Generating an Unfoldase from Thioredoxin-like Domains

Protein-disulfide isomerase (PDI), an endoplasmic reticulum (ER)-resident protein, is primarily known as a catalyst of oxidative protein folding but also has a protein unfolding activity. We showed previously that PDI unfolds the cholera toxin A1 (CTA1) polypeptide to facilitate the ER-to-cytosol retrotranslocation of the toxin during intoxication. We now provide insight into the mechanism of this unfoldase activity. PDI includes two redox-active (a and a′) and two redox-inactive (b and b′) thioredoxin-like domains, a linker (x), and a C-terminal domain (c) arranged as abb′xa′c. Using recombinant PDI fragments, we show that binding of CTA1 by the continuous PDIbb′xa′ fragment is necessary and sufficient to trigger unfolding. The specific linear arrangement of bb′xa′ and the type a domain (a′ versus a) C-terminal to bb′x are additional determinants of activity. These data suggest a general mechanism for the unfoldase activity of PDI: the concurrent and specific binding of bb′xa′ to particular regions along the CTA1 molecule triggers its unfolding. Furthermore, we show the bb′ domains of PDI are indispensable to the unfolding reaction, whereas the function of its a′ domain can be substituted partially by the a′ domain from ERp57 (abb′xa′c) or ERp72 (ca°abb′xa′), PDI-like proteins that do not unfold CTA1 normally. However, the bb′ domains of PDI were insufficient to convert full-length ERp57 into an unfoldase because the a domain of ERp57 inhibited toxin binding. Thus, we propose that generating an unfoldase from thioredoxin-like domains requires the bb′(x) domains of PDI followed by an a′ domain but not preceded by an inhibitory a domain.

Protein-disulfide isomerase (PDI) 2 is a multifunctional protein that resides in the endoplasmic reticulum (ER) lumen of all eukaryotic cells (reviewed in Ref. 1). Mammalian PDI was first identified as a catalyst of oxidative protein folding (2), but it is now also known to mediate viral infection (3,4), antigen processing (5), collagen assembly (6), and ERassociated degradation (7)(8)(9). To participate in this variety of cellular processes, PDI performs multiple activities. For example, during oxidative protein folding, PDI catalyzes the oxidation and isomerization of disulfide bonds and induces conformational changes in non-native polypeptides (10). Independently of redox chemistry, PDI is a molecular chap-erone, binding polypeptides to prevent their aggregation (11)(12)(13). PDI also acts as a structural subunit of the prolyl 4-hydroxylase (P4H) and microsomal triglyceride transfer protein complexes; however, this function is similar to its chaperone activity (14 -19). In contrast to its protein folding activities, PDI unfolds the catalytic A1 polypeptide of cholera toxin (CTA1) in preparation for the retrotranslocation of the toxin from the ER lumen into the cytosol (8,20).
Cholera toxin (CT) is a pathogenic factor that causes secretory diarrhea in animals (reviewed in Ref. 21). The holotoxin includes a single catalytic A subunit (CTA) and a homopentameric B subunit (CTB) joined noncovalently (22). Upon secretion from the bacterium Vibrio cholerae, CTA is cleaved into the A1 and A2 polypeptides, which are joined by a disulfide bond and noncovalent interactions (22,23). To intoxicate a cell, CTB binds the ganglioside GM1 on the surface of the cell, and the holotoxin is transported in a retrograde manner to the ER lumen (24). In the ER, CTA is reduced to generate CTA1, and PDI unfolds and dissociates CTA1 from the holotoxin (20). The unfolded toxin is subsequently transported across the ER membrane (25,26). Upon reaching the cytosol, CTA1 refolds and induces toxicity (27,28).
We showed previously that PDI acts as a redox-dependent chaperone to unfold CTA1 (20). In the reduced state of PDI, it binds and unfolds the toxin. Subsequent oxidation of PDI by ER oxidase 1 causes PDI to release unfolded CTA1 (25). Aside from this information, nothing is known about the mechanism of the unfolding activity of PDI.
PDI is a modular protein comprising two a-type thioredoxinlike domains (a and aЈ), two b-type thioredoxin-like domains (b and bЈ), a flexible linker (x), and an extended C-terminal domain (c) arranged as abbЈxaЈc (29 -31). The a-type domains are characterized by the presence of the catalytic sequence CXXC and are therefore redox-active, whereas the b-type domains lack this sequence and are redox-inactive (32). The thioredoxin-like domains of PDI differ from each other in primary structure despite having a common fold. The crystal structure of yeast PDI shows the bbЈ domains form a rigid base from which the a-type domains extend like flexible arms (33,34). This base is thought to be the core of a substrate-binding groove formed by all four thioredoxin-like domains (30,33,35).
