Reconstitution of Human Ero1-Lα/Protein-Disulfide Isomerase Oxidative Folding Pathway in Vitro

Protein-disulfide isomerase (PDI), a critical enzyme responsible for oxidative protein folding in the eukaryotic endoplasmic reticulum, is composed of four thioredoxin domains a, b, b′, a′, and a linker x between b′ and a′. Ero1-Lα, an oxidase for human PDI (hPDI), has been determined to have one molecular flavin adenine dinucleotide (FAD) as its prosthetic group. Oxygen consumption assays with purified recombinant Ero1-Lα revealed that it utilizes oxygen as a terminal electron acceptor producing one disulfide bond and one molecule of hydrogen peroxide per dioxygen molecule consumed. Exogenous FAD is not required for recombinant Ero1-Lα activity. By monitoring the reactivation of denatured and reduced RNase A, we reconstituted the Ero1-Lα/hPDI oxidative folding system in vitro and determined the enzymatic activities of hPDI in this system. Mutagenesis studies suggested that the a′ domain of hPDI is much more active than the a domain in Ero1-Lα-mediated oxidative folding. A domain swapping study revealed that one catalytic thioredoxin domain to the C-terminal of bb′x, whether a or a′, is essential in Ero1-Lα-mediated oxidative folding. These data, combined with a pull-down assay and isothermal titration calorimetry measurements, enabled the minimal element for binding with Ero1-Lα to be mapped to the b′xa′ fragment of hPDI.

Disulfide bonds play an essential role in the folding and assembly of secretory proteins and membrane proteins. The formation of non-natural disulfide bonds leads to protein misfolding and further aggregation. Both prokaryotes and eukaryotes possess specialized cellular compartments in which disulfide bonds are formed, complete with enzymes devoted to catalyzing disulfide formation and rearrangement (1).
In eukaryotes, oxidative protein folding occurs in the endoplasmic reticulum (ER), 2 facilitated by protein-disulfide isomerase (PDI, EC 5.3.4.1) and/or its homologues, ERp57, PDIp, ERp72, PDIr, P5, ERp44, etc. (2,3). PDI comprises four thioredoxin (Trx) domains, two catalytic domains, a (9 -116) and aЈ (352-462), separated by two homologous non-catalytic domains, b (120 -217) and bЈ (219 -332), plus a C-terminal region (463-491). In addition, there is a short linker region "x" (333-351) between the bЈ and aЈ domains (4,5). As an abundant multifunctional protein within the lumen of the ER, PDI acts as both an enzyme and a chaperone in many physiological processes (6 -8). PDI can act in vitro to catalyze both proteindisulfide isomerization and net protein disulfide formation, but the latter process requires a further source of oxidizing equivalents to reoxidize PDI. Hence a net source of oxidizing equivalents is required in vivo to sustain formation of protein disulfides and oxidative protein folding. Ero1p was found to be directly responsible for the production of disulfide bonds in the Saccharomyces cerevisiae ER (9,10). The catalytic cycle carried out by Ero1p involves direct oxidation of PDI by the shuttle cysteine pair (Cys 100 -Cys 105 ), and re-oxidation of the reduced shuttle cysteine residues through internal dithiol-disulfide exchange with the active site (Cys 352 -Cys 355 ). The latter are re-oxidized by transfer of their electrons to the flavin adenine dinucleotide (FAD) cofactor and further to molecular oxygen, resulting in the formation of hydrogen peroxide. Ero1p activity can be modulated through the redox state of two non-essential "regulatory" cysteine pairs (Cys 90 -Cys 349 and Cys 150 -Cys 295 ) by controlling the range of motion for the shuttle-cysteine-containing loop region (11).
