E3 ligase Rad18 promotes monoubiquitination ratherthan ubiquitin chain formation by E2 enzyme Rad6Richard G. Hibberta,1, Anding Huangb,1, Rolf Boelensb,2, and Titia K. Sixmaa,2

aDivision of Biochemistry and Center for Biomedical Genetics, Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands; andbDepartment of Nuclear Magnetic Resonance Spectroscopy, Bijvoet Center for Biomolecular Research, Utrecht University, Padualaan 8, 3584 CH Utrecht,The Netherlands

Edited by Axel T. Brunger, Stanford University, Stanford, CA, and approved February 18, 2011 (received for review November 24, 2010)

In ubiquitin conjugation, different combinations of E2 and E3enzymes catalyse either monoubiquitination or ubiquitin chainformation. The E2/E3 complex Rad6/Rad18 exclusively monoubi-quitinates the proliferating cell nuclear antigen (PCNA) to signalfor “error prone” DNA damage tolerance, whereas a different setof conjugation enzymes is required for ubiquitin chain formationon PCNA. Here we show that human E2 enzyme Rad6b is intrinsi-cally capable of catalyzing ubiquitin chain formation. This activity isprevented during PCNA ubiquitination by the interaction of Rad6with E3 enzyme Rad18. Using NMR and X-ray crystallography weshow that the R6BD of Rad18 inhibits this activity by competingwith ubiquitin for a noncovalent “backside” binding site on Rad6.Our findings provide mechanistic insights into how E3 enzymes canregulate the ubiquitin conjugation process.

biochemistry ∣ translesion synthesis ∣ protein crystallography ∣NMR spectroscopy

Ubiquitin conjugation to a lysine residue on a target is cata-lyzed by an E1, E2, E3 enzyme cascade. The E1 is required

for activation of the E2, but the E2/E3 complex is the activeligase that transfers the ubiquitin moiety to the target. Within thisligase complex, over 30 E2 enzymes provide the enzyme activitywhile hundreds of E3s define the target specificity (1). Ubiquitinconjugation can take several different forms. The most simple ismonoubiquitination, but because ubiquitin itself contains sevenlysine residues, it can be ubiquitinated to form many differentlinkages of ubiquitin chains. The different ubiquitin structuresprovide different molecular signals for processes such as protea-somal degradation, endocytosis, and DNA repair (2, 3).

Mechanistic studies have revealed that ubiquitin chains can beassembled on a target in different ways. A single E2 enzyme canbe responsible for synthesizing target-linked chains, as has beenshown for CDC34 with the Skp1/Cul1/F-box protein E3 enzyme(4). Alternatively, two E2 enzymes can act sequentially on a singletarget to form specific chains, with one E2 enzyme initiatingthe chain formation and another extending the chains. Such a me-chanism has been shown for modifications of yeast Anaphasepromoting complex/cyclosome (APC) targets by Ubc4 then Ubc1(5) and modification of PCNA by Rad6 with E3 Rad18, thenUbc13/MMS2 with E3 Rad5 (6, 7). To perform these modifica-tions, isolated E2 enzymes may be highly specific for monoubi-quitination or different linkages of ubiquitin chains. Ube2S,E2-25K, and Ubc13/MMS2 are specialized enzymes that catalysethe formation of K11, K48, and K63 chains, respectively (8–11),whereas UbcH5c can formmany linkages of ubiquitin chains (12).

The molecular mechanisms of ubiquitin chain formation byE2 enzymes are only starting to be understood. The specificityof the E2 enzyme Ubc13, with the catalytically inactive E2 variant(E2v) MMS2, for K63-linked ubiquitin chain synthesis has beenexplained in atomic detail (13). MMS2 simultaneously binds toUbc13 and orients an acceptor ubiquitin molecule such that onlyK63 is brought into close proximity to the ubiquitin thioester con-jugate of Ubc13 (13). Interestingly chain formation by the humanE2 enzyme UbcH5 with E3 Brca1 depends upon a noncovalent

interaction with ubiquitin (14), and a similar small ubiquitin-likemodifier (SUMO) interaction is required for the E2 enzymeUbc9 (15–17). Both of these interactions are remarkably similarto the ubiquitin interaction of MMS2 because they occur on the“backside” of the enzymes, distant from the active site cysteine.Loops close to the active site cysteine are important in the chain-formation activity of Ubc1 (18) and Cdc34 (4), whereas Ube2C,together with the APC/C, recognizes TEK boxes found both onthe target securin and close to K11 on ubiquitin (19). It remainsto be resolved if backside ubiquitin binding is also important forthe activity of these enzymes. Only Ube2g2 has been shown tofunction via an entirely different mechanism that results in longubiquitin chains being formed directly on its active site cysteinewhen in complex with the E3 enzyme gp78 (20–22). Althoughmuch of the specificity of the conjugation process is providedby E2 enzymes, E2/E3 enzyme complexes are required for fullactivity and specificity in vivo.