To understand the mechanism of the unfoldase activity of PDI, we analyzed the contribution of each domain to the ability of PDI to bind and unfold CTA1 using recombinant PDI fragments. Unfolded CTA1 was detected by an established in vitro trypsin sensitivity assay that relies on tryptic cleavage sites hidden in the folded toxin to be exposed in the unfolded toxin (20). Because CTA1 likely mimics a misfolded host cell protein for its recognition and unfolding by PDI (22,36,37), this study has implications for how PDI unfolds endogenous misfolded proteins in preparation for their retrotranslocation and subsequent ER-associated degradation.
There are nearly 20 mammalian PDI-like proteins, characterized by the presence of one or more thioredoxin-like domains and ER localization (reviewed in Refs. 38,39). We previously demonstrated that two PDI-like proteins, ERp72 and ERp57, do not facilitate CTA1 retrotranslocation (8). In contrast to PDI, ERp72 retains CTA1 in the ER and either stabilizes its native conformation or renders it more compact (8). To understand how these structurally homologous proteins are functionally unique, we tested whether the various thioredoxin-like domains of ERp57 and ERp72 could functionally replace the corresponding PDI domains to unfold CTA1. Thus, in addition to suggesting a general mechanism for the unfoldase activity of PDI, our data indicate functional similarities and differences among thioredoxin-like domains of PDI family proteins.
Trypsin Sensitivity Assay to Detect Unfolded CTA1-CTA (50 nM; EMD Biosciences) was incubated with purified PDI or BSA (5 M) in physiological buffer with GSH (3 mM) at 30°C for 30 min. The samples were next incubated with or without trypsin (10 M; Sigma) at 4°C for 30 min followed by incubation with the trypsin inhibitor tosyl-L-lysyl-chloromethane hydrochloride (10 mg/ml; Sigma) for 10 min at 4°C. Samples were resolved by reducing SDS-PAGE and analyzed by immunoblotting with an anti-CTA antibody generated in the Tsai laboratory.
CTA1 Release Assay-20 l of GM1-coated polystyrene beads (50% bead volume; 20) were incubated with 20 l of CT (200 nM) for 30 min at room temperature. The beads were washed with phosphate-buffered saline (PBS) and resuspended in 20 l of PBS. 15 l of the indicated PDI fragment (18 M) were added to the toxin-coated beads in the presence of GSH (3 mM). The samples were incubated for 60 min at room temperature and pelleted, and the supernatant was removed. The pellet and supernatant fraction were subjected to reducing SDS-PAGE and immunoblotted with antibodies against CTA and CTB. GM1-coated polystyrene beads were generous gifts from Wayne Lencer (Children's Hospital, Boston).
Co-immunoprecipitation-CTA (50 nM) was incubated with purified His-tagged proteins (5 M) in physiological buffer containing GSH or GSSG (3 mM) for 30 min at 30°C to stimulate the unfolding reaction. Next, His-tagged proteins were immunoprecipitated using an anti-His antibody (H-15, Santa Cruz Biotechnology) for 1 h at 4°C. Immune complexes were isolated by incubation with protein A-agarose beads (Invitrogen) for 30 min at 4°C and resolved by reducing SDS-PAGE. CTA1 was detected by immunoblotting with our anti-CTA antibody, whereas PDI proteins were detected by Western blotting with an anti-PDI antibody (H-160, Santa Cruz Biotechnology) or by Coomassie staining the gel.
Limited Proteolysis of Recombinant Proteins-Purified recombinant proteins (7 M) were incubated with the trypsin concentrations indicated in the figures for 30 min at 30°C in physiological buffer containing GSH (3 mM). Proteolysis was inhibited by incubation with tosyl-L-lysyl-chloromethane hydrochloride (10 mg/ml) for 10 min at 4°C. Samples were subjected to reducing SDS-PAGE, and proteins were visualized with Coomassie stain.

