|Home | About | Journals | Submit | Contact Us | Français|
Ubiquitin (Ub)-conjugating enzymes (E2s) and ubiquitin ligases (E3s) catalyze the attachment of Ub to lysine residues in substrates and Ub during monoubiquitination and polyubiquitination. Lysine selection is important for the generation of diverse substrate-Ub structures, which provides versatility to this pathway in the targeting of proteins to different fates. The mechanisms of lysine selection remain poorly understood, with previous studies suggesting that the ubiquitination site(s) is selected by the E2/E3-mediated positioning of a lysine(s) toward the E2/E3 active site. By studying the polyubiquitination of Sic1 by the E2 protein Cdc34 and the RING E3 Skp1/Cul1/F-box (SCF) protein, we now demonstrate that in addition to E2/E3-mediated positioning, proximal amino acids surrounding the lysine residues in Sic1 and Ub are critical for ubiquitination. This mechanism is linked to key residues composing the catalytic core of Cdc34 and independent of SCF. Changes to these core residues altered the lysine preference of Cdc34 and specified whether this enzyme monoubiquitinated or polyubiquitinated Sic1. These new findings indicate that compatibility between amino acids surrounding acceptor lysine residues and key amino acids in the catalytic core of ubiquitin-conjugating enzymes is an important mechanism for lysine selection during ubiquitination.
Protein ubiquitination plays a fundamental role in most cellular processes and involves three classes of enzymes (22). First, the 8-kDa protein ubiquitin (Ub) forms a thioester bond with the E1 Ub-activating enzyme. Ub is then transferred from E1 to the active-site cysteine of E2. Finally, E2s in conjunction with an E3 transfer Ub to a substrate lysine to form an isopeptide bond. E3 ligases are important for substrate recognition, and two major families exist. The RING (really interesting new gene) finger E3s, which lack catalytic activity, recruit the substrate and E2 into one complex to facilitate ubiquitination (19). Catalytic HECT (homologous to E6-AP carboxyl terminus) E3s accept Ub from E2 via a catalytic cysteine and then transfer Ub to a substrate lysine (22). The versatility of Ub in regulating different processes derives from its ability to be conjugated as a monomer (monoubiquitination) or polymer (polyubiquitination) to substrate lysines. Since Ub contains seven lysines, polyubiquitination can generate chains with different topologies (18). Monoubiquitination can regulate DNA repair, viral budding, trafficking, and gene expression (17). Polyubiquitination through Ub K11, K29, or K48 results in proteasomal degradation, while K63-linked Ub chains function in kinase activation, DNA damage tolerance, signal transduction, and endocytosis (17). In RING E3s the mode of ubiquitination (mono- or polymeric) and Ub linkage specificity are determined by E2, while HECT E3s specify the mode of ubiquitination with this family (11, 12).
The mechanisms that control mono- or polyubiquitination, chain topology, and lysine selection are poorly understood. Insights into this area have come from studies of the RING E3 ligase Skp1-Cdc53/cullin F-box (SCF) protein and its cognate E2, Cdc34, (19). One critical substrate of the budding yeast SCFCdc4/Cdc34 complex (superscript indicates the F-box protein) is Sic1. Sic1 is an inhibitor of the cyclin-dependent kinase (CDK) Cdc28 (29), containing a CDK-inhibitory domain within its C-terminal 70 amino acids (9). In Saccharomyces cerevisiae, G1-S-phase cell cycle progression depends on the SCFCdc4/Cdc34-mediated polyubiquitination of Sic1 on at least one of the six N-terminal lysine residues (K32, K36, K50, K53, K84, and K88) via K48-linked Ub chains, leading to its proteasomal degradation (9, 20). A Sic1 mutant missing these six lysines is still ubiquitinated on the remaining 14 lysines in vitro but is not degraded and is stable in vivo, leading to cell cycle arrest (20).
While a Ub chain on any one of six Sic1 N-terminal lysines sustains proteolysis, the turnover rates vary by 5-fold between the K36 and K84 sites, indicating that the context of the Ub chain is important (20). How SCFCdc4/Cdc34 ubiquitinates different Sic1 lysines and whether a lysine preference exists are poorly understood, but a “positioning model” has been proposed, where E3 positions substrate lysines favorably for the attack of the E2~Ub thioester bond (19, 37). Previous studies suggested that the number of substrate ubiquitination sites correlates with the number of its F-box protein binding sites (8, 33, 37). For example, the ubiquitination of β-catenin and IκBα is directed by a single high-affinity F-box protein binding site, and changes to the distance between the binding site and the lysine affect catalysis by ~2-fold (37). The phosphorylation of Sic1 on multiple CDK sites at the N terminus generates several low-affinity Cdc4 binding sites that dynamically bind Sic1 to the single site of Cdc4 in various geometries, leading to ubiquitination on numerous lysines. The directed binding of Sic1 to Cdc4 through a single artificial high-affinity binding site at the extreme N terminus confines efficient ubiquitination to lysines K84 and K88 (33). Sic1 lysine selection flexibility may also be achieved by the dimerization of the RING E3/E2 complex (8, 33). In addition, the neddylation of the cullin subunit induces conformational variation to the E2, enhances the recruitment of E2 to SCF, and brings substrate lysines toward the Cdc34~Ub thioester to accommodate the ubiquitination of different lysines (6, 25). Lysine selection flexibility may also be achieved by the release of the ubiquitin-charged Cdc34~Ub from SCF to transfer Ub to different substrate lysines, as proposed by the “hit-and-run” hypothesis (5). Apart from higher-order structures contributing to lysine selection flexibility, studies with the human anaphase-promoting complex (APC/C) RING E3 and its E2, UbcH10, have identified a sequence motif adjacent to acceptor lysines, termed the TEK box, which is important for lysine selection (11).
