In vitro assembly and recognition of Lys-63 polyubiquitin chains.

Polyubiquitin chains assembled through lysine 48 (Lys-48) of ubiquitin act as a signal for substrate proteolysis by 26 S proteasomes, whereas chains assembled through Lys-63 play a mechanistically undefined role in post-replicative DNA repair. We showed previously that the products of the UBC13 and MMS2 genes function in error-free post-replicative DNA repair in the yeast Saccharomyces cerevisiae and form a complex that assembles Lys-63-linked polyubiquitin chains in vitro. Here we confirm that the Mms2.Ubc13 complex functions as a high affinity heterodimer in the chain assembly reaction in vitro and report the results of a kinetic characterization of the polyubiquitin chain assembly reaction. To test whether a Lys-63-linked polyubiquitin chain can signal degradation, we conjugated Lys-63-linked tetra-ubiquitin to a model substrate of 26 S proteasomes. Although the noncanonical chain effectively signaled substrate degradation, the results of new genetic epistasis studies agree with previous genetic data in suggesting that the proteolytic activity of proteasomes is not required for error-free post-replicative repair.

Polyubiquitin chains assembled through lysine 48 (Lys-48) of ubiquitin act as a signal for substrate proteolysis by 26 S proteasomes, whereas chains assembled through Lys-63 play a mechanistically undefined role in post-replicative DNA repair. We showed previously that the products of the UBC13 and MMS2 genes function in error-free post-replicative DNA repair in the yeast Saccharomyces cerevisiae and form a complex that assembles Lys-63-linked polyubiquitin chains in vitro. Here we confirm that the Mms2⅐Ubc13 complex functions as a high affinity heterodimer in the chain assembly reaction in vitro and report the results of a kinetic characterization of the polyubiquitin chain assembly reaction. To test whether a Lys-63-linked polyubiquitin chain can signal degradation, we conjugated Lys-63-linked tetraubiquitin to a model substrate of 26 S proteasomes. Although the noncanonical chain effectively signaled substrate degradation, the results of new genetic epistasis studies agree with previous genetic data in suggesting that the proteolytic activity of proteasomes is not required for error-free post-replicative repair.
The conserved protein ubiquitin (Ub) 1 regulates critical intracellular functions in eukaryotes, including transit through the cell cycle and the induction of the inflammatory response (1,2). Ub executes these and other functions through conjugation to other cellular proteins, either as mono-Ub or in the form of a polyUb chain. C termini of Ub molecules (Gly-76) form covalent isopeptide bonds with ⑀ amino groups of lysine residues of target proteins or, in the case of polyUb chains, of Ub itself. The best-understood function of Ub is that of a signal which targets specific proteins for degradation by 26 S proteasomes (1,2). In Ub/proteasome-dependent proteolysis, a polyUb chain is first assembled on the target protein through a series of three reactions: 1) the active-site cysteine residue of E1 forms a thiol ester with the C terminus of Ub in an ATPdependent reaction; 2) Ub is trans-esterified to the active-site cysteine residue of one of a family of E2s; and 3) in a reaction catalyzed by an E3 enzyme, multiple Ubs are transferred from the E2 to an internal lysine residue of the target protein. Specificity in Ub-dependent proteolysis arises during this substrate polyubiquitination phase and principally reflects the properties of the E3 enzyme (1,2). The known E3 enzymes fall into two mechanistic classes: HECT domain E3s and RING finger E3s (reviewed in Refs. 3 and 4)). Substrates destined for degradation by proteasomes are conjugated to many molecules of Ub, usually in the form of a specific polyUb chain (5).
There are seven lysine residues in Ub, of which five, Lys-6, Lys-11, Lys-29, Lys-48, and Lys-63, have been implicated in polyUb chain assembly (5)(6)(7)(8)(9)(10). Chains linked through Lys-48 are the principal signal for proteolysis by proteasomes (5,11,12). Model proteins conjugated to a single Ub are poor substrates for degradation by 26 S proteasomes in vitro, whereas a model substrate bearing a Lys-48-linked chain comprising at least four Ubs is efficiently recognized (5,13,14). As expected from the importance of Ub/proteasome-dependent proteolysis, the K48R mutation in Ub is lethal in the yeast Saccharomyces cerevisiae (12). In contrast, yeast expressing K63RUb degrade short-lived proteins and abnormal proteins, the canonical substrates of the Ub-proteasome system, at a normal rate (9). However, these yeast display other phenotypes, including sensitivity to UV light (9), stress conditions (6), and translational inhibitors (15), that are likely to reflect the inability to assemble polyUb chains through Lys-63. Thus, different types of polyUb chains appear to function as distinct signals.