RESULTS
Recombinant PDI Unfolds CTA1 Efficiently-Native PDI derived from bovine liver was shown previously to unfold the CTA and CTA1 polypeptides, rendering them sensitive to tryptic digestion (20). Using a similar assay, we tested whether Histagged, full-length, mouse PDI (PDI-FL) expressed in and puri-fied from bacteria ( Fig. 1A) can unfold CTA1. Hence, we incubated CTA with purified PDI-FL or the control protein BSA in the presence of the reducing agent glutathione (GSH) at 30°C followed by incubation with trypsin at 4°C. Under this condition, the majority of CTA is reduced by GSH to generate CTA1 (data not shown). The samples were then subjected to SDS-PAGE and immunoblotted with an antibody against CTA1. Whereas CTA1 was mostly resistant to proteolysis when incubated with BSA (Fig. 1B, compare lane 2 with 1), it became sensitive to tryptic digestion when incubated with PDI-FL, indicating that PDI-FL unfolded the toxin (Fig. 1B, compare lane 4 with 3). We next compared the unfoldase activity of recombinant PDI-FL to native PDI using two concentrations of the PDI proteins. At both concentrations, recombinant and native PDI caused CTA1 to become sensitive to tryptic digestion with a similar efficiency (Fig. 1C, compare lane 2 with 3 and lane 5 with 6), indicating that native and recombinant PDI have similar unfoldase activity. Furthermore, when the toxin was incubated with recombinant PDI-FL at 4°C instead of 30°C (followed by incubation with trypsin at 4°C), CTA1 was resistant to tryptic digestion (Fig. 1D, compare lane 4 with 2) as expected for a protein-mediated activity. These experiments establish the feasibility of using recombinant PDI to study the unfoldase activity of PDI in vitro.
PDIbbЈxaЈ Is the Minimum Unit Required to Unfold CTA1 Efficiently-To determine the domain requirements for the unfoldase activity of PDI, we expressed and purified the PDI fragments depicted in Fig. 2A and tested their abilities to render CTA1 trypsin-sensitive. Each fragment was isolated as a soluble protein consistent with previous observations that each thioredoxin-like domain autonomously forms a folded structure (40 -42). When PDIabbЈxaЈ was incubated with the toxin followed by incubation with trypsin, CTA1 was digested efficiently (Fig.  2B, compare lane 3 with lanes 1 and 2), indicating the c domain is not required for the unfoldase activity of PDI. Subsequent deletion of domain aЈ, generating fragment PDIabbЈx, resulted in a loss of the ability of PDI to render CTA1 trypsin-sensitive (Fig. 2B, compare lane 5 with 4), whereas deletion of domain a, generating PDIbbЈxaЈ, did not affect activity (Fig. 2B, compare lane 6 with 4). Thus, domain aЈ but not a is required for the unfoldase activity of PDI. PDIab did not render CTA1 trypsinsensitive (Fig. 2B, compare lane 7 with 4), as expected. However, in contrast to PDIbbЈxaЈ, PDIbЈxaЈ did not make CTA1 trypsin-sensitive (Fig. 2B, compare lane 8 with 6), indicating the b domain is required for unfoldase activity. To determine whether the domains of PDI must be linked to facilitate unfolding, we tested whether fragments PDIab and PDIbЈxaЈ could cooperate to unfold CTA1 and found that together they had very little activity (Fig. 2C, compare lane 3 to 2). We conclude the linked bbЈxaЈ domains form the minimum unit required for efficient unfoldase activity.