In mammalian cells, there are two homologues of Ero1p, Ero1-L␣, and Ero1-L␤, which show different tissue distribution and transcriptional regulation (12,13). Both Ero1-L␣ and Ero1-L␤ accelerate oxidative protein folding in the ER of human cells through their action on PDI (14). Although it was suggested, by analogy to the case in Ero1p, that the two conserved cysteine triads (Cys 85 -Cys 94 -Cys 99 and Cys 391 -Cys 394 -Cys 397 ) in Ero1-L␣ molecule may cooperate in electron transfer (15), little is known about the biochemistry of human Ero1 proteins in catalyzing oxidative folding and the interaction with human PDI (hPDI). Indeed, very little work has been reported on the reconstitution of this process in vitro.
In this work, we report that Ero1-L␣ utilizes oxygen as a terminal electron acceptor, producing one disulfide bond and one molecule of hydrogen peroxide using the prosthetic group FAD as an electron transfer component. Furthermore, we have reconstituted the in vitro Ero1-L␣/hPDI oxidative folding system, and demonstrated by mutagenesis studies on hPDI and domain-swapped hPDI (DShPDI) that the a and aЈ domains of hPDI have distinct activities in Ero1-L␣-mediated oxidative folding and that this is primarily due to their positions in the PDI molecule. Calorimetry and pull-down determinations revealed that the bЈxaЈ region is the minimal fragment of hPDI that is able to bind with Ero1-L␣. Our studies provide insights into the interaction of these two important catalysts of the mammalian oxidative protein folding machine.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-The cDNA encoding the Ero1-L␣ sequence 24 -468 without the signal sequence from pcDNA3.1ERO1-L␣-Myc (a kind gift from Prof. R. Sitia, Università Vita-Salute San Raffaele, Minalo, Italy) was inserted into pGEX-6P-1 vector (Amersham Biosciences) at BamHI and XhoI sites. The glutathione S-transferase (GST)-Ero1-L␣ fusion construct was expressed in BL21 (DE3) pLysS (Novagen). Cultures were grown in 2 liters of 2ϫ YT medium at 37°C for 4 h and shaken for an additional 20 h at 25°C after isopropyl-␤-Dthiogalactoside was added to a final concentration of 40 M. The supernatant of the cell lysate was applied onto a glutathione-Sepharose 4 Fast Flow column (Amersham Biosciences) pre-equilibrated with phosphate-buffered saline solution (PBS, 140 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 , pH 7.3). After washing with 5-10 column volumes of PBS, the GST-Ero1-L␣ fusion protein was eluted with 50 mM Tris-HCl, 10 mM GSH, pH 8.0. The elution was dialyzed against buffer A (50 mM Tris-HCl and 150 mM NaCl, pH7.6) to remove GSH, and then digested with 100 l (200 units) of PreScission Protease (Amersham Pharmacia Biosciences) at 4°C for 4 h followed by chromatography on a glutathione 4 Fast Flow column pre-equilibrated with buffer A to remove the GST moiety and the PreScission Protease. The flow-through was concentrated and stored as aliquots in buffer A at Ϫ80°C. The final purified Ero1-L␣ protein has an additional GPLGS pentapeptide at the N terminus after removing the GST tag.
Recombinant full-length hPDI was purified as described (16). hPDI-bbЈx, hPDI-a and hPDI-aЈc proteins were expressed from clones kindly provided by Prof. L. W. Ruddock (University of Oulu, Finland). The appropriate coding regions were inserted into pLWR51p (a modified version of pET23b (Novagen)), and the proteins were expressed and purified as described previously (5); the resulting proteins all included an N-terminal His tag (MHHHHHHM). All other hPDI and DShPDI mutants were constructed by overlap extension PCR using pQE30-hPDI plasmid as a template and subsequently inserted into pQE30 vector (Qiagen). The mutated proteins, containing an N-terminal His tag (MRGSHHHHHHGS), were expressed and purified using the same protocol as for the full-length hPDI.
The sequences of all the constructs were verified by DNA sequencing. Protein concentrations were determined by the Bradford method with bovine serum albumin as a standard (17).