Post-replicative DNA repair fills in gaps opposite damage sites in newly synthesized DNA (16,17). Ub plays a central, but poorly understood role in this process, as evidenced by a requirement for Lys-63 of Ub together with multiple proteins that catalyze or regulate Ub conjugation. The defining component of this pathway, encoded by the RAD6 gene (18), is an E2 enzyme (19) that executes its repair function as part of a complex with the DNA-binding protein Rad18 (20,21). Rad18 has a RING finger and is likely to be an E3, but its specific substrates remain to be identified. We recently discovered a second heterodimeric complex consisting of Mms2 and Ubc13 that apparently functions in the assembly of Lys-63 chains for RAD6-dependent repair (22). Ubc13 is an E2 enzyme, whereas Mms2 is a ubiquitin E2 variant (UEV) (23) that resembles an E2 but lacks the defining E2 cysteine residue (24). Mms2 positively regulates the assembly of Lys-63 chains by an unknown mechanism (22). Another RING finger protein, the Rad5 helicase, interacts with the Mms2⅐Ubc13 complex (25). It is not known if Rad5 is an E3 enzyme, a substrate of ubiquitination, or both.
Three lines of evidence suggest that the function of Lys-63 chains in DNA repair does not rely on degradative signaling via proteasomes. First, none of the three known in vivo substrates of Lys-63 chains is degraded by 26 S proteasomes; L28, a subunit of the ribosome, is metabolically stable (15), whereas ubiquitination of the yeast membrane receptors Gap1p and Fur4p is a signal for endocytosis (26,27). Second, the ability of K63RUb to support the viability of yeast and the absence of a detectable proteolytic defect in these cells (9) argue that Lys-63 chains do not serve a generalized proteolytic targeting function. However, it is difficult to exclude the possibility that Lys-63 chains target a small set of substrates for degradation by proteasomes. Finally, catalytically impaired mutants of the 20 S proteasome are relatively insensitive to UV light (28,29), but given the mild phenotypes of mms2 and ubc13 mutants (22,23), this property does not definitively exclude a role for proteasomes in post-replicative repair.
To address the role of Lys-63 polyUb chains in DNA repair, we first used recombinant Mms2 and Ubc13 to characterize the chain assembly reaction in vitro. The ability to make large quantities of Lys-63 chains in vitro enabled us to test whether a Lys-63 chain is competent to signal proteasomal degradation. The positive outcome of this experiment prompted us to conduct a more rigorous test of whether Lys-63 chains target DNA repair-relevant substrates for degradation by proteasomes in vivo.

EXPERIMENTAL PROCEDURES
Reagents and proteins were from Sigma unless stated otherwise. Wild-type Ub was radioiodinated to ϳ8000 cpm/pmol with chloramine T (30). E1 was purified from bovine erythrocytes (31). The expression and purification of recombinant Ubs have been described (22,32). Ubiquitin aldehyde was a gift of R. Cohen (University of Iowa). Isopeptidase T was purified from bovine erythrocytes as described (33).
Plasmids-All coding regions were verified by automated DNA sequence analysis. The P73L or C64A mutation was introduced into pET16b-ScMms2 (22) by the QuikChange mutagenesis method (Stratagene) or standard polymerase chain reaction-based methods, respectively.
Expression and Purification of Mms2, Ubc13, and Mms2⅐Ubc13 Complex-Plasmids pET3a-ScUbc13 and pET16b-ScMms2 (specifying His 10 Mms2) were expressed in BL21(DE3)pJY2 Escherichia coli cells (37) in a 135-liter fermenter under conditions described previously (22). Cell pellets were frozen in liquid nitrogen. N-terminal sequencing and mass spectrometric analysis of the purified complex (see below; after Factor Xa treatment to release the His 10 tag of Mms2) confirmed that the proteins had been expressed and purified without incident. 2 All of our studies were done with His 10 -tagged Mms2. The tagged protein is fully functional in DNA repair in vivo when expressed from a centromeric plasmid under the control of its endogenous promoter (22).
Cell pellets (ϳ15 g) were suspended in 40 ml of binding buffer (5 mM imidazole, 50 mM NaCl, 20 mM Tris-HCl, pH 8) containing 0.1% Nonidet P-40, 1 mM phenylmethylsulfonyl fluoride, 20 g/ml soybean trypsin inhibitor, 20 g/ml leupeptin, and 0.1 mM tosyllysylchloromethyl ketone. Cells were lysed by the addition of 0.4 mg/ml lysozyme, and DNA was digested by adding 10 mM MgCl 2 and 10 g/ml DNaseI. Cell debris was pelleted by centrifugation at 12,000 ϫ g for 15 min. The supernatants (soluble lysates) from the His 10 Mms2-and Ubc13-expressing cells were mixed in a 1:1 ratio and loaded onto a 5-ml Ni 2ϩ -nitrilotriacetic acid column (Novagen). The column was washed with 5 volumes of binding buffer and eluted with 2 volumes of elution buffer (0.5 M imidazole, 50 mM NaCl, 20 mM Tris-HCl, pH 8). EDTA was added to 1 mM, followed by DTT to 0.1 mM. The eluate was concentrated by ultrafiltration and repeatedly diluted in buffer A (20 mM Tris-HCl, pH 7.6, 0.1 mM EDTA, 0.2 mM DTT) until the concentration of imidazole was Ͻ50 mM. The complex was further purified on a 1-ml fast protein liquid chromatography MonoQ column (Amersham Pharmacia Biotech) using a linear gradient of 0 -0.25 M NaCl (in 50 mM Tris-HCl, pH 7.6, 0.1 mM EDTA, and 0.5 mM DTT). The final yield of complex was ϳ2 mg. Uncomplexed Mms2 was obtained by the same method except that Ubc13 lysate was omitted. The concentration of Mms2p was determined assuming ⑀ 280 nm ϭ 16,600 M Ϫ1 cm Ϫ1 . Elution positions from the MonoQ column were His 10 Mms2⅐Ubc13 complex, 0.11 M NaCl and His 10 Mms2, 0.13 M NaCl.