It was demonstrated previously that the unfolding of CTA1 correlates with its dissociation from CTB (20). We therefore tested whether PDIbbЈxaЈ could induce release of CTA1 from CTB using a previously established "release" assay (19). Accordingly, CT holotoxin was bound to GM1-coated beads that were then incubated with various PDI fragments in the presence of GSH. After sedimenting the beads, the pellet and supernatant fractions were analyzed by SDS-PAGE and immunoblotting with anti-CTA or anti-CTB antibodies. When the CT-coated beads were incubated with the control protein BSA, both CTA and CTB were entirely in the pellet fraction (Fig. 2D, compare lane 1 with 2). However, when the beads were incubated with PDIbbЈxaЈ, a fraction of CTA1 appeared in the supernatant, whereas CTB remained entirely in the pellet fraction (Fig.  2D, compare lane 3 with 4) indicating that PDIbbЈxaЈ induced the release of CTA1 from CTB. In contrast to incubation with PDIbbЈxaЈ, little CTA1 appeared in the superna-FIGURE 1. Recombinant PDI unfolds CTA1 efficiently. A, Coomassie stain of His-tagged full-length PDI (PDI-FL) purified from bacteria and subjected to SDS-PAGE. B, CTA was incubated with BSA or PDI in the presence of GSH at 30°C followed by incubation with or without trypsin at 4°C. Samples were resolved by reducing SDS-PAGE and analyzed by immunoblotting with an anti-CTA antibody that recognizes CTA1. C, as in B, except native PDI isolated from bovine liver was used, and all samples were treated with trypsin. D, as in B, except CTA was incubated with PDI at 4 or 30°C before incubation with trypsin at 4°C. tant after incubation with PDIabbЈx (Fig. 2D, compare lane 6 with 4) or PDIbЈxaЈ (Fig. 2D, compare lane 12 with 10). Thus, we conclude that PDIbbЈxaЈ is more efficient than PDIabbЈx and PDIbЈxaЈ at inducing dissociation of CTA1 from the holotoxin. These data are consistent with the results of our trypsin sensitivity assay, which indicates that PDIbbЈxaЈ has the most unfoldase activity compared with the other PDI fragments. A PDI-Toxin Interaction Is Necessary but Not Sufficient to Induce Unfolding-To further elucidate the unfolding reaction, we sought to identify the PDI domains that interact with CTA1. Hence, we incubated CTA with PDIabbЈxaЈ or His-tagged ERp29, a control protein that does not unfold the toxin (data not shown), under reducing conditions to stimulate the unfolding reaction and then immunoprecipitated PDIabbЈxaЈ or ERp29, along with any bound toxin, using an anti-His antibody. Immunoprecipitates were analyzed by SDS-PAGE and Coomassie staining or immunoblotting with an antibody against CTA or PDI. As expected, CTA1 co-precipitated efficiently with PDIabbЈxaЈ, but only minimally with ERp29 (Fig. 3A, compare lane 4 with 3), demonstrating the toxin binds specifically to PDIabbЈxaЈ. Repeating the same experiment using the vari- ous PDI fragments revealed that PDIabbЈx and PDIbbЈxaЈ bound to the toxin with a similar efficiency as PDIabbЈxaЈ (Fig.  3B, top panel, compare lanes 8 and 9 with 7), whereas the PDI fragments containing only two domains (PDIab, PDIbЈxaЈ and PDIbbЈx) bound CTA1 less efficiently (Fig. 3B, top panel, compare lanes 10 -12 with lanes 7-9). There was some variability in the relative binding abilities among the fragments (Fig. 3B,  graph); however, the fragments with only two domains consistently co-precipitated less CTA1 than fragments with three domains. These results indicate that each PDI thioredoxin-like domain can interact with CTA1, but either PDIabbЈx or PDI-bbЈxaЈ is required for a strong interaction. Because PDIabbЈx and PDIbbЈxaЈ bound to CTA1 with similar efficiencies, whereas only PDIbbЈxaЈ unfolded the toxin (Fig. 2B), we conclude that an efficient PDI-toxin interaction is required but not sufficient to trigger unfolding.
The reduced but not oxidized state of full-length PDI was shown previously to bind CTA1 (20). To test whether PDIabbЈx or PDIbbЈxaЈ also binds in a redox-dependent manner, we incubated CTA with PDIabbЈx or PDIbbЈxaЈ in the presence of either GSH or oxidized glutathione (GSSG) followed by immunoprecipitation of the PDI fragments. Analysis of the immunoprecipitates revealed that both PDI fragments bound CTA1 under the reducing, but not oxidizing, condition (Fig. 3C, top  panel, compare lanes 5 and 6 with lanes 7 and 8). Thus, we conclude PDIabbЈx and PDIbbЈxaЈ interact with CTA1 in a redox-driven manner similar to full-length PDI (20).
