Characterization of the endoplasmic reticulum – resident peroxidases GPx7 and GPx8 shows the higher oxidative activity of GPx7 and its linkage to oxidative protein folding

Oxidative protein folding occurs primarily in the mammalian endoplasmic reticulum, enabled by a diverse network compris-ing more than 20 members of the protein disulfide isomerase (PDI) family and more than five PDI oxidases. Although the canonical disulfide bond formation pathway involving Ero1 a and PDI has been well-studied so far, the physiological roles of the newly identified PDI oxidases, glutathione peroxidase-7 (GPx7) and -8 (GPx8), are only poorly understood. We here demonstrated that human GPx7 has much higher reactivity with H 2 O 2 and hence greater PDI oxidation activity than human GPx8. The high reactivity of GPx7 is due to the presence of a catalytic tetrad at the redox-active site, which stabilizes the sulfenylated species generated upon the reaction with H 2 O 2 . Although it was previously postulated that GPx7 catalysis involved a highly reactive peroxidatic cysteine that can be sulfenylated by H 2 O 2 , we revealed that a resolving cysteine instead regulates the PDI oxidation activity of GPx7. We also determined that GPx7 formed complexes preferentially with PDI and P5 in H 2 O 2 -treated cells. Altogether, these results suggest that human GPx7 functions as an H 2 O 2 -dependent PDI oxidase in cells, whereas PDI oxidation may not be the central physiological role of human GPx8.

Ero1a and PDI play central role in the canonical pathway of disulfide bond formation at the cost of one molecular oxygen per disulfide bond, resulting in the generation of H 2 O 2 as a byproduct (2,6). Overproduction of H 2 O 2 as a consequence of the Ero1a overwork causes oxidative stress in the ER, eventually leading to cell death. Therefore, the oxidative activity of Ero1a needs to be tightly controlled (20). In the mechanism of operation of mammalian Ero1a, four cysteine residues in the electron shuttle loop act also as components of a regulatory switch, which ensures strict regulation over Ero1a activity (8,9,(20)(21)(22)(23)). An ER-resident peroxiredoxin, Prx4, metabolizes H 2 O 2 and oxidizes PDI family enzymes upon return to the reduced state (24,25). Our prior in vitro studies showed that ERp46 and P5, with their partner oxidase Prx4, are engaged in the rapid but promiscuous disulfide bond introduction in the initial phase of oxidative protein folding (10), whereas PDI, in concert with Ero1a, efficiently catalyzes the correction of non-native disulfide bonds and the subsequent selective formation of native disulfide bonds (26). Other oxidative pathways involving vitamin K epoxide reductase were also reported to potentially operate in the PDI reoxidation cycle (27,28).
More recently, glutathione peroxidase-7 (GPx7) and -8 (GPx8) were identified as ER-resident PDI oxidases using H 2 O 2 as a source of oxidative power (29)(30)(31). GPx family was historically named after its first identified member that catalyzes the reaction of peroxide with reduced glutathione (GSH) on its selenocysteine residue (32). Despite the absence of a GSHbinding motif and the substitution of the selenocysteine residue by Cys in GPx7 and GPx8, both the enzymes are capable of acting as peroxidases, which reduce H 2 O 2 to water (29,31,33). GPx7-deficient cells were found to accumulate endogenous reactive oxygen species, lowering cellular viability (34,35). Another study demonstrated that GPx7 plays a role in the alleviation of oxidative stress in breast cancer cells, esophageal cells, and adipose tissue, emphasizing the importance of GPx7 in the oxidative stress response (35)(36)(37)(38). Both GPx7 and GPx8 were suggested to interact with Ero1a in vivo and possibly modulate the peroxide-generating activity of Ero1a by scavenging H 2 O 2 in its proximity (29). Additionally, the Ero1a-GPx8 ‡ These authors contributed equally to this work. * For correspondence: Kenji Inaba, kenji.inaba.a1@tohoku.ac.jp. complex may serve to prevent the diffusion of H 2 O 2 from the ER (39).
GPx7 and GPx8 commonly possess an essential peroxidatic cysteine residue (C P ) in the typical motif, NVASX(C/U)G (Cys 57 in GPx7 and Cys 79 in GPx8), and a resolving cysteine residue (C R ) in the highly conserved motif of GPx family, FPCNQF (Cys 86 in GPx7 and Cys 108 in GPx8) (40,41). Whereas GPx7 and GPx8 share a highly similar overall structure, their topologies are distinct in the following points: GPx7 is purely a luminal protein, whereas GPx8 has a small N-terminal cytoplasmic domain, a short transmembrane domain, and a catalytic-active domain facing the lumen (named the "luminal domain" in this work) (33). Although some structural and molecular insights have been gained into GPx7 and GPx8, the mechanistic details of GPx7/8-mediated PDI oxidation and their physiological roles in cells, especially their involvement in oxidative protein folding, still remain unclear.
