Rpn1 and Rpn2 Coordinate Ubiquitin Processing Factors at Proteasome*

Background: Proteasome substrates are recognized by (poly)ubiquitin receptors or mediated by ubiquitin-like domain-containing shuttles. Results: Multiple ubiquitin-like domain-containing proteins bind transiently to different sites at proteasome subunit Rpn1. Conclusion: Rpn1 and Rpn2 coordinate substrate shuttles, ubiquitin receptors, and deubiquitination enzymes in proximity at the proteasome. Significance: Targeting of (poly)ubiquitin conjugates to the proteasome depends as much on auxiliary factors and substrate shuttles as on direct recognition of the ubiquitin chain. Substrates tagged with (poly)ubiquitin for degradation can be targeted directly to the 26 S proteasome where they are proteolyzed. Independently, ubiquitin conjugates may also be delivered by bivalent shuttles. The majority of shuttles attach to the proteasome through a ubiquitin-like domain (UBL) while anchoring cargo at a C-terminal polyubiquitin-binding domain(s). We found that two shuttles of this class, Rad23 and Dsk2, dock at two different receptor sites embedded within a single subunit of the 19 S proteasome regulatory particle, Rpn1. Their association/dissociation constants and affinities for Rpn1 are similar. In contrast, another UBL-containing protein, the deubiquitinase Ubp6, is also anchored by Rpn1, yet it dissociates slower, thus behaving as an occasional proteasome subunit that is distinct from the transiently associated shuttles. Two neighboring subunits, Rpn10 and Rpn13, show a marked preference for polyubiquitin over UBLs. Rpn10 attaches to the central solenoid portion of Rpn1, although this association is stabilized by the presence of a third subunit, Rpn2. Rpn13 binds directly to Rpn2. These intrinsic polyubiquitin receptors may compete with substrate shuttles for their polyubiquitin-conjugate cargos, thereby aiding release of the emptied shuttles. By binding multiple ubiquitin-processing factors simultaneously, Rpn1 is uniquely suited to coordinate substrate recruitment, deubiquitination, and movement toward the catalytic core. The broad range of affinities for ubiquitin, ubiquitin-like, and non-ubiquitin signals by adjacent yet nonoverlapping sites all within the base represents a hub of activity that coordinates the intricate relay of substrates within the proteasome, and consequently it influences substrate residency time and commitment to degradation.

a dramatic enhancement of Ubp6 and Usp14 deubiquitinating activities in vitro, although the exact mechanism of this activation remains unclear (66). Activation upon proteasome incorporation suggests that in cells some DUBs (such as Ubp6/ USP14) function most effectively in the proteasome, whereas in cytosol their function is inhibited or reduced. As the case may be, Ubp6/USP14 apparently helps remove the (poly)ubiquitin tag from protein substrates, either before or during their translocation into the catalytic chamber of the proteasome for degradation (50,67,68), with an overall net effect of delaying proteolysis (64).
Mediated delivery implies a designated UBL receptor in the proteasome (Fig. 1A). Two proteasome subunits shown to interact with a subset of these shuttles are Rpn1 and Rpn2 (31). Using a series of deletions constructed from a recombinant GST-Rpn1 fusion, association of Rpn1 with Rad23 was roughly characterized (32). Following up on these findings, we set out to chart UBL recognition sites on Rpn1 and Rpn2, prioritize the binding partners of these platforms, and place them in context of competing/augmenting polyUb binding by neighboring subunits. Comparing binding data for UBL and (poly)ubiquitin signals unfurls a general outline of substrate relay events and points to Rpn1 and Rpn2 as the main coordinators of ubiquitinprocessing factors at the proteasome.
Plasmids-Full-length coding sequences of the indicated genes were amplified by PCR using Pfu DNA polymerase from genomic DNA (S. cerevisiae wild-type strain Sub62) (3,69). The 5Ј and 3Ј amplification primers were designed to add restriction sites as indicated in supplemental Table S1. The fragments were cut with the indicated enzymes and subsequently cloned into a specified vector (supplemental Table S1, column 2) for expression with a His 6 tag. Clones were verified by DNA sequencing and then transformed into Escherichia coli M15/BL21 strain for bacterial expression.
Purification of Recombinant Proteins-Transformants for expression of protein of interest were grown in liquid LB media supplemented with 1 mM ampicillin to A 600 0.6 -0.8 at 37°C, followed by 30 min of heat shock at 42°C. Next, 0.1 mM to 1 nM isopropyl 1-thio-␤-D-galactopyranoside was added for induction, and cultures were grown overnight at 16°C, after which cells were harvested. Cells were lysed, and lysates were clarified at 16,000 rpm for 20 min at 4°C, and the supernatant was loaded onto a Ni 2ϩ -nitrilotriacetic acid column (GE Healthcare) that was previously equilibrated with 50 mM Tris, pH 7.4, 500 mM NaCl, and 10 mM imidazole. Washes were performed with same buffer, and bound proteins were eluted using 50 mM Tris, pH 7.4, 500 mM NaCl, and 250 mM imidazole. Samples were then dialyzed for appropriate buffer, and the resulting proteins were stored at Ϫ80°C. Purity was tested by Coomassie staining and immunoblotting. Protein concentration was determined using an ND-1000 spectrophotometer (Nano-Drop). In some cases proteins were subjected to gel filtration for additional step of purification (24 ml of Superdex200).
