RING (really interesting new gene) and U-box E3 ligases bridge E2 ubiquitin-conjugating enzymes and substrates to enable the transfer of ubiquitin to a lysine residue on the substrate or to one of the seven lysine residues of ubiquitin for polyubiquitin chain elongation. Different polyubiquitin chains have different functions. Lys48-linked chains target proteins for proteasomal degradation, and Lys63-linked chains function in signal transduction, endocytosis and DNA repair. For this reason, chain topology must be tightly controlled. Using the U-box E3 ligase CHIP [C-terminus of the Hsc (heat-shock cognate) 70-interacting protein] and the RING E3 ligase TRAF6 (tumour-necrosis-factor-receptor-associated factor 6) with the E2s Ubc13 (ubiquitin-conjugating enzyme 13)–Uev1a (ubiquitin E2 variant 1a) and UbcH5a, in the present study we demonstrate that Ubc13–Uev1a supports the formation of free Lys63-linked polyubiquitin chains not attached to CHIP or TRAF6, whereas UbcH5a catalyses the formation of polyubiquitin chains linked to CHIP and TRAF6 that lack specificity for any lysine residue of ubiquitin. Therefore the abilities of these E2s to ubiquitinate a substrate and to elongate polyubiquitin chains of a specific topology appear to be mutually exclusive. Thus two different classes of E2 may be required to attach a polyubiquitin chain of a particular topology to a substrate: the properties of one E2 are designed to mono-ubiquitinate a substrate with no or little inherent specificity for an acceptor lysine residue, whereas the properties of the second E2 are tailored to the elongation of a polyubiquitin chain using a defined lysine residue of ubiquitin.
The post-translational modification of proteins with monoubiquitin or polyubiquitin chains has important functions in almost every aspect of cell biology . The best-studied example is the targeting of proteins for proteasomal degradation, which involves the formation of Lys48-linked polyubiquitin chains comprising at least four ubiquitins . More recently it has been appreciated that polyubiquitin chains can be formed using one of the other six lysine residues of ubiquitin. In yeast, all seven lysine residues of ubiquitin seem to be utilized for chain formation . Lys63-linked polyubiquitin chains play roles in DNA repair, signal transduction and endocytosis and do not appear to target proteins for proteasomal degradation. Thus polyubiquitin chains with different topologies determine different responses and the topology must be tightly regulated.
The canonical mechanism of polyubiquitin chain formation involves three steps. After an initial activation step catalysed by the E1, the ubiquitin is transferred in a second step to the E2 ubiquitin-conjugating enzyme. The third step is dependent on the nature of the E3 ubiquitin ligase: RING (really interesting new gene), PHD (plant homeodomain) and U-box E3 ubiquitin ligases bridge E2 and substrate to enable ubiquitination to occur, whereas HECT (homologous to E6-AP C-terminus) E3s form a thiol ester with the ubiquitin before transferring it to the substrate. Target protein selectivity is provided by the E3, whereas the E2 determines the specificity for the lysine residue of the acceptor ubiquitin in polyubiquitin chain formation mediated by RING, PHD and U-box E3s, but probably not by HECT E3s .
E2s that direct specificity for Lys48-linked chains that mark proteins for degradation via the proteasome are for example human Cdc34 (cell division cycle 34) or E2-25K, while the formation of Lys63-linked chains requires a heterodimer of Ubc13 (ubiquitin-conjugating enzyme 13) and either of the Ubc-variants Uev1a (ubiquitin E2 variant 1a) or Mms2 (methyl methanesulfonate sensitivity 2) . Nevertheless, despite more than two decades of research in the field of protein ubiquitination, the details of polyubiquitin chain formation remain elusive . The addition of a polyubiquitin chain of a certain topology involves two steps with very different requirements: substrate (mono-)ubiquitination and chain elongation (polyubiquitination). Whereas mono-ubiquitination seems to lack an inherent specificity for a particular lysine residue on a substrate, chain elongation can occur specifically on a particular lysine residue of ubiquitin, for example Lys48 or Lys63. Two models have been suggested to explain how different E3–E2 complexes operate. The first model suggests that an E3–E2 complex that can elongate Lys48-linked polyubiquitin chains efficiently is also able to add the first ubiquitin to the substrate, albeit more slowly, making this the rate-limiting step . A modified version of this model was also suggested for the TRAF6 (tumour-necrosis-factor-receptor-associated factor 6)–Ubc13–Uev1a complex, in which the first ubiquitin was added to TRAF6 in a Ubc13-dependent, but Uev1a-independent reaction, whereas the elongation of Lys63-linked polyubiquitin chains occurs in a Ubc13–Uev1a-dependent manner . The other model suggests that one E3–E2 pair adds the first ubiquitin and another E3–E2 pair elongates the Lys48- or Lys63-linked polyubiquitin chains [8–10].
