Covalent modification of proteins with ubiquitin is a common regulatory mechanism in eukaryotic cells. Typically, ubiquitinated proteins are targeted for degradation by the 26 S proteasome. However, more recently the ubiquitin signal has also been connected with many other cell processes, including endocytosis, vesicle fusion, DNA repair and transcriptional silencing. Hence ubiquitination may be comparable with phosphorylation in its importance as an intracellular switch, controlling various signal-transduction pathways. Similar to the regulation of the extent of phosphorylation by kinases and phosphatases, specific sets of ubiquitinating/deubiquitinating enzymes control the degree of ubiquitination. A large number of ubiquitin-binding proteins act at different steps in the downstream pathways, followed by the ubiquitinated protein. Different families of ubiquitin-binding proteins have been described. UBA (ubiquitin-associated) domain-containing proteins is the largest family and includes members involved in different cell processes. The smaller groups of UIM (ubiquitin-interacting motif), GAT [GGA (Golgi-associated γ-adaptin homologous) and Tom1 (target of Myb 1)], CUE (coupling of ubiquitin conjugation to endoplasmic reticulum degradation), UEV [ubiquitin E2 (ubiquitin-conjugating enzyme) variant] and NZF (nuclear protein localization gene 4 zinc finger) domain-containing proteins appear to have more specialized functions. Here we discuss functional and structural properties of ubiquitin-binding proteins.
An efficient and common way to alter protein function is by post-translational modifications, such as phosphorylation, glycosylation and acetylation, or through the addition of a ubiquitin molecule in a process called ubiquitination . Typically, ubiquitination of a protein leads to its degradation in the 26 S proteasome. However, ubiquitination has also been linked to a number of other cellular activities, including endocytosis, vesicle fusion, DNA repair and gene silencing, to name a few .
The 26 S proteasome is an abundant protein complex found in the cytoplasm and nucleus of all eukaryotic cells. It is composed of two subcomplexes: a 20 S cylindrical core particle, which provides the proteolytic activity, and a 19 S regulatory complex, which mediates substrate recognition and substrate unfolding. The 19 S particle is itself composed of two smaller subcomplexes: the base complex, which binds the 20 S core particle and contains ATPase subunits, and the lid complex, which covers the base and is functionally not as well characterized .
Ubiquitin–substrate ligation is catalysed by the sequential action of three enzymes, E1 (ubiquitin-activating enzyme), E2 (ubiquitin-conjugating enzyme) and E3 (ubiquitin ligase) . Ubiquitin is first activated by the E1 and then, in collaboration between an E2 and E3, transferred to a specific target protein, forming an isopeptide bond between the C-terminus of ubiquitin and a primary amino group in the target protein. In some cases, the ubiquitination process stops at this point . However, many ubiquitinated proteins undergo several rounds of ubiquitination, so that a new ubiquitin moiety is conjugated to a lysine residue in the last ubiquitin moiety of the growing ubiquitin chain. Some proteins require the action of a ubiquitin-chain-elongation factor, E4 (polyubiquitin chain conjugation factor), to become polyubiquitinated efficiently . Ubiquitin contains seven lysine residues, and chains connected through most of these residues have been detected in vivo [2,6]. Lys48-linked chains are the most common signal for proteasomal degradation, while the other lysine-linkages are mostly involved in non-proteolytic functions. For instance, Lys63-linked ubiquitin chains regulate tolerance to DNA damage .
In combination, the E2 and E3 enzymes determine the type of chain linkage and substrate specificity of the ubiquitination process, and human cells probably contain hundreds of different E3s. However, at least two other levels of regulation can be envisioned. Firstly, regulation may also occur by deubiquitination of substrates, a process catalysed by a large family of DUBs (deubiquitinating enzymes) . Secondly, regulation may occur at the level of recognizing the ubiquitinated protein as a signal for inducing a downstream cellular event .
In this review, we describe certain ubiquitin-binding proteins. However, the number of potential and proven ubiquitin-binding proteins is steadily increasing. We have assembled a list of ubiquitin-binding proteins ordered according to function (Table 1), but this account is in no way comprehensive.
