|Home | About | Journals | Submit | Contact Us | Français|
The SUMO ubiquitin-like proteins play regulatory roles in cell division, transcription, DNA repair, and protein subcellular localization. Paralleling other ubiquitin-like proteins, SUMO proteins are proteolytically processed to maturity, conjugated to targets by E1-E2-E3 cascades, and subsequently recognized by specific downstream effectors containing a SUMO-binding motif (SBM). SUMO and its E2 from the budding yeast S. cerevisiae, Smt3p and Ubc9p, are encoded by essential genes. Here we describe the 1.9 Å resolution crystal structure of a noncovalent Smt3p–Ubc9p complex. Unexpectedly, a heterologous portion of the crystallized complex derived from the expression construct mimics an SBM, and binds Smt3p in a manner resembling SBM binding to human SUMO family members. In the complex, Smt3p binds a surface distal from Ubc9's catalytic cysteine. The structure implies that a single molecule of Smt3p cannot bind concurrently to both the noncovalent binding site and the catalytic cysteine of a single Ubc9p molecule. However, formation of higher-order complexes can occur, where a single Smt3p covalently linked to one Ubc9p's catalytic cysteine also binds noncovalently to another molecule of Ubc9p. Comparison with other structures from the SUMO pathway suggests that formation of the noncovalent Smt3p–Ubc9p complex occurs mutually exclusively with many other Smt3p and Ubc9p interactions in the conjugation cascade. By contrast, high-resolution insights into how Smt3p–Ubc9p can also interact with downstream recognition machineries come from contacts with the SBM mimic. Interestingly, the overall architecture of the Smt3p–Ubc9p complex is strikingly similar to recent structures from the ubiquitin pathway. The results imply that noncovalent ubiquitin-like protein–E2 complexes are conserved platforms, which function as parts of larger assemblies involved many protein post-translational regulatory pathways.
Covalent attachment of ubiquitin-like proteins (Ubls) is emerging as a predominant mechanism for regulating protein function in eukaryotes.1; 2 In this process, the C-terminus of a Ubl becomes covalently attached to the target.3 Protein modification by the prototypic Ubl, ubiquitin, is often associated with altered target half-life, subcellular localization, or protein–protein interactions. In addition to ubiquitin, several structurally related Ubls, such as ISG15, NEDD8, and members of the SUMO family have been found to modify a vast number of targets to alter protein functions in a variety of ways. For example, SUMO, or Smt3p (S. cerevisiae SUMO) modifications have been associated with altering target protein subcellular localization, enzymatic activity, and ability to interact with protein or DNA partners.4; 5; 6; 7 In general, Ubls, including ubiquitin and SUMO, are conjugated by their own enzymatic cascades, which share many common features.3 Ubls are initially translated with extended C-termini, which become proteolytically processed to generate the functional C-terminal Gly-Gly motif. Once processed, a Ubl's C-terminus becomes covalently attached to its targets via an enzymatic cascade involving Ubl-specific E1, E2, and E3 enzymes. The E1 first binds noncovalently to the Ubl and catalyzes adenylation of the Ubl's C-terminus in a MgATP-dependent reaction. The E1 subsequently forms a covalent thioester intermediate between its catalytic cysteine and the Ubl's C-terminus. The E1 binds E2 for a transthiolation reaction, where the Ubl is transferred from the E1's catalytic cysteine to that of E2. The product of this reaction is a covalent E2~Ubl complex (~ denotes covalent interaction), linked by a thioester bond between the E2's catalytic cysteine and the Ubl's C-terminus. The covalent E2~Ubl complex is released from E1, and then generally interacts with an E3, which facilitates Ubl transfer, most often to a target's lysine's ε-amino group. The modified target interacts with Ubl recognition machinery, resulting in altered function of the target. Finally, the dynamic nature of this process is maintained by Ubl deconjugation, carried out by Ubl specific proteases.
