|Home | About | Journals | Submit | Contact Us | Français|
Proteasomal receptors that recognize ubiquitin chains attached to substrates are key mediators of selective protein degradation in eukaryotes. Here we report the identification of a new ubiquitin receptor, Rpn13/ARM1, a known component of the proteasome. Rpn13 binds ubiquitin via a conserved N-terminal region termed the Pru domain (Pleckstrin-like receptor for ubiquitin), which binds K48-linked diubiquitin with an affinity of ~90 nM. Like proteasomal ubiquitin receptor Rpn10/S5a, Rpn13 also binds ubiquitin-like domains of the UBL/UBA family of ubiquitin receptors. A synthetic phenotype results in yeast when specific mutations of the ubiquitin binding sites of Rpn10 and Rpn13 are combined, indicating functional linkage between these ubiquitin receptors. Since Rpn13 is also the proteasomal receptor for Uch37, a deubiquitinating enzyme, our findings suggest a coupling of chain recognition and disassembly at the proteasome.
In eukaryotes, selective protein degradation is performed primarily by the ubiquitin-proteasome pathway. The 26S proteasome is a huge macromolecular machine that contains a proteolytically active 20S core particle (CP) capped at one or both ends by a 19S regulatory particle (RP)1. The RP recognizes ubiquitinated substrates, deconjugates ubiquitin chains, and unfolds substrates prior to their translocation into the CP. Proteasome subunit Rpn10/S5a was shown to bind ubiquitin chains via ubiquitin-interaction motifs (UIMs)2. Receptors were subsequently identified that are not integral proteasome subunits, but deliver ubiquitinated targets to the proteasome (for reviews, see 5 and 6). Canonical members of this UBL/UBA family of receptors are Rad23 (hHR23a/b in humans), Dsk2 (hPLIC-1/2 in humans) and Ddi13,4,7-9. UBA domains bind ubiquitin10-12 and UBL domains interact reversibly with the proteasome, principally via Rpn1, but potentially also via Rpn1013-15.
Another interesting component of the proteasome is Rpn13/ADRM1/ARM116-21, which docks at the RP via an N-terminal region that binds Rpn218,21-23. Its C-terminal region binds deubiquitinating enzyme Uch37/UCHL5 and enhances its isopeptidase activity18,20,21. Uch37 may function as an editing isopeptidase that rescues poorly ubiquitinated substrates from being degraded24.
Using a yeast two-hybrid screen, with a bait of ubiquitin lacking the last two glycines to prevent its conjugation25, we identified the N-terminal segment of human Rpn13 (hRpn13) as a ubiquitin-binding partner. The interaction was confirmed using murine Rpn13 (mRpn13) as bait against monoubiquitin and Rpn2 as prey (Figure 1A). Rpn13 from S. cerevisiae (scRpn13) aligns with the ubiquitin-binding N-terminal region of hRpn13 (Figure 1B). Comprehensive sequence analysis using profiles and Hidden Markov Models failed to reveal similarity to known ubiquitin- or proteasome-binding motifs (Figure 1C and data not shown). Deletion mutants encompassing residues 1-150 were tested for tetraubiquitin binding, thus mapping the minimal binding domain to residues 1-130 (Figure 1D). Although smaller fragments of mRpn13 also showed detectable binding to ubiquitin, they were unstable and expressed poorly as GST-fusions.
The significance of the ubiquitin-Rpn13 interaction would be supported if it were conserved from yeast to mammals, particularly as budding yeast Rpn13 is truncated and the conserved N-terminal region (Figure 1C) only 25% identical to mammalian forms (Figure 1B). The existence of an unidentified ubiquitin receptor in yeast had been evident to us from the viability of a rpn10-uim rad23Δ dsk2Δ ddi1Δ mutant (data not shown). rpn13Δ mutants, which are viable but show defects in protein degradation16,26, were used to test whether Rpn13 binds ubiquitin chains in the context of intact, purified proteasomes.
