The COP9 signalosome-mediated deneddylation is stimulated by caspases during apoptosis

In concert with the ubiquitin (Ub) proteasome system (UPS) the COP9 signalosome (CSN) controls the stability of cellular regulators. The CSN interacts with cullin-RING Ub ligases (CRLs) consisting of a specific cullin, a RING protein as Rbx1 and substrate recognition proteins. The Ub-like protein Nedd8 is covalently linked to cullins and removed by the CSN-mediated deneddylation. Cycles of neddylation and deneddylation regulate CRLs. Apoptotic stimuli cause caspase-dependent modifications of the UPS. However, little is known about the CSN during apoptosis. We demonstrate in vitro and in vivo that CSN6 is cleaved most effectively by caspase 3 at D23 after 2–3 h of apoptosis induced by anti-Fas-Ab or etoposide. CSN6 processing occurs in CSN–CRL complexes and is followed by the cleavage of Rbx1, the direct interaction partner of CSN6. Caspase-dependent cutting of Rbx1 is accompanied by decrease of neddylated proteins in Jurkat T cells. Another functional consequence of CSN6 cleavage is the enhancement of CSN-mediated deneddylating activity causing deneddylation of cullin 1 in cells. The CSN-associated deubiquitinating as well as kinase activity remained unchanged in presence of active caspase 3. The cleavage of Rbx1 and increased deneddylation of cullins inactivate CRLs and presumably stabilize pro-apoptotic factors for final apoptotic steps.


Introduction
The COP9 signalosome (CSN) of mammalian cells consists of eight polypeptides, CSN1 to CSN8 [1]. The CSN interacts with components of the ubiquitin (Ub) proteasome system (UPS) known as cullin-RING Ub ligases (CRLs) [2]. These enzyme complexes select proteins for ubiquitination and are responsible for substrate specificity of the UPS. Besides a specific cullin, most of the CRLs contain a RING domain protein called ROC1/Rbx1 involved in the Ub ligation reaction as well as substrate recognition components [2,3]. The CSN removes the Ub-like protein Nedd8 from its covalent linkage to cullins, a process called deneddylation [4][5][6]. As it has been shown, CSN-mediated deneddylation prevents the assembly of a specific CRL [4]. In cooperation with the UPS the CSN participates in processes such as DNA repair [7], cell cycle [8], angiogenesis [9] and development [10][11][12]. However, its role in apoptosis is unknown. Although the UPS and the CSN determine the stability of regulatory molecules involved in apoptosis such as p53, Bax, Bak and Smac [13][14][15], the exact function of the system in programmed cell death is still obscure. Inhibitors of CSN-associated kinases induce accumulation of p53 and apoptosis in tumor cells [16]. The UPS in concert with the CSN controls the degradation of IjBa, a major regulator of the key surviving factor NF-jB [17,18]. The fact that proteasome inhibitors induce apoptosis is already successfully used in tumor therapy [19,20].
Two major pathways of apoptosis, the intrinsic and extrinsic pathways, have been described [21,22]. Many of the effects that occur during apoptosis result from activation of caspases. Caspase 8 (Casp8) and 9 (Casp9) are the apical caspases in the extrinsic and intrinsic pathway, respectively [21,22]. In the intrinsic pathway, mitochondrial perturbation leads to cytochrome c release into the cytosol where it binds to Apaf-1, facilitating the binding of ATP/dATP and oligomerization of Apaf-1 to form the apoptosome. This oligomerization activates Casp9 [23,24]. The extrinsic pathway is triggered by binding ligands (e.g. FAS, TRAIL) to their respective cell-surface death receptors, which after oligomerization recruits adaptor molecules and the initiator caspase Casp8 [25]. These two pathways converge with the activation of effector caspases such as caspases 3 and 7 (Casp3 and Casp7) and the activation of endonucleases resulting in DNA fragmentation.