Similar to the E3-mediated positioning of substrate lysines, structural aspects of E2s position Ub lysines to generate Ub chains of different topologies by RING E3/E2 complexes (11, 12). Structural studies of Mms2/Ubc13-Ub demonstrate that this complex assembles so that K63 of Ub is positioned proximal to the Ubc13-Ub thioester bond during Ub chain formation (35). Similarly, Cdc34 dimerization, a noncovalent Ub binding domain (UBD), and structural constraints such as an acidic loop region were suggested previously to position K48 of Ub for the attack of the Cdc34~Ub thioester bond (21, 36). However, E2s such as human UbcH5 utilize all Ub lysines but display a preference for K11, K48, and K63, indicating less structural constraint and that other mechanisms contribute to Ub lysine preference (12).
To further elucidate the mechanisms of lysine specificity, we assessed the SCFCdc4/Cdc34-mediated ubiquitination of Sic1's N-terminal lysines and K48 in Ub. SCFCdc4/Cdc34 displayed a strong preference for Sic1 K53. Changes to proximal amino acids regulated the rate of ubiquitination in Sic1 and K48 of Ub. This was linked to key residues composing the catalytic core of Cdc34. The mutation of these core sites differentially affected Cdc34 preference toward lysines on Sic1 or Ub K48. Our studies demonstrate that in addition to the importance of the SCF/Cdc34-mediated positioning of lysines, efficient catalysis is dependent on residues surrounding ubiquitinated lysines and their compatibility with the Cdc34 catalytic core residues. We propose that this compatibility is an important mechanism for lysine selectivity during ubiquitination.
The plasmids and yeast strains used in this study are described in Tables Tables11 and and2.2. Mutations were introduced by site-directed mutagenesis according to the manufacturer's instructions (QuikChange; Stratagene) and confirmed by DNA sequencing. To assess the ability of the Cdc34 mutants to support growth, Δcdc34 deletion strain YMS034 containing maintenance plasmid pURA3-CDC34 was transformed with tester plasmids (pLEU2) expressing wild-type Cdc34(YSA) or the mutant alleles under the control of the CDC34 promoter. Cells were grown on medium selective for both plasmids or on medium selective for the tester plasmid and the loss of the maintenance plasmid (5-fluoroorotic acid [5-FOA]) (4) for 3 days at 30°C. For drop test analysis, yeast cells were grown to an optical density at 600 nm (OD600) of 0.8 at 30°C. Serial 10-fold dilutions were spotted onto plates and incubated at the indicated temperatures for 2 to 5 days.
Strain YMS035 was transformed with the indicated single-lysine Sic1 derivatives in plasmid p416Gal1 under the control of the strong GAL1 promoter. A drop test analysis was performed with plates supplemented with either glucose (2%, wt/vol) or galactose (2%, wt/vol) for 2 to 5 days at 30°C. For analysis of Sic1 stability, the expression of Sic1 was induced in exponentially growing cells with galactose (2%, wt/vol) for 4 h at 30°C. Concurrently, the cells were arrested in G1 phase by adding 15 μg/ml α-factor for 2 h and then adding a further 15 μg/ml α-factor for the remaining 2 h. Sic1 expression was stopped, and cells were released into the cell cycle by centrifugation to collect the cells and then resuspending them in yeast extract-peptone-dextrose (YPD) medium. Samples were taken at the indicated times, and cell extracts were prepared and analyzed by immunoblotting with His5 antibody to detect His6-tagged Sic1, Clb5 (sc-6704; Santa Cruz), and actin antibody as a loading control. For quantitation, Sic1 levels were normalized according to actin levels and are expressed as the remaining Sic1 relative to the time point of 0 min (100%).
Recombinant Cdc34, Sic1, and Ub were expressed in Escherichia coli strain Rosetta BL21(DE3)pLysS (Novagen). Cultures were grown in LB with 100 μg/ml ampicillin and 50 μg/ml chloramphenicol at 37°C to an OD600 of 0.8, and expression was induced by the addition of 0.75 mM isopropyl-β-d-thiogalactoside (IPTG) for 3 h at 37°C. Cells were harvested and lysed, and proteins were purified by affinity chromatography on Ni-nitrilotriacetic acid (NTA) agarose (Qiagen) as described previously (24). Recombinant SCFCdc4 was prepared as described previously (24).
CDK-mediated phosphorylation to generate 32P-labeled Sic1 was performed as described previously (24, 26). Ub was phosphorylated and 32P labeled by cyclic AMP (cAMP)-dependent protein kinase (PKA; NEB) on the PKA motif in the presence of [γ-32P]ATP. Cdc34 was phosphorylated by protein kinase CK2 (NEB) in the presence of [γ-32P]ATP (24).
Ubiquitination reaction mixtures contained 100 nM yeast E1 (AG Scientific), 1 μM Cdc34, ~25 nM SCFCdc4, 40 μM Ub or Ub(K0) (Boston Biochem), and 0.15 μM 32P-labeled Sic1 in UB buffer (50 mM Tris-HCl [pH 7.5], 50 mM potassium acetate, 2.5 mM magnesium acetate, 1 mM dithiothreitol [DTT], 2 mM ATP). After incubation for 60 min at 26°C, reactions were stopped by the addition of SDS-Laemmli buffer to the mixtures and boiling, and proteins were separated by SDS-PAGE. Sic1 was visualized by autoradiography and quantified by PhosphorImager analysis. To prepare monoubiquitinated 32P-labeled Sic1 K32 and 32P-labeled Sic1 K53, reactions were carried out as described above, with 1.5 μM Cdc34(YDA) and Ub or Ub(K48o). After SDS-PAGE, monoubiquitinated 32P-labeled Sic1 K32 and 32P-labeled Sic1 K53 were eluted from gel slices in 0.1 M Tris-HCl (pH 6.8) at 23°C overnight, concentrated by acetone precipitation in the presence of 5% (wt/vol) sucrose, and resuspended in UB buffer. As a control, 32P-labeled Sic1 K53 was prepared in the same way except that Ub(K48o) was omitted from the preubiquitination reaction.