In addition to many conserved features of Ubl cascades, there are several unique features distinguishing the SUMO pathway. First, relative to the structure of ubiquitin, SUMO contains a ~20 residue disordered N-terminal extension.8 Second, whereas the ubiquitin cascade is significantly expanded at the level of E2, a single E2 (Ubc9) is responsible for all SUMO modification.9; 10 Third, in some cases, Ubc9 directly recognizes a target's SUMO modification consensus motif (ϕ-K-X-D/E, ϕ, hydrophobic; K, target lysine; and D/E, acidic) either in the absence or presence of an E3.11 Fourth, whereas the sequence of ubiquitin is highly conserved across phyla, in mammals there are four reported SUMO family members, with the best-studied being SUMO-1, -2, and -3.7 These have distinct features, but all share ~45% sequence identity with the single S. cerevisiae SUMO, Smt3p. S. cerevisiae Smt3p, and mammalian SUMO-2 and SUMO-3, can form polymeric chains, in which a lysine from ϕ-K-X-D/E consensus sites found in the unstructured N-terminal extension from one SUMO is linked to the C-terminus of the next in the chain.12; 13 However, unlike ubiquitin, chain formation is not an essential feature of Smt3p, as neither the entire N-terminal extension nor any Smt3p lysines are essential for S. cerevisiae viability.14 A fifth unique feature of the SUMO pathway is a distinct set of E3s: all eukaryotes appear to have SUMO E3s of the Siz/PIAS class.15; 16; 17 These contain a Siz/PIAS-RING (SP-RING) domain, which resemble RING and U-box sequences found in ubiquitin E3s. SP-RINGs are thought to interact with Ubc9 in a manner resembling RING/U-box interactions with ubiquitin E2s. Higher eukaryotes are also known to have additional families of SUMO E3s, such as Nup358/RanBP2 and PC2,18; 19 which do not resemble any known ubiquitin E3s.
SUMO also has its own distinct recognition machineries, which contain a “SUMO Binding Motif” (SBM, also called a “SUMO Interaction Motif” (SIM)) of sequence ϕ-X-ϕ-ϕ or ϕ-ϕ-X-ϕ, where ϕ is hydrophobic and generally Leu, Ile, or Val.20; 21; 22 SBMs adopt a β-strand structure, and bind a groove in SUMO with orientation depending on whether the pair of hydrophobic side-chains is at the N- or C-terminus of the motif.21 SBM binding by SUMO can exert either negative or positive effects on intermolecular interactions, depending on context. For example, SUMO binding to the SBM of its covalently attached thymine DNA glycosylase disrupts DNA binding,23 whereas SUMO-1 modification of RanGAP1 promotes interaction with the SBM of the nuclear pore component Nup358/RanBP2.24; 25 Notably, Nup358/RanBP2 is also a SUMO E3, and for this activity the SBM can position SUMO-1 relative to Ubc9 for the ligation reaction.26 After the reaction, the Nup358/RanBP2 SBM remains bound to SUMO-1, allowing recruitment of modified RanGAP1 to the nuclear pore.
Several high-resolution structures have provided insights into both general characteristics of Ubl conjugation, and specialized features of the SUMO cascade. These include multiple structures of Ulp/SENP proteases bound to Smt3p/SUMO representing different steps in Smt3p/SUMO processing and deconjugation (for examples, see references 27; 28; 29), the human SAE1–SAE2 E1 complexed with SUMO-1 at the adenylation active site,30 and several structures of Ubc9 bound to the ϕ-K-X-D/E-sequence-containing target RanGAP-1, including SUMO-1-modified RanGAP-1 in complex with Ubc9 and Nup358/RanBP2.26; 31; 32 SUMO recognition by SBMs was revealed by this structure, and also by structures of SUMO-bound to thymine DNA glycosylase, and by the structure of a SUMO–PIASX peptide complex.21; 26; 33 Further insights into SUMO conjugation can been inferred from structures of complexes from the ubiquitin or NEDD8 pathways that represent interactions common among Ubl cascades. Nonetheless, an intriguing complex remains less characterized: SUMO and its E2, Ubc9, bind each other, noncovalently, with high affinity.13; 34; 35 The functions of noncovalent SUMO–Ubc9 interactions remain poorly understood. To gain high-resolution insights into this complex, and to understand the potential for SUMO and Ubc9 in the noncovalent complex to interact with their other binding partners in the cascade, we determined the structure of a noncovalent Smt3p–Ubc9p complex from S. cerevisiae.
