The endoplasmic reticulum (ER) provides an environment optimized for oxidative protein folding through the action of Ero1p, which generates disulfide bonds, and Pdi1p, which receives disulfide bonds from Ero1p and transfers them to substrate proteins. Feedback regulation of Ero1p through reduction and oxidation of regulatory bonds within Ero1p is essential for maintaining the proper redox balance in the ER. In this paper, we show that Pdi1p is the key regulator of Ero1p activity. Reduced Pdi1p resulted in the activation of Ero1p by direct reduction of Ero1p regulatory bonds. Conversely, upon depletion of thiol substrates and accumulation of oxidized Pdi1p, Ero1p was inactivated by both autonomous oxidation and Pdi1p-mediated oxidation of Ero1p regulatory bonds. Pdi1p responded to the availability of free thiols and the relative levels of reduced and oxidized glutathione in the ER to control Ero1p activity and ensure that cells generate the minimum number of disulfide bonds needed for efficient oxidative protein folding.

Introduction

A key step in the maturation of many secreted proteins and extracellular domains of membrane proteins is the formation of disulfide bonds in the ER. Disulfide bonds, which stabilize native and functional conformations of proteins, are formed by pairing and oxidation of cysteines during the initial folding process in the ER. In Saccharomyces cerevisiae, most biosynthetic disulfide bonds are formed by dithiol/disulfide transfer reactions with the oxidized form of protein disulfide isomerase (PDI), Pdi1p (Sevier and Kaiser, 2002; Wilkinson and Gilbert, 2004). Pdi1p, in turn, requires other oxidizing molecules to be recycled because Pdi1p cannot generate disulfide bonds by itself. We and other colleagues identified the major disulfide-generating flavoenzyme Ero1p in S. cerevisiae (Frand and Kaiser, 1998; Pollard et al., 1998). In both yeast and mammalian cells, Ero1 directly transfers disulfide bonds to PDI (Frand and Kaiser, 2000; Mezghrani et al., 2001). The oxidative capacity of the yeast ER depends primarily on the activity of Ero1p: a temperature-sensitive ero1-1 mutation decreases the resistance of yeast to the reducing agent DTT and induces the unfolded protein response with accumulation of secretory proteins, and, after prolonged incubation at the restrictive temperature, a strain with this mutation loses viability (Frand and Kaiser, 1998, 1999; Cuozzo and Kaiser, 1999; Tu and Weissman, 2002), and overexpression of ERO1 from a multicopy plasmid increases the oxidizing capacity of yeast cells, as shown by their increased resistance to DTT (Frand and Kaiser, 1998).

The catalytic cycle of Ero1p depends on a relay of disulfide bonds from the conserved active-site cysteine pair, C352-C355, which is proximal to bound flavin adenine dinucleotide (FAD) to the second shuttle cysteine pair, C100-C105, which is responsible for direct disulfide transfer to Pdi1p (Frand and Kaiser, 2000; Gross et al., 2004; Sevier and Kaiser, 2006b). In addition to these two catalytic cysteine pairs, Ero1p contains three cysteine pairs (C90-C349, C143-C166, and C150-C295) that form regulatory bonds (Sevier et al., 2007). Several lines of evidence indicate that Ero1p is inactive with regulatory bonds formed, whereas it is active with regulatory cysteines in the reduced state. Ero1p is converted from an oxidized (inactive) state to the reduced (active) state in the presence of reduced thioredoxin-1 (Trx1) as a substrate; once Trx1 is fully oxidized through the action of active Ero1p, Ero1p returns to the oxidized (inactive) form (Sevier et al., 2007). A double mutant, Ero1p-C150A-C295A, which cannot form the crucial C150-C295 regulatory bond, exhibits increased Ero1p activity in vitro and in vivo (Sevier et al., 2007).

A byproduct of Ero1p activity is the generation of hydrogen peroxide as a result of a two-electron reduction of oxygen per disulfide generated. Thus, although Ero1p activity is essential for biosynthetic disulfide bond formation, uncontrolled Ero1p activity could produce too much reactive oxygen species that would be detrimental to the cell (Gross et al., 2006). Indeed, overexpression of the hyperactive Ero1p-C150A-C295A mutant inhibits cell growth, highlighting the physiological importance of the regulatory bonds in keeping Ero1p activity in check. Mammalian Ero1-α has been shown to modulate its activity through a similar mechanism involving regulatory bonds, showing that autoregulation in response to the redox environment is a general property of Ero1 (Appenzeller-Herzog et al., 2008; Baker et al., 2008). An additional layer of protection against reactive oxygen accumulation in the ER of mammalian cells is provided by an ER resident, peroxiredoxin, which reduces hydrogen peroxide generated by Ero1 to limit peroxide accumulation (Tavender et al., 2008, 2010; Zito et al., 2010). An obvious sequence homolog to peroxiredoxin is absent in yeast; whether additional pathways functionally similar to the peroxiredoxin system exist in the ER of fungi remains to be explored.

The precise mechanism by which the regulatory bonds of Ero1 are reduced and then reoxidized will be crucial to understand how Ero1p activity is controlled and which features of the redox environment of the ER are sensed by Ero1p. Pdi1p is the most abundant oxidoreductase of the yeast ER and is the preferred physiological substrate for oxidation by Ero1p (Nørgaard et al., 2001; Vitu et al., 2010); therefore, Pdi1p is likely to play an important part in regulating Ero1p. In the mammalian ER, the disulfide relay between Ero1-α and PDI is regulated at least in part by the availability of PDI (Appenzeller-Herzog et al., 2010; Inaba et al., 2010). Interaction between Ero1-α and PDI has been shown to be facilitated by a protruding β-hairpin in Ero1-α and the b′ domain of PDI, which facilitates the interconversion between the Ox1 and Ox2 redox-active forms of Ero1-α (Masui et al., 2011). Differences between Ero1-α and Ero1p, including the lack of a β-hairpin in Ero1p and the preference of Ero1p for the a domain of Pdi1p (versus the preference of Ero1-α for the a′ domain), suggest that the yeast and mammalian Ero1-PDI systems may operate differently.

In vitro assays showing activation of Ero1p have typically used the nonphysiological substrate Trx1, which has a single CxxC-containing domain with a highly reducing reduction potential rather than the physiological substrate Pdi1p because reduced Trx1 is more effective than Pdi1p at activating Ero1p by reduction of regulatory bonds (Sevier et al., 2007; Baker et al., 2008; Vitu et al., 2010). Here, we examine the role of Pdi1p in the reduction and reoxidation of the Ero1p regulatory bonds to reveal how the activity of Ero1p is set in a living cell. We have also sought to understand the role of glutathione in this process as the most abundant redox buffer of the ER. Although glutathione itself does not activate Ero1p nor is glutathione efficiently used as a substrate for Ero1p (Tu et al., 2000; Sevier and Kaiser, 2006b), reduced glutathione can stimulate Ero1p activity when used in conjunction with Pdi1p as a substrate (Sevier et al., 2007; Baker et al., 2008). We have reconstituted this coupled reaction in vitro and show that it recapitulates properties of the ER in vivo; as long as an excess of reduced glutathione is available, Pdi1p is maintained in a state enabled to activate Ero1p, which in turn provides oxidizing equivalents to oxidize glutathione.

Results

Ero1p is capable of autonomous inactivation

Because the regulatory bonds link distant sites of the Ero1p polypeptide chain, the presence or absence of these bonds can be readily discerned by mobility differences on nonreducing SDS-PAGE (Fig. 1 A). Activity of Ero1p can be inferred from the redox state of the regulatory bonds. In live yeast cells, Ero1p is converted to the reduced (active) form in cells treated with the reducing agent DTT and then returns to the oxidized (inactive) form after DTT removal (Fig. 1 B). Activation of Ero1p by reduction of the regulatory bonds and subsequent inactivation by reformation of the regulatory bonds could also be readily observed in an in vitro assay containing recombinant Ero1p (Ero1pc, 56–424, with an N-terminal (1–55) and a C-terminal (425–563) fragment removed) and an excess of reduced Trx1, serving as an effective substrate for Ero1p (Fig. 1 C). First, Ero1pc is inactive with the regulatory bonds in a fully oxidized state. In ∼0.5 min after addition of reduced Trx1, the regulatory bonds of Ero1pc become reduced, and Ero1pc is thus activated for the oxidation of Trx1. As a consequence of Ero1pc, after ∼60 min, all of the Trx1 is oxidized, and Ero1pc returns to the oxidized inactive form (Fig. 1 C).

