Research article

The endoplasmic reticulum sulfhydryl oxidase Ero1β drives efficient oxidative protein folding with loose regulation

Lei Wang, Li Zhu, Chih-chen Wang


In eukaryotes, disulfide bonds are formed in the endoplasmic reticulum, facilitated by the Ero1 (endoplasmic reticulum oxidoreductin 1) oxidase/PDI (protein disulfide-isomerase) system. Mammals have two ERO1 genes, encoding Ero1α and Ero1β proteins. Ero1β is constitutively expressed in professional secretory tissues and induced during the unfolded protein response. In the present work, we show that recombinant human Ero1β is twice as active as Ero1α in enzymatic assays. Ero1β oxidizes PDI more efficiently than other PDI family members and drives oxidative protein folding preferentially via the active site in the a′ domain of PDI. Our results reveal that Ero1β oxidase activity is regulated by long-range disulfide bonds and that Cys130 plays a critical role in feedback regulation. Compared with Ero1α, however, Ero1β is loosely regulated, consistent with its role as a more active oxidase when massive oxidative power is required.

  • disulfide-bond formation
  • endoplasmic reticulum oxidoreductin 1 (Ero1)
  • oxidative protein folding
  • protein disulfide-isomerase (PDI)
  • redox regulation


The formation of correct disulfide bonds is essential for the folding, stability and function of many secretory and membrane proteins. In eukaryotes, oxidative folding occurs mainly in the ER (endoplasmic reticulum) and to a lesser extent in mitochondria [1]. The ER provides an optimal environment for efficient oxidative folding of proteins, with a lower GSH/GSSG ratio than that in the cytosol and with a large number of folding catalysts [2]. A conserved electron-transfer pathway from PDI (protein disulfide-isomerase) to the sulfhydryl oxidase Ero1 (ER oxidoreductin 1) has been identified in the ER of various eukaryotes, first in yeast [3,4] and then in mammals [5,6] and plants [7].

Briefly, yPDI (yeast PDI) introduces disulfide bonds into substrates and becomes reduced. It is then oxidized by a shuttle cysteine pair (Cys100–Cys105) of Ero1p. Through an internal dithiol–disulfide exchange with the active site (Cys352–Cys355), electrons are transferred to the FAD cofactor and further to molecular oxygen, resulting in the formation of hydrogen peroxide. The activity of Ero1p has been demonstrated to be controlled through non-catalytic regulatory disulfides, which connect the extended ‘loop cap’ region (containing the shuttle cysteine residues) to the surface of the helical core of the molecule. The oxidation of the regulatory cysteine pairs leads to inactivation of the enzyme [8]. In mammalian cells, there are two homologues of Ero1p, Ero1α and Ero1β, both of which function to accelerate oxidative protein folding in the ER through interactions with PDI [9]. A novel regulatory disulfide bond Cys94–Cys131 in Ero1α has been recently shown to regulate redox flux, directly inactivating the shuttle cysteine pair in contrast with the regulatory mechanism in Ero1p [1012].

Ero1β is abundantly expressed in professional secretory tissues (e.g. in the pancreas). Its transcription is induced in the course of the unfolded protein response [6,13]. Like Ero1α, Ero1β plays an important role in disulfide-bond formation [14]. Recently, Ero1β was discovered to exhibit a selective and non-redundant function in the oxidative folding of pro-insulin in the islets of Langerhans. However, in the compound Ero1 mutant mouse, only a modest delay in the oxidative folding of IgM was observed [15]. This result suggests that there are other Ero1-independent mechanisms for disulfide-bond formation in mammals [16,17]. Referring to the published disulfide-bond mapping in Ero1p [18] and human Ero1α (hereafter referred to as Ero1α) [19], several well-conserved cysteine pairs in human Ero1β (hereafter referred to as Ero1β) can be found: Cys393–Cys396, the active site to directly accept oxidizing equivalents from FAD; Cys90–Cys95, the shuttle cysteine pair to accept electrons from the active site of hPDI (human PDI); Cys81–Cys390, the longest-range non-catalytic disulfide bond linking the potential ‘loop cap’ and ‘helical core’; Cys207–Cys240, the counterpart of which exists as a structural disulfide bond in Ero1α; and Cys90–Cys130, which could be formed like Cys94–Cys131 in the inactive state of Ero1α. Moreover, Ero1β contains an additional cysteine residue (Cys262) in the helical core domain, providing the possibility of forming a unique disulfide bond in this region. However, the catalytic biochemistry of Ero1β has been poorly characterized and no intramolecular regulatory mechanisms have been described so far for Ero1β. In addition, why mammals need two ERO1 genes is still an unsolved question.

In the present work, we prepared recombinant full-length Ero1β protein, and showed that its in vitro enzymatic activity is approximately twice that of Ero1α. Our findings demonstrate that Ero1β oxidizes preferentially the a′ domain of hPDI, and that other hPDI family members examined except PDIp (pancreas-specific PDI homologue) are inefficient substrates for Ero1β. Furthermore, we established for the first time that the oxidase activity of Ero1β is regulated by long-range disulfide bonds and that Cys130 plays a critical role in feedback regulation. Taken together, the regulation of Ero1β seems to occur in a looser manner compared with Ero1α.


