The Degradation Pathway of COP9 Signalosome–Cullin-RING Ubiquitin Ligase Complexes via Autophagy

TL;DR

This study reveals that specific COP9 signalosome-Cullin-RING complexes are degraded through autophagy, a process that can be inhibited by certain drugs.

Contribution

The study identifies self-ubiquitylation of cullins as a signal for selective macroautophagy of CSN-CRL complexes.

Findings

CSNCSN7A-CRL3 and CSNCSN7B-CRL4A complexes are degraded via autophagy in the absence of serum.

Self-ubiquitylation of cullins acts as a signal for selective macroautophagy of CSN-CRL complexes.

CSN-CRL complexes are expelled from the nucleus and localized in autophagosomes for degradation.

Abstract

In Mammalia, the COP9 signalosome (CSN) is associated with cullin-RING ubiquitin ligases (CRLs). This study focuses on the variants CSNCSN7A and CSNCSN7B, which form complexes with CRL3 and CRL4A, respectively. Although some research has been conducted on the assembly of the complexes, little is known about their breakdown. Here, we show that entire CSNCSN7A-CRL3 and CSNCSN7B-CRL4A complexes are degraded via autophagy. CSN-CRL complexes are degraded in the absence of serum via bulk autophagy and in the presence of the specific inhibitor of CSN, CSN5i-3, via selective macroautophagy. Surprisingly, the self-ubiquitylation of cullins in the CRLs was identified as a specific signal for selective macroautophagy. The self-ubiquitylation of cullins takes place in the presence of CSN5i-3, and CSN-CRL complexes are expelled from the nucleus to be degraded in the cytosol. Selective macroautophagy can be blocked by chloroquine, a specific inhibitor of autophagy. Interestingly, the process can also be inhibited by MLN4924, a neddylation inhibitor. Confocal fluorescence microscopy illustrates the interaction of CSN subunits with ATG8, as well as with RAB7, both in HeLa and in LiSa-2 cells. Confocal fluorescence microscopy produces images that suggest the localization of CSN-CRL particles in autophagosomes. Our data place CSN-CRL in the category of large complexes that are degraded through autophagy.

Figures

Full text

1. Introduction

The COP9 signalosome (CSN) belongs to a group of paralog particles called ZOMES, complexes that include the 26S proteasome LID and the eukaryotic initiation factor 3 (eIF3). In Mammalia, the core CSN consists of six Proteasome lid-CSN-Initiation factor 3 (PCI) domain subunits (CSN1-CSN4, CSN7, and CSN8) and of two MPR1/PAD1 N-terminal (MPN) domain (CSN5 and CSN6) subunits [1]. The CSN is more heterogeneous than indicated by its eight-core-subunit structure [2]. A fraction of the cellular CSN contains a non-canonical subunit [3,4]. Many human CSNs are associated with deubiquitylating enzymes (DUBs) [5] or proteins like p27 and p53 [1]. CSN subunits exist as paralogs forming CSN variants. We are focused on CSN7, which is expressed by the paralog COPS7A/CSN7A, making the CSN variant CSN^CSN7A^ and the paralog COPS7B/CSN7B, which occurs in the CSN variant CSN^CSN7B^. The variants CSN^CSN7A^ and CSN^CSN7B^ appear simultaneously in most human cells. The CSN is a multi-DUB complex [5] removing NEDD8 from the cullins of cullin-RING ubiquitin ligases (CRLs) [2,5,6,7]. The CSN is associated with the CRL complex, and interaction between the two is the reason for conformational changes to CSN2, CSN4, and CSN7, which move the CSN5-CSN6 dimer into position for deneddylation [6,8,9]. The aforementioned CSN variants and specific CRLs form permanent complexes that represent a reservoir of different cellular functions [1,10]. CSN^CSN7A^ interacts with CRL3, and CSN^CSN7B^ forms permanent complexes with CRL4A. The complexes CSN^CSN7A^-CRL3 and CSN^CSN7B^-CRL4A exist side by side in all studied cells [1]. To be active, CSN-CRL complexes need substrate receptors (SRs) and the neddylation of cullins. The appropriate SRs occur in a manner that is dependent on the available substrates, and neddylation is a highly regulated process [1,11]. Under conditions of differentiation, such as during adipogenesis, SRs are mostly exchanged quickly using cullin-associated and neddylation-dissociated protein 1 (CAND1) [1].