Families of ubiquitin-binding proteins
Several different families of ubiquitin-binding proteins have been described. The largest group, the UBA (ubiquitin-associated) domain-containing proteins, includes members involved in different cellular processes. The families of UIM (ubiquitin-interacting motif), CUE [coupling of ubiquitin conjugation to ER (endoplasmic reticulum) degradation], GAT [GGA (Golgi-associated γ-adaptin homologous) and Tom (target of Myb)], UEV (ubiquitin E2 variant) and NZF [Npl4 (nuclear protein localization gene 4) zinc finger] domain-containing proteins appear to some extent to be more specialized in their functions .
The structures of the ubiquitin-binding domains [9–15] have been determined for representative members of the families (Figure 1). Except for the somewhat similar UBA (Figure 1A) and CUE (Figure 1B) domains, their three-dimensional structures are disparate. However, in all cases, the domains appear to interact with a hydrophobic patch on ubiquitin, formed by residues Leu8, Ile44 and Val70 .
UBA domains have been shown also to bind UBL (ubiquitin-like) domains , which show only limited sequence similarity with ubiquitin, but share the ubiquitin fold. The function of the UBL domain is also connected with the ubiquitin–proteasome system, since it interacts with the 26 S proteasome [18–20], but the significance of the interactions between ubiquitin-binding domains and UBL domains is yet unknown.
Ubiquitin-binding proteins in proteolysis
Ubiquitin binding by the 26 S proteasome
Two 26 S proteasome subunits, S6′/Rpt5  and S5a/Rpn10/Pus1 , have been shown to interact directly with ubiquitin chains. The interaction with the ATPase subunit S6′/Rpt5 appears to be modulated by ATP hydrolysis . However, the primary sequence within S6′/Rpt5, which is responsible for the interaction, has not yet been defined.
S5a/Rpn10/Pus1 was the first protein found to interact non-covalently with ubiquitin chains. In human S5a/Rpn10/Pus1, which contains two UIM domains (whereas the yeast orthologues contains only one), the ubiquitin-binding region was mapped to an approx. 20-amino-acid region near the C-terminus , now known as the UIM domain, which is found in several proteins . The UIM domain forms a helix (Figure 1D)  that is amphipathic and on one side engages the hydrophobic patch in ubiquitin.
Most proteasome subunits are essential, but curiously yeast deleted for the RPN10/pus1 gene proved viable [24,25]. More recent studies (described below) have clarified this initially surprising result.
In the N-terminus, S5a/Rpn10/Pus1 contains a VWA (von Willebrand factor A) domain, a domain found in cell adhesion molecules in higher eukaryotes  but encountered only rarely in intracellular proteins. The VWA domain in S5a/Rpn10/Pus1 interacts with PC (proteasome/cyclosome) repeats . PC repeats are found exclusively in the S2/Rpn1/Mts4 and S1/Rpn2 subunits of the 26 S proteasome and in the Apc1/Cut4 subunit of the E3, called the APC/C (anaphase-promoting complex/cyclosome) . Therefore an attractive model for the function of S5a/Rpn10/Pus1 is that it may shuttle from the APC/C to the proteasome to pick-up and deliver, respectively, ubiquitinated protein cargo for degradation . This observation perhaps explains the finding that a substantial amount of S5a/Rpn10/Pus1 exists in a free form in yeast cells [24,25]. However, in HeLa cells, S5a/Rpn10/Pus1 is found exclusively in high molecular mass fractions, and proteasome-bound S5a/Rpn10/Pus1 does not appear to dissociate from the 26 S proteasome more readily than any other subunit . Together, these data point to a functional difference between human and yeast S5a/Rpn10/Pus1.
The UBA domain is a widespread, approx. 50-residue, protein module, which structurally forms a bundle of three helices (Figure 1A) [9,30]. Even in the relatively small fission and budding yeast genomes, about 15 different UBA domain-containing proteins are encoded. Most of these genes have orthologues in higher eukaryotes, which, in addition, contain many other more specialized proteins with UBA domains. The most well-characterized UBA domain proteins are the UBL/UBA proteins Rad23/Rhp23 and Dsk2/Dph1.
