Protein folding and quality control at the endoplasmic reticulum (ER) of a cell is a highly regulated process that ensures proper cell functioning. Despite specialised mechanisms, a considerable fraction of newly synthesised polypeptides fail to attain their native conformation and need to be targeted for degradation [1
]. This process involving the recognition, retro-translocation, and proteolysis in the cytosol by the ubiquitin-proteasome system is generally termed endoplasmic reticulum-associated protein degradation (ERAD) [2
]. While the last step is relatively well documented, the recognition and retro-translocation of ERAD substrates are currently under intense scrutiny. Following the model of protein folding assisted by lectin chaperones, such as calnexin and calreticulin, it has been proposed that the mannose trimming of N-glycans exposed on misfolded polypeptides is a signal for degradation via ERAD [3
]. Mannose processing proteins such as ER mannosidase I and EDEM (ER-degradation enhancing α-mannosidase-like) family of proteins have been shown to catalyse mannose trimming for glycans exposed on partially folded or misfolded glycoproteins, therefore accelerating their degradation [6
ERAD is a dynamic process that has been shown to involve both ER luminal (lectins: OS-9 and XTP3-B; disulfide isomerases: ERdj5; or co-chaperones: ERdj3 and ERdj4), as well as ER-membrane proteins (adaptor protein-SEL1L, E3 ubiquitin ligase-HRD1), which work in concert to select and dislocate misfolded polypeptides from the ER to the cytosol for proteasomal degradation [9
]. Many other proteins have been proposed to function as part of ERAD, and different clusters mainly concentrated around E3-ubiquitin ligases could function independently in protein degradation [10
]. To date, EDEM proteins have been functionally associated with the HRD1-nucleated complex and have been shown to associate with the adaptor protein of HRD1: SEL1L [11
Assigning a role for the mammalian homologues of the yeast Htm1, EDEM1 and the other two members of the EDEM family in ERAD has been highly controversial. EDEM1 has been described as an ER resident protein whose expression is under the control of the unfolded protein response (UPR) and extracts misfolded polypeptides from the calnexin cycle as a first step of ERAD [3
]. Its structural homology with the ER mannosidase has led to the hypothesis that the mannosidase-like domain recognises high-mannose N-glycans attached to proteins [4
]. However, several reports have suggested that EDEM1 associates with ERAD substrates independently of its mannosidase domain [15
]. We showed that besides the structured mannosidase domain, this protein has an intrinsically disordered region (IDR) that has been predicted to facilitate protein–protein interactions [19
]. In addition, more recent reports have stated that recognition of the glycosylated substrates by EDEMs is favoured by their unfolded status, thus supporting the idea of the protein–protein interaction of EDEM1 with misfolded proteins [20
EDEM1 is also associated with processes such as the degradation of orphan oligomeric subunits, the formation of aberrant oligomeric structures [21
] and the formation of LC3-positive structures after virus infection, required for virus replication [25
] or, in other cases, for its own turnover [28
]. All these suggest that EDEM1 might also function in concert with other proteins independent of the canonical ERAD pathway.
Here we show that EDEM1 is found in dynamic complexes with auto-regulatory function and associates with several canonical ERAD proteins. Deletion of EDEM1 N-terminal IDR impaired its capacity to bind misfolded ERAD substrates and implicitly blocked their EDEM1-induced degradation. We also found that the absence of IDR moderately reduced the association of EDEM1 with some ERAD components whilst enhancing others, supporting the hypothesis that IDR mediates misfolded ERAD client degradation with a higher specificity.
Additionally, we found that EDEM1 overexpression accelerated the degradation of misfolded polypeptides even when proteasomal degradation was severely impaired. We propose this takes place by the generation of protein aggregates and recruitment of the cytosolic autophagy machinery to degrade these structures via ER-phagy receptors, a process coordinated by overexpressed EDEM1 to alleviate ER burden.