Reaction mixtures contained 150 nM yeast E1 (AG Scientific), 1 μM Cdc34, and 15 μM 32P-labeled Ub in UB buffer. For the autoubiquitination of 32P-labeled Cdc34, 0.5 μM Cdc34 and 40 μM Ub were employed. After incubation for 60 min at 26°C, reactions were stopped by the addition of SDS-Laemmli buffer to the mixtures and boiling. Proteins were separated by SDS-PAGE and visualized by autoradiography.
It is not clear if Sic1's physiologically relevant N-terminal lysines (K32, K36, K50, K53, K84, and K88) are differentially ubiquitinated by SCFCdc4/Cdc34. Hence, we compared the monoubiquitinations of phosphorylated Sic1 derivatives that contained only a single lysine at these positions with SCFCdc4/Cdc34 and lysine-less Ub, Ub(K0), to prevent Ub chain elongation. Monoubiquitinations of the Sic1 mutants varied substantially. While the K53 variant displayed the highest level of ubiquitination, K32, K36, K50, K84, and K88 were substantially less ubiquitinated (normalized activities relative to those of K53 ± standard deviations [SD] of 17.0% ± 4.0% for K32, 15.3% ± 5.9% for K36, 56.2% ± 3.6% for K50, 100% for K53, 24.8% ± 11.8% for K84, and 23.0% ± 3.1% for K88) (Fig. (Fig.1A1A).
Previous work has shown that altering the spacing between the ubiquitinated lysine and an F-box binding site in the substrate by 4 to 8 residues affected the rate of ubiquitination by ~2-fold, with a lysine positioned 9 or 10 residues away being optimal (37). We examined whether the spacing between the five N-terminal CDK phosphorylation sites that are important for Cdc4 binding (T5, T45, S69, S76, and S80) of Sic1 (14) and the six lysines may explain the differences in their levels of ubiquitination (Fig. (Fig.1B).1B). The binding of Sic1 to Cdc4 requires the phosphorylation of at least five of its seven N-terminal CDK sites, trapping Sic1 on the single receptor site of Cdc4 (14-16). All the single-lysine Sic1 derivatives in this study displayed the characteristic slower-migrating electrophoretic mobility similar to that of fully phosphorylated wild-type Sic1 (Sic1 Wt) or mutant lysine-less Sic1 (20, 21), indicating similar phosphorylation levels. In addition, the lysine-less Sic1 mutant binds SCFCdc4 with the same affinity as that of Sic1 Wt (21). Examination of the spacing between the ubiquitinated lysines and the different CDK phosphorylation sites (Fig. (Fig.1B)1B) revealed no obvious relationship to their degree of ubiquitination. For example, while K36 and K53 are positioned at a similar distance relative to T45 (9 and 8 residues), K53 is ubiquitinated 6.5-fold-more efficiently than K36. In addition, K84 and K88 are 8 residues upstream of S76 and S80, respectively, which is identical to the spacing between T45 and K53, yet the ubiquitination of K84 and K88 is 4-fold lower than that of K53 (Fig. (Fig.1A).1A). Given the highly dynamic equilibrium binding of the phospho-Sic1/Cdc4 interaction (14-16), it is plausible that the ubiquitination level of a particular lysine is not determined by its spacing to a single low-affinity Cdc4 binding site. However, it is also possible that ubiquitinated residues located further from the phosphorylation sites in the primary sequences may actually be located closer in the tertiary structure. These observations prompted us to investigate if other determinants may be important; e.g., whether the local amino acid sequence environment of the lysine contributes to the catalysis efficiency remains uncharacterized. Therefore, we changed the flanking residues of poorly ubiquitinated Sic1 lysines (K32, K84, and K88) to those of efficiently ubiquitinated sites (K50 and K53) and vice versa. The rationale was to determine if keeping the lysine position unchanged but altering its proximal amino acids can alter the level of ubiquitination. We generated a set of single-lysine Sic1 derivatives, including K53-like K32(Q31F/T33N) and K88(V87F/R89N), K50-like K88(V87T/R89S), K84-like K53(F52P/N54S) and K50(T49P), and K88-like K53(F52V/N54R) and K50(T49V/S51R). We also generated the K32 derivatives K32(Q31D/T33D), K32(T33D), and K32(G30D) based on our observation that a lysine embedded in an aspartate-rich sequence motif (DDKDP) was efficiently ubiquitinated by Cdc34 (data not shown). While the K53-like K32(Q31F/T33N) derivative and the T33D and Q31D/T33D mutants displayed between 2- and 3-fold-higher K32 ubiquitination levels, the G30D mutation reduced activity by 2-fold relative to that of K32 within the wild-type sequence [normalized activities relative to wild-type K32 ± SD of 100% for K32, 294.9% ± 66.3% for K32(Q31F/T33N), 200.2% ± 11.1% for K32(Q31D/T33D), 241.5% ± 24.3% for K32(T33D), and 51.1% ± 1.3% for K32(G30D)] (Fig. (Fig.1C,1C, lanes 1 to 5). Similarly, the level of K88 ubiquitination with the K50-like K88(V87T/R89S) and the K53-like K88(V87F/R89N) was increased ~50% relative to that of K88 within its wild-type sequence context [100% for K88, 148.8% ± 18.4% for K88(V87T/R89S), and 156.4% ± 41.7% for K88(V87F/R89N)] (Fig. (Fig.1C,1C, lanes 12 to 14). Conversely, the levels of ubiquitination of K50 were decreased by 5- and 2-fold with the K84-like K50(T49P) and the K88-like K50(T49V/S51R) mutants, respectively, compared to wild-type K50 [100% for K50, 19.3% ± 5.4% for K50(T49P), and 51.8% ± 19.5% for K50(T49V/S51R)] (Fig. (Fig.1C,1C, lanes 6 to 8). The level of ubiquitination of the K84- and K88-like K53 derivatives also decreased by 5- and 2-fold, respectively, compared to wild-type K53 [100% for K53, 17.4% ± 6.1% for K53(F52P/N54S), and 43.9% ± 18.4% for K53(F52V/N54R)] (Fig. (Fig.1C,1C, lanes 9 to 11). Hence, amino acids proximal to the lysine regulated the efficiency of ubiquitination in a predictable manner. To clarify this issue further, we performed kinetic studies. Assessment of the initial reaction rates against the concentration of single-lysine Sic1 K32 and its K53-like and T33D mutants revealed reaction rates, Vmax values, of 7.0, 18.1, and 15.1 fmol/min, respectively, while the estimated Km values were 0.31, 0.30, and 0.32 μM, respectively (Fig. (Fig.1D),1D), consistent with the previously estimated Km of 0.25 μM for Sic1 binding to Cdc4 (33). Although T33 is one of the seven CDK consensus phosphorylation sites at the N terminus (T2, T5, T33, T45, S69, S76, and S80) (15), T33 phosphorylation does not contribute to Cdc4 binding (14), consistent with our results that mutations of this site did not alter the apparent Km. With Sic1 K53 and its K84- and K88-like mutants, the estimated Vmax values were 49.4, 21.3, and 28.8 fmol/min, while the estimated Km values were 0.19, 0.32, and 0.28 μM, respectively (Fig. (Fig.1D).1D). Since the major effect was on the Vmax rather than the Km, this suggests that mutations of residues around Sic1 lysines affected the rate at which the lysine acted as a nucleophile to attack the Cdc34~Ub thioester bond rather than the affinity of Sic1 toward the Cdc4 subunit of SCF.
