Research article

Demonstration that endoplasmic reticulum-associated degradation of glycoproteins can occur downstream of processing by endomannosidase

Nikolay V. Kukushkin, Dominic S. Alonzi, Raymond A. Dwek, Terry D. Butters


During quality control in the ER (endoplasmic reticulum), nascent glycoproteins are deglucosylated by ER glucosidases I and II. In the post-ER compartments, glycoprotein endo-α-mannosidase provides an alternative route for deglucosylation. Previous evidence suggests that endomannosidase non-selectively deglucosylates glycoproteins that escape quality control in the ER, facilitating secretion of aberrantly folded as well as normal glycoproteins. In the present study, we employed FOS (free oligosaccharides) released from degrading glycoproteins as biomarkers of ERAD (ER-associated degradation), allowing us to gain a global rather than single protein-centred view of ERAD. Glucosidase inhibition was used to discriminate between glucosidase- and endomannosidase-mediated ERAD pathways. Endomannosidase expression was manipulated in CHO (Chinese-hamster ovary)-K1 cells, naturally lacking a functional version of the enzyme, and HEK (human embryonic kidney)-293T cells. Endomannosidase was shown to decrease the levels of total FOS, suggesting decreased rates of ERAD. However, following pharmacological inhibition of ER glucosidases I and II, endomannosidase expression resulted in a partial switch between glucosylated FOS, released from ER-confined glycoproteins, to deglucosylated FOS, released from endomannosidase-processed glycoproteins transported from the Golgi/ERGIC (ER/Golgi intermediate compartment) to the ER. Using this approach, we have identified a previously unknown pathway of glycoprotein flow, undetectable by the commonly employed methods, in which secretory cargo is targeted back to the ER after being processed by endomannosidase.

  • endomannosidase
  • endoplasmic reticulum-associated degradation (ERAD)
  • free oligosaccharide
  • N-glycosylation
  • protein quality control
  • secretory pathway


ERAD [ER (endoplasmic reticulum)-associated degradation] is an essential cellular mechanism for disposal of aberrant and terminally misfolded proteins produced in the lumen of the ER. ERAD is tightly linked to quality control, which, for N-glycosylated proteins, comprising the bulk of the secretory pathway cargo, involves glycan trimming and recognition (for a review, see [1]). Although substantial progress has been made in studying glycoprotein ERAD targeting over the last decade, the process is far from being fully understood. The classic view of ERAD attributes major roles in glycoprotein disposal solely to ER-localized enzymes and lectins, including ER mannosidase I [25], EDEM (ER degradation-enhancing α-mannosidase-like protein) 1–3 [68], OS9 [810], XTP3-B [1013] and others.

Previous evidence suggests the involvement of post-ER compartments in the quality control of glycoprotein and ERAD. Elevation of Golgi mannosidase activity can increase the rates of degradation of mutated glycoproteins [14]. VIP36, a lectin binding high-mannose glycans, was found to recycle between the Golgi and ER together with its glycoprotein cargo, suggesting a role for the lectin in ER quality control [15]. A similar role has been proposed for Bap31, a cargo receptor present in both ER and post-ER compartments [16]. Retrograde Golgi–ER transport of ERAD substrates has been demonstrated in several other studies in mammalian cells [1719] and in yeast [20]. Among others, endomannosidase was proposed to have a role in ERAD [21,22].

Glycoprotein MANEA (endo-α-mannosidase) (EC is a Golgi/ERGIC (ER/Golgi intermediate compartment) enzyme that acts in the early secretory pathway by cleaving an internal α1,2-glucosidic bond between the glucose-substituted mannose residue and rest of the N-linked oligosaccharide. As a result, the glucosyl cap is released from the glycan on the substrate glycoprotein [23,24]. This glucosidase-independent deglucosylation allows for evasion of the glycan maturation blockage imposed by glucosidase inhibitors such as castanospermine or NB-DNJ (N-butyldeoxynojirimycin) [25,26]. Following treatment with these inhibitors, endomannosidase has been shown to restore the ability of secreted glycoproteins to acquire fully processed glycans, whereas in cells deficient in endomannosidase activity, glucosidase inhibition results in secretion of glucosylated oligomannose glycans [27].

Endomannosidase has been proposed to act in the post-ER quality control system on the basis of: (i) the implication that the removal of the terminal mannose residue by endomannosidase would not allow for re-entry of the glycoprotein into the calnexin/calreticulin cycle, mediated by UDP-Glc:glycoprotein glucosyltransferase [28]; (ii) co-localization of endomannosidase with calreticulin and co-purification of the two proteins on an affinity matrix [29]; and (iii) the mutually exclusive subcellular distribution of endomannosidase and glucosidase II, an enzyme participating in the quality control of proteins in the ER involving the calnexin/calreticulin cycle. However, a more recent finding that under the conditions of glucosidase inhibition, endomannosidase non-selectively processes mutated variants of α1-antitrypsin, causing the secretion of the misfolded null-Hong Kong α1-antitrypsin, established that endomannosidase does not discriminate between misfolded and properly folded glycoproteins, therefore denying endomannosidase a role in quality control [30].