Activation of Ero1pc by reduction of the regulatory bonds initiates on addition of reduced Trx1, indicating that the regulatory bonds can be reduced by direct dithiol/disulfide exchange with reduced Trx1. For the end phase of the reaction in which most of the Trx1 is oxidized and the regulatory bonds of Ero1pc become reoxidized, we wished to know whether reoxidation of the regulatory bonds occurred by direct oxidation by oxidized Trx1 or instead by autonomous reoxidation by Ero1p when all of the reduced Trx1 had been consumed. To distinguish between these possibilities, we first tested the ability of active Ero1pc to oxidize regulatory cysteines in the absence of Trx1. For this experiment, 1 µM Ero1pc was incubated with 100 µM His6-tagged reduced Trx1 for 2 min, a time sufficient for complete activation of Ero1pc by reduction of the regulatory bonds. Subsequently, His6-tagged Trx1 was removed by affinity to Ni2+ beads (Fig. 2 A). Trx1 depletion was confirmed by the significant absence of Trx1 in the Ni2+ unbound fraction and corresponding presence of Trx1 bound to the Ni2+ beads (unbound faction; Fig. 2 A, bottom). A small amount of residual Trx1 was observed in the unbound fraction (Fig. 2 A, bottom), but it was anticipated that this limited fraction of Trx1 would be rapidly oxidized by Ero1pc and not contribute to our Ero1pc inactivation assay. After depletion of reduced Trx1, Ero1pc was converted from a fully reduced to an oxidized form over a period of 60 min with a half time for completion of ∼10 min (Fig. 2 B, top). The bands migrating between the fully reduced and oxidized forms likely correspond to partially oxidized states of Ero1pc. Parallel reactions using Ero1pc mutants lacking the shuttle disulfide bond (Ero1pc-C100A-C105A) or the catalytic disulfide bond (Ero1pc-C355S) in the absence of Trx1 exhibited greatly slowed or no autonomous reoxidation of Ero1pc, respectively (Fig. 2 B, middle and bottom). In addition, when the reoxidation reaction was performed in the absence of O2 as an electron acceptor (Fig. S1 A, right), Ero1pc exhibited slowed autonomous reoxidation. These observations show that Ero1pc is capable of autonomous oxidation to form regulatory bonds and that this oxidation has catalytic requirements similar to biosynthetic disulfide bond formation.

We considered the possibility that autonomous oxidation of the Ero1p regulatory bonds was promoted by the hydrogen peroxide generated by Ero1pc before removal of Trx1. Therefore, we examined the effects of hydrogen peroxide. Addition of catalase to the reaction, which should convert any hydrogen peroxide into water and oxygen (Fig. S1 B), did not significantly slow autonomous oxidation of the Ero1p regulatory bonds (Fig. S1 A, middle). Conversely, addition of 100 µM hydrogen peroxide, the maximum concentration that could be obtained from complete oxidation of 100 µM Trx1, did not enhance the oxidation of Ero1pc-C100A-C105A (Fig. S1 C).

Autonomous oxidation of Ero1p occurs via an intermolecular reaction

Autonomous oxidation of Ero1p could conceivably occur by intramolecular (in cis) or intermolecular (in trans) disulfide bond transfer. In the crystal structure of Ero1pc (Gross et al., 2004), one of the regulatory bonds, C90-C349, is close enough to the catalytic site to be oxidized in cis, whereas the other two regulatory bonds, C150-C295 and C143-C166, are located >20 Å away from the catalytic site, suggesting that the oxidation of these bonds would require involvement of additional dithiol/disulfide exchanges with other cysteines. To determine whether one Ero1p protein can oxidize the regulatory bonds of another Ero1p, we asked whether Ero1pc-C355S, which is catalytically inactive and is not capable of autonomous oxidation, could nevertheless be oxidized in trans by active Ero1pc. A maltose binding protein (MBP) fusion to Ero1pc was used in this experiment so that MBP-Ero1pc and Ero1pc-C355S could be distinguished by their distinct mobilities on SDS-PAGE. An autonomous oxidation assay was established by combining MBP-Ero1pc and Ero1pc-C355S with reduced Trx1 to reduce the regulatory bonds on both forms of Ero1p. After Trx1 was removed by Ni2+ affinity, both Ero1pc-C355S and MBP-Ero1pc were reoxidized at nearly equal rates, showing that efficient trans oxidation can occur (Fig. 2 C). Consistent with the oxidation of Ero1pc-C355S being a product of MBP-Ero1pc catalytic function, disruption of the MBP-Ero1p catalytic cycle by mutating the C100-C105 pair (MBP-ero1-C100A-C105A) prevented Ero1pc-C355S oxidation (Fig. 2 C). We also tested the effect of dilution of Ero1pc in our autonomous oxidation assay and found that a fourfold dilution of Ero1pc slowed autonomous oxidation of Ero1pc (unpublished data). Together, the trans oxidation experiment and the dilution experiment show that autonomous oxidation of the Ero1p regulatory disulfides occurs predominantly, if not exclusively, by one Ero1p protein oxidizing the regulatory bonds of another Ero1p. The simplest mechanism for such dithiol/disulfide exchange would be for successive direct transfer of the C100-C105 shuttle cysteine disulfide bond to each of the three regulatory cysteine pairs. However, we cannot rule out the possibility that more complicated dithiol/disulfide rearrangements occur. Slow oxidation of Ero1pc facilitated by the Ero1pc-C100A-C105A (but not the Ero1pc-C355S) mutant could be observed also at late time points of the autonomous oxidation assay (>20 min), suggesting that the C352-C355 active site may contribute directly to oxidation of Ero1pc.

Ero1p is activated slowly but inactivated rapidly by Pdi1p

As we have shown, reduced Trx1 is a potent activator of Ero1p, and the use of Trx1 as an activator and substrate for Ero1p in in vitro assays made it possible to document the dramatic effect that reduction and oxidation of regulatory cysteines have on Ero1p activity (Fig. 1 C; Sevier et al., 2007). However, Pdi1p is the major physiological substrate of Ero1p (Frand and Kaiser, 1999), and we were interested to use the in vitro assay for Ero1p regulation to make a comparable assessment of the ability of Pdi1p to either activate or inactivate Ero1p activity through reduction or oxidation of regulatory bonds.

First, we tested the ability of oxidized Pdi1p to inactivate Ero1p. We prepared activated Ero1pc-C100A-C105A by treatment with reduced Trx1 followed by removal of Trx1. The use of the Ero1pc-C100A-C105A mutant prevented efficient autonomous reoxidation of Ero1p, allowing us to test the ability of exogenous agents to reoxidize Ero1p. Oxidized Pdi1p and Trx1 were prepared by air oxidation during purification from Escherichia coli. The endpoint of air oxidation yielded Pdi1p that had the a domain almost fully oxidized and the a′ domain ∼50% oxidized (Fig. S2) and Trx1 that was almost fully in the oxidized form (not depicted). We found that addition of a 20-fold molar excess of oxidized Trx1 had no effect on the rate of reoxidation of Ero1pc-C100A-C105A (Fig. 3 A, top), showing that oxidized Trx1 has little or no capability to oxidize the regulatory bonds of Ero1p. In contrast, oxidized Pdi1p dramatically increased the rate of Ero1pc-C100A-C105A reoxidation (Fig. 3 A, top right). The ability of oxidized Pdi1p to efficiently oxidize the regulatory bonds of Ero1p was not a peculiarity of the Ero1pc-C100A-C105A mutant, as we could also show that oxidized Pdi1p accelerated the oxidation of wild-type Ero1pc as well as Ero1pc-C355S, which is unable to oxidize its regulatory bonds (Fig. S3). These results show that Ero1p can be inactivated by direct feedback from Pdi1p.

Next, we tested the converse process, the ability of reduced Pdi1p to reduce and thus activate Ero1p. To assess the activation process selectively, we again used Ero1pc-C100A-C105A because of this mutant’s greatly diminished capacity for autonomous oxidation of regulatory cysteines once reduced with an exogenous agent (Fig. 2 B, middle). Reduced Pdi1p requires 60 min to reduce the Ero1p regulatory bonds completely, whereas reduced Trx1 requires <5 min, indicating that Pdi1p is a relatively poor activator of Ero1p (Fig. 3 B, top). In summary, Trx1 is a potent activator but poor inactivator of Ero1p, whereas Pdi1p is a weak activator but potent inhibitor of Ero1p. These contrasting abilities to reduce or oxidize the regulatory bonds of Ero1p can be accounted for by the much greater reducing potential of Trx1 as compared with Pdi1p. By assaying the rate of activation or inactivation of Ero1p in separate reactions, we could show that reduced Pdi1p activates Ero1p more slowly than oxidized Pdi1p can inactivate Ero1p. This explains why in an in vitro Ero1p oxidation reaction that contains reduced Pdi1p as a substrate, Ero1p rarely, if ever, becomes fully activated, and the reaction proceeds relatively slowly.