Next, we assessed whether the sequence-dependent ubiquitination of Sic1 lysines is physiologically relevant. Since the proliferation of S. cerevisiae is dependent on the ubiquitination and proteolysis of Sic1 (29), we compared the growth rates of yeast strains overexpressing single-lysine Sic1 derivatives. As controls, we analyzed the growth of cells containing either empty plasmid, wild-type Sic1 (Sic1 Wt), or lysine-less Sic1 (Sic1 K0). Immunoblotting confirmed that all Sic1 constructs were expressed to similar levels (results not shown). The induced expression of ectopic Sic1 Wt slightly reduced cell growth compared to the empty vector, while the expression of Sic1 K0 led to growth arrest (Fig. (Fig.2A).2A). Cells expressing the Sic1 single-lysine K32, K36, K50, K53, K84, and K88 mutants grew at rates between these two extremes (Fig. (Fig.2A).2A). Compared to cells with the empty vector, all of the cells expressing the different Sic1 mutants displayed reduced growth rates, indicating that they retain biological activity and that mutations do not lead to major structural defects. This is consistent with previous studies demonstrating that the N-terminal region (amino acids 1 to 90) of Sic1 has no appreciable stable secondary structure (14, 15), that the CDK-inhibitory region of Sic1 is localized within the C-terminal 70 amino acids (9), and that a Sic1 mutant with all 20 lysine residues replaced with arginines retains full CDK-inhibitory activity (20).
Cells expressing Sic1 K53 grew the fastest, followed by those expressing Sic1 K50, while cells with the remaining Sic1 mutants grew slower (Fig. (Fig.2A).2A). While this growth correlated with the rates of ubiquitination of the different Sic1 single-lysine derivatives by SCFCdc4/Cdc34 (Fig. (Fig.1A),1A), it is possible that the position of the Ub chain within Sic1 might affect their proteasomal degradation and growth effects (20). To exclude this possibility, we kept the lysine position the same and tested whether changes to the proximal amino acids surrounding Sic1 K53 has an impact on the rate of proliferation in S. cerevisiae. Cells expressing the K84- or K88-like K53 mutant grew slower than cells expressing the K53 wild-type derivative (Fig. (Fig.2A).2A). Hence, the growth rate of cells expressing Sic1 K53 and their respective proximal amino acid mutants correlated with their relative ubiquitinations in vitro.
Finally, we tested whether the differences in the rates of ubiquitination of different Sic1 variants affected their degradation rates in vivo. We tested wild-type Sic1 K53 and its poorly ubiquitinated mutant counterpart, K84-like Sic1 K53. In addition, we tested wild-type Sic1 K32 and its improved ubiquitinated mutant counterpart, K53-like Sic1 K32. Cells were first arrested in the G1 phase by α-factor arrest and by Sic1 expression, which was induced with galactose. Following the removal of α-factor and galactose to induce G1-S-phase cell cycle progression and stop ectopic Sic1 expression, pulse-chase time course experiments revealed that the degradation of the K84-like Sic1 K53 mutant was substantially slower than that of wild-type Sic1 K53 (Fig. (Fig.2B).2B). Conversely, the degradation of K53-like Sic1 K32 was noticeably faster than that of wild-type Sic1 K32 (Fig. (Fig.2B).2B). Analysis of the kinetics of the appearance of the S-phase cyclin Clb5 by Western blotting confirmed that Sic1 was degraded when cells were progressing from G1 to S phase. Therefore, the rate of degradation of these Sic1 variants correlates with their rate of ubiquitination in vitro (Fig. (Fig.1C1C and and2B2B).
The above-described findings suggested that key residues in the catalytic region of Cdc34 were involved. The catalytic domain among E2s is relatively conserved, including those of Ub-like conjugating enzymes (22). Studies of the SUMO E2-substrate complex of human Ubc9 and RanGAP1 revealed that Y87, D127, P128, and A129 of Ubc9 around the active-site cysteine are in close proximity to the RanGAP1 lysine and that the mutation of these sites impairs sumoylation (2, 38). E2 sequence alignments revealed that Y89, S139, P140, and A141 are the analogous sites in yeast Cdc34 (Fig. (Fig.3A).3A). In contrast to P140, which is almost entirely conserved, divergence is found at the positions corresponding to Y89, S139, and A141 of yeast Cdc34 (Fig. (Fig.3A).3A). Site 89 has the highest divergence, with tyrosine, asparagine, aspartate, serine, threonine, glycine, or lysine at this position. Site 139 contains either serine or aspartate (and rarely glutamate), and site 141 features alanine, leucine, or glutamine. E2s feature only 11 different triplets from a possible 63 (Fig. (Fig.3A),3A), suggesting that only certain combinations are permissive for catalysis. To define if yeast Cdc34 Y89, S139, and A141 (here referred to as the core sites) are important for sequence-dependent ubiquitination, we replaced these sites in Cdc34, individually or in combination, with residues found at the corresponding positions in yeast Ubc4 (N80, D118, and L120), since these two E2s diverge at all of these sites (Fig. (Fig.3A).3A). We also replaced S139 of Cdc34 with glutamate, alanine, and threonine, since our previous work showed that the corresponding site in hHR6A (S120) regulated the activity of this E2 (27). For simplicity, Y89, S139, and A141, are here indicated by a triplet single-letter amino acid code, and a mutation is marked by the underlined letter; e.g., YDA indicates the single-site mutation S139D. Combinatorial mutagenesis generated a panel of single (NSA, YDA, YAA, YEA, YTA, and YSL), double (NDA, YDL, and NSL), and triple (NDL) mutants. Here, wild-type Cdc34 is referred to as Cdc34.
