Title: Mechanistic analysis of PCNA poly-ubiquitylation by the ubiquitin protein ligases Rad18 and Rad5
Abstract: Article22 October 2009Open Access Mechanistic analysis of PCNA poly-ubiquitylation by the ubiquitin protein ligases Rad18 and Rad5 Joanne L Parker Joanne L Parker Clare Hall Laboratories, Cancer Research UK London Research Institute, Blanche Lane, South Mimms, UK Search for more papers by this author Helle D Ulrich Corresponding Author Helle D Ulrich Clare Hall Laboratories, Cancer Research UK London Research Institute, Blanche Lane, South Mimms, UK Search for more papers by this author Joanne L Parker Joanne L Parker Clare Hall Laboratories, Cancer Research UK London Research Institute, Blanche Lane, South Mimms, UK Search for more papers by this author Helle D Ulrich Corresponding Author Helle D Ulrich Clare Hall Laboratories, Cancer Research UK London Research Institute, Blanche Lane, South Mimms, UK Search for more papers by this author Author Information Joanne L Parker1 and Helle D Ulrich 1 1Clare Hall Laboratories, Cancer Research UK London Research Institute, Blanche Lane, South Mimms, UK *Corresponding author. Clare Hall Laboratories, Cancer Research UK London Research Institute, Blanche Lane, South Mimms, Herts EN6 3LD, UK. Tel.: +44 1707 62 5821; Fax: +44 1707 62 5750, E-mail: [email protected] The EMBO Journal (2009)28:3657-3666https://doi.org/10.1038/emboj.2009.303 PDFDownload PDF of article text and main figures. ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InMendeleyWechatReddit Figures & Info Poly-ubiquitylation is a common post-translational modification that can impart various functions to a target protein. Several distinct mechanisms have been reported for the assembly of poly-ubiquitin chains, involving either stepwise transfer of ubiquitin monomers or attachment of a preformed poly-ubiquitin chain and requiring either a single pair of ubiquitin-conjugating enzyme (E2) and ubiquitin ligase (E3), or alternatively combinations of different E2s and E3s. We have analysed the mechanism of poly-ubiquitylation of the replication clamp PCNA by two cooperating E2–E3 pairs, Rad6–Rad18 and Ubc13–Mms2–Rad5. We find that the two complexes act sequentially and independently in chain initiation and stepwise elongation, respectively. While loading of PCNA onto DNA is essential for recognition by Rad6–Rad18, chain extension by Ubc13–Mms2–Rad5 is only slightly enhanced by loading. Moreover, in contrast to initiation, chain extension is tolerant to variations in the attachment site of the proximal ubiquitin moiety. Our results provide information about a unique conjugation mechanism that appears to be specialised for a regulatable pattern of dual modification. Introduction Like other post-translational modifiers, the small, highly conserved protein ubiquitin mediates its biological functions by reversibly altering the properties of its targets (Ciechanover et al, 2000; Hochstrasser, 2000). However, unlike simple modifications such as methylation or acetylation, ubiquitylation has the potential for a widely expanded range of signalling by means of its ability to form polymeric chains (Pickart and Fushman, 2004; Ikeda and Dikic, 2008). Despite the significance of poly-ubiquitylation for cellular regulation and the abundance of poly-ubiquitylated substrates, surprisingly little is known about the mechanism of ubiquitin chain assembly (Hochstrasser, 2006). Both mono- and poly-ubiquitylation are mediated by a cascade of enzymes, involving activation of ubiquitin as a high-energy thioester intermediate by an activating enzyme (E1), transfer of the ubiquitin thioester to a conjugating enzyme (E2) and attachment of ubiquitin's carboxyl (C)-terminus to a lysine residue in the target protein with the help of a ubiquitin protein ligase (E3) (Kerscher et al, 2006). Whereas in the case of mono-ubiquitylation, E3 is generally responsible for selecting one or more lysines directly on the substrate for ubiquitin conjugation, poly-ubiquitylation requires formation of at least two distinct types of linkage, that between the substrate and the proximal ubiquitin and those between the individual ubiquitin moieties within the chain. Mechanistic studies of selected conjugation factors