Neutral FOS (free oligosaccharides) have been identified previously as biomarkers of ERAD [31]. FOS are generated by a cytosolic PNGase (peptide:N-glycanase) following translocation of degrading glycoproteins from the ER to the cytosol via the sec61 channel [32]. The subsequent catabolism of FOS involves rapid processing by ENGase (endo-β-N-acetylglucosaminidase), cleaving one terminal GlcNAc residue from the reducing end of the free glycan, resulting in GlcNAc1-terminating FOS [33,34], which are then trimmed further by cytosolic mannosidases up to a Man5GlcNAc1 structure [35]. The latter is then transported into the lysosome, where the final stages of FOS degradation are mediated by lysosomal mannosidases [36,37]. There is also evidence of an alternative ERAD pathway, in which glycan release and possibly the subsequent degradation of the polypeptide takes place in the lumen of the ER [3841]. In this case, the resulting GlcNAc2-terminating FOS are then transported into the cytosol and are processed in the same way as the cytosolically generated free glycans [42,43].

Under physiological conditions, the detectable FOS represent the normal cellular level of ERAD. Free glycans are transiently produced and degraded as described above. However, different populations of FOS can be stabilized by various glycosidase inhibitors. ER glucosidase inhibition results in the inability of glucosylated cytosolic FOS to access the lysosome [44], and also the failure of the ER-generated GlcNAc2-terminating glucosylated FOS to escape the compartment [43,45]. Cytosolic mannosidases can be inhibited to block the catabolism of FOS following their release in the cytosol. Finally, inhibition of lysosomal mannosidases results in the accumulation of ERAD-derived oligomannose intermediates that fail to progress further in the degradation pathway [46]. FOS analysis has been employed for the evaluation of ERAD under conditions of glucosidase or mannosidase inhibition, as well as to study ERAD in yeast [20,31,46].

In the present study, we employed FOS analysis as a method to gain a global view of glycoprotein disposal, and provide evidence that following processing by endomannosidase, a fraction of glycoproteins is targeted further for ERAD, as opposed to a separate population of substrates that is stabilized by endomannosidase-mediated deglucosylation.



Tissue culture media were from Gibco/Invitrogen or Sigma. Cell lines were from the A.T.C.C. (Mannassas, VA, U.S.A.). A.R.- and HPLC-grade solvents were from VWR International. Restriction, ligation and PCR enzymes were from NEB UK. Water was Milli-Q grade. pLEXm vector was a gift from Dr Chris Scanlan (Oxford Glycobiology Institute). PEI (polyethyleneimine) was from Polysciences. NB-DNJ was provided by Celltech. YB148 was provided by Dr Yves Blériot (Université de Poitiers, Poitiers, France). NAP-DNJ {N-[6′-(4″-azido-2″-nitrophenylamino)hexyl]-1-deoxynojirimycin} was provided by M. Lee (Oxford Glycobiology Institute). All other reagents were from Sigma, unless otherwise stated.

Cell culture

CHO (Chinese-hamster ovary)-K1 and HEK (human embryonic kidney)-293T cells were cultured at 37 °C in a humidified atmosphere containing 5% CO2. Cultures were maintained in DMEM (Dulbecco's modified Eagle's medium) supplemented with fetal bovine serum (10%), penicillin (100 μg/ml) and streptomycin (100 units/ml).


The following primary antibodies were used: mouse monoclonal against PDI (protein disulfide-isomerase) (2 μg/ml); mouse monoclonal against giantin (1 μg/ml) (both Enzo Life Sciences); mouse monoclonal against GM130 (cis-Golgi matrix protein of 130 kDa) (1.25 μg/ml), mouse monoclonal against BiP (immunoglobulin heavy-chain-binding protein)/GRP78 (glucose-regulated protein of 78 kDa) (0.5 μg/ml) (both BD Biosciences); mouse monoclonal against ERGIC53 (10 μg/ml) (Santa Cruz Biotechnology); rabbit polyclonal against HA (haemagglutinin) tag (1 μg/ml for immunofluorescence, 0.5 μg/ml for immunoblotting) (Sigma); and mouse monoclonal against β-actin (2 μg/ml) (Abcam). The following secondary antibodies were used: goat polyclonal against mouse IgG, HRP (horseradish peroxidase)-conjugate (2 μg/ml); goat polyclonal against rabbit IgG, HRP-conjugate (1 μg/ml) (both Abcam); goat polyclonal against mouse IgG, Alexa Fluor® 633-conjugate (4 μg/ml); and goat polyclonal against rabbit IgG, Alexa Fluor® 488-conjugate (2 μg/ml) (both Invitrogen).

Plasmid construction

Full-length human endomannosidase open reading frame (MANEA) was amplified from an IMAGE cDNA clone (IMAGE:9021009, GenBank® BC137016.1; Geneservice) using the primers 5′-CATTGAATTCTGTGGAATTGTGAGCGGATA3′ (P1) and 5′-TAGCTCTCGAGTTAAGCGTAATCTGGAACATCGTATGGGTAAGAAACAGGCAGCTGGCGATCT-3′ (P2).

The resulting HA-tagged product was PCR-purified and digested with EcoRI and XhoI, and the fragment obtained was ligated into the pLEXm vector following gel purification of both digested vector and insert. Blank plasmid was used as the mock vector.

