 0.5 mM isopropyl-β-D- thiogalactopyranoside and 100 μg/mL ampicillin at 37°C for 3 h. Recombinant Trx1 was purified using Ni-NTA chromatography (Invitrogen) followed by gel filtration chromatography on a TSK gel G3000SW column (Tosoh).

Recombinant Trx1 was reduced using dithiothreitol and desalted using a Zeba Spin Desalting column (Thermo Fisher Scientific). Reactions with Trx1 were initiated by adding 1 μM GmERO1a to 50 μM reduced Trx1 in 50 mM Tris-HCl (pH 7.5) at 25°C. At the indicated times, reactions were quenched by addition of SDS-PAGE sample buffer containing 1 mM 4’-acetamido-4’- maleimidylstilbene-2,2’-disulfonic acid and incubation for 2 h at 37°C. After quenching, samples were separated on 15% polyacrylamide gels and stained with Coomassie Brilliant Blue G-250. GmERO1a redox status was quenched by addition of SDS- PAGE sample buffer containing 25 mM N- ethylmaleimide and incubation for 2 h at 37°C. After quenching, samples were separated on 15% polyacrylamide gels and stained with silver-staining kit (Wako Co.) Soybean DG330 (rpc00051) cells were provided by RIKEN BioResource Center which is participating in the National Bio- Resource Project of the MEXT, Japan (53). Cells were cultured, propagated, and passaged in a 200-mL Erlenmeyer flask containing 80 mL of 0.25% gellan gum-solidified Murashige and Skoog medium at 25 °C under 5000 lx continuous light.

The VUVCD measurements of GmERO1a and its cysteine-mutants were measured over the wavelength range of 172–260 nm by the VUVCD spectrophotometer in Hiroshima Synchrotron Radiation Center (HiSOR) (54). Details regarding the optical and electronic setup of this spectrophotometer are described elsewhere (54, 55). Calibration of the ellipticity amplitude and wavelength position were confirmed by monitoring the CD spectra of an aqueous solution of ammonium d- camphor-10-sulfonate. The spectra were recorded with a 1.0-mm slit using a 4-s time constant and 20 nm/min scan speed, and final spectra were obtained by averaging 8 accumulations and subtracting signals associated with the buffer solution. The secondary structure contents and numbers of segments of GmERO1a and its cysteine- mutants were estimated by the SELCON3 program (56) with the VUVCD spectra and secondary structure dataset of 31 reference proteins (57, 58). The positions of the secondary structures on the amino acid sequences of GmERO1a and its mutants were predicted by combining VUVCD analysis with the NN algorithm. This combination method was detailed previously (54, 59).Oxidation activity was assessed using a coupled assay following the decrease in absorbance at 340 nm due to the consumption of NADPH (Oriental Yeast Co., ltd.) by GSH reductase (Sigma-Aldrich, Inc.), as described previously (10, 35, 60).

A molar extinction coefficient for NADPH of 6,200 M−1cm−1 was used for calculations. GmERO1a (or respective mutant) (1 μM) was incubated with each recombinant PDI family protein (3 μM) in 100 mM N-2-hydroxyethylpiperazine-N-2- ethane sulfonic acid (HEPES) buffer (pH 7.5) containing 2 mM CaCl2, 150 mM NaCl, 3 mM GSH, 120 μM NADPH, and 1 U/mL GSH reductase (Sigma-Aldrich, Inc.) at 25°C. Dot far-Western blot analyses were carried out as described previously (10). Briefly, purified recombinant PDI family protein and bovine serum albumin (as prey proteins) were spotted onto a nitrocellulose membrane (GenScript Biotech Corp.). The membrane was blocked with 20 mM Tris-HCl (pH 8) containing 150 mM NaCl, 0.05% Tween 20, and 5% nonfat dry milk (blocking solution) at 4°C for 16 h and incubated in 0.2 μM GmERO1a as bait protein in blocking solution without nonfat dry milk (TBST) for 3 h at 4°C. The membrane was incubated with anti- GmERO1a antiserum (10) and then horseradish peroxidase-conjugated anti– guinea pig IgG as the secondary antibody (Promega Corp.) diluted with blocking solution as the secondary antibody. Blots were washed four times with TBST for 20 min/wash and developed using Western Lightning chemiluminescence reagent (Perkin Elmer Life Sciences).

