Disassembly of Lys11 and Mixed Linkage Polyubiquitin Conjugates Provides Insights into Function of Proteasomal Deubiquitinases Rpn11 and Ubp6*

Background: Deconjugation of polyubiquitin is an essential step in preparing substrates for proteolysis by the 26S proteasome. Results: Proteasome-associated DUBs, Rpn11 and Ubp6, process long Lys11- or Lys63-linked polyUb more efficiently than Lys48 linkages. Conclusion: 26S proteasomes can completely disassemble a mixed/branched polyUb conjugate. Significance: These observations call into question what constitutes an efficient signal for proteasome targeting versus proteolysis. Protein homeostasis is largely dependent on proteolysis by the ubiquitin-proteasome system. Diverse polyubiquitin modifications are reported to target cellular proteins to the proteasome. At the proteasome, deubiquitination is an essential preprocessing event that contributes to degradation efficiency. We characterized the specificities of two proteasome-associated deubiquitinases (DUBs), Rpn11 and Ubp6, and explored their impact on overall proteasome DUB activity. This was accomplished by constructing a panel of well defined ubiquitin (Ub) conjugates, including homogeneous linkages of varying lengths as well as a heterogeneously modified target. Rpn11 and Ubp6 processed Lys11 and Lys63 linkages with comparable efficiencies that increased with chain length. In contrast, processing of Lys48 linkages by proteasome was inversely correlated to chain length. Fluorescently labeled tetra-Ub chains revealed endo-chain preference for Ubp6 acting on Lys48 and random action for Rpn11. Proteasomes were more efficient at deconjugating identical substrates than their constituent DUBs by roughly 2 orders of magnitude. Incorporation into proteasomes significantly enhanced enzymatic efficiency of Rpn11, due in part to alleviation of the autoinhibitory role of its C terminus. The broad specificity of Rpn11 could explain how proteasomes were more effective at disassembling a heterogeneously modified conjugate compared with homogeneous Lys48-linked chains. The reduced ability to disassemble homogeneous Lys48-linked chains longer than 4 Ub units may prolong residency time on the proteasome.

genome, classified into five different families; for comprehensive reviews, see Refs. 9 and 10) sculpt the heterogeneous Ub landscape. Thus, some DUBs display broad specificity, whereas others are relatively selective for substrate or for linkage type; some can completely disassemble a chain, whereas others amputate an entire polyUb chain "en bloc" from the substrate (10 -12). All linkage types have been detected in vivo. Lys 48 and Lys 63 linkages and, to a lesser extent, Lys 11 account for the bulk of linkage types (13,14), although relative levels may change in response to specific conditions. Lys 63 linkages have been reported to serve in primarily regulatory non-degradative processes, such as DNA repair and membrane trafficking, whereas Lys 48 is the canonical signal that targets substrates to the proteasome (3,4,15,16). Lys 11 linkages alone or branched with Lys 48 linkages are thought to serve as a degradation signal enhancer of anaphase-promoting complex substrates during cell (17,18) or on ERAD pathways (19).
Several DUBs associate with the 19S RP, where they participate in preparing substrates for degradation and rescue Ub from a similar fate (22,23). Ubp6/USP14 and Rpn11/PSMD14 (yeast/human, respectively) are the primary DUBs studied in this context. UCH37/UCHL5, a third proteasome-associated DUB, is not found in all eukaryotes (24). Ubp6 is a cysteine protease that transiently associates with the proteasome by docking to the Rpn1 subunit in the 19S RP (25,26). In contrast, Rpn11, a MPN ϩ /JAMM zinc metalloprotease (23,27), is an integral component of the 19S RP. Recent crystal structures of Rpn11 MPN ϩ domain in combination with cryo-EM position Rpn11 at the center of the 19S RP (28 -30). Further analysis by a cryo-EM study has hinted that the position of Rpn11 in the proteasome differs by 18 Å between the free and substrateengaged forms, pointing to a dramatic reaction-induced conformational change (31). Rpn11 and Ubp6 activities within the proteasome are synchronized to ATP consumption by neighboring Rpt ATPases and substrate degradation, and therefore complicate understanding of their respective contributions (23,(31)(32)(33)(34). So far, Ubp6/USP14 and UCH37 are thought to trim polyUb chains from the distal end (11,25), gradually decreasing affinity of the conjugate as the chain shortens, whereas Rpn11 amputates entire chains at the proximal side coupled to ATPdependent degradation (11,35). Alternatively, Rpn11 has also been demonstrated to disassemble Lys 63 -linked unconjugated polyUb chains (36). Deciphering DUB action of both Ubp6 and Rpn11 is key to understanding how they influence residency time of polyUb substrates at the proteasome (32,33,(37)(38)(39)(40).
Our main objective was to understand the properties associated with each proteasomal DUB uninfluenced by the many other interlinked ATP-dependent processes of the proteasome. With this in mind, we deliberately studied their enzymatic activities in isolation and in complex by utilizing non-degradable substrates: a panel of polyUb conjugates with defined properties, chain length and linkage type. With this experimental setup, we found that the overall DUB activity of proteasome primarily reflects properties of Rpn11. When incorporated into proteasome, Rpn11 is activated over 100-fold compared with unattached subcomplexes. Due to the broad specificity of Rpn11, proteasomes were able to completely remove all Ub from a model substrate Ubch5b-Ub n , which included diverse Ub modifications on multiple lysines. In summary, proteasomes coordinate multiple DUBs and polyUb receptors to act on a wide range of substrates for both degradative and nondegradative outcomes.