have shown that this task can be accomplished in several distinct ways. As exemplified by the action of the ubiquitin ligase SCFCdc4 with the E2 Cdc34 on the cyclin inhibitor Sic1 (Petroski and Deshaies, 2005) or the human anaphase-promoting complex (APC) with the E2 UbcH10 on securin or cyclin-B1 (Jin et al, 2008), a single E2–E3 pair is able to mediate both initiation and elongation of the ubiquitin chain. In contrast, budding-yeast APC or the virally encoded ligase K3 sequentially cooperate with two distinct E2s for mono-ubiquitylation and chain extension (Duncan et al, 2006; Rodrigo-Brenni and Morgan, 2007). Whereas in these examples the poly-ubiquitin chain is presumably assembled in a stepwise manner, some E2s, such as the human Ube2g2 or yeast Ubc7, are capable of preforming poly-ubiquitin chains on their active-site cysteine (Li et al, 2007; Ravid and Hochstrasser, 2007). Ube2g2 then transfers these en bloc onto a substrate protein. In a variation of this mechanism, the conjugating enzyme E2-25K is able to use unanchored poly-ubiquitin chains activated by E1 (Piotrowski et al, 1997). Modification of the eukaryotic replication clamp PCNA in the context of DNA damage tolerance is as of now the only reported case where separate E2–E3 pairs appear to be responsible for mono- and poly-ubiquitylation of a common substrate (Moldovan et al, 2007; Ulrich, 2009). In response to DNA-damaging agents, PCNA is mono-ubiquitylated at a single, conserved lysine, K164, by the E2 Rad6 in complex with the RING-finger E3 Rad18, whereas poly-ubiquitylation at the same site additionally requires the heterodimeric E2 Ubc13–Mms2 and a second RING-finger E3, Rad5 (Hoege et al, 2002). As a consequence, rad5, ubc13 and mms2 mutants can mono-, but not poly-ubiquitylate, PCNA, whereas in rad6 and rad18 mutants, ubiquitylation is completely abolished. The two modifications label PCNA for alternative functions: mono-ubiquitylation activates translesion synthesis through damage-tolerant DNA polymerases (Stelter and Ulrich, 2003; Kannouche et al, 2004; Watanabe et al, 2004), and poly-ubiquitylation is required for an error-free pathway of damage avoidance possibly involving a template switch (Hoege et al, 2002; Zhang and Lawrence, 2005). Poly-ubiquitin chain assembly by Ubc13–Mms2 has been studied in detail (Hofmann and Pickart, 1999, 2001; McKenna et al, 2001, 2003; Moraes et al, 2001; VanDemark et al, 2001; Eddins et al, 2006; Yin et al, 2009). The Ubc13–Mms2 complex is unusual among E2 enzymes in that it polymerises ubiquitin exclusively through lysine 63 (Hofmann and Pickart, 1999), its specificity dictated by a ubiquitin-binding site within the Mms2 subunit (Moraes et al, 2001; VanDemark et al, 2001; Eddins et al, 2006). Moreover, it is particularly active at catalysing the synthesis of free, unanchored chains (Hofmann and Pickart, 1999, 2001). In contrast, the mechanism of cooperation between the two E2–E3 pairs with PCNA as a substrate has not been addressed. In addition to interacting with their cognate E2s, both Rad18 and Rad5 interact with PCNA, with each other and with themselves (Bailly et al, 1994; Ulrich and Jentsch, 2000; Hoege et al, 2002). These interrelations suggest several alternative models of how the enzymes may cooperate in PCNA poly-ubiquitylation. The models, shown schematically in Figure 1, differ with respect to the questions of whether the two E2–E3 pairs act sequentially (A, B) or in concert (C, D), and whether the ubiquitin moieties are added in a stepwise manner (A, C) or transferred en bloc to PCNA (B, D). We have now reconstituted the poly-ubiquitylation of budding-yeast PCNA with purified components in order to differentiate between these models. In addition, we have analysed the relevance of DNA and the ubiquitin attachment site for the process. Our results lend support to model A where Rad6–Rad18 and Ubc13–Mms2–Rad5 act sequentially and mediate PCNA poly-ubiquitylation by stepwise addition of ubiquitin monomers. Figure 1.Alternative models for the mechanism of PCNA poly-ubiquitylation by Rad6–Rad18 and Ubc13–Mms2–Rad5. (A) Sequential action, stepwise assembly: Rad6–Rad18 and Ubc13–Mms2–Rad5 act independently and sequentially, each attaching the ubiquitin moieties in a stepwise