The 26S proteasome is a 2.5-MDa complex responsible for the selective, ATP-dependent degradation of ubiquitylated proteins in eukaryotic cells. Substrates in hundreds cellular pathways are timely ubiquitylated and converged to the proteasome by direct recognition or by multiple shuttle factors. Engagement of substrate protein triggers conformational changes of the proteasome, which drive substrate unfolding, deubiquitylation and translocation of substrates to proteolytic sites. Recent studies have challenged the previous paradigm that Lys48-linked tetraubiquitin is a minimal degradation signal: in addition, monoubiquitylation or multiple short ubiquitylations can serve as the targeting signal for proteasomal degradation. In this review, I highlight recent advances in our understanding of the proteasome structure, the ubiquitin topology in proteasome targeting, and the cellular factors that regulate proteasomal degradation.
The ubiquitin–proteasome system (UPS) regulates multiple cellular processes by selectively removing regulatory proteins and misfolded or injured proteins (1, 2). Recent advances of comprehensive approaches shed light on the vast UPS network involved in maintenance of the cellular proteome (3, 4). Indeed, in human cells, a diverse array of UPS components consisting of ∼1,000 proteins, regulate the fate of ∼5,000 substrate proteins. Due to the central role of this system in protein homeostasis, malfunctions of UPS components lead to a broad array of human diseases, including cancer, developmental disorders, neurodegenerative diseases, and autoimmune diseases. The 26S proteasome is a 2.5-MDa complex responsible for the selective degradation of ubiquitylated proteins. Cell proliferation depends on proteasome activity, and cancer cells require higher proteasome levels to cope with chronic proteotoxic stress; accordingly, proteasome inhibitors such as bortezomib and carfilzomib are used clinically for cancer therapy (5, 6). The successful development of proteasome inhibitors has drawn attention to other UPS components as promising drug targets.
In this review, I will highlight recent progress on the basic mechanisms of proteasomal degradation and the ubiquitin signals for the proteasome. Following that, I will highlight affiliated upstream factors of the proteasome including substrate shuttling factors, Cdc48/p97/VCP, and ubiquitin ligases.
Ubiquitylation: A Multifaceted Protein Modification with Diverse Structures
Substrates to be degraded are marked with ubiquitin, a small (76-amino acid) globular protein, by the concerted action of ubiquitin-activating enzymes (E1), ubiquitin-transferring enzymes (E2) and ubiquitin ligases (E3) (7). Ubiquitin can be ubiquitylated on any of seven lysine (K) residues or the N-terminal methionine (M), leading to eight structurally distinct types of polyubiquitin chains (M1-, K6-, K11, K27-, K29-, K33-, K48- and K63-linked chains) with various lengths. Among them, K48-linked chains are most abundant and play an essential role in proteasomal degradation. Ubiquitylation, like phosphorylation, also functions as a signal in DNA repair, protein trafficking, and NF-κB activation. Monoubiquitylation or K63- and M1-linked polyubiquitylation regulates protein interactions and enzyme activity in proteasome-independent processes, but recent studies have revealed that such chains also play proteolytic roles. Removal of these signals is achieved by deubiquitylating enzymes (DUBs), which antagonize the ubiquitylation machinery by removing ubiquitylation modifications from their substrates. Effector proteins harbouring ubiquitin-binding domains (UBDs) are assumed to function as readers/decoders by discriminating specific chain types (8). In addition to homotypic chains, cells contain heterotypic ubiquitin chains in which multiple linkages form mixed or branched chains. Furthermore, ubiquitin undergoes phosphorylation and acetylation at specific S/T and K residues (9). Ohtake and Tsuchiya describe the emerging complexity and signalling roles of ubiquitylation in this review series.