Ubiquitin chain binding by purified proteasomes can be assayed by native gel electrophoresis3,14. Proteasomes are visualized in this system by an activity stain, using a fluorogenic peptide substrate. For wild-type, the predominant proteasome species contains one RP on either end of the CP cylinder (RP2CP). Ubiquitin chains, produced by the E2 enzyme Cdc34, bind to the proteasomes and confer reduced mobility (Figure 2A). This shift is not dependent on UBL-UBA proteins, since the proteasomes were prepared from rad23Δ dsk2Δ ddi1Δ mutants. A block substitution within the UIM in Rpn10 results in attenuation of the shift, reflecting Rpn10's known ubiquitin receptor function (Figure 2A; 3,4). However, the existence of marked residual electrophoretic retardation by added chains (lane 4) indicates the presence of at least one additional ubiquitin receptor in purified proteasomes.
Addition of conjugates to proteasomes lacking Rpn13 resulted in an electrophoretic shift comparable to that of rpn10-uim samples (Figure 2A). Thus, the yeast ortholog of mRpn13 is active in ubiquitin chain binding and can bind ubiquitin in the context of intact proteasomes. Indeed, its capacity for chain binding in this assay compares well with that of Rpn10. Remarkably, we observed an ostensibly complete abrogation of chain-dependent electrophoretic retardation when rpn10-uim rpn13Δ proteasomes were used (Figure 2A), suggesting that Rpn10 and Rpn13 are the two major ubiquitin receptors in the yeast proteasome. However, by varying the conditions of this assay, we could, as shown below, detect residual ubiquitin chain binding with rpn10-uim rpn13Δ proteasomes, consistent with the existence of a still unidentified proteasomal ubiquitin receptor. The greater abundance of the RP1CP and CP bands in rpn13Δ samples suggests that Rpn13 contributes to the stability of the RP-CP interaction in vitro26.
To determine whether rpn13Δ proteasomes were properly assembled, they were analyzed by SDS-PAGE (Figure 2B). Apart from the absence of Rpn13 itself, the mutant proteasomes appeared to be wild-type in composition. When recombinant Rpn13 was reconstituted onto mutant proteasomes after purification, their chain-binding defect was corrected (Figure 2C). Thus, the chain-binding assay appears to report on a specific Rpn13-ubiquitin chain interaction, and not a gross structural defect of rpn13Δ proteasomes. GST pull-down assays also indicated direct interaction of scRpn13 with ubiquitin (data not shown). In summary, these data indicate that Rpn13 is a novel proteasomal ubiqutin receptor.
To create mutants of scRpn13 deficient in ubiquitin binding and to subsequently study the functional significance of the interaction in vivo, we initially used NMR to solve the structure of full-length scRpn13. These studies revealed that Thr6-Leu101 forms two contiguous, antiparallel β-sheets with a configuration similar to the PH structural domain (Figure 3A). In particular, a β-sheet comprised of four antiparallel β-strands formed by I8-R11, E32-P37, W46-W50, and I64-L66 packs against a 3-stranded sheet formed by M74-V76, I86-V90, and R96-W100. Juxtaposed to the 3-stranded sheet are two β-strands formed by C15-N18 and L23-P26. The configuration of the β-strands centres around interactions between conserved aromatics within the protein core, including F10, F48, W50, W75, F87, F91, F98, F99, and W100. These findings are consistent with the crystal structure27 of mRpn13. We thus named this domain Pleckstrin-like receptor for ubiquitin (Pru). In a canonical PH domain, K117-N130 of scRpn13 would be α-helical, consistent with secondary structure predictions for K119-G127. Residues S106-G127 are absent, however, from all acquired spectra, suggesting that this region undergoes conformational exchange and does not form a rigid helix. In the accompanying manuscript27, we find that the cognate residues in mRpn13 and hRpn13 do form helices. K117-N130 of scRpn13 shares 35.7% sequence identity with mRpn13, but the presence of a glycine at position 127 likely destablizes the helix, as might substitution of the sequence QDE at the beginning of the helix with the more basic sequence KDK in scRpn13. Also, a salt bridge27 between E119 and R122 of mammalian Rpn13 is lost, as R122 is substituted with N123 in scRpn13.