Recently, we identified CHIP [C-terminus of the Hsc (heat-shock cognate) 70-interacting protein] as a binding partner for the Lys63-specific ubiquitin-conjugating enzyme Ubc13–Uev1a, and demonstrated that it functions as an E3 ubiquitin ligase with Ubc13–Uev1a, and determined the structure of a heterotrimeric complex between Ubc13–Uev1a and the CHIP U-box E3 ligase domain . CHIP is a dimeric ∼35 kDa protein containing a C-terminal U-box domain and an N-terminal domain with three TPRs (tetratricopeptide repeats) that mediates interaction with Hsp (heat-shock protein) 90 and Hsp70 chaperones possessing the C-terminal EEVD-TPR-binding motif . In vitro, we showed that CHIP formed unanchored Lys63-linked polyubiquitin chains using Ubc13–Uev1a as the E2 . We also demonstrated that CHIP forms polyubiquitin chains in vitro that are neither specific for Lys63 or Lys48 when paired with UbcH5a, another E2-conjugating enzyme and that these chains are linked to CHIP . In the present study, we show that the CHIP–UbcH5a complex lacks specificity for any of the seven lysine residues of ubiquitin, but is able to attach the first ubiquitin to CHIP or the CHIP-interacting protein Hsp90. In contrast, the CHIP–Ubc13–Uev1a complex synthesizes only unanchored Lys63-linked polyubiquitin chains that are not attached to CHIP or Hsp90. Similar results were obtained using TRAF6 as the E3 ligase. These findings support a model in which the addition of a polyubiquitin chain of a specific topology will frequently require at least two E3–E2 complexes: one for the addition of the first ubiquitin, the second for elongation with a polyubiquitin chain of a specific topology.
An antibody that recognises CHIP was obtained from Abcam, the anti-ubiquitin antibody was obtained from DakoCytomation, the anti-TRAF6 antibody was from Santa Cruz and the Coomassie Blue staining kit was from Invitrogen. The E2 protein UbcH5a was purchased from Biomol. Wild-type ubiquitin was purchased from Sigma, and methylated ubiquitin was from Calbiochem.
The constructs encoding Ubc13, Uev1a, UBE1 (ubiquitin-activating enzyme 1), wild-type ubiquitin and CHIP have been described previously . Hsp90 (NCBI AAQ63401) was amplified from pRSETa His–Hsp90, a gift from Professor Laurence Pearl (Institute for Cancer Research, London, U.K.) using KOD Hot Start DNA Polymerase (Novagen) and was then cloned into pCR2.1 (Invitrogen) and sequenced to completion. It was then digested with Not1 and ligated into the same site in pGEX6P-2 to produce pGEX6P-2 Hsp90. UbcH4 was amplified from IMAGE 3835609 using KOD Hot Start Polymerase (Novagen), cloned into pCR2.1 (Invitrogen) and sequenced to completion. The insert was then cloned into the Not1 site of pET28b (Novagen) to make pET28 UbcH4. Ubiquitin was RT (reverse transcriptase)–PCR-amplified from peripheral blood total RNA using the Access RT–PCR System (Promega). It was cloned into pCR2.1, sequenced and subcloned into the BamH1 site of pGEX6P-1. Each of the mutations was made using the Stratagene QuikChange® method but using KOD Hot Start Polymerase. TRAF6 was amplified from IMAGE 5210798 and cloned as described above into pCR2.1 and then into the BamH1/Not1 sites of pGEX6P-1 to form pGEX6P-1 TRAF6.
