Role of EDEM in the Release of Misfolded Glycoproteins from the Calnexin Cycle

See allHide authors and affiliations

Science  28 Feb 2003:
Vol. 299, Issue 5611, pp. 1397-1400
DOI: 10.1126/science.1079474


The mechanisms that determine how folding attempts are interrupted to target folding-incompetent proteins for endoplasmic reticulum–associated degradation (ERAD) are poorly defined. Here the α-mannosidase I–like protein EDEM was shown to extract misfolded glycoproteins, but not glycoproteins undergoing productive folding, from the calnexin cycle. EDEM overexpression resulted in faster release of folding-incompetent proteins from the calnexin cycle and earlier onset of degradation, whereas EDEM down-regulation prolonged folding attempts and delayed ERAD. Up-regulation of EDEM during ER stress may promote cell recovery by clearing the calnexin cycle and by accelerating ERAD of terminally misfolded polypeptides.

Proteins synthesized in the endoplasmic reticulum (ER) are N-glycosylated by addition of triglucosylated, branched oligosaccharides at asparagine residues. Rapid glucose trimming by ER glucosidases generates monoglucosylatedN-glycans that combine glucose (Glc), mannose (Man), andN-acetylglucosamines (Glc1-Man9-GlcNAc2-) that facilitate protein folding by mediating association with the lectin chaperone calnexin. Glucosidase II cleaves the last glucose, thereby dissociating glycoproteins from calnexin. If folding has not been completed, the ER-folding sensor uridine diphosphate (UDP)-glucose:glycoprotein glucosyl transferase (UGGT) reglucosylates immature polypeptides for another round of folding attempts in association with calnexin (1). Because of prolonged retention in the ER, folding-incompetent glycoproteins eventually become substrates of the slow-acting α-mannosidase I that tag ERAD candidates with Man8-glycans [reviewed in (2,3)]. Specific inhibition of this mannose-trimming event by kifunensine (Kif) delays both release of ERAD candidates from calnexin and degradation (4–6). Overexpression of EDEM, a putative Man8-binding lectin, accelerates degradation of misfolded glycoproteins in mammalian and yeast ER (7–9).

To understand in more detail the role of EDEM during protein quality control, and to acquire mechanistic insight into how unproductive folding cycles are interrupted to ensure ERAD, we first analyzed degradation of a membrane-bound [the β-site amyloid precursor protein cleaving enzyme isoform expressed in the human pancreas [BACE457 (10)] a lumenal (BACE457Δ), and a nonglycosylated (BACE457ΔNOG) ERAD substrate. The proteins expressed in HEK 293 cells with normal or elevated expression of EDEM were metabolically labeled for 10 min with 35S-labeled methionine and cysteine. After various chase times with unlabeled amino acids, detergent extracts were immunoprecipitated with specific antibodies and the remaining labeled BACE was quantified by band densitometry in reducing SDS-polyacrylamide gels (11).

In control cells, degradation of BACE457 started after a lag phase of about 90 min and then proceeded with a half-life of 4 hours (Fig. 1A). The lag phase is determined by association with calnexin, which assists folding attempts of BACE457 consisting of formation of intramolecular disulfide bonds, and offers temporary protection from the ER disposal machinery (6). In cells with a threefold elevation in expression of EDEM [comparable to that observed after EDEM induction upon unfolded protein response (UPR) (Fig. 1B)] (7), the lag phase before onset of BACE457 degradation was shortened to less than 30 min, and the protein half-life was reduced to about 90 min (Fig. 1A). Earlier onset of degradation and significant reduction of the protein half-life from about 45 to 30 min were also observed for the lumenal glycoprotein BACE457Δ upon EDEM up-regulation (Fig. 1C). Degradation of the lumenal, nonglycosylated BACE457ΔNOG was not affected by EDEM overexpression (Fig. 1E). Thus, EDEM up-regulation accelerated degradation of glycosylated membrane-bound and soluble ERAD substrates but had no effect on the same proteins lacking N-glycans.

Figure 1

Consequences of variations in the amount of lumenal EDEM on BACE degradation. (A) The amount of labeled BACE457 remaining at the end of a chase in control (Control) and EDEM overexpressing (+ EDEM) HEK 293 cells was quantified by SDS–polyacrylamide gel electrophoresis and gel densitometry. Shown is one of at least three representative experiments. (C) Same as in (A) but for BACE457Δ. (E) Same as in (A) but for BACE457ΔNOG. (B) The intracellular concentration of EDEM mRNA has been determined by semiquantitative reverse transcriptase–polymerase chain reaction in cells transfected with a plasmid encoding an EDEM-targeted siRNA and a GFP-targeted siRNA and in cells expressing a level of EDEM comparable to that found upon UPR. Actin mRNA was used as a standard for normalization. (D) Kinetics of BACE457 degradation has been compared in cells with a physiologic amount of EDEM expression (siGFP) and in cells with a reduced amount of EDEM (siEDEM). (F) Kinetics of BACE457Δ degradation has been compared in cells with a physiologic amount of EDEM expression (Mock and siGFP), in cells with elevated expression of EDEM (+EDEM), and in cells with a reduced concentration of EDEM (siEDEM).