manner. After conjugation of the first ubiquitin, Rad6–Rad18 is no longer required, and Rad5 recognises the mono-ubiquitylated PCNA as a substrate for chain elongation through K63. (B) Sequential action, preformed chains: Rad6–Rad18 and Ubc13–Mms2–Rad5 act independently and sequentially. A K63-linked poly-ubiquitin chain is assembled by Ubc13–Mms2–Rad5 and transferred en bloc to the mono-ubiquitylated PCNA. The chain may be assembled either free in solution or as a thioester on the active-site cysteine of Ubc13. (C) Separate complexes, stepwise assembly: A dedicated Rad6–Rad18 complex mono-ubiquitylates PCNA. Independently, a complex containing Rad6, Rad18, Ubc13, Mms2 and Rad5 mediates K63-poly-ubiquitylation. Rad6–Rad18 within this complex attaches the first ubiquitin moiety and enhances the contact of PCNA to Ubc13–Mms2–Rad5, which catalyses chain elongation in a stepwise manner. (D) Separate complexes, preformed chains: As in model C, a dedicated Rad6–Rad18 complex mediates PCNA mono-ubiquitylation. A separate Rad6–Rad18–Ubc13–Mms2–Rad5 complex poly-ubiquitylates PCNA by the assembly of a K63-linked poly-ubiquitin chain through Ubc13–Mms2–Rad5, which is then transferred en bloc to PCNA by Rad6–Rad18. Download figure Download PowerPoint Results In vitro reconstitution of PCNA poly-ubiquitylation In order to analyse the mechanism of PCNA poly-ubiquitylation, Rad5 and the Rad6–Rad18 complex were purified from Saccharomyces cerevisiae strains overexpressing the relevant genes. Bovine ubiquitin and recombinant human E1 were obtained from commercial sources, and PCNA, Ubc13 and Mms2 were produced in Escherichia coli. Rad5, Rad18 and Ubc13 were produced with an N-terminal His6-epitope to aid purification. Previous studies had indicated that budding-yeast Rad18 is active only towards PCNA that is loaded onto DNA (Garg and Burgers, 2005). We, therefore, included a nicked plasmid and recombinant clamp loader, Replication Factor C (RFC), in our reactions. Figure 2A shows efficient poly-ubiquitylation of PCNA in the presence of all components, in accordance with analogous experiments using human enzymes (Unk et al, 2006, 2008). High-molecular-weight species of PCNA were produced under these conditions, indicating assembly of long poly-ubiquitin chains on the loaded clamp. As expected from the behaviour of the respective mutants in vivo, Rad6–Rad18 alone produced mono-ubiquitylated PCNA, and omission of Rad6–Rad18 from the reaction prevented both mono- and poly-ubiquitylation. The confinement to mono-ubiquitylated PCNA in reactions containing a K63R mutant of ubiquitin confirmed the linkage specificity of Ubc13–Mms2. Moreover, we observed a low amount of di-ubiquitylated PCNA in the absence of Rad5, indicating that Ubc13–Mms2 was marginally active towards mono-ubiquitylated PCNA even without its cognate E3. Taken together, these results are consistent with in vivo data indicating requirement of Rad6–Rad18 for K63 poly-ubiquitylation of PCNA by Ubc13–Mms2–Rad5, but they do not allow distinction between the different models depicted in Figure 1. In order to assess the kinetics of chain formation on PCNA, we, therefore, followed a time course of poly-ubiquitylation in reactions that had been preincubated with Rad6–Rad18. Figure 2B shows chains of intermediate lengths at early time points that were chased into higher molecular weight species in the course of the reaction. This pattern appears to indicate a stepwise addition of ubiquitin monomers according to models A and C. However, we cannot exclude a combination of stepwise and en-bloc transfer, as the di- and tri-ubiquitylated forms of PCNA may well have been converted to higher forms by the addition of a chain instead of monomers. Figure 2.In vitro reconstitution of PCNA poly-ubiquitylation. (A) Modification of PCNA by mono- and poly-ubiquitylation using purified enzymes. Reactions were performed in the presence of a nicked plasmid and the clamp loader RFC in order to provide loaded PCNA as a substrate. All reactions contained E1 and ATP. Enzymes for PCNA mono-ubiquitylation (Rad6–Rad18) were added first where indicated and the reaction mixture was incubated for 40 min at 30°C before addition of the poly-ubiquitylation factors (Ubc13–Mms2–Rad5) where