Glycerol Gradient Fractionation-11-ml glycerol gradients were prepared manually using 10 and 40% (v/v) glycerol stock solutions supplemented with 10 mM MgCl 2 , 25 mM Tris, pH 7.4, 1 mM ATP, 1 mM DTT. 1 ml of whole fractionated yeast cell extract was applied on top of the gradients and resolved by centrifugation at 28,000 ϫ g for 18 h in TH-641 rotor in ultracentrifuge. Twelve 1-ml fractions were collected and assayed for peptidase activity using N-succinyl-Leu-Leu-Val-Tyr-7amino-4-methylcoumarin as was described (3). Protein samples were further resolved by 10 -12% SDS-PAGE or 16.5% Tris/Tricine-PAGE and immunoblotted.
Pulldown Assays-CH-Sepharose 4B (Sigma) was swollen according to the manufacturer's recommendation. Proteins to be immobilized were first dissolved in 50 mM PBS and then were added to activated Sepharose to a final concentration of 0.6 mg/ml and incubated with agitation at 4°C. After 2 h, the gel was washed, and its remaining active groups were blocked with 1 M ethanolamine, pH 8.0, for 2 h. Routinely, about 90% of protein was bound to the matrix. The resin was washed as recommended by the producer, equilibrated in 50 mM Tris-HCl, pH 8.0, 150 mM NaCl, and 0.1% (w/v) Triton X-100, and stored at 4°C. A control resin was prepared in the same way, but instead of protein, the active groups were derivatized with ethanolamine. Each soluble protein was diluted to a final concentration of 100 g/ml in incubation buffer and then incubated with 0.1 ml of slurry of immobilized proteins equilibrated in incubation buffer (50 mM Tris-HCl, pH 8.4, 150 mM NaCl, 0.1 mg/ml BSA, 10% glycerol, and 1% (w/v) Triton X-100). The mixture was agitated at 4°C for 2.5 h and washed extensively with the equilibration buffer followed by additional five washes with equilibration buffer supplemented with 1% (w/v) Triton X-100. Bound proteins were eluted with elution buffer (incubation buffer supplemented with 1 M NaCl and/or 6 M urea) and analyzed by SDS-PAGE and immunoblotting.
26 S Proteasome Pull-out-Purified recombinant proteins were bound to nickel-nitrilotriacetic acid columns (Qiagen) according to standard procedure. The columns were incubated with 26 S proteasomes lacking Ubp6, Rad23, Dsk2, and Ddi1 proteins and tumbled for 2-4 h. The columns were washed with PBS buffer supplemented with 1% Triton and 0.5 M NaCl, and bound proteins were eluted with 250 mM imidazole. The eluates were immunoblotted with antibodies against various proteasomal subunits.
Surface Plasmon Resonance-All experiments were performed using ProteOn XPR36 instruments developed by Bio-Rad. Purified Rpn1 and Rpn2 were diluted in 10 mM sodium acetate at pH 3.0 or 4.5, respectively, and coupled to a general layer medium sensor surface activated with 37.5 mg/ml N-ethyl-NЈ-(3Јdimethylaminopropyl) carbodiimide hydrochloride and 7.5 mg/ml N-hydroxysuccinimide. Proteins to be immobilized were injected in the vertical orientation ProteOn XPR36 fluidics for 5 min (150 l) at 30 l/min and then followed by blocking excess reactive esters with ethanolamine-HCl (Sigma) for 5 min. All of the SPR binding measurements were performed with PBST as the continuous running buffer at 25°C. Rpn13, Ubp6, Rad23, Dsk2, and Ddi1 at the specified concentrations were injected in the horizontal orientation of the Pro-teOn XPR36 fluidics using a flow rate of 40 l/min for 75 s (50 l). In every separate injection, five different concentrations were injected in channels 1-5, and running buffer was injected simultaneously in the sixth channel for signal normalization. All binding sensorgrams were collected, processed, and analyzed using the integrated ProteOn Manager software (Bio-Rad).