Do the UPS and the CSN exert specific regulatory functions during the apoptotic process? The activity of the 26S proteasome is reduced by caspase-dependent cleavage of the regulatory subunits Rpt5, Rpn10, and Rpn2, resulting in the stabilization of the pro-apoptotic protein Smac. This effect provides a feed-forward amplification loop of apoptosis [26]. We were interested in the role of the CSN complex during apoptosis. Here we demonstrate that apoptotic stimuli cause the caspase-dependent cleavage of subunit CSN6, which is accompanied by the activation of CSN-mediated deneddylation. In addition, CSN6 modification is followed by Rbx1 cleavage resulting in the inactivation of CRLs.

Glycerol gradients and non-denaturing electrophoresis
Glycerol gradient centrifugation was performed as outlined before [32]. In brief, 2 9 10 7 cells were lyzed in monolysis buffer [33] and equal amounts (3 mg) were loaded onto a 12 ml glycerol gradient (10-40%). Fractions of 600 ll were collected and aliquots used for Western blot analysis with the indicated antibodies.
For non-denaturing electrophoresis 20 lg of cell lysates were separated on a 4-15% (w/v) Phast-gel (Amersham Pharmacia Biotech.) at 300 V/h. Proteins were blotted onto nitrocellulose and probed with anti-CSN6 antibody.
Recombinant CSN6 and site-directed mutagenesis CSN6 and CSN6D29 was obtained by PCR using clone IRAUp969A108D (Lib.969) purchased from RZPD as template and cloned into pQE expression vector (Qiagen). Site-directed mutagenesis was performed with CSN6 cDNA in pQE using the Quick-Change strategy (Stratagene). The cDNA of all constructs was verified by sequencing. Protein purification was carried out with the Ni-NTA purification kit (Qiagen).

Cleavage of CSN complex or recombinant CSN subunits by caspases
The CSN complex was purified from human erythrocytes as described before [32]. The purified CSN complex (200 nM) or 500 nM of recombinant His-tagged CSN subunits were incubated with equimolar amounts of either Casp3, Casp7 or Casp8 with or without 50 lM z-DEVD-fmk (z-DEVD, Biomol) in a buffer containing 30 mM Tris, pH 7.8, 10 mM KCl and 5 mM DTT for indicated times at 37°C. Recombinant processed Casp3, Casp7 and Casp8 were prepared as described previously [26].

CSN activity assays
Purified CSN complex (750 nM) was pre-incubated with or without Casp3 (750 nM) for 1 h at 37°C in PBS with 10 mM DTT. Casp3 was inhibited with 1.5 lM z-DEVD (z-DEVD-CHO, Biomol) in a 30 min pre-incubation at 4°C before the CSN was added (inhib. Casp3) or was added to all samples after pre-incubation to block Casp3 during cleavage of the Nedd8-Cul1 peptide. Deneddylating activity was determined by adding 1.6 lg of a biotin-coupled Nedd8-Cul1 peptide (see below). After 3 h the cleavage of the peptide was terminated by addition of loading buffer. SDS-PAGE and Western blotting with streptavidin-POD (ExtrAvidin-peroxidase, Sigma) was performed. CSN5-dependent deneddylation was inhibited by the addition of 5 mM o-phenanthroline (OPT, Roth).
To determine the deubiquitinating activity of the purified CSN complex, 1 lg of tetra-Ub (Ub 4 , Biomol) was added to the pre-incubated CSN complex. SDS-PAGE and Western blotting using the anti-Ub antibody was performed after 1 h incubation. Kinase activity was detected by addition of 0.5 lg recombinant c-Jun and [c-32 P]-ATP to the pre-incubated CSN complex. Phosphorylation of c-Jun was determined as described previously [30].

Subunit CSN6 is cleaved by effector caspases
In order to study changes of the CSN during apoptosis Jurkat T cells were treated for 6 h with the anti-Fas-antibody (Fas-Ab) or with etoposide to activate the extrinsic or intrinsic pathways of apoptosis, respectively. Under these conditions, more than 50% of cells are apoptotic [28]. Apoptosis was blocked with the caspase inhibitor z-VAD. After 6 h treatment cells were lyzed and subunits of the CSN were analyzed by Western blotting. Probing of cell lysate with antibodies against all eight subunits of the CSN revealed that most of the CSN subunits remained stable after 6 h of exposure to anti-Fas-Ab or etoposide (see Fig. 1a). Cleavage of subunit CSN6 but not of other CSN subunits was clearly observed after both apoptotic stimuli. The cleavage was caspase-specific, since z-VAD inhibited the formation of the CSN6 fragment completely. In Fig. 1b the time course of CSN6 cleavage during apoptosis in HeLa as well as in Jurkat cells is shown. In this case apoptosis in HeLa cells was induced by TRAIL and in Jurkat cells by etoposide. The cleavage of CSN6 was clearly detected after 2-3 h in both cell lines. This indicated that cleavage of CSN6 was not an early event in apoptosis and was most likely mediated following the activation of effector caspases. After 6 h of apoptosis more than 50% of CSN6 was cleaved.