Protein expression and purification
A histidine-tagged UBE1 was expressed in insect Sf21 cells and purified by chromatography on Ni-NTA (Ni2+-nitrilotriacetate) agarose. His-tagged CHIP were expressed in bacteria and also purified by chromatography on Ni-NTA–agarose. Ubc13, Uev1a, CHIP, TRAF6, Hsp90, ubiquitin and lysine mutants of ubiquitin were expressed in bacteria with a GST (glutathione transferase) tag at their N-terminus followed by a PreScission protease cleavage site. Each protein was purified from bacterial extracts by affinity chromatography on glutathione–Sepharose (Amersham). To remove the GST tags, the glutathione–Sepharose containing the bound GST-fusion was incubated with PreScission protease to release the cleaved proteins. Each protein was then dialysed against 50 mM Tris/HCl (pH 7.5), 270 mM sucrose, 150 mM NaCl, 0.1 mM EGTA, 0.1% (v/v) 2-mercaptoethanol, 0.2 mM PMSF and 1 mM benzamidine and stored in aliquots at −80 °C.
The formation of polyubiquitin chains was carried out at 30 °C for 1 h in a 20 μl reaction mixture containing His6(hexahistidine)–E1 (50 nM), E2 ubiquitin-conjugating enzyme (1 μM), ubiquitin (100 μM), CHIP or TRAF6 (1 μM or as indicated in the Figure legends), 50 mM Tris/HCl (pH 7.5), 5 mM MgCl2 and 2 mM ATP. The reaction was stopped by addition of SDS sample buffer.
CHIP–Ubc13–Uev1a and CHIP–UbcH5a complexes produce unanchored and anchored polyubiquitin chains respectively
We have reported previously that CHIP can catalyse the formation of polyubiquitin chains using the different E2 ubiquitin-conjugating enzymes Ubc13–Uev1a and UbcH5a in vitro ( and Figure 1). Interestingly, CHIP–Ubc13–Uev1a formed free polyubiquitin chains not attached covalently to any protein present in the incubation mixture, as shown by the formation of ubiquitin oligomers smaller than CHIP (Figure 1A, upper panel) and by the lack of appearance of slower-migrating forms of CHIP during the ubiquitination reaction (Figure 1A, lower panel). In contrast, CHIP–UbcH5a catalysed the auto-ubiquitination of CHIP, as shown by the lack of any ubiquitin oligomers smaller than CHIP (Figure 1B, upper panel) and by the appearance of a ladder of ubiquitinated CHIP species differing in the number of ubiquitin molecules attached (Figure 1B, lower panel). Although highly polyubiquitinated chains were the major species detected by immunoblotting with anti-ubiquitin antibody (Figure 1B, upper panel), the major CHIP species formed was mono-ubiquitinated as shown by immunoblotting with anti-CHIP antibody (Figure 1B, lower panel).
The E3 ubiquitin ligase CHIP catalyses the formation of polyubiquitin chains using the E2 ubiquitin-conjugating enzymes Ubc13–Uev1a and UbcH5a
CHIP–Ubc13–Uev1a catalyses the formation of Lys63-linked polyubiquitin chains whereas the polyubiquitin chains created by the CHIP–UbcH5a complex exhibit no specific topology
The complex formed between CHIP and the E2 ubiquitin-conjugating enzyme Ubc13–Uev1a is known to direct the formation of Lys63-linked polyubiquitin chains. To determine which type of polyubiquitin chain was formed by the CHIP–UbcH5a complex, we used a range of ubiquitin mutants with one of the seven lysine residues mutated to an arginine residue. Surprisingly, all seven mutants were utilized in the auto-polyubiquitination of CHIP, suggesting that the polyubiquitin chains made by CHIP–UbcH5a have no specific topology (Figure 2B). The K6R mutant was used less efficiently than the other mutants for both mono- and polyubiquitination, but was still able to support polyubiquitination at a significant level (Figure 2B, lane 2). The polyubiquitin chains formed by CHIP–UbcH5a and wild-type ubiquitin were identified directly by MS, which revealed that they were linked via at least five different lysine residues, namely Lys6, Lys11, Lys33, Lys48 and Lys63 (E. Carrick, M. Windheim and N. Morrice, unpublished work). In contrast, polyubiquitin chains formed by CHIP–Ubc13–Uev1a are very specific for Lys63, since only the mutation of this lysine residue to arginine abolished the formation of polyubiquitin chains (Figure 2A, lane 8).