Rad6 and its human homologues, Rad6a/b, are particularlyimportant E2 enzymes that have been shown to be involved inDNA damage tolerance (DDT), histone modification, and pro-teasomal degradation, although other functions of Rad6 may ex-ist (23). DDT has been extensively characterized as a two-steppathway. In the first step, Rad6/Rad18 monoubiquitinate PCNAon K164 at stalled replication forks to signal for recruitment ofdamage-tolerant polymerases. Next, Ubc13/MMS2/Rad5 canform K63-linked polyubiquitin chains on modified PCNA to in-itiate error-free DNA repair (6, 7). In this process Rad6 mono-ubiquitinates only PCNA, even in vitro (24–26). Rad6a/b, withRNF20/40 (Bre1) as E3 ligase, will monoubiquitinate histoneH2B in human cells on K120 (27) as an essential step prior toH3 methylation by Set1 and Dot1. Finally, Rad6 together withUbr1 is associated with ubiquitin chain formation within theN-rule pathway. In this role, Rad6 is essential for the formationof K48 ubiquitin chains on proteins bearing an N degron, whichmarks the target for proteasomal degradation (28). Interestingly,Rad18, Bre1, and Ubr1 all contain additional Rad6 recognitionsites outside a canonical RING interface (27, 29, 30). The func-tional significance of Rad6 has been firmly established, but themolecular mechanisms of Rad6’s activity are not well understood.

Here we analyze the structure and activity of Rad6b. We showthat Rad6b is capable of forming ubiquitin chains via a non-

Author contributions: R.G.H., A.H., R.B., and T.K.S. designed research; R.G.H. performedthe enzyme assays and crystallography; A.H. performed the NMR experiments; R.G.H.,A.H., R.B., and T.K.S. analyzed data; and R.G.H., R.B., and T.K.S. wrote the paper.

The authors declare no conflict of interest .

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.

Data deposition: The crystallography, atomic coordinates, and structure factors have beendeposited in the Protein Data Bank, www.pdb.org [PDB ID codes: Native Rad6b, 2yb6;Rad6b-Rad18 (R6BD) complex, 2ybf].1R.G.H. and A.H. contributed equally to this work.2To whom correspondence may be addressed. E-mail: r.boelens@uu.nl or t.sixma@nki.nl.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1017516108/-/DCSupplemental.

5590–5595 ∣ PNAS ∣ April 5, 2011 ∣ vol. 108 ∣ no. 14 www.pnas.org/cgi/doi/10.1073/pnas.1017516108

Dow

nloa

ded

by g

uest

on

May

31,

202

0

covalent interaction with ubiquitin. Rad18 inhibits this activityduring PCNA monoubiquitination because its major interactionsurface with Rad6 can compete with ubiquitin binding and pre-vent ubiquitin chain formation by Rad6. Our results provide me-chanisms by which different E3 enzymes can produce differentmolecular signals with the same E2 enzyme: Rad6 is able to formtarget-linked ubiquitin chains, presumably catalyzed by appropri-ate E3 enzymes, but is directed toward PCNA monoubiquitina-tion by Rad18.

ResultsRad6b Can Form Ubiquitin Chains. When E2 enzymes can formubiquitin chains in vivo, they typically retain some activity for ubi-quitin chain formation in solution, even in the absence of an E3enzyme and a target. We studied if isolated Rad6b has the abilityto form ubiquitin chains in solution, using an in vitro reconsti-tuted enzyme reaction. Rad6b showed a clear intrinsic ability toform ubiquitin chains (Fig. 1A and Fig. S1A). The specificity ofthe ubiquitin chain linkages was studied using mutants of ubiqui-tin with a single lysine mutated to arginine. The activity wasmodestly reduced in all of the mutants compared with wild-typeubiquitin, but the K11R mutant showed a more dramatic effect:The molecular weight of the ubiquitinated species was stronglyreduced and a diubiquitin species was no longer visible. This isindicative of Rad6b forming mixed ubiquitin chains, with a fairlystrong preference for K11-linked chain formation (Fig. 1B).

Some monoubiquitination of Rad6 was detected but not Rad6-linked ubiquitin chains (Fig. S1B), so we interpret that most ofthe chains are free in solution. We cannot exclude some level ofRad6-linked ubiquitin chains. When Rad6b was analyzed in theabsence of reducing agents, we observed thioester formation(Fig. S1B), but no chains are formed on the active site cysteineof Rad6b.

The Mechanism of the Chain Formation Is via a Noncovalent Interac-tion with Ubiquitin. To understand the mechanism of ubiquitinchain formation by Rad6 we used NMR to study the interactionsof Rad6 with ubiquitin. Rad6 has previously been shown tobind noncovalently to ubiquitin as well as forming the covalentthioester intermediate that is conserved among all catalytic E2enzymes (31). We followed both events with NMR using the pre-viously published assignment of Rad6b that was applicable to ourprotein preparations (31, 32). The thioester formation on Rad6was followed by incubating 15N-labeled Rad6b with ubiquitin,ATP∕Mg2þ, and E1 activating enzyme. Within 1 h after additionof ATP, changes in chemical shifts were observed for residues sur-rounding the active site cysteine of Rad6b, consistent with Rad6bforming a thioester bond with ubiquitin (Fig. 1C and Fig. S2A).