In this study, we investigated the mechanistic basis of human GPx7-and GPx8-catalyzed PDI oxidation and identified their preferential substrates among PDI family members both in vitro and in vivo. We thus demonstrated that GPx7 was a much more efficient PDI oxidase than GPx8 because of even higher susceptibility of C P to H 2 O 2 , in which Gln 92 contained in the catalytic tetrad plays a critical role. Although the GPx7 catalysis was known to involve the C P residue (31), we found that the C R residue acts to regulate the PDI oxidation activity of GPx7. Our extensive studies also revealed that GPx7 bound PDI and P5 in H 2 O 2 -treated cells. Thus, human GPx7 is a potent PDI oxidase, whereas the role of human GPx8 seems largely distinct from that of GPx7 despite their structural similarity.

Results
Higher reactivity with H 2 O 2 and PDI of GPx7 than that of GPx8 Although both GPx7 and GPx8 are presumed to work as PDI-dependent peroxidases (29), the reactivity of these two enzymes with H 2 O 2 have not been assessed precisely. To compare their peroxidase activities, we first mixed the fully reduced form of GPx7 or the luminal domain of GPx8 (1 mM each) with different concentrations (10, 50, and 200 mM) of H 2 O 2 and followed their redox state changes (Fig. 1). The aliquot samples were treated with 1 mM maleimidyl PEG-2000 (mal-PEG 2k) for quenching at the indicated time points. Although the topmost and bottom bands are identifiable as reduced and oxidized species, respectively, the intermediate band is most likely a hyperoxidized species (SO 2 H or SO 3 H form), in which C R is mal-PEG modifiable, but C P is no longer alkylated because of the hyperoxidation with H 2 O 2 . In support of this, the intermediate bands significantly increased at higher H 2 O 2 concentrations as observed particularly with GPx8 ( Fig. 1, right panels).
The present assay revealed much higher reactivity or affinity of GPx7 for H 2 O 2 than that of GPx8; 10 mM H 2 O 2 was sufficient to fully oxidize 1 mM GPx7 within 15 s after mixing, whereas most of the GPx8 remained reduced even after 5 min in the presence of 10 mM H 2 O 2 . Higher concentrations of H 2 O 2 significantly accelerated GPx8 oxidation concomitant with increased generation of the hyperoxidized species, suggesting that the K d for H 2 O 2 value of GPx8 is much above 10 mM.
To visualize H 2 O 2 -driven PDI oxidation by GPx7 and the luminal domain of GPx8, reduced PDI (10 mM) was reacted with GPx7 or GPx8 (1 mM each) in the presence of 10, 50, or 200 mM of H 2 O 2 , and the redox-state change of PDI was monitored by mal-PEG 2k modification of free cysteines followed by nonreducing SDS-PAGE ( Fig. 2A). Consequently, GPx7 generated significant amounts of partially and fully oxidized PDI at early time points, whereas GPx8 hardly accelerated PDI oxidation compared with H 2 O 2 alone. This result is in line with the present observation that GPx7 was oxidized by H 2 O 2 much more efficiently than GPx8 (Fig. 1). Of note, GPx7-mediated PDI oxidation was slower at 200 mM H 2 O 2 than at 50 mM H 2 O 2 . This is likely explained by the greater generation of a hyperoxidized GPx7 species during the catalysis of PDI oxidation in the presence of higher concentrations of H 2 O 2 . In support of this notion, most of GPx7 was converted to a hyperoxidized species during the catalysis of PDI oxidation at 200 mM H 2 O 2 (Fig. 2B). The generation of a hyperoxidized species of GPx7 was greater in the presence of reduced PDI ( Fig. 2B versus Fig. 1, left, middle, and bottom panel). This observation suggests that the catalytic cycle significantly increased the hyperoxidized species via redox reactions between H 2 O 2 and reduced GPx7 generated by oxidizing PDI every turnover.