To determine how scRpn13 binds ubiquitin, we performed an NMR titration series (Supplementary Figure 1), which implicated E41, E42, G44, F45, L66, E72, F91, S93 and R96 as being at the ubiquitin contact surface (Figure 3B). Interestingly, these residues are in the S2-S3, S4-S5, and S6-S7 loops (Figure 3C). The S4-S5 loop is strongly conserved in higher eukaryotes, as is F91, which is in the S6-S7 loop (Figure 1C). scRpn13 binding to monoubiquitin is in “fast exchange” by NMR, which is ideal for determining binding affinity by this method, and the affinity of scRpn13 for monoubiquitin was determined to be 65 μM (Figure 5F).
We used NMR titration experiments to determine the stoichiometry of hRpn13 for monoubiquitin, K48-linked diubiquitin, and tetraubiquitin (see Supplementary Information). Monoubiquitin and diubiquitin bound to Rpn13 with 1:1 stoichiometry, whereas two Rpn13 molecules bound one tetraubiquitin (Figure 4A and 4B). Therefore, although three potential diubiquitin elements exist within tetraubiquitin, no more than two Rpn13 molecules can be accommodated simultaneously. The exclusion of a third Rpn13 molecule is consistent with model structures of Rpn13:tetraubiquitin, in which steric clashes arise when three hRpn13 molecules bind neighbouring K48-linked ubiquitin subunits (Supplementary Figure 2). That Rpn13 binds diubiquitin elements of K48-linked chains is further validated in the accompanying manuscript27.
In contrast to scRpn13, resonances broaden and shift as hRpn13 Pru binds monoubiquitin (Supplementary Figure 3). This behavior is associated with stronger Kd values, but prohibits their accurate calculation by the method used to determine the scRpn13:ubiquitin affinity. Fluorescence spectrophotometry was used to determine hRpn13's affinity for monoubiquitin and diubiquitin, since hRpn13 (1-150) contains two tryptophan residues and ubiquitin none. hRpn13 showed a surprisingly strong affinity, with a Kd for monoubiquitin of ~300 nM (Figure 4C and 4D) and for diubiquitin of ~90 nM (Figure 4C and 4D). The value for diubiquitin binding is ~15-fold lower than that of hHR23a for tetraubiquitin28. The stronger affinity of hRpn13 for ubiquitin, as compared to scRpn13, reflects amino acid substitutions at the contact surface. For example, in the accompanying paper, residues F76 and D78 of hRpn13 were implicated in hRpn13 binding to monoubiquitin. In scRpn13, however, these residues are substituted with isoleucine and glycine, respectively.
We next analyzed whether Rpn13 exhibits specificity for ubiquitin or broadly recognizes ubiquitin family members. Using GST pull-down assays, we confirmed that ubiquitin binds to full-length mRpn13 and that this interaction requires ubiquitin's hydrophobic patch, consisting of L8, I44, and V70 (Figure 4E). mRpn13 bound more potently to linear tetraubiquitin expressed as a GST fusion (GST 4×Ub) or to purified K48-linked chains than to monoubiquitin (Figure 4E and data not shown). Under the same experimental conditions, no binding was observed between the mRpn13 Pru domain and SUMO, Nedd8, ISG15 or FAT10 (Figure 4F). In contrast, mRpn13 appeared to bind to the UBL domains of multiple UBL/UBA proteins (Supplementary Figures 4 and 5). We verified that hRpn13 Pru binds directly to the hHR23a and hPLIC2 UBL domains by NMR (Figure 4G), and determined a Kd value of 36 μM for the hRpn13 Pru:hHR23a UBL domain complex (Figure 4D and Supplementary Figure 5). Overlapping residues in hRpn13 were affected by the addition of these UBL domains (Figure 4G) or ubiquitin27, suggesting that these interactions may be mutually exclusive.
Experiments described above implicated residues in Rpn13's S2-S3, S4-S5, and S6-S7 loops in binding ubiquitin (Figure 3). After introducing non-conservative substitutions for these residues individually or in combination, the resulting proteins were expressed in E. coli, purified, and characterized. We sought mutants that were defective in ubiquitin chain binding while being properly folded and proficient in proteasome binding. Separation of these functions is critical, as exemplified by previous studies of proteasomal ubiquitin receptor Rpn10. The rpn10Δ phenotype does not accurately reflect its function in ubiquitin recognition, because Rpn10 plays additional roles in the proteasome4,29. The proteasome is destabilized in the absence of Rpn1029, as is also observed in rpn13Δ mutants, at least under certain in vitro conditions (Figure 2; 26). For Rpn13, unlike Rpn10, proteasome and ubiquitin binding are conferred by the same structural domain, and thus can be effectively separated only by precisely targeted mutations. Moreover, ubiquitin contact residues in Rpn13 are distributed over three distinct loops, and thus differ from those of Rpn10 by being non-contiguous and thus not subject to simple block mutagenesis.