In general, the equilibrium dissociation constant (K D ) evaluating the protein/protein binding affinity could be determined using the 1:1 Langmuir binding model or heterogeneous ligand-parallel reaction (HLPR) model (70). For both models, the association rate constant (k on ) and dissociation rate constant (k off ) were derived from a global fit of primary response data over the entire concentration range for each protein-ligand pair to the corresponding binding model using rate Equation 1, and the equilibrium dissociation constant K D was calculated according to Equation 2. The difference between the two binding models is that the HLPR model has two association and two dissociation reactions as schematically shown in Equation 3.
In these equations, C is the concentration of the soluble ligand (referred to as analyte in the SPR literature); R represents the response unit; for the HLPR model R (observed) ϭ R1 ϩ R2. For generation of concentration-dependent binding curves, response levels at equilibrium were plotted against the concentration of the soluble protein (the ligand) and calculated via a nonlinear least squares fit of the Langmuir binding Equation 4,

Rpn1 and Rpn2
Interact with Ubiquitin-processing Proteins-A number of ubiquitin-processing factors known from previous studies to interact with the proteasome were compared for their ability to interact directly with either Rpn1 or Rpn2. Despite high sequence and structural similarities between Rpn1 and Rpn2, the two exhibit very distinctive preferences for the potential binding partners (Fig. 1C). Although Rpn2 was found to bind only Rpn13, Rpn1 showed strong interaction with Ubp6, Rad23, and Dsk2. Additionally, in this assay, Rpn1 showed weak affinity for Rpn10 and Ddi1 (Fig. 1C). It is noteworthy that although Rpn1 bound a number of UDPs, no interaction was measured with ubiquitin, indicating that Rpn1 is able to distinguish between ubiquitin and ubiquitin-like proteins.
To achieve a deeper understanding of the differences in the binding properties of Rpn1 and Rpn2 and how they may differentiate among various ubiquitin-like signals, we quantified all pairwise binding interactions by SPR. Example titrations of single binding experiments are shown in Fig. 2, A-D. Association (k on ) and dissociation (k off ) rate constants were extracted by fitting binding curves over a range of concentrations using the Langmuir model for a one-to-one binding stoichiometry. Association of Ubp6 with Rpn1 was an exception to single-site binding, but we were able to model the data in Fig. 2C using the HLPR model (70). In this model, a single protein (A) interacts with two independent binding sites/states on its partner/receptor (B); each interaction is characterized by its own set of k on and k off parameters (the general scheme is shown in Equation 3 above). One way to rationalize the heterogeneous binding model with what we know of Ubp6 and the proteasome is that the single protein "A" is a bivalent molecule with two domains (depicted as A 1 -A 2 in supplemental Scheme Si), each adapted for one of two binding sites (B 1 and B 2 on the receptor "B"; see Equation 3 above and supplemental Scheme Si). Initial binding of Ubp6 either via A 1 or via A 2 contributes to the SPR results, and hence two binding parameters were observed for the fulllength Ubp6 protein attributed to two different binding modes of Ubp6-Rpn1 association (summarized in Table 1; see further analysis below and a more detailed description in supplemental Schemes Si and Sii).
Rate constants k on and k off and the dissociation constant K D for each binding pair were calculated from at least three independent SPR experiments (Fig. 2, A-D), averaged, and summarized in Table 1. Two pairwise interactions were found to be particularly tight, displaying K D values in the nanomolar range, Ubp6-Rpn1 and Rpn13-Rpn2 (62 and 12 nM; Table 1). Rad23 and Dsk2 showed similar kinetic behavior vis-à-vis Rpn1, yielding affinities in the micromolar range (3.6 M versus 12.1 M, respectively). By contrast, the affinity of Ddi1 for Rpn1 was much weaker and therefore not quantifiable in the experimental concentration range. Likewise, we were unable to quantify associations of ubiquitin or Rpn10 with either Rpn1 or with Rpn2. Dissociation constants obtained by SPR (Table 1) point to a hierarchy of interactions that is completely compatible with results obtained from reciprocal binding experiments ( Fig.  1). Interestingly, the proteasome interacting proteins that we assayed had similar "on" rates, yet they differed in their dissociation rates. Thus, release of Ubp6 from Rpn1 and release of Rpn13 from Rpn2 were 3 orders of magnitude slower than that of Rad23 or Dsk2 from their receptor Rpn1. These binding properties explain how transiently associated Rad23 and Dsk2 are members of the substrate delivery family, whereas proteins such as Ubp6 and Rpn13 with slow dissociation rates behave more like "sometimes proteasome subunits" (2). The fast exchange rates of shuttle proteins such as Rad23 or Dsk2 are a necessary feature if they are to unload ubiquitinated cargo and dissociate for the next round.
