ERdj5 Is Required as a Disulfide Reductase for Degradation of Misfolded Proteins in the ER

See allHide authors and affiliations

Science  25 Jul 2008:
Vol. 321, Issue 5888, pp. 569-572
DOI: 10.1126/science.1159293


Membrane and secretory proteins cotranslationally enter and are folded in the endoplasmic reticulum (ER). Misfolded or unassembled proteins are discarded by a process known as ER-associated degradation (ERAD), which involves their retrotranslocation into the cytosol. ERAD substrates frequently contain disulfide bonds that must be cleaved before their retrotranslocation. Here, we found that an ER-resident protein ERdj5 had a reductase activity, cleaved the disulfide bonds of misfolded proteins, and accelerated ERAD through its physical and functional associations with EDEM (ER degradation–enhancing α-mannosidase–like protein) and an ER-resident chaperone BiP. Thus, ERdj5 is a member of a supramolecular ERAD complex that recognizes and unfolds misfolded proteins for their efficient retrotranslocation.

In eukaryotic cells, secretory and membrane proteins are cotranslationally translocated into the endoplasmic reticulum (ER), acquire N-glycans and disulfide bonds, and become folded with the help of ER-resident molecular chaperones. Correctly folded proteins exit the ER and traffic through the Golgi to their final destinations. However, if these proteins fail to acquire their correct conformation, they are recognized by ER “quality-control” mechanisms (1). Terminally misfolded proteins are retrotranslocated from the ER into the cytosol for degradation via the ubiquitin-proteasome system, a process known as ER-associated degradation (ERAD) (1).

The enzymatic modification of the N-linked Glc3Man9GlcNAc2 oligosaccharide of glycoproteins (i.e., glucose, mannose, and N-acetylglucosamine) is recognized not only by the calnexin-calreticulin cycle for productive folding (1), butalsobyER degradation–enhancing α-mannosidase–like protein EDEM (Htm1/Mnlp1 in yeast), or Yos 9 (yeast osteosarcoma 9) for ERAD (2). EDEM enhances the degradation of misfolded proteins in a mannose-trimming–dependent manner (3). EDEM accepts Man8GlcNAc2 substrates from calnexin (4, 5) and also associates with the transmembrane proteins Derlin 2 and 3 (6), which inturnassociate withthe cytosolic p97 complex. However, EDEM itself does not bind properly folded proteins, nor does it affect their secretion or degradation (7).

Disulfide bonds not only stabilize protein tertiary structure, but, in the ER, they also create large oligomers of misfolded proteins (8) that may not be accommodated by the retrotranslocation channel. Thus, the reduction of such disulfide bonds is required for the unfolding and retrotranslocation of misfolded proteins. ERAD is accelerated by treatment with reductants, such as dithiothreitol (DTT), and is inhibited by oxidants (7, 9). No protein having reductase activity in the ER has been reported to be involved in ERAD.

To elucidate the precise role of EDEM in ERAD, we screened for EDEM-binding ER proteins by an ER-membrane yeast two-hybrid system (ER-MYTHS) (10, 11). In this method, the ER luminal portion of the yeast Ire1p was replaced by rat EDEM as bait and tested against a library of ER proteins as prey. Association of prey and bait causes oligomerization of Ire1p and activates its cytosolic ribonuclease activity, which leads to splicing of the mRNA of the transcriptional activator HAC1 and, in turn, results in induction of the reporter gene (ADE2 in this study) under control of unfolded protein–responsive elements (UPRE).

In a screen of a focused library of ER-resident proteins, we identified ERdj5 (JPDI) (12, 13) as an EDEM-binding protein (fig. S1A). The ERdj family is comprised of five ER proteins, each containing a DnaJ domain (14), and ERdj5 is the only member that has thioredoxin-like domains with CXXC motifs (active cysteines with various amino acid residues between them) (Fig. 1A) (15). The specificity of the binding of ERdj5 to EDEM was confirmed with ER-MYTHs by using ERdj5 as bait (fig. S1B), and EDEM did not bind to any other ERdj proteins, including ERdj3 and ERdj4 (fig. S1C).

Fig. 1.

Characteristics of recombinant ERdj5. (A) Schematic representation of the domain structures of ERdj5, ERdj family proteins, and thioredoxin superfamily proteins. (B) Coomassie brilliant blue (CBB) stain of purified recombinant ERdj5 and its cysteine mutant ERdj5/SS. (C) Insulin reductase activities of ERdj5 and PDI. The enzyme-catalyzed reduction of the insulin disulfide bonds by GSH is coupled to the reduction of GSSG to GSH by glutathione reductase (GR). The insulin reductase activity of oxido-reductases was measured in the presence of 8 mM GSH at 25°C by spectrophotometrically monitoring NADPH consumption, which is concomitant with GSSG reduction by GR. (D) Redox equilibrium assay with glutathione at 30°C. The free sulfhydryl groups of the cysteine residues were modified with fluorescein 5-maleimide (top) after incubation with different [GSH]2/[GSSG] ratios. (Bottom) A CBB stain of the fluorescence-labeled proteins. (E) Measured redox equilibrium constant of ERdj5.