To purify Ubc13, cells (12 g) were lysed in 50 ml of lysis buffer, and DNA was digested as described above. The soluble lysate was fractionated with ammonium sulfate; proteins precipitating between 30 and 70% saturation were collected and dialyzed against 20 mM Tris-HCl, pH 7.6, 0.1 mM EDTA, and 0.5 mM DTT. The dialysate (ϳ700 mg of protein) was loaded onto a 25-ml Q-Sepharose column (Amersham Pharmacia Biotech) that had been equilibrated with 50 mM Tris-HCl, pH 7.6, 0.1 mM EDTA, and 0.5 mM DTT. The loaded column was washed with one volume of equilibration buffer and then eluted with a 380-ml linear gradient of NaCl (0 -0.3 M). Ubc13 eluted at ϳ0.13 M NaCl. The peak fractions were concentrated and dialyzed by dilution into column equilibration buffer. This procedure yielded ϳ23 mg of Ubc13 at a purity of about 60%. The concentration of Ubc13 was determined by quantitating 125 I-Ub present in the Ubc13ϳUb thiol ester (31) against a standard of known concentrations of human UbcH5A (provided by J. You, Johns Hopkins University) Steady-state Ub 2 Synthesis Assay-The principle of the assay has been described; it utilizes two differentially blocked Ub molecules that can only give rise to Ub 2 (22,32). The conditions were 0.1 M E1, 4 M Mms⅐Ubc13 complex, 60 M 125 I-K63CUb (ϳ300 cpm/pmol), pH 8, and 37°C. The concentration of Asp-77Ub was varied to determine steadystate K m and k cat values.
Pulse-chase Ub 2 Synthesis Assay-Assays were performed essentially as described previously, at pH 7.6 and 37°C (32). A substoichiometric concentration of 125 I-Ub (the donor Ub, 2 M) was incubated for 3 min (37°C) with E1 (50 -250 nM), Ubc13 (4 M), and varying concentrations of His 10 Mms2 to form the Ubc13ϳUb thiol ester in a pulse. A high concentration of unlabeled Ub (the acceptor Ub) was then added together with 20 mM EDTA to initiate the chase. Ub 2 formed during the chase is exclusively Lys-63-linked (22). Aliquots were withdrawn at increasing times between 0 and 5 min and quenched in sample buffer without ␤-mercaptoethanol. The reaction products were electrophoresed, and the Ubc13ϳUb bands were excised and quantitated by ␥-counting. Pseudo-first order rate constants were obtained from semilog plots of the percent E2ϳUb thiol ester radioactivity remaining versus incubation time. The data were corrected by subtracting counts obtained in a control reaction quenched with sample buffer containing mercaptoethanol (a low level of stable E2-Ub adduct is produced through autoubiquitination). Hydrolysis of the E2ϳUb thiol ester bond was negligible during the time course of the reactions studied here (data not shown).
In the first round of Lys-63 chain assembly, K48C, K63CUb was conjugated to K48RUb-NAL in a 1-ml incubation containing 5 mg of each protein, 4 M His 10 Mms2⅐Ubc13 complex, and 0.1 M E1. After incubation for 4 h at pH 7.6 (38), enzymes were removed on an anion exchange column. Ub 2 was then purified by gradient elution from a 1-ml S-Sepharose column (Amersham Pharmacia Biotech) at pH 4.5 (38). The NAL moiety was removed using YUH1 (38); DTT, 1.7 mM, was then added for 15 min followed by iodoacetamide, 10.2 mM, for 50 min more (all at 37°C). Alkylating the two cysteines in the distal Ub stabilized the chain against intermolecular aggregation via disulfide bond formation. YUH1 was removed on an anion exchange column.
This proximally deblocked dimer (4.7 mg) was conjugated to K48RUb-NAL (2.7 mg) essentially as described above (0.6-ml reaction). This step and the next one employed a slight molar excess of the monomeric reactant to drive most of the chain reactant into product, thereby facilitating purification by a stepwise method. After 4 h of incubation, 100 l of a 50% slurry of Q-Sepharose was added. After 2 A. VanDemark and C. Wolberger, personal communication.
mixing, the resin was pelleted, and the supernatant was transferred to a separate tube. The resin was washed with 100 l of buffer. The combined supernatants were adjusted to pH 4.5, and Ub species were loaded onto a 0.5-ml S-Sepharose column. The column was eluted first with 2 ml of the same buffer containing 0.33 M NaCl and then with 2 ml of the same buffer containing 0.4 M NaCl. All of the Ub 3 was in the second fraction. This fraction was concentrated and dialyzed, the NAL moiety was removed using YUH1, and YUH1p was removed by anion exchange (paragraph above). In the final round of the reaction (0.5 ml), the proximally deblocked trimer (3.6 mg) was conjugated to K48RUb-NAL (1.5 mg) in a 1.5-h incubation. Conjugating enzymes were removed by anion exchange (paragraph above), and Ub 4 was purified from the unbound fraction essentially as described in the paragraph above, except that the first elution employed 0.35 M NaCl, and the second employed 0.7 M NaCl. The yield of Lys-63-Ub 4 was 4.8 mg. Lys-48-Ub 5 DHFR and Lys-63-Ub 5 DHFR were prepared as described by conjugating the respective polymers to 35 S-UbDHFR using E2-25K (13).