indicated and further incubation for 40 min. Replacement of ubiquitin by a K63R mutant is indicated as 'R'. Products were detected by Western blotting with a PCNA-specific antibody. Asterisks indicate cross-reactive bands visible upon prolonged exposure of the blots. (B) Time course of the poly-ubiquitylation reaction. Reactions were set up with the mono-ubiquitylation enzymes as described above, and Ubc13, Mms2 and Rad5 were added after a 60 min incubation at 30°C. Starting from this point, samples were taken at the indicated times and analysed by Western blotting as above. Download figure Download PowerPoint Continued presence of Rad18 is not required for poly-ubiquitin chain extension on PCNA The interaction between Rad18 and Rad5 suggests that they might act as a complex in PCNA poly-ubiquitylation, according to models C and D (Ulrich and Jentsch, 2000). In this case, presence of Rad18 would be required throughout the poly-ubiquitylation reaction, either to enhance the contact between PCNA and Ubc13–Mms2–Rad5 (model C) or for catalytic transfer of an entire poly-ubiquitin chain (model D). Alternatively, Rad6–Rad18 might solely be required to attach the first ubiquitin moiety onto PCNA. In the latter case, the complex would be dispensable for the subsequent action of Ubc13–Mms2–Rad5 (models A and B). In order to distinguish between these possibilities, we used a purified preparation of partially mono-ubiquitylated PCNA in chain extension reactions either containing or lacking Rad6–Rad18. Figure 3A shows that the extent of poly-ubiquitylation as judged by the disappearance of mono-ubiquitylated PCNA and the appearance of higher molecular weight forms was unaffected by the Rad6–Rad18 complex. This was true both in the presence of DNA and RFC, where Rad6–Rad18 is in principle capable of modifying PCNA, and in their absence, where Rad18 would interact with PCNA and Rad5 without being able to modify the clamp (Figure 3A). Titration of the concentration of Rad6–Rad18 in the reaction confirmed that the complex had no effect on the activity of Ubc13–Mms2–Rad5 towards mono-ubiquitylated PCNA (Figure 3B). Intriguingly, the unmodified PCNA in the preparation was not further ubiquitylated by the newly added Rad6–Rad18. We suspect that this may be due to either a failure to modify PCNA within a trimer already bearing one or two ubiquitin moieties, or simply due to the low concentration of unmodified PCNA in the reaction. Along a similar line of observation, Rad6–Rad18 was previously reported to be rather sensitive to the scale of the reaction (Garg and Burgers, 2005). Overall, however, our data suggest that despite their physical interactions, the two E2–E3 pairs, Rad6–Rad18 and Ubc13–Mms2–Rad5, act sequentially and independently of each other, in support of models A and B (Figure 1). Figure 3.Influence of Rad6–Rad18 and DNA/RFC on PCNA poly-ubiquitylation. (A) Rad6–Rad18 is dispensable for chain extension by Ubc13–Mms2–Rad5. Purified, partially mono-ubiquitylated PCNA was used in reactions containing Rad5, Ubc13 and Mms2 where indicated. All reactions contained E1 and ATP, and Rad6–Rad18 was added where indicated. The assay was performed both in the presence and absence of nicked plasmid DNA and RFC. (B) Titration of the concentration of Rad6–Rad18 in poly-ubiquitylation reactions on purified, partially poly-ubiquitylated PCNA confirms that Rad6–Rad18 is dispensable for chain extension. Rad6–Rad18 concentration was varied from 200 to 800 nM. (C) Loading of PCNA enhances the efficiency of chain extension by Ubc13–Mms2–Rad5. Poly-ubiquitylation reactions on purified, partially mono-ubiquitylated PCNA were performed as described above, but adding DNA and RFC separately as indicated. Download figure Download PowerPoint Loading of PCNA onto DNA enhances the efficiency of PCNA poly-ubiquitylation A second important conclusion from the experiments shown in Figure 3A and B is that chain elongation by Ubc13–Mms2–Rad5—in contrast to Rad6–Rad18-dependent mono-ubiquitylation—does not require PCNA to be loaded onto DNA. Nevertheless, comparison of the modification efficiencies in the presence and absence of DNA and RFC (Figure 3A) indicated that loading of PCNA might stimulate the poly-ubiquitylation