Plasmid construction and mutagenesis

The cDNA encoding the Ero1β sequence 34–467 without the signal sequence from pcDNA3.1ERO1β−myc (generously provided by Professor R. Sitia, Division of Genetics and Cell Biology, Università Vita-Salute San Raffaele, Milan, Italy) was inserted into the pGEX-6P-1 vector (Amersham) at the BamHI and XhoI sites. All of the Ero1β constructs containing Cys→Ala mutations and the Ero1α mutants were created using a site-directed mutagenesis kit (Transgene), and all constructs were verified by DNA sequencing.

Protein expression and purification

The glutathione transferase–Ero1β fusion construct was expressed in Escherichia coli BL21(DE3)pLysS cells (Novagen). Cell cultures were grown in 2 litres of 2YT medium [1.6% (w/v) tryptone/1% (w/v) yeast extract/0.5% NaCl] at 37 °C for 4 h and shaken at 200 rev./min for an additional 20 h at 16 °C after adding isopropyl β-D-thiogalactopyranoside to 40 μM. After centrifugation (4670 g for 30 min), pelleted cells were resuspended in lysis buffer [50 mM sodium phosphate buffer, pH 6.5, 150 mM NaCl and 0.1% Triton X-100] and the supernatant of the cell lysate was applied on to a glutathione–Sepharose 4 Fast Flow column (Amersham) pre-equilibrated with lysis buffer. After washing with 5–10 column vol. of lysis buffer, the glutathione–Sepharose beads were incubated with 100 μl (200 units) of PreScission Protease (Amersham) at 4 °C overnight. The eluted Ero1β protein was then buffer-exchanged into 50 mM Tris/HCl, pH 8.5, with 150 mM NaCl, concentrated and stored at −80 °C as aliquots. The finally purified Ero1β protein had an additional GPLGS pentapeptide at the N-terminus. Recombinant WT (wild-type) Ero1α and recombinant full-length hPDI were expressed and purified as described previously [20]. PDIp, ERp57, ERp72, P5 and ERp18 proteins were expressed and purified in the same way as hPDI. All mutant proteins were expressed and purified using the same protocol as for WT proteins.

Protein concentrations were determined using the Bradford method with BSA as a standard [21]. The absorption spectrum of Ero1β was recorded using a UV-2501 PC spectrophotometer (Shimadzu). The cofactor/enzyme ratio was determined based on the molar absorption coefficient of 12.5 mM−1·cm−1 at 454 nm for enzyme-bound FAD as previously reported for Ero1p [22].

Oxygen consumption assay

Oxygen consumption was measured by using an Oxygraph Clark-type oxygen electrode (Hansatech Instruments) [20]. E. coli Trx (thioredoxin) was reduced by DTT (dithiothreitol) as described previously [22]. All components of each reaction, except for Ero1β, were freshly mixed in a total volume of 0.5 ml, and the reaction was initiated by injection of Ero1β into the reaction vessel of the oxygen electrode. Assays with hPDI mutants or hPDI family members were carried out by monitoring oxygen consumption immediately after the injection of 1 μM Ero1β protein into 100 μM or 50 μM hPDI protein and 10 μM FAD, with 10 mM GSH as a reducing equivalent [11]. Oxidation rate was calculated by measuring the slope of the linear phase of the oxygen consumption curve.

RNase A assay

Re-activation of denatured and reduced RNase A was assayed quantitatively by monitoring the absorbance increase at 296 nm at 25 °C due to the hydrolysis of cCMP [23], and the detailed calculation for isomerase and oxidoreductase activities of hPDI can be found elsewhere [20,24].

Gel-based oxidation assays

Gel-based Trx oxidation analyses were performed by adding 2 μM Ero1β to 100 μM Trxred (reduced Trx) followed by incubation at 25 °C. At various time points, 10-μl samples were removed for immediate mixing with 5× SDS loading buffer [250 mM Tris/HCl buffer, pH 6.8, 10% (w/v) SDS, 50% (v/v) glycerol and 0.5% Bromophenol Blue] containing 10 mM AMS (4-acetamido-4′-maleimidylstilbene-2,2′-disulfonic acid; Invitrogen), and 50 μl of the same samples were mixed with 5× SDS loading buffer without AMS. Samples for reducing SDS/PAGE contained 2% (v/v) 2-mercaptoethanol. The samples were analysed by SDS/15% (or 10%) PAGE and visualised by Coomassie Blue staining to determine the relative oxidation states of Trx (or Ero1β). Alkylation of each free thiol (sulfhydryl) group of the protein by AMS adds 0.5 kDa to the molecular mass, resulting in a lower migration rate on SDS/PAGE.