The CHIP–Ubc13–Uev1a complex catalyses the formation of unanchored Lys63-linked polyubiquitin chains, whereas CHIP–UbcH5a attaches polyubiquitin chains without any specific topology to CHIP
When wild-type ubiquitin was replaced by methylated ubiquitin in which all seven lysine residues were modified, incubation with CHIP and Ubc13–Uev1a did not result in the formation of any polyubiquitin chains or multi-ubiquitinated CHIP (Figure 2C, compare lane 2 with lane 3). In contrast, incubation of CHIP, UbcH5a and methylated ubiquitin did not produce the high-molecular-mass forms of polyubiquitinated CHIP seen with wild-type ubiquitin (Figure 2C, compare lane 5 with lane 6), and instead mono-ubiquitinated CHIP was formed with smaller amounts of a di-ubiquitinated derivative (Figure 2C, bottom panel, lane 6). This latter result suggested that CHIP was mainly ubiquitinated at a single lysine residue, which has been reported to be Lys22, in the presence of UbcH5a . Our group confirmed this result (E. Carrick, M. Windheim and N. Morrice, unpublished work) and also identified Lys221 and Lys255 as sites of polyubiquitination, consistent with the presence of small amounts of di-ubiquitinated CHIP (Figure 2C, bottom panel, lane 6).
Deletion of the N-terminal TPR domain of CHIP suppresses ubiquitination by the CHIP–UbcH5a complex, without affecting ubiquitination by the CHIP–Ubc13–Uev1a complex
We showed recently that CHIP dimerization involves the helical hairpin domain and the U-box . We found that the U-box alone had ubiquitination activity using Ubc13–Uev1a as an E2 (Figure 3, lanes 4 and 14), demonstrating that the U-box is the minimum structural domain needed to support polyubiquitination. Nevertheless, a CHIP mutant consisting of the U-box and the helical hairpin domain, but lacking the TPR domain, had a significantly higher activity, which was indistinguishable from that of the full-length protein (Figure 3, lanes 5 and 15 compared with lanes 2 and 12 respectively). On the other hand, the CHIP U-box alone had no detectable ubiquitination activity using UbcH5a as an E2 (Figure 3, lanes 9 and 19) and, even with the CHIP mutant containing the U-box and the helical hairpin domain, the ubiquitination activity was far lower than with the full-length protein (Figure 3, lanes 10 and 20 compared with lanes 7 and 17 respectively). CHIP is ubiquitinated at Lys22 in the presence of UbcH5a  and this lysine residue is missing in the CHIP mutant containing the U-box and the helical hairpin domain only. The UbcH5a-catalysed ubiquitination is greatly reduced in this CHIP mutant compared with the full-length protein, because it lacks the lysine residue required to anchor the polyubiquitin chains on CHIP. The N-terminal TPR-domain has no effect on ubiquitination mediated by Ubc13–Uev1a, because this E2 does not require Lys22 to anchor the polyubiquitin chains and forms unanchored polyubiquitin chains.
Truncated forms of CHIP lacking the TPR domain are as active as full-length CHIP when paired with Ubc13–Uev1a, but far less active with UbcH5a
CHIP–UbcH5a but not CHIP–Ubc13–Uev1a ubiquitinates Hsp90 in vitro
CHIP is known to bind to Hsp90 via its N-terminal TPR domain . In order to find out whether Hsp90 binding has an effect on the ubiquitination activity of CHIP, we included recombinant Hsp90 in the ubiquitination assays. Interestingly, Hsp90 had no significant effect on the CHIP–Ubc13–Uev1a-dependent ubiquitination (Figure 4, lanes 1–4 and 9–12), but was used as a substrate for UbcH5a-dependent ubiquitination very efficiently (Figure 4, lanes 5–8 and 13–16). Strikingly, almost all of the Hsp90 included in the assay mixture became ubiquitinated. As observed before, the CHIP–Ubc13–Uev1a complex only formed unanchored polyubiquitin chains, whereas the CHIP–UbcH5a complex formed polyubiquitin chains anchored either to CHIP or to Hsp90 (Figure 4, right-hand panels).