EXPERIMENTAL PROCEDURES
Purification of Yeast Proteasome-The 26S proteasome was purified from wild type Saccharomyces cerevisiae strain using conventional chromatography as described previously (20). The 26S proteasomes were purified from wild type or isogenetic ⌬ubp6 Saccharomyces cerevisiae strains using conventional chromatography as described previously (20). Because Ubp6 is a labile proteasome-associated protein (37,41,42), equimolar levels of Ubp6 were added to purified 26S proteasomes to guarantee stoichiometric levels. Full incorporation of Ubp6 into 26S proteasomes was validated in each case (Fig. 1B). Intact 26S proteasomes were confirmed by non-denaturing PAGE and peptidase activity with Suc-LLVY-AMC as described (20).
Preparation of Proteasomal Deubiquitinases and Subcomplexes-pQE30 plasmids were transformed to Escherichia coli M15 cells, whereas pET28b and pETDuet1 plasmids were transformed to BL-21 Rosetta 2 cells (Novagen). To ensure that Rpn11 was contained in all complexes, Rpn11 was N-terminally His 6 -tagged following a published methodology (43). Proteins were expressed in LB medium at 37°C supplemented with 60 g/ml ampicillin or 50 g/ml kanamycin and induced with 0.5 mM isopropyl ␤-D-1-thiogalactopyranoside at A 600 of ϳ0.7. Following induction, 2-liter cultures were grown overnight at 16°C, harvested, and stored at Ϫ80°C. Cells were lysed using a French press, the lysates were clarified by centrifugation at 18,000 rpm for 30 min at 4°C, and the supernatant was loaded onto a 10-ml His-Trap (GE Healthcare) column that was preequilibrated with 50 mM Tris, pH 7.4, 300 mM KCl, 5% glycerol, 10 mM imidazole in order to isolate recombinant His 6 -Ubp6 and His 6 -Rpn11. A solution of 50 mM pH 7.4, 100 mM NaCl, 100 mM KCl, 5% glycerol, 10 mM imidazole was used for Rpn11 subcomplexes. Proteins were eluted in the same buffer with 280 mM imidazole. Following elution, fractions containing proteins of interest were pooled and dialyzed in PBS, pH 7.4, buffer for His 6 -Ubp6 and His 6 -Rpn11, whereas Rpn11 subcomplexes were dialyzed into 50 mM, pH 7.4, 100 mM NaCl, 100 mM KCl, 5% glycerol. To obtain the highest purity and isolate stoichiometric Rpn11-containing subcomplexes, gel filtration was performed with a Sephacryl 400 column (GE Healthcare), and for monomeric proteins, it was performed with a Superdex 200 column (GE Healthcare). After confirming the purity with SDS-PAGE, all enzymes were aliquoted and stored at Ϫ80°C.
In vitro ubiquitination reactions for preparative assembly of fluorescently labeled Lys 48  Reactions were performed at 30°C for 18 h in the dark. At the end point, each reaction was centrifuged at 20,000 ϫ g for 30 min at 4°C, and tetra-Ub species were isolated using gel filtration (HiLoad 26/60 Superdex 75 pg, GE Healthcare) in PBS, pH 7.4, and then concentrated using an Ultra-15, 10 kDa cut-off (Millipore).
Preparation of Dimeric Ub, Ub 4 , Ub 6ϩ , and Ubch5b-Ub n Conjugates-Monomeric Ub mutants, E2 conjugating enzymes, and human E1 were obtained recombinantly as described (8,48). Enzymatically synthesized Lys 11 -, Lys 48 -, and Lys 63 -linked Ub dimers were assembled by combining a proximally blocked Ub mutant (Ub D77 or Ub-His 6 ) in combination with a distally blocked Lys to Arg Ub variant as described (48,49). Lys 11 -linked dimers were obtained from a reaction containing 10 mg of each Ub-His 6 and Ub K11R,K63R , 500 nM Uba1, 30 M Ube2s, 5 mM tris(2-carboxyethyl)phosphine, and 15 mM ATP in a volume of 2 ml with a 50 mM Tris, pH 8.0, buffer incubated at 30°C for 20 h. Lys 48 -linked dimers were obtained in a similar reaction with E2-25K as the sole E2 and Ub D77 and Ub K48R,K63R monomers. In a similar fashion, reactions to generate Lys 63 -linked dimers contained a 20 M concentration of each Ubc13 and Uev1a with Ub D77 and Ub K48R/K63R monomers. Following the completion of each reaction, 10 ml of cation buffer A (50 mM ammonium acetate, pH 4.5) was added, the solution was centrifuged at 13,000 rpm for 10 min to remove precipitated E1 and E2 enzymes, and the supernatant was slowly injected onto a 5 ml HiTrap SP column (GE Healthcare) at 0.2 ml/min. The polyUb species were eluted with cation buffer B (50 mM ammonium acetate, 1 M NaCl, pH 4.5); exchanged into PBS, pH 7.4, buffer; and concentrated to a final volume of 1 ml. Monomeric and dimeric Ub species were separated on a Superdex 75 size exclusion column (GE Healthcare) in PBS, pH 7.4, buffer with a flow rate of 0.35 ml/min. Fractions containing pure dimers were detected using SDS-PAGE. Non-fluorescent Lys 48 -Ub 4 and Lys 63 -Ub 4 were prepared as described above, without Ub cysteine variants.