Oxygen Consumption Assay-Escherichia coli Trx (a kind gift from Dr. Xi Wang in this group) was reduced by DTT as described (18). Oxygen consumption was measured by using an Oxygraph Clark-type oxygen electrode (Hansatech Instruments). All components of each reaction, except Ero1-L␣, were fresh mixed in a total volume of 0.5 ml, and the reaction was initiated by injection of Ero1-L␣ into the reaction vessel of the oxygen electrode. For detection of hydrogen peroxide, 5 l of catalase at 2 mg/ml was injected into the reaction vessel at 10 min, when the thiol oxidation reaction was complete. All experiments were performed in buffer B (100 mM Tris-HAc, 50 mM NaCl, and 1 mM EDTA, pH 8.0).
RNase A Assay-Denatured and reduced RNase A was prepared as described (19). RNase A reactivation was assayed quantitatively by monitoring the absorbance increase at 296 nm at 25°C due to the hydrolysis of cCMP, and the isomerase and oxidoreductase activities of hPDI proteins are calculated as previously reported (20). The relative activity (%) was calculated as (A-A0)/(A1-A0) ϫ 100%, where A, activity of hPDI mutants or DShPDI mutants; A1, activity of hPDI or DShPDI; A0, activity of hPDIC1/2 or DShPDIC1/2, respectively. Data are expressed as mean Ϯ S.D. (n ϭ 3).
Gel-based RNase A re-oxidation analyses were performed by addition of Ero1-L␣, hPDI, and/or FAD into buffer B containing 15 M denatured and reduced RNase A. At the indicated time points, free thiols were blocked by the addition of 6ϫ SDS-loading buffer containing 12 mM 4-acetamido-4Ј-maleimidylstilbene-2,2Ј disulfonic acid (AMS, invitrogen). The oxidized and reduced forms of RNase A were detected by Coomassie Blue staining after non-reducing SDS-PAGE.
GST-Pull-down Assay-GST-Ero1-L␣ fusion protein on glutathione-Sepharose resin was incubated with hPDI proteins for 2 h at 4°C in PBS, washed with PBS for three times, and bound proteins were detected by Coomassie Blue staining after reducing SDS-PAGE.
Isothermal Titration Calorimetry (ITC) Measurement-Thermodynamic parameters associated with the reactions between hPDI proteins and Ero1-L␣ were measured on a VP-ITC titration calorimeter (MicroCal, Inc.) at 25°C. A solution of 12 M Ero1-L␣ was loaded into the sample cell (1.43 ml), and a solution of 75-120 M hPDI proteins was placed in the injection syringe (295.7 l). The first injection (5 l) was followed by 29 injections of 10 l. Dilution heats of the hPDI proteins were measured by injecting the hPDI proteins into buffer alone and were subtracted from the experimental curves prior to data analysis. Data were analyzed using MicroCal ORIGIN software supplied with the instrument. All experiments were performed in buffer A. The stirring rate was 300 rpm.

RESULTS
Characterization of Ero1-L␣-After cleavage of the GST tag, the Ero1-L␣ protein was bright yellow; it exhibited a single band of about 55 kDa on reducing SDS-PAGE (supplemental Fig. S1) and two peaks, corresponding to apparent molecular mass of 44 kDa and 100 kDa, respectively, on a Superdex 75 HR 10/30 column, indicating that the preparation is largely composed of monomer with less than 20% dimer (Fig. 1A). Because the content of dimer decreased but did not disappear after DTT treatment, the dimer is mainly a disulfide-linked homodimer with a small fraction of non-covalent homodimer (also see supplemental Fig. S1).
As shown in Fig. 1B, the absorption of Ero1-L␣ preparation exhibits a typical unresolved flavin spectrum with the maximum absorbance at 454 nm with a shoulder at 485 nm and another peak at ϳ370 nm, in addition to the protein peak at 280 nm. Each Ero1-L␣ molecule contains 0.9 FAD determined by using the extinction coefficient of 12.5 mM Ϫ1 cm Ϫ1 at 454 nm (18). Dialysis of Ero1-L␣ against 2 mM EDTA buffer revealed that FAD binds tenaciously to the enzyme. After treatment by 6 M guanidine hydrochloride, the spectrum became like that of free FAD indicated by the shift of the maximum absorbance from 454 nm to ϳ447 nm and disappearance of the shoulder at 485 nm.