The Proteasomes: An Intricate Molecular Machine for Degradation
The 26S proteasome is a 2.5-MDa complex of 33 different subunits, which are arranged into an elongated structure composed of a central 20S core particle (CP) with one or two terminal 19S regulatory particle(s) (RP) (10–12) (Fig. 1). The CP consists of two outer α rings and two inner β rings, which are composed of seven structurally similar α and β subunits, respectively. Three of the seven β subunits have proteolytic sites (β1, β2, and β5; caspase-like, trypsin-like, and chymotrypsin-like activities, respectively) on the inner surface of the chamber formed by two abutting β rings. The combination of multiple active sites with different selectivity contributes to rapid and processive degradation of substrates that have entered the catalytic chamber. In mammalian cells, the proteolytic subunits can be replaced with immunotype subunits to form tissue-specific proteasomes, called the ‘immunoproteasome’ and ‘thymoproteasome’ that generate specific peptides for antigen presentations (13). The central narrow channel of the α ring forms an entrance gate for substrates, which is essentially closed in CP alone. Association of the RP or other proteasome activators such as PA28α/β (encoded by PSME1/2), PA28γ (PSME3) and PA200 (PSME4) induces gate opening, converting the CP into the degradation-competent state. Because the translocation channel is very narrow, a substrate protein must be unfolded before it can enter the channel. Hence, even in the open state, well-folded proteins do not enter the chamber of the CP. Accordingly, the proteasome is referred to as a ‘self-compartmentalizing protease’ (14). The RP consists of six ATPase subunits (Rpt1–6) and 13 non-ATPase subunits (Rpn1–3, 5–13 and 15), which play roles in processing ubiquitylated substrates. The ubiquitin receptor subunits (Rpn10, Rpn13 and the recently characterized Rpn1) capture ubiquitylated substrates directly or indirectly, whereas the ATPase subunits form a ring structure that promotes substrate unfolding and translocation into the CP, driven by ATP hydrolysis. Because PA28α/β, PA28γ, and PA200 do not utilize ATP and ubiquitin, the substrates of these alternative activators are generally unstructured proteins or peptides (15).
Recent advances in cryoelectron microscopy (cryoEM) and extensive efforts devoted to sample preparation, cross-linking and computational modelling have revealed the full structure of human and yeast 26S proteasomes, nearly at the atomic level (16–19). The RP can be biochemically separated into two subassemblies called the base and lid, both of which constitute assembly intermediates in cells (20, 21). The ATPase ring, two large scaffold subunits (Rpn1 and Rpn2), and the canonical ubiquitin receptor Rpn13, all form the base, which directly contacts the top of the CP. The N-terminal coiled-coil helices of ATPase subunits protrude from the ATPase ring and act as pillars to organize the overall architecture of the RP by contacting Rpn1 and Rpn2. Because Rpn2 is integrated into the lid, the ATPase can communicate with peripheral lid subunits. The lid, which consists of nine non-ATPase subunits, surrounds the ATPase ring. The most important subunit in the lid is Rpn11, a metalloprotease DUB, which hovers above the substrate entrance of the ATPase ring and removes the polyubiquitin chains from ubiquitylated substrates coupling with substrate translocation. Six non-ATPase subunits (Rpn3, 5, 6, 7, 9 and 12) harbouring a PCI domain assemble into a horseshoe-like structure to display Rpn11 above the ATPase mouth. The extended N-terminal domain of Rpn6 directly contacts with α2 subunit to stabilize the RP–CP interaction. The well-characterized ubiquitin receptors, Rpn10 and Rpn13, are localized in the apical region of the RP near the periphery, where they are well positioned to catch ubiquitylated substrates (22). In addition to the constitutive subunits, multiple cofactors, commonly referred to as proteasome-interacting proteins (PIPs), are transiently associated with the 26S proteasome. These cofactors, which include substrate shuttling factors, deubiquitylating enzymes, and ubiquitin ligases, all dynamically regulate proteasome function in cells. A large subunit, Rpn1 is the primary receptor for ubiquitin-like (UBL) domain-containing proteins (23).
The structure of the 26S proteasome is not static. To date, three conformational states of the 26S proteasome have been observed in cryoEM analyses: the substrate-free ground state (s1), substrate-engaged high-energy state (s3) and an intermediate between them (s2) (24, 25). These snapshots are interpreted as conformational switches of the RP during the substrate degradation cycle. In the ground state, the central axis of the ATPase ring is slightly tilted; consequently, the substrate-entrance channels of the ATPase ring and the CP α ring do not align. In the substrate-engaged state, which induces ATP hydrolysis, all the channels coaxially align to generate the degradation-competent conformation of the proteasome. After complete digestion of substrate proteins, the proteasome returns to the ground state. The conformational switch is originated by the alignment of six ATPase subunits. In the ground state, the C-terminal AAA+ domains of six Rpt proteins form an open-ring structure resembling a split washer. When a substrate enters the ATPase channel, the aromatic hydrophobic (Ar-Φ) pore-1 loops in the channel grab the extended unstructured region of the substrate, leading to rearrangement of the AAA+ domains into a planar ring configuration. ATP loading probably facilitates this conformational change, because the same structure was observed in proteasome preparations containing non-hydrolyzable nucleotides (ATPγS). The ATPase rearrangement transmits to Rpn2, resulting in a 25° rotation of the lid subunits. Concomitantly, the key subunits move to the proper positions for substrate processing. For substrate deubiquitylation, the active site of Rpn11 shifts near the ATPase mouth, whereas Rpn10 also moves closer to the mouth. Substrate protein is then translocated and gradually unfolded via a paddling movement of the six pore-1 loops, which are also aligned along the channel axis during the ATP cycle.