Little is known about the turnover of CSN-CRL complexes; however, the assembly of fungal CSN from two trimeric intermediates was recently published [12]. These novel findings provide insight into the assembly of CRL1 and CRL3 complexes [13,14,15], and the accompanying electron microscopy images present views of CSN-CRL interaction [6]. However, nothing is known about CSN-CRL catabolism, though recent data indicate that degradation may take place via autophagy [16,17]. Some large protein complexes, like the 26S proteasome [18,19,20] and CDC48 [21], are degraded via autophagy, in a process in which both nutrient starvation and 26S proteasome inhibition lead to autophagy, also called proteaphagy, implying that bulk and selective routes for using the ubiquitylation of RPN10 as an autophagic receptor, tethering the complex to ATG8, do exist [18,19].

In this study, we show that CSN-CRL is degraded via autophagy as an entire complex. Whether via bulk autophagy, using serum starvation, or selective macroautophagy, induced by a CSN5i-3 inhibitor, CSN-CRL complexes are finally degraded by lysosomes. In the event of selective macroautophagy, CSN-CRL complexes are first tethered to ATG8 by phagophores prior to autophagy. The self-ubiquitylation of cullins in CRL complexes serves as a trigger, and can be inhibited through blocking neddylation.

2. Materials and Methods

2.1. Generation of Stable Cell Lines

The generation of FLAG, FLAG-CSN7A, and FLAG-CSN7B plasmids has been described previously [1]. The HA-tagged ubiquitin (HA-Ub) plasmid was purchased from Addgene (Watertown, MA, USA).

To generate stable transfectants, the HA-Ub plasmid was transfected into HeLa cells using Lipofectamine 2000 (ThermoFisher, Waltham, MA, USA), according to the manufacturer’s protocol. After transfection, HeLa cells permanently expressing HA-Ub were obtained via 0.5 mg/mL G418 administration for 1 to 3 weeks. HA-Ub-containing individual clones were isolated and propagated in the selection medium. CSN7A and CSN7B knockout cells were used and were generated using CRISPR-Cas9 technology as described previously [1].

2.2. RNA Interference

SiRNA transfection against GFP (control) and ATG7 was performed with the Lipofectamine 2000 transfection reagent (ThermoFisher) according to the manufacturer’s protocol. The transfection of siRNAs into HeLa cells was performed at a final concentration of 100 nM of siRNA, 72 h prior to further experiments.

2.3. Protein Extraction, Immunoblotting, and Immunoprecipitation

Cell lysates were obtained in triple-detergent buffer (50 mM Tris-HCl, pH 8.5, 150 mM NaCl, 0.02% (w/v) sodium azide, 0.1% (w/v) SDS, 1% (v/v) NP-40, 0.5% (w/v) sodium deoxycholate) as outlined previously [1]. Western blots with appropriate samples were performed and analyzed as described [1].