Our cloning, expression, and purification of Smt3p, Ubc9p, and Smt3p's heterodimeric E1 complex (Aos1p–Uba2p) have been described previously,13 although additions to protocols for proteins in the crystallized complex are described here. For the crystallized complex, Smt3p and Ubc9p were generated by coexpression in BL21(DE3), from the pRSF-Duet vector, where a thrombin cleavage site-containing linker sequence (DPLVPRGS) was inserted between the His6SQ-tag and Smt3p's residue 13 (residue 13 was based on previous findings that residues 1-13 are disordered, and that Smt3p starting at residue 13 or 20 is functional14; 36), and full-length Ubc9p was expressed untagged. The complex was purified by Ni2+-affinity chromatography (His-Select resin, Sigma) and by gel filtration chromatography (SD200 column, GE Healthcare) in 25 mM Tris-Cl, 150 mM NaCl, 2 mM β-mercaptoethanol, pH 7.6, and concentrated to 28 mg/ml by ultrafiltration (Amicon). Crystals were obtained only for the uncleaved His6SQ/thrombin site linker/Smt3p–Ubc9p complex, by the hanging drop vapor diffusion method at 4°C in 18% PEG 3350, 0.2M MgCl2, 0.1 Bis-Tris pH 5.5. Crystals were cryoprotected in mother liquor supplemented with 4% PEG 3350 and 30% glycerol prior to flash-freezing in liquid nitrogen and data collection. The structure was refined to a resolution of 1.9 Å (Table 1, Fig. 1). In an unexpected coincidence, the thrombin cleavage site-containing linker sequence loosely mimics an SBM sequence. Crystal packing explains the requirement for the uncleaved thrombin-site linker for forming these crystals: the linker from one Smt3p forms a β-strand that docks in the groove from the adjacent symmetry-related Smt3p, in a manner resembling SBM binding to human SUMO family members.21; 26; 33 The model contains two complexes per asymmetric unit. One contains all residues except the His6SQ-tag, Smt3p residues 96-98, and Ubc9 residues 1-2 and 157, which were not visible in the electron density. These and Smt3p residues 15-18 and 95 are not visible in the other complex.
In order to precisely characterize the association state of a Smt3p and Ubc9p mixture, we performed analytical ultracentrifugation analysis. Sedimentation equilibrium and velocity experiments provide experimentally determined molecular weights of 29.75 and 29.88 kDa, respectively, which agree well with the calculated molecular weight of 30.08 kDa for a 1:1 Ubc9p–Smt3p complex (Supplementary Fig. 1). Therefore, as both complexes in the asymmetric unit superimpose well (rmsd 0.738 Å), only one is discussed below.
Smt3p residues 21-94 adopt the canonical, globular ubiquitin-like β-grasp fold, with a twisted β-sheet forming one face, and an α-helix forming the other, resembling the previous solution structure of free Smt3p (1.572 Å rmsd)36 and the previous crystal structure of Smt3p in complex with the processing/deconjugating enzyme Ulp1p (0.658 Å rmsd) (Fig. 1).27 Ubc9p from the complex superimposes well with the previous crystal structure of free Ubc9p (0.406 Å rmsd),37 adopting the canonical E2 domain fold comprising a central anti-parallel 4-stranded β-sheet surrounded by 4 α-helices (Fig. 1).
Overall, the Smt3p–Ubc9p complex adopts a “V” shape (Fig. 1c). Smt3p contributes one side of the “V”, with Ubc9p comprising the base and the other side. A broad, concave surface involving Smt3p's conserved β-sheet cradles a complementary, relatively localized surface from Ubc9p's N-terminal α-helix and β-strand (Fig. 1c). This N-terminal Ubc9p face is completely separate from the surface containing the catalytic Cys93: Cys93 is on the opposite side, and >25 Å away (Fig. 1c). In total, Smt3p–Ubc9p interactions bury ~1400 Å2. Several observations suggest that these noncovalent interactions are conserved between Ubc9 and SUMO from other organisms. Many interface side chains are conserved in SUMO and Ubc9 family members (Fig. 1a, b). Also, NMR studies of human SUMO-1–Ubc9 and SUMO-3–Ubc9 complexes revealed the greatest chemical shift perturbations upon complex formation for Ubc9 resonances from the N-terminal helix and strand, and from the SUMO β-sheet.34; 38; 39 Although an NMR-based docking model suggested that these regions of Ubc9 and SUMO-3 undergo significant structural rearrangement as compared to existing crystal structures,39 we find no evidence of such rearrangement in the corresponding proteins from S. cerevisiae. High-resolution insights into a corresponding complex between human orthologs is described by Lima and colleagues.