Ubiquitin Receptors in the Proteasome: More than Two
The first proteasomal ubiquitin receptor to be identified was Rpn10, also known as S5a/PSMD4 in mammals, which contains a short helical sequence called the ubiquitin-interacting motif (UIM) (26, 27). Yeast Rpn10 harbours a single UIM, whereas homologs in multicellular eukaryotes contain two or three of these domains. The additional UIMs can be the binding site of ubiquitin-like (UBL) domain-containing shuttle factors such as RAD23A/B in mammal cells and the autophagic vesicle protein ATG8 in plants (28, 29). Deletion of RPN10 in yeast results in only modest sensitivity to amino-acid analogues, which perturb the proteome balance, and these phenotypes are associated with the N-terminal von Willebrand factor A (VWA) domain, which plays roles in activating the RP (27, 30). In mice, RPN10 knockout is embryonic lethal, but deletion of the UIM alone results in a milder phenotype (31). These observations indicate that Rpn10 is not the only ubiquitin receptor in the proteasome. Subsequent studies identified several candidate subunits by multiple techniques, in the following historical order: Rpt5, by cross-linking with ubiquitin; Rpn13, by yeast two-hybrid screening with ubiquitin; Rpn15/Sem1, by genetic screening; and Rpn1, by structural analysis (32–35). Among these subunits, Rpn10 and Rpn13 are generally thought to function as the primary ubiquitin receptors of the proteasome because the proteasome harbouring simultaneous mutations in these subunits do not bind efficiently to polyubiquitylated substrates in vitro (36). In mice, liver-specific deletion of either RPN10 or RPN13 showed modest impairment, but simultaneous loss of both RPN10 and RPN13 caused severe liver injury accompanied by massive accumulation of ubiquitin conjugates, suggesting redundant roles of RPN10 and RPN13 in ubiquitin recognition of the proteasome (37). Rpn13 recognizes ubiquitin through a globular Pru (pleckstrin-like receptor for ubiquitin) domain, whose opposite surface is involved in the interaction with Rpn2. The distance between Rpn10 and Rpn13 is approximately 90 Å in both the ground and substrate-engaged states (17, 22), sufficient to accommodate tetraubiquitin. In multicellular eukaryotes, Rpn13 contains an additional sequence called the DEUBiquitinase Adaptor (DEUBAD) domain, which is involved in binding the DUB UCH37/UCHL5 (38, 39). UCH37 also associates with the INO80 chromatin remodelling factor via an interaction with the DEUBAD domain of INO80G (NFRKB). Interestingly, UCH37 is stimulated 8-fold by Rpn13, but is inhibited in the INO80 complex. UCH37 is thought to remodel ubiquitin chains within the proteasome by trimming the chains from their distal ends; thus, UCH37 can limit proteasomal degradation of certain substrates. Another proteasome-associated DUB, Ubp6/USP14, is also activated in the proteasome. Ubp6/USP14 is tethered to Rpn1 via its N-terminal UBL domain, and when ubiquitin is loaded, its C-terminal catalytic domain contacts the ATPase mouth (40, 41). In contrast to UCH37, USP14 can remove polyubiquitin chains en bloc from multiply polyubiquitylated substrates, reducing the burden on Rpn11 and promoting efficient degradation (42).