Recruiting substrates to proteasomes via UBL/UBA adaptors
The UBA domain interacts with ubiquitin chains, and the UBL/UBA proteins therefore link ubiquitinated substrates to the 26 S proteasome . Single knock-out strains proved viable. However, a triple mutant in Rhp23, Dph1 and Pus1 was inviable, indicating that the UBL/UBA proteins function in parallel to the proteasome's substrate-receptor subunit S5a/Rpn10/Pus1 . Thus, Rad23/Rhp23 and Dsk2/Dph1 may function as the 26 S proteasome's substrate receptors, or by carrying ubiquitinated proteins from the E3 enzymes to the proteasome .
The first attempt to verify this ‘substrate shuttle model’ by biochemical assays was unsuccessful, but revealed that UBA domains protect ubiquitin chains from disassembly by DUBs, and as chain disassembly is coupled to proteasomal degradation, UBA domains were inhibitory for proteasomal degradation in vivo . Recently, however, an elegant cell-free system was developed, which provided direct evidence that the UBL/UBA domain proteins actually fulfil the role of bona fide substrate receptors/carriers .
The human genome encodes two Rad23/Rhp23 orthologues, hHR23a and hHR23b. The mammalian orthologues of yeast Dsk2/Dph1 are known as ubiquilins or PLICs. The protein Nub1 [Nedd8 (neural precursor cell expressed developmentally down-regulated 8) ultimate buster-1] has a domain organization similar to that of Rad23/Rhp23, but has not been linked to degradation of ubiquitinated proteins, but rather to the degradation of proteins conjugated to the small ubiquitin-like modifier Nedd8 . This is a controversial observation, as Nedd8-conjugated proteins are not normally targeted for degradation. More recently, Nub1 was also shown to mediate the degradation of another ubiquitin-like modifier, FAT10 .
Ubiquitin-binding adaptors of the Cdc48 ATPase
A subfamily of UBA-domain proteins contains both UBA and UBX (ubiquitin regulatory X) domains . The UBX domain is a general Cdc48-interacting module [37–39]. Cdc48, also known as VCP (valosin-containing protein) and p97, is a hexameric ATPase complex thought to unfold proteins and is involved in fusion of ER and Golgi membranes, but also in spindle disassembly, DNA synthesis, and degradation of ubiquitinated proteins . The question of how Cdc48 is involved in such a diverse set of cellular processes remains open, but it is probably connected with a host of Cdc48 cofactors, including the UBA/UBX domain proteins.
The UBA/UBX domain protein p47 (Ubx3 in Schizosaccharomyces pombe and Shp1 in Saccharomyces cerevisiae) interacts with ubiquitin chains via its N-terminal UBA domain , a binding required for its function in Golgi-membrane fusion. As proteasome activity is not required for membrane fusion, the recruitment of certain ubiquitinated proteins to Cdc48 by p47 is not connected with proteasomal degradation . However, the recruitment of ubiquitinated substrates to Cdc48 via p47 is, in at least some cases, an important step, upstream of proteasome degradation [38,39,43]. Like the rhp23 dph1 double mutant, an shp1/ubx3 null mutant is synthetically lethal with null mutants in RPN10/pus1 [38,39]. Also, shp1/ubx3 null mutants display slowed turnover of certain proteasomal substrates, and increased levels of ubiquitinated proteins are found when p47 is knocked down with RNAi (RNA interference) . Hence, it appears that p47 can recruit ubiquitinated proteins to Cdc48 for both proteolytic and non-proteolytic purposes.
A null mutant in budding yeast Ubx2, another UBA/UBX domain protein, also displays retarded protein degradation . However, Sacch. cerevisiae Ubx5 null mutant, which corresponds to Schiz. pombe Ubx2 null mutant, does not display any obvious defects in the ubiquitin–proteasome pathway [38,39].
Human Npl4 is another ubiquitin-binding Cdc48 cofactor. It interacts with ubiquitin via an NZF domain , which is not conserved in yeast. The NZF domain is composed of four β-strands stabilized by a Zn2+ ion . Similar to the UBA domain, the NZF domain also interacts with the hydrophobic patch on ubiquitin. Although another NZF domain protein, Vps36 (vacuolar protein sorting 36), also binds ubiquitin, the NZF domain is not a general ubiquitin-binding domain, as the NZF domain from RanBP2 (Ran-binding protein 2) does not interact with ubiquitin .