The Smt3p–Ubc9p complex is stabilized by a combination of hydrophobic and electrostatic elements. The interface is anchored by hydrophobic contacts between Ubc9p's His20, Phe22, Gly23, and Tyr25 associating with Smt3p's Asp30, Gly31, Asp68 and Gly69 backbone carbons, Glu90 side-chain carbons, and Ile88 and His92 side-chains (Fig. 1d). The complex is further secured by a network of electrostatic interactions. A basic patch from Ubc9p's Arg13, Arg17, Lys27, and Lys30 side-chains interacts with an acidic patch involving Smt3p's Tyr67 side-chain hydroxyl group, Asp68, Asp82, Glu84 and Asp87. Much of this electrostatic network is water-mediated (not shown), but there are also many direct interactions, including between side-chains from Ubc9p's Arg17 and Smt3p's Asp82 and Asp87 (Fig. 1d). There are additional electrostatic contacts between Ubc9's Asp19 and Smt3p's Lys27 side-chains, and also between Smt3p's Glu90 side-chain oxygens with Ubc9p's His20 side-chain and Gly23 backbone nitrogens (Fig. 1d).
The relevance of the structurally observed contacts is confirmed by a previous mutational analysis of Ubc9p, where Ubc9p surface residues were swapped with sequences found on ubiquitin E2s, with the goal of identifying protein binding sites without perturbing the E2-fold structure.13 In agreement with key roles indicated by our Smt3p–Ubc9p structure, an R17A mutant Ubc9p, and a triple mutant version of Ubc9p harboring F22A, G23Q, and Y25S substitutions, showed significant defects in noncovalent interactions with Smt3p.13 These residues are contained within a 38-residue N-terminal stretch of Ubc9p's sequence we found previously to be important for function in vivo.37 Indeed, we find that neither the R17A nor the F22A/G23Q/Y25S mutants were able to complement the viability defect in ubc9Δ cells (Fig. 1e), or the temperature-sensitivity of the ubc9-10 strain of S. cerevisiae in the presence of the genotoxic agents hydroxyurea (HU) or methyl methane sulfonate (MMS) (Fig. 1f).
The structure reveals that interactions observed in the noncovalent Smt3p–Ubc9p complex are distinct from those in the covalent Ubc9p~Smt3p thioester-linked conjugate. The large distance between Ubc9's noncovalent Smt3p binding site and its catalytic cysteine, which is the site of the thioester linkage with Smt3p's C-terminus in the conjugation intermediate, would preclude a thioester-bound Smt3p from simultaneously occupying the noncovalent binding site of the same molecule of Ubc9p.
Despite the topological incompatibility of a single molecule of Smt3p interacting at both sites with a single molecule of Ubc9p, comparison of our noncovalent Smt3p– Ubc9p structure with a structural analog of Ubc9~SUMO reveals that in the thioester intermediate, the noncovalent interaction surfaces are exposed in both Ubc9 and SUMO (Fig. 2a-c). A model for covalent interactions between SUMO and Ubc9's catalytic Cys93 face comes from the structure of the human SUMO-1~RanGAP1–Ubc9– Nup358/RanBP2 complex.26 Although this contains a trapped-product complex where SUMO-1 is conjugated to a lysine from RanGAP1, the SUMO-1 C-terminus is adjacent to Ubc9's catalytic Cys93, suggesting a good structural model for a covalent Ubc9~SUMO thioester complex,26 in which interactions primarily involve Ubc9's catalytic Cys93 face and SUMO's C-terminal tail and β1β2-loop (Fig. 2b-c), which do not overlap the noncovalent Smt3p–Ubc9p binding sites. Thus, it is possible that higher-order complexes could form, where a single Smt3p covalently linked to one Ubc9p's catalytic cysteine also binds noncovalently to another molecule of Ubc9p. Such higher-order complexes (Smt3p~Ubc9p–Smt3p~Ubc9p and so on in S. cerevisiae, SUMO~Ubc9–SUMO~Ubc9 in higher eukaryotes) could also interact with targets. Indeed, an analogous higher order complex has been found between ubiquitin and the ubiquitin E2, UbcH5: ubiquitin bound covalently via a thioester at the active site of one UbcH5 is also bound noncovalently to the “backside” of another UbcH5.40 In the case of ubiquitin and UbcH5, gel filtration analysis was used to confirm the nature of these interactions, where a higher-order assembly of the wild-type UbcH5~ubiquitin thioester elutes earlier than the corresponding complex containing a UbcH5 mutant defective in noncovalent binding to ubiquitin.40 We performed such an analysis for the Smt3p pathway, taking advantage of