Contribution of Ubiquitin-like Domains to Interaction of UDPs with Rpn1-Four UDP proteins (Rad23, Dsk2, Ubp6, and Ddi1) have very different affinities for the UBL receptor Rpn1 ( Table 1). Recognition of UDPs is generally thought to occur via their structurally similar UBL domains (Fig. 1). Therefore, we compared the contribution of each UBL domain to the overall interaction with Rpn1. Association of purified Rpn1 with recombinant UBL domains from each of the four UDPs was assayed by SPR. Dissociation constants were derived from equilibrium analysis (Fig. 2E) because the on/off rates were too fast FIGURE 1. Rpn1 and Rpn2 dock ubiquitin processing factors. A, a number of UDPs serve as bifunctional shuttles by interacting simultaneously with polyubiquitinated cargo and with receptors at the proteasome. PolyUb conjugates can also dock at the proteasome directly. Four proteasome subunits (Rpn1, Rpn2, Rpn10, and Rpn13) have been implicated in recruiting ubiquitin chains or UBLs. B, most UDPs contain an N-terminal UBL followed by a unique functional domain. Some bind polyUb through a UBA. Rad23 contains two UBA domains, UBA1 and UBA2, which share 27% sequence identity, and an XPC (xeroderma pigmentosum group C complementing protein) domain between the two (92). Dsk2 and Ddi1 contain only one UBA domain, with Ddi1 also having an internal retroviral aspartyl protease domain (93). Ubp6 associates with ubiquitin via its UBP. Protein domains are drawn roughly to scale. C, immobilized ubiquitin processing factors (Rpn13, Rpn10, Ubp6, Rad23, Dsk2, and Ddi1) as well as ubiquitin were each tested for binding Rpn1 or Rpn2 proteins. Rpn is regulatory particle non-ATPase. IB, immunoblot.
for accurate kinetic quantification. The resulting K D values are summarized in Table 2.
Three UBLs (Rad23, Dsk2, and Ubp6) bound Rpn1 with comparable micromolar range affinities; Ddi1 UBL associated very weakly (Ͼ100 M), whereas neither ubiquitin nor Lys-48linked polyUb chains bound at detectable levels (Fig. 2E). Isolated UBLs bound Rpn1 with the same affinity as their corresponding full-length proteins (Table 2) suggesting that association is carried out through the UBL domain. A notable exception was Ubp6 UBL , which bound Rpn1 weaker than Ubp6 (K D ϭ 2 M versus 60 nM; Table 2), yet with comparable affinity to those of Rad23 UBL and Dsk2 UBL (Fig. 2E). Moreover, Ubp6 UBL binding to Rpn1 (Fig. 3A) was fit to a single bindingsite model (Langmuir kinetics). The dissociation constant K D for Ubp6 UBL corresponds within experimental error to the K D of the weaker binding site of Ubp6 ( Table 2). As the net dissociation rate of this UBL domain is faster than that observed for full-length Ubp6, it seems likely that additional properties of the full-length protein slow down the apparent dissociation. In this respect, Ubp6 differs from other members of the UDP family that do not exhibit complicated association kinetics and consequently do not bind as tightly to Rpn1 receptor.
Coordinated Binding of Ubp6 Domains to the Rpn1 Receptor-To dissect parameters that contribute to Ubp6-Rpn1 binding, we generated truncated constructs of Ubp6. One fragment spanned just the UBL domain, and another encompassed the enzymatic UBP domain (Ubp6 UBL versus Ubp6 ⌬UBL ; Fig. 3, A and B). Whereas Ubp6 UBL exhibited fast association and dissociation rates similar to those for other Rpn1-associating UDPs ( Table 3 versus Table 1), the C-terminal portion of Ubp6 associated slower and dissociated much slower than other UDPs studied ( Fig. 3B; Table 3). Independent binding of two proteins (A 1 and A 2 ) to two binding sites (on a platform B) can be Sensograms in C (Rpn1-Ubp6 response) were fit to an HLPR, assuming two-binding sites. Association (k on ) and dissociation (k off ) rate constants derived from a global fit of primary response data over the entire concentration range for each pair to the corresponding model are summarized in Table 1. E, binding isotherms for Ubp6 UBL (circles), Rad23 UBL (triangles), Dsk2 UBL (diamonds), Ddi1 UBL (ϫ), and ubiquitin (squares) derived from equilibrium SPR measurements. Normalized equilibrium response is plotted as a function of soluble protein concentration; the fitting curves correspond to dissociation constants shown in Table 2. Summary of Rpn2-Rpn13, Rpn1-Ubp6, Rpn1-Rad23, and Rpn1-Dsk2 kinetic parameters determined by SPR analysis. All response curves were fitted to a single site binding model (Langmuir kinetics) with the exception of Ubp6 for which the full-length protein required a two-site binding model (2ϫ Ubp6 ϩ Rpn1(site 1) ϩ Rpn1(site 2) 7 Ubp6-Rpn1(site 1) ϩ Ubp6-Rpn1(site 2)). Values for dissociation constant K D were calculated as the ratio of k off over k on .    Table 3. RU, response units. C, immobilized Ubp6 and Rad23 lacking their UBL domains (Ubp6 ⌬UBL and Rad23 ⌬UBL ) were tested for binding to 26 S proteasome-purified free of all UDPs or other transient proteasome-interacting proteins. Association of full-length Ubp6 and Rad23 is included as positive control. Eluate from each column was tested for proteasome activity (20 S CP) as well as immunoblotted (IB) for presence of 19 S subunits from the Base and Lid subcomplexes. D, truncated Ubp6-expressing strains were molecular weight-fractionated, and proteins copurifying with peak of proteasome activity were identified by immunoblotting. E, potency of Ubp6 to remove ubiquitin from Ub-GFP was monitored over 60 min, and % reaction progression was plotted over time (diamonds). The reaction was repeated in presence of equimolar Rpn1 (squares) or purified base-CP (lidless) proteasome lacking both endogenous Ubp6 and Rpn11 (circles). Neither Rpn1 (star) nor base-CP alone (ϫ) exhibit deubiquitination activity in the absence of added Ubp6 (dotted lines). F, experiment similar to E but for truncated Ubp6 ⌬UBL . Ubp6 activation is not dependent on UBL.