Additional Determinants of PDI Unfoldase Activity-Because a strong PDI-toxin interaction was not sufficient to unfold the toxin, we tested whether the specific linear arrangement of bbЈxaЈ and/or the identity of the a-type domain (aЈ versus a) C-terminal to bbЈx are also determinants of the unfoldase activity of PDI. Thus, we exchanged the positions of the a and aЈ domains of PDI, generating the purified PDIbbЈxa and PDIaЈbbЈx (Fig. 4A, lanes 1 and 2) and tested their abilities to unfold CTA1 using the trypsin sensitivity assay. Neither PDI-bbЈxa nor PDIaЈbbЈx rendered CTA1 sensitive to tryptic digestion (Fig. 4B, compare lanes 3 and 4 with 1 and 2), indicating neither fragment unfolded the toxin. In contrast, co-immunoprecipitation experiments revealed that PDIbbЈxa bound the toxin with an efficiency similar to PDIbbЈxaЈ (Fig. 4C, top panel,  compare lane 7 with 6). Thus, domain a can promote stable binding to CTA1 regardless of its position relative to the bbЈx domains ( Fig. 3B and Fig. 4C), although neither interaction induces the unfolding of the toxin (Fig. 2B and Fig. 4B). In contrast, PDIaЈbbЈx bound less CTA1 compared with PDIbbЈxaЈ (Fig. 4C, top panel, compare lane 8 to 6), suggesting that domain aЈ must be C-terminal to bbЈx to efficiently bind and unfold the toxin.
We next determined whether PDIbbЈxa and PDIaЈbbЈx might have structural alterations that could explain their inability to unfold the toxin using limited proteolysis. PDIbbЈxa was insensitive to proteolysis, similar to PDIbbЈxaЈ (Fig. 4D, compare lanes 5-8 with lanes 1-4), indicating that PDIbbЈxa is structurally intact. In contrast, PDIaЈbbЈx was more susceptible to proteolysis than PDIbbЈxaЈ or PDIabbЈx (Fig. 4D, compare  lanes 9 -12 with lanes 1-4 and 13-16); however, the generation of defined proteolytic fragments derived from PDIaЈbbЈx (Fig.   4D, lanes 11 and 12, * and **) suggests it is destabilized locally rather than misfolded globally. Thus, we conclude that, in addition to an efficient PDI-toxin interaction, both the specific linear arrangement of bbЈxaЈ and the type of a domain (aЈ versus a) C-terminal to bbЈx are determinants of the unfoldase activity of PDI. We also tested whether the flexibility between the bЈ and aЈ domains (34), mediated by the x linker (43), is important for this activity. PDIbbЈaЈ, lacking x, was unable to unfold CTA1 (data not shown); however, because PDIbbЈaЈ was significantly more sensitive to proteolysis than PDIbbЈxaЈ, this result is difficult to interpret.
Functional Differences between PDI and ERp57 Thioredoxinlike Domains-We next tested whether the thioredoxin-like domains from ERp57, the closest known PDI homologue (1), can functionally substitute the thioredoxin-like domains of PDI to generate an unfoldase. ERp57 has the same overall architecture as PDI (Fig. 5A). We showed previously that ERp57 is not involved in CTA1 retrotranslocation (8), consistent with its inability to unfold CTA1 in vitro (data not shown). In the context of the minimum domains required to unfold CTA1, bbЈxaЈ, we tested whether replacing the bbЈ or aЈ domains of PDI with the corresponding domains from ERp57 produces an unfoldase. Hence, we generated purified PDIbbЈ-ERp57xaЈ and ERp57bbЈx-PDIaЈ, along with ERp57bbЈxaЈ (Fig. 5B) and tested their unfoldase activity using the trypsin sensitivity assay (Fig.  5C). PDIbbЈ-ERp57xaЈ unfolded CTA1, whereas ERp57bbЈxaЈ and ERp57bbЈx-PDIaЈ were inactive (Fig. 5C, compare lanes 4  and 2 with lanes 5 and 3). (We used the x linker from ERp57, rather than from PDI, in the PDI-ERp57 hybrid because a hybrid containing the x linker of the PDI showed signs of structural instability.) Furthermore, when purified PDIbbЈ-ERp57xaЈ or ERp57bbЈxaЈ was incubated with CT bound to GM1-coated beads, PDIbbЈ-ERp57xaЈ but not ERp57bbЈxaЈ induced release of CTA1 from CTB into the supernatant fraction (Fig. 5D, compare lane 4 with 6). This is consistent with our conclusion that PDIbbЈ-ERp57xaЈ but not ERp57bbЈxaЈ is able to unfold CTA1.
