BIMEL, an intrinsically disordered protein, is degraded by 20S proteasomes in the absence of poly-ubiquitylation

BIM-extra long (BIMEL), a pro-apoptotic BH3-only protein and part of the BCL-2 family, is degraded by the proteasome following activation of the ERK1/2 signalling pathway. Although studies have demonstrated poly-ubiquitylation of BIMEL in cells, the nature of the ubiquitin chain linkage has not been defined. Using ubiquitin-binding domains (UBDs) specific for defined ubiquitin chain linkages, we show that BIMEL undergoes K48-linked poly-ubiquitylation at either of two lysine residues. Surprisingly, BIMELΔKK, which lacks both lysine residues, was not poly-ubiquitylated but still underwent ERK1/2-driven, proteasome-dependent turnover. BIM has been proposed to be an intrinsically disordered protein (IDP) and some IDPs can be degraded by uncapped 20S proteasomes in the absence of poly-ubiquitylation. We show that BIMEL is degraded by isolated 20S proteasomes but that this is prevented when BIMEL is bound to its pro-survival target protein MCL-1. Furthermore, knockdown of the proteasome cap component Rpn2 does not prevent BIMEL turnover in cells, and inhibition of the E3 ubiquitin ligase β-TrCP, which catalyses poly-Ub of BIMEL, causes Cdc25A accumulation but does not inhibit BIMEL turnover. These results provide new insights into the regulation of BIMEL by defining a novel ubiquitin-independent pathway for the proteasome-dependent destruction of this highly toxic protein.


Introduction
linkage on BIM EL has not been previously defined. Up to eight different types of Ub chain linkage might exist and this complexity is interpreted within cells by various UBDs, some of which have exquisite specificity for individual types of Ub chain linkage Komander, 2009). To define the Ub chain linkage on BIM EL we used the immobilized Ub-associated (UBA) domains of GST-Dsk2 and GST-Mud1, which specifically bind K48-linked Ub chains (Ohno et al., 2005;Trempe et al., 2005;Lowe et al., 2006;Komander et al., 2009), and the NZF domain of GST-Tab2c, which is specific for K63-linked chains (Kulathu et al., 2009).
GST-UBDs on GSH-agarose beads were used in a pulldown assay to capture and partially purify poly-Ub BIM constructs from cell extracts. We first confirmed the specificity of GST-Dsk2 and GST-Mud1 for K48-linked chains using antibodies specific for K48-or K63-linked poly-ubiquitin chains (Newton et al., 2008). HEK293 cells were transfected with empty vector or BIM EL , and cell extracts were incubated with GST-Dsk2 UBA immobilized on GSH-agarose beads. These samples were readily detected by the K48-specific poly-ubiquitin antibody, which revealed a characteristic smear of poly-Ub species running up the gel, whereas the K63-specific poly-ubiquitin antibody failed to detect any material in the GST-Dsk2 pulldowns except the weak 'ghost' of the GST-Dsk2 protein itself, which is present in excess (Fig. 1A). The overexpression of BIM EL made no difference to the pattern of K48-linked poly-Ub species, indicating that it represents a tiny fraction of the total K48-linked poly-ubiquitin in cells. When these same samples were blotted with antibodies to BIM, we could readily detect full-length BIM EL and its poly-Ub species running as a smear up the gel in the sample from BIM EL -transfected cells. We could also detect endogenous BIM EL binding to GST-Dsk2 UBA, although the much lower levels meant that poly-Ub species of the endogenous BIM EL were difficult to detect (Fig. 1A). Identical results were obtained when we used GST-Mud1 UBA to precipitate BIM EL from control and transfected HEK293 cells (supplementary material Fig. S1A).
To define the specificity of the interaction between BIM EL and GST-Dsk2 UBA, we performed several further experiments. First, it is known that activation of the ERK1/2 pathway promotes the phosphorylation and enhances the turnover of BIM EL Ewings et al., 2007). Consistent with this, when expressed in HR1 cells (HEK293 cells stably expressing the conditional kinase ⌬RAF-1:ER*) (Ewings et al., 2007), the basal level of poly-Ub BIM EL , detected by binding to GST-Dsk2, was greatly enhanced when BIM EL phosphorylation was promoted by 4hydroxytamoxifen (4-HT)-dependent activation of the ⌬RAF-1:ER-MEK1/2-ERK1/2 pathway (Fig. 1B). Second, we compared BIM EL binding to GST-Dsk2 with that of BIM L and BIM S (Fig.  1C), which lack the ERK1/2 and RSK phosphorylation sites that are thought to be the primary signal for poly-Ub and turnover. We again observed good binding of BIM EL , and poly-Ub species were readily apparent. However, binding was very weak with BIM L and particularly BIM S and no poly-Ub species were detected (Fig. 1D), consistent with reports that BIM L and BIM S stability is not regulated by ERK1/2 (Luciano et al., 2003;Ley et al., 2004;Wickenden et al., 2008). We also observed that the binding of BIM EL was drastically reduced when an inactive mutant form of GST-Dsk2 UBA was used in these pulldown experiments (Fig. 1D), and similar results were obtained with an inactive, mutant form of the GST-Mud1 UBA domain (supplementary material Fig. S1B). Finally, the GST-Tab2c NZF domain, which binds K63-linked poly-ubiquitin chains but not K48-linked chains (Kulathu et al., 2009), failed to precipitate poly-Ub BIM EL whereas GST-Dsk2 was again effective (supplementary material Fig. S1C). Thus, using the K48 linkage specificity of the Dsk2 and Mud1 UBA domains we demonstrate for the first time that BIM EL is subject to K48linked poly-Ub in cells and this is enhanced following activation of the ERK1/2 pathway. After 18 hours, cells extracts were precipitated with GST-Dsk2 UBA beads. Pulldowns were divided into three, fractionated by SDS-PAGE, and immunoblotted with antibodies specific for K48-linked Ub, K63-linked Ub, BIM or GST; the input samples were immunoblotted with antibodies for BIM or ERK1. (B)Cycling HR1 cells were transfected with empty vector (EV) or HA-BIM EL for 16 hours before being switched to serum-free medium and treated with or without 100nM 4-HT + 10M MG132 for 1 hour. Whole cell lysates were generated and used for precipitation of poly-ubiquitylated protein using GST-Dsk2 UBA beads. Pulldowns were fractionated by SDS-PAGE and immunoblotted with antibodies specific for BIM or GST; the input samples were immunoblotted with antibodies for BIM, P-ERK1/2 or ERK1. (C)Representation of BIM EL , BIM L and BIM S indicating the ERK1/2-responsive domain, DLC binding domain, BH3 domain, membrane binding motif (M) and the position of the two lysine residues, K3 and K108, numbered according to rat BIM EL . (D)Cycling HEK293 cells were transfected with empty vector (EV) or pcDNA3 plasmids encoding HA-BIM EL (EL), HA-BIM L (L) or HA-BIM S (S). After 18 hours, cells extracts were precipitated with wild-type or mutant GST-Dsk2 UBA beads. Pulldowns were fractionated by SDS-PAGE and immunoblotted with antibodies specific for BIM or GST; the input samples were immunoblotted with antibodies for BIM or P-ERK1/2.