Gel-based RNase A re-oxidation analyses were performed by incubation of 3 μM Ero1β, 3 μM hPDI and/or 100 μM FAD with 15 μM denatured and reduced RNase A at 25 °C. At different time points, free thiols were blocked by the addition of 5× SDS loading buffer containing 10 mM AMS. The samples were resolved by non-reducing SDS/15% PAGE stained with Coomassie Blue thereafter.


Characterization of recombinant Ero1β

Full-length Ero1β produced in the BL21(DE3)pLysS strain of E. coli displayed a yellowish colour and an apparent molecular mass of 50 kDa on reducing SDS/PAGE with a purity >95% (see Supplementary Figure S1A at Unlike the three discrete redox states (R, OX1 and OX2) observed for Ero1α, the recombinant Ero1β preparation shows two predominant redox bands on non-reducing SDS/PAGE. The upper band corresponds to the reduced form (R) and the lower band to the oxidized form (OX) with a more compact conformation (Supplementary Figure S1B), which mirrors the redox states observed in vivo [13]. Recombinant Ero1α and Ero1β proteins were eluted mainly as a single peak on reverse-phase HPLC (Supplementary Figure S1C), and far-UV CD spectra suggested that both proteins were well folded and stable during storage (Supplementary Figure S1D).

The absorbance spectrum of Ero1β showed a peak at 454 nm characteristic of a bound FAD moiety (Figure 1A). After treatment with 6 M guanidinium chloride, the spectrum in the visible region became similar to that of free FAD (Figure 1A) as indicated by the shift of the maximum absorbance from 454 nm to ~448 nm and the disappearance of the shoulder at 485 nm. Unlike Ero1α, which contains a molar equivalent of FAD molecules [20], one Ero1β contains on average only 0.5 FAD, as determined using the molar absorption coefficient of 12.5 mM−1·cm−1 at 454 nm. Addition of a 10-fold excess of exogenous FAD fulfilled the maximum oxygen consumption rate in the oxidation of DTT (Figure 1B). Ero1p overexpressed from yeast has been reported to be active only in the presence of exogenous FAD [25,26], and recombinant Ero1p [22] and Ero1α [20] produced from bacteria contain an equimolar prosthetic group. The different FAD-binding ratios observed among these Ero1 preparations probably resulted from the different purification protocols.

Figure 1 Biochemical characterization of the Ero1β preparation

(A) Absorbance spectra of Ero1β at 12.8 μM were measured in the absence (continuous line) or presence (dashed line) of 6 M guanidinium chloride, with the spectrum of 13.3 μM free FAD (dotted line) (11.3 mM−1·cm−1 for the molar absorption coefficient at 446 nm) as a control. Inset: enlarged spectra in the visible region. (B) Oxygen consumption was monitored immediately after injection of 1 μM Ero1β into 10 mM DTT in the absence and presence of exogenous FAD at the molar ratios indicated. (C) Size-exclusion chromatography of Ero1β in the absence (continuous line) or presence (dashed and dotted line) of 10 mM DTT, and of the isolated monomer (dotted line) and dimer (dashed line) fractions, was performed on a Superdex 75 HR 10/30 column (Amersham) at room temperature (25 °C) using Tris/HCl buffer, pH 8.5, containing 0.2 M NaCl at 0.5ml/min. Molecular-mass markers (Amersham) were: albumin, 67 kDa; ovalbumin, 43 kDa; chymotrypsinogen A, 25 kDa; and RNase A, 13.7 kDa.

As shown in Figure 1(C), the gel-filtration profile of purified Ero1β showed a major peak of 50 kDa and a minor one (~30%) of 85 kDa. Since the content of dimer decreased after DTT treatment, the dimer is mainly a disulfide-linked homodimer. To address their functional relevance, the two Ero1β peaks were isolated from the gel-filtration apparatus for a subsequent activity assay. Both species were catalytically active, although dimers were somewhat less active than the unfractionated preparation and the monomer fraction (see Supplementary Figure S2 at

Ero1β shows higher in vitro sulfhydryl oxidase activity than Ero1α

We studied the in vitro catalytic chemistry of Ero1β by monitoring oxygen consumption. During the oxidation of 50 μM Trxred catalysed by Ero1β, 46 μM oxygen was consumed. Addition of catalase restored 25 μM oxygen in the solution (resulting from the production of 50 μM hydrogen peroxide) (Figure 2A). The Km of 5.5 μM (Figure 2B) for oxygen is very similar to that of Ero1α [20], suggesting that Ero1β can also function efficiently at low oxygen levels. Thus Ero1β and Ero1α catalyse the same thiol oxidation reaction (2R−SH+O2→R−S−S−R+H2O2).