the previously described F22A/G23Q/Y25S triple mutant version of Ubc9p, which does not bind noncovalently to Smt3p, but which maintains wild-type ability to form a covalent thioester complex with Smt3p.13 The F22A/G23Q/Y25S-Ubc9p~Smt3p thioester complex elutes from the gel filtration as a compact, symmetric peak, consistent with one molecule of each F22A/G23Q/Y25SUbc9p and Smt3p in the conjugated complex (Supplementary Fig. 2). However, as with the higher-order UbcH5~ubiquitin complexes,40 the wild-type Ubc9p~Smt3p thioester complex elutes earlier, as a broad, asymmetric peak (Supplementary Fig. 2). These results are consistent with formation of larger molecular weight, higher order complexes.
Smt3p and Ubc9p interact with numerous other proteins in the SUMO conjugation cascade. We sought to understand which other proteins and complexes in the pathway might also involve Smt3p–Ubc9p noncovalent interactions, and which assemblies cannot. Comparison with existing structures shows that residues at the Smt3p–Ubc9p interface also play critical roles in forming many other Smt3p and Ubc9p complexes with roles early in the cascade. During Smt3p conjugation, Smt3p is first processed by a Ulp protease, and then Smt3p is activated by Aos1p–Uba2p, Smt3p's dedicated heterodimeric E1 enzyme. The noncovalent Ubc9p binding site overlaps with 44% of the Smt3p surface that interacts with the Ulp1p protease (Fig. 2a, d). Although there is no structure of Smt3p in complex with Aos1p–Uba2p, a high degree of sequence identity suggests that this will resemble the corresponding structure containing human SUMO-1 (45% identical to Smt3p) bound to the human SUMO E1 (the heterodimeric SAE1–SAE2 complex, 36% identical to Aos1p–Uba2p).30 50% of Smt3p's Ubc9 binding surface overlaps the corresponding SAE1–SAE2 binding surface of human SUMO-1 (Fig. 2a, e). Together, Ulp1 and E1 bind the key Smt3p residues involved in the noncovalent interaction with Ubc9p.
During Smt3p activation, the Aos1p–Uba2p E1 also interacts with Ubc9p, to generate the covalent Smt3p~Ubc9p thioester-linked intermediate, which is distinct from the noncovalent Smt3p–Ubc9p complex. Although there is no structure of Ubc9p in complex with Aos1p–Uba2p, structures of E1–E2 complexes are available from the pathway for the Ubl NEDD8,41; 42 and the E2-binding domains from the SUMO/Smt3p and NEDD8 E1s share related folds.30; 43 The structures show N-terminal helix residues from NEDD8's E2 (Ubc12) binding to the E1.41; 42 Some of these E1-binding residues from Ubc12 correspond to Smt3p-binding residues from Ubc9p (Fig. 2a, f). This suggests that Smt3p and Aos1p–Uba2p bind partially overlapping surfaces on Ubc9p. Indeed, a previous study showed that mutation of Ubc9p's N-terminal helix residue Arg17 reduces binding to both Smt3p and Aos1p–Uba2p.13 Interestingly, there may be subtle species-specific differences in SUMO E1-Ubc9p interactions, because mutation of basic residues, including Arg17, from human Ubc9's N-terminal helix reduces binding to SUMO-1 without significantly affecting binding to SAE1-SAE2,35 and Aos1p-Uba2p charges human Ubc9 very poorly37. To further probe whether S. cerevisiae Smt3p and Aos1p–Uba2p bind overlapping, but non-identical surfaces on Ubc9p, we used an in vitro gel mobility shift binding assay. We found previously that in such an assay, titration of excess free Smt3p leads to displacement of Ubc9p from an Aos1p–Uba2p–Ubc9p complex into a Smt3p–Ubc9p complex.13 To test whether this comes from Smt3p and Aos1p–Uba2p competing to bind overlapping sites on Ubc9p, we used the F22A/G23Q/Y25S triple mutant version of Ubc9p. Phe22, Gly23, and Tyr25 are all located at the Smt3p–Ubc9p interface, but are distal from Ubc9p's N-terminal helix implicated in binding to both Smt3p and Aos1p–Uba2p. Therefore, if Smt3p and Aos1p–Uba2p compete for binding to Ubc9p's N-terminal helix, then the F22A/G23Q/Y25S triple mutant version of Ubc9p, which does not bind noncovalently to Smt3p, would be able to bind Aos1p–Uba2p equally well in the absence or presence of Smt3p. In competition experiments, we found that unlike the case with wild-type Ubc9p, excess free Smt3p is unable to displace the F22A/G23Q/Y25S triple mutant Ubc9p from the complex with Aos1p–Uba2p (Supplementary Fig. 3). These results support the notion that parts of the Smt3p and Aos1–Uba2p binding surfaces on Ubc9p overlap.