BIM EL ⌬KK undergoes normal ERK1/2-driven proteasomedependent turnover
Covalent attachment of Ub typically takes place at lysine residues. BIM EL contains only two lysine residues at K3 and K108 (numbered according to the rat sequence) (Fig. 1C) so we mutated either K3 (BIM EL ⌬K) or K3 and K108 (BIM EL ⌬KK). These mutants were transfected into HR1 cells that were then treated with 4-HT+MG132 ( Fig. 2A). Wild-type BIM EL again exhibited a basal level of poly-Ub that was enhanced by 4-HT treatment. Mutation of K3 reduced the degree of basal and 4-HT-driven poly-Ub and caused the loss of certain poly-Ub species, whereas we failed to detect poly-Ub of BIM EL ⌬KK despite overexposure of the blots ( Fig. 2A). Thus, poly-Ub can take place at both lysine residues in BIM EL and mutation of both is required to generate a non-poly-Ub form.
Because BIM EL ⌬KK was not poly-Ub in cells, we anticipated that it would accumulate at higher levels and so elicit greater cell death than wild-type BIM EL . However, we found that BIM EL and BIM EL ⌬KK were equally effective at killing when transiently expressed in HEK293 cells (Fig. 2B). Furthermore, western blots revealed that wild-type BIM EL and BIM EL ⌬KK were expressed at similar levels in transfected HR1 cells ( Fig. 2A). These results prompted us to evaluate the turnover of the BIM EL ⌬KK protein directly. When HR1 cells were transfected in parallel with haemagglutinin (HA)-BIM EL (Fig. 2C) or HA-BIM EL ⌬KK (Fig.  2D) the two proteins again expressed at similar levels and exhibited very similar turnover in response to activation of the ERK1/2 pathway by 4-HT (Fig. 2C,D). As a further control, we observed that the ⌬RAF-1:ER*-driven turnover of both HA-BIM EL and HA-BIM EL ⌬KK was inhibited when ERK1/2 activation was prevented by the MEK1/2 inhibitor U0126 (supplementary material Fig. S2A,B). Thus, both wild-type BIM EL and BIM EL ⌬KK exhibited very similar ⌬RAF-1:ER*-driven, MEK1/2-dependent turnover, despite BIM EL ⌬KK being defective for poly-Ub.