We performed a structural characterization of the noncovalentcomplex as a basis for studying its role in ubiquitin chain forma-tion. In the absence of a high-resolution structure, we validatedpublished NMR titrations (31), then performed a data-driven

Fig. 1. Rad6b can form ubiquitin chains via a noncovalent interaction withubiquitin. (A) Time-course experiment of ubiquitin chain formation by Rad6b.Antiubiquitin Western blot using an in vitro system containing purified E1(90 nM), His-Rad6b (3 μM), ubiquitin (12 μM), and ATP∕Mg2þ (3 mM∕10 mM). The time points are as follows: 8 h, 0 h, 30′, 1 h, 2 h, 4 h, and 8 h (lanes1–7). (B) Specificity of Rad6-catalyzed chain formation. Antiubiquitin Westernblot of in vitro ubiquitin chain-formation assays using E1 (90 nM), His-Rad6b(3 μM), and wild-type ubiquitin or single lysine to arginine point mutants ofubiquitin (12 μM) after 8 h. The ubiquitin mutants are as follows: WT, WT,K6R, K11R, K27R, K29R, K33R, K48R, and K63R (lanes 1–9). The experiment re-veals a preference of K11-linked ubiquitin chains. (C) Rad6b forms a thioesterbond with ubiquitin in the presence of E1 and Mg2þ∕ATP. NMR experimentsshowed significant CSPs of residues D50, G51, T69, N80, I87, C88, L89, D90,I91, L92, Q93, A122, and N123 (colored green) near the active site cysteine,C88, following thioester formation of Rad6b with ubiquitin. (D) The Rad6b-ubiquitin noncovalent complex. Cartoon representation of the lowest energydocking solution of Rad6b-ubiquitin noncovalent complex. Significant CSP ofRad6b (>0.05; residues 22–26, 38–39, 41, 51–52, 142, and 146–147) and ubiqui-tin (CSP > 0.025; residues 7–8, 13–14, 32, 41–42, 47–49, and 68–72) are high-lighted in green and blue. Residues G23 and T52 are highlighted on thecartoon representation in red. The affinity of the interaction is estimated tobe approximately 600 μM. (E) A structural superposition of UbcH5/ubiquitin[cyan; PDB ID code 2FUH(14)], MMS2/ubiquitin [green; PDB ID code 2GMI(13)], and Rad6b/ubiquitin (blue). The mode of ubiquitin binding is highlysimilar between these E2 and E2v enzymes. (F) Point mutations at “backside”of Rad6b interfere with ubiquitin chain formation. The point mutants wereselected to interfere with ubiquitin binding but not Rad18 binding, basedon our structural data (see below). Assay follows chain formation after 4 h withpurified E1 (90 nM) and ubiquitin (12 μM) and a 2-fold dilution series of wild-type His-Rad6b or a G23A T52R His-Rad6b mutant, analyzed by antiubiquitinWestern blotting. The Rad6b concentrations are as follows: 2 μM, 1 μM,0.5 μM, 0.25 μM, 0 μM, 2 μM, 1 μM, 0.5 μM, 0.25 μM, and 0 μM (lanes 1–10).(G) The back side G23A T52R Rad6 mutant is not defective in E1 interaction.Thioester formation experiment assay with His-Rad6b (3 μM), ubiquitin(12 μM), and a 3-fold dilution series of E1, analyzed by anti-His (Rad6) Westernblot after 10 min under nonreducing conditions. The E1 concentrations were asfollows: 100 nM, 33 nM, 11 nM, 3.7 nM, 1.2 nM, 0.4 nM, 0 nM, 100 nM, 33 nM,11 nM, 3.7 nM, 1.2 nM, 0.4 nM, and 0 nM (lanes 1–14). No significant isopeptidebond formation was observed under these conditions. (H) Model showing thechain-formation activity of Rad6b depends upon a covalent and noncovalentinteraction with ubiquitin.

Hibbert et al. PNAS ∣ April 5, 2011 ∣ vol. 108 ∣ no. 14 ∣ 5591

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

May

31,

202

0

docking approach to generate models of the complex (33). Thesignificant chemical shift perturbations (CSPs) in the NMR titra-tions mapped to β-strands β1-3 and helix α4 on Rad6b and toβ-strands β1-5 of ubiquitin (Fig. 1D and Fig. S2B). Our highestscoring docking models pack the hydrophobic core of ubiquitin,surrounding Ile44, against a small hydrophobic patch on the“backside” of Rad6b, around Val39 and Phe41 (Fig. 1D). Inter-estingly, when our model is compared with existing structuresof E2s and E2 variants making noncovalent complexes with ubi-quitin (13, 14), we find the complexes are highly similar to eachother and differ only by small rotations in the orientation of theubiquitin (Fig. 1E). This confirms that the backside of Rad6binteracts with ubiquitin via a canonical mode of interaction.

To understand the functional relevance of the Rad6b-ubiquitincomplex in ubiquitin chain formation, we constructed single pointmutants on the backside of Rad6b to perturb the complex, basedon our structural data, and tested the recombinantly expressedproteins in our in vitro chain-formation assay. Both a G23Rand T52A mutation (Fig. 1D) show reduced activity for ubiquitinchain formation (Fig. S1C) and a G23R T52A double mutant ofRad6b was further compromised in ubiquitin chain formation(Fig. 1F), without any detectable effect on E1 interaction(Fig. 1G). We conclude that the noncovalent backside interactionis required for chain formation (Fig. 1H).

Rad18 Inhibits the Rad6′s Chain-Formation Activity. We studied theactivity of Rad6 toward PCNA. Consistent with previous reports(24–26), Rad6 will modify PCNA only when in complex withE3 ligase Rad18, with no evidence for ubiquitin chains formedby the Rad6/Rad18 complex (Fig. 2 A and C). Because Rad6bis capable of forming ubiquitin chains, but Rad6/Rad18 com-

plexes only monoubiquitinate PCNA, we studied whether Rad18inhibits the chain-formation activity of Rad6b upon forming abinary Rad6/Rad18 complex. We tested the chain-forming activ-ity of Rad6 in the presence of Rad18. Remarkably, the chainformation is abrogated when Rad18 is present in stoichiometricamounts (Fig. 2B).