reaction to some extent. Considering that Rad5, like Rad18, is a DNA-binding protein (Johnson et al, 1994), a similar behaviour would not be surprising. In order to verify that the observed stimulation was really due to PCNA loading and not to the mere presence of either DNA or RFC, we repeated the chain extension reactions, adding DNA and RFC separately (Figure 3C). Based on the depletion of the mono-ubiquitylated substrate, we found that stimulation of the reaction required both DNA and RFC, suggesting that mono-ubiquitylated PCNA is a better substrate for Ubc13–Mms2-Rad5 when residing on DNA than free in solution. Surprisingly, while RFC alone had no effect, addition of DNA alone reproducibly resulted in slight inhibition of PCNA poly-ubiquitylation. This effect might be due to sequestering of Rad5 on DNA, which could render modification of soluble PCNA less efficient. Importantly, however, the scaffolding function of DNA that was found indispensable for mono-ubiquitylation by Rad6–Rad18 appears to be beneficial, but not essential, for the activity of Rad5 towards PCNA. Rad5 tolerates variations in the site of ubiquitin attachment on PCNA In the absence of PCNA, Rad5 strongly stimulates the synthesis of unanchored poly-ubiquitin chains by Ubc13-Mms2 (Figure 4A). By means of its interaction with PCNA, Rad5 may therefore simply act as an enhancer of E2 activity that mediates proximity of the Ubc13–Mms2 complex to the substrate. Alternatively, Rad5 may be involved in the specific recognition of PCNA mono-ubiquitylated at K164. In order to determine to what extent Rad5 is selective with respect to the site of modification on PCNA, we generated fusions of ubiquitin to the N- or C-terminus of PCNA as artificially 'mono-ubiquitylated' substrates. As shown in Figure 4B and C, both constructs were modified by Ubc13–Mms2–Rad5. This indicates that Rad5 does not require the first ubiquitin moiety to be attached to a specific site on PCNA. However, the two constructs were modified with very different efficiencies. The fusion protein bearing ubiquitin at its N-terminus was virtually depleted in the course of a 30-min reaction, thus exhibiting substrate qualities comparable to or even better than the physiologically K164-modified PCNA (Figure 4B). In contrast, modification of the C-terminal fusion was rather inefficient (Figure 4C). Although the inefficiency of Ubc13–Mms2–Rad5 towards the C-terminal fusion construct could possibly indicate some preference with respect to the modification site, it is more likely due to the arrangement of the fusion partners: as K63 is spatially adjacent to the N-terminus of ubiquitin, modification at this site could easily be impeded by a partial obstruction by means of the fusion. Time-course and titration experiments with the N-terminal fusion protein again showed a stimulatory effect of PCNA loading (Figure 4D and E). Interestingly, we also observed inhibition of poly-ubiquitylation by concentrations of RFC approaching stoichiometric amounts in the absence of DNA (Figure 4E). It is unclear whether this effect was due to an inability of Ubc13–Mms2–Rad5 to modify PCNA in the ring-opened conformation or whether RFC simply sequestered the clamp away from the modifying enzymes. Figure 4.Influence of the ubiquitin attachment site on poly-ubiquitin chain formation. (A) Rad5 stimulates the polymerisation of free ubiquitin by Ubc13 and Mms2. Chain synthesis assays were performed in the absence of PCNA under our standard reaction conditions, using Ubc13–Mms2 at concentrations where chains above di-ubiquitin are formed inefficiently (200 nM). Rad5 concentration was varied from 25 to 200 nM. Replacement of ubiquitin by a K63R mutant is indicated as 'R'. (B–E) Linear N- and C-terminal fusions of ubiquitin to PCNA can be poly-ubiquitylated by Ubc13–Mms2–Rad5. (B) Chain extension reactions with Rad5, Ubc13 and Mms2 as indicated were carried out for 30 min using the N-terminal ubiquitin fusion, HisUb–PCNA, as a substrate. DNA and RFC were added for loading of the fusion protein where indicated. (C) The same reactions as in panel B were performed with the C-terminal ubiquitin fusion, HisPCNA–Ub, for 60 min. (D) Time-course experiments were performed with HisUb–PCNA in order