We hypothesized that the a domain of ERp57 disrupts toxin binding because the a-b junction of ERp57a-PDIbbЈ-ERp57xaЈ is more flexible than the corresponding junction of PDIabbЈxaЈ, enabling the a domain of ERp57 to collapse onto the bbЈ domains of PDI, thereby interfering with binding. Thus, we subjected ERp57a-PDIbbЈ-ERp57xaЈ and PDIabbЈxaЈ to limited proteolysis. Discrete fragments derived from ERp57a-PDIbbЈ-ERp57xaЈ but not PDIabbЈxaЈ appeared (Fig. 5I, top  panel, compare lane 6 with 3). Importantly, fragment 3 reacted with an anti-PDI but not an anti-His antibody (Fig. 5I, compare  middle and bottom panels, lane 6), indicating that fragment 3 lost the N-terminal His tag appended to domain a. Because fragment 3 is approximately the same size as PDIbbЈ-ERp57xaЈ (Fig. 5B), the simplest explanation is that the a domain of ERp57 was removed by proteolysis. This implies the a-b junction of ERp57a-PDIbbЈ-ERp57xaЈ is more flexible than the corresponding junction of PDIabbЈxaЈ, which might explain why addition of a domain of ERp57 to PDIbbЈ-ERp57xaЈ inhibited toxin binding. Thus, we conclude that aЈ domain of PDI but not its abbЈ domains can be partially substituted by the corresponding ERp57 domains to generate an unfoldase. A, Coomassie stain of the indicated His-tagged proteins purified from bacteria and subjected to SDS-PAGE. B, CTA was incubated with the indicated proteins in the presence of GSH followed by incubation with trypsin. CTA1 was detected by SDS-PAGE followed by immunoblotting with an anti-CTA antibody. C, CTA was incubated with the indicated proteins in the presence of GSH to stimulate the unfolding reaction. PDI-toxin complexes were isolated by co-immunoprecipitation with an antibody against the His tag on PDI and resolved by reducing SDS-PAGE. CTA1 was detected by immunoblot analysis using an anti-CTA antibody (top), and PDI proteins were detected by Coomassie stain (bottom). D, Coomassie stain of the indicated proteins subjected to SDS-PAGE after incubation with the indicated amounts of trypsin.

Functional Differences between PDI and ERp72 Thioredoxin-
like Domains-We also tested whether any of the bbЈxaЈ domains from the PDI-like protein ERp72 can be used to gen-erate an unfoldase. Full-length ERp72 includes three a-type, two b-type thioredoxin-like domains, an x linker, and an acidic extension (c) arranged ca°abbЈxaЈ (Fig. 6A). We purified ERp72bbЈxaЈ, PDIbbЈx-ERp72aЈ, and ERp72bbЈx-PDIaЈ and tested their abilities to unfold CTA1 using the trypsin sensitivity assay (Fig. 6B). ERp72bbЈxaЈ was inactive (Fig. 6C, compare lane 3 with 2), consistent with our previous report showing that full-length ERp72 does not unfold the toxin (8). PDIbbЈx-ERp72aЈ was partially able to unfold CTA1 (Fig. 6C, compare lane 6 with 5), whereas ERp72bbЈx-PDIaЈ was inactive (Fig. 6C, compare lane 7 with 5).
The inactive ERp72bbЈx-PDIaЈ hybrid was only slightly more sensitive to proteolysis than ERp72bbЈxaЈ or PDIbbЈx-ERp72aЈ (Fig. 6D, compare lanes 5-8 with lanes 1-4 and 9 -12). Moreover, the activity of PDIbbЈx-ERp72aЈ correlates with its ability to bind CTA1, albeit with slightly reduced efficiency compared with PDIbbЈxaЈ (Fig. 6E, compare lane 3 with 2). Thus, the various PDI-ERp72 and PDI-ERp57 hybrids display a similar trend in their abilities to unfold CTA1, indicating the PDI bbЈx domains are indispensable for the unfolding reaction, whereas the aЈ domain from other PDI-like proteins can be used to generate an unfoldase.

DISCUSSION
This study identifies the general structural features of PDI required for the unfolding reaction, thus providing insight into the mechanism of the unfoldase activity of PDI. Specifically, we found that binding of CTA1 by the bbЈxaЈ domains of PDI is necessary and sufficient to unfold CTA1 efficiently. However, a strong PDI-toxin interaction is not sufficient to trigger unfolding; the specific linear arrangement of bbЈxaЈ and identity of the a-type domain (aЈ versus a) C-terminal to bbЈx are additional determinants of activity. We also show that the bbЈ domains of PDI are indispensable to the unfolding reaction, whereas the function of its aЈ domain can be substituted partially by domain aЈ from ERp57 or ERp72. However, the bbЈ domains of PDI were not sufficient to convert full-length ERp57 into an unfoldase because the a domain of ERp57 inhibited the unfolding reaction.