The a and a′ domains of Pdi1p differ in their ability to regulate Ero1p

Yeast Pdi1p has two redox-active thioredoxin-like domains with different reduction potentials: the a domain (−195 mV) and a′ domain (−164 mV; Fig. 1 A; Tian et al., 2006; Vitu et al., 2010). We recently showed that Ero1p preferentially oxidizes the a domain both in vivo and in vitro, suggesting Ero1p provides oxidizing equivalents to the ER mainly via the a domain (Vitu et al., 2010). Thus, the a′ domain may play a lesser role in the oxidation of ER substrates despite its higher reduction potential than the a domain. To fully understand the interactions between Pdi1p and Ero1p, we wished to evaluate the contribution of each of the active Pdi1p domains to reduction and oxidation of Ero1p regulatory bonds.

Recombinant Pdi1p purified from E. coli had a fully oxidized a domain and a partially oxidized a′ domain (Fig. S2), consistent with the different reduction potentials of the two domains. Moreover, naturally oxidized purified domain mutants Pdi1p-CCSS (a domain cysteines intact) and Pdi1p-SSCC (a′ domain cysteines intact) had similar oxidation states to the corresponding domains in wild-type Pdi1p (unpublished data). The same assay for Pdi1p oxidation of the Ero1p regulatory bonds was performed with Pdi1p-CCSS or Pdi1p-SSCC mutants. Both Pdi1p mutants accelerated oxidation of the Ero1p regulatory bonds. Despite the less-active disulfides, Pdi1p-SSCC reproducibly accelerated Ero1p oxidation more rapidly than Pdi1p-CCSS (Fig. 3 A), consistent with the higher reduction potential of the a′ CxxC. A Pdi1p-SSSS that lacked both redox-active sites did not accelerate oxidation of Ero1p regulatory bonds to the extent of the other Pdi1p variants, but a modest acceleration of Ero1p oxidation was observed (Fig. 3 A, bottom right). It is possible that the structural (CX6C) disulfide in Pdi1p (Fig. 1 A) has some effect or that Pdi1p may stabilize the oxidized form of Ero1p through a mechanism not involving disulfide exchange with the redox-active motifs.

Next, we compared the ability of reduced Pdi1p-CCSS or Pdi1p-SSCC to reduce the regulatory bonds of Ero1pc-C100A-C105A. Although reduction potentials measured against a glutathione standard suggest that the a domain is a better reductant, both reduced Pdi1p-CCSS and reduced Pdi1p-SSCC reduced Ero1p regulatory bonds at similar rates (Fig. 3 B, bottom). As expected, the single-domain mutants reduced the Ero1p regulatory bonds somewhat less efficiently than did wild-type Pdi1p, which has both active sites intact, and the double domain mutant Pdi1p-SSSS did not reduce the Ero1p regulatory bonds (Fig. 3 B, bottom right).

The redox state of Ero1p can be set by a mix of reduced and oxidized Pdi1p

Reduced and oxidized Pdi1p promote activation and inactivation of Ero1p, respectively, suggesting that the activity of Ero1p in steady state may be set by the ratio of reduced to oxidized Pdi1p. By taking advantage of Ero1pc-C100A-C105A, which lacks the ability to oxidize Pdi1p efficiently, we could test admixtures of oxidized and reduced Pdi1p for their effect on the steady-state redox balance of the regulatory bonds of Ero1pc-C100A-C105A. Although accurate quantitative amounts cannot be obtained as a result of the complications that oxidized Pdi1p contains a partially reduced domain and also nonlinear detection by immunoblot analysis, we found that when the fraction of reduced Pdi1p was >0.8, most Ero1pc-C100A-C105A was fully reduced, whereas when the fraction of reduced Pdi1p was <0.3, most Ero1pc-C100A-C105A was oxidized (Fig. 3 C). In a steady state, in vivo Pdi1p is partially reduced (Vitu et al., 2010) like air-oxidized recombinant Pdi1p, and almost all Ero1p is in a fully oxidized form (see 0 time point of Figs. 1 B and 5 E). Therefore, we could conclude that partially reduced Ero1pc is at least partially active and will continue to oxidize Pdi1p until a ratio of oxidized-to-reduced Pdi1p is achieved that no longer facilitates reduction and activation of Ero1p.

An Ero1p regulatory mutant is rapidly activated by Pdi1p

Previously, we showed that an Ero1p mutant that lacks only one of the three regulatory bonds, Ero1p(c)-C150A-C295A, has high enzymatic activity both in vivo and in vitro (Sevier et al., 2007; Vitu et al., 2010). This suggests that in the absence of the C150-C295 regulatory bond, Ero1p is either more readily activated by reduced Pdi1p or less easily inactivated by reaction with oxidized Pdi1p or by autonomous oxidation.

First, the autonomous oxidation of Ero1p regulatory bond mutants Ero1pc-C150A-C295A, Ero1pc-C90A-349A, and Ero1pc-C143A-166A was evaluated for their ability to undergo autonomous reoxidation after being reduced by Trx1. This comparison is complicated by differing band mobility for the oxidized and partially oxidized states of each of the mutants (Fig. 4 A). As a measure of the oxidation rate that could be consistently applied to each of the mutants, we represented the progression of the oxidation process as the decline in the fraction of Ero1p that was in a fully reduced state. All regulatory mutants showed similar overall autonomous oxidation rates (Fig. 4 B), indicating that formation of the C150-C295 bond does not significantly affect oxidation of other regulatory bonds. Because the pattern of partially oxidized states was more complex for wild type (Fig. 2 B) than for the mutants, it was not possible to make a precise quantitative comparison of rates.

Next, we tested each of the three regulatory bond mutants for their ability to be activated by reduced Pdi1p. This test required addition of the shuttle cysteine mutant to each of the three regulatory bond mutants. Reduction of Ero1pc-C100A-C105A-C150A-C295A by reduced Pdi1p was significantly faster than for Ero1pc-C100A-C105A or for the other regulatory bond mutants (Fig. 4, C and D). Ero1pc-C100A-C105A-C90A-C349A also exhibited slightly faster reduction than Ero1pc-C100A-C105A or Ero1pc-C100A-C105A-C143A-C166A (Fig. 4, C and D). These results suggest that removal of the C150-C295 regulatory bond lowers the threshold for Ero1p activation. Interestingly, both the Ero1pc-C100A-C105A-C150A-C295A and Ero1pc-C100A-C105A-C90A-C349A mutants exhibited some reoxidation of Ero1p (Fig. 4 C) and oxidation of Pdi1p (not depicted) at the 60-min time point, suggesting that Ero1p may have alternative pathways to relay disulfide bonds to Pdi1p when the shuttle disulfide is unavailable (see Discussion).

Pdi1p is the primary in vivo regulator of Ero1p

If Pdi1p regulates Ero1p through reduction and oxidation of its regulatory bonds, it should be possible to trap mixed disulfide complexes between Pdi1p and the regulatory cysteines of Ero1p in vivo. To specifically isolate mixed disulfides with regulatory cysteines, we used FLAG epitope–tagged Ero1p-C100A-C105A (expressed ectopically in cells that carry wild-type ERO1). Ero1p-C100A-C105A is unable to support yeast viability as a result of the disruption of its ability to oxidize Pdi1p in what has been shown to be the primary pathway for electron exchange between Pdi1p and Ero1p in living cells (Frand and Kaiser, 2000). Therefore, any observable mixed disulfides trapped between Ero1p-C100A-C105A and Pdi1p in living cells would likely represent thiol/disulfide exchange between the Ero1p regulatory cysteines and Pdi1p.