Very recently, it was revealed that Rpn1 could recognize not only UBL domains but also ubiquitin (35). Rpn1 has two toroidal regions formed by multiple short hairpin repeats known as PC repeats. Within the first toroid, which is positioned near the ATPase mouth, Shi et al. identified two distinct ligand-binding sites, T1 and T2. The T1 site is utilized for ubiquitin and UBL-containing shuttle factors, whereas the T2 site is used for Ubp6/USP14. Thus, the Rpn1 toroid provides adjacent receptor sites for ubiquitylated substrates, shuttling factors, and deubiquitylating enzymes proximal to the substrate entry mouth. That study also revealed that Rpn1, Rpn10 and Rpn13 have redundant functions as ubiquitin/UBL receptors, suggesting that substrate recognition by the proteasome is surprisingly complex, i.e. a single ubiquitylated substrate can be bound in a multivalent manner by multiple ubiquitin receptors. Curiously, simultaneous mutation of RPN1 T1, RPN10 UIM and RPN13 Pru, in yeast cells results in only a modest phenotype, indicating that proteasome might contain as-yet-unidentified ubiquitin-binding sites in its depths. One strong candidate is Rpt5, which can be cross-linked with ubiquitin chains only within the intact proteasome (32). The ubiquitin receptor function of Rpn15/Sem1/Dss1 remains controversial (34, 35). Because Rpn15 is localized on the side opposite the ubiquitin-processing site in the RP, its proposed ubiquitin-binding activity might contribute to proteasome anchoring to ubiquitin-rich structures such as injured mitochondria.
Ubiquitin Signals for the Proteasome: K48 Is Not Everything
K48-linked ubiquitin chains are the most abundant chain type in yeast, plant and mammalian cells (43–47). The K48 linkage was originally identified in a short-lived model substrate of the N-end rule pathway (48). Subsequent genetic studies in yeast revealed that only the K48 residue, among seven K residues of ubiquitin, is essential for cell growth and degradation of cell-cycle regulators (49, 50). In 2000, using a fully defined reconstituted system containing ubiquitylated substrate and purified 26S proteasomes, Thrower et al. (51) provided direct evidence that the proteasome recognizes K48-linked tetraubiquitin as a minimal targeting signal, and that binding strength increases markedly as chain length increases up to octaubiquitin (Fig. 2A, i). Although the study was performed using a well-folded model substrate, dihydrofolate reductase (DHFR), subsequent elegant reconstitution studies showed that K48-linked chains also promote rapid proteasomal degradation of an endogenous substrate, the CDK inhibitor Sic1 (52, 53). At that time, researchers noticed that ubiquitylation is not sufficient for efficient degradation of tightly folded proteins. Prakash et al. (54) reported that an unstructured disordered region within a substrate protein is involved in initiating proteasomal degradation. Indeed, even when ubiquitylated, well-folded globular proteins such as GFP are not degraded, whereas the addition of a disordered polypeptide tail, typically 20–30 amino acids length, stimulates rapid degradation of GFP (55). At the proteome level, initiation sites for efficient engagement by the proteasome seem to have unbiased sequences with complex amino-acid compositions (56).
Recent advances in proteomic approaches have provided a draft map of the ubiquitin proteome in cells (3). A combination of quantitative proteomics and use of the anti-diGly antibody, which recognizes ubiquitylated peptides within tryptic fragments, revealed that a large number of proteins (∼5,000) are ubiquitylated in human cultured cells (46). Unexpectedly, >60% of these proteins contained multiple ubiquitylation sites, and ∼58% of the ubiquitylated peptides became more abundant following treatment with a proteasome inhibitor. Another quantitative proteomics study using isotope-labelled ubiquitin revealed that ubiquitylated substrates mainly exist in the monoubiquitylated form (47). Thus, these proteomics studies revealed that substantial numbers of endogenous proteasome substrates are modified with multiple mono/polyubiquitins rather than a single chain. The K48 linkage rapidly and significantly accumulates in cells treated with a proteasome inhibitor, and other linkages such as K6, K11, K27, K29, K33 and to a lesser extent K63 also became more abundant, suggesting the involvement of these chain types in proteasomal degradation (43, 46, 47).