As shown in Fig. 1C the absorbance at 454 nm of 10 M Ero1-L␣ started decreasing after 4 min of 2 mM DTT treatment, suggesting the reduction of bound FAD to FADH 2 . The lag corresponds to the time required to remove solution-dissolved oxygen. In contrast, no absorbance change was detected for 10 M free FAD over the course of the experiment. The results indicated that electrons can be transferred from reducing agent to Ero1-L␣-bound FAD but not to free FAD. Thus, Ero1-L␣ contains an equimolar FAD as its prosthetic group, and the bound FAD is redox active.
Oxygen Is Reduced to Form Hydrogen Peroxide during Ero1-L␣-catalyzed Thiol Oxidation-There were conflicting reports on whether hydrogen peroxide was a byproduct of the thiol oxidation driven by yeast Ero1p (18,21). We monitored the oxygen consumption of a system containing Ero1-L␣ and thiol reagents. As shown in Fig. 2A, a drop of 49 M oxygen (after subtracting the background oxygen consumption by a nonenzymatic reaction) was detected during Ero1-L␣-catalyzed oxidation of 50 M reduced E. coli Trx (Trx red ), and addition of catalase at 10 min restored 20 M oxygen in the solution, indicating a production of 40 M hydrogen peroxide ( catalyzed by flavoprotein sulfhydryl oxidases applies also to Ero1-L␣. Table 1 compares the catalytic activities of our recombinant Ero1-L␣ with recombinant Ero1p. GSH is a rather poor substrate of Ero1-L␣, similar to Ero1p (22), while DTT and Trx red are good substrates of Ero1-L␣. The K m for oxygen was determined by observing the rate of reaction as oxygen was depleted from the system (Fig. 2B). The low K m of 5 M for oxygen is similar to that of Ero1p (Table 1), suggesting that Ero1-L␣ can function efficiently at low oxygen tension, minimizing the risk of reductive stress in the ER. Moreover, the turnover number of Ero1-L␣ for oxygen is about 3-fold that of Ero1p.
Effect of Exogenous FAD on the Reaction Catalyzed by Ero1-L␣-Ero1p purified from yeast has been reported to be active only in the presence of exogenous FAD (21,22), however, two C-terminal-truncated Ero1p mutants expressed in E. coli showed activity without addition of FAD (18). We tested the effect of exogenous FAD on the activity of recombinant Ero1-L␣ using oxygen consumption assay. As shown in Fig. 3, the presence of 100 M exogenous FAD did not affect the oxygen consumption rate in Ero1-L␣-catalyzed oxidation of either 50 M Trx red or 10 mM DTT. Also, exogenous FAD was not required for the reactivation of RNase A driven by Ero1-L␣ (see below). Hence the FAD found bound to recombinant Ero1-L␣ is fully functional.
Reconstitution of the Ero1-L␣/hPDI Oxidative Folding Pathway in Vitro-By monitoring the reactivation of denatured and reduced RNase A, which requires the formation of four specific disulfides necessary for its activity, we reconstituted the Ero1-L␣/hPDI system in vitro, and attempted to resolve the isomerase and oxidoreductase activities of hPDI in this system. As shown in Fig. 4A, either Ero1-L␣ or hPDI alone did not promote reactivation of RNase A. The presence of Ero1-L␣ and hPDI together did catalyze reactivation of RNase A, and the reactivation rate increased with increasing concentrations of Ero1-L␣. Exogenous FAD is not required for the reactivation of RNase A in this system. The direct observation of distribution of reduced and oxidized RNase A during reactivation (Fig. 4B) supported the above results. The data show that the reactivation of RNase A results from the oxidative formation of disulfide bonds, which requires both hPDI as catalyst and either Ero1-L␣ or a glutathione buffer to deliver oxidizing equivalents. The transfer of oxidizing equivalents in the system of Ero1-L␣/hPDI is similar to what has been observed in the case of Ero1p (22). We then isolated the monomer Ero1-L␣ fraction by gel filtration and subsequently assayed this monomer fraction in the reconstituted system. The monomer Ero1-L␣ is somewhat more active than the unfractionated preparation (supplemental Fig.  S2), indicating that the monomer fraction contributes the dominating activities in our Ero1-L␣ preparation and the dimer fraction is either inactive or less active than the monomer fraction. However, this does not exclude the possibility that Ero1-L␣ homodimer could have a potential function in vivo, which needs further investigation and is beyond the scope of this research.