Degradation Assays-26 S proteasomes were purified (39) and assayed as described previously (13). Briefly, reactions containing 10% glycerol, 0.4 mM DTT, 1 M ubiquitin aldehyde, 2 mg/ml bovine serum albumin (to prevent absorption), and Lys-63-or Lys-48-35 S-Ub 5 DHFR (0 -200 nM) were preincubated at 37°C for 8 min. Purified 26 S proteasomes (2.5 nM) were added. Aliquots of 7 l were withdrawn at increasing times and quenched by mixing with 10 l of 10 mg/ml bovine serum albumin, followed immediately by 7 l of 40% trichloroacetic acid. Tubes were chilled for 5 min on ice, and the precipitated proteins were collected by centrifugation. Acid-soluble radioactivity was determined by counting an aliquot of the supernatant in a liquid scintillation counter. We confirmed that unmodified 35 S-UbDHFR is inert to proteasomes (13). A saturating concentration of unanchored Lys-63-Ub 4 did not stimulate the degradation of unmodified UbDHFR, indicating that the chain cannot signal degradation in trans. A similar result was obtained previously with Lys-48-Ub 4 (13).

Catalytically Active Mms2⅐Ubc13 Complex Is a Heterodimer-We reported previously that recombinant yeast
Mms2 and Ubc13 co-eluted from a gel filtration column at an apparent molecular mass of ϳ35 kDa, suggesting that the two proteins form a heterodimeric complex (22). Several new observations now confirm this stoichiometry. First, the two proteins can be co-purified as an apparent 1:1 complex. The Mms2-and Ubc13-encoding cDNAs were cloned into the bacterial expression vectors pET16b and pET3a, respectively (22), and the proteins were individually expressed (each at a low level; Fig.  1A, lanes 1 and 2). Preliminary control experiments showed that His 10 Mms2 bound efficiently to nickel resin, whereas Ubc13 did not bind detectably (data not shown). When the two lysates were mixed (lane 3) and passed through a nickel column, untagged Ubc13 bound to the column by virtue of its association with His 10 Mms2 and could be eluted with imidazole. The two proteins also co-eluted during subsequent gradient elution from an anion exchange column (lane 5), although they eluted at distinct positions when chromatographed individually ("Experimental Procedures"). Based on the Coomassie staining intensities of the two bands, His 10 Mms2 and Ubc13 appeared to purify in a 1:1 ratio (Fig. 1A) (for simplicity, we refer to the His 10 -tagged protein as Mms2 in the remainder of the paper and number residues according to the native protein sequence.).
We also purified Mms2 and Ubc13 separately to provide an independent estimate of the stoichiometry of the catalytically active complex. We titrated increasing concentrations of Mms2 against a fixed concentration of Ubc13 (4 M) and measured the rate constant for a single turnover of Ub 2 synthesis by pulsechase ("Experimental Procedures"). The concentration of Mms2 was determined by UV absorbance, whereas the concentration of Ubc13 was determined by quantitative Ub thiol ester assay (see "Experimental Procedures"). Plotting the rate constant for Ub 2 synthesis against the concentration of Mms2 showed that k obsd reached saturation when the two proteins were present in approximately a 1:1 ratio (Fig. 1B). The limiting value of k obsd seen in Fig. 1B, 1.6 min Ϫ1 , agrees reasonably well with the value of 1.1 min Ϫ1 predicted from studies described in the next paragraph. These results are most simply explained if the catalytically active complex is a heterodimer. This conclusion is consistent with the apparent molecular mass of the complex determined by gel filtration (22) with our finding that the two proteins are readily cross-linked by dimethylpimelimidate to yield a product of ϳ40 kDa (data not shown) and with preliminary x-ray crystallographic results. 2 A fraction of the intracellular pool of each protein is present in a complex in yeast cells together with several other components of the RAD6 repair pathway (see "Discussion" (25)).
Kinetics of Chain Assembly-The experiment shown in ϭ 0.2 min Ϫ1 , and n H ϭ 2 ( Fig. 2A). An independent experiment with a different preparation of the complex gave a similar value of K m,app and a slightly larger value of k cat (ϳ0.5 min Ϫ1 , data not shown). In a second set of studies, we used pulse-chase assays to measure the rate constant for a single round of Ub 2 synthesis as a function of acceptor (Asp-77) Ub concentration. The steady-state and pulse-chase assays should give similar results if the chemical step (isopeptide bond formation) is ratelimiting in the steady state. Instead, the dependence observed in the pulse-chase assays was hyperbolic and could be fitted assuming K m ϭ 437 M and k cat ϭ 2.2 min Ϫ1 (Fig. 2B) (this k cat value predicts a turnover number of 1.1 min Ϫ1 at [S] ϭ K m , in reasonable agreement with the value observed in Fig. 1B). The 4 -10-fold larger k cat value observed in the pulse-chase (versus steady-state) assay suggests that a non-chemical step could contribute to rate limitation in the steady state. Further studies will be necessary to confirm this inference and to determine the basis of the apparent cooperativity seen in steady-state conjugation.