Consistent with the in vivo observations, recent in vitro studies show that purified proteasomes can recognize a broader range of ubiquitylation topologies on various degradation substrates. Using a single-molecule approach, Lu et al. showed that the proteasome efficiently degrades cyclin B1 with multiple short ubiquitin chains (57, 58) (Fig. 2A, ii). In their reconstituted system, a large multiprotein E3 ligase, the APC/C (anaphase-promoting complex/cyclosome) rapidly modifies its substrates with mono-, di- and tri-ubiquitins on multiple sites via a processive affinity amplification mechanism (57, 59). Lu et al. engineered cyclin B1 with various ubiquitylations of defined chain length, and monitored the resident time of these proteins on the proteasome at the single-molecule level using total internal reflection fluorescent (TIRF) microscopy. Substrate binding increased exponentially for the first three ubiquitins in any configuration, i.e. multiple mono- or short ubiquitin chains, and linearly from four to nine ubiquitins. Among possible configurations with four ubiquitins, substrates modified with two K48-linked di-ubiquitin chains provided a stronger degradation signal than those modified with a single K48-linked tetraubiquitin chain. The observations were confirmed by a more recent study using the model substrate GFP-tail (55). Multiple short chains linked through K11, K27 and K63 also support substrate binding, indicating that the proteasome does not discriminate among chain types on multiply ubiquitylated substrates. Therefore, the local concentration of substrate-attached ubiquitin groups, rather than linkage type, is important for substrate recognition by the proteasome. Presumably, weak multivalent interactions between three ubiquitin receptors and ubiquitin might be advantageous not only for stochastic increase of substrate recognition but also for release of ubiquitin from the proteasome after substrate deubiquitylation.
The multivalent recognition mode might also be observed for branched ubiquitin chains. The APC/C, along with two different E2s, UBE2C and UBE2S, can assemble branched ubiquitin chains with K11 and K48 linkages in cells (60). APC/C substrates modified with K11/K48 branched chains are more efficiently degraded than those with homotypic K48-linked chains (60, 61). Although the topology of the branched chains remains unclear because single homotypic K11-linked ubiquitin chains does not bind the proteasome or stimulate substrate degradation (55, 61), chain branching with K11 linkages might increase the number of K48-linked chains. Among possible branched ubiquitin chains, we recently found that cells contain substantial levels of K48/K63-branched ubiquitin chains (62). The cellular abundance of such chains increased upon treatment with a proteasome inhibitor, suggesting that branched chains of this type constitute a proteasome targeting signal.
Although K63-linked chains are not usually involved in proteasomal degradation in cells, purified proteasomes can bind and process both K48- and K63-linked ubiquitylated substrates at comparable levels (36, 63, 64). We previously developed a simple ubiquitylation system using the HECT-type E3 Rsp5 and its degron PY-motif (65). Sic1 containing the PY motif (Sic1PY) is highly ubiquitylated, with long K63-linked chains, by Rsp5. The ubiquitylated Sic1PY is rapidly degraded by the proteasome; accordingly, this system is widely used for assaying the proteasome function (35, 63, 66). Why then, do K63-linked chains not support proteasomal degradation in cells? One explanation might be the participation of K63-chain-selective UBD proteins in vivo. Nathan et al. demonstrated that the ESCRT-0 complex, which functions in the MVB sorting pathway, binds K63-linked ubiquitylated substrates and prevents their degradation by the proteasome (64). Alternatively, K63-linked chains, which may be single chains, on endogenous substrates are too short to be recognized by the proteasome and shuttle factors.
Interestingly, the requirements of ubiquitylation levels for proteasomal degradation correlate with the sizes and structural features of the substrate proteins themselves. Although attachment of a single ubiquitin may provide a weaker proteasomal targeting signal, in reticulocyte lysates, monoubiquitylation is sufficient to induce proteasomal degradation for structurally disordered proteins shorter than 150 amino acids (67, 68). A recent proteomics study using yeast and human cells in which all ubiquitin was replaced with lysine-less ubiquitin to restrict polyubiquitin synthesis (69) revealed that numerous proteins (∼25% in yeast and ∼50% in human, respectively) are degraded by mono- or multiple monoubiquitylation, and that these substrate proteins share certain features; in particular, most are smaller and more structurally disordered than polyubiquitylated substrates. Physiologically, monoubiquitylated substrates are enriched among proteins with roles in carbohydrate transport and oxidative stress response pathways, suggesting that specific E3s are responsible for their monoubiquitylation.
The aforementioned studies reveal unexpected complexities of the ubiquitin code for the proteasome, but we still do not have a comprehensive picture of proteasomal degradation pathways. In cells, many upstream factors, including shuttle factors such as UBL–UBA (ubiquitin-associated domain) proteins, the ubiquitin-selective chaperone p97/VCP (Cdc48 in yeast) and proteasome-associated ubiquitin ligases, are involved in proteasomal degradation, but the relative contributions of these factors are still unclear. In the latter part of this review, I summarize current knowledge about the shuttle factors, p97/VCP/Cdc48, and proteasome-associated ubiquitin ligases.