Protein quality control in the endoplasmic reticulum is a tightly regulated process, and any dysregulation of this process can lead to pathological conditions. In particular, the process of recognition and targeting for the degradation of misfolded proteins is not completely elucidated, despite the sustained efforts of many labs in the last few years. ERAD has been described as a tightly regulated and dynamic process that ensures the endoplasmic reticulum disposes the excess folding-incompetent proteins to overcome proteomic burden. EDEM proteins have been described as key players for ERAD by acting as mediators between the folding and degradation machinery, prevalently recognising glycoproteins. EDEM1 was initially proposed to extract glycoproteins from the calnexin folding cycle and target them for proteasomal degradation by association with ERAD components to facilitate their dislocation to the cytosol [4
]. However, our lab and others have shown that EDEM1 is able to bind and accelerate the degradation of non-glycosylated proteins or in conditions where glycan recognition is blocked [15
]. This has been attributed to the presence of an intrinsically disordered region at its N-terminus, which mediates the association with misfolded proteins based on their capacity to expose hydrophobic patches [19
We documented using mass spectrometry, that the proteins co-enriched with EDEM1 are involved in folding and glycosylation, quality control, and protein degradation, suggesting EDEM1 could form different complexes, thus having versatile functions. It is worth mentioning that using LC–MS/MS analysis, we found a relatively reduced number of interactors involved in the ER folding system, such as calnexin, glucosyltransferase (UGGT1 and UGGT2), folding enzymes (Erp44), and the lectins calreticulin and ERGIC-53. In exchange, a significantly higher number of ER resident proteins associated with the ERAD pathway were detected as EDEM1-associated proteins. Our experiments suggest that the stability endogenously expressed EDEM1 is modulated by the abundance and complex stoichiometry of proteins shown to be involved in ERAD substrate degradation. Knock-down experiments indicated that the level of endogenously expressed EDEM1 is modulated by the OS-9-SEL1L-HRD1 complex, with SEL1L and HRD1 being described as key players in ERAD [31
]. Similarly, some proteins maintaining ER homeostasis are also subjected to continuous turnover by the ERAD pathway, as previously reported for IRE1α and ATF6 [41
]. The hypothesis that ERAD is dynamic with auto-regulatory function has also been recently proposed, thus strengthening our observations that EDEM1 could form auto-regulatory complexes along with OS-9, SEL1L, and HRD1 [34
Further, we analysed the association of EDEM1 and its IDR-lacking mutant with several ERAD substrates, since our previous results showed this region was essential for EDEM1–tyrosinase interaction [19
]. Our results indicate the intrinsically disordered region of EDEM1 mediates protein–protein interactions with ERAD clients, and the deletion of this domain does not abolish, but does significantly reduce the association with α-1AT, NHK, BACE-476, and Ri-332. A reduced binding of ∆-EDEM1 to ERAD clients was correlated with a less efficient degradation induced by the overexpression of ∆-EDEM1 compared to EDEM1, thus confirming our previous report that weak binding of the substrate leads to a lower efficiency in degradation. Moreover, we also confirmed these results using an EDEM1-deficient cell line, which clearly confirmed the dependence of misfolded protein degradation on the presence of EDEM1-IDR. We do not exclude that the mannosidase-like domain may also bind the substrates, as previously reported [58
]; however, the IDR of EDEM1 ensures the specificity and efficiency of degradation.
Considering our results showed that the IDR of EDEM1 is important for binding ERAD substrates, we also explored the idea it might affect association with components of the ERAD pathway. We found—using mass spectrometry, immunoprecipitation, and Western blotting—that some interactions with ERAD components are reduced for ∆-EDEM1 (e.g., SEL1L, OS-9, ERLEC1/XTP3-B, GRP94/HSP90B1, and HRD1/SYVN1), while others are either not-affected or even consolidated (e.g., DNAJC10, CCDC47, ANKZF1, ERLIN1, and ERLIN2). This might be explained by the less efficient association of ∆-EDEM1 with ERAD substrates, which suggests a reduced activity of ∆-EDEM1 in protein degradation and, implicitly, a lower association with partner proteins from ERAD or that it associates with selected proteins to increase its stability.