A pK a value, which determines the degree of protonation at a given pH, represents a fundamental character of protein thiol groups (42,43). To examine whether the pK a values of GPx7 and GPx8 can explain their different peroxidatic activities, we analyzed crystal structures of the enzymes (Protein Data Bank codes 2P31 for GPx7 and 3CYN for GPx8) using the program PROPKA3. The in silico analysis yielded the pK a values of peroxidatic cysteines (C P ) of GPx7 and GPx8 to be 9.3 and 11.5, respectively. Thus, C P in GPx8 is predicted to have exceptionally low reactivity compared with C P in GPx7 and free cysteine in solution (pK a of 8.3). Altogether, the results tempt us to suspect a functional role of GPx8 as a H 2 O 2 -scavenging peroxidase in the ER.

Essential role of a Gln residue in the catalytic tetrad of GPx7
Effective thiol-based peroxidases commonly contain a catalytic triad or tetrad including a highly reactive, C P (40,41). Indeed, GPx7 contains a catalytic tetrad consisting of Cys 57 , Trp 142 , Asn 143 , and Gln 92 at the active site, where Gln 92 is predicted to stabilize thiolated or sulfenylated Cys 57 through a hydrogen bond upon reaction with H 2 O 2 (40). Although GPx8 also conserves the first three residues, Cys 79 , Trp 164 , and Asn 165 , at the positions corresponding to Cys 57 , Trp 142 , and Asn 143 in GPx7, respectively, we surmised that the replacement of Gln 92 (in GPx7) to Ser 114 (in GPx8) (Fig. 3, A and B) was a primary reason for much lower peroxidase activity of GPx8 than that of GPx7. In this context, the distance between the S g atom of GPx7 Cys 57 and O e1 of Gln 92 is 3.1 Å, close enough to stabilize the thiolated GPx7 C P via a hydrogen bond to the O e1 ; meanwhile the S g atom of GPx8 Cys 79 is separated from the O g atom of Ser 114 by 4.8 Å, far beyond a hydrogen-bonding distance (Fig. 3B).
To explore the importance of the C P -neighboring residues in the different peroxidatic activities of GPx7 and GPx8, we constructed swap mutants where Gln is mutated to Ser in GPx7 (GPx7 Q92S) and vice versa in the luminal domain of GPx8 (GPx8 S114Q). In the result, the GPx7 swap mutation greatly compromised the GPx7 reactivity with H 2 O 2 (Fig. 3C, left panels), leading to the abolishment of PDI oxidation activity (Fig. 3D, left panel). Notably, GPx7 Q92S exhibited a similar rate in H 2 O 2 -dependent oxidation as GPx8, although the latter generated larger amount of hyperoxidized species than the former at both 10 and 50 mM H 2 O 2 (Figs. 1 and 3C, left panel). It is also noteworthy that GPx8 S114Q greatly increased the GPx8 reactivity with H 2 O 2 , leading to more efficient PDI oxidation in the presence of 10 mM Even without reduced PDI, GPx8 S114Q indeed generated a considerable amount of hyperoxidized species upon reaction with 50 mM H 2 O 2 (Fig. 3C, lower right panel). Collectively, we conclude that Gln 92 acts as an essential residue for high reactivity of GPx7 with H 2 O 2 and hence effective oxidation of PDI by the enzyme.
GPx7 Cys 86 plays a critical role in the regulated oxidative activity of GPx7 Three-dimensional structures of reduced forms of GPx7 and GPx8 have high similarity, in which the locations and orientations of C P and C R in GPx7 are almost identical to those in GPx8 (29). The S g atoms of C P and C R are separated by ;11 Å in both the enzymes, suggesting that a large conformational change is required for formation of an intramolecular disulfide bond between these two cysteines (30). Given the situations, two alternative mechanisms have been proposed for GPx7/ GPx8-mediated PDI oxidation: a one-cysteine mechanism in which the C P acts as the sole redox center that reacts with both H 2 O 2 and PDI and a two-cysteine mechanism in which an intramolecular disulfide bond between C P and C R is formed preceding the oxidation of PDI (30,33).
To investigate which of the mechanisms is primarily exerted by GPx7 and GPx8, we constructed single-Cys mutants of the enzymes in which C R was mutated to alanine: C86A for GPx7 and C108A for GPx8. Of interest, the results demonstrated that GPx7 C86A oxidized PDI more efficiently than GPx7 WT (Figs. 2A, left middle panel, and 4A, left panel), suggesting that a onecysteine mechanism is more competent in PDI oxidation than a two-cysteine mechanism and that GPx7 WT exerts primarily a two-cysteine mechanism or both one-and two-cysteine. By contrast, GPx8 C108A oxidized PDI even less efficiently than GPx8 WT (Figs. 2A, middle right panel, and 4A, right panel), indicating the necessity of C R for the GPx8-mediated PDI oxidation. Thus, GPx8 is likely to utilize a two-cysteine mechanism exclusively.