Transfection of plasmids

Conditions were optimized for both cell lines to ensure highest transfection efficiencies. Cells were seeded into six-well plates at densities of 0.3×106 (CHO-K1) or 0.5×106 (HEK-293T) cells/well. The following day, plasmids were transfected using PEI for CHO-K1 or FuGENE™ 6 (Roche) for HEK-293T cells. For CHO-K1, media were replaced with 4 ml of serum-free DMEM. PEI was mixed with DNA (3 μg of DNA per well) at a phosphate/nitrogen ratio of 1:14 in a NaCl solution (150 mM) and incubated at room temperature (23 °C) for 10 min. The resulting mixture (400 μl) was added dropwise into the wells. Cells were then incubated for 4 h at 37 °C, after which fetal bovine serum was added to the wells to 10%. Compound treatment was started in fresh medium following an overnight incubation. For transfection of HEK-293T cells with shRNA (short hairpin RNA) plasmids, FuGENE™ 6 was mixed with DNA in serum-free medium (4 μl of reagent, 1 μg of DNA per well, total volume 100 μl), incubated for 45 min at room temperature and added dropwise to the cells. Treatment was started the next day by adding the compound from a concentrated stock solution.

Compound treatments

In glucosidase-inhibition experiments, cells were treated with either NB-DNJ at a high concentration (1 mM), or NAP-DNJ at a low concentration (5 μM). The latter compound was reported previously to be a more potent inhibitor of both glucosidases I and II than NB-DNJ [47]; however, in cultured cells at low concentrations, it has been noted to cause the accumulation of primarily monoglucosylated GlcNAc1-terminating FOS (M.J. Lee, personal communication). Additionally, cells were treated with swainsonine (100 μM) to inhibit lysosomal FOS processing, YB148 (Compound 1 [46] at 100 μM), an inhibitor of cytosolic mannosidase, to stabilize the cytosolic pool of FOS, and DTT (dithiothreitol) (2 mM), to induce the UPR (unfolded protein response).

FOS analysis

FOS were extracted from cells, labelled with 2-AA (2-anthranilic acid), and purified by affinity chromatography using ConA (concanavalin A)–Sepharose 4B as described previously [31]. NP (normal-phase)-HPLC was used to separate the purified 2-AA-labelled FOS following a method published previously [48]. GU (glucose unit) values were determined by interpolation from a standard curve obtained by analysing a 2-AA-labelled glucose oligomer standard (partial hydrolysate of dextran).

Western blotting

Following SDS/PAGE of crude cell lysates, proteins were transferred on to PVDF membranes, which were then blocked in 5% (w/v) non-fat dried milk powder in PBS supplemented with 0.2% Tween 20 at 4 °C overnight. Primary and secondary antibodies were diluted in the blocking buffer and incubated with the membrane for 1–2 h, followed by three brief washes and three 5 min washes in PBS supplemented with 0.05% Tween 20. ECL (enhanced chemiluminescence) substrate (GE Healthcare) was used for detection in all cases.

Indirect immunofluorescence

Transfected CHO-K1 cells were grown overnight on coverslips pre-treated with poly-L-lysine bromide (50 μg/ml). Cells were washed with PBS, fixed with 2% (v/v) paraformaldehyde for 30 min, blocked in 5% (w/v) BSA in PBS for 30 min, permeabilized with 1% (v/v) Triton X-100 for 20 min, followed by an incubation with primary antibodies diluted in the blocking buffer for 1 h. Cells were then washed in PBS supplemented with 1% (v/v) Tween 20 three times for 15 min, followed by incubation with secondary antibodies diluted in the blocking buffer for 1 h. After washing, coverslips were mounted on microscopy slides using Vectashield mounting medium with DAPI (4′,6-diamidino-2-phenylindole) (Vector Laboratories). Fluorescence images were obtained using the DeltaVision deconvolution-enhanced wide-field fluorescence microscope (Olympus APO 100×/1.4) coupled to a CoolSNAP HQ camera (Photometrics). Deconvolution was carried out using the Applied Precision Resolve 3D algorithm. Images were analysed using ImageJ software (NIH).

Endomannosidase assay

Endomannosidase assay was carried out as described previously [24,27]. Cells were harvested 48 h post-transfection. Cell pellets were resuspended in 0.1 M sodium/Mes buffer (pH 6.5) containing 0.5 mM DTT and 0.02% sodium azide. Cells were disrupted using a Branson sonifier (three 10 s bursts, setting 3). Postnuclear supernatants (600 g, 10 min, 4 °C) were centrifuged at 50000 rev./min for 20 min at 4 °C using a Beckman TLA 100.3 rotor to obtain the membrane pellets. After washing with the homogenization buffer, pellets were resuspended in same buffer to a concentration of 15 mg/ml. The assay solutions contained EDTA (10 mM), NB-DNJ (2 mM), DMJ (1-deoxymannojirimycin) (2 mM) and Triton X-100 (0.2%). Cell membrane suspensions were pre-incubated with the inhibitors for 1 h, followed by addition of a 2-AA-labelled substrate and subsequent incubation for various durations. Reaction mixtures were then deproteinated by ultrafiltration using 10-kDa cut-off spin filters (Millipore) and analysed by HPLC as described above.