Ero1p-C100A-C105A-FLAG, isolated from cysteine-alkylated cell extracts by immunoprecipitation with FLAG antibody, formed a predominant mixed disulfide complex that also reacted with Pdi1p antiserum (Fig. 5 A, lane 2). Analysis of the crude immunoprecipitate by mass spectroscopy identified Pdi1p peptides but failed to identify additional PDI family members, including Eps1p, Eug1p, Mpd1p, or Mpd2p (unpublished data). This result suggests that Pdi1p is the primary oxidoreductase engaged in oxidation or reduction of the Ero1p regulatory bonds. Loss of any single regulatory cysteine pair in Ero1p-C100A-C105A did not prevent trapping of a mixed disulfide bond between Ero1p-C100A-C105A and Pdi1p, indicating that Pdi1p interaction with Ero1p-C100A-C105A is not limited to a single regulatory cysteine pair (Fig. 5 A). The mixed disulfides between Pdi1p and the various Ero1p-C100A-C105A regulatory bond mutants showed distinct mobilities on SDS-PAGE (Fig. 5 A), which likely corresponded to different variable combinations of cysteines in Ero1p and Pdi1p that engaged in oxidation and reduction of Ero1p regulatory bonds.

If reduced Pdi1p were essential for activating Ero1p, depletion of Pdi1p from the ER would be expected to decrease activation of Ero1p by reduction of regulatory bonds even under prevailing reducing conditions in the ER. We have established conditions to deplete a cell of Pdi1p, which is essential for viability, by glucose repression for 15 h of PGAL1-PDI1 in a pdi1Δ genetic background (Fig. 5 B; Tachibana and Stevens, 1992; Frand and Kaiser, 1999). As expected, Pdi1p depletion disrupts folding in the ER, as shown by the accumulation of the ER form of carboxypeptidase Y (CPY), similar to that in cells that had been treated with DTT (Fig. 5 C). Radiolabeling of cells to specifically follow CPY synthesized after Pdi1p depletion confirmed a complete block of biosynthetic disulfide bond formation in newly synthesized CPY (Fig. S4). However, under the same conditions of Pdi1p depletion, most Ero1p remains oxidized (Fig. 5 D), indicating that the signal for an insufficiently oxidized ER cannot be transduced to activate Ero1p in the absence of Pdi1p. Pdi1p-depleted cells showed very slow activation of Ero1p even when they were treated with the reducing agent DTT, a condition that leads to rapid activation of Ero1p in Pdi1p-expressing cells (Fig. 5 E).

In the Pdi1p-depleted cells, a small amount of activated (reduced) Ero1p was detected (Fig. 5 D), and catalytically inactive Ero1 mutants Ero1p-C355S and Ero1p-C100A-C105A exhibited significantly more of the reduced forms (Fig. S5), suggesting that under conditions of Pdi1p depletion, although Ero1p can be reduced to some extent, most of the Ero1p remains in an oxidized state if autonomous reoxidation is possible. We tested the possibility that the residual reduction of Ero1p in cells that had been depleted for Pdi1p was catalyzed by another PDI homolog. By conducting the depletion experiment in a strain depleted for the other homologs (mpd1Δ, mpd2Δ, eug1Δ, eps1Δ, pdi1Δ, and CEN-PGAL1-PDI1), we found that reduced Ero1p formed to at least the same extent as in a pdi1Δ PGAL1-PDI1 strain (Figs. 5 D and S5), indicating that no other PDI family members contribute significantly to reduction of Ero1p regulatory bonds.

Glutathione acts synergistically with Pdi1p to activate Ero1p

The rapid Pdi1p-dependent reduction of Ero1p regulatory bonds that occurs in cells treated with DTT could be the result of a pool of reduced Pdi1p formed by direct reduction with DTT that had entered the ER. Alternatively, DTT treatment may produce a pool of reduced glutathione (GSH) in the ER, which in turn could lead to reduction of Pdi1p. To test the possibility that DTT acts indirectly through glutathione, we tested the effect of depleting cells of glutathione by treating cells with 5 mM buthionine sulfoximine (BSO) for 15 h, which blocks glutathione synthesis by inhibiting γ-glutamylcysteine synthetase. Reduction of Ero1p regulatory bonds by DTT treatment occurred more slowly in glutathione-depleted cells than in untreated cells (Fig. 5 E). This result suggests that the presence of a reducible pool of glutathione may accelerate the overall process of reduction of Pdi1p by DTT. Alternatively, depletion of glutathione by BSO treatment may subtly affect Pdi1p expression or have some other indirect effect on the redox state of the ER.

To address the role of glutathione in Ero1p activation more directly, we reconstituted Ero1p activation in vitro using both Pdi1p and GSH as substrates. By itself, Pdi1p is a poor activator of Ero1p because once Pdi1p has become oxidized through Ero1p oxidase activity, there is no longer sufficient reduced Pdi1p available to keep Ero1p in an active state, and oxidized Pdi1p can accelerate inactivation of Ero1p. Therefore, fully activated Ero1p was rarely observed during reaction with reduced Pdi1p (Fig. 6 A). Glutathione alone was also a poor activator of Ero1p. 10 mM GSH did not reduce regulatory bonds in either wild-type Ero1pc (Fig. 6 B, left) or Ero1pc-C100A-C105A (not depicted). However, a mixture of 10 mM GSH and Pdi1p maintained a pool of at least partially activated Ero1p (Fig. 6 B, right). The basis of Ero1p activation in the presence of Pdi1p and GSH appeared to be that provided by a pool of GSH in great stoichiometric excess to Pdi1p, which allowed for a significant pool of reduced Pdi1p to be maintained. This pool of reduced Pdi1p in turn could reduce the regulatory bonds of Ero1p, thus overcoming the kinetic barrier for GSH to reduce the Ero1p regulatory bonds. In this coupled reaction, oxidized glutathione (GSSG) was generated by the combined action of Pdi1p and Ero1p (Fig. 6 C), and production of GSSG depended on Ero1p enzymatic activity because even in the presence of Pdi1p, catalytically inactive Ero1pc-C100A-C105A did not produce GSSG actively, whereas hyperactive Ero1pc-C150A-C295A produced GSSG more rapidly than wild type (not depicted).

To investigate the effects of the balance of GSH to GSSG on Ero1p activity, we reconstituted an in vitro reaction in which recombinant Ero1p and Pdi1p were incubated in the presence of three different ratios of GSH/GSSG. When the reaction was begun in the presence of GSH only, the rate of glutathione oxidation was initially rapid but then slowed. Evidently, the concomitant generation of GSSG by the action of Ero1p and Pdi1p eventually leads to inactivation of Ero1p (Fig. 6 D, 5 mM GSH curve). Introduction of GSSG into the coupled reaction slowed the initial rate of GSH oxidation (compare the curves of 5 mM GSH with 8:1 and 3:1 GSH/GSSG; Fig. 6 D). Taking the background of slow nonenzymatic generation of GSSG into account, the Ero1p and Pdi1p catalyzed oxidation of GSH stops at when the molar ratio of GSH/GSSG is ∼3:1. Thus, this ratio of GSH/GSSG appears to be the approximate steady-state set point for the coupled reaction.

Discussion

A conserved protein disulfide relay system of the ER, consisting of Ero1 and PDI, supports correct disulfide bond formation of secretory proteins (Sevier et al., 2001; Sevier and Kaiser, 2006a; Appenzeller-Herzog et al., 2008; Braakman, 2009). In S. cerevisiae, the Ero1-PDI relay is essential for biosynthetic disulfide bond formation and is thus essential for cell viability (Frand and Kaiser, 1998; Pollard et al., 1998; Vitu et al., 2010). We previously showed that the activity of S. cerevisiae Ero1p is controlled by an autoregulatory feedback loop that involves the reduction and the reoxidation of three regulatory disulfide bonds (Sevier et al., 2007). Here, we show that not only is Pdi1p the major substrate of Ero1p in the disulfide relay system but that Pdi1p is also the primary regulator of Ero1p activity by its ability carry out dithiol/disulfide exchange with the regulatory bonds. We also show that the redox state of glutathione in the ER is important for setting the regulatory state of Pdi1p, and, in this way, the disulfide bond–generating activity of Ero1p is controlled by the redox status of both Pdi1p and of glutathione.

In a reaction initiated by adding to Ero1p a stoichiometric excess of reduced Trx1, a model substrate, the regulatory bonds in Ero1p are rapidly reduced, and Trx1 is oxidized by active Ero1p. Once Trx1 has been oxidized, the regulatory bonds in Ero1p are reformed, and Ero1p returns to an inactive state. In this paper, we separated the activation and inactivation phases of this cycle to determine the biochemical requirements for each phase. As expected, the reduced form of Trx1 readily reduces regulatory bonds in Ero1p and is thus a potent activator of Ero1p. However, oxidized Trx1 has no capacity to reoxidize the regulatory cysteines of Ero1p. Instead, oxidation of the regulatory cysteines is accomplished by Ero1p itself in a reaction that probably involves the transfer of a disulfide bond from the shuttle cysteines of one Ero1p protein molecule to the regulatory cysteines of another. Thus, the inactivation phase occurs by autonomous oxidation of Ero1p once reduced Trx1 has been depleted by being converted to the oxidized form.