If proteins that are to be secreted are unable to attain a proper structure within the ER, these proteins are transported back into the cytoplasm, where they are ubiquitinated and degraded. Together with another protein, Ufd1 (ubiquitin fusion degradation protein-1), Npl4 interacts with Cdc48, and the Cdc48–Ufd1–Npl4 complex is involved in this ERAD (ER-associated degradation) pathway , presumably pulling the proteins through a channel in the ER membrane. However, the Cdc48–Ufd1–Npl4 complex is involved not only in degrading ERAD substrates; certain membrane bound transcription factor precursors are also processed via the Cdc48–Ufd1–Npl4 complex. Recently, it was shown that the complex prevents excessive formation of ubiquitin chains and works upstream from, but in co-operation with, the UBL/UBA proteins Rad23/Rhp23 and Dsk2/Dph1 .
Ubiquitin binding in ubiquitination and deubiquitination
Although they bind ubiquitin, only a modest subset of ubiquitinating enzymes and DUBs contain recognizable ubiquitin-binding domains.
Gp78 and c-Cbl are ubiquitinating enzymes. The c-Cbl proto-oncoprotein is an E3 of the RING (really interesting new gene) finger family which recognizes activated RTKs (receptor tyrosine kinases), and terminates their signalling by promoting their destruction. Gp78 is an E3 enzyme in the ER membrane, where it is involved in the ERAD pathway . The gp78 enzyme contains a CUE domain near the C-terminus. As no mammalian orthologue of the yeast ERAD component called Cue1 has been found, gp78 perhaps also carries out the function of Cue1 in mammalian cells.
Usp5 (ubiquitin-specific protease 5), or isopeptidase T-1, is a DUB enzyme which cleaves Lys48-linked ubiquitin chains released from target proteins , but it also shows affinity for Lys29-linked chains and for the ubiquitin-like protein ISG (interferon-stimulated gene)-15 . Hence, isopeptidase T probably utilizes its UBA domains to grasp the ubiquitin chains prior to catalysis. In yeast, isopeptidase T is not essential, but in accordance with its role in ubiquitin recycling, a null mutant accumulates free ubiquitin chains . No other yeast DUBs contain UBA domains, so they must interact with their substrates in a different manner. In humans, a close homologue of isopeptidase T-1, Usp13 (isopeptidase T-3), also contains a UBA domain, but Usp13 largely remains to be functionally characterized.
Usp25 and Usp28 are two related human DUBs, which both contain UIM domains.
Ubiquitin-binding proteins in membrane events
Ubiquitin also works as a signal for endosomal sorting of many receptors. At least some targeted proteins carry ubiquitin chains, but monoubiquitination is often sufficient to promote lysosomal protein targeting. In many, but not all, cases, UIM-domain proteins both bind ubiquitin and promote their own mono- and poly-ubiquitination. However, UIM-dependent ubiquitination does not lead to degradation of the protein, and since the ubiquitination of UIM-domain proteins inhibits their interaction with other ubiquitinated proteins, it appears that their ubiquitination may be a regulatory mechanism [49,50].
During endocytosis of the EGFR (epidermal growth factor receptor), it becomes ubiquitinated by the UBA domain-containing c-Cbl E3 enzyme, enabling the protein Eps15 (EGFR pathway substrate 15) to recruit the EGFR to clathrin-coated pits . Epsins and Eps15 are UIM-domain proteins involved in endocytosis. Their role in this process is critical and derives from their capacity to bind ubiquitin and to undergo ubiquitination.
Generally, ubiquitin-tagged substrate proteins are sorted into the interior of MVBs (multi-vesicular bodies). These MVBs subsequently fuse with lysosomes, thereby delivering both hydrolytic enzymes and substrate proteins to the lysosome/vacuole. The biogenesis of MVBs requires the activity of Vps proteins, some of which are subunits of the ESCRT (endosomal sorting complexes required for transport) . The ESCRT complexes are recruited to the endosome membranes, where ubiquitinated proteins are sorted and vesicles are formed. Several of the proteins involved in these events are ubiquitin-binding proteins .