GlcNAc1/GlcNAc2 analysis of glucosylated FOS

GlcNAc1- and GlcNAc2-terminating forms of FOS with longer retention times can co-elute on HPLC. Therefore, to accurately determine the ratio between the two groups of glucosylated FOS, a series of enzyme digests was carried out. To isolate glucosylated FOS from deglucosylated species, total 2-AA-labelled FOS were first digested with jackbean α-mannosidase. Briefly, purified dried FOS were resuspended in citrate buffer (10 mM, pH 4.5) containing zinc acetate (0.2 mM) and 100 μ-units of jackbean α-mannosidase (Oxford Glycobiology Institute), and incubated at 37 °C for 24 h. Samples were deproteinated by ultrafiltration as above, and glucosylated FOS were isolated using preparative HPLC. Eluates were dried, and the resulting FOS were digested 37 °C for 24 h using a recombinant human endomannosidase catalytic fragment generated by expressing a polyhistidine-tagged truncated version of human endomannosidase in Escherichia coli, followed by inclusion body isolation, nickel-affinity purification of GdmCl (guanidinium chloride)-solubilized protein and subsequent refolding, concentration and buffer exchange to 0.1 M sodium/Mes (pH 6.5) (N.V. Kukushkin, unpublished work). Approximately 0.1 μg of purified protein was used per reaction. Following deproteination, reaction products were analysed by HPLC.


Recombinant endomannosidase is functional and localized to the Golgi apparatus

CHO-K1 cells, naturally lacking functional endomannosidase [49], were transiently transfected with HA-tagged human MANEA. Expression was verified by indirect immunofluorescence and Western blotting (Figure 1). Recombinant endomannosidase showed some co-localization with ERGIC53 (ERGIC marker) and to a greater extent GM130 (cis-Golgi marker), but most significantly with giantin, a more distal (cis–medial) Golgi marker. No co-localization with the ER marker PDI was detected (Figure 1A). Overall, the observed distribution of recombinant endomannosidase in CHO-K1 cells agreed with the previously reported localization of the endogenous enzyme in vivo [21].

Figure 1 Recombinant MANEA is functional in CHO-K1 cells

(A) Co-localization of HA-tagged MANEA in transfected CHO-K1 cells with ER (PDI), ERGIC (ERGIC53), cis-Golgi (GM130) and cis–medial Golgi (giantin) markers. Scale bars, 5 μm. (B) Immunoblot showing expression of endomannosidase in MANEA-transfected cells. Bands are likely to represent unmodified (1) and post-translationally modified (2) endomannosidase (not verified experimentally). (C) Endomannosidase activity in transfected cell extracts. 2-AA-labelled Glc1Man5GlcNAc1 (1 and 3) or Glc3Man5GlcNAc1 (2 and 4) were used as substrates in an overnight reaction in the presence of NB-DNJ, DMJ and EDTA. Reaction mixtures were deproteinated and analysed by HPLC. Peak shifts were observed in MANEA-transfected (3 and 4), but not in mock-transfected (1 and 2) cell extracts. (D) HPLC analysis of 2-AA-labelled ConA flowthrough of transfected cell lysates shows an average 5.4-fold increase in Glc3Man1 levels (derived from the results of three independent experiments; the result was significant with P=0.0034)

It has been reported previously that endomannosidase is post-translationally modified by phosphorylation, which occurs in the Golgi apparatus [50]. Western blotting of cell lysates following expression of endomannosidase resulted in a double band similar to the one observed previously, suggesting post-translational processing and therefore probably confirming Golgi localization exhibited by the engineered MANEA construct (assumption without direct experimental proof; Figure 1B).

Enzymatic activity of recombinant endomannosidase was confirmed by an enzyme assay. Post-nuclear membrane extracts of CHO-K1 cells were incubated with fluorescently labelled oligosaccharide substrates Glc1Man5GlcNAc1 and Glc3Man5GlcNAc1 in the presence of glucosidase and mannosidase inhibitors. Following an overnight reaction, HPLC analysis showed no peak shift caused by mock-transfected CHO-K1 extracts. A complete shift of the substrate peaks to a single peak corresponding to Man4GlcNAc1, the predicted endomannosidase product, was observed in the case of MANEA-transfected cells (Figure 1C).

Finally, the action of recombinant endomannosidase in the glycoprotein-processing pathway was verified by analysing ConA affinity chromatography flowthrough following purification of FOS extracted from cells treated with NB-DNJ (1 mM). ConA binds most FOS, with only low-molecular-mass sugars with no more than one α-mannosyl residue generally detected in the flowthrough. A peak with GU=3.47 corresponding to Glc3Man1 [26], a characteristic endomannosidase product in the presence of glucosidase inhibitors, was shown to increase over 5-fold in MANEA-transfected cells (Figure 1D). This peak was confirmed further to correspond to a tetrasaccharide by MALDI–TOF (matrix-assisted laser-desorption/ionization–time of flight) (results not shown).