The Rad18 (R6BD) Competes with Ubiquitin Binding and UbiquitinChain Formation. To understand the mechanism of this inhibitionof chain formation by Rad18, we studied the interactions betweenRad6 and Rad18. Rad18 recognizes Rad6 via an N-terminalRING domain and a separate C-terminal Rad6 binding domain(R6BD) (Fig. 3A) (29, 34). Because the RING-E2 interaction ishighly conserved among E2-E3 complexes regardless of theirchain-formation activity, and Rad18 binds via this canonicalinteraction, we studied the interaction of Rad6 with the Rad18(R6BD) using NMR titrations. The observed CSPs mappedthe binding site of the R6BD at the backside of Rad6b, oppositeto the active site cysteine, C88 (Fig. 3 B and C and Fig. S2C),with major shifts on β-strands β1, β2, and β3 and the C-terminalresidues of Rad6b. The titrations are consistent with a 62-μMaffinity and fast-on/fast-off rates for the interaction that wereobserved using SPR (Fig. S3A).

We solved the crystal structure of Rad6 alone and in complexwith the R6BD peptide as a basis to understand the interactionin atomic detail (Fig. 3 C and D, Table 1, and Fig. S4). In bothstructures the Rad6b has the canonical E2 fold, and the twomonomers are highly similar (Fig. S4). The R6BD structure isthat of a kinked helix with residues Ser348-Arg358 forming along α-helix and residues Lys341-Lys344 forming a second helixrotated by about 60° with respect to the first. Consistent withthe NMR data, the peptide interacts with the backside of Rad6b,via a series of hydrogen bonds, salt bridges, and van der Waalscontacts (Fig. 3 C and D). The isolated R6BD peptide formsloosely folded helices in solution (Fig. S3 B and C and Table S1).The R6BD adopts the more compact structure upon complexformation.

Our structural studies show that the major binding site ofRad18 (R6BD) on Rad6b overlays with the noncovalent ubiquitininteraction site that is important for ubiquitin chain formation(Fig. 4A). Therefore we postulated that the R6BD of Rad18could compete with the ubiquitin chain-formation activity ofRad6 by preventing binding to ubiquitin. Our affinity estimatesfor the individual interactions imply that the Rad18 (R6BD)should efficiently compete with ubiquitin for binding to the back-side of Rad6b. We tested the competition between the two inter-actions using our NMR assay. This experiment showed that theR6BD can compete with ubiquitin for Rad6 binding (Fig. 4B).Next, we studied if R6BD has an effect on the Rad6’s chain-formation activity. We followed the ubiquitin chains formed byRad6b alone or in the presence of R6BD. The R6BD peptidecould compete for Rad6’s ubiquitin chain-formation activity(Fig. 4 C and D). As a control, we used an R6BD* peptide withfour point mutations (H346A, L353A, V354A, and A357R) thatmaintained the same overall charge as the native peptide andcontained the same number of lysine residues, but bound with alower affinity. R6BD* showed a highly reduced inhibition of theubiquitin chain-formation activity (Fig. 4 C and D and Fig. S1D).A small inhibition remained, however, probably because R6BD*still has some affinity for Rad6b, or because peptide modificationitself competes for ubiquitin in the assay. As an additional controlwe studied if the secondary interaction site of chain-forming E3ligase Ubr1, the basic residues-rich (BRR) domain, was capableof competing with Rad6b for ubiquitin binding and ubiquitinchain formation. We show that the BRR binds to the same siteas ubiquitin on the backside of Rad6b, but with only a millimolaraffinity. The BRR is not able to compete for ubiquitin bindingor ubiquitin chain formation (Fig. S5). In summary, the Rad18

Fig. 2. Rad18 inhibits Rad6’s ubiquitin chain-formation activity to directRad6 toward PCNA monoubiquitination. (A) Time-course experiment ofPCNA monoubiquitination by Rad6b or Rad6b/Rad18 complexes. Anti-PCNAWestern blot following an in vitro assay containing purified E1 (90 nM),Rad6b (20 μM), or Rad6b/Rad18 (20 μM), ubiquitin (12 μM) and PCNA (3 μM).The following time points were used: 0, 1 h, 4 h, 16 h, 0, 1 h, 4 h, and 16 h(lanes 1–8). The Rad6/Rad18 dependent PCNA monoubiquitination is almostcomplete under these conditions, with no evidence for chain formation onPCNA. Rad6b shows little detectable activity for PCNA in the absence ofRad18. (B) The full-length Rad18 inhibits Rad6-mediated ubiquitin chainformation. Time-course experiment of ubiquitin chain formation by E1(90 nM), ubiquitin (12 μM), and His-Rad6b (3 μM) alone, or reconstitutedwith stoichiometric His-SUMO-Rad18, followed by antiubiquitin Westernblotting. The following time points were used: 3 h, 0, 20′, 1 h, 3 h, 0, 20′, 1 h,and 3 h (lanes 1–9). Rad18 was expressed in the absence of Rad6b and recon-stituted into the reaction. Significant ubiquitination of Rad18 was detected.(C) Cartoon showing PCNA monoubiquitination by Rad6/Rad18 complexes.

5592 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1017516108 Hibbert et al.