We purified recombinant mouse ERdj5 from Escherichia coli (E. coli) (Fig. 1B) and determined its reductase activity using oxidized insulin as a substrate. ERdj5 catalyzed the reduction of the insulin disulfide bonds in a dose-dependent manner in the presence of reduced glutathione (GSH), although the specific activity of ERdj5 was about one-third that of recombinant human protein disulfide reductase (PDI) (Fig. 1C). This reductase activity was not displayed by ERdj5/SS mutant, in which all cysteines of the four CXXC motifs of ERdj5 are replaced by serines (Fig. 1C). In addition, ERdj5 had neither oxidase nor isomerase activity for ribonuclease A (RNase A) and lysozyme (fig. S2).

Recombinant ERdj5 was incubated with different ratios of GSH and oxidized glutathione (GSSG) in order to determine its redox potential. The redox equilibrium constant K of ERdj5 was determined from the alkylation of free cysteines with a fluorescent maleimide (Fig. 1D). The apparent equilibrium constant (Kapp) of ERdj5 (190 mM) was ∼100 times that of the redox state of the ER lumen (16) (0.5 to 2.3 mM) (Fig. 1E), consistent with ERdj5's having only reductase activity. Because the redox state of the ERdj5/SS mutant did not change within the measured redox ranges (Fig. 1E), we could attribute the observed Kapp values of ERdj5 to its CXXC motifs. The Kapp of ERdj5 is the most reducing equilibrium constant reported for any ER oxidoreductase to date, which suggests that ERdj5 acts as a potent disulfide reductase under the redox conditions of the ER.

Because ERdj5 could act as a disulfide reductase and bind to EDEM, we examined the effect of ERdj5 on ERADin vivo using two representative ERAD substrates, the null Hong Kong (NHK) variant of human α1-antitrypsin and the J chain of mouse immunoglobulin M (8). NHK contains one cysteine, but it aberrantly forms a disulfide-linked dimer that can be reduced to a monomer by DTT treatment, which accelerates its ERAD (7). Transfection of ERdj5 decreased the formation of NHK dimers after chase periods (Fig. 2, A and C). NHK was degraded much faster in ERdj5-transfected cells than in mock-transfected cells (Fig. 2, A and B). In contrast, transfection of the ERdj5/SS mutant did not inhibit dimer formation or promote the ERAD of NHK (Fig. 2, A and B, and fig. S3). Similarly, transfection with the mutants ERdj5/AA, wherein the cysteines of the four CXXC are replaced by alanine (AXXA), or ERdj5/CA, wherein the four CXXC are mutated to CXXA, did not affect the ERAD of NHK (fig. S4). An NHK/CS mutant in which the cysteine residue in NHK was mutated to serine did not form dimers and was degraded much more rapidly than wild-type NHK, regardless of the presence or absence of ERdj5 (Fig. 2D). ERdj5 overexpression did not induce the unfolded protein response (UPR) in HEK293 cells; nor was it able to induce UPR in HeLa cells, which are more stress-sensitive than HEK293 cells (fig. S5).

Fig. 2.

Overexpression of ERdj5 accelerates ERAD for NHK. HEK293 cells were labeled for 15 min with [35S]methionine-cysteine after 24 hours of transfection and chased for the periods indicated. The metabolically labeled NHK were immunoprecipitated with an α1-antitrypsin–specific antibody (A) and the band intensities under reducing conditions were quantified (B). R, reducing condition; NR, nonreducing condition. (C) Relative amount of the NHK dimer during the chase periods. (B) and (C) Means ± SD of three independent experiments. (D) Degradation of NHK cysteine-less mutant (NHK/CS) was similarly quantified. (E) Ratio of the free SH form of NHK to its SS form in pulse-chase experiments. HEK293 cells were labeled for 15 min with [35S] after 24 hours of transfection and were chased for 1 hour in the absence or presence of 0.5 mM DTT. MG132 was added 4 hours before pulse-label to inhibit proteasome activity. Immunoprecipitation and modification of SH groups of NHK were performed as described in (10). The immunoprecipitants were subjected to reducing SDS–polyacrylamide gel electrophoresis, and the radioactivities of the bands of modified (SH form) and unmodified (SS form) NHK were quantified.