A similar analysis for a reaction that contains Ero1p and its natural substrate Pdi1p revealed that Ero1p is never fully activated in the presence of Pdi1p (Fig. 6 A). Whereas we initially thought that the inability of Pdi1p to activate Ero1p robustly was a consequence of some deficiency in the in vitro reaction, we now have evidence that the balance of the reduced and oxidized forms of Pdi1p is crucial in regulating Ero1p enzyme activity and faithfully reproduces the redox homeostasis set point of the ER.

Pdi1p has two redox-active domains containing the CxxC motifs. The a′ domain is not as efficiently oxidized by Ero1p as the a domain (Vitu et al., 2010). However, the a′ domain does appear to play a greater role than the a domain in oxidizing the Ero1p regulatory bonds; the partially oxidized a′ domain (Pdi1p-SSCC) accelerates oxidation of Ero1p regulatory bonds more rapidly than the fully oxidized a domain (Pdi1p-CCSS). Given the relative redox potentials of the two domains, their propensity to be oxidized by Ero1p and their relative ability to oxidize the regulatory bonds of Ero1p provide a consistent explanation for why the a′ domain is only partly oxidized, whereas the a domain is fully oxidized in cells under standard conditions (Vitu et al., 2010). In contrast to the yeast Pdi1p, the two active domains of the human PDI have similar reduction potentials, and the a′ domain is the one preferentially oxidized by Ero1-α (Wang et al., 2009; Chambers et al., 2010). Therefore, it seems possible that the mechanism of Ero1-α regulation via the two domains of the human PDI is distinct from the one in yeast. Specificity in interaction between human and yeast Ero1 and PDI proteins has been demonstrated (Vitu et al., 2010; Masui et al., 2011). Comparison of the activity of human and yeast proteins in vitro shows that the highest enzyme activity is achieved using protein combinations of Ero1 and PDI from the same species (Araki and Nagata, 2011). Low catalytic activity was traced in part to a weaker binding affinity between human Ero1-α and yeast Pdi1p, relative to human PDI (Araki and Nagata, 2011). Whether the different redox potentials of the PDI domains in human and yeast PDI also contribute to the differences in Ero1 regulation between species remains to be assessed.

Several lines of evidence point to the existence of multiple disulfide relay pathways from Ero1p to Pdi1p. Best characterized is the flow of disulfides from the C352-C355 pair to Pdi1p via the C100-C105 shuttle cysteines. This pathway is necessary for full activity of Ero1pc in vitro and is necessary for cell viability. Nevertheless, Ero1pc-C100A-C105A retains some activity in vitro and shows slow autonomous reoxidation, showing that a disulfide bond can be transferred, albeit slowly, from the active site to the regulatory disulfides without the action of the C100-C105 shuttle (Fig. 2 B). Moreover, in vivo experiments have shown that yeast cells with an ero1-C100A mutation are viable if they have the capacity to up-regulate Ero1p expression through unfolded protein response induction (Frand and Kaiser, 2000), and overexpression of Ero1p-C100A-C105A-C150A-C295A was able to partially suppress the temperature sensitivity of the ero1-1 strain (unpublished data).

The pathway of dithiol/disulfide exchange reactions used by Ero1p-C100A-C105A to oxidize Pdi1p remains to be determined. Considering that autonomous oxidation of the regulatory cysteines by Ero1p-C100A-C105A is possible, we hypothesize that the active-site disulfide bond C352-355 may first be transferred to regulatory cysteine pairs, and then Pdi1p could be oxidized by reducing these regulatory bonds during the activation process of Ero1p. These pathways were revealed by experiments with Ero1pc-C100A-C105A but are not limited to mutant Ero1p and are likely also used by wild-type Ero1p. By transferring one or two of the regulatory bonds to Pdi1p, Ero1p would have some capacity to oxidize a stoichiometric quantity of Pdi1p without any Ero1p becoming fully active. The C143-C166 bond may be the most labile regulatory bond, as its reduction is the first step of Ero1 activation by Pdi1p in vitro (Heldman et al., 2010) and in turn could readily contribute to Pdi1p oxidation. This pathway may allow the disulfide relay system to respond to modest increases in load with a minimum of excess hydrogen peroxide generation.

When Ero1p is incubated with reduced Pdi1p and GSH, Ero1p can be shown to be catalytically active by its ability to oxidize a stoichiometric excess of GSH, although little or no Ero1p can be detected in a state in which all regulatory cysteines are in a reduced state (Fig. 6 B). As previously noted (Heldman et al., 2010), this observation indicates that the partially reduced forms of Ero1p have at least some oxidase activity. Thus, the regulatory bonds in Ero1p may act less like an on/off switch than a proportional controller with different degrees of activity according to the redox states of the three regulatory bonds. We have not been able to devise a quantitative test of this attractive hypothesis because it is not possible to isolate in pure form Ero1p in a partially reduced state.

Toward a biochemical understanding of the mechanism of Ero1p regulation in a highly simplified reaction, we successfully reconstituted the disulfide relay system and the basic properties of redox homeostasis in vitro with just their components, Ero1pc, Pdi1p, and glutathione (Fig. 6). The capacity of Pdi1p to sustain Ero1p in an activated state can be increased by addition of an excess of GSH in the in vitro reaction. This coupled reaction appears to approximate conditions in the ER (Hwang et al., 1992), in which transfer of disulfide bonds to nascent secretory proteins and to GSH keeps enough Pdi1p in a reduced state to allow for some capacity to activate Ero1p. In the coupled reaction, Ero1p enzyme activity is controlled by the ratio of GSH/GSSG, and a 3:1 ratio of GSH/GSSG appears to be close to the steady-state set point at which Ero1p can be activated. Pdi1p, which has both reductase and isomerase activities, can serve as the key link between thiol substrates, including glutathione, and Ero1p to activate the disulfide relay system only when substrates are present (Fig. 7). In the mammalian cell, it has also been suggested that oxidation of PDI and glutathione is well regulated by an equilibrium between Ero1-α and PDI (Appenzeller-Herzog et al., 2010). It seems likely that ER redox homeostasis is maintained by the same basic mechanism in yeast and mammalian cells.

Materials and methods

Protein expression and purification

The sequence coding Ero1pc (56–424) or Ero1pc mutants (Ero1pc-C100A-C105A, Ero1pc-C355S, Ero1pc-C150A-C295A, Ero1pc-C90A-C349A, Ero1pc-C143A-C166A, and Ero1pc-C100A-C105A-C143A-C166A) was fused at the C terminus to His6-tagged MBP linked by the Tobacco etch virus (TEV) protease cleavage sequence (ENLYFQGS) inserted into the pET17b vector (EMD) between NdeI and XhoI restriction sites. We found that C355S was the only point mutation of the active-site cysteine pair (C352-C355) that could be produced in a soluble FAD-bound state. Ero1pc and all variants were overexpressed in Origami B(DE3) competent cells (EMD) by induction with 0.4 mM IPTG supplemented with 10 µM FAD overnight at room temperature. Ero1pc fusion proteins with His6-tagged MBP were purified from cell lysates by affinity to amylose (New England Biolabs, Inc.) and then dialyzed into TEV cleavage buffer (50 mM Tris, pH 7.8, 0.5 mM EDTA, and 0.5 mM tris(2-carboxyethyl)phosphine). MBP-fused Ero1pc (MBP-Ero1pc) and Ero1pc-C100A-C105A were also expressed and purified as described for Ero1pc purification. Ero1pc was cleaved from His6-tagged MBP by incubating with recombinant His6-tagged TEV protease (1:100 wt/wt protease to substrate; the TEV protease–encoding plasmid was provided by C. Hill, University of Utah, Salt Lake City, UT) for 15 h at 25°C. Ero1pc and variants cleaved from His6-MBP were collected by flowing through the HiTrap Ni2+ column (GE Healthcare) to remove uncleaved fusion protein, cleaved His6-tagged MBP, and TEV protease. Some Ero1pc variants (Ero1pc-C100A-C105A-C150A-C295A and Ero1pc-C100A-C105A-C90A-C349A) were subcloned in pGEX-4T vector purified by flowing though GSTrap (GE Heathcare) followed by thrombin cleavage for elution and secondary purification with the HiTrap Ni2+ column for removal of thrombin (Gross et al., 2004). All purified Ero1pc proteins were dialyzed against PBS.