Proteasome-dependent degradation of BIM EL in the absence of poly-Ub
In considering other pathways that might contribute to the rapid turnover of BIM EL in the absence of poly-Ub we examined autophagy, a catabolic process in which the cell's own components are degraded by recruitment to autophagolysosomes. We compared immortalized mouse embryo fibroblasts (iMEFs) from wild-type and Atg5-/-mice (Kuma et al., 2004), which are defective for autophagy as judged by LC3 processing (Fig. 3A). The basal level of BIM EL was higher in Atg5-/-iMEFs compared to wild-type cells, and Atg5-/-iMEFs also exhibited a more pronounced increase in BIM EL compared to wild-type when the cells were serum starved (Fig. 3A). However, when we added cycloheximide, which both inhibits protein synthesis and activates ERK1/2, to serum-starved cells we found that the turnover of BIM EL at 3 and 6 hours was essentially identical between the two cell types (Fig.  3A). These results suggest that autophagy might contribute to determining the basal level of BIM EL but plays little or no role in acute ERK1/2-driven turnover of BIM EL .
Since the first description , dozens of laboratories have shown that ERK1/2-driven turnover of BIM EL is proteasomedependent in a wide variety of cell types, as judged by the use of small molecule proteasome inhibitors including MG132, bortezomib (Velcade), and lactacystin. We used these same inhibitors to examine the acute turnover or long-term accumulation of HA-BIM EL ⌬KK. MG132 was able to effectively inhibit the ⌬RAF-1:ER*-driven turnover of HA-BIM EL ⌬KK (Fig. 3B) and similar results were obtained with bortezomib ( Fig. 3C). Furthermore, when cells were transfected with BIM EL constructs and treated chronically with lactacystin, both BIM EL and BIM EL ⌬KK proteins accumulated to the same degree and with the same kinetics (Fig. 3D). Together these results suggested the presence of an alternative poly-Ub-independent pathway for proteasome-dependent degradation of BIM EL ⌬KK following ERK1/2 activation.

BIM EL and BIM EL ⌬KK are degraded by 20S proteasomes
A recent structural study reported that BIM is an IDP , although the functional consequences of this were not 971 Degradation of BIM EL by 20S proteasomes