With all these results in hand, we were next interested to know whether EDEM1 activity specifically requires partner proteins from ERAD, since, as described above, it forms functional complexes with other ERAD proteins. To our surprise, we found that the overexpression of EDEM1, and not ∆-EDEM1, was able to accelerate the degradation of ERAD substrates when either SEL1L, HRD1, OS-9, or XTP3-B were silenced, suggesting EDEM1 most likely targets ERAD substrates for degradation, independent of the SEL1L-HRD1 complex in this case. Furthermore, we discovered that EDEM1 was able to accelerate substrate degradation even when proteasomal degradation was severely impaired by silencing of p97/VCP. Therefore, we hypothesised that EDEM1-induced degradation, when ERAD is impaired, may occur via autophagy, the alternative degradation pathway in eukaryotes.
To test this hypothesis, we used differential fractionation and immunoprecipitation to investigate whether EDEM1 has functional connections with the autophagy machinery. We found several of the proteins involved in autophagy to either co-fractionate or co-precipitate with EDEM1 and ∆-EDEM1. Our results show that EDEM1, and ∆-EDEM1 with less efficiency, induced the formation of protein aggregates, likely due to the presence of its IDR, which has the capacity to drive self-assembly, as extensively reported for other systems [53
]. Furthermore, we identified three of the six reported ER-phagy receptors as potential interactors of EDEM1 that could mediate the degradation of EDEM1-induced aggregates by recruiting the autophagy cytosolic machinery [47
]. All these results allowed us to speculate and propose that EDEM1 functions in dynamic ERAD complexes; however, in the absence of a functional ERAD or efficient proteasomal degradation, EDEM1 overexpression leads to the efficient formation of protein aggregates that are recognised by the autophagy pathway, thus restoring ER homeostasis.
In conclusion, we identified EDEM1 as part of auto-regulatory complexes in ERAD and the crucial role of the N-terminal IDR for substrate binding and accelerating their degradation, as well as productive association with some ERAD components. We also showed that EDEM1 overexpression bypasses the requirement for proteasomal degradation by driving the formation of amyloid-like oligomers and most likely recruiting the cytosolic autophagy machinery to degrade these aggregates, via association with ER-phagy receptors.
4. Materials and Methods
Reagents, antibodies, and plasmids: pcDNA3.1-α-1AT-HA and pcDNA3.1-NHK-HA, and pcDNA3.1-BACE-476 were a kind gift of M. Molinari (IRB, Bellinzona, Switzerland), pCI-neo-Ri-332 was generated in the lab of N.E. Ivessa, and all other plasmids were described previously [19
Rabbit α-EDEM1 (E84060-Sigma-Aldrich, St. Louis, MO, USA) was used for Western blotting, and goat α-EDEM1 (sc-27891) was used for immunoprecipitation; goat α-SEL1L (sc-48081), MAN1B1 (sc-393145), BiP (sc-166490), GAPDH (sc-81545) were from Santa Cruz Biotechnology (Dallas, TX, USA); rabbit α-BACE1 (ab2077), rabbit α-calnexin (ab22595), rabbit α-OS-9 (ab19853), rabbit α-XTP3-B (ab181166), rabbit α-tubulin (ab18251), and rabbit α-Grp94 (ab3674) were from Abcam (Cambridge, UK); mouse α-LC3 (0231-Nanotools, Teningen, Germany) and rabbit α-1AT (A0012) was from Dako (Jena, Germany); rabbit α-ATG5 (12994) and rabbit α-HRD1 (14773) were from Cell Signaling (Leiden, Netherlands); and mouse α-actin (612657-BD Biosciences, San Jose, CA, USA) and rabbit α-Ribophorin I antibodies were described previously [59
]. All siRNAs were from Santa Cruz Biotechnology, as follows: siRNA SEL1L (sc-61514), siRNA OS-9 (sc-96230), siRNA XTP3-B (sc-94979), siRNA HRD1 (sc-76620)., siRNA VCP (sc-37187), MG132 (sc-201270), and kifunensine (sc-201364). All other chemicals were from Santa Cruz Biotechnology (Dallas, TX, USA) unless specified.