The ubiquitin signal is recognized by the protein Hrs (hepatocyte growth factor-regulated tyrosine kinase substrate; Vps27 in yeast), which also recruits the ESCRT complex to endosomes through an interaction with TSG101 (tumour susceptibility gene 101; Vps23 in yeast) . When Hrs/Vps27 is recruited to the endosome, it is in a complex with the protein STAM (signal transducing adaptor molecule; Hse1 in yeast). At the endosome, it is localized to clathrin-coated areas and interacts directly with clathrin. It is thought that Hrs/Vps27 serves to concentrate receptors modified with ubiquitin prior to inward vesiculation . Hrs, TSG101 and STAM are all ubiquitin-binding proteins, albeit utilizing different protein modules for the interaction. Hrs and STAM contain UIM domains, whereas TSG101 interacts with ubiquitin via a UEV domain.
Vps9 is a guanine nucleotide exchange factor for the yeast Rab5 orthologue Vps21  and is necessary for the delivery of proteins to the vacuole. Mis-sorting is seen when Vps9 is absent, and the CUE domain in Vps9 is required for its monoubiquitination . Vps9 binds ubiquitin via its CUE domains and this interaction is necessary for Vps9's endocytotic activity.
Interestingly, the human neuronal protein HSJ1 contains both a UIM domain and a DnaJ domain, the latter being a module present in cofactors of Hsp70 (heat-shock protein 70)-type chaperones. The role of the UIM domain in HSJ1 is unknown but HSJ1 has been linked to the processing of rhodopsin and in clathrin uncoating , which are both processes where ubiquitin plays a role. Hence it is possible that HSJ1 also functions in endocytosis, perhaps in analogy to another chaperone-binding protein, Swa2. The yeast protein Swa2 contains a C-terminal auxillin-type DnaJ domain, and, like auxillin, Swa2 is able to activate the ATPase activity of Hsc70 (heat-shock cognate 70 stress protein) and stimulate the uncoating of clathrin-coated vesicles . The structure of the Swa2 UBA domain has been determined, and it has been confirmed that Swa2 interacts with ubiquitin chains . This indicates either that Swa2 is a multifunctional protein involved in other processes related to ubiquitin, or that ubiquitin-binding is part of Swa2's function in clathrin-uncoating.
The human protein called Tom1 is another ubiquitin-binding protein. Tom1 binds ubiquitin via a GAT domain, which in a mutually exclusive manner also interacts with another ubiquitin-binding protein, called tollip (toll-interacting protein) . Tollip contains a C2-like domain, and a CUE domain. C2 domains are typically associated with various phospholipids. Tollip localizes to early endosomes and can recruit Tom1 , suggesting that Tom1 and tollip are involved in endosome trafficking of ubiquitinated proteins.
Other ubiquitin-binding proteins
The budding yeast protein Snf1 is a serine/threonine kinase, which mediates glucose repression of several genes and is involved in the cellular response to various forms of stress . Snf1 interacts with ubiquitin . However, the physiological significance of this interaction is presently unknown. The Snf1 orthologue in humans is the AMP-activated protein kinase, which is one of at least 13 related kinases, of which some contain UBA domains . The functional relevance of the UBA domains in these kinases is still unknown.
Other putative ubiquitin-binding proteins
NAC (nascent polypeptide-associated complex) α-subunit is an abundant UBA domain protein which interacts with nascent polypeptide chains emerging from the ribosome. NAC has been proposed to protect the nascent chains from premature interaction with other cell proteins . It is therefore attractive to speculate whether NAC is involved in directing nascent chains with folding defects to the 26 S proteasome. However, the presence of a UBA domain in archaeal orthologues of NAC makes this hypothesis less attractive, since archaea lack a ubiquitin system. Since this UBA domain seems more ancient than ubiquitin, its function is intriguing also from an evolutionary perspective. Recently, the structure of the NAC UBA domain was determined, revealing that the hydrophobic ubiquitin-binding area is structurally conserved in archaea .