Figure 2 Thiol oxidation reaction catalysed by Ero1β

(A) Oxygen consumption was monitored immediately after injection of 1 μM Ero1β and 10μM FAD into solution with (continuous line) or without (dotted line) 50 μM Trxred. Catalase was added at the indicated time (arrow). The turnover rate for Ero1β with Trxred as a substrate is ~0.3 s−1. (B) Km determination for oxygen. Ero1β at 1 μM was injected into solution containing 10 mM DTT and 10 μM FAD. The graph shows the end of the reaction, when the oxygen has been depleted to <10% of its initial value. Inset: the first derivative of the data, which was used to identify the point at which the rate was half-maximal. (C) Oxygen consumption of 10 mM DTT in the presence of 1 μM Ero1β proteins and 10 μM FAD. (D) Oxygen consumption of 100 μM PDI with 10 mM GSH as reducing equivalents in the presence of 1 μM Ero1β proteins and 10 μM FAD. Curve 1, WT Ero1β; curve 2, Ero1β C90/95A; curve 3, Ero1β C393/396A; curve 4, Ero1β C90/95A plus Ero1β C393/396A.

Table 1 compares the catalytic activities of recombinant Ero1β and Ero1α. In the presence of a 10-fold excess of exogenous FAD, using either an artificial substrate (DTT) or a physiological substrate (hPDI), the turnover number of Ero1β for oxygen is approximately twice that for Ero1α. It is generally believed that the active-site sequences play an important role in the enzymatic activities of the oxidoreductase family [27], so the unusual phenylalanine residue in the Cys394-Phe-Lys-Cys397 active site of Ero1α might underlie differences in the redox activities of Ero1α and Ero1β, which has a Cys393-Asp-Lys-Cys396 active site [6]. Accordingly, Ero1α F395D was prepared, and this mutant indeed exhibited an increased ability to catalyse disulfide-bond formation (Table 1) and to drive oxidative refolding of RNase A (see Supplementary Figure S3 at Thus, the Cys-acidic-basic-Cys motif makes Ero1β a more active sulfhydryl oxidase than Ero1α.

View this table:
Table 1 Comparison of substrate kinetics of Ero1α and Ero1β

FAD (10-fold) was added to Ero1α, Ero1β and Ero1α F395D. GSH (10 mM) was used as a reducing source for hPDI. Turnover numbers are expressed in terms of oxygen molecules consumed per s; results shown are means±S.D. (n≥3). N.d., not determined

The effects of the two conserved redox-active cysteine pairs Cys90–Cys95 and Cys393–Cys396 on the catalytic activity of Ero1β were studied using oxygen consumption assays. First, the low-molecular-mass reducing agent DTT was used as the electron donor, as it can directly reduce the active site near FAD. As expected, the active-site C393A/C396A mutant showed no or very little activity, whereas the C90A/C95A mutant showed an activity similar to that of WT Ero1β (Figure 2C), confirming that Cys393–Cys396 is the inner active site. However, when the native substrate hPDI was used, both mutants were inactive (Figure 2D). As hPDI cannot access and reduce the active-site disulfide bond of Ero1β directly, this result suggests that Cys90–Cys95 is essential for transferring the electrons from hPDI to Cys393–Cys396. Moreover, the mixture of these two mutants showed no detectable recovery of oxidase activity (Figure 2D), suggesting that, at least in vitro, there is no intermolecular electron transfer for Ero1β. Taken together, the disulfide-transfer mechanism in Ero1β is conserved compared with its relatives: following disulfide transfer to hPDI, the Ero1β shuttle cysteine pair is re-oxidized by disulfide exchange with its inner active site [19].

Ero1β drives oxidative protein folding in vitro efficiently

A general Ero1β/hPDI oxidative protein-folding pathway was previously identified in vivo [9]. By monitoring the re-activation of denatured and reduced RNase A, we reconstituted the Ero1β/hPDI system in vitro in order to study the molecular interactions of Ero1β with hPDI. As shown in Figure 3(A), RNase A cannot be re-activated in the absence of either Ero1β or hPDI. In the presence of hPDI, the re-activation rate increased with increasing concentrations of Ero1β, at 3 μM reaching the levels obtained with the optimal glutathione redox buffer (1 mM GSH/0.2 mM GSSG). Ero1β was approximately twice as active as Ero1α as an oxidase. However, in contrast with the oxidoreductase activity of hPDI measured by the rate of RNase A oxidation, the isomerase activity of hPDI measured by the rate of RNase A re-activation was basically independent of the nature of the oxidizing equivalents (see Supplementary Figure S4 at A gel-based RNase A re-oxidation analysis (Figure 3B) confirmed that the re-activation of RNase A requires both hPDI and Ero1β as catalysts to deliver oxidizing equivalents, and that the presence of exogenous FAD accelerated RNase A re-oxidation (see Supplementary Figure S5 at Taken together, the results confirm that hPDI can work as an efficient oxidoreductase/isomerase in a system driven by the sulfhydryl oxidase Ero1β.