Endomannosidase causes a decrease in total FOS and a relative increase in small FOS concentrations following glucosidase inhibition

Glucosylated FOS generated in the presence of glucosidase inhibitors are a valuable analysis tool primarily due to their arrested catabolism. In our experiments, however, glucosylated FOS were also employed to discriminate between the conventional ERAD, in which the degradatory pathway is confined to the ER, and a potential route involving shuttling the ERAD substrates between ER and Golgi before their degradation. In the former case, processing by glucosidases is the only possible way of removing glucose residues from glycoproteins destined for degradation, resulting in the production of glucosylated FOS following glucosidase inhibition. In contrast, the involvement of post-ER compartments would allow for deglucosylation by endomannosidase. In wild-type CHO-K1 cells, however, endomannosidase activity is absent and therefore FOS produced in the presence of glucosidase inhibitors were expected to be predominantly glucosylated whether or not circulation of ERAD substrates between the ER and post-ER compartments takes place. We therefore asked whether endomannosidase expression in CHO-K1 cells would be reflected in the production of glucosylated/deglucosylated FOS. To address this question, we analysed ConA-binding FOS generated in MANEA-transfected CHO-K1 cells in the presence of glucosidase inhibitors. Following 24 h of treatment with NB-DNJ (1 mM), the major peaks in CHO-K1 FOS HPLC profiles corresponded to triglucosylated oligosaccharides Glc3Man5GlcNAc1 and Glc3Man7GlcNAc2 (Figure 2A, and see Supplementary Table S1 at Endomannosidase expression did not change the ratio between the dominant FOS peaks. However, the total levels of FOS were decreased significantly (Figure 2B). This decrease was composed of oligomannosidic glycans with five or more mannose residues, but primarily all glucosylated oligosaccharides (‘large FOS’, Figure 2A). At the same time, the relative contribution of FOS with shorter retention times (‘small FOS’) to the total FOS pool appeared to be increased in endomannosidase-transfected CHO-K1 cells (Figures 2C and 2D). FOS status of untransfected cells, untreated or treated with NB-DNJ, was also assessed and in most cases was not different from mock-transfected cells. A decrease in the levels of Glc3Man5GlcNAc1 and a concomitant increase in the levels of Glc3Man4GlcNAc1 were significant in mock-transfected CHO-K1 cells; however, the differences observed were negligible compared with those between mock- and MANEA-transfected cells (see Supplementary Table S2 at

Figure 2 Endomannosidase effect on FOS in NB-DNJ-treated cells

(A) HPLC analysis of FOS extracted from mock- or MANEA-transfected CHO-K1 cells incubated for 24 h without inhibitors (u/t), with NB-DNJ or with NB-DNJ and swainsonine (SW). Identified large FOS peaks are numbered and structurally annotated as shown in Supplementary Table S1 at Small FOS were analysed further as discussed below. Peaks were labelled and structurally annotated as shown in Table 1. (B) Total amounts of FOS in the cells, plotted as pmol/mg of total protein. (C) Amounts of small FOS in the cells, plotted as in (B). (D) Proportion of small FOS in the cells, plotted as percentages relative to total FOS. All results shown are representative of three independent experiments. Results in histograms are means±S.E.M. (n=3). **P<0.01, ***P<0.001 (two-tailed unpaired Student's t tests).

Small FOS correspond to putative endomannosidase products

To determine the origin of the small FOS that were relatively increased following endomannosidase transfection in CHO-K1 cells, the corresponding GU values were determined and used to compare with those previously obtained for a range of cellular FOS [31]. To confirm further the structure of the glycans, the eluted small FOS were digested with α1,2-mannosidase, and the migration of the peaks was analysed by HPLC. Five peaks were identified as different isomers of GlcNAc1-reducing Man2-Man4 oligosaccharides (Table 1). Of the FOS identified, only a single species bore the D1 mannose residue (Man4GlcNAc1, linear isomer; see Table 1, D1 mannose shown in bold). In sharp contrast with the other identified Man2-Man4 oligosaccharides, this linear isomer of Man4GlcNAc1 was not increased in endomannosidase-expressing cells (Table 1). Since the D1 mannose residue is removed by endomannosidase, the selective increase in the production of FOS lacking this residue suggests that they have originated from endomannosidase-processed glycoproteins. It was suggested previously that endomannosidase-mediated deglucosylation acts as a ‘bypass mechanism’ allowing glycoproteins to evade quality control without compromising their end glycosylation. However, our results hereby confirm the existence of an alternative ERAD pathway in which glycoproteins are exposed to post-ER compartments (ERGIC, possibly Golgi) and specifically endomannosidase before their degradation.

View this table:
Table 1 Characterization of small FOS in transfected CHO-K1 cells treated with NB-DNJ

Transfected CHO-K1 cells were treated with NB-DNJ (1 mM) for 24 h, followed by FOS extraction, fluorescent labelling and analysis by HPLC (Figure 2). Peak areas were analysed and compared for mock- and MANEA-transfected cells using two-tailed Student's t tests. The nomenclature used refers to the structure below. Results are means±S.E.M.

In contrast with FOS produced in the conventional ER-confined pathway under conditions of glucosidase blockage, the glycans released from ERAD substrates processed by endomannosidase are deglucosylated and therefore do not accumulate in the cell. They are rapidly processed by cytosolic and then lysosomal enzymes. The small FOS detected are therefore transient and their levels could not be quantitatively compared with those of glucosylated FOS generated as a result of the conventional glycoprotein ERAD. Hence, the decrease in total FOS observed in endomannosidase-expressing CHO-K1 cells could be explained by a greater deglucosylated/glucosylated FOS ratio and consequently higher susceptibility of the total FOS pool to cytosolic and lysosomal glycosidases. However, endomannosidase-mediated deglucosylation could also prevent some glycoproteins from entering the degradation pathway, resulting in decreased overall FOS release. Such stabilization of glycoproteins by endomannosidase in the conditions of glucosidase inhibition has been demonstrated previously for mutated variants of α1-antitrypsin [30].

