Identification of a multifunctional binding site on Ubc9p required for Smt3p conjugation

which conjugates the SUMO family member Smt3p. We identified Ubc9p surfaces involved in thiolester bond and Smt3p-Smt3p chain formation. The residues involved in thiolester bond formation map to a surface we show is the E1 binding site, and E2s for other ub-lps bind to their E1s at a homologous site. We also find that this same binds Smt3p. A mutation that impairs binding to E1 but not Smt3p impairs thiolester bond formation, suggesting that it is the E1 interaction at this other E2s and their also use this same for binding to ubiquitin, E3s and other revealing this to be a multipurpose binding site, and suggesting that the E1-E2-E3 pathway has for


INTRODUCTION
Post-translational covalent attachment of ubiquitin and ubiquitin-like proteins (ub-lps) has emerged as a predominant cellular regulatory mechanism (reviewed in 1) that plays an important role in a variety of biological processes including the immune response, development, endocytic trafficking, cell division (2,3) and cancer (4)(5)(6). The best understood function of these modifications is ubiquitin-mediated proteolysis, in which polymeric chains containing four or more ubiquitins, linked between K48 on the surface of one ubiquitin to the C-terminus of the next, direct proteins for degradation by the proteasome (reviewed in 7). Polyubiquitin chains with linkages via K63, rather than K48, activate the IkB kinase (8), and monoubiquitylation plays a role in processes ranging from protein trafficking to transcriptional activation (9). New ubiquitin-like proteins that are conjugated to macromolecules in vivo are being discovered at a rapid rate. For example, NEDD8 (Rub1p in budding yeast) is involved in cell cycle control (10)(11)(12), Apg8p modifies the lipid phophatidylethanolamine to modulate membrane dynamics (13), and SUMO family members, including Smt3p in budding yeast, modify a number of proteins involved in cell division, nuclear transport, the stress response and signal transduction (reviewed in 14,15). Polymeric chains of SUMO family members have been found to modify a number of proteins in vivo and in vitro, including septins, p53, c-Jun, histone deacetylase, and RanBP2 (16)(17)(18)(19)(20).
All of these ub-lps are conjugated at their C-termini to protein substrates by an enzymatic cascade with common features (11,12,15,(21)(22)(23)(24)(25)(26)(27). A number of ub-lps are proteolytically processed at their C-termini in order to end with the sequence Gly-Gly, required for subsequent steps of the pathway. Ub-lps are adenylated at the C-terminus by an activating Bencsath et al., JBC M2:07442, A multifunctional binding site on Ubc9p 3 enzyme or E1. The C-terminus of the ub-lp subsequently forms a thiolester with the catalytic cysteine of E1. The ub-lp then forms a thiolester between its C-terminus and the catalytic cysteine of a conjugating enzyme, or E2. The sequences of E1s are similar, and the structures of E2s and ub-lps are conserved, suggesting that the mechanisms of activation and thiolester bond formation will be similar for all ub-lps.
There are three known mechanisms for transfer of the ub-lp to a lysine residue on the target. In many cases, transfer is facilitated by an E3 protein or protein complex that selects the target and acts as an ub-lp protein ligase (reviewed in 7,28). In the predominant mechanism of ubiquitin transfer, the E3 appears to act as a scaffold that recruits the target protein and the E2, facilitating ligation from the E2 to the target. For some E3s, ubiquitin forms a thiolester with the active site cysteine of an E3, which itself transfers ubiquitin. It is also possible for the E2 to transfer a ub-lp in the absence of an E3. For example, chimeric E2s for ubiquitin, fused to protein binding peptides, ubiquitylate the associated proteins (29), and Ubc9 binds a number of its targets directly and transfers SUMO family members to a lysine in the tetrapeptide consensus For ubiquitin there is one E1, tens of E2s, and hundreds of E3s playing a major role in substrate conjugation, although a number of ub-lps only have one known E2. Genetic studies in S. cerevisiae indicate that E2s have evolved to orchestrate modification of proteins in a common pathway. For example, Rad6p and Cdc34p, which transfer ubiquitin, are involved in the DNA damage response and the G1-S transition of the cell cycle respectively (31,32). Ubc9p, which transfers Smt3p, is also involved in cell cycle control (22,33,34). Therefore, the ability of a particular E2 to accept and transfer its ub-lp is crucial to the organization of the pathway. This PCR products encoding Smt3, Ubc9 and mutants were subcloned into pGEX4T3 (Pharmacia) and expressed as GST fusions in BL21(DE3) (Novagen). The cDNAs encoding Aos1p and Uba2p were subcloned into a bicistronic vector described previously (42), and