Dow

nloa

ded

by g

uest

on

May

31,

202

0

(R6BD) can compete with the activities of Rad6 for ubiquitinbinding and ubiquitin chain formation under these conditions,but the Ubr1 (BRR) cannot.

We studied the effect of mutating the R6BD on the activity offull-length Rad18. We produced full-length Rad18* protein thatcontained the same four-point mutants in the R6BD domain(H346A, L353A, V354A, and A357R) and compared the activityof this protein with wild-type Rad18 in the Rad6-dependentmodification of PCNA. Mutations in the R6BD strongly reducethe PCNA ubiquitination because an intact R6BD is requiredfor transfer of ubiquitin onto PCNA (Fig. S1E). Consequentlywe do not observe ubiquitin chains on PCNA (Fig. S1E). Finally,we studied the ability of Rad18* to inhibit the chain-formationactivity of Rad6. In contrast to the wild-type protein, our mutantRad18 is no longer able to inhibit the chain-formation activityof Rad6 (Fig. S1F), confirming that a functional Rad18 (R6BD)is required to inhibit Rad6’s chain-formation activity.

DiscussionWe show that Rad6 has an intrinsic capability for ubiquitin chainsynthesis that depends upon a noncovalent interaction withubiquitin. Rad18 is capable of modulating this activity by compet-ing with the noncovalent ubiquitin interaction. We propose thatRad18 utilized this mechanism to direct Rad6 toward monoubi-quitination rather than chain formation on PCNA.

Rad18 can recognize Rad6 via its RING and R6BD domains.RING-E2 interactions are highly conserved among RING E3ligases and a functional interaction is a prerequisite for activity

(1). Consistent with this, Rad18 RING domain mutants show re-duced activity in vivo (35, 36). The R6BD interaction is requiredin vivo in higher eukaryotes (36), and in our in vitro reactions.The interaction is also important in yeast because cells thatproduced Rad18 with a truncated R6BD showed increased sen-sitivity to UV irradiation (29). Our high-resolution structure ofthe Rad6-R6BD complex is highly consistent with the mappingexperiments in the yeast study but reveals that their in vivoconstructs (29) still contained a critical part of the R6BD. It islikely that constructs lacking the complete R6BD will be evenmore sensitive to UV light.

We demonstrate that human Rad6b can form ubiquitin chains,as has been shown for Saccharomyces cerevisiae Rad6 (37). Weshow that chain formation by human Rad6 depends upon anoncovalent interaction with ubiquitin, using mutagenesis aswell as R6BD peptides that inhibit the ubiquitin interaction andthe chain-formation activity of Rad6 in a dose-dependent man-ner. This places the chain-formation activity of Rad6 in the sameclass as several other ubiquitin E2s, SUMOE2s, and E2/E2v com-plexes (13–15, 17). Models that use such a noncovalent interac-tion are attractive because they provide an explanation for howthe E2 is associated with the end of the growing ubiquitin chain,which becomes increasingly distant from the original conjugationsite on the target as the length of the chain increases.

In contrast to Rad6, Ube2g2 has been shown to synthesizeubiquitin chains via a very different mechanism from other E2s,because it forms long ubiquitin chains on its active site cysteine,when in complex with E3 enzyme gp78 (20). Ube2g2 utilizes a

A C

B

D

Fig. 3. The interaction of R6BD with Rad6b. (A) Domain structure of Rad18 with the amino acid sequence of the R6BD indicated. (B) NMR spectrum (1H-15N-HSQC) of Rad6b alone (black) with increasing amounts of R6BD to a final Rad6b:R6BD molar ratio of 1∶4 (red). (C) Crystal structure of the complex of R6BDand Rad6b. R6BD (orange) is a kinked helix that binds to the back of Rad6b. Significant (>0.1) CSPs of Rad6b (residues 23–27, 37–41, 44, 50–52, 54–57, 146, and148–149) are highlighted in green. The crystal structure is highly consistent with the NMR titration data. (D) Details of the Rad6b interaction with R6BD. Theinteraction of the peptide with Rad6b is stabilized by potential hydrogen bonds to Tyr342, Arg343, and His346 at the kinked N terminus of the peptide andTyr361 at the C terminus, although this residue may be somewhat mobile because it is less well resolved in the electron density. The interaction is also stabilizedby a number of hydrophobic interactions including Val354 and Ala357, which are buried in a hydrophobic patch on Rad6b surrounding Val39 and Phe41.Residues at the interface on R6BD and Rad6b are colored yellow and green, respectively.

Hibbert et al. PNAS ∣ April 5, 2011 ∣ vol. 108 ∣ no. 14 ∣ 5593

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

May

31,

202

0

similar mode of interaction with gp78, to the complex betweenRad6 and Rad18 (21, 22). The gp78 peptide forms a continuoushelix on the backside of Ube2g2 and the binding location of thishelix is similar to the R6BD, but it is oriented approximatelyperpendicular to that of Rad18 on Rad6. Ube2g2 has evolvedwith differences close to its active site, and allosteric activationof these residues upon gp78 interaction activates its unique me-chanism (20, 21).

We show that Rad18 can compete for Rad6’s activity towardubiquitin chain formation by competing with the noncovalentinteraction of Rad6 with ubiquitin. In vivo Rad18 is itself mod-ified with K48-linked ubiquitin chains that are recognized by theproteasome (34). This modification is likely to be indirect, withthe chains formed by other E2/E3 enzymes, because a functionalRad18-RING domain is not required.