High molecular weight Lys 48 -Ub 6ϩ and Lys 63 -Ub 6ϩ chains were created using reaction conditions identical to the dimers with 30 mg of wild type Ub as the only monomer. Following the Superdex 75 column, fractions containing the desired chain lengths were detected by SDS-PAGE and pooled. The autoubiquitinating properties of the E2 Ub-conjugating enzyme, Ubch5b, were exploited to generate Ubch5b-Ub n conjugates under similar reaction conditions using 2 mM of human wild type Ub, 5 mM DTT, 15 mM ATP, and 75 M Ubch5b-His 6 . Ubch5b-Ub n conjugates were purified as described (50). DTT was used in place of tris(2-carboxyethyl)phosphine to minimize the accumulation of thioester-linked Ubs on the active site cysteine of Ubch5b. Desired oligomeric states of Ubc5b-Ub n were detected using 15% SDS-PAGE; exchanged into PBS, pH 7.4, without DTT; and stored at Ϫ20°C.
In Vitro Deubiquitination Reactions of Fluorescently Labeled Tetra-Ub Chains-In vitro deubiquitination time course experiments containing 25 M Ubp6, Rpn8⅐Rpn11 heterodimer, or lid-core were incubated with 25 M Lys48-Ub 4 or Lys 63 -Ub 4 fluorescent substrates in PBS, pH 7.4. Both tetra-Ub substrates included 1% A488-labeled Ub. All deubiquitination reactions were performed at 30°C in the dark. Samples were taken at given time points and analyzed by 18% SDS-PAGE. Gels were analyzed using fluorescence scanning on a Typhoon TM FLA 9500 scanner (GE Healthcare) employing emission filter BPB1 (530DF20) with excitation at 473 nm and Coomassie staining. Fluorescence intensity was quantified using ImageQuant TL software (GE Healthcare).
Gel-based Deubiquitination Assays-All cleavage assays were performed in vitro. 50 nM 26S proteasome was incubated with a 500 nM concentration of either Lys 48 -or Lys 63 -Ub 6ϩ , whereas 1 M lid-core, Rpn8⅐Rpn11 heterodimer, or Ubp6 was incubated with 10 M Ub 6ϩ . 10 M isolated Rpn11 was incubated with 100 M Ub 6ϩ in order to observe activity. All reactions were incubated at 30°C. Samples of the deubiquitination assay were taken at the indicated points, mixed with 5ϫ protein loading dye, boiled at 95°C for 5 min, and kept at room temperature until analysis by 15% SDS-PAGE. Gels were visualized with Coomassie staining or transferred and immunoblotted with anti-ubiquitin antibody polyclonal rabbit anti-Ub Dako (z0458) in a 1:1,000 dilution. For detection of Lys 11 , Lys 48 , and Lys 63 linkages, rabbit monoclonal Lys 11 (Millipore catalog no. 2021885), Lys 48 (Millipore catalog no. 2197314), and Lys 63 (Millipore catalog no. 2063204) antibodies were used in a 1:1,000 dilution. IgG goat anti-rabbit HRP conjugate (Bio-Rad catalog no. 170-6515) was used as the secondary antibody in a 1:50,000 dilution. Gel bands were quantified using ImageJ analysis. Enzymatic efficiency in units of nM Ϫ1 ϫ min Ϫ1 was determined using Equation 1.
All reactions were repeated in triplicate. Average values with S.D. are summarized in Table 1.
Mathematical Analysis of Fluorescent Tetra-Ub Disassembly Intermediates-At each time point of tetra-Ub disassembly, the fluorescent signal for each Ub species was quantified. Disassembly was analyzed by fitting simultaneously the intensities of the gel bands corresponding to Ub 4 , Ub 3 , Ub 2 , and Ub 1 over the time course to Equations 2-5) and treating the rate constants as fitting parameters. Here the tetramer is initially cleaved at an exo position (rate constant k 433 ) or the endo (k 432 ), as schematically illustrated in supplemental Fig. S2. The trimer is cleaved to a dimer (k 332 ), and finally, the dimer is converted to monomeric Ub (k 231 ). Each rate constant was determined numerically, and the results are summarized in (supplemental Fig. S2).

MS/MS Analysis of Ubch5b-Ub n Modifications-The
Ub linkages on the Ubch5b-Ub n substrate were determined by tryptic digestion in solution followed by LC-MS/MS analysis. Ubch5b-Ub n in PBS buffer was reduced by 10 mM DTT and reacted with 40 mM iodoacetamide (at 25°C) and trypsinized (modified trypsin (Promega)) at a 1:100 enzyme/substrate ratio for 18 h at 37°C. The resulting tryptic peptides were desalted and resolved by reverse-phase chromatography on 0.075 ϫ 200-mm fused silica capillaries (J&W) packed with Reprosil reversed phase material (Dr. Maisch GmbH, Ammerbuch-Entringen, Germany). The peptides were eluted with linear 65-min gradients of 5-45% and 15 min at 95% acetonitrile with 0.1% formic acid in water at flow rates of 0.25 l/min. Mass spectrometry was performed by "hybrid" mass spectrometer (Orbitrap, Thermo) in a positive mode using a repetitively full MS scan followed by collision-induced dissociation of the seven most dominant ions selected from the first MS scan. The mass spectrometry data were analyzed using Trans Proteomic Pipeline (TPP) version 4.6.3 (51). TPP-processed centroid fragment peak lists in mzML format were searched against a database composed of S. cerevisiae yeast and E. coli proteins (Uniprot) and human Ub supplemented with their corresponding decoy sequences (as described on the Matrix Science Web site). The database searches were performed using X! Tandem with the k-score plugin through TPP. Search parameters include trypsin cleavage specificity with two missed cleavage sites; cysteine carbamidomethyl as a fixed modification; and lysine ubiquitination, methionine oxidation, and protein N-terminal acetylation as variable modifications. Peptide tolerance and MS/MS tolerance were set at 20 ppm and 0.8 Da, respectively. X! Tandem refinement included semistyle cleavages and variable lysine GG modification. Peptide and protein lists were generated following Peptide Prophet and Protein Prophet analysis using a protein false discovery rate of Ͻ1%.