The experiments shown in Fig. 4A represent a one-pot assay in which reduced RNase A is oxidized and re-activated, and the active RNase A then hydrolyzes its substrate, giving rise to a change in absorbance; the data shown represent the instantaneous RNase A activity as a function of time (19,23). The data show a lag phase followed by a period in which RNase A activity increases linearly. The lag is interpreted as a period during which reduced RNase A is oxidized to produce a mix of inactive disulfide isomers, while the linear phase represents isomerization of these species to native active RNase A. We have calculated the rate of RNase A reactivation (essentially the isomerase activity of hPDI) by measuring the rate of increase of RNase A activity during the linear phase, and the rate of RNase A oxidation (the oxidoreductase activity of the complete system) by measuring the reciprocal of the lag time before the appearance of active RNase A. As shown in Fig. 4C, increasing concentrations of Ero1-L␣ significantly enhanced the oxidoreductase activity, but did not significantly influence the isomerase activity of hPDI. The isomerase and oxidoreductase activities of hPDI in the presence of the optimal glutathione redox buffer (1 mM GSH/0.2 mM GSSG) are shown in Fig. 4C for comparison, and it is apparent that the isomerase activity is approximately invariant and is basically independent of the nature of the source of oxidizing equivalents.
Ero1-L␣-mediated and hPDI-catalyzed Oxidative Folding Is Strongly Influenced by the Positions of Catalytic Trx Domains in hPDI-To explore the structural and functional requirements for the interaction of hPDI with Ero1-L␣, we examined the FIGURE 2. Oxygen is reduced by Ero1-L␣ to form hydrogen peroxide. A, oxygen consumption was monitored immediately after injection of 1 M Ero1-L␣ into buffer B with (--) or without (⅐⅐⅐⅐) 50 M Trx red . Catalase was added at the indicated time (arrow). B, K m determination for oxygen. Ero1-L␣ at 1 M was injected into buffer B containing 10 mM DTT. The graph shows the end of the reaction, when the oxygen has been depleted to Ͻ10% of its initial value. The inset is the first derivative of the data, used to identify the point, at which the rate was half-maximal.   reactivation of RNase A in the reconstituted system containing Ero1-L␣ and different hPDI domain combinations and mutants (Fig. 5, upper panel). The activities of the various PDI constructs were studied in the presence either of the glutathione redox buffer or of Ero1-L␣. In the optimal glutathione redox buffer, hPDI-⌬C showed similar activities to those of the fulllength hPDI. Those constructs that contain only one intact dithiol active site (hPDI-bbЈxaЈ and hPDI-abbЈx) or one intact site plus one -CGHC-mutated to -SGHS-(hPDIC1 and hPDIC2), all showed similar abilities to catalyze the reactivation of RNase A. Hence, all combinations of one active Trx domain plus the core bbЈx region showed roughly equivalent isomerase and oxidoreductase activities in GSH/GSSG-driven oxidative protein folding. The single domains (hPDI-a and hPDI-aЈc) showed low oxidoreductase activity (Ͻ20%) and very low isomerase activity (Ͻ5%). The mutant full-length PDI with two active sites both mutated to -SGHS-(hPDIC1/2) showed no activity at all and was taken as a negative control.