Rad6 is an abundant E2 enzyme, present in micromolar con-centrations in vivo (38) comparable to our assays. The chain-formation activity of Rad6 that we detect is modest; however,one or more specific E3 enzymes would be expected to activateRad6’s ubiquitin chain-formation activity toward specific targets.Among the E3 partners of Rad6b, a role of Rad6-mediatedubiquitin chain formation in the Rad18 and Bre1 pathways isunlikely because Rad6 is currently believed to exclusively mono-ubiquitinate the targets in these systems (7, 27). Interestingly,Rad6/Ubr1-dependent ubiquitin chains are directly produced bythe N-rule degradation pathway (39), so it is credible that Rad6’snoncovalent interaction with ubiquitin is used to synthesize thechains. We studied whether the BRR secondary interaction siteof Ubr1 was capable of modulating Rad6b-mediated ubiquitinchain formation. This domain bound to the backside of Rad6b,but the affinity was so low that it was not able to compete withubiquitin binding or ubiquitin chain formation, in contrast to theR6BD of Rad18. It is plausible that the Ubr1 (BRR)-Rad6b in-teraction is dynamic during the activity of Ubr1 in N-rule degra-dation and that ubiquitin chains are synthesized via the backsideinteraction of Rad6, although we cannot rule out a higher Rad6b-BRR affinity in the context of the full-length proteins. It is alsopossible that Rad6/Ubr1 complexes form ubiquitin chains onN-rule targets via a yet unknown mechanism.

It is likely that more functions of Rad6 remain to be identified,on previously undescribed targets or with different E3 ligases,which could utilize the chain-formation propensity of Rad6.Because Rad18, Bre1, and Ubr1 all bind to Rad6 via unusual bi-valent interactions, E3 enzymes that bind via weaker interactionsthat require only a RING domain may not have been discoveredyet. Consistent with this, a recent reconstruction of the ubiquitinnetwork in yeast identified Rad6 as a major hub, with manyGolgi-, vesicle-, or endosome-associated roles for Rad6 thatare not explained by currently known targets of Rad6 (23). These

Table 1. Data collection and refinement statistics

Rad6b* Rad6b + Rad18 (339–366)*

Data collectionSpace group P43212 P6522Cell dimensions

a, b, c (Å) 48.1, 48.1, 124.8 58.2, 58.2, 167.1α, β, γ (°) 90.0, 90.0, 90.0 90.0, 90.0, 120.0

Resolution (Å) 44.84–1.50(1.58–1.50)

24.92–2.00(2.11–2.00)

Rmerge 0.086 (0.431) 0.128 (0.718)Mean [ðIÞ∕σðIÞ] 13.0 (4.0) 16.0 (4.7)Completeness (%) 98.1 (97.2) 99.3 (95.5)Redundancy 5.9 (6.0) 12.9 (12.1)RefinementResolution (Å) 1.50 2.00No. reflections 23,655 11,963Rwork∕Rfree 0.166∕0.205 0.201∕0.242No. atoms

Protein 1,198 1,390Ligand/ion 5 15Water 135 85

B factorsProtein 21.6 29.8Ligand/ion 29.0 42.6Water 33.3 16.8

rms deviationsBond lengths (Å) 0.0118 0.008Bond angles (°) 1.409 0.967

*One crystal.

A C

B D

Fig. 4. Ubiquitin chain formation by Rad6b is inhibited by competitionwith Rad18 (R6BD). (A) The binding sites of ubiquitin and R6BD on Rad6boverlap. The crystal structures of ubiquitin with UbcH5 (cyan) and MMS2(green) and our docking-based model of the Rad6b-ubiquitin complex (blue)were superimposed upon the crystal structure of Rad6b in complex withRad18 (R6BD) using only the E2s for the superposition. The cartoon represen-tations of ubiquitin are shown as a semitransparent representation. Rad6b(blue) and Rad18 (R6BD) are shown as opaque cartoons. (B) R6BD can com-pete for backside ubiquitin binding. 1H-15N-HSQC peaks of 15N-labeledRad6b were studied that had been perturbed in a different directionupon addition of ubiquitin (ii) and R6BD (i) during the separate titrations.A 15N-labeled sample of Rad6b (200 μM) was preincubated with a 3-foldmolar excess of ubiquitin (ii) and R6BD was titrated into the sample to thesame final concentration as ubiquitin (iii). These peaks were observed tomove to the same final positions (iii) as the single titration of R6BD intoRad6b (i), confirming that R6BD can compete with ubiquitin for binding toRad6b. (C) Time-course experiment showing that R6BD can inhibit Rad6-mediated ubiquitin chain formation. The assay contains purified E1 (90 nM),His-Rad6b (20 μM), and ubiquitin (20 μM) alone or together with R6BD(100 μM) or a mutated R6BD* (128 μM; H346A, L353A, V354A, and A357R)peptide, followed by antiubiquitin Western blotting. The following timepoints were used: 0, 30′, 2 h, 8 h, 0, 30′, 2 h, 8 h, 0, 30′, 2 h, and 8 h (lanes1–12). Both peptides undergo significant monoubiquitination. (D) A concen-tration series of R6BD shows dose-dependent inhibition of Rad6-mediatedubiquitin chain formation. The assay contains purified E1 (90 nM), His-Rad6b(3 μM), and ubiquitin (12 μM) alone or together with a twofold dilutionseries of R6BD (lanes 3–7: 64 μM, 128 μM, 256 μM, 512 μM, and 1,024 μM)or R6BD* (lanes 9–13: 64 μM, 128 μM, 256 μM, 512 μM, and 1,024 μM) at8 h. R6BD shows a far greater activity than R6BD* for inhibiting thechain-formation activity of Rad6.