We assayed Rpn13-proteasome binding by adding GST-Rpn13 to purified proteasomes. Due to the GST moiety, the fusion protein results in strong electrophoretic retardation in native gels; this effect is seen with rpn13Δ but not wild-type proteasomes (Figure 5A). Thus, Rpn13 assembled into the proteasome was not exchangeable with added GST-Rpn13, indicating that scRpn13 is a true proteasome subunit. Two putative ubiquitin contact site mutants of Rpn13 were demonstrated to be proficient in proteasome assembly, as shown in Figure 5B. Several other mutants failed to pass this and other preliminary assays, due typically to global folding defects (data not shown). E41 and E42 are in the S2-S3 loop, and S93 in the S6-S7 loop (Figure 5C). These sites, though greater than 22 Å apart, are both situated proximally to bound ubiquitin in a model based on the mRpn13:monoubiquitin structure27 (Figure 5C).
To assay mutational effects on ubiquitin chain binding, we used the native gel-based assay introduced in Figure 2A. Note that the mobility shift resulting from addition of GST-Rpn13 to the proteasome (Figure 5A and 5B) is irrelevant to the chain-binding assay, since Rpn13 itself does not affect proteasome migration in gels. Only the larger GST-tagged form of Rpn13 can do so, and, in the ubiquitin chain-binding assay, untagged Rpn13 was used. Neither the S93D mutant nor the E41K, E42K mutant conferred a strong defect in the proteasome-ubiquitin interaction, although both conferred reduced mobility shifts in comparison to wild-type Rpn13 (Figure 5D). To further impair ubiquitin binding, we combined the S2-S3 and S6-S7 mutations. The resulting protein -- a E41K, E42K, S93D triple mutant, referred to as scRpn13-KKD -- was comparable to a buffer control in its influence on the proteasome's electrophoretic mobility in the presence of ubiquitin conjugates (Figure 5E). NMR titrations revealed that scRpn13-KKD binds monoubiquitin ~8-fold more weakly than wild-type (Figure 5F; see also Supplement). To ensure that these mutations did not affect Rpn13 structural integrity, we compared a 1H,15N HSQC spectrum recorded on 15N labeled scRpn13-KKD with that of wild-type (Figure 5G). Only resonances corresponding to the mutated residues or their immediate neighbors were shifted, indicating that these surface mutations did not affect Rpn13's structure. In addition, we identified the hRpn13 surface that binds Rpn2, which is remote from the substituted residues27. The corresponding surface in scRpn13 is preserved in Rpn13-KKD as none of the residues within it were affected. Thus, Rpn13-KKD appeared to be suitable for in vivo analysis of the physiological function of ubiquitin recognition by Rpn13.
To test the biological significance of the Rpn13-ubiquitin interaction, we integrated the triple mutant allele into yeast in place of the wild-type chromosomal sequence. Functional defects in proteasomes can be revealed by plate assays such as sensitivity to the arginine analog canavanine, whose incorporation into proteins causes misfolding and accelerated degradation. The enhanced substrate load is lethal to mutants lacking proper proteasome function (e.g., 3). rpn13-KKD mutants proved sensitive to 8 μg/ml of canavanine when in the genetic background of an rpn10-uim mutation (Figure 6A). Thus, the rpn13-KKD mutation leads to a defect in proteasome function, and interacts synthetically with another specifically targeted proteasomal ubiquitin receptor mutation. Since Rpn13 can bind UBL/UBA proteins (Figure 4G), we also investigated its genetic relationship with Dsk2 and Rad23. rpn13-KKD also showed a strong synthetic interaction with a null allele of proteasomal ubiquitin receptor Dsk2 (Fig. 6A). In the case of Rad23, the genetic interaction was comparatively modest. These data support the view that the docking of ubiquitin-conjugates at the proteasome via UBL/UBA proteins is not mediated obligatorily via Rpn13. In addition, binding assays performed with purified proteasomes and the UBL domains of Rad23 and Dsk2 indicate that Rpn13 is not the sole receptor for UBL/UBA proteins on the proteasome (Supplementary Figure 7). The results of the binding assays are consistent with our previous report that proteasome subunit Rpn1 binds UBL/UBA proteins14. Further work is required to define more precisely the extent to which Rpn13–dependent docking of ubiquitin-conjugates at the proteasome is mediated or regulated by UBL/UBA proteins.