Thiolester bond formation assays
Thiolester bond formation assays were performed in 50 µl volumes containing 4.5 nM

Ubc9p/E1(S) and Ubc9p/Smt3p protein-protein interaction assays
Binding was determined by a native gel mobility shift assay. Binding was performed Reactions were terminated by boiling half of the reaction in SDS sample buffer containing DTT.
Reaction products were fractionated on 15% SDS polyacrylamide gels and visualized by Coomassie staining. The identity of proteins in the gel bands and the location of the Smt3p-Smt3p linkage was confirmed by mass spectrometry following tryptic digestion (see below).

Smt3p-Smt3p isopeptide complexes, and the native gel mobility shift assays by tryptic digests and MALDI-TOF or LC mass spectrometric analysis
Following electrophoresis, Coomassie-stained proteins were excised from gels, reduced,  used for concentration also appear in the spectra.

Identification of surfaces of Ubc9p
To identify the surfaces on Ubc9p available for mediating protein-protein interactions and catalysis, we performed structural analysis. Although the structure of S. cerevisiae Ubc9p has not been determined, three crystal structures are available of the human ortholog (40,43,44), whose sequence is 56% identical to the yeast protein, with most of the differences being conservative substitutions. We first analyzed the crystal structures of human Ubc9, and compared them with the crystal structures of human UbcH7 (35,36), which transfers ubiquitin.
We reasoned that the residues involved in binding to E1 and SUMO family members should be conserved among Ubc9 family members, but not among E2s that transfer ubiquitin, and residues involved in common features of catalysis should be conserved around the active sites of both types of E2s. We identified 62 residues that cover most protein-protein interaction sites on the surface, and mutated these to either alanine or the corresponding residue in E2s that transfer ubiquitin and not Smt3p. Our goal was to make mutations that would disrupt protein-protein interactions or catalysis, without significantly disrupting the structure. One of the greatest structural differences between Ubc9 and E2s that transfer ubiquitin is a 5-residue insertion in the loop between β-strand 1 and β-strand 2 (43,44), so some deletion mutants were also made that truncate this loop to the size of that found in E2s for ubiquitin to test the function of this loop.
The location of amino acids that were mutated in this study is shown on the structure of human Ubc9 in Figure 1.

Surfaces involved in thiolester bond formation
In order to identify regions of Ubc9p other than the active site cysteine required for activity, we developed an in vitro assay using purified recombinant Ubc9p, Smt3p and the heterodimeric Aos1p/Uba2p complex, which is the E1 for Smt3p, hereafter referred to as E1(S).
This assay is similar to that described previously for human Ubc9, with reactants and products Because of the essential role that activating enzymes play in transferring ub-lps to E2s, we wanted to know whether this surface is involved in binding to E1(S). In addition, previous yeast 2-hybrid and NMR studies have shown that Ubc9 family members interact noncovalently with SUMO family members (38,(47)(48)(49), and this region overlaps with the region of human Ubc9 previously identified by NMR as the SUMO-1 binding site (38). Therefore, we tested whether these or other mutations in Ubc9p impair noncovalent binding to either E1(S) or to Smt3p.