For GPx7 to exert a one-cysteine mechanism with high efficiency, its C P must be very reactive with H 2 O 2 . To verify that it is indeed the case, we investigated the susceptibility of GPx7 C86A to H 2 O 2 -mediated hyperoxidation. As expected, GPx7 C86A was rapidly hyperoxidized with 50 mM H 2 O 2 , whereas its counterpart, GPx8 C108A, was much less sensitive to H 2 O 2 (Fig. 4B). To further compare the oxidative activities of GPx7 WT and C86A, we investigated oxidative folding of RNase A catalyzed by these two using gel-shift assay (Fig. 4C). The results demonstrated that GPx7 C86A oxidized RNase A much more rapidly than GPx7 WT. Accordingly, GPx7 C86A restored RNase A activity more quickly than GPx7 WT (Fig. 4D). Thus, GPx7 can be a more competent H 2 O 2 -dependent oxidase by exerting a one-cysteine mechanism exclusively on the highly reactive peroxidatic cysteine.
To explore whether GPx7 WT primarily employs a one-or two-cysteine mechanism, we examined the PDI concentration dependence of the rate of PDI oxidation by GPx7 WT and C86A by performing the NADPH consumption assay. We predicted that, at higher PDI concentrations, a one-cysteine mechanism gets more predominant over a two-cysteine mechanism because, under such conditions, reduced PDI should more likely win the race of nucleophilic attack to a sulfenylated C P against the C R residue of the same GPx7 molecule (Fig. 5A). Given that, the oxidative activity of GPx7 WT will approach that of GPx7 C86A as the concentration of PDI increases. However, higher PDI concentrations did not render the initial NADPH consumption rates of GPx7 WT and C86A closer to each other (Fig. 5, B and C). It is thus conceivable that, regardless of the PDI concentration, GPx7 WT primarily employs a two-cysteine mechanism, resulting in the slower oxidation of PDI and hence the slower oxidative folding of a downstream substrate than GPx7 C86A (Fig. 4).

Preferential oxidation substrate of GPx7 among PDI family members
Previous studies by us and others suggest that PDI oxidases have high selectivity against the PDI family members in the disulfide bond formation network (10, 12, 21, 29, 44, 45). The present study indicated that whereas the PDI oxidation activity of GPx8 is minimal, GPx7 has significant PDI oxidation activity. To investigate GPx7 preference for the PDI family members, five representative members of the PDI family (PDI, ERp46, P5, ERp57, and ERp72), either independently or in a mixture of five, were subjected to oxidation by GPx7 in the presence of H 2 O 2 , and the time-dependent changes of their redox states were monitored by immunoblotting with an antibody against each PDI (Fig. 6A). The results of GPx7-catalyzed single-PDI oxidation demonstrated that the enzyme was a versatile oxidase to all PDI family members tested (Fig. 6, A and B, left panels). By contrast, when GPx7 was reacted with the mixture of PDIs, ERp46 and ERp72 were oxidized in greatest amount compared with the other three PDIs (Fig. 6, A and B, right panels), sug-gesting that GPx7 has significant preference for these two members. Interestingly, even though P5 seemed a good substrate of GPx7 in the single-PDI oxidation assay, P5 remained fully reduced in the mixture of PDIs, probably because of the competitive inhibition by the other PDI family members. Unlike P5, ERp46 and ERp72 were significantly oxidized by GPx7 even in presence of the other representative PDIs (Fig.  6, A and B, right panels).