5594 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1017516108 Hibbert et al.

Dow

nloa

ded

by g

uest

on

May

31,

202

0

roles may use Rad6b to mark a target for proteasomal degra-dation, because we observed some specificity for K11-linkedubiquitin chains.

We find that the chain-formation activity of Rad6 can beinhibited by E3 complex formation with Rad18 and predictthat other E3 enzymes can promote or inhibit chain formationin different biological contexts. This functional diversity of asingle E2 is not surprising given the low number of E2 enzymescompared with E3s and targets. Understanding how E3s modu-late the activity of E2s toward chain formation on a particulartarget becomes an exciting area for future research.

MethodsRad18-R6BD (339-366), R6BD* (as R6BD with H346A, L353A, V354A, andA357R), and Ubr1-BRR (1013-1028) were produced as synthetic peptides.Rad6b, Rad18, Uba1, Ubiquitin, and PCNA were overexpressed in Escherichiacoli and purified by successive chromatography steps. In vitro ubiquitinationassays were performed using conditions stated in the figure legends, resolvedon SDS-PAGE gels, immunoblotted, and analyzed. NMR titrations were per-

formed on a Bruker Avance2 750-MHz NMR and followed by 1H-15N hetero-nuclear sequential quantum correlation (HSQC) spectra. Docking of thenoncovalent Rad6b/Ub complex was performed using the Haddock Webserver (33). Crystals of Rad6 and the Rad6 (R6BD) complex were grown byvapor diffusion and solved using molecular replacement. The final modelwas refined to Rwork∕Rfree of 16.6%∕20.5% at 1.5-Å resolution (Rad6) orRwork∕Rfree of 19.9%∕24.2% at 2.0-Å resolution (Rad6-R6BD complex).Further details for all experiments are available in SI Methods.

ACKNOWLEDGMENTS. We are grateful to European Synchrotron RadiationFacility beam line scientists, Henk Hilkman for peptide synthesis, DeneLittler and Alex Faesen for crystallography data collection, and AnastassisPerrakis and Robbie Joosten for advice during data processing and refine-ment. Valerie Notenboom provided useful discussions and Anastassis Perra-kis, Judith Smit, and Peter Krijger provided comments on the manuscript.Funding was from the European Union SPINE2COMPLEXES (T.K.S. andR.B.), the Dutch Organization for Scientific Research (NWO) TOP grant(R.B.), and European Union Role of Ubiquitin and Ubiquitin-like Modifiersin Cellular Regulation (T.K.S.).

1. Deshaies RJ, Joazeiro CA (2009) RING domain E3 ubiquitin ligases. Annu Rev Biochem78:399–434.

2. Ikeda F, Dikic I (2008) Atypical ubiquitin chains: New molecular signals. EMBO Rep9:536–542.

3. Ye Y, Rape M (2009) Building ubiquitin chains: E2 enzymes at work. Nat Rev Mol CellBiol 10:755–764.

4. Petroski MD, Deshaies RJ (2005) Mechanism of lysine 48-linked ubiquitin-chainsynthesis by the cullin-RING ubiquitin-ligase complex SCF-Cdc34. Cell 123:1107–1120.

5. Rodrigo-Brenni MC, Morgan DO (2007) Sequential E2s drive polyubiquitin chainassembly on APC targets. Cell 130:127–139.

6. Hoege C, Pfander B, Moldovan GL, Pyrowolakis G, Jentsch S (2002) RAD6-dependentDNA repair is linked to modification of PCNA by ubiquitin and SUMO. Nature419:135–141.

7. Parker JL, Ulrich HD (2009) Mechanistic analysis of PCNA poly-ubiquitylation by theubiquitin protein ligases Rad18 and Rad5. EMBO J 28:3657–3666.

8. Chen Z, Pickart CM (1990) A 25-kilodalton ubiquitin carrier protein (E2) catalyzes mul-ti-ubiquitin chain synthesis via lysine 48 of ubiquitin. J Biol Chem 265:21835–21842.

9. Garnett MJ, et al. (2009) UBE2S elongates ubiquitin chains on APC/C substrates topromote mitotic exit. Nat Cell Biol 11:1363–1369.

10. Hofmann RM, Pickart CM (1999) Noncanonical MMS2-encoded ubiquitin-conjugatingenzyme functions in assembly of novel polyubiquitin chains for DNA repair. Cell96:645–653.

11. Williamson A, et al. (2009) Identification of a physiological E2 module for the humananaphase-promoting complex. Proc Natl Acad Sci USA 106:18213–18218.

12. Brzovic PS, Klevit RE (2006) Ubiquitin transfer from the E2 perspective: Why is UbcH5so promiscuous? Cell Cycle 5:2867–2873.

13. Eddins MJ, Carlile CM, Gomez KM, Pickart CM, Wolberger C (2006) Mms2-Ubc13covalently bound to ubiquitin reveals the structural basis of linkage-specific polyubi-quitin chain formation. Nat Struct Mol Biol 13:915–920.