We first assessed the importance of these sites for Cdc34 function in vivo. Since Cdc34 is essential for S. cerevisiae viability, we monitored the mutants for their abilities to support growth in the absence of endogenous Cdc34 by plasmid shuffling in a Δcdc34 deletion strain (4). Immunoblotting confirmed the expression of the mutants at levels similar to that of Cdc34 (Fig. (Fig.3D).3D). Cells expressing either Cdc34 or the NSA, YDA, NSL (Fig. (Fig.3B),3B), YAA, YTA, and YEA mutants (results not shown) were viable, while cells expressing the YSL, NDA, YDL, and NDL mutants did not grow (Fig. (Fig.3B).3B). We analyzed the survivors by drop test analysis. While the growth of cells expressing Cdc34 was similar to that of cells expressing the NSA or NSL mutant at 23°C, incubation at 37°C induced a temperature-sensitive growth defect in the NSL mutant (Fig. (Fig.3C).3C). Although cells with YSL were not viable (Fig. (Fig.3B),3B), the expression of the NSL double mutant rescued viability at 23°C (Fig. 3B and C), indicating that the Y89N mutation compensates conditionally for the lethal effect of the A141L mutation. Cells expressing YAA, YTA, YDA, or YEA displayed a moderate-to-severe slow-growth phenotype at 23°C compared to Cdc34. Incubation at 37°C exacerbated the growth defect of cells expressing the YDA, YEA, or YTA mutant (Fig. (Fig.3C).3C). Therefore, Y89, S139, and A141 play important roles, and only certain combinations of amino acids support Cdc34 function in vivo (Table (Table33).
We next assessed whether these core sites are important for the ubiquitination of Sic1. Sic1 Wt has 20 lysines, and almost all of them are ubiquitinated by SCFCdc4/Cdc34 in vitro (20). To evaluate only the ubiquitination of Sic1, we employed the chain-terminating Ub derivative Ub(K0). Cdc34 and YDA efficiently ubiquitinated Sic1 on multiple sites, generating conjugates of ~100 to 250 kDa (Fig. (Fig.4A,4A, lanes 2 and 4, and B). YAA and YTA ubiquitinated fewer Sic1 sites, as evidenced by conjugates of ~100 kDa, and the remaining mutants failed to generate conjugates of >75 kDa (Fig. 4A and B). Interestingly, the efficiency of the Cdc34 mutants in ubiquitinating Sic1 did not correlate with their efficiency in supporting growth in S. cerevisiae. For example, YDA was severely impaired in supporting growth of the Δcdc34 strain but was as active as Cdc34 in ubiquitinating Sic1 (Fig. 3B and C and 4A and B). Conversely, NSA supported growth to the same extent as Cdc34 but was substantially less efficient in ubiquitinating Sic1 (Fig. 3B and C and and4A).4A). To determine if this was due to a change in the Sic1 lysine preference, we assayed the Cdc34 mutants for their abilities to monoubiquitinate the N-terminal single-lysine Sic1 derivatives. While YDA was as active as Cdc34 toward the single-lysine Sic1 substrates, the other mutants showed reduced activity (Fig. (Fig.4C).4C). The overall activity profiles of the different mutants toward Sic1 Wt and the single-lysine Sic1 mutants were similar (Fig. 4A and C; summarized in Table Table3),3), indicating that these mutants did not noticeably alter their preference for particular Sic1 lysine residues.
The recognition and proteasomal degradation of proteins requires a Ub chain of at least four Ubs (34). Hence, the efficiency of the different Cdc34 mutants in supporting the growth of the Δcdc34 yeast strain might be correlated with their ability to polyubiquitinate via K48 of Ub. To address this, we first monitored the polyubiquitination of Sic1 Wt with Ub(Wt). Sic1 Wt was efficiently polyubiquitinated by Cdc34 and the YAA and YTA mutants, generating conjugates of >250 kDa (Fig. (Fig.5A,5A, lanes 2, 10, and 12). The other mutants generated lower-molecular-weight (MW) intermediates. Interestingly, while the YDA mutant ubiquitinated all of the Sic1 Wt substrate, the majority of these conjugates were <150 kDa, and time course experiments revealed that the YDA mutant was impaired in chain elongation (Fig. (Fig.5B).5B). To distinguish between the ubiquitination of Sic1 lysines and K48 of Ub during polyubiquitination, we assayed polyubiquitination with the single-lysine derivative Sic1 K53. We specifically tested the ability of the Cdc34 mutants to utilize K48 of Ub for chain synthesis by employing Ub(K48o), which contained only a single lysine at position 48. Cdc34 generated chains with up to seven Ub molecules on Sic1 K53 (Fig. (Fig.5C,5C, lane 2). While the YAA and YTA mutants conjugated chains of similar lengths, the YDA and YEA mutants were impaired in Ub chain extension, accumulating predominantly mono- and diubiquitinated Sic1 K53 (Fig. (Fig.5C,5C, lanes 4 and 10 to 12). Interestingly, the NSA mutant generated chains of approximately four Ub molecules but did not accumulate monoubiquitinated Sic1 K53, suggesting that this mutant was efficient in Ub chain extension. All of the other mutants were severely impaired in Sic1 K53 polyubiquitination (Fig. (Fig.5C,5C, lanes 5 to 9).