Figure 3 Reconstitution of Ero1β/hPDI oxidative protein-folding system in vitro

(A) Re-activation of denatured and reduced 8 μM RNase A in the presence of: 3 μM hPDI and 100 μM FAD (+); 3 μM Ero1β and 100 μM FAD (×); and 3 μM hPDI, 100 μM FAD and 3 μM Ero1α (●) or Ero1β at 1 μM (▲), 2 μM (▼) or 3 μM (◆) respectively. RNase A refolding in the optimal glutathione redox buffer (1 mM GSH/0.2 mM GSSG) containing 3 μM hPDI (■) was used for comparison. (B) Direct observation of Ero1β-mediated catalysis of disulfide formation in RNase A by hPDI. Refolding of 15 μM RNase A was carried out in the absence or presence of 3 μM Ero1β, 3 μM hPDI and/or 100 μM FAD as indicated, quenched with AMS at various time points and analysed by non-reducing SDS/15% PAGE. The position of each species is indicated. OX, oxidized form; R, reduced form. (C) Ero1β preferentially oxidizes the active site in the a′ domain of hPDI. Re-activation of 8 μM RNase A was determined in the presence of 3 μM Ero1β, 100 μM FAD and 3 μM hPDI proteins: hPDI (■), hPDI-ΔC (□), hPDI-bbxa′ (●), hPDI-bxa′ (◆), hPDI-abbx (○), hPDIC1 (▲), hPDIC2 (△), hPDIC1/2 (+). Schematic representation of WT hPDI and mutant proteins is shown on the right, and the Cys-Gly-His-Cys active sites (★) and the mutated Ser-Gly-His-Ser sites (☆) are shown; residue numbering is for mature hPDI (without the signal sequence).

We investigated further the details of the Ero1β−hPDI interaction using reconstituted systems containing Ero1β and different hPDI domain combinations and mutants. We have previously shown that all hPDI combinations of one redox-active Trx domain plus the core bbx region exhibited roughly equivalent activities in GSH/GSSG-driven oxidative protein folding [20]. However, when Ero1β was used to supply the oxidizing equivalents, the various hPDI constructs containing a single redox-active Trx domain were not equivalent. As shown in Figure 3(C), the activity of hPDI-ΔC is very close to that of the full-length hPDI, indicating that the C-terminal tail is not involved in the interaction with Ero1β. Deleting domain a′ (hPDI-abbx) or mutating its Cys-Gly-His-Cys active site to Ser-Gly-His-Ser (hPDIC2) abolished the activity almost completely. Conversely, deletion of domain a (hPDI-bbxa′) or mutation of the Cys-Gly-His-Cys active site to Ser-Gly-His-Ser in domain a (hPDIC1) retained substantial activity, suggesting that the a′ domain is dominantly responsible for the electron flow to Ero1β. Another domain combination, bxa′, which has been identified as the minimal element for efficient interaction with Ero1α [20], displayed similar activity to bbxa′ in the Ero1β-driven system. As a negative control, the mutant full-length hPDI with two active sites both mutated to Ser-Gly-His-Ser (hPDIC1/2) showed no activity at all. This asymmetry between domains a and a′ of hPDI in Ero1β-mediated oxidative folding suggests that Ero1β preferentially oxidizes the active site in the a′ domain of hPDI, as does Ero1α [20]. Moreover, the electron-transfer efficiency between the isolated a or a′ domain of hPDI and Ero1β is very poor (L. Wang, L. Zhu, and C.-c. Wang, unpublished work), whereas the non-native substrate Trx (with low reduction potential) is a good electron donor to Ero1β (Figure 2A), further implying the critical role of the non-catalytic domain of hPDI in the interaction of hPDI with Ero1β. Our conclusion from these results is that the interaction patterns of the two hEro1 proteins with hPDI are very much conserved: stable binding of hPDI to hEro1 proteins requires both b′ and a′ domains, and the subsequent transient redox exchange of hEro1 proteins with hPDI occurs via dithiols in the a′ domain of hPDI. Experimental results from cells also supports this model that non-covalent interactions precede the formation of covalent linkages between hPDI and hEro1 proteins [28]. Intriguingly, a recent study in yeast showed that the oxidizing equivalents flow from Ero1p primarily through the a domain but not the a′ domain of yPDI [29]. This difference can be explained by different conformational plasticity in hPDI and yPDI, i.e. the bxa′ part is more flexible than the ab part in hPDI [30]; in contrast, the ab part is more flexible than the bxa′ part in yPDI [31].

Currently, 20 members of the hPDI family have been identified, and most of them contain at least one Trx-like catalytic domain [32]. Whether hEro1s could transfer oxidative equivalents to other hPDI family members has been an open question. Therefore we examined the rates of oxidation of hPDI, PDIp, ERp57, ERp72, P5 and ERp18 by Ero1β using oxygen consumption assays. The rate of oxidation of PDIp was ~50% of that of hPDI, implying the presence of a pancreas-specific Ero1β/PDIp oxidative protein-folding pathway. However, other hPDI family members tested in the present study were oxidized at markedly lower rates (<10%) by Ero1β (Table 2). Thus hPDI is the most efficient target of Ero1β among its family members. The presence of a specific hydrophobic b′ domain may be the precondition for hPDI proteins to associate with hEro1 proteins, and this concept is also supported by the crystal structure of ERp44, which strongly suggests that the binding of ERp44 to Ero1α is very probably through a hydrophobic pocket in domain b′ [33].