Fig. 2. Mutation of lysine residues within BIM EL abolishes poly-Ub but does not prevent ERK1/2-dependent phosphorylation and degradation.
(A)HR1 cells were transfected with HA-BIM EL , HA-BIM EL K (K3R) or HA-BIM EL KK (K3R+K108R) for 16 hours. Cells were then treated with serum-free medium + 10M U0126 (SF U0) or 100 nM 4-HT + 10M MG132 (4HT MG) for one hour. Cells were harvested and poly-Ub proteins precipitated using GST-Dsk2 UBA. Input and precipitations were separated by SDS-PAGE and immunoblotted with antibodies to BIM, P-ERK1/2 and ERK1. (B)Cycling HR1 cells were co-transfected with empty vector (EV), HA-BIM EL or HA-BIM EL KK in combination with EGFP-spectrin in a 5:1 ratio to allow the selection of BIM EL -positive cells; UTF, untransfected. Cells were fixed and stained with propidium iodide, and the sub-G1 DNA content of GFP-positive cells measured by flow cytometry. Data is expressed as mean ± s.d. of triplicate cell replicates from a single experiment representative of three. (C,D)HR1 cells were transfected with HA-BIM EL (C) or HA-BIM EL KK (D) for 16 hours before being switched to serum-free medium containing emetine. Cells were then treated with vehicle control (control) or 100 nM 4-HT for 1, 2 or 4 hours. Samples were prepared for BIM, P-ERK1/2 and ERK1/2 analysis by western blot. In each case, BIM EL levels were quantified by densitometry, normalized to the loading control (ERK1), expressed as a percentage of BIM EL present at time zero and represented graphically.
investigated. There is a growing appreciation that some disordered proteins can be degraded by uncapped 20S proteasomes independently of poly-Ub (Tsvetkov et al., 2008;Baugh et al., 2009); indeed, cleavage by the 20S proteasome has been proposed as an operational definition for IDPs (Tsvetkov et al., 2008). We used FoldIndex (Prilusky et al., 2005) to assess the distribution of folded and unfolded regions in BIM EL using p21 CIP1 and PCNA (proliferating cell nuclear antigen) as comparators. This confirmed that p21 CIP1 was extensively unfolded (Kriwacki et al., 1996), whereas PCNA was almost exclusively folded ( Fig. 4A), consistent with the PCNA crystal structure (Gulbis et al., 1996). In comparison, BIM EL was largely unfolded, notably at the N-terminus and towards the C-terminus, though not including the C-terminal hydrophobic tail. Similar results were obtained when we used IUPred (Dosztanyi et al., 2005), to predict regions of disorder (supplementary material Fig. S3), confirming that BIM EL is an IDP .
Prompted by this, we investigated whether BIM EL was degraded by 20S proteasomes. We first generated BIM EL in vitro in a coupled transcription and translation (T&T) reaction and analysed the products using the GST-Dsk2 pulldown assay. Wild-type BIM EL synthesized in vitro was poly-ubiquitylated and this was reduced in the BIM EL ⌬K mutant and abolished in the BIM EL ⌬KK mutant (Fig. 4B), reflecting previous observations in cells ( Fig. 2A). To assess their degradation, [ 35 S]methionine-labelled BIM EL or BIM EL ⌬KK were synthesized in the T&T reaction and incubated with purified 20S proteasomes. In common with p21 CIP1 (Touitou et al., 2001;Tsvetkov et al., 2008), both BIM EL and BIM EL ⌬KK were rapidly degraded by 20S proteasomes whereas PCNA, a folded, globular protein was not (Fig. 4C). Thus, BIM EL is an IDP

972
Journal of Cell Science 124 (6)   that can be degraded by 20S proteasomes in the absence of poly-Ub.
One of the earliest consequences of ERK1/2-dependent phosphorylation of BIM EL is promotion of its dissociation from pro-survival BCL-2 proteins such as MCL-1 or BCL-x L (Ewings et al., 2007); indeed, the BIM EL ⌬BH3 mutant, with three point mutations in its BH3 domain, is defective for binding to MCL-1 or BCL-x L and exhibits accelerated turnover in cells in the absence of ERK1/2 signalling (Ewings et al., 2007). Some IDPs are protected from 20S proteasomal degradation by interactions with their partner proteins (Alvarez-Castelao and Castaño, 2005;Tsvetkov et al., 2008) and, consistent with this, we observed that when incubated with recombinant MCL-1, recombinant HA-BIM EL was protected from degradation by 20S proteasomes (Fig.  4D,E). This protection was not complete, but then it is likely that not all the MCL-1 protein produced in the T&T reaction was correctly folded to allow binding of all the BIM EL . In addition, we noted that MCL-1 was itself partially degraded by 20S proteasomes (supplementary material Fig. S4A), consistent with recent reports (Stewart et al., 2010). Thus the effects of MCL-1 in this assay are probably an underestimate. The specificity of this effect was underlined by the demonstration that BIM EL ⌬BH3, which is defective for MCL-1 binding, was not protected by preincubation with recombinant MCL-1 (supplementary material Fig. S4B). Thus, binding of the BH3 domain of BIM EL to one of its biological targets, MCL-1, protects it from degradation by 20S proteasomes.
Because activation of ERK1/2 can promote the poly-Ub of BIM EL and its rapid turnover we speculated that Ub-dependent degradation by the 26S proteasome (UD/26S) and Ub-independent degradation by the 20S proteasome (UI/20S) pathways might operate in parallel, with the UD providing rapid and efficient targeting and the UI pathway serving as a failsafe, back-up pathway. Indeed, degradation of IDPs by the 20S proteasome has been described as 'degradation by default' (Asher et al., 2006;Tsvetkov et al., 2009b) and the fact the BIM EL ⌬KK turned over following ERK1/2 activation served as some support for this model. On the basis of this, we reasoned that inhibition of the UD/26S pathway might not prevent BIM EL turnover. To test this we knocked down the Rpn2 subunit of the 19S proteasome cap to prevent the assembly of 26S proteasomes without affecting 20S proteasomes (Tsvetkov et al., 2009a). In cycling cells, knockdown of Rpn2 was effective and caused the accumulation of Cdc25A, a folded protein that is degraded by the UD/26S pathway (Fig. 5A); however, this had no effect on basal BIM EL expression, suggesting the possibility of 26S-independent degradation. However, we were concerned that the low basal level of ERK1/2 and RSK activity in cycling HEK293 cells was not sufficient to provide a strong signal for BIM EL turnover so that conditions were not optimal for observing any effects of Rpn2 knockdown. Instead, HR1 cells were transfected with RNAi oligonucleotides and subsequently stimulated with 4-HT to promote rapid BIM EL turnover. Knockdown of Rpn2 caused some delay in the turnover of BIM EL at early time points but ultimately did not prevent it, so that after 6 hours of stimulation the BIM EL turnover was equal in both cell populations (Fig. 5B,C). Because BIM EL ⌬KK is turned over in a proteasome-dependent manner (Fig. 3) these results support the hypothesis that UD/26S and UI/20S pathways for BIM EL turnover can operate in parallel, and that when the UD/26S pathway is inhibited BIM EL turnover can still proceed by 20S-dependent degradation, albeit after a delay.