The steady-state K m,app is ϳ10-fold higher than the concentration of Ub in cells, suggesting that chain assembly by the isolated complex would occur at a slow rate in vivo. However, chain assembly could occur more rapidly than expected from these data if chains are favored as substrates over Ub 1 . Such an effect would not have been detected in Fig. 2A because Ub 2 is the terminal product in the steady-state assay. As a qualitative test of processivity, we used wild-type Ub as a substrate for the assembly of long chains. We used E2-25K, a non-processive enzyme that produces Lys-48 chains (41, 42) as a negative control to define the product spectrum expected for a nonprocessive reaction. The reactions were carried out at a subsaturating concentration of substrate, with enzyme concentrations normalized so that the initial rate of Ub 1 consumption was identical in each reaction. As shown in Fig. 3A, E2-25K assembled mainly chains of n Յ 3 over a 60-min incubation (lanes 7-12), whereas most chains assembled by Mms2⅐Ubc13 were four or more Ubs in length (lanes 1-6). Thus, Mms2⅐Ubc13 catalysis favors the formation of longer chains, at least when compared with catalysis by E2-25K. This result suggests that there is modest processivity in chain assembly catalyzed by the complex. Further studies will be necessary to determine whether this behavior reflects slow product dissociation (allowing for multiple turnovers per binding event) or a length-dependent increase in chain affinity (leading to preferential binding of longer acceptors).
Disassembly of Lys-63 Chains-Assembly of unanchored chains by the Mms2⅐Ubc13 complex could acquire added sig- nificance if Lys-63 chains are more stable than other chains. Isopeptidase T (Ubp14p in yeast) is the principal enzyme that disassembles unanchored Lys-48 chains in vivo (33,43). To determine if isopeptidase T can act on Lys-63 chains, we used the purified Mms2⅐Ubc13 complex to assemble Lys-63-Ub 4 (see "Experimental Procedures"). As shown in Fig. 3B, Lys-63-Ub 4 and Lys-48-Ub 4 were similarly susceptible to disassembly over a range of concentrations of purified isopeptidase T (lanes 1-6 versus [7][8][9][10][11][12]. This result is in agreement with a report that stable analogs of Lys-63-Ub 2 and Lys-48-Ub 2 are similarly potent inhibitors of chromogenic substrate hydrolysis by this enzyme (44). The absence of a linkage preference in isopeptidase T catalysis contrasts with the absolute linkage specificity shown by several E2s and E3s (22,41,45,46). Competitive inhibition of proteasomes by unanchored Lys-48 chains provides a biological rationale for the rapid disassembly of these chains by isopeptidase T (13,43). The same rationale probably applies to Lys-63 chains (see Fig. 5

below).
Structure/Function Studies-In considering the mechanism by which Mms2 (22) and its mammalian relative Uev1 (47) positively regulate Lys-63 chain assembly, we noted that several UEVs harbor a conserved cysteine residue (Cys-64 in Mms2) at a site distinct from the canonical E2 active site (24). As a first test of whether this conserved cysteine acts as a site of thiol ester formation with Ub, we determined if Mms2 was sensitive to the thiol-alkylating agent NEM. Mms2 was completely inactivated by a low concentration of NEM (Fig. 4A, lane 2); in fact, Mms2 was inactivated by lower concentrations of NEM than was Ubc13 (data not shown). However, mutating Cys-64 to alanine did not inactivate Mms2 (Fig. 4A, lane 3); moreover, C64A-Mms2 remained sensitive to NEM (lane 4). These results exclude Cys-64 as a catalytically relevant site and indicate that inactivation by NEM involves a residue other than Cys-64. The other two cysteines of Mms2, Cys-26 and Cys-85, are poorly conserved among UEVs. This fact and our inability to observe an Mms2ϳUb thiol ester (not shown) make it unlikely that Cys-26 or Cys-85 is a site of Ub thiol ester formation. To address whether NEM affects heterodimer formation, we mixed NEM-treated (wild-type) Mms2 with Ubc13 and subjected the mixture to gel exclusion chromatography. The (inactive) Mms2 eluted in the void volume, whereas Ubc13 eluted at the position expected for the uncomplexed E2 (data not shown). This result suggests that alkylation by NEM leads to the formation of soluble Mms2 aggregates (NEM treatment did not cause Mms2 to precipitate).