To directly assess the ability of the Cdc34 Y89 and S139 mutants to utilize K48 of Ub for chain elongation independent of the initial attachment on Sic1, we prepared a preubiquitinated substrate with a single molecule of Ub with lysine only at position 48, Ub(K48o), attached to Sic1 K53 [Sic1 K53-Ub(K48o)1]. We assayed Ub chain elongation with lysine-less Ub(K0) to terminate conjugation after the attachment of the first Ub(K0) onto the Sic1 K53-Ub(K48o)1 substrate to generate only diubiquitinated Sic1 (Sic1-Ub2). Cdc34 readily conjugated Ub(K0) onto Sic1 K53-Ub(K48o)1, generating Sic1-Ub2 (Fig. (Fig.5D,5D, lane 2). While the NSA, YAA, and YTA mutants showed reduced activities, the YDA and YEA mutants were almost entirely inactive (Fig. (Fig.5D,5D, lanes 2 to 7). As a control, we assayed the Cdc34 derivatives toward Sic1 K53 as a substrate. Consistent with the results shown in Fig. Fig.4C,4C, Cdc34 and the YDA mutant were equally active; the YEA, YAA, and YTA mutants were less active; and the NSA mutant was substantially impaired in Sic1 K53 monoubiquitination (Sic1-Ub1) (Fig. (Fig.5D,5D, lanes 9 to 14). Strikingly, the YDA and NSA mutants displayed polar-opposite activities toward K53 of Sic1 and K48 of Ub. While the YDA mutant was efficient in Sic1 K53 ubiquitination but impaired in polyubiquitin chain extension via Ub K48 (Fig. (Fig.5D,5D, lanes 4 and 11; summarized in Table Table3),3), the NSA mutant was impaired in Sic1 K53 ubiquitination but readily ubiquitinated K48 of ubiquitin (Fig. (Fig.5D,5D, lanes 3 and 10; summarized in Table Table33).
We next tested whether the mutation of S139 might change linkage specificity by utilizing a non-K48 lysine(s) of Ub. We preubiquitinated Sic1 K32 with Ub(Wt) to generate Sic1 K32-Ub(Wt)1 as a substrate. To block the utilization of K48 in the chain-extending Ub, we added Ub(K48R), which had six lysines remaining. A change in linkage specificity would allow the YDA mutant to conjugate Ub(K48R) to Sic1 K32-Ub(Wt)1 or the YAA mutant to extend a Ub chain. Cdc34 conjugated predominantly a single molecule of Ub(K48R) onto Sic1 K32-Ub(Wt)1, with minimal levels of triubiquitinated Sic1 K32 (Sic1-Ub3) (Fig. (Fig.5E,5E, lane 2). YDA failed to efficiently attach a single Ub(K48R) onto the preubiquitinated substrate, while the YAA and YTA mutants terminated Ub chain elongation after the attachment of the first molecule of Ub(K48R) (Fig. (Fig.5E,5E, lanes 3 to 5). Hence, the mutation of S139 in Cdc34 did not change the K48 linkage specificity.
Since amino acids proximal to lysines in Sic1 determine the efficiency of ubiquitination (Fig. (Fig.1C),1C), amino acids proximal to K48 of Ub may also be important during polyubiquitination. Since the YDA mutant readily ubiquitinated Sic1 lysines but did not utilize Ub K48, we investigated whether the mutation of residues around Ub K48 (GKQL) affected polyubiquitination. Given the robust activity of the YDA mutant toward Sic1 K36 (Fig. (Fig.4C),4C), we chose the sequence surrounding K36 as a template (QKPS) and generated the single-site Ub derivatives Ub(G47Q), Ub(Q49P), and Ub(L50S). We assessed these mutants with Cdc34 and Sic1 K36 as substrates. Incubation with Ub(Wt), Ub(Q49P), or Ub(L50S) generated high-molecular-mass conjugates of >250 kDa, while polyubiquitination was severely impaired with Ub(G47Q) (Fig. (Fig.6A,6A, lanes 1 to 4). With Ub(Wt) and Ub(G47Q), the YDA mutant was impaired in polyubiquitination, generating predominantly mono- and diubiquitinated Sic1 K36 (Fig. (Fig.6A,6A, lanes 6 and 7). Conversely, Ub(Q49P) and Ub(L50S) supported polyubiquitination, generating high-molecular-mass conjugates of >250 kDa, similar to that observed with Cdc34 (Fig. (Fig.6A,6A, lanes 8 and 9). Since these mutations in Ub did not affect ubiquitination with Cdc34 and restored the activity of the YDA mutant, this strongly suggested that these changes did not have major structural effects on Ub. Time course experiments revealed that the YDA mutant was almost as active as Cdc34 in the polyubiquitination of Sic1 Wt with Ub(L50S) (Fig. (Fig.6B).6B). Therefore, G47 of Ub is critical during the catalysis of Ub K48. Studies with YDA showed that, similar to Sic1 lysine ubiquitination, compatibility between residues in the catalytic core of Cdc34 and residues surrounding K48 of Ub is important for catalysis.
We next tested whether alterations in the activities of the Cdc34 mutants were independent of SCFCdc4 by assessing them in autoubiquitination reactions without SCFCdc4 and Sic1. Autoubiquitination and the formation of unanchored Ub chains by E2s are frequently used to measure their intrinsic activity in the absence of E3s (13). Cdc34 autoubiquitinates four C-terminal lysine residues (K273, K277, K293, and K294) (1) and can generate unanchored K48-linked Ub chains (21). Cdc34 efficiently generated unanchored di-Ub (Ub2) with Ub(Wt) and autoubiquitinated with high-molecular-mass Ub conjugates of ~75 to 250 kDa (Fig. (Fig.7A,7A, lane 2). All the Cdc34 mutants showed reduced activities in at least one of the two reactions. The YDA mutant was active in autoubiquitination, with conjugates of ~70 to 250 kDa, but generated very low levels of di-Ub, indicating that it was impaired in the catalysis of K48 Ub linkages (Fig. (Fig.7A,7A, lane 4). Conversely, the YAA, YTA, and YEA mutants were less active in autoubiquitination but generated di-Ub levels intermediate to those of Cdc34 and the YDA mutant. The remaining Cdc34 mutants were less active in both reactions. These differential activities were similar to those observed in the presence of SCFCdc4 (Fig. (Fig.4A4A and and5A).5A). Therefore, sequence-dependent lysine specificity is intrinsic to the Cdc34 mutants and independent of SCF.