View this table:
Table 2 Ero1β oxidizes hPDI more efficiently than other hPDI family members

The oxidation rate was calculated as the amount of oxygen molecules consumed per s with 10 mM GSH used as a reducing source. The percentage relative oxidation rate was calculated as (AA0)/(A1−A0)×100, where A is the rate of oxidation of hPDI family members, A1 is rate of oxidation of hPDI, and A0 is the background oxidation rate. Results shown are means±S.D. (n≥3).

Long-range disulfide(s) are reduced during the Ero1β-catalysed reaction

Our in vitro assays demonstrated that Ero1α [20] and Ero1β generate hydrogen peroxide in equimolar amounts to the number of disulfide bonds formed, the same as for Ero1p [22]. Recent in vivo experiments showed that the accumulation of reactive oxygen species correlated with the oxidation of sulfhydryl groups catalysed by Ero1 in the ER [7,32]. Although peroxide was reported to be able to drive efficient oxidative protein folding in vitro [34], excess reactive oxygen species can be highly detrimental for the cells and thus Ero1 activity must be appropriately controlled. Distinct feedback-regulation mechanisms controlled by long-range disulfide bonds in Ero1p [35] and Ero1α [10,11] have been discovered, which stimulated us to check what kind of regulation mechanism exists for Ero1β. As shown in Figure 4(A), the substrate Trxred was gradually oxidized in the presence of catalytic amounts of Ero1β, and the oxidation reached completion after 5 min. Conversely, at the beginning of the reaction, Ero1β existed in two redox states (Figure 4B, lane 1) and the oxidized form (OX) disappeared after 20 s, indicating reduction of long-range disulfide(s) in Ero1β molecules (Figure 4B, lane 2). At 5 min, the oxidized form re-emerged, and most of Ero1β was oxidized after 20 min, indicating the re-formation of the long-range disulfide(s) once all of the Trx had been oxidized (Figure 4B, lane 7). Again, during the oxidative refolding of the physiological substrate RNase A, Ero1β was initially reduced and thereafter became re-oxidized after all RNase A molecules were refolded (Figure 3B). The above results clearly show that Ero1β has long-range disulfide(s), which are reduced during the Ero1β-catalysed reaction and may participate in feedback regulation.

Figure 4 The redox status of Ero1β changes transiently during the oxidation of Trxred

(A) The change of redox status of Trx during oxidation by Ero1β was analysed by non-reducing SDS/15% PAGE after alkylation with AMS. (B) The change of redox status of Ero1β during incubation with Trxred was analysed by non-reducing SDS/10% PAGE. OX, oxidized form; R, reduced form.

Identification of long-range disulfide(s) of Ero1β

To date, neither structural information nor a systematic study of cysteine-residue mutants of Ero1β is available. Although only two predominant bands were observed on the non-reducing gel, this does not mean that there is only one long-range disulfide bond in Ero1β, as reduction of different long-range disulfide bonds may result in very similar migration rates. To further identify the long-range disulfide bond(s), we constructed a group of Ero1β mutants with cysteine residues mutated individually or in combinations and examined the mobility changes of the panel of mutants on non-reducing gels. The two adjacent cysteine residues (Cys42 and Cys44) were not analysed, as their equivalents in Ero1α (Cys46 and Cys48) have been shown to form a disulfide bond in the OX2 state [10].

All cysteine mutants of Ero1β migrated identically with the WT protein on reducing SDS/PAGE gels (Figure 5A, lanes 1–15), but migration was distinctly different on the non-reducing gels. The mutation to an alanine residue of Cys165, Cys207 or Cys240 did not cause any change from an oxidized to a reduced state, indicating that none of them forms long-range disulfide bonds (Figure 5A, lanes 20, 29 and 30). As expected, the C81A mutant underwent a significant migration shift from an oxidized to a reduced state because Cys81 was predicted to form a long-range disulfide bond with Cys390 (Figure 5A, lane 17). C100A migrated with an intermediate mobility between the reduced and oxidized WT proteins, with a large conformational change (Figure 5A, lane 18). C130A was found in a predominantly reduced state (Figure 5A, lane 19), suggesting that Cys130 may form a regulatory disulfide bond with Cys90, as occurs for their equivalents in Ero1α [10,11]. Furthermore, the mutation of the additional cysteine, Cys262, which is absent from Ero1α, resulted in a shift from the oxidized to the reduced form (Figure 5A, lane 21), implying that this non-conserved cysteine residue could be involved in the formation of a long-range disulfide bond. Intriguingly, compared with C130A the two double-cysteine mutants, C81A/C130A and C100A/C130A, shifted the oxidized/reduced ratio back to that of the WT protein (Figure 5A; compare lanes 24 and 25 with lanes 16 and 19). This observation indicates that new long-range disulfide bonds were generated between the original partners of the two mutated cysteine residues. In other words, non-native disulfide bonds were formed in these double mutants, between Cys90 and Cys390 (the partner of Cys81) in C81A/C130A, and between Cys90 and the partner of Cys100 in C100A/C130A, with Cys262 being the most probable candidate.