and Ubp6 ⌬UBL . Coordinated binding of two Ubp6 domains could explain the higher affinity of the full-length protein; Ubp6 would be targeted to Rpn1 via its UBL domain and locked in place by interactions involving the C-terminal portion of the protein.
To summarize, of the various UDPs in this study Ubp6 had the highest affinity for Rpn1. The overall affinity of Ubp6 for Rpn1 is comparable with that of a bona fide proteasome subunit such as Rpn13 (Fig. 2), thus placing Ubp6 in the category of occasional proteasome subunits rather than fast-exchange shuttle proteins.
Incorporation of Ubp6 into Proteasomes Does Not Require the UBL Domain-It has been suggested that distribution of Ubp6 between proteasome-bound and proteasome-unbound pools reflects an equilibrium between association and dissociation (44). To test whether free Ubp6 can incorporate into preassembled proteasomes, a purified proteasome sample free of UDP proteins (lacking Ubp6, Rad23, Dsk2, and Ddi1) was assayed for recruiting Ubp6, with or without its UBL domain. Incorporation of Ubp6 or Ubp6 ⌬UBL into the 26 S proteasome was equally efficient. By contrast, binding to Rad23 required its UBL domain (Fig. 3C). Moreover, even in cells, proteasomes isolated from a strain expressing a construct of Ubp6 lacking the UBL domain (ubp6 ⌬UBL ) contained Ubp6 ⌬UBL as a subunit (Fig. 3D). Interestingly, the UBL domain alone was not retained during the purification process (Fig. 3D, right column) supporting the in vitro observation that the UBL alone had fast on/off kinetics. Thus, we conclude that stickiness of the C-terminal UBP-containing segment contributes toward Ubp6 as a proteasome subunit.
Next, we assessed whether association with its receptor Rpn1 was the underlying cause of enzymatic activation of Ubp6 upon proteasome incorporation (66). Premixing Ubp6 with a proteasome sample devoid of all known DUBs (Lidless Base-CP lacking Rpn11 purified from a ⌬ubp6 strain) enhanced basal Ubp6 activity to the same extent as premixing with isolated Rpn1 (Fig.  3E). In this assay, we made use of a ubiquitin-fusion substrate used previously as a model for deubiquitination (4,71). Even a Ubp6 version lacking its N-terminal UBL domain was activated by either proteasome or Rpn1 to essentially the same extent as was full-length Ubp6 (Fig. 3F). These findings indicate that activation of Ubp6 is a direct consequence of association with Rpn1 yet does not require its UBL.
Domain Mapping of Rpn1 and Rpn2-Having quantified association of Rpn1 or Rpn2 with a number of ubiquitin-processing factors, we then asked how they discern between multiple binding partners. Are there dedicated binding sites, or do UDPs compete for a single receptor? Both Rpn1 and Rpn2 proteins can be similarly dissected into three major segments. The central portions of each are composed of 11 ␣-helical rich repeats (also known as PC repeats) that fold into highly curved solenoids (26). Truncated versions of Rpn1 or Rpn2, encompassing only the solenoid portion, fold into stable structures (27) and are sufficient to maintain interactions with each other (Fig. 4A) or with the 20 S CP (Fig. 4B), as has been published for the full-length proteins (28). This central solenoid domain appears to be sufficient for association with Rpn10 (Fig. 4C). No other proteasome auxiliary factor or transient subunit tested in this study was found to associate stably with this region. Association of Rpn10 with Rpn1 is stabilized in presence of Rpn2 (Fig. 4D). Coupling of Rpn1 and Rpn2 stabilizes their PC folds (28) and thus may explain the stronger association of Rpn10 with a joint Rpn1-Rpn2 complex (Fig. 4C). Indeed, all three have been found in a stable complex together with the 20 S CP and a few additional factors (7,36).