Shuttling Factors: Indirect Pathways to the Proteasome
Another mode of ubiquitin recognition of the proteasome involves shuttle factors such as Rad23 (RAD23A/B in mammals), Dsk2 (UBQLN1–4 and L) and Ddi1 (DDI1/2) (Fig. 2B). These extrinsic ubiquitin receptors contain a UBL domain and one or two UBA domains (2). Without substrates, it is generally assumed that UBL–UBA proteins exist in an autoinhibitory state mediated in cis or trans by interactions between UBL and UBA domains. When ubiquitylated substrates are captured by the UBA domain, the structure opens, and the released UBL domain can bind to the UBL receptors of the proteasome. Although the UBL–UBA proteins play overlapping roles with each other and the intrinsic receptors, they exhibit some substrate specificity. Rad23 was originally characterized as a DNA repair factor that forms a complex with Rad4 (XPC in human) in the nucleotide excision repair (NER) pathway. Subsequently, Rad23 was shown to have a proteolytic function in cell-cycle progression, stress response and ER-associated degradation (ERAD). Because Rad23 can bind ubiquitin chains through its two UBA domains, although it is specific for K48-linked chains, a broad range of substrates might be shuttled by this receptor (64, 70,). In addition, yeast Rad23 can collaborate with the ubiquitin elongation factor Ufd2 in the so called ‘escort pathway’ (71, 72). Ufd2 was originally identified in a screen for mutants defective in ubiquitin-fusion degradation (Ufd) pathway (73), and subsequently as a cofactor of the ubiquitin-selective chaperone Cdc48 (71). In the escort pathway, oligo-ubiquitylated substrates in membrane or protein complexes are recognized and extracted by Cdc48, and Ufd2 modifies the client proteins with K48-linked chains of four to six ubiquitins. Rad23 is recruited to the ternary complex, in which its UBL domain associates with Ufd2, and then Cdc48 remodels the complex to release the Rad23-ubiquitylated substrate for degradation. The escort pathway is well characterized in yeast, but remains elusive in mammalian cells. Interestingly, Rad23 and RAD23A/B can directly stabilize several proteins. In the NER pathway, Rad4/XPC is constitutively degraded by the UPS, but Rad23 stabilizes Rad4 by associating with it through a unique XPC-binding domain, which shares homology with the STI1 (heat-shock chaperone-binding) domain. Moreover, in fly and mammalian cells, RAD23A and B directly associate with Ataxin-3/ATXN3 and regulate its stability (74).
Yeast Dsk2 has a redundant function with Rad23 in ERAD and the escort pathway. Its mammalian homologues are the five ubiquilins (UBQLN1-4 and L), of which UBQLN1, 2 and 4 might have redundant functions (75, 76). The importance of UBQLNs is highlighted by the fact that mutations in UBQLN2 are linked to amyotrophic lateral sclerosis (ALS) and frontotemporal lobar degeneration (77). Dsk2 and UBQLNs share identical architectures that contain UBL–STI1–UBA domains. Curiously, the single UBA domain can bind any ubiquitin chain type, suggesting that Dsk2/UBQLNs are non-selective UBD proteins (70, 78). The central STI1-like domain plays specific role in the recognition of newly synthesized transmembrane proteins in the cytosol, to prevent aggregation and facilitate membrane targeting to the ER or mitochondria (76, 79). In addition, the STI1-like domain can bind HSP70, activating UBQLN to shuttle protein aggregates to the proteasome (80). Importantly, pathogenic mutations of UBQLN2 caused protein aggregation, and the resultant aggregates are cleared by collaboration between the proteasome and HSP70. Thus, the HSP70–UBQLN–proteasome axis plays a pivotal role in protein homeostasis.
Ddi1 and its mammalian counterparts, DDI1 and DDI2, exhibit properties unique among the canonical shuttle factors. Ddi1 and DDI1/2 each contain a retroviral aspartyl protease (RVP) domain found in HIV. Curiously, DDI1 and DDI2 lack the C-terminal UBA domain, but instead bind ubiquitin via a UIM (81). Yeast Ddi1 is not involved in the escort and ERAD pathways, but is specifically involved in degradation of the F-box protein Ufo1 and its substrate, the HO endonuclease (82). In addition, its UBL domain binds only weakly to the proteasome, but can bind with ubiquitin (81). Very recently, human DDI2 emerged as a critical transcriptional regulator of proteasome genes (83, 84). Expression of human proteasome genes is controlled by an ER-anchored transcriptional factor, Nrf1 (NFE2L1) (85). When proteasome activity is compromised, Nrf1 is cleaved by DDI2 and activated as a transcriptional factor. Although the mechanism of specific recognition of Nrf1 by DDI2 and the function of the UBL domain remain unclear, DDI2 represents an attractive target for anti-proteasome drugs.