We also tested the activities of Cdc34 and the YDA mutant with Ub(Wt), Ub(G47Q), Ub(Q49P), Ub(L50S), or Ub(K48R) in autoubiquitination reactions. Cdc34 efficiently autoubiquitinated K48-linked Ub chains with Ub(Wt), Ub(Q49P), or Ub(P50S), generating high-MW conjugates, which were absent in the reaction with Ub(K48R). Conversely, Ub(G47Q) severely impaired autoubiquitination (Fig. (Fig.7B,7B, lanes 1 to 5). While the YDA mutant autoubiquitinated the same number of lysines as Cdc34 with Ub(K48R), it failed to form Ub chains with Ub(Wt) and Ub(G47Q) (Fig. (Fig.7B,7B, lanes 5 to 7 and 10). In contrast, the YDA mutant readily assembled high-MW Ub chains with Ub(Q49P) and Ub(L50S) (Fig. (Fig.7,7, lanes 8 and 9). Time course experiments revealed that the YDA mutant displayed a similar activity in polyubiquitination with Ub(Q49P) and Ub(L50S) compared to that of Cdc34 (Fig. (Fig.7C7C).
Assessment of the initial reaction rates against the concentration of Ub(Wt), Ub(Q49P), and Ub(L50S) in autoubiquitination reactions with Cdc34 revealed similar apparent Vmax values (339, 387, and 279 fmol/min, respectively) and Km values (5.7, 8.3, and 5.7 μM, respectively) (Fig. (Fig.7D).7D). The YDA mutant had a 3-fold-lower Vmax (107 fmol/min) and a lower Km (3.4 μM) with Ub(Wt) than did Cdc34, while the values with Ub(Q49P) and Ub(L50S) were similar to those observed for Cdc34 (Vmax values of 342 and 294 fmol/min, respectively, and Km values of 5.8 and 5.2 μM, respectively) (Fig. (Fig.7D).7D). The reduced activity of the YDA mutant with Ub(Wt) suggested that the impaired utilization of K48 in Ub chain synthesis was not due to the reduced binding and docking of Ub K48 to the YDA catalytic region, since the Km was also slightly reduced, but rather that K48 was a weaker nucleophile in attacking the Cdc34~Ub thioester bond. These studies show that compatibility between the catalytic core of Cdc34 and residues around Ub K48 are important for polyubiquitination independent of SCF.
Despite the importance of ubiquitination, the mechanisms that control the selection of lysines on substrates and Ub are poorly understood. Our studies demonstrate that compatibility between amino acids proximal to acceptor lysines and residues in the catalytic core of Cdc34 ultimately determines lysine ubiquitination in substrates and Ub. This new concept expands upon the current view that ubiquitination is dependent primarily on the proximity of the lysine to the E2-thioester mediated through positioning by E2/E3 (21, 37) and has important implications for our understanding of the mechanisms of substrate monoubiquitination and polyubiquitination.
Substrates such as Sic1, which are ubiquitinated on numerous lysines, indicate that mechanisms for selecting spatially separated lysines must exist. As described in the introduction, several higher-order structural mechanisms contribute to the flexibility in the positioning of different lysines for substrate ubiquitination, such as the E3-mediated positioning of lysines toward the E2~Ub thioester. Lysine selection flexibility may also be achieved by the release of the ubiquitin-charged Cdc34~Ub from SCF to transfer Ub to different substrate lysines, as proposed by the “hit-and-run” hypothesis (5). Our studies now demonstrate that in addition to these higher-order effects, amino acids immediately proximal to lysines also play an important role in catalysis, suggesting that the interplay of both mechanisms ultimately defines the efficiency of the ubiquitination of a lysine. SCFCdc4/Cdc34 displayed a clear preference toward Sic1 K53. The importance of the sequence environment was first exemplified by our studies mutating residues around Sic1 lysines, where changes to proximal amino acids consistently reduced or increased ubiquitination. Importantly, the growth rate of S. cerevisiae expressing the different single-lysine Sic1 mutants or the K53 sequence mutants correlated with their level of ubiquitination in vitro, indicating that sequence-dependent ubiquitination is physiologically important (Fig. (Fig.11 and and2).2). This was due to the altered stability of Sic1 in S. cerevisiae, as exemplified by Sic1 K53 and Sic1 K32 and their derivatives. The importance of Sic1 lysine selectivity during ubiquitination is exemplified by studies demonstrating that the half-life of Sic1 ubiquitinated on K36 or K84 is 1 min or 5 min, respectively, indicating that the context of Sic1 lysine ubiquitination is important for proteolysis (20).
In addition to Sic1 lysines, we observed that alterations around Ub K48 can affect polyubiquitination by Cdc34. Hence, catalysis was severely impaired with Ub(G47Q), while Ub(Q49P) and Ub(L50S) did not affect the utilization of K48 for polyubiquitination. Previous reports have shown that G47 of Ub is important for yeast cell growth and Ub conjugation via K48 (21, 31). Similar to the E3-mediated positioning of substrate lysines, structural aspects of some E2s, such as Ubc13 and Cdc34, position Ub lysines to generate Ub chains of a specific topology (11, 12, 21, 36). However, E2s such as human UbcH5 can utilize all Ub lysines but display a preference for K11, K48, and K63, indicating less structural constraint and that other mechanisms contribute to Ub lysine specificity (12). Our work suggests that amino acids proximal to lysines in Ub may provide a further level of specificity control with certain E2s.