The flexible terminal extensions of Rpn1 were found to be required for UDPs to bind (Fig. 4E). Specifically, binding of either Rad23 or Dsk2 required the presence of the N-terminal region, whereas deleting this region did not abolish association of Ubp6 (Fig. 4E) suggesting that Rpn1 harbors multiple UBLbinding sites. Similarly, we mapped the interaction of Rpn13 with Rpn2 to the C-terminal segment of Rpn2 (Fig. 4F). This result is in full agreement with previously published data for Rpn2-Rpn13 association (7,35).
Simultaneous Binding of Rad23 and Dsk2 to Rpn1-Having observed that both Rad23 and Dsk2 require the N-terminal region of Rpn1 for binding (Fig. 4E), we wished to explore whether these two shuttle proteins compete for the same site. We immobilized Rad23 UBL and monitored its ability to retain pre-bound Rpn1-Dsk2 UBL . The results provided evidence for a ternary complex of Rad23 UBL -Rpn1-Dsk2 UBL (Fig. 5A), indicative of separate binding sites on Rpn1. The reverse experiment in which immobilized Dsk2 UBL retained pre-bound Rpn1-Rad23 UBL also generated the same conclusion (Fig. 5B).
The two UBL-binding sites may differ in their binding specificities as each may show preference for one UBL over the other, thereby clarifying why a 50-fold molar excess of Dsk2 was utilized to out-compete Rad23 bound to Rpn1 (40). In search of a minimally required binding region for each UBL, we tested N-terminal segments of Rpn1 for pairwise interactions. Truncating the first 440 residues abolished association with both Rad23 and Dsk2 (Fig. 4E); however, the detached segment alone was competent to bind Dsk2 but not Rad23, confirming that different sets of residues are required to anchor either of these two UBLs (Fig. 5C, top). Using the meta Protein DisOrder prediction System (PrDOS (72)), we dissected the N-terminal region of Rpn1 into three segments separated by natively disordered linkers (Fig. 5C). The first segment (Rpn1(1-192)) did not display affinity for any UDP in our assay, whereas a Dsk2binding site was positively identified in the second segment (Rpn1(194 -360)). None of these N-terminal segments bound Rad23 (Fig. 5C),and neither did the immediately downstream PC repeat region (residues 440 -863, see Fig. 4E). Therefore, we hypothesized that contributions from residues in flanking  APRIL 27, 2012 • VOLUME 287 • NUMBER 18

Rpn1 Differentiates between Ubiquitin and Ubiquitin-like
regions may be required in tandem. An earlier study identified Rad23 as associating with a region of Rpn1 spanning the N terminus as well as the first two of 11 PC repeats (32). Therefore, we generated an additional fragment starting at the second disordered interceding linker and extending into the flanking PC region (Rpn1(372-518)) that spans the first two "covert" PC repeats predicted to differ in their properties from the subsequent nine PC repeats (26). This segment (entirely distinct from the Dsk2-binding region; Fig. 5C) was sufficient to associate with Rad23. The tertiary fold of Rpn1 may provide the necessary contact points for Rad23 binding. These contact points may be contributed by relatively few amino acid residues from different structural domains positioned far apart in the primary sequence. Identification of UBL receptors at the previously poorly characterized N terminus of Rpn1 illuminates how Rpn1 may serve as a scaffold for ubiquitin shuttles at the proteasome. Moreover, the existence of two distinct binding sites within this region, each dedicated for a preferred ligand (Fig. 5C), suggests that Rpn1 may coordinate binding of multiple ubiquitin-processing factors simultaneously (Fig. 5, A and B), each with its own affinity for UBLs targeting the proteasome (Tables 1 and 2). Nonoverlapping binding sites within the Base, each with different specificities, raise the possibility that multiple proteasome-interacting proteins co-reside on a single proteasome (supplemental Fig. S1).