Inhibition of cullin-based E3 ligases has no effect on basal BIM EL expression or ERK1/2-driven BIM EL turnover
It has recently been shown that coordinated phosphorylation of BIM EL by ERK1/2 and RSK provides a binding site for SCF -TrCP1/2 , which promotes BIM EL poly-Ub (Dehan et al., 2009). The Skp1/Cul1/F-box protein (SCF) complexes are perhaps the bestunderstood RING-type E3s (Cardozo and Pagano, 2004;Nakayama and Nakayama, 2006;Frescas and Pagano, 2008) and consist of a catalytic core (Cul1 and a RING protein) linked by an adaptor (SKP1) to a substrate-specific receptor subunit (the F-box protein). Recognition by F-box proteins often requires phosphorylation of the substrate, providing a link between signalling, poly-Ub and protein turnover. Examples of SCF complexes include SCF FBXW7 , which promotes the destruction of cyclin E (Koepp et al., 2001); SCF Skp2 , which promotes the destruction of p21 CIP1 and p27 KIP ; and SCF -TrCP1/2 , which promotes destruction IB and Cdc25A (for a review, see Frescas and Pagano, 2008). If the UI/20S pathway could substitute for UD/26S, then we reasoned that disrupting BIM EL poly-Ub by targeting its E3 ligase might not affect the levels of endogenous BIM EL . To address this we inhibited cullin function using interfering mutants or RNAi-mediated knockdown of Cul1.

973
Degradation of BIM EL by 20S proteasomes First, we expressed a dominant-negative interfering mutant of cullin1 (dnCul1) (Shirogane et al., 2005) in cycling HEK293 cells and examined the impact on basal protein expression. This approach was validated by showing that dnCul1 caused a substantial accumulation of p27 KIP1 , cyclin E and Cdc25A, all of which are recognized SCF substrates (Fig. 6A); indeed, like BIM EL , Cdc25A is a target of SCF -TrCP1/2 (for a review, see Frescas and Pagano, 2008). Despite this, dnCul1 did not cause an increase in the basal levels of BIM EL (Fig. 6A). Because activation of ERK1/2 promotes BIM EL turnover, we again reasoned that cycling cells were not optimal for BIM EL turnover. As an alternative we again used HR1 cells, where activation of the ⌬RAF-1:ER*-MEK1/2-ERK1/2 pathway results in a rapid and robust turnover of BIM EL (Ewings et al., 2007). For these experiments we also used a dominant-negative mutant of Ubc12 (dnUbc12) (Amir et al., 2001), a protein that catalyses the NEDD8 conjugation of a conserved lysine residue that is required for the function of all cullins. The efficacy of dnUbc12 was validated by showing that it also caused accumulation of p27 KIP1 when expressed in cycling HEK293 cells, just like dnCul1 (Fig. 6B). Despite this, neither dnCul1 or dnUbc12 could block ⌬RAF-1:ER*-driven turnover of BIM EL (Fig. 6C). Finally, knockdown of Cul1 in HR1 cells by RNAi was also without effect on ⌬RAF-1:ER*-driven turnover of BIM EL (Fig. 6D).
These results revealed that selective inhibition of cullin1 (by two strategies) or inhibition of all cullins (using dnUbc12) failed to impair ERK1/2-dependent BIM EL turnover. This data could suggest the existence of other E3 Ub ligases for BIM EL in addition to SCF -TrCP1/2 (Dehan et al., 2009), but they are also consistent with our demonstration of an alternative Ub-independent pathway for BIM EL degradation.