Cell culture and transfection: HEK293T, HEK293T-KO, and HeLa cells were cultivated in DMEM (cat: 10566-032) supplemented with 10% FBS (cat: 10270-098) from Gibco (Life Technologies, Paisley, UK). 24 h post-seeding, the cells were transfected using Lipofectamine 2000 (Invitrogen-Life Technologies, Paisley, UK) or polyethylenimine (PEI) (Sigma-Aldrich St. Louis, MO, USA,) (2:1 v/w ratio transfection reagent:DNA), according to manufacturer’s protocol, and they were harvested after 48 h. For siRNA transfection, Lipofectamine RNAiMAX (Invitrogen-Life Technologies, Paisley, UK) was used as the transfection reagent (1:1 v/v ratio transfection reagent:siRNA), and the cells were harvested after 72 h.
CRISPR/Cas9 generation of EDEM1-KO cell line: HEK293T cells were co-transfected with CRISPR/Cas9 KO-specific plasmids from Santa Cruz Biotechnology (Dallas, TX, USA: EDEM CRISPR/Cas9 KO Plasmid (h) (sc-401946) and EDEM HDR Plasmid (h) (sc-401946-HDR), according to manufacturer’s instruction. At 48 h post transfection, the media were changed with fresh DMEM supplemented with 4 µg/mL puromycin, a selection antibiotic. The cells were kept in the media supplemented with puromycin for 3 passages, after which they were transferred onto media supplemented with 2 µg/mL puromycin. Selection efficiency was verified by expression level of EDEM1 in normal versus KO EDEM1 cell line assessed by Western blot. For a homogenous expression, the cells were cloned using a FACS Aria III system (BD Biosciences, San Jose, CA, USA). The expression level of EDEM1 in clones was verified also by Western blotting. The clone selected for further work was supplementary tested by immunoprecipitation with EDEM1 specific antibodies followed by Western blotting detection.
Sucrose gradient fractionation: HEK293T cells transfected with EDEM1 and ∆-EDEM1, treated or not with kifunensine, were lysed either in a TritonX-100-containing buffer (1% Triton X-100 (v/v), 150 mM NaCl, 1.5 mM MgCl2, and 1 mM EDTA) or a Digitonin-containing buffer (1% Digitonin (w/v) 50 mM Tris-HCl pH 7.3, 5 mM EDTA, and 150 mM NaCl) supplemented with protease inhibitors (Roche-Basel, Switzerland), and the lysates that were cleared at 14,000 g for 30 min were loaded on a continuous 10–40% sucrose gradient, prepared in an 8× diluted lysis buffer. The samples were centrifuged in an SW41 Ti rotor (Beckman, Brea, CA, USA) for 16 h, 39,000 rpm, at 4 °C. The collected fractions were precipitated with a 100% TCA (Sigma-Aldrich, St. Louis, MO, USA) solution, 1:4 ratio, centrifuged at 4 °C, and washed 3 times with cold acetone. Dried pellets were resuspended in a 4% SDS-containing buffer (100 mM TRIS-HCl, pH: 7.60) and sonicated, and equal volumes from each fraction were separated by SDS-PAGE in reduced conditions. The proteins were transferred onto nitrocellulose membranes and probed with specific antibodies.
Inhibitors treatment: The cells were seeded in 12 well plates, transfected as described above, and incubated with 12.5 µM MG132 or 30 µM kifunensine for approximately 16 h before harvesting; for pulse-chase experiments, MG132 was added in the starvation, pulse, and chase period at a 20 μM concentration.