We have proposed that the Mms2⅐Ubc13 heterodimer functions in the assembly of Lys-63 chains for DNA repair in the RAD6 pathway (22). If this model is correct, in vitro chain assembly activity should correlate with in vivo DNA repair activity. We tested this hypothesis using the original mms2 mutant allele, which specifies a proline-to-leucine mutation at the highly conserved residue Pro-73 (23). We tested purified P73L-Mms2 in a chain assembly assay with purified Ubc13 and found that the mutant Mms2 was devoid of activity (Fig. 4B). Thus, the in vitro chain assembly activity of Mms2 correlates with in vivo activity in DNA repair. When the P73L-Mms2⅐Ubc13 complex was analyzed by gel filtration, the results were similar to those obtained with NEM-treated Mms2 (see preceding paragraph). Evidently, the P73L mutation also caused Mms2 to form soluble aggregates.
Lys-63 Chain Signals Proteasomal Degradation-Although yeast cells expressing the K63R mutant as the sole form of Ub degrade short-lived proteins at a normal rate (9), inhibition of the turnover of a small number of proteins would not have been detected in these whole-cell pulse-chase studies. Our ability to assemble Lys-63 chains in vitro and the availability of a model substrate (UbDHFR (13)) allowed us to test rigorously whether a Lys-63 chain can signal degradation by proteasomes. We used the Mms2⅐Ubc13 complex to assemble Lys-63-Ub 4 ("Experimental Procedures" ; Fig. 3A). We intended to use the complex to conjugate this chain to the fused Ub of UbDHFR, but we were unable to do so because the activity of the complex was severely inhibited by the N-terminal His 10 tag of UbDHFR. 3 We therefore used the Lys-48-specific enzyme E2-25K to conjugate Lys-63-Ub 4 to 35 S-UbDHFR. In the resulting substrate, termed Lys-63-Ub 5 DHFR, the Lys-63 chain is linked to Lys-48 of the fused Ub moiety. The results of previous studies suggested that in a similar substrate carrying a Lys-48 chain (Lys-48-Ub 5 DHFR), the fused Ub moiety of UbDHFR is recognized as part of the chain (13). If the same is true of Lys-63-Ub 5 DHFR, then the "chain" has one proximal Lys-48 linkage and three distal Lys-63 linkages. We reasoned that this single discrepant linkage would not affect the qualitative outcome of the degradation experiment because Lys-48-Ub 2 is not detectably recognized by 26 S proteasomes (13). We tested the validity of this assumption as described below.
Degradation of Lys-63-Ub 5 DHFR was readily detectable; it was linear with time and directly proportional to the concentration of 26 S proteasomes (data not shown). This substrate was degraded with a saturating concentration dependence (Fig.  5A, circles). Its K m value, 70 nM, was ϳ4-fold larger than the K m of Lys-48-Ub 5 DHFR determined in the same experiment, whereas the two substrates displayed similar V max values (Fig.  5A, circles versus triangles).
We employed a competition approach to address whether the single Lys-48 linkage in Lys-63-Ub 5 DHFR caused an artificial gain-of-function effect. Unanchored Lys-48-Ub 4 inhibits the degradation of Lys-48-Ub 5 DHFR by displacing the substratelinked chain from its binding site on the 26 S proteasome (13). Lys-63-Ub 4 and Lys-48-Ub 4 similarly inhibited the turnover of Lys-48-Ub 5 DHFR (Fig. 5B), indicating that a Lys-63 homopolymer can indeed bind to 26 S proteasomes, most likely at the same site that binds Lys-48 chains. Consistent with this interpretation, Lys-63-Ub 4 inhibits the cross-linking of Lys-48-Ub 4 to the 19 S regulatory complex of the proteasome. 4 Also, linear Ub 5 , which may resemble a Lys-63 chain, competitively inhibits Lys-48-Ub 5 DHFR degradation (13) (Met-1 and Lys-63 are spatially adjacent on the Ub surface (48)). We are uncertain which of the two approaches used in Fig. 5 most accurately reports the relative affinities of Lys-63 and Lys-48 chains. The results of the degradation assay (Fig. 5A) suggest that the Lys-63 chain binds ϳ4-fold more weakly than the Lys-48 chain, but we cannot rule out that the affinity of the substrate-linked Lys-63 chain was weakened by the Lys-48 linkage between the chain and UbDHFR. The results of the competition assay (  5B) suggest that the two chains bind with similar affinities. However, the combined data leave no doubt that a Lys-63 chain is a competent signal for degradation by proteasomes.
Evidence against Degradative Signaling by Lys-63 Chains in Vivo-This unexpected in vitro result (Fig. 5) led us to test more rigorously whether Lys-63 chains function as a degradation signal in post-replicative repair. If proteolysis by proteasomes is required in the RAD6 repair pathway, then mutating the proteasome active sites should make yeast cells UV-sensitive, and these mutations should map to the RAD6 epistasis group. Although proteasome mutants are weakly UV-sensitive (28,29), this phenotype is not very informative because the UV sensitivity of mms2 or ubc13 mutants is also quite modest (22,25,35,49).