Identification of the E1(S) binding site on Ubc9p
To address whether mutations that impair thiolester bond formation are important for Ubc9p binding to E1(S) or Smt3p, we developed a nondenaturing gel mobility shift assay to examine protein-protein interactions between E1(S) and Ubc9p, and Smt3p and Ubc9p. Briefly, E1(S) and Smt3p are acidic, so they enter a native gel at pH 8. However, the pI of Ubc9p and all of the mutants in this study is greater than 8, so they do not enter a pH 8 gel. Binding between  Fig. 3A and   B, respectively). We tested the ability of all of the Ubc9p mutants to bind to E1(S) and Smt3p (summarized in Table 1). Using this approach we identified the E1(S) binding site on Ubc9p to correspond to the surface important for Smt3p thiolester bond formation, involving the Nterminal helix and the loop between β-strands 1 and 2. To our surprise, the binding sites for E1(S) and Smt3p partially overlap each other, as this region is also involved in noncovalent interactions with Smt3p, consistent with a previous NMR study of human Ubc9 and SUMO-1 (38). To exclude the possibility that mutations such as K14E, which impair thiolester bond formation and binding to both E1(S) and Smt3p, lead to a protein folding defect, we examined the one-dimensional 1 H NMR spectrum of the K14E mutant (Fig. 4). The spectrum contains extensive chemical shift dispersion, including resonances shifted upfield of 0 ppm and downfield of 10 ppm that resemble those of the wild-type protein. These features indicate proper folding.

Smt3p and E1(S) compete for binding to Ubc9p
Because the binding sites for E1(S) and Smt3p are partially overlapping, we tested whether Smt3p can compete with E1(S) for binding to Ubc9p. We first tested whether Smt3p interacts with E1(S) using our native gel shift assay and gel filtration chromatography (data not shown), and we see no evidence for interaction in the absence of MgATP, consistent with classic findings for the E1 for ubiquitin (50,51 and MgATP (Fig. 6A), and find that Smt3p also forms polymeric chains in vitro under similar conditions to those described for the human proteins. Ubc9p is also automodified via an isopeptide linkage with Smt3p, and this same modification has been observed in vivo (Pamela Meluh, personal communication). We isolated the Smt3p-Smt3p species from the SDS-PAGE gel and determined the isopeptide linkage to be through K15 by a combination of proteolysis and mass spectrometry (see Experimental Procedures). To confirm that K15 is the site of the Smt3p-Smt3p linkage, we tested the ability of K to R mutants to form the Smt3p-Smt3p conjugates.
Both the K15R mutation and the triple mutation of all three possible sites severely impaired for formation of the Smt3p-Smt3p conjugate, without affecting formation of the Ubc9p-Smt3p complex (Fig. 6B). There is no effect of mutating the other sites (Fig. 6B), consistent with the identification of K15 as the site of Smt3p-Smt3p isopeptide bond formation. This suggests that the mechanism of polymeric chain formation will be conserved among SUMO family members.
We next tested the Ubc9p mutants for their ability to form Smt3p-Smt3p conjugates ( SUMOylation of a number of substrates (40), and a mutation in this region of Ubc13 plays a role in recognition of the acceptor ubiquitin molecule in polyubiquitin chain assembly (53). The N124A and N140A/A142E/E143K mutants are more severely impaired for Smt3p transfer, as they are also defective in Ubc9p-Smt3p isopeptide bond formation (data not shown).
Interestingly, N124 is highly conserved as either an Asn or a Gln in the sequences of other E2s, suggesting that it may play a general role in catalysis of transfer of ub-lps.