The mono-detergent lysis buffer (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, pH 8.0, 1% Triton X-100) was outlined previously [1]. For denaturing, cells were lysed with triple-detergent lysis buffer with 1% SDS and heated at 95 °C for 10 min. Then, the samples were diluted to a ratio of 1:10 with triple-detergent buffer. Cellular debris was removed via centrifugation for 10 min at 13.000 rpm and 4 °C. Appropriate antibodies, incubated with supernatants, were used at 1–5 µg for 2 h at 4 °C. Protein-A Sepharose was used, and the mixture was rotated overnight at 4 °C. After washing the beads three times with the appropriate lysis buffer, they were boiled with 1× Laemmli SDS-PAGE sample buffer. Immunocomplexes were separated on SDS-PAGE and analyzed via immunoblotting with the following antibodies: anti-CSN3 (Abcam, Cambridge, UK), anti-CSN5 (Cell Signaling, Danvers, MA, USA), anti-CSN7A (Santa Cruz, Santa Cruz, CA, USA), anti-CSN7B (Abcam, Cambridge, UK), anti-CUL3 (BD biosciences, San Jose, CA, USA), anti-CUL4A (Abcam), anti-CDC48 (Abcam), anti-FLAG (Sigma, St. Louis, MO, USA), anti-HA (Santa Cruz), anti-NEDD8 (Invitrogen, Carlsbad, CA, USA), anti-C23 (Santa Cruz), anti-RAB7 (Santa Cruz), anti-Ubiquitin (Santa Cruz), p62 (Abcam), TOLLIP (Abcam), and anti-γ-tubulin (Santa Cruz).

2.4. Density Gradient Centrifugation

As described above, HeLa cells (10^7^) untreated and treated with 1 μM of CSN5i-3 were lyzed with mono-detergent lysis buffer. Cell extracts were loaded onto a 10–30% density glycerol gradient, and ultracentrifugation (27,000 rpm for 24 h at 4 °C) was performed as described [22]. Following centrifugation, the gradients were fractionated into 600 μL volumes. Each fraction (10 μL) was analyzed via SDS–PAGE and immunoblotting with different antibodies. CDC48 served as a marker protein.

2.5. FLAG Pulldowns

LiSa-2 cells with FLAG-tagged CSN7A paralogs were lyzed in mono-detergent buffer as outlined above. Cell lysates were loaded into a pre-equilibrated ANTI-FLAG M2 affinity column (Sigma, St. Louis, MO, USA). Following washing with 20 column volumes of mono-detergent lysis buffer, bound proteins were eluted through competition with 100 µg/mL of the FLAG peptide, as recommended by the manufacturer’s protocol. SDS-PAGE was used to separate proteins, which were later analyzed via immunoblotting with the indicated antibodies.

2.6. Subcellular Fractionation

The fractionation of subcellular compartments has previously been outlined [1].

2.7. Confocal Fluorescence Microscopy and Immunostaining

HeLa and LiSa-2 cells with different backgrounds were exposed in chamber slides. After treatment with different inhibitors, the cells were washed three times with PBS and fixed in 4% paraformaldehyde for 15 min. Fixed samples were permeabilized with 0.1% TritonX-100/PBS for 10 min at room temperature and blocked with 3% filtered BSA in PBS for 1 h. The samples were incubated with primary antibodies directed against LAMP2 (rabbit), ATG8 (rabbit), RAB7 (mouse), CSN3 (rabbit), or CSN7A (mouse) in the indicated combinations in 1% BSA in PBS overnight at 4 °C. Thereafter, the samples were washed three times in PBS and stained for 2 h at room temperature with secondary antibodies coupled to FITC 647 (anti-rabbit) and Cy3 (anti-mouse) fluorophores with 1% BSA in PBS. The cell nuclei were counterstained using DAPI. The slides were visualized and analyzed as described previously [1].

2.8. Statistics

Bands from SDS PAGE were visualized with a ChemoCam Imager (Intas, Göttingen, Germany) and quantified using ImageJ software (version 1.51d). GraphPad Prism 8.0.1 software was used to calculate statistical significance. Error bars indicate standard deviations (SDs). For statistical analysis, unpaired Student’s t-tests were applied. n represents the number of independent experiments. Statistical details of individual experiments can be found in figure legends.