To investigate whether both of the ‘FOS-destabilizing’ and the ‘glycoprotein-stabilizing’ mechanisms take place in the cell at the same time, we employed swainsonine, a mannosidase inhibitor primarily used for late-Golgi glycoprotein processing perturbation, but also known to be a potent lysosomal mannosidase inhibitor [46]. Swainsonine therefore arrests the catabolism of FOS at late stages, predominantly causing the accumulation of the core-type Man3GlcNAc1 (Figure 2A). Following co-treatment of transfected CHO-K1 cells with swainsonine and NB-DNJ, the levels of FOS were decreased in endomannosidase-expressing cells, but to a significantly lesser extent than in cells treated with NB-DNJ alone (approximately 44% of the decrease seen in the latter case; Figure 2B). As expected, the proportion of small FOS was increased in MANEA-expressing cells in both cases (Figures 2C and 2D). Since the catabolism of FOS is blocked in the presence of swainsonine [46], the extent of FOS depletion still observed upon swainsonine/NB-DNJ co-treatment was indicative of decreased production rather than increased degradation of FOS, supporting the previous finding that the secretion of glycoproteins is enhanced by endomannosidase under conditions of glucosidase inhibition [30]. We therefore conclude that both stabilization of glycoproteins and destabilization of FOS contribute to the decrease in FOS levels observed in endomannosidase-transfected CHO-K1 cells.

Glucosylated luminal FOS are not the primary source of small FOS

Glucosidase inhibition by NB-DNJ in a high concentration results in the accumulation of two types of glucosylated FOS, namely cytosolic and luminal. The latter fail to get transported across the ER membrane and could potentially become subject to ER/Golgi vesicular trafficking and endomannosidase-mediated deglucosylation [51]. Owing to the rapid action of ENGase in mammalian cells, cytosolic and luminal FOS are distinguished by the presence of either one or two GlcNAc residues on the reducing terminus respectively. To confirm that the small FOS up-regulated following endomannosidase expression under conditions of NB-DNJ treatment (see Figure 2) are not derived from luminal FOS, we employed NAP-DNJ, a potent NB-DNJ analogue which at low concentrations has been noted to cause the accumulation of primarily monoglucosylated GlcNAc1-terminating oligosaccharides, with hardly any GlcNAc2-terminating FOS produced (Figure 3). To verify this observation, we evaluated the ratio between GlcNAc1- and GlcNAc2-terminating glucosylated oligosaccharides in NB-DNJ- (1 mM) and NAP-DNJ- (5 μM) treated cells. For this purpose, glucosylated FOS were separated from deglucosylated species, and, following digestion with recombinant endomannosidase, the levels of the resulting Man3/4GlcNAc1/2 were compared. These final products are clearly distinguishable by HPLC, facilitating discrimination between GlcNAc1 and GlcNAc2-terminating FOS (Figure 3B). In cells treated with 5 μM NAP-DNJ, GlcNAc1-terminating glycans comprised approximately 90% of total glucosylated FOS, compared with approximately 55% in cells treated with 1 mM NB-DNJ. These results confirm the observation that treatment of CHO-K1 cells with NAP-DNJ at a low concentration results in the accumulation of glucosylated FOS that are almost exclusively cytosolic. At the same time, endomannosidase expression together with the treatment causes an approximately 50% relative increase in non-D1-terminating Man2-Man4 oligosaccharides (Figures 3D and 3E). We therefore conclude that the primary source of small FOS is not glucosylated luminal FOS, but glycoproteins undergoing ERAD through the route involving shuttling between the ER and post-ER compartments.

Figure 3 Independence of endomannosidase effect from glucosylated luminal FOS

(A) HPLC analysis of FOS extracted from mock- or MANEA-transfected CHO-K1 cells incubated for 24 h with NAP-DNJ (5 μM). (B) HPLC analysis of 2-AA-labelled oligosaccharides obtained by digesting the FOS extracted from NAP-DNJ- or NB-DNJ-treated cells with jackbean α-mannosidase, followed by isolation of glucosylated FOS and further digestion with recombinant endomannosidase catalytic fragment. The resulting peaks are structurally annotated. (C) Quantification of total FOS (pmol/mg of total protein). (D) Amounts of small FOS in the cells, plotted as in (B). (E) Proportion of small FOS in the cells, plotted as percentages relative to total FOS. All results shown are representative of three independent experiments. Results in histograms are means±S.E.M. (n=3). **P<0.01 (two-tailed unpaired Student's t tests). u/t, untreated.