RESULTS
Proteasome-associated DUBs Are Active in Isolation-To assess the individual properties of proteasome-associated DUBs, we prepared isolated enzymes to compare their activity with that of 26S proteasome holoenzyme (Fig. 1, A and B). Recent studies have demonstrated that a truncated Rpn8⅐Rpn11 heterodimer as well as an Rpn8⅐Rpn11 fusion used for crystallization maintain catalytic activity (28,29). For in depth investigation of Rpn11 activity, we recombinantly purified full-length Rpn11, an Rpn8⅐Rpn11 "heterodimer," and another subcomplex containing Rpn5, -6, -8, -9, and -11 (referred to as "lid-core"), which has been demonstrated to be a stable complex (43,52,53). We remind the reader that Rpn8 is an inactive paralogue, of Rpn11; both are adjacently located at the center of the 19S RP (28 -30, 52). The composition of each preparation was confirmed by SDS-PAGE and MS analysis ( Fig. 1A and supplemental Table  S1). Purified proteasomes from yeast (54,55) were used to compare the activity of these enzymes in isolated subcomplexes with that of intact proteasome holoenzymes. We found that Ubp6, a non-essential subunit transiently associated with proteasome (37,41,42), easily dissociated during purification, leading to inconsistent levels of Ubp6. Therefore, to ensure full incorporation (verified by gel shift (41)), recombinant Ubp6 was added back at a 1:1 molar ratio (Fig. 1B).
Given that both Lys 48 and Lys 63 linkages are processed by the proteasome (56,57), all recombinant enzyme preparations were tested on polymeric Lys 48 and Lys 63 chains representative of proteasome targeting signals. In complete isolation as a standalone enzyme, Rpn11 was competent to process polymeric Lys 63 linkages (Fig. 1C). Despite not having been previously documented, this property is consistent with other members of the MPN ϩ /JAMM metalloprotease family, such as AMSH and BRCC36 (58,59), which have strict specificity for Lys 63 linkages. Lending further support to the idea Rpn11 as a member of the MPN ϩ /JAMM metalloprotease, the zinc ion chelator 1,10-phenanthroline (OPA) abolished all measurable DUB activity (Fig. 1C, right), reaffirming that Rpn11 is a MPN/ JAMM metalloprotease family member. To demonstrate that Rpn11 was solely responsible for DUB activity in the lid-core, we introduced the D122A substitution (37) to the active enzyme, which resulted in ablation of DUB activity (Fig. 1D). In contrast to Rpn11, significant levels of Ubp6 have been documented unassociated from proteasomes (26,42). And indeed, Ubp6 efficiently disassembled both Lys 48 and Lys 63 polyUb chains uncomplexed, on its own (Fig. 1E). This DUB activity was arrested by N-ethylmaleimide (NEM), a general inhibitor of cysteine proteases. Moreover, substitution of the active site cysteine C118A rendered the enzyme inactive (Fig. 1F).
Ubiquitin Dimers Are Processed by 26S Proteasome but Are Poor Substrates for Associated DUBs-The ease of synthesis and homogeneity of dimeric Ub facilitates direct comparison of linkage types. We initiated our investigation of proteasomeassociated DUBs by synthesizing the most abundant linkage types: Lys 11 , Lys 48 , and Lys 63 . The K m of dimeric substrates for Rpn11 has been reported to be in the hundreds of M (29), which makes it impractical to assess conditions where the substrate concentration exceeds K m . Therefore, by working well below the K m , we calculated the apparent rate of conversion of dimer to monomer and extracted the bimolecular rate constant (nM Ϫ1 ϫ min Ϫ1 ) of enzymatic efficiency. This analysis allowed the direct quantitative comparison of different enzymes for an array of substrates.
All preparations of proteasome-associated DUBs had measurable activity for all three dimers ( Fig. 2A). Both Rpn11 containing complexes showed a significant preference for Lys 11 linkages; however, the lid-core converted each substrate at approximately double the efficiency (Table 1). From this analysis, the lid-core and Rpn8⅐Rpn11 heterodimer exhibited similar linkage preferences, with efficiency for Lys 11 approximately twice that for Lys 48 and Lys 63 ( Table 1). Regardless of linkage type, the larger lid-core complex had enhanced activity demonstrated by the greater efficiency. In contrast, under the identical conditions, the preferred linkage for Ubp6 was Lys 48 ( Fig. 2A and Table 1).
It is notable that DUBs intimately associated with the proteasome process a Lys 11 linkage on par with Lys 48 and Lys 63 . This led us to test whether the preference of the whole proteasome reflects those of its constituent DUBs. Disassembly of Lys 11 was remarkably faster than of the other linkage types by 26S holoenzymes (Fig. 2B). With an increase in more than 2 orders of magnitude, the enzymatic efficiency of proteasome dwarfs its   FEBRUARY 20, 2015 • VOLUME 290 • NUMBER 8 isolated constituents (Table 1). We attribute the bulk of this activity to the presence of Rpn11 in the proteasome because proteasomes lacking Ubp6 had comparable activity (Fig. 2B and Table 1). Processing of Lys 11 is a new undocumented property of Rpn11 beyond the characteristic Lys 63 cleavage activity of other MPN ϩ /JAMM family members.