H 2 O 2 -induced complex formation between PDIs and GPx7/8 in cells
To gain insights into physiological redox partners of GPx7 and GPx8, we next transfected HeLa cells with plasmids expressing FLAG-tagged Prx4, GPx7, or full-length GPx8; treated the cells with or without 0.5 mM H 2 O 2 for 10 min; and alkylated the free cysteines with N-maleimide (NEM) as described under "Experimental procedures," followed by immunoprecipitation and immunoblotting with an antibody against each PDI family member (Fig. 7). We thus reproduced our previous observation that Prx4 most preferentially bound ERp46 and P5 in an H 2 O 2 -independent manner in cells (10). Notably, GPx7 bound PDI only slightly and did not bind the other PDI family members without H 2 O 2 (Fig. 7). However, H 2 O 2 treatment significantly increased amounts of PDI and P5 bound to GPx7. A similar observation was made with GPx8 despite even smaller amounts of PDI and P5 co-immunoprecipitated with GPx8 (Fig.  7). Thus, GPx7 displayed PDI preference different from Prx4, Figure 5. Effects of PDI concentration on the oxidative activities of GPx7 WT and C86A. A, alternative (one-or two-cysteine) mechanism of GPx7 in the catalysis of PDI oxidation. A peroxidatic cysteine (C P ) of GPx7 is readily sulfenylated upon reaction with H 2 O 2 , followed by the race between reduced PDI and a resolving cysteine (C R ) of the same GPx7 molecule against the sulfenylated C P . B, NADPH consumption coupled to PDI oxidation catalyzed by GPx7 WT or GPx7 C86A was monitored by measuring absorbance change at 340 nm. All experiments were initiated at 30°C by mixing 50 mM H 2 O 2 into the mixture of reduced PDI (5-50 mM), GSH (1 mM), GR (1 unit), NADPH (200 mM), and GPx7 WT or GPx7 C86A (1 mM) in buffer containing 50 mM Tris/HCl (pH 7.4) and 300 mM NaCl. Representative data of three experiments at each condition are displayed. C, plots of the initial NADPH consumption rate (mM/min) during the catalysis of GPx7 WT or GPx7 C86A as a function of PDI concentrations. Note that the rates were calculated by subtracting the rate of noncatalyzed reaction (No GPx7) from those of GPx7-catalyzed reactions (GPx7 WT or C86A) (n = 3, means = 6S.D.).
although both the peroxidases bound their preferential PDIs more tightly in the presence of H 2 O 2 . Notably, the intracellular redox partners of GPx7 did not coincide with its preferential substrates identified by in vitro experiments (Fig. 6), suggesting the presence of mediator proteins that link GPx7 to its physiological redox partners (see also "Discussion").

Discussion
Previously, GPx7 and GPx8 were demonstrated to be capable of utilizing Ero1a-derived H 2 O 2 in cells (29). Although GPx8 was reported to interact directly with Ero1a and scavenge the neighboring H 2 O 2 molecule (39), the present work revealed that GPx7 had much higher reactivity with H 2 O 2 than GPx8, leading to the more efficient PDI oxidation. Although the affinity for H 2 O 2 of GPx7 was not precisely determined in this study, 10 mM H 2 O 2 was sufficient to oxidize GPx7 rapidly (Fig. 1). It is thus conceivable that GPx7 could compete with Prx4, another ER-resident peroxidase with a micromolar range of affinity for H 2 O 2 , for reacting with H 2 O 2 if these two enzymes co-localize in the same compartment of the ER. However, considering that unlike GPx7, Prx4 does not scavenge Ero1a-generated H 2 O 2 efficiently in cells (39), these two enzymes seem likely to have distinct physiological roles in the ER.
Reactivity of a given cysteine residue depends on the propensity of its thiolate form, which is quantitatively expressed as pK a (46). However, pK a is not the sole factor that determines the reactivity of a peroxidatically active cysteine with H 2 O 2 (43,47). The active sites of GPx7 and GPx8 consist of a catalytic tetrad including Cys, Trp, Asn, and Gln/Ser residues. We here demonstrated that the Gln 92 residue in the catalytic tetrad of GPx7 is essential for the enhanced reactivity of GPx7 C P with H 2 O 2 , probably because Gln 92 serves to stabilize the sulfenylated C P species via hydrogen bonding (Figs. 3 and 4B). Gln 92 is highly conserved among all GPx family members but GPx8, in which Gln is substituted by Ser 114 . In this regard, Ser 114 can be interpreted to have an inhibitory role in the H 2 O 2 -scavenging activity of GPx8 (39), possibly providing GPx8 with other physiological functions in cells.