Our structure suggests that noncovalent Smt3p–Ubc9p interactions may coexist in some complexes involved in latter steps conjugation cascades. During conjugation, following formation of the covalent Ubc9p~Smt3p thioester intermediate, Ubc9p can interact with an E3, which promotes Smt3p transfer from Ubc9p's catalytic Cys93 to the target's lysine. All presently characterized S. cerevisiae Smt3p E3s (Siz1p, Siz2p, Mms21p, and Zip3p) are members of the SP-RING family.7 Although there are no reported structures of any Ubc9–SP-RING E3 complexes, there are related structures of E2–RING E3 complexes from the ubiquitin pathway.44; 45 There is no structural overlap upon superposition of Ubc9p–Smt3p onto structures of ubiquitin E2s in complex with RING and RING-related domains (Fig. 2g). By contrast, many human Ubc9 residues involved in binding the mammalian E3 Nup358/RanBP226 overlap with the corresponding noncovalent Smt3p-binding site on S. cerevisiae Smt3p. Thus, there may be variation in the abilities of different E3s to interact with a noncovalent SUMO–Ubc9 complex.
Following conjugation, SUMO binds specific recognition machineries, resulting in altered function of the SUMOylated target. Structural insights into interactions of Smt3p with endogenous recognition machineries remain poorly characterized, although detailed binding and structural studies are available for some human SBM–SUMO complexes.21; 26; 33 Like human SUMO-1, Smt3p was found to bind an SBM peptide derived from human PIASX with high affinity.21 Our structure suggests that the noncovalent Smt3p–Ubc9p complex can interact with Smt3p recognition machineries, and provides insight into how Smt3p recruits its S. cerevisiae protein partners like mammalian SUMOs recruit SBMs. Interestingly, a portion of the uncleaved thrombin-site linker sequence (DPLVPRGS) inserted between the His6SQ-tag and Smt3p loosely resembles the ϕ–X-ϕ-ϕ sequence of an SBM, with Leu, Pro, and the substantial aliphatic portion of an arginine side-chain all hydrophobic. This engineered sequence attached to one Smt3p binds the symmetry-related Smt3p in the crystal lattice, in a manner resembling SBM binding to human SUMO family members (Fig. 3a-e). The Leu and Pro, and the hydrophobic portion of the Arg side-chain, bind in a conserved hydrophobic groove between Smt3p's β2-strand α-helix comprising Smt3p's Ile35, Phe36, Phe37, Ile39, Arg47, Leu48, Ile51, and Arg55. The hydrophobic regions of corresponding side-chains in mammalian SUMO family members also play critical roles in recruiting bonafide mammalian SBMs (Fig. 3a-e).20; 21; 22; 26; 33
Just as the three structurally characterized mammalian SBMs make unique contacts with their SUMO partners, there are distinguishing features of the linker–Smt3p interface (Fig. 3c-e). The proline in the linker sequence induces a kink in the strand not found in the existing SBM-SUMO structures. As a result, the proline backbone oxygen is able to form a hydrogen bond with Smt3p's Arg55. Critical electrostatic interactions are also mediated by the corresponding Arg conserved in human SUMO family members, but in some cases with SBM acidic side-chains in structural locations corresponding to the proline oxygen.21; 26; 33 Also, Smt3p's Ser33 side-chain forms a hydrogen bond with the side-chain from the Glu corresponding to the adjacent Smt3p's residue 13.