Figure 5 Identification of the cysteine residues involved in the formation of long-range disulfides in Ero1β

(A) Reduced (R) and non-reduced (NR) WT Ero1β and its cysteine-residue mutants were analysed by SDS/10% PAGE as indicated. OX, oxidized form; R, reduced form. (B) Schematic illustration of the cysteine-residue connectivities in oxidized (OX) Ero1p [18,35,36], Ero1α OX2 [10,11] and oxidized Ero1β (the present study). The cysteine residues are shown as white, grey (shuttle cysteine pair) and black (active site) circles with numbering, and disulfide bonds are indicated as straight or curved (regulatory function) lines. The unique trypsin cleavage site at Lys119 in Ero1β is indicated by an asterisk. The ‘loop cap’ region is represented by a white bar. Ero1p has an additional transmembrane C-terminal tail. Note that in hEro1 proteins the shuttle cysteine pairs form disulfide bonds in the active state when the regulatory disulfide bonds are reduced.

To test this deduced disulfide-bond pattern, we took advantage of the unique cleavage of Ero1β by trypsin. In Ero1β there is a unique trypsin cleavage site C-terminal to Lys119 and a ~40 kDa fragment can be generated only when all the long-range disulfide bonds that bridge across the cleavage site at Lys119 were reduced. Unsurprisingly, after trypsin treatment, a ~40 kDa fragment appeared on reducing gels in addition to the 50 kDa full-length band for WT Ero1β and for all the cysteine mutants (see Supplementary Figure S6, lanes 1–12, at However, under non-reducing conditions, neither the WT protein nor the single/double cysteine mutants were split (Supplementary Figure S6, lanes 13–22). The large fragment only clearly showed up on non-reducing gels for the cleaved triple-cysteine mutant C81A/C100A/C130A (Supplementary Figure S6, lane 23), suggesting that Cys81, Cys100 and Cys130 participate in three different long-range disulfide bonds across Lys119 respectively. Moreover, the ~40 kDa fragment was not released in the C81/130/165A mutant (Supplementary Figure S6, lane 24), suggesting that Cys165 is not the partner of Cys100. The only possible cysteine residue C-terminal to Lys119 to pair with Cys100 is Cys262. Molecular modelling also supported this idea, as the distance between these two cysteine residues could be as close as 8.7 Å (1 Å=0.1 nm), implying the potential of a covalent linkage (results not shown). Attempts to purify the triple mutants C81A/C130A/C262A, C81A/C90A/C100A and C130A/C262A/C390A did not succeed, as simultaneous mutation of three residues involved in long-range disulfide bonds would probably make the protein molecule unstable. Taking our new data on Ero1β with previous studies on Ero1p [18,35] and Ero1α [10,11] together, we deduced that Ero1β contains three long-range disulfide bonds: Cys81–Cys390, Cys90–Cys130 and Cys100–Cys262. A schematic illustration summarizing the positions of the disulfide bonds in yeast and human Ero1 proteins is shown in Figure 5(B).

Loose regulation of Ero1β activity

To study whether the long-range disulfide bonds identified in the experiments decribed above regulate enzymatic activity of Ero1β, we carried out oxygen consumption assays and RNase A re-oxidation/re-activation assays to determine the activities of the WT protein and the mutant proteins lacking specific long-range disulfide bonds. The yield, activity and stability on storage of the C81A mutant significantly decreased compared with the WT protein (results not shown), indicating that the longest-range disulfide bond Cys81–Cys390 is critical for the structural stability of Ero1β rather than for activity regulation. In this respect, Cys90–Cys349 in Ero1p was first suggested to be a regulatory disulfide [35], but was later reported to be intact during the substrate oxidation reaction and not essential for Ero1p activity [36]. In Ero1α, Cys85–Cys391 was suggested to be a regulatory disulfide, although mutations at Cys85 and/or Cys391 made the protein unstable [11]. C100A showed slightly decreased activity, whereas C130A showed a small, but significant, increase in activity compared with WT protein. In addition, the activity of C100A/C130A towards hPDI was even higher than that of C130A (Figures 6A and 6B). These results imply that the Cys90–Cys130 disulfide bond functions as the predominant regulatory switch to modulate the activity of Ero1β, and that Cys100–Cys262 plays an auxiliary role in regulating Ero1β activity. Mutation of Cys130 results in a significant mobility retardation on non-reducing SDS/PAGE gels and an increase of catalytic activity, indicating that the less-compact form resulting from the absence of a disulfide bond defined by Cys130 is more active. This result showed that although Ero1β lacks a redox form corresponding to OX2 (defined by Cys94–Cys131) observed in Ero1α, it does have the equivalent regulatory disulfide switch, which was not defined in previous studies.