The effect of endomannosidase on FOS is dependent on substrate glucosylation

To verify that the processing of glucosylated substrates by endomannosidase is the cause of the changes in the FOS profiles of CHO-K1 cells expressing the enzyme, we replaced the glucosidase inhibitors with either DTT, a commonly used UPR-inducing agent, or YB148, a cytosolic mannosidase inhibitor arresting the catabolism of FOS following their release in the cytosol [46]. The effects of these compounds on FOS paralleled those of glucosidase inhibitors, namely induction of UPR and increasing the rates of ERAD [46] or stabilizing the cytosolic pool of FOS respectively. However, glucosidase-mediated deglucosylation of N-linked glycans was unaffected (Figure 4 and see Supplementary Figure S1 at Consistent with the proposed mechanism of endomannosidase impact on FOS, the effect of the enzyme on FOS in cells treated with YB148 or DTT was much less pronounced than in the case of glucosidase inhibitors. Endomannosidase expression in YB148-treated CHO-K1 cells caused a subtle but significant decrease in the amounts of total FOS (Figures 4A and 4B). No changes in the relative amounts of Man2-Man4 oligosaccharides were detected, consistent with the YB148-mediated arrest of FOS catabolism on the early stages (Figures 4C and 4D). The effect of endomannosidase in DTT-treated cells was not significant, although the expected up-regulation of BiP/GRP78 was observed, indicating the induction of UPR (see Supplementary Figure S1). Taken together, these findings confirm that the observed effect of endomannosidase expression in CHO-K1 cells is a specific response to the up-regulation of glucosylated substrates induced by glucosidase inhibitors in our experimental system.

Figure 4 Endomannosidase effect on FOS in the presence of YB148

(A) HPLC analysis of FOS extracted from mock- or MANEA-transfected CHO-K1 cells incubated for 24 h with YB148. (B) Quantification of total FOS (pmol/mg of total protein). (C) Amounts of small FOS in the cells, plotted as in (B). (D) Proportion of small FOS in the cells, plotted as percentages relative to total FOS. All results shown are representative of three independent experiments. Results in histograms are means±S.E.M. (n=3). *P<0.05, **P<0.01 (two-tailed unpaired Student's t tests). u/t, untreated.

Endomannosidase knockdown in HEK-293T cells leads to increased FOS production

To verify our findings of the effect of endomannosidase on FOS production, we performed shRNA-mediated knockdown of the enzyme in HEK-293T cells. Four shRNA-encoding vectors (pGFP-V-RS, #1–#4, OriGene) were tested for their ability to knock down endomannosidase. Before the knockdown experiments, two commercially available antibodies (Sigma catalogue numbers HPA011046 and SAB2101423) were tested in Western blotting experiments. We were, however, unable to detect specific bands at the predicted molecular mass for any of the cell lines tested, suggesting that endogenous endomannosidase in the cells used was below the level of immunological detection. The latter forced us to verify efficient knockdown by co-transfecting HEK-293T cells with the shRNA-encoding vectors and the MANEA–HA plasmid, similar to the strategy recently employed for ER mannosidase I [7]. HA-tagged endomannosidase expression was then assayed using Western blotting (Figure 5A).

Figure 5 Effect of endomannosidase knockdown on FOS production

(A) Western blot analysis of MANEA–HA overexpression in HEK-293T cells upon co-transfection with MANEA/mock plasmids and pGFP-V-RS vectors containing shRNA-encoding inserts (#1–#4), a scrambled insert (scr) or no insert (empty). (B) Endomannosidase activity in transfected cell membrane extracts. 2-AA-labelled Glc3Man5GlcNAc1 (1 pmol in 10 μl) was used as a substrate in the presence of NB-DNJ, DMJ and EDTA. Reactions were carried out for various durations, and samples were deproteinated and analysed by HPLC. (C) HPLC analysis of 2-AA-labelled FOS extracted from HEK-293T cells transfected with vectors containing shRNA insert #4 or the scrambled insert and treated with NB-DNJ (1 mM) for 24 h. (D) Relative amounts of Man6GlcNAc1/2, Glc3Man5GlcNAc1, Glc3Man7–9GlcNAc1/2 in NB-DNJ-treated HEK-293T cells transfected with #4/scrambled, plotted as percentages relative to total FOS, and total levels of FOS extracted from the cells, plotted in pmol/mg of total protein. Results shown are representative of three independent experiments. (E) HPLC analysis of 2-AA-labelled FOS extracted from HEK-293T cells treated with NB-DNJ (1 mM) and incubated with either PBS (control) or human endomannosidase catalytic fragment (digest) for 16 h. Results in histograms are means±S.E.M. (n=3). *P<0.05, **P<0.01 (two-tailed unpaired Student's t tests).

Out of the four shRNA constructs tested, #4 appeared to be most efficient (>90% attenuation of MANEA–HA overexpression). Knockdown was verified by an activity assay using a triglucosylated substrate. Breakdown of Glc3Man5GlcNAc1 (1 pmol in 10 μl) in cell extracts was slowed down dramatically following transfection with #4. At the linear phase of the reaction, the slope of the kinetic curve was decreased 9.7-fold (Figure 5B).