to compare the efficiency of chain elongation in the loaded versus the unloaded state. (E) Titration of RFC in the presence and absence of DNA indicates that loading stimulates poly-ubiquitylation of HisUb–PCNA, whereas high concentrations of RFC in the absence of DNA inhibit the reaction. RFC concentration was varied from 2.5 to 20 nM. Download figure Download PowerPoint Rad18 and Rad5 can catalyse the transfer of preassembled poly-ubiquitin chains onto PCNA Models B and D (Figure 1) postulate a Ube2g2-like mechanism in which the E2 transfers an entire poly-ubiquitin chain en bloc to a substrate (Li et al, 2007). In the case of Ube2g2, the chain is built as a thioester upon the active-site cysteine of the E2 (Li et al, 2007); however, other E2s such as E2-25K are also known to accept activated preassembled chains from E1 (Piotrowski et al, 1997). In order to establish whether a similar mechanism could apply to mono- and/or poly-ubiquitylation of PCNA, we assessed the ability of the two E2–E3 pairs to accept and transfer ubiquitin dimers and tetramers of different linkages. A prerequisite for this activity is the ability of the E2 to form a thioester with the respective ubiquitin derivatives. Figure 5A shows that Rad6 formed thioesters with wild-type (WT) ubiquitin, a K63R mutant and di-ubiquitin of K48- and K63-linkage with comparable efficiency. Tetra-ubiquitin of K48 linkage was accepted equally well, and only K63-linked tetra-ubiquitin was used less efficiently for thioester formation. The presence of Rad18 did not significantly affect this pattern (Supplementary Figure S1A). Modification reactions with Rad6–Rad18 on loaded PCNA with the ubiquitin derivatives indicated that all forms were attached to the substrate with similar efficiency when present as the only source of ubiquitin. The reduced thioester formation with K63-tetra-ubiquitin was not limiting for PCNA modification. However, when an equimolar mixture of mono-, di- and tetra-ubiquitin of K63 linkage was used, di- and particularly mono-ubiquitin were strongly preferred over the tetramer. Likewise, minor contaminations of mono-ubiquitin in the K48-di- and tetra-ubiquitin preparations that had negligible effect in the thioester assay (Figure 5A) gave rise to noticeable quantities of mono-ubiquitylated PCNA (Figure 5B). These observations suggest that despite the ability to use polymeric ubiquitin for thioester formation and transfer, mono-ubiquitin is the preferred moiety for transfer to PCNA by Rad6–Rad18, arguing against model D (Figure 1). Figure 5.Transfer of preformed ubiquitin chains by Rad6–Rad18 and Ubc13–Mms2–Rad5. Abbreviations for ubiquitin derivatives are as follows: 'R', K63R mutant; 'di' and 'tet', di-ubiquitin and tetra-ubiquitin chains of K48- or K63-linkage as indicated; 'M', an equimolar mixture of mono-, di- and tetra-ubiquitin (K63 linkage). (A) Rad6 is capable of forming thioesters with di- and tetra-ubiquitin. Reactions containing E1, ATP, the indicated ubiquitin derivatives and Rad6 were analysed by Western blotting with a Rad6-specific antibody under reducing and non-reducing conditions in order to assess thioester formation. (B) Rad6–Rad18 can transfer preformed di- or tetra-ubiquitin moieties onto loaded PCNA. Mono-ubiquitylation reactions were performed under standard conditions in the presence of DNA and RFC with the indicated ubiquitin derivatives. (C) Ubc13 is capable of forming thioesters with di- and tetra-ubiquitin. Thioester assays were performed with Ubc13 as in panel A and analysed by Western blotting with a Ubc13-specific antibody. (D) Ubc13–Mms2–Rad5 can attach preformed di- or tetra-ubiquitin to the N-terminal ubiquitin-PCNA fusion, HisUb–PCNA. Chain extension reactions were performed under standard conditions with the indicated ubiquitin derivatives. (E) Ubc13–Mms2–Rad5 can attach preformed di- or tetra-ubiquitin to PCNA mono-ubiquitylated at K164. Chain extension reactions were performed with purified, partially mono-ubiquitylated PCNA as in panel D, either in the presence or in the absence of DNA and RFC. Download figure Download PowerPoint In order to assess the validity of model B, we performed analogous experiments for the chain extension step. The pattern of thioester formation