The present study also demonstrates that GPx7 WT preferentially employs a two-cysteine mechanism in oxidizing PDI, whereas the GPx7 C86A mutant oxidizes PDI at a much higher   1, 2, 9, and 10), pCMV-Gg PRX4 (encoding PRX4-FLAG) (10) (lanes 3, 4, 11, and 12), pES104 (encoding GPx7-FLAG) (lanes 5, 6, 13, and 14), or pES105 (encoding full-length GPx8-FLAG) (lanes 7, 8, 15, and 16), grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum for 24 h, treated with 0.5 mM H 2 O 2 (lanes 9-16) for 10 min or not (lanes 1-8), washed twice with PBS, treated with 10% TCA, and subjected to alkylation with NEM (51). The NEM-treated cell lysates were subjected to immunoprecipitation with anti-FLAG M2 magnetic beads (Sigma) to purify complexes containing FLAG-tagged proteins. The immunoprecipitates were separated by reducing SDS-PAGE, and PDI family members and FLAGtagged constructs contained in the immunoprecipitates were detected with antibodies to the indicated PDI family members or horseradish peroxidase-conjugated antibody to FLAG. Lanes 1, 3, 5, 7, 9, 11, 13,  rate via a one-cysteine mechanism. GPx7 C P (Cys57) is oxidized by H 2 O 2 to generate sulfenylated C P , which can in turn react with either reduced PDI or the C R residue of GPx7 itself. The predominant usage of a two-cysteine mechanism by GPx7 suggests faster nucleophilic attack by the C R residue of the same molecule than by the active-site cysteines of PDI (Fig. 6A). Obviously, two-cysteine mechanism drives a much slower catalytic cycle than one-cysteine. This observation suggests that the intermolecular disulfide transfer from a Cys 57 -Cys 86 pair of GPx7 to the PDI active sites is a rate-limiting step of the GPx7-mediated PDI oxidation cycle. In other words, the presence of a C R residue serves to regulate the oxidative activity of GPx7. In this context, the selection of oneor two-cysteine mechanism by GPx7 could be an effective strategy to maintain the redox homeostasis in the ER, as is the similar case with Ero1a, which tightly modulates its oxidative activity through the disulfide bond rearrangement among the four regulatory cysteines in response to the redox environment in the ER (7-9, 23). However, the higher concentrations of reduced PDI did not greatly enhance the oxidative activity of GPx7 WT (Fig. 5), making it unlikely that the switching from a two-cysteine to one-cysteine mechanism is induced in GPx7 when the ER environment becomes more reducing. Another possible physiological meaning of the twocysteine mechanism is to prevent overoxidation of the peroxidatic Cys in GPx7, thereby keeping the enzyme active. Collectively, the two-cysteine mechanism of GPx7 is likely to play significant roles in maintaining the ER in the moderately oxidizing environment.
In the previous study, Ero1a was shown to preferentially oxidize PDI, and to a lesser extent ERp46, whereas Prx4 preferential oxidation substrates were P5 and ERp46 (10,12,44). Here we demonstrated that whereas GPx7 preferentially oxidized ERp46 and ERp72 in vitro, the enzyme bound PDI and P5 in H 2 O 2 -treated cells. The complex formation between GPx7 and P5 in cells suggests that some third-party proteins may mediate the interaction between these two. In agreement with this notion, GPx7 was reported to form a disulfidelinked complex with BiP, an ER-resident chaperone under oxidative stress conditions (35), whereas P5 targets BiP client proteins (48). Moreover, the preceding work reported that large ER-localized multiprotein complexes including various molecular chaperones such as BiP, GRp94, PDI, ERdj3, Hsp40, cyclophilin B, ERp72, GRP170, UDP-glucosyltransferase, and SDF2-L1 are formed in the ER (49). GPx7 may be one of the as-yet-unidentified components contained in this multiprotein complex, although further extensive studies are required to clarify the diverse interaction network involving GPx7 in cells.
In conclusion, the present work demonstrates that whereas GPx7 is an efficient PDI peroxidase, its phylogenic cousin GPx8 does not possess an efficient catalytic activity. GPx7 and GPx8 are similar in overall structure, such that they could share the same H 2 O 2 -scavenging mechanism involving their catalytic cysteine pairs. However, the different local environments near the cysteines give rise to the strikingly different reactivity with H 2 O 2 of these two peroxidases. Although these two enzymes appear to fulfill distinct physiological roles in cells, it will be interesting to see whether GPx8 can function as a backup enzyme of GPx7 under some special situations such as a highly reducing redox environment or a highly H 2 O 2 abundant condition in the ER.

Experimental procedures Plasmids
Plasmids for overexpression of GPx7 and the luminal domain of GPx8 in Escherichia coli (pKEHS780 and pVD54, respectively) (29) were kind gifts from Dr. Lloyd Ruddock (University of Oulu). GPx7 and GPx8 Cys-Ala mutants were constructed by using the QuikChange site-directed mutagenesis kit (Agilent Technologies).