Western blotting: HEK293T was co-transfected with EDEM1 mutants or/and ERAD substrates, as described above. Cells were lysed in buffers containing either 1% Triton-X100, 2% CHAPS, or 1% Digitonin for 30 min on ice and centrifuged at 14,000 rpm for 30 min. Equal amounts of proteins from each sample, detected by bicinchoninic acid method (BCA), were separated by SDS-PAGE and transferred onto nitrocellulose membranes. The membranes were probed with appropriate primary antibodies for 2 h and diluted in 5% milk or BSA in phosphate buffered saline (PBS)-0.1% Tween at room temperature (RT) or ON at 4 °C, and they were washed and incubated with secondary antibodies coupled with HRP for 1h at RT. The results were viewed by chemiluminescent reaction.
Pulse-chase and immunoprecipitation: The cells were starved in the cysteine/methionine-free medium (Sigma-Aldrich, St. Louis, MO, USA) for 30 min, pulse-labelled for 20 or 30 min with 50–75 mCi of [35S]-methionine/cysteine (Tran35S-label, Perkin Elmer, Waltham, MA, USA), and chased for the indicated time points; for some of the samples MG132 20 µM was added. Labelled cells were washed with ice cold PBS and lysed with a CHAPS buffer (50 mM HEPES, 200 mM NaCl, and 2% CHAPS). The lysates were incubated overnight with antibodies for Ribophorin I, EDEM1, or α-1AT, immobilized on protein A-Sepharose beads for 2 h at 4 °C, and eluted with a Laemmli buffer, 5× stock diluted to 1× with TE (50 mM Tris, 150 mM NaCl, and 1 mM EDTA) at 95 °C for 5 min. The samples were separated by SDS-PAGE, and proteins were visualized by autoradiography.
Immunoprecipitation and Western blotting: HEK293T or HEK293T-KO EDEM1, previously transfected with EDEM1 mutants and/or ERAD substrates treated or not with kifunensine, were harvested, lysed in a CHAPS-containing buffer and subjected to immunoprecipitation with specific antibodies, as indicated in each figure panel ON and captured on protein A/G-Sepharose beads for 2 h at 4 °C. The proteins were eluted with a Laemmli buffer, separated in polyacrylamide gels, transferred onto nitrocellulose membranes, and probed with specific antibodies. Results were visualised by chemiluminescent reaction.
Immunoprecipitation of sucrose gradient fractions: Sucrose gradient fractions of HEK293T-overexpressing EDEM1 and digitonin lysates were concatenated to have 8 out of 15 fractions initially harvested. Each fraction was diluted 10 times with a 0.1% Digitonin buffer and incubated ON with polyclonal anti EDEM1 antibodies (1:1000 v/v dilution). The antibodies were captured on protein A-Sepharose beads for 2 h at 4 °C. After 3 washes with a 0.1% Digitonin-containing buffer, the proteins were eluted with a Laemmli buffer and separated in polyacrylamide gels. Furthermore, the gel was cut in small pieces that were subject to trypsin digestion.
Immunofluorescence microscopy and co-localization analysis: HeLa cells were seeded onto coverslips and transfected with the corresponding plasmids for 24 h using Lipofectamine 2000 (Invitrogen-Life Technologies, Paisley, UK) according to the manufacturer’s instructions. Afterwards, the cells were fixed by incubation with 1% PFA in PBS (13 mM NaCl, 2.7 mM KCl, 10 mM Na2HPO4, and 1.8 mM KH2PO4) for 1 h and permeabilised for 3 min with 0.005% Digitonin in a blocking buffer (2% horse serum in PBS). The samples were incubated with a blocking buffer for 2 h and overnight with the primary antibodies at the indicated dilutions in humidified atmosphere. The next day, the coverslips were washed and incubated with a combination of secondary antibodies and fluorescent dye (Proteostat-Enzo Life Sciences, Farmingdale, NY, USA) to detect aggregates for 30 min at RT. Samples were extensively washed and subsequently mounted on glass slides. Images were acquired using the Zeiss LSM 700 (63X, 1.4 NA, oil) microscope using the LSM acquisition software (Zeiss, Oberkochen, Gemany).