The sequence coding Pdi1p (31–522) or Pdi1p variants (C406S-C409S for CCSS, C61S-C64S for SSCC, and C61S-C64S-C406S-C409S for SSSS) tagged with His6 at the N terminus was inserted into pET20b between NdeI and SacI restriction sites. Recombinant His6-tagged Pdi1p, Pdi1p variants, and His6-tagged E. coli thioredoxin (Trx1; Sevier and Kaiser, 2006b) were overexpressed in BL21(DE3)pLysS competent cells (EMD) by induction with 0.4 mM IPTG for 5 h at 37°C and purified from cell lysates by affinity to the HiTrap Ni2+ column. Purified, naturally oxidized Pdi1p and Trx1 were dialyzed into PBS. Reduced forms of Pdi1p and Trx1 were prepared by incubating with 10 mM DTT for 30 min at room temperature and removing DTT with Micro Bio-Spin 6 (Bio-Rad Laboratories).

Yeast strains and plasmids

CKY1044 and CKY1045 and MATa and MATα pdi1Δ::KanMX strains covered by URA3 CEN PDI1 plasmid pCS213 have been previously described (Vitu et al., 2010). To facilitate scoring of crosses with the pdi1Δ strain, a NatMX-marked pdi1Δ strain was constructed by swapping the KanMX marker in CKY1045 with a NatMX4 cassette by homologous recombination (Goldstein and McCusker, 1999). Individual disruptions of the genes encoding the yeast PDI homolog proteins were made by one-step gene replacement of the entire ORF with KanMX by homologous recombination. The compound deletion strain CKY1086 with the genotype MATa GAL2 ura3-52 leu2-3,112 pdi1Δ::NatMX mpd1Δ::KanMX mpd2Δ::KanMX eug1Δ::KanMX eps1Δ::KanMX [pCS213] was constructed by sequential rounds of mating, sporulation, and scoring of the initial individual disruption mutant strains. Disruption of all five genes encoding the PDI1 homologs in CKY1086 was confirmed by PCR analysis. Some pdi1Δ strains also contained a genomic-tagged copy of BGL2. Tagging of BGL2 with a triple HA epitope at its C terminus was performed by homologous recombination with a 3HA-KanMX6 module (Longtine et al., 1998). To construct the MATa GAL2 ura3-52 leu2-3,112 pdi1Δ::NatMX BLG2:HA-KanMX [pCS213] double mutant (CKY1087), individual pdi1Δ and BLG2:HA strains were mated, and an MATa NatMX+ KanMX+ segregant was selected after sporulation. Viability of CKY1086 and CKY1087 depends on a plasmid-borne PDI1 gene provided by plasmid pCS213. CKY1088 and CKY1089 were generated by transformation of CKY1086 or CKY1087, respectively, with the pSK65 [CEN LEU2 PGAL1-PDI1] followed by selection against pCS213 by plating on supplemented minimal medium (SMM) with 5-FOA. CKY1090 was created by transformation of CKY1044 with pSK2 [CEN LEU2 His6-PDI1] followed by selection against pCS213 with 5-FOA.

To create pSK2 [CEN LEU2 His6-PDI1], the sequence coding HHHHHGG was inserted right after the signal sequence (1–22) by introducing the NheI restriction site in pCS463, which contains the PDI1 coding region and ∼875 and ∼150 bp of the 5′ and 3′ untranslated regions. Plasmid pSK65 [CEN LEU2 PGAL1-PDI1] contains the PDI1 coding region and ∼60 and ∼150 bp of the 5′ and 3′ untranslated regions under the GAL1 promoter. ERO1-coding URA3 2µ plasmids pSK63 [ERO1-flag], pSK64 [ero1-C100A-C105A-flag], pDPS51 [ero1-C100A-C105A-C150A-C295A-flag], pDPS49 [ero1-C100A-C105A-C90A-C340A], pDPS50 [ero1-C100A-C105A-C143A-C166A], and pSK66 [ero1-C355S-flag] were generated by insertion of sequence-encoding flag tag right after the last codon using the NotI restriction site and subcloning or site-directed mutagenesis based on the previously reported plasmid pAF84 (Frand and Kaiser, 1998).

Analysis of protein oxidation states

For reduction of Ero1p regulatory bonds in vitro, 1 µM Ero1pc or its variants were incubated with 50 or 20 µM reduced Trx1, reduced Pdi1p, or reduced Pdi1p mutants in PBS containing 0.5 mM EDTA. After incubation at 25°C, the reactions were quenched by mixing with 1× SDS-AMS (4-acetamido-4′-maleimidylstilbene-2,2′-disulfonic acid) buffer (1× SDS sample buffer without reducing agent and 1 mM AMS [Invitrogen]), which alkylates free cysteine thiols and adds one ∼500-D moiety to each free thiol. For reactions that contained glutathione or DTT, proteins were first precipitated by 10% chilled TCA. Protein precipitants were washed with 0°C acetone and were redissolved with 1× SDS-AMS buffer. Ero1p and Pdi1p were analyzed by nonreducing SDS-PAGE followed by immunoblotting with anti-Ero1p serum or anti-HDEL (Santa Cruz Biotechnology, Inc.). The oxidation state of Trx1 was analyzed by nonreducing SDS-PAGE followed by staining with SimplyBlue SafeStain (Invitrogen).

The in vivo oxidation state of Ero1p was assessed in cells growing in SMM medium incubated with 0, 0.2, or 1 mM DTT for 30 min and then lysed in the presence of chilled 10% TCA by agitating with glass beads. Protein pellets from cell lysates were washed with chilled acetone and redissolved with 1× SDS-AMS buffer. Oxidation states of Ero1p and variants were analyzed by nonreducing SDS-PAGE and immunoblotting with anti-Ero1p. Immunoblots were visualized with an Odyssey imaging system (LI-COR Biosciences). The fully reduced or the oxidized band of Ero1pc was quantified by the Odyssey program (v2.0; LI-COR Biosciences), and reduced fraction was calculated by dividing reduced band intensity with the sum of reduced and oxidized band intensity.

To determine the effect of hydrogen peroxide on the oxidation state of Ero1p regulatory bonds (Fig. S1), 1 µM Ero1pc was incubated with 100 µM reduced His6-tagged Trx1 in PBS containing 0.5 mM EDTA for 2 min, and Trx1 was selectively removed by affinity for Ni2+ beads. To remove hydrogen peroxide generated during action of Ero1p oxidase, 80 µg/ml catalase was added to the reaction. Removal of hydrogen peroxide by catalase was confirmed by measuring the amount of hydrogen peroxide in the reaction with Amplex UltraRed reagent (Invitrogen). 1 µM Ero1pc was incubated with 100 µM reduced thioredoxin, 50 µM Amplex red, and 0.1 U/ml HRP in the presence or absence of 80 µg/ml catalase for 30 min at room temperature. Fluorescence was measured in a microplate reader with wavelength settings of 530/590 nm. Inactivation of Ero1p under anaerobic conditions was analyzed as described above except being performed in the anaerobic chamber flowing N2 and CO2 gases. To test whether exogenous hydrogen peroxide or oxidized Pdi1p can promote inactivation of Ero1p, 100 µM hydrogen peroxide or 20 µM oxidized Pdi1p was added to Ero1pc, Ero1pc-C100A-C105A, or Ero1pc-C355S after Trx1 depletion. Oxidation states of Ero1pc and its mutants were analyzed after cysteine modification with 1 mM AMS by nonreducing SDS-PAGE and immunoblotting with anti-Ero1p serum.

The oxidation states of each catalytic domain of recombinant Pdi1p were determined by introducing a thrombin cleavage site that allows separation of the two catalytic domains (Fig. S2). The sequence coding N-terminal His6-tagged Pdi1p (31–522) with a thrombin recognition sequence in the X-linker was subcloned from pSK42 (Vitu et al., 2010) for E. coli expression. His6-Pdi1p-thrombin was overexpressed in BL21(DE3)pLysS competent cells by IPTG and purified from cell lysates with the HiTrap Ni2+ column. Purified His6-Pdi1p-thrombin was dialyzed into PBS. 10 µg His6-Pdi1p-thrombin was incubated with 1 mM AMS for 30 min at room temperature followed by treatment with 0.1 µg thrombin supplemented with 1 mM CaCl2 for 1 h at 37°C. Fully reduced Pdi1p control was prepared by treatment with 10 mM DTT for 30 min at room temperature. DTT was removed by using Bio-Spin 6. AMS-modified and thrombin-digested samples were analyzed by nonreducing SDS-PAGE.