Consistent with our discovery that PDIbbЈxaЈ is necessary and sufficient to unfold CTA1, we showed previously that oxidation of the aЈ domain of PDI by ER oxidase 1 causes PDI to release the unfolded toxin (25). Whether PDIbbЈxaЈ is sufficient to facilitate retrotranslocation remains to be determined and will reveal whether domains a and c are required in vivo, perhaps to interact with components of the retrotranslocation machinery (26).
Although PDIbbЈxaЈ was sufficient to bind and unfold CTA1 with maximum efficiency, our data show that all four PDI thioredoxin-like domains can interact with the toxin. Interestingly, PDIbbЈxaЈ and PDIabbЈx bound CTA1 with similar efficiencies, but PDIabbЈx did not unfold the toxin, indicating an efficient interaction is not sufficient to trigger unfolding. However, because domain a was not required to unfold CTA1, it is unclear whether it binds the toxin in the context of full-length PDI. Nevertheless, hydrophobic regions suitable for binding non-native proteins are present on each PDI thioredoxin-like domain (30,33), and each domain has been shown to contribute to binding substrates (35,44). Moreover, specific hydrophobic residues that mediate substrate interactions have been identified on the a, bЈ, and aЈ domains of PDI (31,45). Future studies are required to test whether these same sites are critical for the PDI-toxin interaction, revealing whether differences in the nature of the interactions of PDI with its substrates might explain its multiple activities.
In addition to the requirement for a strong PDI-toxin interaction, we show the specific linear arrangement of the bbЈxaЈ domains is required to trigger the unfolding of CTAI. Similarly, the arrangement of domains is important for the oxidoreductase functions of PDI during protein folding (46,47). That the bbЈxaЈ domains, in this specific arrangement, are required to interact efficiently with CTA1 indicates the binding site on each domain must be aligned to bind CTA1 concurrently, and this continuous binding site is asymmetric. Thus, we propose a general mechanism for the unfolding reaction: the concurrent binding of the bbЈxaЈ domains to particular regions along the CTA1 molecule triggers the unfolding of the toxin.
The a domain of PDI cannot functionally replace its aЈ domain in the unfolding reaction, indicating domain aЈ is intrinsically suited for this activity. This is also true for the function of PDI as a P4H subunit (44). Because either a-type domain positioned C-terminal to bbЈx promoted a stable PDI-toxin interaction, their functional difference is likely due to a difference in precisely how they interact with CTA1. Indeed, domain a but not aЈ promoted stable toxin binding regardless of its relative position to bbЈx, implying domain aЈ has a higher binding specificity than domain a. The notion that domain aЈ is intrinsically suited for the unfolding reaction is further supported by our finding that the aЈ domain of PDI can be substituted partially by the aЈ domain from ERp57 or ERp72. Comparison of the amino acid sequence of the aЈ domain of PDI to its a domain (ϳ40% identical) and to the ERp57 and ERp72 aЈ domains (each ϳ50% identical) will help to identify unique features of the aЈ domain of PDI that are critical for the unfolding reaction.
Analysis of the ability of the thioredoxin-like domains from PDI, ERp57, and ERp72 to unfold CTA1 indicates a principal functional difference lies in their bbЈ domains. Indeed, these domains share low sequence homology (ϳ20 -25% identical) and have different binding specificities (44, 48 -50). Nonetheless, the bbЈ domains of PDI were unable to convert full-length ERp57 into an unfoldase, showing that their a domains also differ functionally. Our data suggest this difference is because the a-b junction of ERp57a-PDIbbЈ-ERp57xaЈ is more flexible than the corresponding junction in PDIabbЈxaЈ, enabling the a domain of ERp57 to inhibit toxin binding. Therefore, although the a domain of PDI does not contribute directly to the unfolding reaction, it appears to indirectly play a role by not inhibiting the PDI-toxin interaction. Furthermore, these results imply that potential differences in the flexibilities of PDI and ERp57 (34,51,52) might contribute to their functional differences. Together these results indicate the abbЈ domains of PDI, ERp72, and ERp57 account for their differences in unfoldase activity. Moreover, the results suggest that generating an unfoldase from thioredoxin-like domains requires the bbЈ(x) domains of PDI followed by an aЈ domain but not preceded by an a domain that inhibits toxin binding.