The notion that effector caspases cleave CSN6 was confirmed by experiments shown in Fig. 2. The purified CSN complex was incubated with activated recombinant Casp3 or Casp7. CSN6 was completely cleaved after 4 h in the presence of Casp3, whereas it was only partially processed by Casp7 (Fig. 2a). Under these conditions, Casp7 is active as we have recently shown that it can cleave some substrates, such as CENP-C and INCENP more efficiently than Casp3 [34]. Proteolysis was specific, because it was inhibited by z-DEVD, a caspase inhibitor (Fig. 2a). These in vitro results demonstrated that complex-bound CSN6 was cleaved and supported our observations with cells (see Fig. 1) where free CSN6 does not occur.
Similar to the complex-bound protein, the recombinant full-length CSN6 can be cleaved by Casp3 or Casp7 (see Fig. 2b). However, we did not see any cleavage of a truncated version of CSN6, the CSN6D29 mutant protein, missing the N-terminal amino acids 1-29. Therefore the consensus sequence for caspase cleavage M 20 EVD 23 was the most likely target for Casp3 and Casp7. By mutating D 23 to A the full-length CSN6 protein became completely resistant against caspase cleavage. Interestingly, although CSN subunits possess numerous caspase consensus sequences, M 20 EVD 23 of CSN6 was the only one cleaved during apoptosis. Apoptosis (2008) 13:187-195 189 During the preparation of this manuscript it has been published that CSN6 can be also cleaved in SK-BR3 and in various epithelial cell lines [35]. In these studies Casp8 seems to be critical for CSN6 processing, whereas in our hands Casp8 was less effective as compared to Casp3 (data not shown).
Caspase cleavage of CSN6 is accompanied by cleavage of Rbx1 The CSN assembles with CRLs into super-complexes [33,36,37], in which CSN6 interacts with Rbx1 and CSN2 with cullins [5,6]. We were interested to see whether besides CSN6 additional components of the CSN-CRL complexes were modified by activated caspases during apoptosis. Therefore Jurkat T cells were treated for 6 h with etoposide resulting in 60% apoptosis. After treatment protein complexes were separated by density gradient centrifugation. As seen in Fig. 3a, CSN subunits in apoptotic cells sedimented into the same fractions as in control cells or in cells incubated in the presence of etoposide and the caspase inhibitor z-VAD. These data indicate that the CSN complex does not fall apart during extensive induction of apoptosis. This was confirmed by non-denaturing gel electrophoresis of Jurkat and of HeLa cell lysates after induction of apoptosis (see Fig. 3b). Migrations of complexes that cross-react with the anti-CSN6 antibody were not affected by apoptosis. This was also true for the associated enzyme USP15, which co-sedimented with the CSN into fraction 8 (data not shown) and partially dissociated during glycerol gradient centrifugation as published before [31]. As seen in Fig. 3a, no cleavage of USP15 was observed despite the induction of extensive apoptosis (60%). In contrast, marked cleavage of the CRL component Rbx1 was obtained similar to that seen with CSN6. Obviously Rbx1 cleavage was caspase-specific, because it was inhibited by z-VAD. The *10 kDa Rbx1 fragment produced by caspases was recognized by the used antibody raised against the C-terminal Rbx1 peptide 98-108 indicating that the caspase-consensus cleavage site is localized in the RING domain of the protein. As seen in Fig. 3c, significant cleavage of Rbx1 was only observed after 4-5 h. These data demonstrate that CSN6 cleavage (see Fig. 1b) preceded caspase-dependent cleavage of Rbx1.