The oxidation state of newly synthesized CPY in the Pdi1p-depleted ER was analyzed in pdi1Δ cells with galactose-inducible PDI1 grown to the exponential phase for 15 h in glucose medium to deplete Pdi1p. Cells were suspended in SMM-met, pulse labeled for 10 min at 30°C, and lysed in the presence of 10% chilled TCA to block disulfide exchange. After AMS modification, CPY was immunoprecipitated and resolved by reducing SDS-PAGE. A fully reduced CPY control was prepared by treating cells with 10 mM DTT for 15 min before cell lysis.

Reoxidation of Ero1pc after thioredoxin depletion

1 µM Ero1pc was reduced by incubation with 100 µM reduced His6-tagged Trx1 for 2 min and then passed through Ni2+ beads to remove Trx1. At times after Trx1 depletion, the reaction was quenched by mixing with SDS sample buffer containing 1 mM AMS. Depletion of Trx1 was confirmed by analysis with nonreducing SDS-PAGE. When necessary, 40 µM oxidized Trx1, 20 µM oxidized Pdi1p, or 20 µM oxidized Pdi1p mutant was added into Ero1pc or Ero1pc mutants immediately after Trx1 depletion.

Purification of mixed disulfides from yeast

CKY1090 expressing His6-Pdi1p was transformed with pSK64 [ero1-C100A-C105A-flag], pDPS51 [ero1-C100A-C105A-C150A-C295A-flag], pDPS49 [ero1-C100A-C105A-C90A-C340A], and pDPS50 [ero1-C100A-C105A-C143A-C166A]. Exponentially grown cells were lysed by agitation with glass beads in the presence of chilled 10% TCA. TCA-precipitated protein pellets were washed with acetone and redissolved with 2% SDS and 50 mM Tris, pH 8.0, containing 50 mM NEM to block free cysteines. For efficient affinity purification, SDS was removed by SDS-Out reagent (Thermo Fisher Scientific). After clarification by centrifugation, Ero1p-C100A-C105A-flag was released from anti-flag affinity beads using 100 µg/ml flag peptide (Sigma-Aldrich) followed by deglycosylation with PNGaseF for 3 h at 37°C. Samples purified by affinity to anti-flag beads were passed through an Ni2+ column to capture mixed disulfides with His6-Pdi1p. Both Ni2+ unbound fraction and bound fraction were collected and analyzed by nonreducing SDS-PAGE and immunoblotting with anti-flag (Sigma-Aldrich) or anti-Pdi1p.

Depletion of Pdi1p and glutathione in yeast

To determine the oxidation states of Ero1p and mutants in Pdi1p- or glutathione-depleted cells, pSK63, pSK64, and pSK66 were transformed into CKY1088 and/or CKY1089. Cells exponentially grown in SMM containing 1.5% galactose and 1% raffinose were incubated in SMM containing 2% glucose for 15 h to shut off expression of Pdi1p. Glutathione was depleted by growing cells in the presence of 5 mM BSO, an inhibitor of glutathione biogenesis, for 15 h. Pdi1p depletion was confirmed by immunoblotting with anti-Pdi1p (Fig. 5 E), and glutathione depletion was confirmed using Ellman’s reagent.

The maturation of CPY was evaluated in pdi1Δ cells with galactose-inducible PDI1 grown to exponential phase for 15 h in glucose medium to deplete Pdi1p. For fully reduced cell environment, cells were incubated with 1 mM DTT for 1 h after a 14-h incubation in glucose medium. Cell lysates was prepared in the presence of chilled 10% TCA, and mature and immature CPY were analyzed by immunoblotting with anti-CPY serum.

Measurement of oxidized glutathione

GSSG produced by Ero1pc and Pdi1p was measured by using Ellman’s reagent (5,5′-dithiobis-(2-nitrobenzoic acid) [DTNB]). 1 µM Ero1pc, Ero1pc-C150A-C295A, or Ero1pc-C355S was incubated with 10 mM GSH with or without 10 µM oxidized Pdi1p. For extended reaction, 5 mM GSH, 4 mM GSH/0.5 mM GSSG, or 3 mM GSH/1 mM GSSG was mixed with Ero1p and Pdi1p. At various time points after mixing, 10 µl of reactants was quenched by chilled 5-sulfosalicylic acid (1% final concentration), and precipitated protein pellet was removed. GSH was alkylated by 2% 2-vinylpyridine after neutralizing sulfosalicylic acid with a drop of triethanolamine (Sigma-Aldrich) for 1 h at 25°C followed by a 20-fold dilution in DTNB reaction buffer containing 100 mM sodium phosphate, pH 7.0, 1 mM EDTA, 0.2 mM DTNB, and 0.2 mM NADPH. The change in absorbance at 405 nm was recorded after adding 1 U of glutathione reductase.

Online supplemental material

Fig. S1 shows that autonomous oxidation and inactivation of Ero1p is not a side effect of hydrogen peroxide produced by aerobic catalysis by Ero1p. Fig. S2 shows that the endpoint of air oxidation yields a form of recombinant Pdi1p with the a domain almost fully oxidized, and the a′ domain is partially oxidized. Fig. S3 shows that air-oxidized Pdi1p can contribute to the oxidation of the Ero1p regulatory bonds. Fig. S4 shows that when cells are depleted of Pdi1p by growing pdi1Δ cells with glucose-repressible PDI1 in glucose medium, newly synthesized CPY remains largely in a reduced state. Fig. S5 shows that when cells are depleted of Pdi1p, autonomous oxidation of the regulatory bonds of Ero1p can occur. This oxidation requires active-site cysteine C355 but does not require the shuttle cysteines C100 and C105.

Acknowledgments

The authors thank Debbie Fass for thoughtful suggestions.

This work was supported by a grant from the National Institutes of Health (GM46941) to C.A. Kaiser.