Mammals have additional UBL–UBA proteins, such as UBAC1, UBL7 and NUB1L. Although the proteasomal functions of UBAC1 and UBL7 are not known, NUB1L has several unique features. NUB1L (NEDD8 ultimate buster-1 long), which contains one UBL and three UBA domains, was originally reported as a negative regulator of neddylation, the covalent modification of proteins with the ubiquitin-like protein NEDD8 (86). NUB1L specifically associates with NEDD8 or neddylated proteins through its UBA domains and induces proteasomal degradation. Although the precise mechanisms are unclear, reduction in NEDD8 levels mediated by NUB1L enhances ubiquitylation and proteasomal degradation of aggregation-prone proteins (87). In addition, NUB1L also stimulates proteasomal degradation of FAT10, another ubiquitin-like modifier, and its conjugates (88). In this case, the three UBA domains recognize FAT10, whereas the UBL domain interacts with the VWA domain of human Rpn10. Because expression of FAT10 and NUB1L is induced by proinflammatory cytokines, proteasomal degradation of FAT10 substrates might be involved in antigen presentation.
Structurally distinct shuttle factors have also been identified. AIRAPL (arsenite-inducible 19S RP-associated protein-like, encoded by ZFAND2B in human) is an evolutionarily conserved UIM-containing protein that associates with the proteasome upon exposure of cells to arsenite (89). Although the precise functions of AIRAPL are unknown, worms lacking this protein have reduced lifespans, and AIRAPL-deficient mice develop cancer as a result of impaired proteostasis (90, 91). AIRAPL can also associate with p97/VCP, and the tandem UIM found in AIRAPL is highly specific for K48-linked chains, suggesting that AIRAPL might recruit a broad range of substrates to the proteasome (92).
Cdc48/VCP/p97: More Upstream Factor
The AAA-ATPase Cdc48/p97/VCP plays a crucial role in substrate processing at the post-ubiquitylation step in multiple ubiquitin-dependent pathways (93, 94). Cdc48/p97/VCP functions as a segregase for the extraction of ubiquitylated proteins from multimeric complexes, membranes and chromatin (Fig. 2B). Importantly, like the proteasome, Cdc48/p97/VCP has multiple cofactors that function as ubiquitin receptors, DUBs and ubiquitin ligases. Although Cdc48/p97/VCP itself can bind ubiquitin, the cofactors are generally required for processing of substrates. Ufd1-Npl4 heterodimer (UFD1L-NPLOC4 in human), the best-characterized essential cofactor, functions in the UFD pathways, ERAD, degradation of nascent misfolded proteins, replication and DNA repair. UFD1L and NPLOC4 possess a UT3 (Ufd1 truncation 3) domain and NZF (Npl4 zinc finger) domain as UBDs, respectively, whereas yeast Npl4 lacks the NZF domain (95). Interestingly, the NPLOC4 NZF domain binds both K48- and K63-linked chains, and p97/VCP also participates in clearance of protein aggregates and stress granules via autophagy, implying that p97/VCP in complex with UFD1L-NPLOC4 may serve as a node that orchestrates the two degradation pathways. Although the precise functions of other cofactors in mammals are less clear, yeast Shp1/Ubx1 functions in membrane fusion processes by recruiting monoubiquitylated substrates to Cdc48, and also regulates autophagosome biogenesis (96, 97), whereas Ufd3 is involved in protein sorting, proteasomal degradation and maintenance of cellular ubiquitin levels (98–100).