To test whether amino acid substitutions in the ubiquitin-binding loops of Rpn13 can lead to global defects in ubiquitin metabolism, whole cell extracts from our mutants were examined by immunoblotting. High molecular weight ubiquitin conjugates, which are enriched in proteasome substrates, accumulated in the rpn13-KKD rpn10-uim double mutant (Fig. 6B). The defect is synthetic, as with the canavanine-sensitivity of the double mutant (Fig. 6A). We also observed an in vivo proteolytic defect in the rpn13-KKD mutant (Fig. 6C), using the model proteasome substrate UbV76-Val-eΔK-ßgal30, which was previously found to be stabilized in an rpn13Δ mutant16.
Defects in proteasome assembly have been observed in rpn13Δ proteasomes (Figure 2; 26), and could potentially account for the phenotypes observed in Figures 6A through 6C. We therefore analyzed the assembly state of rpn13-KKD proteasomes by running native gels on freshly-prepared, unfractionated cell extracts31. We observed no significant change in level, assembly, or peptidase activity in rpn13-KKD proteasomes (Figure 6D). Surprisingly, no assembly defect was observed for rpn13Δ proteasomes (Figure 6D). Thus, assembly defects previously observed for rpn13Δ proteasomes are apparently a result of in vitro handling.
Although the recombinant Rpn13-KKD protein is properly folded and assembled onto purified proteasomes in vitro, it remained possible that the mutant protein would be absent from proteasomes in vivo as a result of its being rapidly degraded. To assess the extent of Rpn13-KKD loading of endogenous proteasomes, we used the GST-Rpn13 add-back assay of Fig. 5A, where purified proteasomes were used. In unfractionated cell extracts, GST-Rpn13 similarly shifted rpn13Δ proteasomes (Figure 6E). rpn13-KKD proteasomes behaved as wild-type in this assay, indicating that they were fully loaded with Rpn13-KKD. We conclude that the phenotypes of the rpn13-KKD mutant do not reflect deficient proteasome assembly or deficient loading of Rpn13 onto proteasomes, but are specifically related to its impaired ubiquitin-binding site.
We report here the identification of a new ubiquitin receptor for the proteasome, Rpn13, which is unrelated to Rpn10 and the three UBL/UBA proteins, and moreover defines a new class of ubiquitin recognition surface. Rpn13 differs dramatically from other proteasomal ubiquitin receptors. First, whereas the UBL/UBA proteins (and perhaps Rpn10) have distinct ubiquitin and proteasome binding domains that are separated by flexible linkers, Rpn13 is docked into the proteasome via a surface that is in close spatial proximity to its ubiquitin binding region. Thus, Rpn13 may provide for a ubiquitin chain with precise positioning and polarity. Second, the UBL/UBA proteins, having a large population free of the proteasome and oftentimes multiple ubiquitin binding domains, are better equipped than Rpn13 to capture ubiquitinated substrates and then deliver them to the proteasome. Third, UBL/UBA proteins are also capable of protecting the chain during transit to the proteasome, as they inhibit deubiquitination4,32,33. In striking contrast, Rpn13 promotes chain deubiquitination18,20,21. Binding to Rpn13 both facilitates Uch37's deubiquitinating activity20,21 and links Uch37 to the proteasome, suggesting that Rpn13 plays a major role in ubiquitin chain disassembly at the proteasome. Third, Rpn13 is exceptionally proficient at binding monoubiquitin and diubiquitin compared to other ubiquitin receptors associated with proteasome-mediated degradation. Although it is widely supposed that extended ubiquitin chains are required for degradation, contrary observations have been reported, in which monoubiquitin targets proteins to the proteasome24,34,35. Such substrates, although possibly rare, may be more strongly dependent on Rpn13 than those marked by chains.