References

References
Appenzeller-Herzog
C.
,
Riemer
J.
,
Christensen
B.
,
Sørensen
E.S.
,
Ellgaard
L.
.
2008
.
A novel disulphide switch mechanism in Ero1alpha balances ER oxidation in human cells
.
EMBO J.
27
:
2977
2987
.
Appenzeller-Herzog
C.
,
Riemer
J.
,
Zito
E.
,
Chin
K.T.
,
Ron
D.
,
Spiess
M.
,
Ellgaard
L.
.
2010
.
Disulphide production by Ero1α-PDI relay is rapid and effectively regulated
.
EMBO J.
29
:
3318
3329
.
Araki
K.
,
Nagata
K.
.
2011
.
Functional in vitro analysis of the ERO1 protein and protein-disulfide isomerase pathway
.
J. Biol. Chem.
286
:
32705
32712
.
Baker
K.M.
,
Chakravarthi
S.
,
Langton
K.P.
,
Sheppard
A.M.
,
Lu
H.
,
Bulleid
N.J.
.
2008
.
Low reduction potential of Ero1alpha regulatory disulphides ensures tight control of substrate oxidation
.
EMBO J.
27
:
2988
2997
.
Braakman
I.
2009
.
Entering a new era with Ero
.
Nat. Rev. Mol. Cell Biol.
10
:
503
.
Chambers
J.E.
,
Tavender
T.J.
,
Oka
O.B.
,
Warwood
S.
,
Knight
D.
,
Bulleid
N.J.
.
2010
.
The reduction potential of the active site disulfides of human protein disulfide isomerase limits oxidation of the enzyme by Ero1α
.
J. Biol. Chem.
285
:
29200
29207
.
Cuozzo
J.W.
,
Kaiser
C.A.
.
1999
.
Competition between glutathione and protein thiols for disulphide-bond formation
.
Nat. Cell Biol.
1
:
130
135
.
Frand
A.R.
,
Kaiser
C.A.
.
1998
.
The ERO1 gene of yeast is required for oxidation of protein dithiols in the endoplasmic reticulum
.
Mol. Cell.
1
:
161
170
.
Frand
A.R.
,
Kaiser
C.A.
.
1999
.
Ero1p oxidizes protein disulfide isomerase in a pathway for disulfide bond formation in the endoplasmic reticulum
.
Mol. Cell.
4
:
469
477
.
Frand
A.R.
,
Kaiser
C.A.
.
2000
.
Two pairs of conserved cysteines are required for the oxidative activity of Ero1p in protein disulfide bond formation in the endoplasmic reticulum
.
Mol. Biol. Cell.
11
:
2833
2843
.
Goldstein
A.L.
,
McCusker
J.H.
.
1999
.
Three new dominant drug resistance cassettes for gene disruption in Saccharomyces cerevisiae
.
Yeast.
15
:
1541
1553
.
Gross
E.
,
Kastner
D.B.
,
Kaiser
C.A.
,
Fass
D.
.
2004
.
Structure of Ero1p, source of disulfide bonds for oxidative protein folding in the cell
.
Cell.
117
:
601
610
.
Gross
E.
,
Sevier
C.S.
,
Heldman
N.
,
Vitu
E.
,
Bentzur
M.
,
Kaiser
C.A.
,
Thorpe
C.
,
Fass
D.
.
2006
.
Generating disulfides enzymatically: Reaction products and electron acceptors of the endoplasmic reticulum thiol oxidase Ero1p
.
Proc. Natl. Acad. Sci. USA.
103
:
299
304
.
Heldman
N.
,
Vonshak
O.
,
Sevier
C.S.
,
Vitu
E.
,
Mehlman
T.
,
Fass
D.
.
2010
.
Steps in reductive activation of the disulfide-generating enzyme Ero1p
.
Protein Sci.
19
:
1863
1876
.
Hwang
C.
,
Sinskey
A.J.
,
Lodish
H.F.
.
1992
.
Oxidized redox state of glutathione in the endoplasmic reticulum
.
Science.
257
:
1496
1502
.
Inaba
K.
,
Masui
S.
,
Iida
H.
,
Vavassori
S.
,
Sitia
R.
,
Suzuki
M.
.
2010
.
Crystal structures of human Ero1α reveal the mechanisms of regulated and targeted oxidation of PDI
.
EMBO J.
29
:
3330
3343
.
Longtine
M.S.
,
McKenzie
A.
III
,
Demarini
D.J.
,
Shah
N.G.
,
Wach
A.
,
Brachat
A.
,
Philippsen
P.
,
Pringle
J.R.
.
1998
.
Additional modules for versatile and economical PCR-based gene deletion and modification in Saccharomyces cerevisiae
.
Yeast.
14
:
953
961
.
Masui
S.
,
Vavassori
S.
,
Fagioli
Cl.
,
Sitia
R.
,
Inaba
K.
.
2011
.
Molecular bases of cyclic and specific disulfide interchange between human ERO1α protein and protein-disulfide isomerase (PDI)
.
J. Biol. Chem.
286
:
16261
16271
.
Mezghrani
A.
,
Fassio
A.
,
Benham
A.
,
Simmen
T.
,
Braakman
I.
,
Sitia
R.
.
2001
.
Manipulation of oxidative protein folding and PDI redox state in mammalian cells
.
EMBO J.
20
:
6288
6296
.
Nørgaard
P.
,
Westphal
V.
,
Tachibana
C.
,
Alsøe
L.
,
Holst
B.
,
Winther
J.R.
.
2001
.
Functional differences in yeast protein disulfide isomerases
.
J. Cell Biol.
152
:
553
562
.
Pollard
M.G.
,
Travers
K.J.
,
Weissman
J.S.
.
1998
.
Ero1p: A novel and ubiquitous protein with an essential role in oxidative protein folding in the endoplasmic reticulum
.
Mol. Cell.
1
:
171
182
.
Sevier
C.S.
,
Kaiser
C.A.
.
2002
.
Formation and transfer of disulphide bonds in living cells
.
Nat. Rev. Mol. Cell Biol.
3
:
836
847
.
Sevier
C.S.
,
Kaiser
C.A.
.
2006a
.
Conservation and diversity of the cellular disulfide bond formation pathways
.
Antioxid. Redox Signal.
8
:
797
811
.
Sevier
C.S.
,
Kaiser
C.A.
.
2006b
.
Disulfide transfer between two conserved cysteine pairs imparts selectivity to protein oxidation by Ero1
.
Mol. Biol. Cell.
17
:
2256
2266
.
Sevier
C.S.
,
Cuozzo
J.W.
,
Vala
A.
,
Aslund
F.
,
Kaiser
C.A.
.
2001
.
A flavoprotein oxidase defines a new endoplasmic reticulum pathway for biosynthetic disulphide bond formation
.
Nat. Cell Biol.
3
:
874
882
.
Sevier
C.S.
,
Qu
H.
,
Heldman
N.
,
Gross
E.
,
Fass
D.
,
Kaiser
C.A.
.
2007
.
Modulation of cellular disulfide-bond formation and the ER redox environment by feedback regulation of Ero1
.
Cell.
129
:
333
344
.
Tachibana
C.
,
Stevens
T.H.
.
1992
.
The yeast EUG1 gene encodes an endoplasmic reticulum protein that is functionally related to protein disulfide isomerase
.
Mol. Cell. Biol.
12
:
4601
4611
.
Tavender
T.J.
,
Sheppard
A.M.
,
Bulleid
N.J.
.
2008
.
Peroxiredoxin IV is an endoplasmic reticulum-localized enzyme forming oligomeric complexes in human cells
.
Biochem. J.
411
:
191
199
.
Tavender
T.J.
,
Springate
J.J.
,
Bulleid
N.J.
.
2010
.
Recycling of peroxiredoxin IV provides a novel pathway for disulphide formation in the endoplasmic reticulum
.
EMBO J.
29
:
4185
4197
.
Tian
G.
,
Xiang
S.
,
Noiva
R.
,
Lennarz
W.J.
,
Schindelin
H.
.
2006
.
The crystal structure of yeast protein disulfide isomerase suggests cooperativity between its active sites
.
Cell.
124
:
61
73
.
Tu
B.P.
,
Weissman
J.S.
.
2002
.
The FAD- and O(2)-dependent reaction cycle of Ero1-mediated oxidative protein folding in the endoplasmic reticulum
.
Mol. Cell.
10
:
983
994
.
Tu
B.P.
,
Ho-Schleyer
S.C.
,
Travers
K.J.
,
Weissman
J.S.
.
2000
.
Biochemical basis of oxidative protein folding in the endoplasmic reticulum
.
Science.
290
:
1571
1574
.
Vitu
E.
,
Kim
S.
,
Sevier
C.S.
,
Lutzky
O.
,
Heldman
N.
,
Bentzur
M.
,
Unger
T.
,
Yona
M.
,
Kaiser
C.A.
,
Fass
D.
.
2010
.
Oxidative activity of yeast Ero1p on protein disulfide isomerase and related oxidoreductases of the endoplasmic reticulum
.
J. Biol. Chem.
285
:
18155
18165
.
Wang
L.
,
Li
S.J.
,
Sidhu
A.
,
Zhu
L.
,
Liang
Y.
,
Freedman
R.B.
,
Wang
C.C.
.
2009
.
Reconstitution of human Ero1-Lalpha/protein-disulfide isomerase oxidative folding pathway in vitro. Position-dependent differences in role between the a and a’ domains of protein-disulfide isomerase
.
J. Biol. Chem.
284
:
199
206
.
Wilkinson
B.
,
Gilbert
H.F.
.
2004
.
Protein disulfide isomerase
.
Biochim. Biophys. Acta.
1699
:
35
44
.
Zito
E.
,
Melo
E.P.
,
Yang
Y.
,
Wahlander
A.
,
Neubert
T.A.
,
Ron
D.
.
2010
.
Oxidative protein folding by an endoplasmic reticulum-localized peroxiredoxin
.
Mol. Cell.
40
:
787
797
.

    Abbreviations used in this paper:
     
  • BSO

    buthionine sulfoximine

  •  
  • CPY

    carboxypeptidase Y

  •  
  • FAD

    flavin adenine dinucleotide

  •  
  • MBP

    maltose binding protein

  •  
  • PDI

    protein disulfide isomerase

  •  
  • SMM

    supplemented minimal medium

  •  
  • TEV

    Tobacco etch virus

Author notes

C.S. Sevier’s present address is Dept. of Molecular Medicine, Cornell University, Ithaca, NY 14853.

This article is distributed under the terms of an Attribution–Noncommercial–Share Alike–No Mirror Sites license for the first six months after the publication date (see http://www.rupress.org/terms). After six months it is available under a Creative Commons License (Attribution–Noncommercial–Share Alike 3.0 Unported license, as described at http://creativecommons.org/licenses/by-nc-sa/3.0/).