As mentioned above, the proteasome cannot degrade tightly folded proteins without unstructured initiation sites in vitro. However, such a substrate, ubiquitin-fused GFP, is widely used for monitoring proteasome activity in vivo, but its degradation requires Cdc48/p97/VCP activity. Godderz et al. pointed out this discrepancy and showed that long polyubiquitin chains are primarily required for Cdc48-dependent unfolding rather than the proteasome itself (101). Although the putative unfolding or unfoldase activity has not been clarified in detail, it is clear that Cdc48/p97/VCP assists proteasomal degradation of difficult-to-degrade substrates and aggregation-prone substrates to reduce the burden on the proteasome. Reduced activity of p97/VCP impairs protein homeostasis and induces formation of protein aggregates; indeed, mutations in p97/VCP cause the degenerative late-onset disorders IBMPFD (inclusion body myopathy associated with Paget disease of bone and frontotemporal dementia) and ALS (102, 103). On the other hand, ATP-competitive inhibitors of p97/VCP, such as CB-5083, are developed for treatment of various cancers including solid tumours (104). Thus, further studies are required in order to fully understand Cdc48/p97/VCP in the context of ubiquitin chain-type selectivity and its relationship with the proteasome.
Ubiquitin Ligases That Function on the Proteasome: Energy Saving?
Several ubiquitin ligases associate with the proteasome, where they might serve to control ubiquitylation of certain substrates on the proteasome (Fig. 2C). The HECT-type ubiquitin ligase Hul5, the most abundant ubiquitin ligase in the yeast proteasome, directly interacts with Rpn2 and elongates K63-linked ubiquitin chains on proteasome-loaded ubiquitylated substrates (105). In addition to Hul5, Ubr1 and Ufd4, two ubiquitin ligases historically shown to target short-lived proteins, interact with the proteasome via Rpt6 and Rpt4, respectively (106). Given that Hul5 and Ubr1 are required for clearance of heat shock-induced misfolded proteins in the cytosol, the interaction with the proteasome might be advantageous for rapid and effective elimination of injured proteins (107, 108).
In humans, the interactions of proteasome-associated ubiquitin ligases are poorly characterized, but mutations of this class of ubiquitin ligases are associated with human diseases. The HECT-type ubiquitin ligase UBE3A, also known as E6AP, ubiquitylates proteasome subunits and RAD23A/B, which deregulates proteasome activity (66, 109, 110). Mutations of UBE3A are linked to Angelman syndrome, a neurodevelopmental disorder characterized by severe intellectual and developmental disability. The physiological substrates and regulation of human Hul5 orthologues, UBE3B and UBE3C/KIAA10, remain unclear, but mutations of either gene are linked to Kaufman oculocerebrofacial syndrome and Kabuki syndrome, both of which cause growth retardation and accompanying intellectual disability. Parkin (encoded by PARK2) is a RING-type ubiquitin ligase in which mutations are linked to familial Parkinson’s disease (PD) (111). Recent extensive studies showed that Parkin ubiquitylates mitochondrial outer membrane proteins to clear injured mitochondria via autophagy, a process called mitophagy. Interestingly, Parkin has a UBL domain that can directly associate with Rpn10 and Rpn13 (112, 113). Impairment of the interaction between Parkin and the proteasome results in significant delay of mitochondrial clearance, suggesting that proteasomes are also involved in mitophagy. Another PD-associated ubiquitin ligase, FBXO7/PARK15, also harbours a UBL domain, which binds α2 (PSMA2) of the proteasome, and seems to regulate proteasome assembly (114). Thus, the proteasome itself functions as the platform for a series of ubiquitin ligases involved in substrate degradation and regulation of the proteasome’s functions. Therefore, it would be of great interest to determine the contributions of these proteasome-associated ubiquitin ligases.
Conclusion and Perspectives
Given that the proteasome is an essential element of protein homeostasis, its content and activity must be tightly controlled by the cellular environment. Although we now have an extensive picture of how ubiquitylated proteins are recognized and degraded by the proteasome, it remains unclear how the vast network surrounding the proteasome machinery orchestrates degradation at the proteome level. In this regard, we still do not know how much of each interacting protein is loaded onto the proteasome, nor do we know their relative contributions to proteasomal degradation. To understand proteasomal degradation more precisely, reconstitution studies with such cofactors and quantitative proteomics approaches may be necessary.
I thank Dr. Tsunehiro Mizushima for providing atomic models of the proteasome. Y.S. is supported by Grants-in-Aid from the Ministry of Education, Culture, Sports, Science, and Technology of Japan (JP24112008), the Sumitomo Foundation for Basic Science Research (130479), and the Takeda Science Foundation.
Conflict of Interest
homologous with E6-associated protein C-terminus
ubiquitin fusion degradation
really interesting new gene