In the system in which Ero1-L␣ supplied the oxidative equivalents, hPDI-⌬C did not show any inferiority to the full-length hPDI, indicating that the C-terminal tail is not involved in the interaction with Ero1-L␣. However, in this system, the various constructs containing a single redox-active Trx domain were not equivalent in activity, in contrast to the situation with the glutathione system as oxidant. Absence of domain aЈ (hPDI-abbЈx) or mutation of the active site in domain aЈ (hPDIC2) decreased the activity markedly (to Ͻ10% of wild-type). Conversely, hPDI-bbЈxaЈ and hPDIC1 with the active site in domain aЈ intact are both ϳ50% active as wild-type, suggesting that the aЈ domain is responsible for the dominating flow of electrons to Ero1-L␣. Moreover, the single domains (hPDI-a and hPDI-aЈc) did not show any detectable activity in this system.
The difference between domains a and aЈ of hPDI in Ero1-L␣-mediated oxidative folding is not likely a result of their distinct abilities to interact with the RNase A substrate or facilitate RNase A reactivation, since they behave similarly in the optimal glutathione redox buffer (Fig. 5, upper panel). Thus, it is reasonable to hypothesize that the difference arises from their interaction with Ero1-L␣ and may depend either on some intrinsic difference between the domains or on their position, i.e. whether the active Trx domain is located N-terminal or C-terminal of the bbЈx fragment. To distinguish between these factors, we produced a series of DShPDI combinations and mutants (Fig. 5, lower panel), with the a and aЈ domains exchanged positionally. Because DShPDI showed defective activity compared with hPDI we then took the activity of DSh-PDI as 100% for evaluating the activities of all DShPDI proteins. In the optimal glutathione redox buffer, compared with DShP-DIC1/2 with both active sites mutated, DShPDI-bbЈxa and DShPDI-aЈbbЈx were both able to promote RNase A reactivation, and the former is about 3-fold more active than the latter. Similar to the case in un-swapped proteins, addition of an inactive aЈ or a domain in the "swapped position" to one of these constructs (DShPDIC1 and DShPDIC2) does not enhance nor reduce activity further. In Ero1-L␣-mediated oxidative folding, DShPDI-bbЈxa is over 10-fold more active than DShPDI-aЈbbЈx, and again the addition of the missing domain in mutated form has little effect. All the above results strongly suggest that either domain a or aЈ is able to donate electrons to Ero1-L␣, provided that it is in the position to the C-terminal of the bbЈx fragment of hPDI, but operates much less efficiently if in the position to the N-terminal of bbЈx.
bЈxaЈ of hPDI Is the Minimal Element for Binding with Ero1-L␣-We have further investigated the direct interaction of hPDI proteins and Ero1-L␣ by GST-pull-down assay. As shown in Fig. 6, full-length hPDI and hPDI-⌬C strongly interacted with Ero1-L␣, whereas hPDI-bbЈxaЈ and hPDI-bЈxaЈ bound to Ero1-L␣ to a lesser extent, and other mutants did not show interaction with Ero1-L␣, consistent with the result obtained by RNase A refolding assay (Fig. 5). The above results indicated that bЈxaЈ is the minimal element for efficient interaction with Ero1-L␣. In addition, hPDI-bЈxaЈ showed similar enzymatic activities to hPDI-bbЈxaЈ in the Ero1-L␣-mediated oxidative refolding of RNase A (Fig. 5).
ITC was then used to describe quantitatively the interaction between Ero1-L␣ and hPDI (Fig. 7). The binding of hPDI with Ero1-L␣ is an exothermic reaction (Fig. 7A), showing favorable van der Waals, hydrogen bond, and electrostatic interactions (24). For the titration of wild-type hPDI and hPDI-⌬C into Ero1-L␣, the integrated binding isotherm was best fit to a four sequential binding sites model, giving a dissociation constant for the first binding event of 0.4 and 1.1 M, respectively (Fig. 7B). The affinity of hPDI-bbЈxaЈ and hPDI-bЈxaЈ to Ero1-L␣ were significantly lower than those of full-length hPDI and hPDI-⌬C, with a dissociation constant for the first binding event 11.1 and 9.7 M (Fig. 7B) derived from a three sequential binding sites model. Because the following binding events may reflect the weak, nonspecific interactions, corresponding to a multisite low affinity binding of substrates to PDI (25), we just presented here the first binding parameters for comparision. No binding reaction for hPDI-abbЈx, hPDI-bbЈx, or hPDI-aЈc with Ero1-L␣ was detected by ITC. The affinity of hPDI/Ero1-L␣ binding from our data is higher than the binding between ERp57 and the P-domain of calnexin (24) or calreticulin (26), also measured by ITC, suggesting a substantial strong interaction between hPDI and Ero1-L␣.