Cleavage of CSN6 by caspases modifies the deneddylation by the CSN but not the activity of the associated USP15 or kinases The isopeptidase activity catalyzing deneddylation has been localized to CSN5, which exhibits a metalloprotease MPN +domain [38]. In addition, the CSN is associated with a deubiquitinating enzyme called USP15 and with kinases such as CK2 and PKD [30,31,39]. To investigate the  impact of CSN6 cleavage by caspases on CSN functions, we studied the influence of activated Casp3 on the enzymatic activities of the purified CSN. For this purpose the CSN was pre-incubated with recombinant active Casp3 for 1 h. Under these conditions CSN6 was cleaved almost completely (Fig. 4a, panel anti-CSN6). In control experiments the CSN and Casp3 were pre-incubated with o-phenanthroline (OPT), an inhibitor of the deneddylase CSN5 [38], which had no influence on CSN6 cleavage. In contrast, pre-incubation in the presence of z-DEVD completely blocked CSN6 cleavage (Fig. 4a, panel anti-CSN6). Deneddylation was measured with a synthesized peptide (Nedd8-Cul1-pep) consisting of the C-terminal amino acids 31-76 of Nedd8, which was linked via an isopeptide bond to the Cul1 peptide 719-724. The Cul1 peptide was biotinylated at K 723 . The isopeptide bond was formed between the C-terminal G 76 residue of Nedd8 and the e-NH 2 group of K 720 of Cul1 (see ''Material and methods''). The peptide was specifically synthesized, because the MPN + -based deneddylating activity of CSN5 prefers isopeptide bonds [7]. The Nedd8-Cul1-pep was added after 1 h preincubation together with the caspase inhibitor z-DEVD. The reaction was stopped after 3 h and Western blots were performed with streptavidin-POD. As demonstrated in Fig. 4a, panel streptavidin-POD, approximately 30% of the Nedd8-Cul1-pep was cleaved after 3 h by purified CSN. Surprisingly, the cleavage was significantly faster upon CSN treatment with activated Casp3 (Fig. 4a, CSN +  Casp3). The Casp3-stimulated CSN-mediated cleavage of Nedd8-Cul1-pep was blocked by OPT as well as by the Casp3 inhibitor z-DEVD (see Fig. 4a, panel streptavidin-POD) demonstrating that (i) the cleavage was catalyzed by the metalloprotease CSN5 and (ii) the activation of deneddylation was induced by Casp3.
As shown before (see Fig. 3a), the associated USP15 was not affected by etoposide treatment. This was confirmed by measuring the tetra-Ub (Ub 4 ) cleavage activity after treatment of the purified CSN with activated or with inhibited Casp3. As seen in Fig. 4b, panel anti-ubiquitin, the Ub 4 cleavage activity did not change during 1 h (Fig. 4c) indicating that USP15 was not affected by the caspase. In addition, the associated kinase activities were not significantly influenced by Casp3 treatment as demonstrated with c-Jun as substrate (Fig. 4b, panel  autoradiography). Fig. 4 Casp3 cleavage of CSN6 stimulates the deneddylating activity of the CSN complex. (a) Purified CSN complex (750 nM) was preincubated with or without equimolar Casp3 (CSN + Casp3) or with Casp3 in the presence of 1.5 lM z-DEVD (inhib. Casp3) for 1 h (-1 to 0 h) at 37°C. Biotin-coupled Nedd8-Cul1 peptide (Nedd8-Cul1-pep) was added after pre-incubation (0 h) to detect deneddylating activity of the CSN. As an inhibitor of CSN5 deneddylase, the metalloprotease inhibitor OPT (5 mM) was used. Western blotting was performed using the anti-CSN6 antibody as well as streptavidin-POD. CSN6 cleavage products were labeled with an asterisk (*). Nedd8-Cul1-pep cleavage was quantified by densitometric analysis. The density at 0 h was 100% and the peptide decay was monitored for 3 h. (b) Casp3 treatment of the CSN complex has no impact on its associated deubiquitinating or kinase activities. Purified CSN complex was pre-incubated as in (a). To detect deubiquitinating activity Activation of CSN-mediated deneddylation should reduce the neddylation status of cullins in cells and might subsequently cause the disassembly of CRLs. To test this hypothesis neddylation was analyzed by Western blotting in Jurkat T cells (see Fig. 4c). After 4-6 h of apoptosis neddylation of endogenous proteins, most likely of cullins, decreased significantly as indicated by the anti-Nedd8 antibody. The effect was caspase-dependent, since z-VAD stabilized neddylated proteins. To analyze whether the Cul1 neddylation status was affected we induced apoptosis in Jurkat T cells with etoposide for 6 h after transfection with Cul1. As shown in the Western blots in Fig. 4d, Cul1 neddylation was reduced during apoptosis. This effect was completely blocked by z-VAD.