The weak phenotypes of mms2 or ubc13 mutants are explained by the existence of at least two branches of RAD6-dependent DNA repair, an error-prone (mutagenesis) pathway and one or more error-free pathways (16,17,23,35,49). Mms2 and Ubc13 function in an error-free repair pathway, whereas Rev3 is part of the error-prone pathway (16,17,22,23,49,50). Mutations in MMS2 or REV3 (an error-prone pathway gene) that inactivate only one type of repair confer a weak phenotype because lesions are repaired through the remaining active pathway (Fig. 6, open squares and open triangles versus open  circles). Simultaneous inactivation of both branches has a synergistic effect, generating a UV sensitivity similar to that seen in rad6 or rad18 mutants (Fig. 6, filled squares versus open  squares and open triangles (23, 35, 49)). Mutations in RAD6 or RAD18, which disable both branches, make cells extremely UV-sensitive (16).
To enhance our ability to detect proteolytic signaling by Lys-63 chains in DNA repair, we tested the effect of inhibiting proteasome active sites in a yeast strain in which error-prone repair had been disabled through deletion of REV3. Proteasome inhibition was accomplished through mutations in the PRE1-and PRE2-encoded 20 S subunits; these mutations severely inhibit the proteasome chymotryptic peptidase activity in vitro and retard the degradation of natural and model proteasome substrates in vivo (36,(51)(52)(53)(54). Although the pre1-1 and pre2-2 mutations are maximally inhibitory at 38°C, inhi-bition of the turnover of many substrates is readily detectable at 30°C (36,52,54). We were constrained to work at 30°C so as to preclude stress effects during the long incubations required for colony formation after UV treatment.
If the primary function of Lys-63 chains in repair is to target specific substrates for degradation by proteasomes, then mutation of the proteasome active sites should have a similar effect to blocking chain assembly through deletion of Ubc13. This model predicts that the UV sensitivity of a pre1-1pre2-2rev3 triple mutant should be similar to that of a ubc13rev3 double mutant. In contrast to this prediction, the pre1-1pre2-2rev3 triple mutant was Ͼ10-fold less UV-sensitive than the ubc13rev3 double mutant (Fig. 6, filled triangles versus filled squares), and the increase in sensitivity from combining the pre1-1pre2-2 and rev3 mutations was additive rather than synergistic (Fig. 6, compare filled triangles to open triangles and filled circles). This conclusion is not definitive because the pre1-1pre2-2 strain retains some proteolytic activity, as indicated by its viability. Nonetheless, the epistasis data ( Fig. 6) are fully consistent with the absence of a proteolytic phenotype in ubiK63R cells (9). The combined in vivo results argue strongly that the proteolytic activity of proteasomes is not required for error-free repair. DISCUSSION Biologically, Mms2 functions in error-free post-replicative repair (23); biochemically, it forms a complex with Ubc13 and confers activity in the assembly of Lys-63 chains (22). However, although Mms2 is arguably the best-characterized member of the UEV protein family, its mechanism of action remains poorly understood. One goal of the studies described here was to probe the relationship between the known biochemical activity of the Mms2⅐Ubc13 complex, namely Lys-63 chain assembly, and the biological role of the complex in DNA repair. A second goal was to rigorously test one potential mechanistic function of the Lys-63 chain signal in DNA repair, that of a signal for proteolysis by proteasomes.
The titration study shown in Fig. 1B indicates that the chain synthesis activity of the Mms2⅐Ubc13 complex is maximal when the two proteins are present in a 1:1 ratio. The two recombinant proteins co-purify in the same apparent ratio (Fig.  1A) and migrate on a size exclusion column at the position expected for a heterodimer (22). Thus, the catalytically active form of the complex is a heterodimer in vitro. Genetic data  6. UV sensitivity of 20 S proteasome mutants. Assays were carried out as described previously (22)  indicate that MMS2, UBC13, and Lys-63 of Ub all participate in the error-free post-replicative repair pathway (9,22,23,49,50). We have proposed that the biological function of the Mms2⅐Ubc13 complex is to facilitate the assembly of Lys-63 chains, which then act as a signal in the DNA repair pathway. Consistent with the idea that the chain assembly activity of the complex is relevant in DNA repair, the P73L mutation in Mms2, which abrogates the activity of Mms2 in DNA repair in vivo (23,55), also abolishes activity in chain assembly in vitro (Fig. 4B). Ulrich and Jentsch show that a fraction of the Ubc13 and Mms2 in yeast cells are present in a nuclear complex (25). These genetic and biochemical data suggest that the Mms2⅐Ubc13 complex is the active entity in DNA repair, although a rigorous confirmation of this model awaits the availability of Mms2 or Ubc13 mutants that are specifically impaired in complex formation.
Recently, a heterodimer composed of the human Uev1A and Ubc13 proteins was shown to assemble Lys-63 chains in a signaling pathway that activates IB␣ kinase (47). Thus, in higher cells Ubc13 partners with more than one UEV protein to assemble Lys-63 chains for distinct purposes. As discussed below, UEV/Ubc13 complexes may be dedicated to the assembly of Lys-63 chains, with specificity of function (e.g. kinase activation versus DNA repair) dictated by the properties of specific partner proteins or by intracellular localization.