All the above results strongly supported the idea that the region bЈxaЈ of hPDI is the minimal element for binding with Ero1-L␣, and all domains together fulfill the maximum binding.

DISCUSSION
In this study, we have shown that the recombinant Ero1-L␣ contains equimolar tightly bound FAD, which is non-covalent but released only on denaturation. Different from Ero1p purified from yeast (21,22) but similar to recombinant Ero1p purified from E. coli (18), the bound FAD in recombinant Ero1-L␣ is functionally active, because addition of free FAD does not enhance its activity. Our Ero1-L␣ preparation resembles yeast Ero1p in substrate kinetic properties (Table 1) and in the generation of stoichiometric hydrogen peroxide in the presence of dithiol substrates and oxygen.
We have reconstituted an in vitro oxidative folding pathway with purified recombinant Ero1-L␣ and hPDI, that to our knowledge is the first reconstitution of the folding pathway using mammalian proteins. By using quantitative calculations we have distinguished the isomerization and oxidoreduction processes. The rate of isomerization is independent of the source of oxidizing equivalents, whereas the rate of oxidoreduction is sensitive to the source of oxidation and enhanced with the increase of the concentration of Ero1-L␣. We have shown an asymmetry of the two active sites of hPDI in  the Ero1-L␣-mediated oxidative folding. The C-terminal active site (in domain aЈ) is much more active than the N-terminal active site (in domain a) when Ero1-L␣ is the source of oxidation. The difference of the two active sites is not thought to be due to their intrinsic catalytic properties, since they are reported to have a similar redox potential of Ϫ175 mV (27) and showed similar abilities to re-activate RNase A when GSSG is the source of oxidizing equivalents. If the two catalytic domains were exchanged positionally, the active site in domain a became more active than the one in domain aЈ. Pull-down and ITC assays confirmed that bЈxaЈ is critical for binding with Ero1-L␣ as well as for functional electron transfer. Thus, we attributed this asymmetry to the specific order of the two catalytic domains in the context of the fulllength hPDI. In other words, the principal Ero1-L␣-binding site bЈxaЈ on hPDI supplies the structural basis for the preference of Ero1-L␣ to oxidize the active site next to the C-terminal of bbЈx (Fig. 8).
Although a similar asymmetric effect was observed in terms of the oxidation rate of the two active sites in yeast PDI (yPDI) by Ero1p (28), it was claimed that the bЈ domain in yPDI is not required for recognition by Ero1p, in contrast to the necessary role of the bЈ domain of hPDI with Ero1-L␣. This claim is rather surprising, since the bЈ domain of hPDI has been characterized to contain the principal substrate binding site (5,29), and a hydrophobic pocket in yPDI bЈ was also identified in the crystal structure (4). In hPDI the bЈ domain is sufficient for the binding of short peptides and essential for the binding of more complex substrates, while the a and aЈ domains also contribute to the binding of larger substrates (29). It was also reported that the minimum requirement for assembly of collagen prolyl 4-hydroxylase tetramer is fulfilled by domains bЈ and aЈ of hPDI, and the presence of the a and b domains greatly enhances assembly (30). Our recent study on ERp44, a PDI family member, has shown that a hydrophobic pocket in domain bЈ and a hydrophobic patch in domain a could be potential binding sites for client proteins, including Ero1-L␣ (20). All the evidence above strongly support our finding that the bЈxaЈ region is the minimal element for binding with Ero1-L␣, and all domains together fulfill the maximum binding. One consequence of this conclusion is that both Ero1-L␣ and substrate proteins share a common binding site on PDI and hence that PDI must interact sequentially with these partners when transferring oxidizing equivalents from the oxidase to a reduced protein substrate (rather than interacting with both in a ternary complex of Ero1-L␣/PDI/reduced protein).