Discussion
Poly-Ub of a target protein determines the fate of that protein; K48-linked chains provide a signal for proteasomal degradation whereas K63-linked chains are important in the assembly of proinflammatory signalling complexes and protein trafficking (Ikeda and Dikic, 2008;Komander, 2009). The nature of the polyubiquitin chain linkage is typically defined by using individual Ub point mutants (K48R, K63R, etc.) to compete with endogenous Ub. However, such approaches have limitations (see Newton et al., 2008), require substantial overexpression of the mutant Ub (which can be difficult to achieve), and in our hands gave variable results. As an alternative we have made use of the ability of certain UBDs to discriminate between different Ub chain linkages . Such specificity underpins the use of ubiquitylation as a regulatory signal in a variety of cellular processes; however, in this instance we have used it simply as a diagnostic tool. The Dsk2 and Mud1 UBA domains have been defined as K48-specific by comparing their ability to bind chemically synthesized K48-linked, K63-linked and linear poly-Ub chains; similar studies have defined the K63 specificity of the TAB2C NZF domain (Kulathu et al., 2009). In addition, crystal structures have revealed the molecular basis by which the UBA domains from Dsk2 and Mud1 utilize the unique conformational features of K48-linked chains for specific recognition (Trempe et al., 2005;Lowe et al., 2006). Although the data are still at an early stage, emerging studies suggest that K11 linkages might be a second proteasomal degradation signal (Xu et al., 2009;Wu et al., 2010). K11-linked chains are compact and structurally distinct from K48-linked chains (Bremm et al., 2010) and neither Mud1 UBA or TAB2C NZF domains are able to interact with K11linked poly-ubiquitin chains in pulldown experiments (Kulathu et al., 2009) (D.K., unpublished). Thus, the results of testing against all currently available Ub chain types strongly suggest that the Dsk2 and Mud1 UBA domains are specific for K48 chains, thereby validating our approach of using these UBA domains as diagnostic tools. Accordingly, our results demonstrate for the first time that BIM EL is subject to K48-linked poly-Ub and define a simple assay for monitoring this linkage specificity in cells. We believe that such an assay might be more generally applicable to the study of K48-linked poly-Ub of proteins and are currently testing this.

Two pathways for ERK1/2-driven proteasomal degradation of BIM EL
The ERK1/2 signalling pathway is the major pathway controlling the proteolytic turnover of BIM EL  (for a review, see Gillings et al., 2009). Coordinated phosphorylation by ERK1/2 and RSK1/2 targets BIM EL for poly-Ub by SCF -TrCP (Dehan et al., 2009). Despite this, we found that a BIM EL ⌬KK mutant failed to undergo poly-Ub but was still subject to ERK1/2-driven, proteasome-dependent turnover. Prompted by the suggestion that BIM EL is an IDP , we found that BIM EL (wildtype or ⌬KK) could be degraded in a Ub-independent manner by 20S proteasomes. Further evidence for Ub-independent turnover of BIM EL came from the demonstration that Rpn2 knockdown delayed, but did not prevent, BIM EL turnover. Finally, inhibition of cullin-based E3 Ub ligases disrupted the turnover of p27 KIP1 , cyclin E and Cdc25A (the latter a validated target of SCF -TrCP1/2 ) but had no effect on BIM EL turnover. Taken together, these results provide strong, independent lines of evidence for an additional Ubindependent pathway for BIM EL turnover by the proteasome. We suggest that ERK1/2-driven degradation of BIM EL proceeds by two pathways: classical Ub-dependent degradation by the 26S proteasome (UD/26S) or Ub-independent degradation by the 20S proteasome (UI/20S). In the former, phosphorylation of BIM EL by ERK1/2 and RSK will allow the E3 ligase SCF TrCP to promote the poly-Ub and 26S-dependent destruction of BIM EL (Dehan et al., 2009). However, because ⌬RAF-1:ER* can still promote the turnover of BIM EL ⌬KK, it is apparent that ERK1/2 activation can also promote BIM EL turnover by the proteasome, independently of poly-Ub. UI/20S degradation would explain why we could not prevent BIM EL turnover by ⌬KK mutation, Rpn2 knockdown or inhibition of cullin-based E3 ligases.
UI/20S is thought to be an important and evolutionarily conserved proteolytic pathway. Uncapped 20S proteasomes are abundant in mammalian cells and degrade up to 20% of cellular proteins (Baugh et al., 2009), including some proteins involved in cell cycle control and apoptosis. For example, the tumour suppressor p53 is degraded by a classical UD/26S mechanism but its N-terminus is disordered, allowing degradation by 20S proteasomes (Asher et al., 2005a;Asher et al., 2006;Tsvetkov et al., 2009a). Similarly, ornithine decarboxylase (Asher et al., 2005b), p21 CIP1 (Touitou et al., 2001) and IBa (Krappmann et al., 1996;Alvarez-Castelao and Castaño, 2005) are all poly-Ub but can also be degraded by 20S. This default degradation pathway might serve as a back-up to ensure timely removal of these biologically important proteins (Asher et al., 2006).