3. Results

3.1. The COP9 Signalosome Is Degraded as a Complex upon Serum Starvation and Specific Inhibition via the CSN5i-3 Inhibitor

In LiSa-2 cells, serum starvation and inhibition by the deneddylation inhibitor CSN5i-3 [23] led to the degradation of the whole CSN protein complex consisting of eight subunits, most likely via autophagy. Control subunits of the 26S proteasome, RPN1 and 20Sa4, were not influenced by CSN5i-3 (1 µM) (Figure 1A). The selected CSN subunits CSN3, CSN5, CSN7A, and CSN7B were characterized by a maximum degradation of 20% following 48 h of starvation. In contrast, the presence of CSN5i-3 led to degradation greater than 50% after 48 h of treatment (Figure 1B). In CSN7A or CSN7B, knockout HeLa cells brought about a similar degradation of CSN subunits as in wild-type (WT) cells (Figure 1C). We conclude that both CSN^CSN7A^ and CSN^CSN7B^ are degraded in a manner that is similar to whole-protein complexes in the presence of CSN5i-3. Figure 1D shows that CSN variants equipped with FLAG-CSN7A or FLAG-CSN7B in LiSa-2 cells were also degraded in a manner dependent on the CSN5i-3 inhibitor concentration, probably via autophagy. A concentration of 0.5 µM CSN5i-3 was sufficient to trigger degradation. Moreover, as shown in Figure 1E, degradation of CSN was induced by 1 µM CSN5i-3 after 28 h in all studied cells, making the process universal. We selected these conditions for further experiments.

3.2. The COP9 Signalosome Is Degraded in Association with Cullin-Ring Ubiquitin Ligases

In human cells, permanent CSN-CRL complexes are localized predominantly to the nucleus [1] due to their functions as cell cycle regulators or chromatin protectors [24]. Using 1 µM CSN5i-3, both CSN subunits as well as CUL3 and CUL4A were expelled from the nucleus (Figure 2A). This was the case in both HeLa and in LiSa-2 cells and indicated that the CSN subunits and cullins had been subjected to similar processes. To ensure that both CSN subunits and cullins were degraded together, we calculated ratios between the CUL4A and CSN subunits upon CSN3 immunoprecipitations before and after the stimulation of degradation (Figure 2B, IP: CSN3). Figure 2C demonstrates that the CUL4A/CSN ratio in the CSN3 immunoprecipitations is similar before and after 28 h of degradation in the presence of 1 µM CSN5i-3, indicating equal degradation rates for CUL4, CSN3, and CSN7B. CUL3 immunoprecipitates confirmed the degradation rates of input and CSN3 immunoprecipitates (Figure 2B). The glycerol gradient in Figure 2D demonstrates that CSN-CRL complexes did not decay during 28 h of stimulation in the presence of 1 µM CSN5i-3. They were not dissected into individual parts and seem to have been degraded as an entire complex. As a control for the density gradient, we used CDC48, which migrated in fractions of 500–600 kDa. The fluorescence microscopy images in Figure 2E illustrate once more how CSN-CRL complexes were expelled from the nucleus in the presence of 1 µM CSN5i-3, where they bound with LAMP2, one of the lysosome-associated membrane glycoproteins [25]. The data indicates that CSN-CRL complexes were exported from the nucleus to the cytosol together, where the autophagy of the entire complex took place.

3.3. COP9 Signalosome–Cullin-Ring Ubiquitin Ligases (CSN-CRLs) Are Degraded as a Whole Complex via Autophagy

As expected from previous experiments, the degradation of CSN-CRL complexes occurs via autophagy. To analyze that, we used the specific autophagy inhibitor chloroquine (CQ) [26]. Under the conditions we selected, p62 and ATG8 were accumulated in the presence of CQ without any influence on cell viability (Figure S2). As shown in Figure 3A, CQ inhibits the degradation both of CUL4A and of selected CSN subunits, CSN3, CSN5, CSN7A, and CSN7B, in the same manner. Degradation in the presence of CSN5i-3 after 28 h incubation is about 50% and can be restored by CQ to almost 100% (Figure 3B). There is no influence of ATG7, an E1 activating enzyme of ATG8 conjugation, on CSN-CRL degradation (Figure S1A).