Acquired images were processed using the ImageJ software. Co-localization analysis was performed using the ImageJ JACoP plugin (NIH, Bethesda, MD, USA). The images were split into separated channels and used for threshold processing. The analysis was performed in three independent experiments, and the total number of fields analysed is indicated in the figure legends.
LC–MS/MS analysis: HEK293T cells expressing an empty vector, EDEM1, or ∆-EDEM1 were harvested at 90% confluence and lysed using either 1% Triton X-100 (TX)- or 1% Digitonin (DIG)-containing buffers for 30 min on ice. The lysates were cleared by centrifugation, followed by immunoprecipitation using polyclonal anti-EDEM1 antibodies, as previously described [19
]. The captured complexes were eluted from the resin with a soft elution buffer (SEB: 50 mM Tris pH 8.0, 0.2% SDS, and 0.1% Tween-20) in a 4:1 (v/v
) ratio at RT [60
]. The eluted proteins were separated by SDS-PAGE and prepared for MS analysis using a previously described protocol for in-gel digestion [61
]. For HEK293T cell proteomic analysis, proteins were extracted using 6M Guanidine hydrochloride, reduced with 10 mM TCEP (Tris(2-carboxyethyl)phosphine hydrochloride), alkylated with 5 mM chloroacetamide, and subjected to in solution overnight digestion with trypsin at 37 °C. The extracted peptides were dried in Speed-Vac, and each sample was reconstituted in mobile phase A (0.1% FA and 2% ACN) and injected on a C18 trap column (20 mm × 100 µm internal diameter) (Proxeon Biosystems, Thermo Scientific, Waltham, MA, USA) connected online to a C18 analytical column (100 mm x 75 µm internal diameter) (Proxeon Biosystems, Thermo Scientific, Waltham, MA, USA) for peptide separation. The chromatographic equipment was connected online to an LTQ-Orbitrap Velos Pro instrument operated in a data-dependent mode. A top 5, 10, or 15 method, depending on sample complexity, was used for data acquisition involving a survey scan at 60,000 resolution (m/z 400) with Orbitrap detection, followed by the consecutive collision-induced dissociation (CID) fragmentation scans in the linear ion trap. A 2–30% B (0.1% FA and 98% ACN) gradient was used for the chromatographic separation of the peptides.
LC–MS/MS data analysis: Raw data files were searched using the SEQUEST/SEQUESTHT algorithms integrated into Proteome Discoverer v1.4 (Thermo Scientific, Waltham, MA, USA) or using the Andromeda integrated in MaxQuant. For both searches, the settings were the following: trypsin as the proteolytic enzyme and a maximum of two missed cleavages, 10 ppm as mass accuracy for precursor ions or 20 ppm (during the first search) and 6 ppm (for the second search) for Andromeda searches, 0.5 Da for fragment ion tolerance, carbamidomethylation on Cys residues as a static modification, and oxidation on Met residues as a dynamic modification. For PSM validation, the percolator node available in Proteome Discoverer v1.4 was used, with the validation based on q value. The results were filtered for 1% FDR (PSM level) and a peptide mass deviation of maximum 5 ppm. Andromeda results were similarly filtered to 1% FDR using the built-in MaxQuant procedure
Statistical analysis: For statistical analysis, data sets were processed using either one-way or two- way ANOVA with a Bonferroni correction using Prism6 (GraphPad software, San Diego, CA, USA). Results with a p value of less than 0.05 were considered significant, as indicated in the figure legends. No criteria of inclusion or exclusion of data were used in this study. Data shown are representative for two-to-four experiments, as specified in the figure legends.
For LC–MS/MS data analysis, a two-sample t-test was used with a permutation-based FDR correction at a significance value of 0.05 using the MaxQuant reported intensity values (LFQs). Shown in figures are either log of LFQ or spectral counts for the SEQUEST/ SEQUESTHT searches.