So far the only known substrate of the Mms2⅐Ubc13 complex is Ub itself. Although it is formally possible that the biological function of the complex is to assemble unanchored Lys-63 chains that are then used by a different E2/E3 enzyme, such a role appears unlikely because the kinetic K m for Ub is ϳ10-fold higher than the intracellular concentration of Ub ( Fig. 2A). Processivity in chain assembly could help to overcome this unfavorable K m , but the observed processivity is modest (Fig.  3A). Induction of Ub expression by DNA damage could also ameliorate the effects of a high K m , but this induction is likewise modest (56,57). Therefore we favor a model in which the Mms2⅐Ubc13 complex acts in conjunction with another factor that helps to overcome the unfavorable Ub binding revealed by our studies. In this model the Mms2⅐Ubc13 complex functions as a special catalytic module that promotes the assembly of a Lys-63 chain on a substrate selected by a specific partner protein. Whether this reaction involves unanchored chains as intermediates is an open question.
Such a partner protein has been identified in the IB␣ kinase activation pathway, where the RING finger protein Traf6 acts together with the Uev1A⅐Ubc13 complex to assemble Lys-63 chains (47). Recent findings suggest that the Rad5 helicase, which functions in error-free repair (49,58), may be biochemically analogous to Traf6; Rad5 can interact through its RING finger with Ubc13 in the Mms2⅐Ubc13 complex (25). A biological significance for these interactions is suggested by the finding that Mms2, Ubc13, and Rad5 can be co-immunoprecipitated from yeast cell extracts and co-localize to chromatin in yeast cells treated with DNA-damaging agents (25). Besides Ubc13, Rad5 also interacts with Rad18 in a manner that is permissive for the Rad6-Rad18 interaction (25). A multicomponent E2⅐E3 complex may thus be present at single-stranded DNA sites recognized by Rad18 (25). An attractive but speculative model postulates that Rad18 recruits Rad5 to DNA damage sites, where the latter is modified with a Lys-63 chain that serves as an allosteric modulator of helicase activity or as a signal to recruit an additional factor(s) to the damage site (see below).
The Mms2⅐Ubc13 complex is a model for what is probably a small number of related UEV/E2 heterodimers. The complex may also represent a tractable system for studying catalysis and specificity in polyUb chain assembly more generally. We speculate that complex formation creates a scaffold upon which the acceptor Ub binds so that Lys-63 is positioned to efficiently attack Gly-76 of the Ubc13-bound donor Ub. A similar model for E3 catalysis has been proposed based on the crystal structure of the c-Cbl RING-UbcH7 complex (59), which revealed no obvious mechanism for rate enhancement besides the possible spatial juxtaposition of the substrate lysine and the UbcH7bound Ub. This is an especially attractive model for the Mms2⅐Ubc13 complex, which, unlike most E3s, targets just one lysine residue of its (Ub) substrate.
The K63R mutation in Ub inhibits error-free repair without detectably inhibiting proteolysis (9). The conjugation of L28 to a Lys-63 chain does not target L28 for degradation (15), and signaling by Lys-63 chains in IB␣ kinase activation can be reconstituted in vitro in the definitive absence of proteasomes (47). We therefore expected the apparent irrelevance of proteasomes in Lys-63 chain signaling in vivo to have its origin in an inability of proteasomes to recognize these chains. Instead, Lys-63-Ub 4 is a competent proteasomal degradation signal (Fig. 5). However, despite the intrinsic ability of the Lys-63 chain to signal proteolysis by proteasomes, we think it unlikely that Lys-63 chains play this role in DNA repair. First, competition effects probably limit the ability of Lys-63 chains to signal degradation in vivo. Studies of yeast expressing specific lysine-to-arginine Ub mutants suggest that only a small fraction of Ub-conjugated substrates carry alternatively-linked chains in vivo (9). In the specific case of Lys-63, nearly all the detectable chains are linked to L28 (9,15). Therefore the total concentration of Lys-63 chains linked to substrates relevant to DNA repair must be extremely low. These substrates will thus compete poorly with the abundant Lys-48 chains linked to normal proteolytic substrates; this effect will be magnified if Lys-63 chains bind somewhat more weakly to proteasomes (see "Results"). Second, the available biological data ( Fig. 6 and (9)) argue strongly against a model in which Lys-63 chains signal degradation in DNA repair. Overall, it appears more likely that Lys-63 chains act as a signal in DNA repair (and IB␣ kinase activation) by modulating the function of the substrate protein (60) or recruiting additional factors that facilitate repair. This latter function is analogous to the function of Lys-48 chains in recruiting 26 S proteasomes for substrate proteolysis. If this model is correct, the postulated recognition factors may sequester Lys-63 chains and further reduce the efficiency with which these chains can signal proteolysis.
The proteolytic signaling competence of Lys-63 chains also has implications for polyUb chain recognition by 26 S proteasomes. The results of a previous analysis of the Lys-48 chain signal led us to suggest that the recognition of this signal depends on a quarternary arrangement of the Ub molecules that displays hydrophobic residues of Ub at specific locations on the chain surface (13,14). If this model is correct, then Lys-63-Ub 4 must be able to adopt a similar conformation. The conformational plasticity of Lys-48 chains (61-64) perhaps argues in favor of this possibility, but a definitive resolution awaits structural studies of different chains bound to their cognate recognition factors.