Specificity of the E1-E2 interaction
Several previous findings have raised the question of how a given E2 selects its cognate ub-lp. First, the sequences of some ubiquitin conjugating E2s are closer to the sequences of both Ubc9 and Ubc12, which transfer SUMO and NEDD8, respectively, than to other ubiquitin conjugating E2s. Second, the three-dimentional structure of human Ubc9 is very similar to the structure of ubiquitin conjugating enzymes (43,44). These observations suggest that the mechanism of conjugation is conserved for all E2s. Our finding that the residues involved in Smt3p-Smt3p conjugate formation map to the region around the active site, which is generally conserved among E2s, is consistent with this notion. Third, NMR studies have revealed few contacts between ubiquitin and ubiquitin conjugating enzymes in thiolester or analogous ester complexes formed by catalytic cysteine to serine mutants of E2s (37,39). In addition, even the Bencsath et al., JBC M2:07442, A multifunctional binding site on Ubc9p 16 ester complexes are labile (37). Therefore it has been suggested that the E1 may play a major role in bringing the E2 together with the correct ub-lp (7). This hypothesis is supported by the kinetics of ubiquitin transfer, which are consistent with high affinity interaction between E1 and E2 (41). Also, a mutant form of NEDD8 activated by the E1 for ubiquitin was found to form a thiolester intermediate with an E2 for ubiquitin (54). Here we provide direct support for this hypothesis, by finding that the mutations disrupting Ubc9p-Smt3p thiolester bond formation map to the binding site for E1(S). Together, these findings suggest that binding to the appropriate E1 by an E2 is involved selection of the appropriate ub-lp.
We have mapped the E1(S) binding site on Ubc9p to involve the N-terminal helix and the loop between the first and second β-strands. Interestingly, in Ubc9 family members, this loop contains an insertion not found in the sequences of E2s for Ub or other Ub-lps (43,44).
This region protrudes away from the Ubc core and is one of the most prominent differences between the structure of Ubc9 and the structures of E2s for Ub (43,44). In addition, K14, in the N-terminal helix, is often replaced by an acidic or polar residue in E2s for ubiquitin. Thus, this surface does contain differences between E2s for different ub-lps that could serve as the basis for specificity in E1 binding.

Conservation of the E1 binding location
The overall location of the E1 binding site, the surface containing the N-terminal helix and the loop between the first two β-strands, is probably conserved among E2s for other ub-lps.
Two previous mutational studies of E2s for ubiquitin have shown that the N-terminal helix is important for thiolester bond formation (55,56), and we have found that the loop between the Bencsath et al., JBC M2:07442, A multifunctional binding site on Ubc9p 17 first two β-strands is important for thiolester bond formation for another E2 (data not shown).
Although these studies did not directly examine E1 binding, they led Cook and colleagues to propose a role for the E2 N-terminus in E1 binding in their description of the first crystal structure of an E2 (57). If this region is involved in E1 binding in general, then other proteinprotein interactions that block access to this surface would be expected to impair E2-ubiquitin thiolester bond formation. This result has recently been reported for human Ubc13. Two recent crystal structures show that MMS2 binding blocks access to the loop between the first two βstrands of Ubc13 (53,58), and Ubc13-ubiquitin thiolester bond formation has recently been reported to be inhibited by the binding of MMS2 (45).

A conserved multifunctional binding site on E2
We have demonstrated that the E1(S) binding site on Ubc9p is also the binding site for Smt3p, and that Smt3p competes with E1(S) for binding to this site. While binding to E1 is important for thiolester bond formation, the function of noncovalent binding of Ubc9 by SUMO family members remains a mystery. Interestingly, tight noncovalent interactions with E2s have not been reported for other ub-lps, so this may be a unique feature of SUMO family members.
A previous NMR study revealed that the SUMO-1 C-terminus, which is the site of conjugation, is not involved in the noncovalent interaction with Ubc9, suggesting that at least the terminal SUMO in a conjugated polySUMO chain could interact with Ubc9 (38). If this were the case, then thiolester-linked SUMO-Ubc9 complexes could bind to polySUMO chains to promote chain elongation, although our mutational analysis suggests that this interaction is not essential for Smt3p-Smt3p chain assembly. of the interaction surface for the complex with c-Cbl (36), and although it is partially exposed in the complex with E6AP (35), it is completely buried in the complex with c-Cbl (36).

Coevolution of the E1-E2-E3 pathway
Because it is not known whether E1 and E3 bind to E2 simultaneously, it is possible that E1 and E3 bind to the same site on E2. Our finding that the corresponding Ubc9p binding site for E1(S) is completely buried in the UbcH7/c-Cbl complex (36) suggests that the E2 is a central coordinator in the pathway, and raises the possibility that E1, E2 and E3 have coevolved as for a given ub-lp as a method of ensuring that the correct ub-lp is directed to the correct target.