J chains have two inter- and three intramolecular disulfide bonds, and overexpression of J chains in HEK293 cells produced intermolecular disulfide-bonded high-molecular-weight (HMW) complexes that accumulated during the chase period (fig. S6). Transfection of wild-type ERdj5 into HEK293 cells suppressed the accumulation of HMW bands (fig. S6C) and slightly increased the levels of monomer and dimer species (fig. S6A). ERdj5 overexpression dramatically enhanced J chain degradation compared with that in mock-transfected controls (fig. S6, A and B). The J chain mutant J/CS, in which two intermolecular disulfide-bonding cysteines were replaced with serines, did not produce HMW complexes and was degraded much more rapidly than wild-type J chains (fig S6, A and B).

If ERdj5 acts as a reductase for misfolded substrates, it should bind to NHK. The formation of a mixed disulfide complex of NHK with wild-type ERdj5 was detected in a coimmunoprecipitation analysis; such complexes were not detected with the ERdj5/AA mutant (fig. S7). We next examined the accumulation of reduced NHK in the presence of ERdj5. In the presence of the proteasomal inhibitor MG132, the ratio of the reduced and oxidized forms of NHK was not altered by ERdj5 overexpression (Fig. 2E). This may be because reoxidation of stalled substrates occurred during the chase period. However, when reducing equivalents were provided by adding low concentrations of DTT, accumulation of reduced NHK increased in the ERdj5-overexpressed cells compared with nontransfected cells (Fig. 2E). Such an increase in the reduced form of NHK was not observed when PDI was overexpressed (Fig. 2E). Thus, overexpression of ERdj5 prevents the covalent multimer formation of misfolded proteins by disulfide bond cleavage and thereby accelerates protein degradation of the ERAD substrates.

We next examined the contribution of endogenous ERdj5 to the ERAD of J chains or of NHK by transfecting HEK293 cells with an ERdj5-specific small interfering RNA (siRNA). The amount of ERdj5 in these cells was reduced to 10% of that in nonspecific siRNA-transfected controls (Fig. 3A). ER stress was not induced in cells in which expression of ERdj5 was reduced (knocked down) (fig. S5). ERdj5 knockdown induced the accumulation of HMW complexes of J chains and of NHK dimers (Fig. 3, B and C). Furthermore, the degradation of J chains and NHK was clearly inhibited in ERdj5-knockdown cells compared with nonspecific siRNA-transfected cells (Fig. 3, B and C). Finally, degradation of an ERAD substrate without cysteine, a soluble form (residues 1 to 332) of ribophorin I, was unaffected by knockdown of endogenous ERdj5 (Fig. 3D). Thus, endogenous ERdj5 is involved in the acceleration of ERAD by cleaving the disulfide bonds of misfolded proteins.

Fig. 3.

Effect on ERAD of endogenous ERdj5 knockdown by the use of siRNA. (A) Immunoblotting of a HEK293 cell lysate with an antibody against ERdj5 shows that endogenous ERdj5 was efficiently knocked down 48 hours after siRNA transfection. (B and C) ERdj5 knockdown using siRNA in HEK293 cells slows ERAD of the J chain (B) and NHK (C). Pulse-chase experiments were performed 24 hours after transfection of ERAD substrates, which were transfected 48 hours after siRNA transfection. This degradation analysis was performed 24 hours after J chain transfection to observe the delay of degradation; this explains why the degradation profile of the J chain is different from that in fig. S6. Each graph shows the quantification of the band intensities under reducing conditions. (D) Knockdown of ERdj5 has no effect on the ERAD of the cysteine-less substrate, soluble ribophorin (1-332) (RI332). Results are means ±SD of three independent experiments. NS, nonspecific siRNA; R, reducing condition; NR, nonreducing condition.

ERdj5 contains a DnaJ domain and was shown to interact with the DnaJ-binding chaperone BiP in the ER-MYTHS system (fig. S1B). We also showed that recombinant BiP binds directly to ERdj5 in an adenosine triphosphate (ATP)–dependent manner, by surface plasmon resonance by using recombinant ERdj5 immobilized on a CM5 sensor chip (Fig. 4A). Because DnaJ domain–containing proteins interact with heat shock protein 70 (Hsp70) family proteins through their His-Pro-Asp (HPD) motifs, we next examined the binding of BiP with a DnaJ-domain mutant of ERdj5 in which histidine 63 in the HPD motif is replaced with glutamine (ERdj5/H63Q) (13). ERdj5/H63Q did not bind to BiP, even in the presence of ATP (Fig. 4A). Thus, ERdj5 interacts through its DnaJ domain with BiP.

Fig. 4.