Primers and templates used to construct following plasmids are listed in Table 1. Plasmid pES101 (encoding GPx7) was constructed by assembling three DNA fragments using the Gibson assembly kit (New England Biolabs). The DNA fragments used were a fragment amplified from pcDNA3.1 1 (Invitrogen) with primers NeoRH1 and gpx7E1, a fragment amplified from pKEHS780 with primers gpx7E2 and gpx7E4, and a fragment amplified from pcDNA3.1 1 with primers NeoRH2 and gpx7E3. Plasmid pES102 that expresses the luminal domain of GPx8 in the ER and plasmid pES103 (encoding GPx8) were constructed by assembling fragments amplified using a template and a primer set indicated on Table 2. Plasmids pES104 (encoding GPx7-FLAG on pcDNA3.1 1 ) and pES105 (encoding GPx8-FLAG on pcDNA3.1 1 ) were constructed by inserting a triple FLAG sequence in front of the KDEL sequence of these proteins. They were constructed also by assembling fragments amplified using a template and a primer set indicated on Table 2.

Protein overexpression and purification
Plasmids for overexpression of GPx7, the luminal domain of GPx8, PDIs, and their mutants were transformed into E. coli BL21(DE3). Cells harboring the plasmids were grown at 30°C and induced by the addition of 0.5 mM isopropyl b-D-thiogalactopyranoside at an A 600 of 0.5. The cells were cultured for an additional 4 h at 30°C and then harvested. To purify GPx7, GPx8, and their mutants, cells in buffer A (50 mM Tris/HCl, pH 8.1, 0.3 M NaCl, and 1 mM phenylmethylsulfonyl fluoride) were disrupted using a microfluidizer (Niro Soavi PA2K). After centrifugation of the cell lysate to remove cellular components at 12,000 rpm for 20 min, the supernatant was loaded onto an open nickel-nitrilotriacetic acid-Sepharose column (Qiagen). The column was washed with buffer A containing 20 mM imidazole, and the proteins were eluted with buffer A containing 200 mM imidazole. The eluted sample was concentrated to 500 ml by filtration using Amicon Ultra filter units (molecular weight cutoff, 10,000; Millipore), applied to a Superdex 75 sizeexclusion column (GE Healthcare) pre-equilibrated with 50 mM Tris/HCl (pH 8.1), 0.3 M NaCl, and 1 mM EDTA and finally eluted with the same buffer. To

Antibodies, reagent, and immunological techniques
Rabbit antibodies to ERp57 (GTX100297), and to ERp72 (GTX115263) were purchased from GeneTex. For the detection of proteins fused with a triple FLAG tag, mouse anti-FLAG M2 antibody conjugated with horseradish peroxidase (Sigma A8592) was used. Rabbit antibody to PDI has been published (50). Antibodies to ERp46 and P5 were used as described previously (10). Immunoloblotting was performed using standard methods. The protein bands were visualized using an appropriate secondary antibody, Clarity Western ECL substrate (Bio-Rad) and ChemiDoc touch imaging system (Bio-Rad). The images were processed with Image laboratory software (Bio-Rad).
Analysis of the redox states of GPx7, GPx8, PDI, and their mutants Purified GPx7, the luminal domain of GPx8, PDIs, and their mutants were separately reduced with 10 mM DTT for 10 min at 30°C followed by DTT removal through a PD-10 column (GE Healthcare) pre-equilibrated with 50 mM Tris/HCl (pH 7.4) buffer containing 300 mM NaCl. To analyze the redox status of GPx7, GPx8, and their mutants, different concentrations (10, 50, and 200 mM) of H 2 O 2 were mixed with 1 mM reduced GPx7, GPx8, or their mutants in degassed buffer containing 50 mM Tris/HCl (pH 7.4) and 300 mM NaCl at 30°C. For PDI oxidation assays, 10 mM reduced PDI was incubated with 1 mM GPx7, GPx8, or their mutants, and reactions were initiated by adding 10, 50, or 200 mM of H 2 O 2 . At the indicated time points, the samples were quenched using 1 mM mal-PEG 2000 (NOF Corporation). The reaction mixture was boiled for 3 min after addition of an equal volume of 23 Laemmli buffer. All of the samples were run through nonreducing SDS-PAGE followed by staining with Coomassie Brilliant Blue (CBB) (Nacalai Tesque). The gel images were captured by using ChemiDoc touch imaging system (Bio-Rad).