Substrate Specificity of Proteasome-associated DUBs
To reconcile the documented activity of Ubp6 on model substrates containing a single Ub domain (37, 60 -62) with our current results using dimeric Ub, we assessed how each proteasomal DUB processed linear Ub fusions lacking an isopeptide bond, using the proven Ub-GFP substrate (37,63,64). Ubp6 deconjugated nearly all of the Ub form GFP, whereas neither Rpn11 complex at equimolar concentration was capable of efficiently hydrolyzing the peptide bond within the same time frame (Fig. 2C). The potency of Ubp6 to remove a substrateattached Ub may explain one of the phenotypes associated with loss of Ubp6 function: low levels of cellular Ub caused by degradation of Ub at the proteasome along with its conjugate (65,66). The results thus far point to Rpn11 accounting for the majority of Lys 11 and Lys 63 processing at the proteasome.
Patterns of Lys 48 -Ub 4 and Lys 63 -Ub 4 -Use of dimeric Ub substrates for DUBs is informative regarding linkage specificity yet probably does not represent natural proteasome targeting signals, which often contain longer Ub polymers (67). Tetramers can provide information on the position preference (i.e. endo/ exo chain cleavage) within a polymer as well as general effect of length on activity outcome. Homogeneously linked Lys 48 or Lys 63 tetramers were obtained in high purity and used to assess DUB action associated with proteasome. Both complexes of Rpn11 continued the trend demonstrating Lys 63 linkage preference over Lys 48 (Fig. 3A). Surprisingly, with tetramers, Ubp6 also exhibited preference for Lys 63 linkages in contrast to dimeric substrates, where Lys 48 was processed with a greater efficiency ( Fig. 3A and Table 1). The resistance of Lys 48 -Ub 4 to disassembly by individual DUBs was carried over to the proteasome, which again showed a dramatic preference for Lys 63 -Ub 4 . As was the case with processing of dimers, Ubp6 did not greatly contribute to the total DUB activity of the proteasome (Fig. 3B). Nonetheless, proteasome holoenzymes were 2 orders of magnitude more efficient in processing and therefore were able to fully convert the Lys 63 -Ub 4 substrate within the time frame of the reaction. Increased Lys 63 efficiency and the minimal effect of Ubp6 are in agreement with early reports that Rpn11 is activated when incorporated in the proteasome. USP2, a DUB unrelated to proteasome function was able to fully convert Lys 48 -Ub 4 to monomeric units, highlighting that this substrate is not problematic for dedicated enzymes (Fig. 3C).
Despite having potentially more binding sites on a per mole basis, cleavage of Lys 48 -Ub 4 by proteasomal DUBs occurred at a lower efficiency than Lys 48 -Ub 2 . Given the unique spatial environments within Ub polymers, individual isopeptide bonds are probably recognized differently ( Fig. 4A and B). To address DUB preference for linkages within a chain, we designed a set of homogeneously linked Lys 48 or Lys 63 tetra-Ub chains in which a randomly positioned Ub unit was labeled with Alexa 488 on position 20 using the Ub S20C mutant. Concentrations of DUBs were increased relative to the previous reactions, in order to collect information on all reaction intermediates. For this reason, this specific experimental setup was not used to extract enzymatic efficiency but was subject to separate kinetic analysis. Changes in concentrations of initial tetra-Ub substrate as well as the mono-, di-, and tri-Ub cleavage products were quantified independently by in-gel fluorescence. The total fluorescent signal emanating from all Ub species remained constant throughout the time course for the reaction, indicating no loss of material and supporting use of this approach for monitoring reaction progression. Nevertheless, the relative contribution of Ub intermediates did differ for certain enzyme substrate pairs. Ubp6 acting on Lys 48 -Ub 4 stood out by initially generating one primary intermediate; Lys 48 -Ub 2 accumulated until the Ub 4 substrate was depleted, and only then was Lys 48 -Ub 2 fully converted into the final product, mono-Ub (Fig. 4C, green curve). The low abundance of Ub 3 as an intermediate suggests that Ubp6 favors the endo isopeptide bond in Ub 4 . Such behavior was not observed for Ubp6 with Lys 63 -Ub 4 (Fig. 4D). Likewise, Rpn11 in either complex equally generated all three Ub intermediates initially, suggesting random action on the three isopeptide bonds in Ub 4 (Fig. 4, E-H). As a note, concentration of Ub intermediates depends on both rate of generation from longer chains and conversion into shorter ones, hence the unique profile of tri-and di-Ub intermediates in each reaction.