BIM EL , MCL-1 and the nanny model
We previously showed that phosphorylation of BIM EL promotes its dissociation from pro-survival BCL-2 proteins; indeed, BIM EL ⌬BH3, which is defective for binding to pro-survival BCL-2 proteins, is turned over more rapidly than wild-type BIM EL suggesting that dissociation contributes to BIM EL turnover (Ewings et al., 2007). Interestingly, some IDPs are protected from 20S degradation by interactions with their partner proteins (Alvarez-Castelao and Castaño, 2005;Tsvetkov et al., 2008;Tsvetkov et al., 2009a) and this has led to the suggestion that such partner proteins serve as 'nannies' (Tsvetkov et al., 2009b). Our demonstration that binding of BIM EL to MCL-1 could protect it from degradation by 20S proteasomes (Fig. 4D,E) suggests that MCL-1 and presumably other pro-survival BCL-2 proteins serve as nannies for BIM EL . In the course of writing up this work, a study reported that MCL-1 is degraded by 20S proteasomes in a Ub-independent manner (Stewart et al., 2010); the authors speculated that this was dependent on the disordered N-terminus of MCL-1. There are remarkable parallels with our study on BIM; both proteins are subject to poly-Ub and turnover by the UD/26S pathway but both proteins can also be degraded by the UI/20S pathway when poly-Ub is prevented by mutation of all lysine residues. Studies have previously shown that the binding of BH3-only proteins can influence MCL-1 stability in cells; Noxa binding promoted MCL-1 degradation , whereas binding of BIM (Czabotar et al., 2007) or PUMA (Mei et al., 2005) promoted MCL-1 stabilization. Although binding of BH3-only proteins could stabilize MCL-1 by displacing the E3 ligase MULE (Warr et al., 2005;Zhong et al., 2005), these new studies suggest an additional explanation whereby BIM EL and MCL-1 serve as nannies for each other to prevent their degradation by 20S. The biological consequences of this are likely to be complex: dissociation of BIM EL from MCL-1 (Ewings et al., 2007) would facilitate degradation of BIM EL , which would support cell survival; conversely, this might also facilitate destruction of MCL-1 by the UD/26S or UI/20S pathways, which would tend to support cell death. The net effect is likely to be determined by the expression of other BCL-2 family proteins.
On the basis of these observations we propose that in addition to poly-Ub by -TrCP (Dehan et al., 2009), phosphorylationdependent dissociation from BCL-2 proteins represents a signal for targeting BIM EL to the 20S proteasome. This might be because 'free' BIM EL is intrinsically disordered or because dissociation of BIM EL unmasks disordered regions or favours a disordered conformation that is a prerequisite for access to 20S proteasomes (Baugh et al., 2009). Phosphorylation of BIM EL might contribute directly, because high net charge is often associated with disorder (Dyson and Wright, 2005). In this scenario, UD/26S would provide a rapid, targeted destruction mechanism, whereas the UI/20S pathway might operate to remove any excess or 'free' BIM EL and serve as a 'failsafe' to ensure the removal of BIM EL if ubiquitylation is disrupted. Quite how BIM EL accesses the 20S proteasome remains to be seen. Non-Ub p21 CIP1 can bind to the REG 20S proteasome regulator (Li et al., 2007;Chen et al., 2007) or might be recognized by the C8 subunit of the 20S proteasome (Touitou et al., 2001). To date, we have not been able to detect specific binding of BIM EL to C8 (M.J., C.M.W. and S.J.C., unpublished observations).
In summary we have used the inherent linkage specificity of the UBA domains of Dsk2 and Mud1 as a diagnostic tool to demonstrate that BIM EL undergoes ERK1/2-driven, K48-linked poly-Ub. Despite this, poly-Ub is not a prerequisite for proteasomal degradation; we suggest that BIM EL can also be degraded in a Ubindependent fashion by 20S proteasomes following its dissociation from pro-survival BCL-2 proteins, by virtue of its intrinsic disorder. These different mechanisms of degradation reflect the biological imperative of destroying BIM EL in a timely fashion. BIM is one of the most toxic BH3-only proteins because of its ability to engage with all the pro-survival BCL-2 proteins . Consequently, its abundance must be tightly regulated by multiple mechanisms to ensure that cell death is only initiated under appropriate conditions. from Noboru Mizushima (Tokyo Medical and Dental University, Tokyo, Japan), were provided by Aviva Tolkovky (Universtiy of Cambridge, Cambridge, UK).