RNA-mediated interference (RNAi) was used to reduce the intracellular concentration of EDEM and to determine effects on the ERAD process. To this end, we transfected HEK 293 cells expressing BACE457 or BACE457Δ with a pcDNA3 vector modified for expression of a small interfering RNA (siRNA) hairpin covering nucleotides 1266 to 1284 of human EDEM in sense and antisense orientations (siEDEM) or a control, green fluorescent protein (GFP)-targeted siRNA (siGFP) (11). Twenty-seven hours after transfection, the amount of EDEM protein (supporting online text) and mRNA (siEDEM, 45% compared with controls) (Fig. 1B) was reduced. In these cells, the lag time between synthesis and degradation of the ERAD substrates was prolonged (Fig. 1D) and degradation of both BACE457 and BACE457Δ was significantly slower than in controls (Mock and siGFP) or in cells with elevated EDEM (Fig. 1, D and F). The opposite effects of up- and down-regulation of EDEM on the length of the lag between synthesis and degradation and on the overall degradation kinetics of folding-incompetent glycoproteins showed that at a physiologic concentration EDEM also plays an active role in regulating ERAD.

The electrophoretic mobility of BACE457 and BACE457Δ increased during the chase because of α-mannosidase I–mediated trimming of the protein's N-glycans (Fig. 2A, fig. S1). Accordingly, reduction of the protein's molecular weight did not occur in cells treated with Kif, which also significantly inhibited BACE degradation (Fig. 2A) (6). Moreover, the proteins isolated from cells overexpressing EDEM reproducibly showed slightly faster electrophoretic mobility after 180 (BACE457) and 75 (BACE457Δ) min of chase than those isolated from control cells after the same chase time (Fig. 2A, fig. S1). The difference in the protein molecular weight was minor, disappeared at later chase times (Fig. 2B), and was eliminated upon EndoH-mediated deglycosylation of BACE. Hence, it could be ascribed to a faster processing of BACE'sN-glycans in cells overexpressing EDEM (12). To assess the structure of BACE457Δ N-glycans, we treated the proteins with jack bean α-mannosidase (JB-αM) (11), an exoglycosidase that removes terminal mannoses from N-linked carbohydrates. Core oligosaccharides that retain terminal glucoses are protected from complete mannose removal (Fig. 2B) (13). After a 5-min chase, the electrophoretic mobility of BACE457Δ isolated from cells with normal or elevated EDEM was the same (Fig. 2A) and was poorly affected by JB-αM treatment because both proteins displayed terminal glucoses (Fig. 2B). After 75 min, the mobility of the proteins differed (Fig. 2A, fig. S1) because in cells overexpressing EDEM BACE457Δ was not glucosylated (b′ in Fig. 2B), whereas at least 25% of the protein still displayed glucosylatedN-glycans in control cells (arrow a′ in Fig. 2B). After 90 min, also in control cells, the entire population of labeled BACE was not glucosylated (b′ in Fig. 2B). Thus, in cells with elevated EDEM, glucose persisted for shorter times on BACE's N-glycans. This was likely to affect the kinetics of association of the ERAD candidate with calnexin.

Figure 2

Faster accumulation of nonglucosylated misfolded glycoproteins in cells with increased EDEM. (A) Mobility of BACE457Δ extracted from control cells (lanes 1 and 3); from EDEM overexpressing cells (lanes 2 and 4); and from control cells treated with Kif during starvation, pulse, and chase (lane 5) was analyzed by reducing SDS–polyacrylamide gel electrophoresis. (B) To assess the glucosylation state of BACE457Δ in control cells and in cells overexpressing EDEM, the protein immunoprecipitated from cell extracts was treated for 4 hours at 37°C with (+) or without (−) JB-αM. The mobility shift is exaggerated for nonglucosylated proteins because of more complete mannose trimming (as in the schematic).