We expanded on the advantages offered by the fluorescent labeling, which enabled a more accurate quantification of changes in each intermediate concentration. Therefore, we were not limited to monitoring solely depletion of initial substrate and could apply our mathematical models to deconvolute parallel reactions. We wrote the reaction progression equations to describe the pattern of all intermediates displayed in panels. Trend lines derived from fitting parallel reactions in Equations 2-5 were overlaid on the experimental data points (Fig. 4,  C-H). Supporting our initial observation for Ubp6 acting on Lys 48 -Ub 4 , the rate constant for the endo reaction (conversion of tetramer to dimer (k 432 )) was more than twice that for each

for proteasome and associated DUBs
Enzymatic efficiency of polyUb chain disassembly by deubiquitination enzymes in this study was calculated using Equation 1 (see "Experimental Procedures"). Disassembly efficiency of Ub 2 substrate was determined by quantifying the Ub 2 gel band intensity. Values for Ub 4 , Ub 6ϩ , and UbcH5b-Ub n were determined by following the decrease in signal intensity of residual substrate relative to the entire signal generated by Ub in all of its forms. examined the disassembly of longer (Ub 6ϩ ) homogeneously linked chains, which we refer to herein as Lys 11 -Ub 6ϩ , Lys 48 -Ub 6ϩ , and Lys 63 -Ub 6ϩ . Processing efficiency was calculated based on the depletion of the starting polymeric substrate at initial time points to avoid the complicated treatment of multiple reactions as the reaction progressed. Both Lys 11 -Ub 6ϩ and Lys 63 -Ub 6ϩ were disassembled by Rpn11-containing complexes with an efficiency exceeding that of tetramers or dimers (Table 1). Rpn11-related DUB activity revealed an interesting trend: increased efficiency with chain length for Lys 11 and Lys 63 polyUb and an inverse relationship for Lys 48 linkages. In fact, disassembly efficiency by either Rpn11 subcomplex for Lys 48 -Ub 6ϩ was not calculated due to slow progression ( Fig. 5A and  B). Once again, the lid-core exhibited a higher efficiency than the Rpn8⅐Rpn11 heterodimer (Table 1). Ubp6 was able to process both Lys 11 -and Lys 63 -linked Ub 6ϩ conjugates with a similar efficiency to the lid-core and could even act on long Lys 48 chains (Fig. 5C and Table 1). However, with these longer substrates, Lys 48 -linked Ub 6ϩ was not the preferred linkage for Ubp6 (Table 1).  FEBRUARY 20, 2015 • VOLUME 290 • NUMBER 8

JOURNAL OF BIOLOGICAL CHEMISTRY 4695
Because polyUb is considered the canonical signal for targeting substrates to the proteasome (3, 68), we continued to assess 26S proteasome action on Ub 6ϩ of each linkage type. Enzymatic reactions were performed at 0.5 M substrate concentration, intentionally below the reported K m for tetra-Ub (56) in order to extract bimolecular rate constants from initial rates at non saturating conditions. The proteasome was competent to disassemble Lys 11 -Ub 6ϩ and Lys 63 -Ub 6ϩ , in contrast to Lys 48 -Ub 6ϩ which was remarkably resilient to proteasome action (Fig. 6, A-C, and Table 1). The highly compact nature of Lys 48 -Ub 6ϩ (supplemental Fig. S1) may explain the relatively poor processing of this substrate by proteasome-associated DUBs, in contrast to another DUB adapted for Lys 48 linkages (Fig.  6D).
With Lys 63 -Ub 6ϩ , we expand on proteasome action for Lys 63 linkages previously documented for Lys 63 -Ub 4 (56). Although proteasome holoenzymes processed Ub polymers faster than did the individual DUBs, the overall preference for Lys 11 and Lys 63 over Lys 48 mimicked that of Rpn11 with a dramatic increase in apparent rate constants (Table 1). This completes   FEBRUARY 20, 2015 • VOLUME 290 • NUMBER 8 the trend observed in this study that larger Rpn11 assemblies possess greater rate constants, in particular for Lys 11 and Lys 63 linkages. To address potential catalytic and non-catalytic contributions of Ubp6 to overall proteasome DUB activity, we compared proteasomes containing active Ubp6 with those incorporating inactive (Ubp6 C118A ) or no Ubp6. Inactivation of or the absence of Ubp6 had a minimal effect on processing of unanchored Ub 6ϩ of the three linkages tested (Fig. 6, A-C). The cumulative observations from our panel of unanchored polyUb chains did not identify a significant contribution of Ubp6 in the proteasome to DUB activity.

Substrate Specificity of Proteasome-associated DUBs
Rpn11 C Terminus Hints at Autoinhibitory Mechanism-Thus far, the bulk of DUB activity measured for proteasome complexes can be attributed to Rpn11. This suggests that Rpn11 is signifi-cantly activated when incorporated into proteasome holoenzyme. Activation of CSN5, a structurally related JAMM/MPN ϩ metalloprotease, is mediated by unblocking the active site zinc ion by repositioning Glu 104 (69). The equivalent position in the ins-1 loop of Rpn11 is not conserved, suggesting involvement of additional resides in autoinhibition isolated from the proteasome. The C terminus of Rpn11 has been documented to facilitate the delicate coordination of Rpn11 into the proteasome (70). This encouraged us to test the impact Rpn11 C terminus has on its DUB activity. For this purpose, we obtained the truncated Rpn8(1-186)⅐Rpn11(1-229) heterodimer spanning the MPN domains (Fig. 7A). As a case in point, we reevaluated Rpn8(1-186)⅐Rpn11(1-229) with Lys 11 -Ub 2 (Fig. 7B). Direct comparison clearly demonstrates significant activation of truncated Rpn8(1-186)⅐Rpn11(1-229) compared FIGURE 6. A and B, proteasomes purified from ubp6 C118A , ⌬ubp6, or wild-type strains were incubated with Lys 48 -Ub 6ϩ (A) and Lys 63 -Ub 6ϩ (B). C, Lys 11 -Ub 6ϩ was assayed with ⌬Ubp6 and wild-type proteasome. D, full disassembly of Lys 48 -Ub 6ϩ was achieved by the addition of OTUB1; all blots were visualized using anti-Ub (IB ␣Ub).
with the full-length heterodimer (Fig. 7C). Activation also extended to Lys 63 chains, yet even with this enhanced activity, truncated Rpn8(1-186)⅐Rpn11(1-229) still had reduced ability to process Lys 48 -Ub 4 (Fig. 7D). Truncation of the C terminus of CSN5 correlates with a 12-Å transition of Glu 104 in the ins-1 loop away from the active site Zn 2ϩ ion (Fig. 7E). Available high resolution structural data position the ins-1 loop of Rpn11 at an intermediate position between the autoinhibited and derepressed forms of CSN5 (Fig. 7E). We speculate that rearrangement of the C terminus of Rpn11 upon incorporation into proteasome is partially emulated by truncation described herein.