The activation of the transcription factor NF-κB (nuclear factor κB) is highly complex and is regulated on several levels by the ubiquitin system. NF-κB activation requires a complex consisting of the kinase TAK1 [TGF (transforming growth factor)-β-activated kinase] and the regulatory proteins TAB1, TAB2 or TAB3 (TAK1-binding protein-1, -2 and -3 respectively). The activity of TAK1 is regulated by the E3 enzyme, TRAF (tumour-necrosis-factor-receptor-associated factor)-6. When TRAF-6 ubiquitinates TAK1 with a chain of Lys63-linked ubiquitin moieties, the kinase is activated . Structurally, Lys63-linked ubiquitin chains have a different conformation from Lys48-linked chains . The N-termini in both TAB2 and TAB3 contain CUE domains, whereas in their C-termini, they contain highly conserved Zn-finger domains. The TAB2 and TAB3 proteins interact with ubiquitinated TAK1. However, as with the NZF domain, the interactions appear to be mediated by the Zn-finger domains, while the CUE domains may co-operate with the Zn-finger domains in providing specificity for Lys63-linked ubiquitin chains . The ubiquitin-binding Zn-finger domains in TAB2 and TAB3 are essential for activation of TAK1, revealing a striking example of how the ubiquitin signal may be involved in signal transduction via a non-proteolytic mechanism.
Ubiquitin-binding proteins in disease
Paget's disease of bone is a disorder affecting about 3% of individuals in their mid-fifties. It is caused by mutation in the UBA domain of the p62 protein, also known as sequestosome 1 and SQSTM1. Among other functions, p62 regulates NF-κB signalling and localizes, along with ubiquitin, to intracellular protein aggregates typical of Alzheimer's and Parkinson's diseases . Interestingly, the N-terminus of p62 contains a PB1 (Phox and Bem1)-domain that has a ubiquitin-like fold. Hence, p62 may be somewhat analogous to the UBL/UBA domain proteins.
Aggregates are also found in MJD (Machado–Joseph disease) and Huntington's disease, where they contain molecular chaperones, ubiquitin and ubiquitin-binding proteins, including tollip and ubiquilin/Dsk2/Dph1 . MJD is caused by the expansion of a polyglutamine (polyQ) stretch in the protein ataxin-3. Ataxin-3 contains a Josephin domain with deubiquitinating activity , and two UIM domains besides the polyQ region . Ataxin-3 is alternatively spliced, and a splicing variant contains a third UIM domain at the C-terminus. Ataxin-3 is conserved in higher eukaryotes, including plants, but no orthologue is present in yeast.
The UIM domains in ataxin-3 bind ubiquitin chains, ubiquitinated proteins and Rad23/Rhp23 . Interestingly, ataxin-3 also binds to Cdc48 , and Cdc48 contributes to the formation of vacuoles in polyQ-associated disorders . Presumably, ataxin-3 plays an important role in the ubiquitin–proteasome system, but its substrates and its precise role remains to be determined.
This review is only a brief account of the increasing amount of literature on ubiquitin-binding proteins, and although the ubiquitin field is progressing at a rapid pace, it is clear that there are still fundamental problems to address.
Much remains to be learned about how poly- and mono-ubiquitin chains are recognized within the cell and how the cell distinguishes between the different types of chains. Biochemical studies have uncovered some detailed aspects of ubiquitin-binding proteins, but more physiological studies on the in vivo function of the different proteins are still needed.
The diversity of the ubiquitin-binding proteins indicates that the ubiquitin–proteasome pathway is more versatile than previously acknowledged, but also reveals how many other cellular functions may require and interplay with the ubiquitin signal.
Several families of ubiquitin-binding proteins have been characterized and their three-dimensional structures reveal that they are a heterogeneous group of proteins.
The UBL/UBA domain proteins are involved in targeting ubiquitinated substrates to the 26 S proteasome for destruction.
The UBA/UBX proteins are involved in targeting ubiquitinated proteins to Cdc48.
An array of ubiquitin-binding proteins is involved at several levels in the regulation of endocytosis and vesicle fusion.
Some ubiquitin-binding proteins are connected to hereditary human diseases: most notable are p62 and ataxin-3.
We thank Dr Colin Gordon, Dr Klavs B. Hendil and Dr Peter S. Walmod for critical comments on the manuscript and apologise to those whose work we were unable to include due to space constraints. R.H.-P. is supported by a grant from the Danish Natural Science Council.
- © 2005 The Biochemical Society