The construct was employed in subsequent experiments alongside the scrambled control vector. HEK-293T cells transfected with the shRNA-encoding plasmid were treated with NB-DNJ (1 mM), and FOS were analysed as described for CHO-K1 cells. Endomannosidase knockdown resulted in both an absolute increase in total FOS produced, and a relative increase in glucosylated oligosaccharides (Figures 5C and 5D). The latter included Glc3Man5GlcNAc1, representing the major product of glucosylated FOS cytosolic processing [31], and Glc3Man7–9GlcNAc1/2, a group of incompletely demannosylated luminal and cytosolic oligosaccharides. At the same time, a relative increase was observed in some of the deglucosylated FOS, most strikingly the Man6GlcNAc1/2 isomer A, representing a characteristic product of Glc3Man7GlcNAc1/2 deglucosylation by endomannosidase, or demannosylation of the equivalent Man7 and Man8 species (generated from Glc3Man8GlcNAc1/2 or Glc3Man9GlcNAc1/2 species respectively). Peak identity was based on the reported GU value for the corresponding oligosaccharide [31], and the structure was confirmed by digesting the total 2-AA-labelled FOS with the recombinant endomannosidase catalytic fragment, resulting in the shift of the Glc3Man7GlcNAc1/2 GU value to that of the peak down-regulated following endomannosidase knockdown (Figure 5E).

Although we did not observe statistically significant changes in small FOS corresponding to Man2–4GlcNac1, the decreased level of Man6GlcNAc1/2 isomer A reflects the reduction in endomannosidase activity in a similar fashion to the increase in the relative Man2–4 FOS levels following recombinant endomannosidase expression in CHO-K1 cells. The less pronounced changes in the levels of FOS in HEK-293T cells may be due to cell-specific kinetics of cytosolic and lysosomal FOS degradation, which is supported by the fact that Man7–9 FOS are primarily associated with the cytosol and become subject to active cytosol–lysosome transport in the <Man7 state [52]. It is therefore likely that Man6GlcNAc1/2 represents the transition point between the two stages of FOS catabolism, resulting in higher levels of the oligosaccharide in the steady state and greater susceptibility to endomannosidase knockdown.

In summary, by employing the method of FOS analysis to gain a generalized insight into ERAD, we have obtained results suggesting that endomannosidase suppresses degradation of a fraction of glycoproteins in the presence of glucosidase inhibitors. This supports the reported evidence on abnormal endomannosidase-mediated secretion in castanospermine-treated cells [30]. However, we additionally demonstrate that a significant population of glycoproteins processed by the enzyme still becomes subject to ERAD (Figure 6). This finding underlines that, on a global level, misfolded proteins are able to recycle between the ER and post-ER compartments (ERGIC and possibly Golgi) before their degradation. Although such recycling probably does not depend on additional quality control checkpoints, the uncertain fate of endomannosidase-processed misfolded glycoproteins does not exclude the possibility that additional sorting mechanisms operate at a post-ER location.

Figure 6 Proposed model for the endomannosidase-mediated ERAD pathway

Under physiological conditions (upper panel), nascent glycoproteins are deglucosylated predominantly by ER glucosidases I and II (ER Gls) as part of ER quality control (1). Aberrant glycoproteins are subjected to the conventional ER-confined degradation pathway, resulting in the production of deglucosylated FOS (2). Normal glycoproteins proceed further to the post-ER compartments for maturation and secretion (3). A fraction of glucosylated glycoproteins escapes the quality control in the ER and is transported to ERGIC/Golgi, allowing for deglucosylation by endomannosidase (MANEA) (4). The resulting glycoproteins then enter normal maturation pathways (5). Alternatively, endomannosidase-processed glycoproteins can be recycled back to ER (6) and become subject to ERAD targeting, producing FOS (7). However, under physiological conditions, the FOS generated in the latter pathway are indistinguishable from those produced in the conventional ER-confined ERAD. This is opposed to the conditions of glucosidase inhibition (lower panel). In this case, conventional ERAD results in the production of glucosylated FOS that accumulate in the cytosol (8). The endomannosidase-mediated pathway retains a functional deglucosylation step, therefore resulting in deglucosylated FOS (7), and is eventually subjected to complete degradation via Man2-Man4 small FOS (9).


Nikolay Kukushkin performed the experiments. Terry Butters, Dominic Alonzi and Nikolay Kukushkin designed experiments, analysed data and wrote the paper. Raymond Dwek oversaw the research and revisions of the paper.


This work was funded by the Oxford Glycobiology Institute. N.V.K. was supported by Clarendon Fund and New College (Robert Lyns) graduate scholarships.

Abbreviations: 2-AA, 2-anthranilic acid; BiP, immunoglobulin heavy-chain-binding protein; CHO, Chinese-hamster ovary; ConA, concanavalin A; DMEM, Dulbecco's modified Eagle's medium; DMJ, 1-deoxymannojirimycin; DTT, dithiothreitol; ENGase, endo-β-N-acetylglucosaminidase; ER, endoplasmic reticulum; ERAD, ER-associated degradation; ERGIC, ER/Golgi intermediate compartment; FOS, free oligosaccharide(s); GM130, cis-Golgi matrix protein of 130 kDa; GRP78, glucose-regulated protein of 78 kDa; GU, glucose unit; HA, haemagglutinin; HEK, human embryonic kidney; HRP, horseradish peroxidase; MANEA, endo-α-mannosidase; NAP-DNJ, N-[6′-(4″-azido-2″-nitrophenylamino)hexyl]-1-deoxynojirimycin; NB-DNJ, N-butyldeoxynojirimycin; PDI, protein disulfide-isomerase; PEI, polyethyleneimine; shRNA, short hairpin RNA; UPR, unfolded protein response


View Abstract