Plants and cell culture Soybean (G. max L. Merrill cv. Jack) seeds were planted in 5-L pots and grown in a controlled environmental chamber at 25 °C under 16-h day/8-h night cycles. Seeds (230 mg) were collected from soybean plants, and the cotyledons were then isolated. Tissues were frozen in liquid nitrogen and then ground with a micro pestle SK-100 (Tokken). Proteins were extracted by boiling for 5 min in Laemmli’s SDS-loading buffer containing a 1% cocktail of protease inhibitors (Sigma-Aldrich). The protein concentration in the samples was measured using an RC/DC protein assay kit (Bio-Rad Laboratories). Extracted proteins were separated along with known amounts of recombinant PDI family proteins and GmERO1 on the same gel by SDS-PAGE and then blotted onto polyvinylidene difluoride membranes (Bio- Rad Laboratories). The blotted proteins were then immuno-stained using specific rabbit or guinea pig antiserum as primary antibodies and horseradish peroxidase-conjugated anti-rabbit IgG (Santa Cruz Biotechnology, Inc.) or anti– guinea pig IgG as the secondary antibody. Signals were visualized using Western Lightning Plus-ECL Enhanced Chemiluminescence Substrate (Perkin Elmer Life Sciences) and an Image Quant LAS4010 imager (GE Healthcare Life Science). The amount of each protein was calculated based on the intensities of the sample and corresponding recombinant protein standard bands using ImageJ software.

Redox equilibrium constants of PDI family proteins and their active center mutants (GmPDIL-1: C418/421A and C73/76A, GmPDIS-1: C176/179A and C57/60A, GmPDIS-2: C175/178A) were determined with wild type proteins and their as described previously (10). Briefly, PDI family proteins or respective mutants were incubated with 0.1 mM GSSG and 0.015–28 mM GSH at 25°C for 1 h in 0.1 M sodium phosphate buffer (pH 7.0) containing 1 mM EDTA and 150 mM NaCl and then incubated under N2 at 25°C for 1 h. Free thiol groups were modified by incubation on methoxypolyethylene glycol-maleimide (Fluka Sigma-Aldrich) at 25°C for 30 min. Proteins were separated by SDS-PAGE and stained with Coomassie Brilliant Blue G-250. The reduced, oxidized, and intermediate fractions were quantified using ImageJ software (U.S. National Institutes of Health). Values for the completely oxidized or reduced states were regarded as 0 or 100 %, respectively, and all intermediate states were recalibrated. The Keq value was calculated by fitting the recalibrated fraction of the apparent reduced form to the following equation: R = ([GSH]2/[glutathione disulfide (GSSG)])/{Keq + ([GSH]2/[GSSG])}, where R represents the relative ratio of the reduced forms. The protein equilibrium redox potential was calculated using the Nernst equation, (E’0 = E’0(GSH/GSSG) – [RT/nF] * ln Keq), using a GSH standard potential E’0(GSH/GSSG) of −0.240 V at pH 7.0 and 25°C.

Thiol oxidative refolding activity was assayed as described previously (61, 62).Briefly, each reaction mixture contained 100 mM HEPES buffer (pH 7.5), 150 mM NaCl, 2 mM CaCl2, GmERO1a, 2 mM cCMP, reduced and denatured RNase A, and recombinant PDI family proteins and respective mutants. The mixtures were incubated at 25°C, and the formation of active RNase A was assessed spectrophotometrically by monitoring the hydrolysis of the RNase A substrate cCMP at 296 nm.Analyses were carried out as described previously (10). In brief, the refolding reactions were performed by the addition of 1 μM GmERO1a or C121/146A hyperactive GmERO1a in 100 mM HEPES buffer (pH 7.5),150 mM NaCl, 2 mM CaCl2, and 3 μM recombinant PDI family proteins and respective EN460 mutants along with 8 μM denatured and reduced RNase A. At the indicated time points, free thiols were blocked by the addition of Laemmli’s SDS-loading buffer containing 8 mM 4-acetamido-4-maleimidylstilbene-2,2- disulfonic acid. Proteins were then separated by SDS-PAGE on a 15% polyacrylamide gel without reducing reagent. Proteins were detected by Coomassie Blue staining.All data are shown as the mean ± SE of at least three replicates. Welch’s t test (two-tailed, unpaired) or the Tukey-Kramer test were used for comparisons of data from more than three experiments.