CSN-CRL degradation, through an autophagic mechanism, was further tested via fluorescence microscopy. The binding of CSN-CRL complexes to ATG8 was initially analyzed using confocal fluorescence microscopy [25]. The resulting images demonstrate a co-localization between CSN7A and ATG8. The process was inhibited by CQ (Figure 3C). CSN-CRL complexes might bind to ATG8 via specific autophagic receptors. Precipitates were tested to identify whether autophagic receptors bind prominent receptors, p62 and TOLLIP [27,28,29], to CSN or to CRL3. There was no association between the analyzed receptors and CSN or CRL3 (Figure S1B). RAB7 is a component that tethers complexes to autophagosomes [30] and plays a key role as a regulator of autophagosomes [26]. Therefore, whether the CSN-CRL complexes co-localize with RAB7 was also tested. The data shows that the marker of autophagosomes RAB7 is co-localized with CSN3, a component of CSN-CRL complexes (Figure 3D). Clearly, CSN3 is expelled from the nucleus and trapped by RAB7 vesicles in both LiSa-2 cells (Figure 3D) and HeLa cells (Figure S1C). Interestingly, in Figure 3D and Figure S1C, the vesicles, probably autophagosomes, are filled with complexes like CSN-CRL. In the presence of CQ, the autophagic process is stopped, and the superposition of CSN3 and RAB7 in the cytosol is even more evident in both LiSa-2 cells (Figure 3D) and HeLa cells (Figure S1C).

3.4. Neddylation-Dependent Self-Ubiquitylation of CUL3 and CUL4 Are Signals for CSN-CRL Degradation via Selective Macroautophagy

Endogenous immunoprecipitation using the anti-CSN7A antibody usually precipitates CSN subunits and cullins [1]. After 28 h of the incubation of LiSa-2 cells in the presence of 1 mM CSN5i-3, NEDD8 and ubiquitin labels were found in the molecular weight regions of the cullins (Figure 4A). Similarly, in FLAG-CSN7A pulldowns of LiSa-2 cells, NEDD8 and ubiquitin were detected using anti-NEDD8 and anti-ubiquitin antibodies in cullin positions (Figure 4B). To ensure that the cullins are ubiquitylated as a signal of the macroautophagy of CSN-CRL complexes, HeLa cells were stably transfected with HA-Ub. After 28 h of incubation without and with CSN5i-3, HA was immunoprecipitated. The results in Figure 4C show that, in the presence of CSN5i-3, the cullins and CSN subunits are immunoprecipitated as well as degraded, and that the cullins are labeled with ubiquitin. Moreover, in Figure S2A, the ubiquitylation of CUL4A is shown directly via immunoprecipitation under denaturing conditions. Clearly, cullin self-ubiquitylation is a signal for selective CSN-CRL macroautophagy. Is neddylation also necessary for autophagy? The neddylation inhibitor MLN4924 blocks the export of CSN-CRL complexes from the nucleus to the cytosol (Figure 4D). Moreover, it can be seen from Figure 4E that MLN4924 inhibits selective autophagy induced by CSN5i-3. Whereas 28 h of incubation with CSN5i-3 leads to the autophagy of about 50% of the CSN-CRL complexes, MLN4924 + CSN5i-3 or MLN4924 alone blocked the process almost completely (Figure 4F). It should be noted that the MLN4924 inhibitor must be added at least 2 h before the CSN5i-3 supplement. Adding MLN4924 at the same time as CSN5i-3 or later has no effect.