14. Brzovic PS, Lissounov A, Christensen DE, Hoyt DW, Klevit RE (2006) A UbcH5/ubiquitinnoncovalent complex is required for processive BRCA1-directed ubiquitination. MolCell 21:873–880.

15. Knipscheer P, van Dijk WJ, Olsen JV, MannM, Sixma TK (2007) Noncovalent interactionbetween Ubc9 and SUMO promotes SUMO chain formation. EMBO J 26:2797–2807.

16. Duda DM, et al. (2007) Structure of a SUMO-binding-motif mimic bound to Smt3p-Ubc9p: Conservation of a non-covalent ubiquitin-like protein-E2 complex as a plat-form for selective interactions within a SUMO pathway. J Mol Biol 369:619–630.

17. Capili AD, Lima CD (2007) Structure and analysis of a complex between SUMO andUbc9 illustrates features of a conserved E2-Ubl interaction. J Mol Biol 369:608–618.

18. Rodrigo-Brenni MC, Foster SA, Morgan DO (2010) Catalysis of lysine 48-specificubiquitin chain assembly by residues in E2 and ubiquitin. Mol Cell 39:548–559.

19. Jin L, Williamson A, Banerjee S, Philipp I, Rape M (2008) Mechanism of ubiquitin-chainformation by the human anaphase-promoting complex. Cell 133:653–665.

20. Li W, Tu D, Brunger AT, Ye Y (2007) A ubiquitin ligase transfers preformed poly-ubiquitin chains from a conjugating enzyme to a substrate. Nature 446:333–337.

21. Das R, et al. (2009) Allosteric activation of E2-RING finger-mediated ubiquitylationby a structurally defined specific E2-binding region of gp78. Mol Cell 34:674–685.

22. Li W, et al. (2009) Mechanistic insights into active site-associated polyubiquitination bythe ubiquitin-conjugating enzyme Ube2g2. Proc Natl Acad Sci USA 106:3722–3727.

23. Venancio TM, Balaji S, Iyer LM, Aravind L (2009) Reconstructing the ubiquitin network:Cross-talk with other systems and identification of novel functions. Genome Biol10:R33.

24. Garg P, Burgers PM (2005) Ubiquitinated proliferating cell nuclear antigen activatestranslesion DNA polymerases eta and REV1. Proc Natl Acad Sci USA 102:18361–18366.

25. Notenboom V, et al. (2007) Functional characterization of Rad18 domains for Rad6,ubiquitin, DNA binding and PCNA modification. Nucleic Acids Res 35:5819–5830.

26. Watanabe K, et al. (2004) Rad18 guides poleta to replication stalling sites throughphysical interaction and PCNA monoubiquitination. EMBO J 23:3886–3896.

27. Kim J, et al. (2009) RAD6-mediated transcription-coupled H2B ubiquitylation directlystimulates H3K4 methylation in human cells. Cell 137:459–471.

28. Varshavsky A (1996) The N-end rule: Functions, mysteries, uses. Proc Natl Acad Sci USA93:12142–12149.

29. Bailly V, Prakash S, Prakash L (1997) Domains required for dimerization of yeastRad6 ubiquitin-conjugating enzyme and Rad18 DNA binding protein. Mol Cell Biol17:4536–4543.

30. Xie Y, Varshavsky A (1999) The E2-E3 interaction in the N-end rule pathway: TheRING-H2 finger of E3 is required for the synthesis of multiubiquitin chain. EMBO J18:6832–6844.

31. Miura T, Klaus W, Gsell B, Miyamoto C, Senn H (1999) Characterization of thebinding interface between ubiquitin and class I human ubiquitin-conjugating enzyme2b by multidimensional heteronuclear NMR spectroscopy in solution. J Mol Biol290:213–228.

32. Miura T, Klaus W, Ross A, Guntert P, Senn H (2002) The NMR structure of the class Ihuman ubiquitin-conjugating enzyme 2b. J Biomol NMR 22:89–92.

33. Dominguez C, Boelens R, Bonvin AM (2003) HADDOCK: A protein-protein dockingapproach based on biochemical or biophysical information. J Am Chem Soc125:1731–1737.

34. Miyase S, et al. (2005) Differential regulation of Rad18 through Rad6-dependentmono- and polyubiquitination. J Biol Chem 280:515–524.

35. Tateishi S, Sakuraba Y, Masuyama S, Inoue H, Yamaizumi M (2000) Dysfunction ofhuman Rad18 results in defective postreplication repair and hypersensitivity tomultiple mutagens. Proc Natl Acad Sci USA 97:7927–7932.

36. Huang J, et al. (2009) RAD18 transmits DNA damage signalling to elicit homologousrecombination repair. Nat Cell Biol 11:592–603.

37. Sung P, Prakash S, Prakash L (1988) The RAD6 protein of Saccharomyces cerevisiaepolyubiquitinates histones, and its acidic domain mediates this activity. Gene Dev2:1476–1485.

38. Siepmann TJ, Bohnsack RN, Tokgoz Z, Baboshina OV, Haas AL (2003) Protein interac-tions within the N-rule ubiquitin ligation pathway. J Biol Chem 278:9448–9457.

39. Turner GC, Du F, Varshavsky A (2000) Peptides accelerate their uptake by activating aubiquitin-dependent proteolytic pathway. Nature 405:579–583.

Hibbert et al. PNAS ∣ April 5, 2011 ∣ vol. 108 ∣ no. 14 ∣ 5595

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

May

31,

202

0