Interactions of ERdj5 with EDEM and BiP have important implications for ERAD acceleration by ERdj5. (A) Surface plasmon resonance experiments. ERdj5 was immobilized on a CM5 chip and BiP was injected as an analyte with nucleotides. The time points of each injection are indicated by arrows. (B) Accumulation of a HMW complex of J chains at 48 hours after transfection was analyzed under nonreducing conditions by immunoblotting with a Myc-specific antibody. (C) Pulse-chase experiments of the J chain in HEK293 cells transfected with ERdj5 or the ERdj5/H63Q mutant were performed 36 hours after transfection. (D) ERAD promotion by ERdj5 was inhibited by adding the ER α1,2-mannosidase inhibitor kifunensine; experiments using NHK as a substrate were performed 24 hours after transfection. Transfected cells were treated with 5 μg/ml kifunensine for 4 hours before pulse-label. Results in (C) and (D) are the means ± SD of three independent experiments.

We examined the in vivo effect of the ERdj5/H63Q mutant on HMW complex formation and on the ERAD of J chains. Accumulation of the J chain HMW complex was considerably decreased 48 hours after transfection with wild-type ERdj5 compared with amounts after mock transfection (Fig. 4B). In contrast, neither transfection with ERdj5/H63Q nor with the ERdj5/SS mutant repressed the accumulation of J chain HMW complexes. Transfection with the ERdj5/H63Q mutant did not accelerate J-chain degradation (Fig. 4C and fig. S3) even though this mutant retains the disulfide reductase activity (fig. S2). Thus, both the reductase activity of ERdj5, which is conferred by its CXXC motifs, and the association of ERdj5 with the molecular chaperone BiP are necessary to prevent multimer formation by the misfolded proteins and also to promote efficient ERAD of such proteins.

We next examined the involvement of ERdj5 in the EDEM-mediated ERAD pathway. EDEM coimmunoprecipitated with both wild-type ERdj5 and the ERdj5/AA mutant in HEK293 cells, which suggests that the binding of EDEM with ERdj5 is CXXC-independent (fig. S8). Acceleration of ERAD by EDEM overexpression is dependent on mannose-trimming and is inhibited by kifunensine, an inhibitor of ER mannosidase I (3). Promotion of ERAD by ERdj5 overexpression was totally abolished in the presence of kifunensine (Fig. 4D), which suggests that ERdj5-mediated ERAD requires EDEM to function.

Thus, ERdj5, either overexpressed or endogenous, can serve as a key component for the ERAD of misfolded proteins. The following are all required for ERdj5-mediated ERAD acceleration: (i) the reductase activity of ERdj5, which is conveyed through its CXXC motif; (ii) the binding of ERdj5, through its DnaJ domain, to the ER-resident Hsp70 family chaperone BiP; and (iii) the functional interaction of ERdj5 with EDEM, a lectin-like molecule that may recognize the Man8 N-glycan on misfolded proteins to be degraded. ERdj5-mediated cleavage of intermolecular disulfide bridges decreased the accumulation of covalent multimeric forms of misfolded proteins in the ER; such accumulations are expected to hinder the retrograde transport of misfolded proteins through the retrotranslocation channel. ERdj5 may prevent disulfide-linked aggregation and/or misfolding of substrates by maintaining them in reduced states and may enhance their ERAD by increasing retrotranslocation-competent misfolded proteins.

The redox potential of ERdj5 is even more reducing than the ER redox status (16) or the redox potential of PDI (17). ERdj5 has three CXPC motifs, consistent with reports that the redox potential of CXPC motifs contained in thioredoxin superfamily proteins is reducing (18). Its reducing redox potential indicates that ERdj5 is thermodynamically stable in an oxidized form in the ER redox environment and that ERdj5 can function as a strong disulfide reductase once it accepts electrons from electron donors. Electrons might be transported into the ER from the reducing cytosol or provided from the high ER concentration of the reduced form of nicotinamide adenine dinucleotide phosphate (NADPH) (19). These possibilities remain to be addressed.

Here, we have established the presence of a supramolecular functional ERAD complex, comprising EDEM, ERdj5, and BiP, which have distinct, but linked and concerted, roles (fig. S9). In this model, after the transfer of terminally misfolded proteins from calnexin to EDEM, ERdj5 bound to EDEM cleaves their disulfide bonds, which results in dissociation of the covalent multimeric substrates. At the same time, ERdj5 activates the conversion of the ATP-form of BiP to adenosine diphosphate form, resulting in dissociation of BiP from ERdj5, which, in turn, strongly binds the substrates (20) and holds them in a dislocation-competent state until they are transferred to the retrotranslocation channel.

Supporting Online Material

Materials and Methods

Figs. S1 to S9


References and Notes

View Abstract

Navigate This Article