by Ubc13 was very similar to that obtained with Rad6, in that the efficiency was reduced with K63-tetra-ubiquitin, but comparable for all other derivaties (Figure 5C). Addition of Mms2 and Rad5 did not significantly change this pattern (Supplementary Figure 1B). Chain extension reactions were initially performed with the HisUb–PCNA fusion, as this was modified by Ubc13–Mms2–Rad5 with similar efficiency as K164-mono-ubiquitylated PCNA. We found that all derivatives were efficiently attached to the substrate (Figure 5D), indicating that preformed chains of varying linkage can be used by Ubc13–Mms2–Rad5 on PCNA for chain extension. As expected, only one moiety of the K63R ubiquitin mutant was attached to HisUb–PCNA. Based on substrate depletion, K48 chains were less effectively used than K63 chains, and whereas multiple units of K48- and K63-di-ubiquitin were attached, HisUb–PCNA was modified by no more than a single unit of K48-tetra-ubiquitin. In the case of K63-tetra-ubiquitin, single modification predominated, but higher forms were detectable as well. Time-course experiments confirmed the slight preference for mono-ubiquitin and for K63- over K48-di-ubiquitin, although the rate of tetra-ubiquitin attachment appeared similar for the two linkages (Supplementary Figure 2). The modification pattern on physiologically K164-mono-ubiquitylated PCNA was very similar and was not influenced by the presence of DNA and RFC, although the overall efficiency of the reaction was enhanced (Figure 5E). These results indicate some linkage specificity with respect to the use of preformed chains by Ubc13–Mms2–Rad5 and a slight preference for attachment of monomers over chains. The notion that K63-tetra-ubiquitin was used by Ubc13–Mms2–Rad5 for chain extension despite the inefficiency of thioester formation indicates that the latter was not rate-limiting in our reactions. Considering that this might be different under physiological conditions, transfer of longer poly-ubiquitin units en bloc by Ubc13–Mms2–Rad5 is rather unlikely. Our data, thus, provide support for model A (Figure 1), although some use of short chains cannot be excluded, as the enzymes involved are in principle capable of transferring poly-ubiquitin units. Discussion Our efforts to reconstitute PCNA modification in vitro have given important insights into a mechanism of poly-ubiquitin chain formation that is distinct from previously analysed examples. In contrast to many reported cases where a single E2–E3 pair mediates both attachment of the first ubiquitin moiety to the substrate and extension to a polymeric chain, poly-ubiquitylation of PCNA involves cooperation of two E2–E3 pairs with distinct properties, that is, Rad6–Rad18 and Ubc13–Mms2–Rad5. Our results indicate that there is a clear separation of tasks between the two complexes. Mechanism of cooperation between Rad6–Rad18 and Ubc13–Mms2–Rad5 Previous information about the properties of the conjugation factors involved in PCNA modification gave rise to several alternative models of how poly-ubiquitylation might be mediated (Figure 1). On the one hand, physical interactions between the E3s and their cognate E2s suggested the existence of a complex in which both E2–E3 pairs are present (Bailly et al, 1994; Ulrich and Jentsch, 2000; Ulrich, 2003). In addition, Rad18 is known to dimerise and form a hetero-tetramer with Rad6 (Ulrich and Jentsch, 2000; Notenboom et al, 2007). These notions suggested that there might be dedicated complexes for mono- versus poly-ubiquitylation of PCNA, consisting of either Rad6–Rad18 alone or of all five components, according to models C and D (Figure 1). On the other hand, both Rad18 and Rad5 interact directly with PCNA and could, thus, in principle recognise their substrate independently (Hoege et al, 2002). It was, therefore, important to determine whether Rad6 both Rad18 was required for the chain extension step or whether purified mono-ubiquitylated PCNA could serve as a substrate for modification by Ubc13 both Mms2 both Rad5 alone. Our experiments have now shown that the latter is clearly the case, lending support to models A and B, and effectively ruling out models C and D (Figure 1). Hence, despite their interaction, Rad18 and Rad5 do not follow t