Hsp90 is a substrate for CHIP–UbcH5a, but not for CHIP–Ubc13–Uev1a in vitro
The RING domain E3 ligase TRAF6 and the U-box domain E3 ligase CHIP have similar activities when paired with the E2 ubiquitin-conjugating enzymes UbcH5a and Ubc13–Uev1a
The results presented so far show that the E2 ubiquitin-conjugating enzymes Ubc13–Uev1a and UbcH5a exhibit strikingly different characteristics when paired with the E3 ubiquitin ligase CHIP. To find out whether these findings were specific to CHIP, we used the E3 ubiquitin ligase TRAF6 and investigated its ubiquitinating activity when paired with Ubc13–Uev1a or UbcH5a. As shown with CHIP, Ubc13–Uev1a-dependent ubiquitination using TRAF6 resulted in the formation of unanchored Lys63-linked polyubiquitin chains, while the TRAF6–UbcH5a complex catalysed TRAF6 auto-ubiquitination (Figure 5). Thus the characteristics of the E2 ubiquitin-conjugating complexes Ubc13–Uev1a and UbcH5a do not depend on the E3 ubiquitin ligase that they are paired with, but are intrinsic to the E2 ubiquitin-conjugating complexes. In addition, we tested CHIP and TRAF6 ubiquitinating activity with another E2 ubiquitin-conjugating enzyme UbcH4. UbcH4 showed the same characteristics as UbcH5a and was able to support the auto-ubiquitination of CHIP and TRAF6 (M. Windheim, unpublished work).
The E3 ubiquitin ligase TRAF6 acts similarly to CHIP when complexed to the E2 ubiquitin-conjugating enzymes Ubc13–Uev1a and UbcH5a
Recently, we  and others  have reported that CHIP is able to form unanchored Lys63-linked polyubiquitin chains with the E2 ubiquitin-conjugating enzyme Ubc13–Uev1a and that the CHIP–UbcH5a complex forms polyubiquitin chains that are anchored to CHIP (Figure 1 and ) or luciferase . Moreover the polyubiquitin chains that are attached to CHIP (Figure 2) or luciferase  have no specific topology. CHIP is a U-box-containing E3 ligase and fulfills its function by bringing the E2 and the substrate into close proximity to allow ubiquitination to occur. Ubiquitination of CHIP by UbcH5a depends on an acceptor lysine residue on CHIP, which is predominantly Lys22in vitro . As a result, the deletion of the N-terminal region of CHIP including the TPR domain, drastically decreased CHIP–UbcH5a-mediated ubiquitination, but had no effect on CHIP–Ubc13–Uev1a-mediated ubiquitination (Figure 3). The CHIP–UbcH5a complex also ubiquitinated CHIP-associated Hsp90 in vitro (Figure 4), but Hsp90 had no effect on CHIP–Ubc13–Uev1a-mediated polyubiquitination (Figure 4).
It is important to point out that the formation of a substrate-linked polyubiquitin chain of a certain topology involves two steps with very different requirements. For the attachment of the first ubiquitin there seems to be little inherent specificity for any particular lysine residue on the substrate provided that it can be accessed by the relevant E2–E3 complex. However, the interaction between the E3 ubiquitin ligase and its substrate may sometimes determine which particular lysine residue(s) becomes ubiquitinated. For example the ubiquitination of IκBα (inhibitor of nuclear factor κBα) occurs at Lys21 and Lys22 because the SCF E3 ligase complex binds specifically to IκBα phosphorylated at Ser32 and Ser36 in such a way that presumably only these two lysine residues can be accessed . In contrast, the elongation of a polyubiquitin chain with a particular topology is a very specific reaction that must ensure that only one unique lysine residue on ubiquitin and no other on either ubiquitin or any substrate is ubiquitinated. Interestingly, the characteristics of UbcH5a and Ubc13–Uev1a, when paired with CHIP or TRAF6 in vitro, exactly match the different requirements that one would predict for the E2s in a two-step model of ubiquitination. UbcH5a has no inherent specificity and can ubiquitinate almost any lysine residue on a substrate if brought into close proximity. This explains why CHIP is auto-ubiquitinated in the CHIP–UbcH5a complex and why Hsp90 is ubiquitinated in the Hsp90–CHIP–UbcH5a complex. The lack of specificity of UbcH5a also explains why the polyubiquitin chains it forms have no specific topology (Figure 2), because UbcH5a does not differentiate ubiquitin from any other substrate. Therefore, after the mono-ubiquitination of a substrate such as CHIP in vitro, the CHIP–ubiquitin conjugate that is formed becomes a substrate for UbcH5a and ubiquitin can be transferred from ubiquitin-charged UbcH5a to any or all of the lysine residues of ubiquitin that come into close proximity (Figure 2 and ). Therefore it is important to distinguish this reaction from the one catalysed by Ubc13–Uev1a, which is able to recognize ubiquitin specifically, but cannot ubiquitinate any other protein substrate. An important implication of this two-step model of protein ubiquitination is that after the mono-ubiquitination of the substrate, the polyubiquitination has to be performed by the same E3 using a different E2 or by a completely different E3–E2 complex (Figure 6). Both models are conceivable, since the binding of E1 and E3 to the E2 is mutually exclusive , and the E3 and E2 have to dissociate after each ubiquitination cycle.