A basic question in the oxidative folding in the ER is the pathways or networks of electron flows. In prokaryotic periplasm, the oxidation (operated by the DsbA/DsbB system) and the reduction/isomerization (operated by the DsbC/DsbD system) pathways are kinetically separated (31). In eukaryotic ER, glutathione (GSH/GSSG) has long been considered to provide the principal redox buffer implicated in the formation of native protein-disulfide bonds (32). The ratio of GSH/GSSG in the secretory pathway is 1:1-3:1, far more oxidizing than the cytosolic ratio of 30:1-100:1 (33,34) and similar to the optimum GSH/GSSG ratio of 5:1 for in vitro folding of disulfide bond-containing proteins (19). However, the discovery of Ero1 as the essential oxidase relaying oxidizing equivalents from molecular oxygen for the formation of protein disulfide bonds (9, 10) raised questions on the role of GSH/GSSG in oxidative folding in the ER. In this work, we clearly showed that the oxidative protein folding process can be facilitated by Ero1-L␣ and hPDI, independent of any bulk redox buffer. However, these data cannot exclude the possibility that GSH/GSSG functions FIGURE 7. ITC data of the hPDI proteins binding to Ero1-L␣. A, upper panel, heat released per second during the addition of 90 M full-length hPDI into the ITC cell containing 12 M Ero1-L␣. Lower panel, squares represent the integrated binding isotherm at each injection after correction for dilution heat effects and normalization for the molar concentration. Solid lines represent the non-linear least squares fit to a four sequential binding sites model. B, ITC thermodynamic parameters of hPDI proteins binding to Ero1-L␣. *, data are for the first binding event with fitting to a four sequential binding sites model. The data in parentheses are based on a single site model. **, data are for the first binding event with fitting to a three sequential binding sites model. NB, no binding observed. FIGURE 8. Schematic model illustrating the interaction between Ero1-L␣ and hPDI. The four Trx domains of hPDI are arranged in a "U" shape, with the two active sites in a and aЈ domains facing each other, according to the crystal structure of yPDI (PDB code: 2B5E) and molecular modeling (SWISS-MODEL) (data not shown). The bЈxaЈ fragment of hPDI provides the essential binding for Ero1-L␣. The Cys 94 -Cys 99 cysteine pair of Ero1-L␣ (corresponding to Cys 100 -Cys 105 in Ero1p) faces the active site in the aЈ domain of hPDI.
as a redox buffer in the ER with GSSG providing oxidizing equivalents to PDI or directly to protein substrates in some circumstances, while GSH is responsible for the reduction of non-native disulfide bonds directly or indirectly in the ER (32), The absence of such a reducing function in the reconstituted Ero1-L␣/hPDI oxidative protein folding system may explain why this system is less active in reactivation of reduced RNase A than the simpler system of PDI plus GSH/GSSG (Fig. 4).
Another intriguing question is how is the redox homeostasis in the ER maintained under various conditions? Recently, a feedback mechanism was proposed to regulate the homeostasis in the ER through modulation of Ero1p activity by changing the redox state of the regulatory cysteine pairs (35). Our results from Ero1-L␣ together with previous report on Ero1p (18,22,35) suggest that Ero1 can oxidize various vicinal dithiols such as DTT, Trx red , and PDI but not monothiols (GSH) or multi-thiols (RNase A) where the thiols cannot form a stable vicinal disulfide. This property of Ero1 could be very important for maintaining the redox homeostasis in the ER by, for example: 1) effectively eliminating hyperreducing agents, such as DTT, thus protecting cells from reductive stress; 2) passing oxidizing equivalents to protein substrates via PDI, an ER oxidoreductase with proofreading (isomerase) activity, rather than directly introducing non-native disulfides into substrates; 3) avoiding the futile consumption of reducing equivalents in GSH, which can itself reduce efficiently non-native disulfide bonds as well as reactive oxygen species generated by Ero1 activity.