Oxidative folding assay of RNase A RNase A from bovine pancreas (Sigma-Aldrich) was reduced and denatured with 100 mM DTT and 6 K guanidinium hydrochloride. The sample was passed through a PD-10 column (GE Healthcare) pre-equilibrated with 10 mM HCl aq. to remove the reducing and denaturing reagents. Reduced/denatured RNase A was diluted to 20 mM (;50-fold dilution) and incubated with 1 mM GPx7 or its mutant and 10 mM PDI in a buffer (50 mM Tris/ HCl, pH 7.4, and 300 mM NaCl) at 30°C, and the reactions were initiated by adding 50 mM of H 2 O 2 . At the indicated time points, the samples were quenched using 1 mM 4-acetamide-49-maleimidylstilbene-2,29-disulfonic acid (AMS; Molecular Probes, Inc.). All samples were separated by nonreducing SDS-PAGE followed by staining with CBB. The gel images were captured by the ChemiDoc touch imaging system (Bio-Rad). The recovery of RNase A activity was measured by monitoring the linear increase in absorbance at 295 nm on a U-3900 spectrophotometer (Hitachi) at 30°C after addition of 1.25 mM cytidine 29,39-cyclic monophosphate monosodium salt (Sigma-Aldrich) to reaction mixtures at the indicated time points.

NADPH consumption assay
NADPH consumption coupled with GPx7-catalyzed PDI oxidation was monitored by measuring absorbance change at 340  nm using a SH-9000 microplate reader (Corona Electric Co.). The solution was first prepared by mixing 1 mM GPx7 WT or GPx7 C86A, 5-50 mM reduced PDI, 1 mM GSH, 1 unit GSH reductase (GR), and 200 mM NADPH in a buffer (50 mM Tris/ HCl, pH 7.4, and 300 mM NaCl), and the reactions were initiated at 30°C by adding 50 mM of H 2 O 2 to the solution. The initial NADPH consumption rates were calculated by considering the decrease of absorbance at 340 nm during the initial reaction time of 30 s with a molar extinction coefficient value of 6200 M 21 cm 21 for NADPH.
Preferential PDI oxidation by GPx7 in vitro 5 mM each of purified PDI family members (PDI, P5, ERp46, ERp57, and ERp72) was mixed with 0.5 mM GPx7. The reactions were initiated by adding 50 mM of H 2 O 2 to the solutions. At the indicated time points, the samples were subjected to 2 mM mal-PEG 2000 (NOF Corporation) followed by immunoblotting with the corresponding anti-PDI antibodies.

Cell culture, transfection, and sample preparation
The cells were transfected with plasmids using Effectene (Qiagen). For the transfection, HeLa cells were plated in 10-cm dishes at 8 3 10 5 cells/well, cultured for 24 h, and transfected with an appropriate plasmid, following the manufacturer's instruction. The amount of plasmid DNA used for the transfection was 2 mg for pcDNA3.1 1 and pCMV-Gg PRX4 (encoding PRX4-FLAG) (10), 6 mg for pES104 (encoding GPx7-FLAG), and 6 mg for pES105 (encoding full-length GPx8-FLAG). The larger amount of plasmid DNA was used for transfection with the latter two plasmids because the levels of protein production from these plasmids were smaller than that from the former one when the same amount of plasmid DNA was used. At 24 h after the transfection, the cells were treated with or without 0.5 mM H 2 O 2 for 10 min, washed twice with PBS, treated with 10% TCA, and subjected to alkylation with NEM essentially as described (51).
Immunoprecipitation 500 mg of NEM-treated cell lysate prepared as described above was diluted with ice-cold KI buffer (2% (w/v) Triton X-100, 50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1 mM EDTA) and centrifuged at 15,000 3 g for 10 min at 4°C to obtain NEM-treated cleared cell lysate. To purify FLAG-tagged proteins (PRX4-FLAG, GPx7-FLAG, or GPx8-FLAG) from the cleared cell lysate, 50 ml of prewashed anti-FLAG M2 magnetic beads (Sigma) were incubated with the NEM-treated cleared cell lysate at 4°C for 3 h. The immune complexes were collected by magnetization and washed four times with 800 ml of ice-cold high-salt buffer (1% (w/v) Triton X-100, 50 mM Tris-HCl, pH 8.0, 1 M NaCl, 1 mM EDTA) and once with 800 ml of 10 mM Tris-HCl (pH 8.0). The immunoisolates were then released by incubating the sample at 37°C for 1 h with 65 ml of 23 Laemmli sample buffer supplemented with 10 mM NEM, 2 mg/ml pepstatin A, 1 mM benzamidine, and 1 mM phenylmethylsulfonyl fluoride.

Data availability
All data described are contained within the article.