4. Discussion

Our results show that the complexes CSN^CSN7A^-CRL3 and CSN^CSN7B^-CRL4A are degraded after serum starvation via bulk autophagy or, in the presence of CSN5i-3, by selective macroautophagy (Figure 5). The data form the basis for a recent publication [17]. The self-ubiquitylation of CUL3 and CUL4, presumably mediated by the RBX protein of the appropriate CRLs, serves as a signal for selective macroautophagy (Figure 5). This was shown through stable HA-ubiquitin transfection and HA immunoprecipitation (Figure 4C). Generally, the special topology of ubiquitylation, such as its mono-ubiquitylation, drives protein complexes, organelles, and pathogens to autophagic degradation [31]. Most CSN^CSN7A^-CRL3 and CSN^CSN7B^-CRL4A complexes are localized in the nucleus. In CSN-CRL complexes, the binding of CSN5i-3 to the active site [23] of the deneddylation-active MPN + protein, CSN5, leads to the expulsion of the complexes from the nucleus (see Figure 2A and Figure 5). The autophagy machinery is localized in the cytoplasm. Interestingly, 26S proteasomes dissociate into subcomplexes prior to export and lysosomal degradation [19]. Regarding CSN-CRL complexes, it is not yet precisely known how stable non-functioning CSN-CRL particles are recognized and expelled from the nucleus. The degradation occurs via autophagy because the specific lysosome and autophagy inhibitor, CQ, blocks this process (Figure 3A) [32]. Furthermore, confocal fluorescence microscopy demonstrates that, in the presence of CSN5i-3, CSN^CSN7A^-CRL3 is most likely co-localized with ATG8 in the cytosol, where selective macroautophagy takes place (Figure 3C, upper images). CQ blocks this process (Figure 3C, lower images). Interestingly, the downregulation of ATG7, the E1-enzyme for ATGylation, does not significantly influence the autophagic degradation of CSN-CRL (Figure S1A), as confirmed by previously published proteaphagy data for mammalian cells [33]. The independence of CSN-CRL degradation from ATG7 could be explained by the occurrence of an as yet unknown further isoform of the E1 enzyme for ATGylation in Mammalia. The co-localization of CSN (CSN3) and RAB7 underlines the association of CSN particles with autophagosomes in HeLa cells (Figure S1C) [34]. In LiSa-2 cells (Figure 3D), as well as in HeLa cells (Figure S1C), confocal fluorescence imaging shows RAB7 vesicles, which might be filled with CSN particles. The loading of RAB7 vesicles is inhibited by CQ both in HeLa (Figure S1C) and in LiSa-2 cells (Figure 3D). Therefore, CSN^CSN7A^-CRL3 and CSN^CSN7B^-CRL4A are degraded via bulk autophagy and specific macroautophagy. We speculate that most CSN-CRL complexes are removed via autophagy. The selective macroautophagy of CSN-CRL complexes is similar to the proteaphagy of the 26S proteasome [19]. In both cases, for the selective macroautophagy of CSN-CRL complexes and of the 26S proteasome, specific ubiquitylation is required. In proteaphagy, three ubiquitin ligases act sequentially to promote nuclear export and autophagy [35]. In contrast, the degradation of CSN-CRL complexes via selective macroautophagy is initiated through the self-ubiquitylation of cullins by the corresponding CRLs. Generally, CRLs are involved in the autophagic degradation of different proteins via ubiquitylation in different organisms [36,37]. However, here we see the CSN-dependent self-ubiquitylation of CRLs, a novel self-regulation mechanism. The neddylation inhibitor MLN4924 blocks self-ubiquitylation. It inhibits two processes: first, the export from the nucleus (Figure 4D), and second, degradation in the cytosol via selective macroautophagy (Figure 4D,F).

As is the case with CSN-CRL, the degradation of the UPS regulator CSN via autophagy demonstrates further functional mutual regulation between two mayor proteolytic pathways, UPS and autophagy. CSN and CRL control autophagy [38,39]. At the same time, under special conditions, they are self-consumed via autophagy. Moreover, both pathways overlap via ubiquitylation and their component, the E3 ligases.

5. Conclusions

Our study reveals the degradation pathway of CSN^CSN7A^-CRL3 and CSN^CSN7B^-CRL4 particles via bulk autophagy or, in the presence of CSN5i-3, via macroautophagy. CSN-CRL complexes signal macroautophagy through self-ubiquitylation, which can be blocked via neddylation inhibition. Ubiquitylated CSN-CRL complexes are tethered by, thus far, unknown autophagy receptors (ARs) and ATG8 to phagophores and are, finally, degraded in lysosomes.