Model for substrate polyubiquitination by RING and U-box E3 ligases
The lack of specificity of the CHIP–UbcH5a complex determined in the present study is consistent with the results of another study  which analysed the polyubiquitin chains formed by UbcH5 paired with three different E3s: CHIP, MuRF1 (muscle-specific RING finger 1) and Mdm2 (murine double minute 2) . The authors not only found by MS that all seven lysine residues of ubiquitin were utilized for chain formation but also identified ‘forked chains’ with two ubiquitin molecules linked to adjacent lysine residues of another ubiquitin. This unexpected result can be explained very easily by the model in the present study, since E3–UbcH5 complexes lack inherent specificity and any lysine residue on ubiquitin which is attached to a substrate can become ubiquitinated if it gets close enough to the ubiquitin-charged UbcH5. Provided that steric constraints do not prohibit it, there seems no reason why ubiquitination should not take place at any lysine residue, or even on more than one lysine residue of the same ubiquitin molecule, resulting in the formation of ‘forked chains’.
The present findings and those of Kim et al.  contrast with two reports that the RING E3 ligase BRCA1 (breast cancer 1, early onset)–BARD1 (BRCA1-associated RING domain 1) mainly forms Lys6-linked polyubiquitin chains with UbcH5c [17,18]. In our experiments the polyubiquitin chains that were formed with CHIP–UbcH5a were linked via almost every lysine residue of ubiquitin, as shown by MS, but the Lys6 mutant of ubiquitin was not used as efficiently as wild-type ubiquitin, as shown not only by decreased formation of polyubiquitin, but also decreased mono-ubiquitination of CHIP (Figure 2), which should be unaffected by the Lys6 mutation. Therefore reliance on the use of this mutant in the absence of MS data would have been misleading and would lead to the erroneous conclusion that the ubiquitin linkages formed were mainly via Lys6. Therefore either the BRCA1–BARD1 complex is able to restrict the specificity of UbcH5c-mediated polyubiquitin chain formation, perhaps for steric reasons, or the BRCA1–BARD1–UbcH5-mediated polyubiquitin chain formation is not as specific as previously suggested. Ubiquitination reactions mediated by CHIP and TRAF6, when paired with Ubc13–Uev1a or UbcH5a, were very similar suggesting that the observed differences between Ubc13–Uev1a and UbcH5a are not dependent on the E3 (Figure 5). We have recently obtained similar results with another E3 ligase, Pellino 1, which produced free Lys63-polyubiquitin chains when paired with Ubc13–Uev1a but polyubiquitinated Pellino or its substrate IRAK1 (interleukin-1-receptor-associated kinase 1) in the presence of UbcH5a, these chains being linked mainly via Lys11, Lys48 and Lys63 .
In another recent publication a different model for TRAF6–Ubc13–Uev1a auto-ubiquitination was proposed . In accordance with our results the authors found that TRAF6 stimulates Ubc13–Uev1a-dependent formation of unanchored Lys63-linked polyubiquitin chains and UbcH5c-dependent TRAF6 auto-ubiquitination. However, they also found that the TRAF6–Ubc13 complex links ubiquitin to TRAF6 in vitro in the absence of Uev1a, a reaction that is prevented in the presence of Uev1a. They therefore proposed a model in which the first ubiquitin is added by the TRAF6–Ubc13 complex whereas the TRAF6–Ubc13–Uev1a complex subsequently adds the Lys63-linked polyubiquitin chains. This model requires that there is either an excess of free Ubc13 in cells or that the rather stable Ubc13–Uev1a complex is capable of dissociating in vivo under particular conditions, neither of which has been established. In addition, the E3 ligase would have to differentiate between the monomeric Ubc13 and the heterodimeric Ubc13–Uev1a complex to switch from substrate ubiquitination to Lys63-linked polyubiquitination, which would be difficult, given the fact that, for example, in the CHIP–Ubc13–Uev1a complex the U-box of CHIP is only in contact with Ubc13 . Moreover, this model overlooks the possibility that another E2 or another E3–E2 complex could be involved in catalysing the initial mono-ubiquitination reaction and several other publications support the two-step model of ubiquitination illustrated in Figure 6. There are two reports which indicate that Ubc13-dependent Lys63-polyubiquitination is preceded by a Ubc13-independent mono-ubiquitination. Genetic evidence from yeast showed that PCNA (proliferating-cell nuclear antigen) is mono-ubiquitinated by RAD6 (E2) and RAD18 (E3) and then Lys63-polyubiquitinated by RAD5 (E3) and Ubc13–Mms2 (E2) . Similarly, the Lys63-linked polyubiquitination of MHC class I molecules in HeLa cells mediated by the K3 gene product of KSHV (Kaposi's-sarcoma-associated herpes virus) involves UbcH5b/c-mediated mono-ubiquitination followed by Ubc13-dependent polyubiquitination . Furthermore, in the case of Lys48-linked polyubiquitination, a two-step model was proposed for the APC (anaphase promoting complex) in yeast with Ubc4 promoting mono-ubiquitination and Ubc1 (the yeast homologue of the human E2-25K protein) promoting chain extension .