DISCUSSION
There are multiple levels of redundancy in targeting ubiquitinated substrates to the proteasome (Fig. 6). Substrate shuttles can mediate recruitment by simultaneously binding both conjugates and the proteasome. Although some of these ubiquitinbinding proteins show selectivity for certain ubiquitin-chain linkages, they appear to be generally redundant. Moreover, the few intrinsic ubiquitin receptors that have been identified at the proteasome also seem to be dispensable, suggesting functional overlap. Moreover, proteasomes lacking most, or all, known associated ubiquitin-binding subunits are functional, at least on some substrates (41,43). Even categorizing ubiquitin-binding proteins neatly into shuttles versus receptors may be an oversimplification, as some ubiquitin receptors at the proteasome apparently also bind ubiquitin-like domains (supplemental Table S1). In this study, we identified Rpn1 as a definitive UBL receptor in the proteasome, able to distinguish UBL from ubiquitin/polyUb and with the capacity to bind multiple UDPs simultaneously. This feature of Rpn1 is distinct from the binding preferences of Rpn2, a paralog also situated within the Base of the 19 S RP, yet together these two large subunits coordinate the activity and placement of multiple ubiquitin-processing FIGURE 5. Dsk2 and Rad23 bind to two different sites at Rpn1. A and B, immobilized Rad23 UBL was incubated with either Dsk2 UBL or a premixed Rpn1 ϩ Dsk2 UBL mixture and washed, and bound proteins were eluted and immunoblotted for the presence of Dsk2. The reciprocal experiment was repeated with Dsk2 UBL immobilized and incubated with either Rad23 UBL or a Rpn1 ϩ Rad23 UBL mixture. The formation of these Rad23-Rpn1-Dsk2 ternary complexes demonstrates that Rad23 and Dsk2 can bind simultaneously to Rpn1, via at least two discrete binding sites. C, purified recombinant N-terminal fragments of Rpn1 were washed over immobilized binding partners, washed, and elution immunoblotted for presence of Rpn1. A Dsk2-binding site is located between residues 196 and 369 of Rpn1, whereas Rad23 binds tighter to an Rpn1 fragment spanning residues 372-518. FIGURE 6. Schematic description of substrate relay. Polyubiquitin-binding proteins such as Dsk2 or Rad23 are transiently associated with proteasomes and thus may serve as shuttles or delivery proteins for polyubiquitin conjugates. These proteins contain an N-terminal UBL (ubiquitin-like) domain through which they dock at the proteasome. The primary UBL receptors for Rad23 and Dsk2 was found to be proteasome subunits Rpn1 (Tables 1-3), although these two UDPs (UBL-domain containing proteins) were also capable of associating directly with two neighboring subunits, Rpn10 and Rpn13 (supplemental Table  ST1). Ddi1 apparently also associates with Rpn1, although in this study we were unable to quantify the intensity of their mutual association. Polyubiquitin chains could also bind to Rpn10 and Rpn13 directly (supplemental Table ST1), thereby bypassing the need for shuttles. The relative affinities of these receptors for the different (poly)ubiquitin or UBL signals influence efficiencies of substrate targeting and coordinate their entry into the proteasome. Another UDP, the deubiquitinating enzyme Ubp6, bound only to Rpn1 yet much tighter than transiently associated shuttle proteins, thus behaving more akin to proteasome subunits Rpn10 and Rpn13. The proximal localization of Ubp6 to these polyubiquitin-chain-receptors may also coordinate trimming, processing, or shaving of conjugated chains with substrate preparation and the proteolytic mechanism (see additional details in supplemental Fig. S1).

Rpn1 Differentiates between Ubiquitin and Ubiquitin-like
factors at the proteasome. Most of the data for this study was collected on isolated Rpn1 and Rpn2 subunits; nevertheless, by quantifying the interactions of Rpn1 and Rpn2 with their binding partners, we provide a possible outline of early steps in substrate processing by the proteasome (Fig. 6 and supplemental Fig. S1).
One proteasome-associated UDP that is not a substrate shuttle but rather harbors a distinct enzymatic activity is Ubp6. Association of Ubp6 with Rpn1 activates the deubiquitinase activity of Ubp6 suggesting that proteasome-incorporated Ubp6 is functionally distinct from its free form. One explanation may be that rearrangement of the UBL domain upon proteasome binding frees up access to the active site cleft in the catalytic core (the UBP domain) (32,73). Such an auto-inhibitory mechanism has been concluded for another DUB, USP4, also with an embedded UBL domain that can compete with substrate by acting as a ubiquitin mimic at the catalytic site (74,75). A somewhat similar model implied conformational changes related to the UBL domain of other UDPs upon proteasome incorporation (76,77). However, as this study observed Ubp6 lacking UBL incorporated into proteasomes, such confrontational changes are probably more subtle and do not necessitate the UBL domain for activation. The distinction between a transient shuttle, whose role is to upload cargo, depart, and make way for subsequent rounds, and a subunit that needs to be properly incorporated is evident from the different on/off kinetics (Table 2). In contrast to the UDPs that serve as delivery proteins by hovering in and out of proteasomes, subunits seem to require more than their UBL to keep them anchored in place.