Rpn13 resembles Rpn10 in its ability to bind ubiquitin-like domains of the UBL/UBA ubiquitin receptors (Figure 4D and 4G, Supplementary Figure 4 and 5), implying that Rpn13 may recruit substrates to the proteasome either directly or indirectly via UBL/UBA proteins. These observations support a hypothetical model whereby conjugates bind UBL/UBA family members, which dock them to the proteasome and pass them to the intrinsic receptors Rpn10 and Rpn13. In addition, compound complexes may be formed, in which a single conjugate is simultaneously bound by an intrinsic ubiquitin receptor and a UBL/UBA protein. Compound complexes may be favored when longer chains are delivered, leading to more stable chain-proteasome interactions, and thus more rapid substrate degradation.
Ubiquitin and the proteasome are both essential, but remarkably the inactivation of all five known proteasomal ubiquitin receptors in the same yeast strain does not appear to be lethal (data not shown). In our assays, the residual chain-binding capacity of rpn10-uim rpn13-KKD rad23Δ dsk2Δ ddi1Δ proteasomes is modest (Figure 5E), suggesting that the highest-affinity intrinsic receptors are now known. The unidentified receptor may be of lower affinity but comparable functionality, or it may be a receptor of typical affinity that is not intrinsic to the proteasome, such as Rad23/hHR23.
The phenotypic properties of multiply receptor deficient strains suggest functional redundancy (Figure 6; 3,32). This may reflect robustness in their principal function of delivering substrate to the proteasome. Given the number of known receptors, it is likely that docking of the chain at any of multiple locations in the proteasome will support breakdown of the target protein. However, there is apparently a deeper and more interesting functional relationship among ubiquitin receptors as well, in which they play distinct roles. For example, they have various phenotypes in isolation, albeit not lethal ones. Furthermore, proteasome function appears substantially compromised in multiply receptor-deficient strains3,32,37. When different receptor mutants are compared, the relative strengths of the degradation defects vary from substrate to substrate4. Thus, the receptors show in vivo specificity, although it remains unclear how specific they are and what the mechanistic basis of this specificity is. Finally, functionally redundant behavior as inferred from mutant phenotypes may not reflect the functioning of the wild-type system in a straightforward manner, since compensation can mask differentiated activities. With more detailed characterization, the individuality of proteasomal ubiquitin receptors and its mechanistic basis should become more clear.
Standard methods were used for yeast genetics, growth assays, and protein turnover assays (see Supplementary Information). A complete list of yeast strains is in the Supplementary Information. Sequences corresponding to mouse ubiquitin lacking two terminal glycines (UbΔGG) were subcloned into pYTH9 vector38, creating fusion proteins with the Gal4 DNA-binding domain (bait). Yeast strain Y190 was transformed with bait vector and human thymus cDNA library (Clontech) was screened as previously described38.
Antibodies used were: anti-myc (9E10) and anti-Ub (P4D1) from Santa Cruz Biotechnology, anti-ADRM1 (anti-Rpn13) from Biomol, and anti-ß-galactosidase (Promega). All constructs used in this study are described in Supplementary Information.
Recombinant proteins were expressed in and purified from Rosetta cells (Novagen). Proteasome was affinity-purified essentially as described39. Immunoprecipitation, immunoblotting, GST pull-down assays were performed as previously described25. Native gel analysis was performed as in 3. More detailed descriptions are available as Supplementary Information.
We thank members of our labs, as well as D. Hoeller, G. Dittmar, J. Lipscomb and M. Schmidt, for discussions, constructive comments and critical reading of the manuscript and the UMN BMBB NMR facility and MSI BSCL. We also thank Grzegorz Zapart for the initial Y2H ubiquitin screening, as well as Michael Groll and Patrick Schneider for allowing us to use the mRpn13:ubiquitin coordinates to generate Figure 5C. This work was supported by grants from Deutsche Forschungsgemeinschaft (DI 931/3-1), the Cluster of Excellence “Macromolecular Complexes” of the Goethe University Frankfurt (EXC115) to ID, and National Institutes of Health (CA097004 to KJW; GM043601 to DF; GM008700-CBITG to LR).
Accession Numbers: Coordinates for scRpn13 are available through the Protein Data Bank with accession number 2Z4D.