Additionally, Smt3p's Asn25 forms a hydrogen bond with the linker Arg side-chain (Fig. 3d). Such interactions might occur with human SUMO-2 and -3, which also contain an Asn at this position, but would likely be destabilized by the corresponding lysine in the sequence of SUMO-1 (Fig. 1a).
The Smt3p–Ubc9p structure globally resembles two recent high-resolution structures of noncovalent complexes for ubiquitin with (i) an active ubiquitin E2 (UbcH5),40 and (ii) a catalytically-inactive Ubiquitin E2 Variant protein (MMS2)46 (Fig. 3b,f, g). All three noncovalent E2 (or UEV)–ubiquitin-like protein complexes use parallel interaction surfaces (Fig. 3b,f, g). The functions of ubiquitin–E2 noncovalent interactions are better characterized, with clear roles in multiprotein complex formation. Higher-order (UbcH5~ubiquitin-UbcH5~ubiquitin)x self-assembled thioester complexes have been implicated in BRCA1 E3-mediated processive polyubiquitin chain assembly.40 Ubiquitin binding to the ubiquitin E2-variant (UEV) protein, MMS2, also plays a role in assembly of a complex for polyubiquitin chain assembly. UEV domains are structural homologs of E2s, but they lack the active site cysteine and C-terminal helices of bone fide E2 conjugating enzymes. MMS2 recruits ubiquitin, noncovalently, and directs ubiquitin's Lys63 toward the active site of the E2, Ubc13, to establish linkage-specificity of MMS2–Ubc13-catalyzed polyubiquitin chain assembly.46; 47; 48; 49 Although the mechanisms underlying polymeric SUMO chain formation, and the functions of polymeric SUMO chains themselves, remain poorly characterized, structural similarity with complexes involved in polyubiquitin chain assembly raise the possibility that the noncovalent Smt3p–Ubc9p complex will eventually be found to play a role in polymeric SUMO chain formation.
Our structure of the noncovalent Smt3p–Ubc9p complex, the structure of the human SUMO-1–Ubc9 complex described by Lima and colleagues, and the previous ubiquitin–E2/UEV structures40; 46 reveal conservation of noncovalent Ubl–E2 interactions. The striking structural similarity - for different Ubls, for divergent E2/E2-like proteins, and across phyla - implies important functions for these complexes. In the ubiquitin pathway, noncovalent ubiquitin–E2/UEV complexes interact with multiple additional partners, as parts of large assemblies involved in polyubiquitin chain formation. Our serendipitous finding of an SBM mimic also bound to Smt3p–Ubc9p suggests that the noncovalent Smt3p–Ubc9p complex (and mammalian SUMO–Ubc9) exists as a part of larger assemblies, linked to other proteins or complexes by simultaneous SBM–Smt3p and Smt3p–Ubc9p interactions. We anticipate that future studies will reveal how larger assemblies involving SBM–Smt3p–Ubc9p interactions fit into the broader context of the Smt3p cascade, to yield the precise post-translational regulation that make this pathway essential.
We thank K. Bencsath for initiating our studies of the noncovalent Smt3p–Ubc9p complex, D. Huang for the crystallized Smt3p–Ubc9p coexpression construct, and C. Ross for computer support. We are grateful to J. Chrzas and Z. Jin for performing mail-in data collection at the SER-CAT (22-BM) beamline at the Advanced Photon Source (APS). Use of the APS was supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. W-31-109-Eng-38. This work was supported by the American Lebanese Syrian Associated Charities (ALSAC), the National Institutes of Health (P30CA21765 for St. Jude Cancer Center support, R01GM069530 to BAS, R01CA111542 to MAB), a Beckman Young Investigator Award to BAS, an American Cancer Society Postdoctoral Fellowship to DMD, and the Howard Hughes Medical Institute. BAS is an Investigator of the Howard Hughes Medical Institute.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Protein Data Bank accession numbers
Coordinates and structure factors have been deposited in the RCSB Protein Data Bank under the filenames 2EKE.