Figure 6 Long-range disulfide bonds regulate the enzymatic activity of Ero1β

(A) Oxygen consumption was monitored immediately after injection of 1 μM Ero1β proteins into 50 μM hPDI and 10 μM FAD, with 10 mM GSH as reducing equivalents. (B) Re-activation of 8 μM RNase A was monitored in the presence of 1 μM hPDI, 10 μM FAD and 1 μM Ero1β proteins. Curve 1, no Ero1β; curve 2, Ero1β C100A; curve 3, WT Ero1β; curve 4, Ero1β C130A; curve 5, Ero1β C100/130A. The relative oxidation rates derived from oxygen consumption assays (C), and the relative isomerase (black bars) and oxidoreductase (white bars) activities of hPDI derived from RNase A assays (D) were compared in the presence of deregulated Ero1 mutants or WT Ero1 proteins as indicated. The values in the presence of WT Ero1 proteins were taken as 100%. Results shown are means±S.D. (n≥3). *P< 0.05, **P< 0.01, ***P< 0.001 by paired one-tailed Student's t test. n.s., not significant.

Next, we compared the activities of deregulated hEro1 mutants and WT hEro1 proteins by calculating the rates of oxidation of hPDI (Figure 6C) or the isomerase/oxidoreductase activities of hPDI (Figure 6D). In both analyses, the C100A/C130A mutation in Ero1β moderately increased the oxidizing ability compared with the WT protein, whereas Ero1α C104A/C131A exhibited markedly raised oxidizing ability compared with WT Ero1α. On the other hand, both of the deregulated mutants, Ero1β C100A/C130A and Ero1α C104A/C131A, showed little effect on the isomerase activity of hPDI. Our observation that hEro1 proteins did not influence the isomerase activity of hPDI (Figure 6D and Supplementary Figure S4) suggests that the reduction/isomerization pathway seems segregated from the Ero1-driven oxidation pathway. From these results, we concluded that Ero1β is a loosely regulated sulfhydryl oxidase and appears to be operating at high speed, whereas the activity of Ero1α seems to be impeded by a tight regulatory disulfide brake. The above regulatory mechanism was concluded on the basis of our recombinant protein, and obviously further studies under physiological conditions are needed to provide conclusive evidence for this model.

Concluding remarks

In the present work, by using the recombinant protein and an in vitro reconstituted system, we have provided the first evidence that the oxidase activity of Ero1β is controlled by long-range disulfide bonds, and Cys130 is the most critical cysteine residue that contributes to activity regulation. It is reasonable to speculate that during the physiological disulfide generation catalysed by Ero1β, Cys130 may compete with reduced substrates, such as PDI, for disulfide exchange with a shuttle cysteine pair, in a similar manner to Ero1α [10]. The relatively looser regulation and higher turnover of Ero1β observed from our in vitro system implies that Ero1β may be of greater physiological relevance for the oxidative folding of secreted proteins.


Lei Wang and Chih-chen Wang conceived the project. Chih-chen Wang provided funding and guided the experimental work. Lei Wang and Li Zhu performed experiments and analysed data. Lei Wang generated the first draft of the paper and the Figures, and all authors contributed to revision and approval of the final paper.


This work was supported by the Chinese Ministry of Science and Technology [grant number 2011CB910303], the Chinese Academy of Sciences [grant numbers KSCX2-YW-R-119 and KSCX2-YW-R-256] and the National Natural Science Foundation of China [grant number 31000351].


We thank Roberto Sitia (Division of Genetics and Cell Biology, Università Vita-Salute San Raffaele, Milan, Italy) for generously providing the Ero1β cDNA, Lloyd Ruddock (Department of Biochemistry, University of Oulu, Oulu, Finland) for the PDIp, ERp72, P5 and ERp18 plasmids, and Xi Wang (National Laboratory of Biomacromolecules, Institute of Biophysics, Chinese Academy of Sciences, Beijing, China) for the ERp57 plasmid and Trx protein. We also thank them and Robert Freedman (Department of Biological Sciences, University of Warwick, Coventry, U.K.) for helpful discussions and critical reading of the manuscript, and Zhensheng Xie (Laboratory of Proteomics, Institute of Biophysics, Chinese Academy of Sciences, Beijing, China) for providing HPLC analysis.

Abbreviations: AMS, 4-acetamido-4′-maleimidylstilbene-2,2′-disulfonic acid; DTT, dithiothreitol; ER, endoplasmic reticulum; Ero1, ER oxidoreductin 1; hEro1, human Ero1; PDI, protein disulfide-isomerase; hPDI, human PDI; PDIp, pancreas-specific PDI homologue; Trx, thioredoxin; Trxred, reduced Trx; WT, wild-type; yEro1p, yeast Ero1p; yPDI, yeast PDI


View Abstract