Structural considerations also support the two-step model of substrate polyubiquitination. Structural analysis of the Ubc13–Mms2 complex revealed the key role of Mms2 in positioning the acceptor ubiquitin to the Ubc13 active site to allow the modification of Lys63 to occur specifically . Therefore the Ubc13–Mms2 complex provides a structural explanation for the ability of this E2 to form Lys63-linked polyubiquitin chains specifically. On the other hand, this characteristic makes it unlikely, if not impossible, for this complex to be engaged in the non-specific substrate mono-ubiquitination. There are reports in the literature that E2s specific for a particular lysine residue of ubiquitin can also ubiquitinate substrates in vitro, e.g. with TRAF6–Ubc13–Uev1a  or CHIP–Ubc13–Uev1a , but this substrate ubiquitination is negligible compared with the formation of unanchored polyubiquitin chains.
Unlike Ubc13–Uev1a, UbcH5a lacks inherent specificity for any particular lysine residue of ubiquitin and makes this E2 a candidate for substrate mono-ubiquitination. We therefore propose that there are different classes of E2s, some that are involved in substrate mono-ubiquitination, e.g. the E2s of the UbcH5 family or UbcH4, and others that are able to form polyubiquitin chains of a specific topology, e.g. human Cdc34 and E2-25K for Lys48-linked chains and Ubc13–Uev1a (Mms2) for Lys63-linked chains. On the basis of this model we predict that E2s exist that are specific for other lysine residues of ubiquitin and have yet to be identified. During the preparation of the present manuscript, a publication appeared which suggested a similar model to that proposed in the present study based on results obtained with the RING domain E3 ligase BRCA1–BARD1 and several E2s .
A major question arising from the present study is how the eukaryotic cell is able to switch from the mono-ubiquitination of the substrate to the elongation of the polyubiquitin chain in such a way as to avoid mixed chain formation or whether under some circumstances mixed chains and ‘forked chains’ are formed in vivo and fulfil a particular function that has yet to be identified. Analysis of the complexity of polyubiquitinated proteins in vivo by MS will be key to a better understanding of these processes.
We thank Professor Laurence Pearl (Institute of Cancer Research, London, U.K.) for the construct encoding Hsp90 and Emma Carrick and Nick Morrice (MRC Protein Phosphorylation Unit, University of Dundee) for the MS analysis. We are grateful to the protein and antibody production teams of the Division of Signal Transduction Therapy, University of Dundee (co-ordinated by Hilary McLaughlan and James Hastie) for His6–E1, His6–UbcH4 and GST–TRAF6 and the DNA Sequencing Service, University of Dundee (www.dnaseq.co.uk). M. W. acknowledges a postdoctoral position from EU Research Training Network Framework 5. This work was supported by the UK Medical Research Council and The Royal Society.
breast cancer 1, early onset
BRCA1-associated RING domain 1
cell division cycle 34
C-terminus of the Hsc (heat-shock cognate) 70-interacting protein
homologous to E6-AP C-terminus
inhibitor of nuclear factor κB α
methyl methanesulfonate sensitivity 2
really interesting new gene
tumour-necrosis-factor-receptor-associated factor 6
ubiquitin-conjugating enzyme 13
ubiquitin-activating enzyme 1
ubiquitin E2 variant 1a