Complete Disassembly of a Heterogeneously Ubiquitinated Substrate Conjugate at the Proteasome-In a cellular environment, proteasomes may encounter more complex substrates heterogeneously modified with multiple Ub linkages rather than unanchored homogeneous chains (7). Moreover, in a conjugate, Ub units distal or proximal to substrate add an extra dimension for deubiquitinating machinery to distinguish. As a representative polymorphic substrate, we utilized the ability of Ubch5b to autoubiquitinate (50, 71) (supplemental Fig. S3). Lys 11 , Lys 48 , and Lys 63 linkages were found across a broad range of modified Ubch5b using linkage-specific antibodies (Fig. 8A). Closer analysis using tryptic MS/MS identified all isopeptide Ub-Ub linkages and even two forked linkages, Lys 6 -Lys 11 and Lys 11 -Lys 27 , simultaneously modifying several lysines on the Ubch5b target (verified on Lys 8 , Lys 128 , and Lys 133 (50)). Ubch5b-Ub n is thereby established as an extreme example of a heterogeneously modified substrate ( Fig. 8B and supplemental  Fig. S4). Processing of such a complex substrate by proteasomal DUBs was an intriguing concept to put to an experimental test.
Rpn11 subcomplexes and Ubp6 were capable of deubiquitinating Ubch5b-Ub n (Fig. 8C). Although enzymatic efficiencies at initial time points were similar ( Table 1), conversion of Ubch5b-Ub n by Ubp6 was more complete (Fig. 8C). This may reflect a broader range of associated activities for Ubp6 as an isolated enzyme. After prolonged incubation with each DUB, we analyzed the residual short modifications on Ubch5b using linkage specific antibodies (Fig. 8D). Ubp6 eliminated nearly all Lys 48 linkages, whereas the lid-core eliminated nearly all of the Lys11 linkages in Ubch5b-Ub n (Fig. 8D). As a side product, accumulation of unanchored dimeric Ub supports the common theme emanating from our current work, that dimers of all linkages are poor substrates compared with longer polyUb chains.
As with homogeneously linked polyUb chains, the proteasome processed Ubch5b-Ub n with an efficiency nearly 100-fold greater than any constituent DUB. Once again, Ubp6 had a minimal effect on overall DUB activity of the proteasome (Fig.  Fig. S4). C, Ubch5b-Ub n was incubated with each indicated enzyme and analyzed with anti-Ub (IB ␣Ubi). D, the 20 h end point for each enzyme was subjected to blotting with Lys 11 , Lys 48 , and Lys 63 linkagespecific antibodies. E, anti-Ub reveals that all polyUb is fully removed from the Ubch5b-Ub n substrate by both ⌬Ubp6 and wild-type proteasome. F, a fluorescent label on Ubch5b emphasizes accumulation of free substrate following action of ⌬Ubp6 and wild type proteasome. Table 1). When presented simultaneously with multiple linkages on the same substrate, the proteasome experienced no limitation in processing, reflecting diverse capabilities harbored in resident DUBs. To determine the fate of Ubch5b modified with this mixture of Ub linkages, we tracked the substrate with a fluorescent label. Indicating a lack of degradation, a Ubfree form of Ubch5b accumulated as a result of proteasomal deubiquitination (Fig. 8F). Consequently, deubiquitination proceeded to completion, removing all modifications from the substrate. Previous studies have demonstrated that branched polyUb chains can serve as substrate for the proteasome (18,48), and expanding on this concept, we demonstrate that mixed and branched modifications render a conjugate as an adequate substrate for deubiquitination (Fig. 8, A-F).

DISCUSSION
The 26S proteasome emerges from this study as a broad spectrum multicatalytic DUB, capable of reducing a heterogeneously modified conjugate containing all linkage types down to mono-Ub. Disassembly of polymeric Ub, especially of well defined linkage type or chain length, poses experimental challenges in the synthesis of substrates and in the isolation, characterization, and quantification of generated products. Due to these experimental constraints, DUBs are routinely studied with synthetic model substrates that allow for easy detection of liberated products or for continuous assays. Hence, fluorescently labeled Ub (e.g. Ub-AMC) or substrates modified by a single Ub are prevalent in experimental setups. The current study found that Ub dimers are poor substrates for deconjugation by proteasomes or their associated DUBs. Not only was processing slow, but Ub dimers failed to expose significant differences between linkages. Purified homogeneously linked Ub tetramers or longer poly-Ub 6ϩ proved useful to flush out differences in the processing of linkage types. For instance, the potency of proteasome complexes or their constituent DUBs for disassembling Lys 11 -linked polyUb became apparent. Longer or more complex chains may provide the three-dimensional spatial information needed for recognition or efficient processing by proteasomes (supplemental Fig. S1).