The most striking feature of Rpn1 and Rpn2 is their toroidal, almost ring-like, structure, at least in the absence of other associated subunits (27). These solenoid portions are responsible for their mutual interaction as well as for association with the ubiquitin receptor, Rpn10 (Fig. 4), and possibly with regulatory particle triple nucleotidase ATPases (30). In absence of additional binding partners, the flanking N-and C-terminal extensions in each protein are likely to be unstructured or flexible (27). This study goes a considerable way to show that these extensions carry the primary proteasome docking sites for substrate delivery proteins and auxiliary ubiquitin processing subunits (Figs. 4 and 5). More specifically, it is the N-terminal region of Rpn1 that anchors UDPs at the proteasome either near to, or partially overlapping with, the central solenoid PC repeat region (Fig. 6), whereas the C-terminal regions of Rpn2 and Rpn1 are required for interaction with the ubiquitin receptor Rpn13 and the deubiquitinating enzyme Ubp6, respectively. Independently, these flexible extensions may also contribute to stability of 19 S RP-20 S CP association (37) or to proper nuclear localization of proteasome (11).
All told, Rpn2 displays specific binding to at least five partners as follows: 20 S CP, Rpn13, Rpn1, Rpt4, and Rpt6 (see this work and Refs. 31,33,78,79). Rpn1 associates with an assortment of ubiquitin processing factors, Rad23, Dsk2, Rpn10, Ddi1, and Ubp6, and neighboring proteasome subunits, Rpn2, Rpt2 and Rpt5 (this work and Refs. 7, 30 -32, 36, 54, 60, 78 -80). In contrast, the adjacent ubiquitin receptors Rpn10 and Rpn13 also exhibit propensity for UBL domains (41, 46, 80, 81), although they do seem adapted for polyUb (supplemental Table S1). Thus, Rpn1 emerges as the primary UBL receptor at the proteasome with a marked preference for UBLs over ubiquitin. Properties of the transported cargo, such as hydrophobic/unstructured patches, or the nature of the polyubiquitin tag (linkage/length) may significantly alter the relationships identified in this study, cargo transported by these shuttles. Nevertheless, the proximity of their potential biding sites at the Base may allow for efficient funneling of polyubiquitinated substrates from shuttle proteins that dock at Rpn1 to downstream polyUb receptors, and ultimately toward the 20 S proteolytic core.
In its capacity as a UBL-binding platform, Rpn1 not only binds multiple UBLs, but does so simultaneously by utilizing specific sites for each UBL. As Rpn1 has no sequence homology to any known ubiquitin-binding domain or any protein known to bind UBLs (such as UIM, UBA, CUE, etc. (82,83)), we believe that we have uncovered a totally new class of receptors critical for proteasome function. Many UDPs are substrate-delivery proteins (84); therefore, finding a pluripotent docking site for UDPs within the 19 S RP is a significant step toward dissecting a proteasome mechanism and the multistep trajectory of substrates within. Placement of Rpn1 at the Base, in the vicinity of proteasomal ATPases, supports a model whereby conjugated substrates may be unloaded from transitory shuttles to the ATPase unit for unfolding, whereas the polyUb tag is anchored by one of the nearby intrinsic ubiquitin receptors.
In prokaryotes and archaea, simplified proteasome analogs consist of a protease core complex and an ATPase regulator; few if any auxiliary factors are known to be required (85,86). In these domains of life, substrates are not ubiquitinated, and the ATPases are sufficient for trapping substrates and transferring the unfolded polypeptide to the protease domain (87,88). However, most substrates in eukaryotes are first conjugated to a polyUb tag. Therefore, an additional layer of factors for substrate recognition and preparation may be called for to mediate recognition, binding, anchoring, and deubiquitination, coupled with unfolding by the ATPases (65,68). Although most observations in this study (with the exception of Ubp6) were for potential binding partners with isolated Rpn1 and Rpn2, we propose that these factors may serve to recruit ubiquitin conjugates to the proteasome. Once at the proteasome, the length of the polyUb modification could act as a threshold, with poorly ubiquitinated substrates released on account of fast dissociation rates, whereas chains of proper length and linkage are retained on the proteasome (89). As a chain-editing enzyme, Ubp6 could help define substrate residency time by trimming the chain, ubiquitin by ubiquitin (1,50,64,67).
Cumulative studies have unearthed a sequence of substrate preparation steps both upstream to and within the proteasome. These include recognition, ubiquitination, delivery, recruitment by 26 S proteasome, binding, anchoring, unfolding, deubiquitination, and translocation into the 20 S CP proteolytic chamber. Curiously, many of these diverse activities have been mapped in one way or another to the Base of the 19 S RP. Nevertheless, the Lid of the 19 S RP is also required for treatment of polyUb conjugates (3), at least partially by providing (redundant) DUB activity in the form of Rpn11 (4, 90, 91). Although the ubiquitinated cargo may influence the specific properties of

Rpn1 Differentiates between Ubiquitin and Ubiquitin-like
receptors and binding proteins measured in this study, Rpn1 and Rpn2 emerge as the primary coordinators of the ubiquitin processing machinery at the Base. As such, Rpn1 and Rpn2 appear to be the inherently eukaryotic subunits that distinguish the Base of the 26 S proteasome holoenzyme from simpler activators of prokaryotic ATP-dependent (yet ubiquitin-independent) proteases.