Discussion
Here we demonstrate that the CSN subunit CSN6 is cleaved during apoptosis by activated caspases. This is in agreement with recent findings [35]. Our experiments, however, clearly show that activation of both intrinsic as well as the extrinsic apoptotic pathways result in CSN6 cleavage after 2-3 h suggesting that effector caspases such as Casp3 might be responsible for the effect in vivo. In fact, our in vitro data confirm Casp3 as the most effective protease. The CSN6 cleavage site D 23 is conserved in higher eukaryotes, which makes it a target in other species too.
We show for the first time that the processing of CSN6 precedes the caspase-dependent cleavage of Rbx1. It has been published that the RING component of many CRL complexes, Rbx1, directly interacts with CSN6 in the CSN-CRL super-complexes [5,6]. Presumably CSN6 modification makes Rbx1 accessible for caspase cleavage. Most likely activated Casp3 binds to the CSN-CRL complex and cleaves CSN6 first, which facilitates the modification of the adjacent Rbx1 protein. In CRL complexes Rbx1 is essential for the ligation of Ub moieties to substrate proteins [2]. The irreversible caspase-dependent cleavage of Rbx1 produces a fragment (see Fig. 3a) indicating cleavage inside the RING domain. Therefore, Rbx1 modification during apoptosis most likely causes inactivation of the affected Ub ligase and should significantly reduce CRL-mediated ubiquitination. Interestingly, as demonstrated before Ub conjugates accumulate up to 4 h in Jurkat T cells in the presence of etoposide due to the inactivation of the 26S proteasome by caspases. However, after 4 h the level of Ub conjugates decreases perhaps as a result of the CRL inactivation (see Fig. 5b in [26]). Moreover, Rbx1 has been identified not only as a ubiquitinating enzyme in CRL complexes, but also as a ligase for Rub1, the yeast homolog of Nedd8 [40]. The decrease of neddylated proteins including Cul1 during apoptosis shown in Fig. 4c and d could be at least in part a result of Rbx1 inactivation. The RING domain protein Rbx1 is highly evolutionarily conserved [41] and, therefore, caspasedependent inactivation of CRLs might be a common event in higher eukaryotes.
Another consequence of CSN6 cleavage caused by apoptosis is the stimulation of the deneddylating activity. In the CSN core complex the MPN-domain protein CSN6 directly interacts with the MPN + -domain subunit CSN5 [42]. The increase of CSN5 isopeptidase activity might be explained by a conformational change induced by cleaving off 23 amino acids of CSN6. Since CSN-mediated deneddylation prevents CRL complex assembly [2,4,43], it is conceivable to assume that the elevation of this activity is another strategy to knockout CRLs during apoptosis. In fact, it has been shown that the typical Cul1-CRL substrate, p27 Kip [44], is accumulated in tumor cells after treatment with etoposide [45]. Another Cul1-CRL substrate, p57 Kip [46], promotes the mitochondrial apoptotic pathway [47].
Interestingly, overexpression of the cleavage-resistant CSN6D23A mutant had no significant effect on apoptosis in MCF-7 cells [35] indicating that CSN6 cleavage is not necessary for the process. Nonetheless, cleavage of CSN6 and its consequences seem to be part of the programmed inactivation of the UPS during apoptosis. As a result proapoptotic factors such as p27 Kip , p53 or Smac are stabilized driving the apoptotic process to its completion. For the first time our data demonstrate that CSN-mediated deneddylation can be regulated by active Casp3 and that the CSN executes a specific function during the apoptotic process.