Importantly, our studies show that ubiquitination is linked to key residues in the catalytic core of Cdc34, which can differentially regulate initial substrate ubiquitination and Ub chain extension via K48. It is unclear if the divergence of residues surrounding the catalytic cysteine in different E2s is functionally important. Our results argue that this divergence may have evolved to contribute to lysine specificity in substrates and Ub. Previous studies underscored the importance of the analogous sites in other E2s. Studies of Ubc9 in complex with the substrate RanGAP1 demonstrated that Ubc9 N85, Y87, and D127 facilitate catalysis through the optimal alignment and nucleophilic activation of the attacking lysine within the active site (38). The human Cdc34(Y87A) and Cdc34(S138A) mutants are impaired in IκBα ubiquitination (7). The yeast Ubc13(D81K) mutant is completely inactive (35). Our studies delineate their specific importance, showing that these residues are important for lysine selectivity and can specify if Sic1 is monoubiquitinated or polyubiquitinated. This was strikingly exemplified by the YDA and NSA mutants, which displayed polar-opposite activities toward lysines of Sic1 and K48 of Ub. Hence, the YDA mutant was as active as Cdc34 in the ubiquitination of Sic1 but essentially inactive toward K48 of Ub. Conversely, the NSA mutant was significantly impaired in Sic1 ubiquitination but active toward K48 of Ub (Fig. (Fig.44 and and55 and Table Table3).3). Therefore, these mutants display dichotomy and are not inactive per se but rather in their specificity toward particular acceptor lysines. These studies strongly suggest that different combinations of residues in the catalytic regions of E2s are not functionally redundant. E2s may have evolved different active-site structures for the optimal ubiquitination of specific lysines in a particular range of substrates and possibly in dictating whether a substrate is mono- or polyubiquitinated. The regulatory nature of the site analogous to Cdc34 S139 is further illustrated in E2s such as hHR6A, which is phosphorylated by CDKs on serine at this position (S120) to regulate activity (27), although not all E2s with serine in this position are phosphorylated at this site, e.g., Cdc34 (3, 24). By recruiting a particular E2 or E2s, E3 may specify the mode of substrate ubiquitination. In support of this notion, although E3s such as SCF and human APC/C use single E2s for attaching Ub to both substrate and Ub lysines, other E3s utilize different E2s for specific roles during polyubiquitination. Hence, yeast APC/C utilizes Ubc4 and Ubc1 for ubiquitinating cyclin B (23). Ubc4 catalyzes cyclin B monoubiquitination, while Ubc1 catalyzes further K48-mediated polyubiquitination.
Our studies with the YDA mutant highlighted the importance of compatibility between the lysine environment and E2 core residues. While this mutant was impaired in its ability to utilize Ub K48, changes to the proximal residues (Q49P and L50S) restored the activity of this mutant in Ub chain assembly (Fig. (Fig.66 and and7B).7B). Although the structure of yeast Cdc34 is not known, the catalytic domain of E2 enzymes is highly conserved (22). Structure-function studies of the E2-substrate complex of human Ubc9 and RanGAP1 showed that the sites analogous to Cdc34 Y89, S139, and A141 (Ubc9 Y87, D127, and A129) surround the catalytic cysteine. Ubc9 Y87 and A129 make van der Waals contacts with L525, S527, and E528 of RanGAP1, which are proximal to the sumoylated K526, while D127 is within hydrogen-bonding distance of sumoylated K526 (2, 38). These interactions are important for sumoylation through optimal alignment and pK suppression for the nucleophilic activation of the attacking lysine (2, 38). Our studies suggest that similar contacts between the Cdc34 core sites and the ubiquitination site sequence are important for catalysis. Our kinetic studies suggest that compatibility is important for lysine activation rather than substrate binding. Alternatively, it is possible that some of the Cdc34 residues examined in this study may be important for facilitating Cdc34 activity via self-association, which is important for its catalytic activity (36). However, these studies demonstrated that the active-site cysteine and serines 73 and 97, which are distinct from the sites evaluated in this study, are important for oligomerization. Future structural studies of the Cdc34-substrate complex will provide further insights into these issues.
In agreement with our findings, recent studies with UbcH10 and its cognate E3, APC/C, demonstrated that a sequence motif, termed the TEK box, surrounding lysines in the substrate and Ub K11 is important for ubiquitination (11). In addition, proteins with Ub binding domains (UBDs) can be ubiquitinated directly by E2s without E3s (10). It is postulated that the substrate UBD binds to the Ub conjugated to the E2 and that Ub is then transferred to the substrate lysine. Seven different E2s displayed differential specificity toward different UBD proteins, raising the issue of how a particular UBD substrate is selected for ubiquitination by a particular E2, since there is no E3 to direct specificity. Our studies suggest that the compatibility of the E2 catalytic core with residues surrounding potential target lysines is important for controlling specificity between particular E2s and UBD proteins.
Altogether, these findings demonstrate that in addition to the importance of higher-order structural features of SCFCdc4 and Cdc34 or models such as the “hit-and-run” hypothesis, which position different lysines for the attack of the Cdc34~Ub thioester bond, compatibility between key residues within the Cdc34 catalytic region and those proximal to acceptor lysines provides a further level of specificity control. E2s may have evolved divergent structures in their catalytic regions to modulate lysine specificity. It will be interesting to comprehensively define all the important determinants in the catalytic region of E2s and residues proximal to acceptor lysines to define if this is a general feature of E2-mediated ubiquitination.
We thank R. Deshaies for Sic1 plasmids, W. Harper for baculoviruses of the SCF subunits, M. Goebl for the Cdc34 antibody, and B. Kuhlman for the PKA-Ub construct. We thank A. Traven, B. Callus, C. House, and T. Mason for critically reading the manuscript.
This research was supported by grants from the Cancer Council Victoria, the Multiple Myeloma Research Foundation, and the Association for International Cancer Research.
Published ahead of print on 1 March 2010.