Regardless of linkage type, processing of chains was generally more efficient than removal of a single conjugated Ub (as in unanchored Ub dimers). However, one notable exception was polyUb linked via Lys 48 (e.g. Lys 48 -Ub 4 or Lys 48 -Ub 6ϩ chains) that were processed with difficulty. The compact conformation of Lys 48 -linked polyUb longer than 4 units (supplemental Fig.  S5) may obstruct access of cleavage sites to certain DUBs (72,73). Could Lys 48 -linked tetra-Ub be the optimal "Goldilocks" proteasomal signal (74) precisely because it is rather resilient to disassembly? In this regard, Lys 63 linkages are thought to be poor proteasome-targeting signals (56,57), presumably because they can be disassembled rapidly and are therefore unable to provide the residency time required to commit a substrate for unfolding. Given that we have now uncovered an ability of proteasomes to rapidly disassemble Lys 11 linkages, it is unclear how they may serve as a targeting signal (17,18,75,76). Beyond homogeneously linked chains, mixing of linkages into a single chain or onto a single substrate may provide for new signaling properties (77, 78), unexpected when considering information from individual linkages alone. Mixed or branched linkages may disrupt a repetitive configuration of a homogeneous chain, exposing initiation sites for DUBs. This may be one explanation for efficient proteasome processing of the heterogeneously modified Ubch5b-Ub n , although it was abundant in Lys 48 linkages. Likewise, a recent study demonstrated that a branched Lys 11 /Lys 48 -linked chain had a greater affinity for proteasome receptor S5a and was also deconjugated faster at the proteasome than either homogeneous Lys 11 -or Lys 48linked chains (18). Nevertheless, linkage branching should be considered on a case-by-case basis, given that some types of forked linkages have been reported to inhibit proteasomal degradation (79), whereas a Lys 48 /Lys 63 -linked branched chain had no effect on deconjugation (48).
In separate studies, activities of recombinant Rpn8⅐Rpn11 or USP14 were demonstrated against short substrates (25,28,29). Further expanding on these findings, Ubp6 and Rpn11 cleaved a range of longer Ub conjugates linked via Lys 11 , Lys 48 , or Lys 63 as well as heterogeneously. Biological roles of Ubp6 or Rpn11 could henceforth be evaluated under the notion that a portion is uncoupled from the proteasome and partakes in biological pathways that do not necessarily involve Lys 48 linkages (80 -82). Involvement of Rpn11 in Lys 63 -linked Ub signaling at DNA double-stranded breaks has been documented, but it was unclear whether Rpn11 acted independently of the 20S (83). Other MPN ϩ /JAMM family DUBs are also known to act in Lys 63 -signaling pathways, such as DNA repair or membrane trafficking (36,58,59,(83)(84)(85). With the development of functional Rpn11 and Ubp6 assays, their direct involvement in these processes could now be reevaluated also for Lys 11 linkages.
Relative to either DUB, proteasome deubiquitination efficiency was enhanced over 100-fold. With a model substrate, Ubch5b-Ub n , a multiubiquitinated globular protein, proteasomes selectively deconjugated rather than proteolyzed the substrate. Under these conditions, removal of Ubp6 had only a marginal effect on overall proteasome DUB activity. One explanation is that proteasome-incorporated Ubp6 remains in a latent form until triggered into action either by substrate fully committed to degradation after initiation of unfolding (60,86,87) or by conformational changes upon engagement of RPT ATPases (32). In either scenario, we propose that the primary contribution of Ubp6 is near the end of the proteasome catalytic cycle. In contrast, deconjugating efficiency of proteasomeincorporated Rpn11 was enhanced against the same model substrate. PolyUb receptors on the proteasome (Rpn10 and Rpn13) may compensate for the low inherent affinity of Rpn11 for Ub (29,56). Rpn10, in particular, is situated adjacent to Rpn11 within the 19S RP (30,43,88,89). Activation of Rpn11 may also be driven through a conformational change upon incorporation into the proteasome involving slight rearrangements of Zn 2ϩ -coordinating residues in the active site (Fig. 7E). That the Rpn8⅐Rpn11 heterodimer was more active as truncated MPN domains rather than as full-length proteins (Fig. 7) points to a role for their C termini in autoinhibition until property incorporated into proteasomes. Indeed, the C termini participate in critical subunit interactions stabilizing mature proteasome (90,91). Given that lid-core was also less active than Rpn11 lacking its C terminus, the full extent of Rpn11 activation probably involves interactions beyond the lid subcomplex, for instance by synchronization with AAA-ATPases (29). In the cell, proteasomes encounter numerous ubiquitinated proteins that are not necessarily intended for degradation. By removing Lys 11 and Lys 63 linkages, the preference of Rpn11 may aid in deflecting these substrates from an unwanted fate, acting independently uncoupled from degradation.
The central role of the proteasome in protein homeostasis, as well as its ability to regulate oncogenic factors and toxic proteins, makes it a lead pharmacological target (92). Inhibition of USP14 increased the turnover of oxidized and damaged proteins by the proteasome (60). In addition, the therapeutic importance of inhibiting USP14 and UCHL5 have been demonstrated for controlling progression of acute myeloid leukemia (63,64) or tumor growth (93). Although its roles in cell viability, life span, and cancers (94 -96) make it a promising drug target (62), there are currently no specific inhibitors for the Rpn11 metalloprotease. An ability to study individual proteasome-associated DUBs in isolation should facilitate intensive ongoing drug discovery efforts (61,63,93,(97)(98)(99) relative to the dynamic, ATP-dependent, multisubunit proteasome complex.