To determine the kinetics of BACE457 and BACE457Δ association with calnexin, we subjected cell extracts to sequential immunoprecipitation with calnexin- and BACE-specific antibody (11). Association of BACE457 (Fig. 3A) and BACE457Δ (Fig. 3B) with calnexin was transient. EDEM overexpression did not affect the initial association of both ERAD candidates with calnexin but accelerated their release from the lectin chaperone (Fig. 3, A and B). This was consistent with the shorter persistence of glucoses onN-glycans in cells with elevated EDEM and resulted in faster onset of BACE degradation. siRNA-mediated downregulation of EDEM delayed not only degradation (Fig. 1, D and F) but also release of BACE457 and BACE457Δ from the calnexin cycle (fig. S2). Because release from calnexin does not occur synchronously for all molecules, variation of EDEM also slightly affected the kinetics of degradation that follows the lag phase (Fig. 1).

Figure 3

Consequences of EDEM overexpression on glycoprotein maintenance in the calnexin cycle. (A) BACE457 associated with calnexin was determined at different chase times by consecutive immunoprecipitation in control and EDEM overexpressing cells. (B) Same as in (A) but for BACE457Δ. (C) Same as in (A) but for BACE501.

Calnexin was associated with glucosylated BACE457Δ, whereas most of the protein associated with EDEM was nonglucosylated, as determined by JB-αM treatment, and had slightly faster mobility in SDS gels (Fig. 4A). Our data support a model whereby trapping by EDEM upon release from calnexin may prevent UGGT-mediated reglucosylation of misfolded glycoproteins, thereby interrupting their folding attempts in the calnexin cycle.

Figure 4

Association of BACE457Δ with calnexin and EDEM. (A) Total cell extracts were immunoprecipitated with antibody to calnexin (AntiCnx) or antibody to hemagglutinin (HA)-tagged EDEM (AntiHA-EDEM). Radioactive calnexin and EDEM and coimmunoprecipitated BACE457Δ are shown with arrows. EDEM appears as a doublet of different glycosylated forms (7). BACE457Δ associated with EDEM has a faster electrophoretic mobility. (B) EDEM overexpression does not accelerate glycoprotein degradation in cells with an inactive calnexin cycle. Consequences of EDEM overexpression (lanes 1 and 2) on protein degradation were determined in cells with an inactive calnexin cycle (Lec23 cells) as described for HEK 293 cells. Slower mobility of BACE457Δ (lanes 1 to 4) compared with the protein expressed in HEK cells (lane 5) confirmed lack of glucose trimming in Lec23 cells.

Variation in the amount of intracellular EDEM did not affect the time of association with calnexin and the overall maturation of glycoproteins with normal folding—for example, BACE501 (Fig. 3C, fig. S3A) (10, 14) and vesicular stomatitis virus G protein (fig. S3C). Moreover, only release from calnexin of ERAD candidates was delayed when Kif was used to inhibit the α-mannosidase I–mediated generation of Man8 N-glycans required for optimal EDEM activity (fig. S3B). Thus, EDEM selectively regulated the release from the calnexin cycle of terminal misfolded glycoproteins and not of glycoproteins undergoing productive folding.

To address the consequences of EDEM overexpression in cells with an inactive calnexin cycle, we investigated degradation of BACE457Δ in Lec23 cells that are defective in glucosidase I (15,16). Lec23 cells had a proliferation rate one-third as fast as HEK 293 cells and a significantly reduced rate of protein synthesis compared with HEK 293 cells [this study and (17)]. Protein degradation was also slower (the half-life of BACE457Δ was about 75 min) (Fig. 4B). Overexpression of EDEM did not significantly accelerate the process (Fig. 4B). Experiments performed in HEK 293 cells treated with the glucosidase inhibitor N-butyl deoxynojirimycin (13) to prevent access to the calnexin cycle confirmed that overexpression of EDEM had little, if any, acceleration effect on ERAD of BACE457 and BACE457Δ glycoproteins when the calnexin cycle was bypassed (fig. S1).

In summary, it can be postulated that, by associating with sugar and misfolded protein determinants (fig. S4), EDEM sequestered ERAD candidates released from calnexin, thereby interfering with the UGGT-mediated reglucosylation that maintains misfolded glycoproteins in the calnexin cycle. In this model, extraction from the calnexin cycle is favored upon trimming of N-glycans to the Man8 form that makes ERAD candidates poorer UGGT substrates (1) and better ligands for EDEM (7) and/or by the overexpression of EDEM. Because EDEM is up-regulated when protein folding is perturbed in the ER (7), these mechanisms may be essential for cell recovery upon UPR. They would prevent accumulation of folding-incompetent glycoproteins in the calnexin cycle, thereby granting more effective assistance during maturation of newly synthesized glycoproteins.

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S4

  • * To whom correspondence should be addressed. E-mail: Maurizio.molinari{at}


View Abstract

Navigate This Article