Assay of BIM EL poly-Ub in cells
HR1 cells were transfected with the indicated plasmids. After 24 hours, cells were serum starved in the presence of U0126 to inactivate ERK1/2 or treated with 4-HT + MG132 to activate ERK1/2 and inhibit the proteasome. Following lysis in TG lysis buffer  cell extracts were retained (input) or incubated end-overend with immobilized GST-UBDs at 4°C for 2 hours, washed and fractionated by SDS-PAGE as described for GST-BIM pulldowns (Ley et al., 2004).

Assay of BIM EL turnover in cells
HR1 cells were transfected with the indicated BIM EL plasmids. After 24 hours, cells were treated with emetine to block new protein synthesis and chased in serum-free medium or in the presence or absence of 4-HT, U0126 or MG132. Cell extracts were fractionated by SDS-PAGE and immunoblotted as described.

In vitro 20S proteasomal degradation assay
Purified 20S proteasomes were generated as described previously (Asher et al., 2005b). [ 35 S]methionine-labelled proteins translated in vitro were incubated with 1 g of purified 20S proteasomes in 150 mM NaCl, 50 mM Tris HCl, pH 7.5 at 37°C for the indicated times. Reactions were then resolved by SDS-PAGE, visualized by autoradiography and quantified by phosphorimaging (Tsvetkov et al., 2008).

RNAi against Cul1 and Rpn2
HR-1 cells (150,000 cells per well plated in a six-well plate) were plated in antibioticfree media. Transfection complexes were prepared according to the manufacturer's instructions using Lipofectamine 2000 (Invitrogen). For Rpn2 knockdown, 200 pmol siRpn2-PT (5Ј-GCTCATATTGGGAATGCTTAT-3Ј) and 200 pmol siRpn2-MJ (5Ј-GGATACTTCTCCAGGATCA-3Ј) were mixed and 400 pmol of a control siRNA used (murine Bim1 5Ј-GGAGGAACCTGAAGATCTG-3Ј). Complexes were applied and cells incubated for 48 hours before being aspirated and fresh transfection complexes applied for a further 48 hours. Cells were then harvested immediately for western blot analysis, or an emetine chase experiment performed. For Cul1 knockdown, a pool of three siRNAs targeting Cul1 was used (Santa Cruz Biotechnology) alongside a human siRNA control (Santa Cruz Biotechnology). HR-1 cells were transfected for 24 hours, the transfection media was then aspirated and serum-free media applied. Following 16 hours of serum-free treatment, emetine (50 M) was added to block protein synthesis followed 30 minutes later by addition of 4-HT for 8 hours.
We thank Rebecca Gilley and other members of the Cook laboratory for advice and encouragement and John Pascal (The Babraham Institute, Cambridge, UK) for initial advice on T&T reactions. We also thank Jane Endicott and Jean-Francois Trempe (University of Oxford, Oxford, UK), for providing the GST-Mud1 UBA constructs; Aaron Ciechanover, for dnUbc12 and J. Wade Harper, for dnCul1. This work was funded by core funding to the Babraham Institute and a response mode project grant from the BBSRC (BB/E02162X/1). C.M.W. was supported by a BBSRC PhD studentship and C.L.J. was supported by a BBSRC/AstraZeneca CASE studentship.