Crystal Structure of UBA2ufd-Ubc9: Insights into E1-E2 Interactions in Sumo Pathways

Canonical ubiquitin-like proteins (UBLs) such as ubiquitin, Sumo, NEDD8, and ISG15 are ligated to targets by E1-E2-E3 multienzyme cascades. The Sumo cascade, conserved among all eukaryotes, regulates numerous biological processes including protein localization, transcription, DNA replication, and mitosis. Sumo conjugation is initiated by the heterodimeric Aos1-Uba2 E1 enzyme (in humans called Sae1-Uba2), which activates Sumo's C-terminus, binds the dedicated E2 enzyme Ubc9, and promotes Sumo C-terminal transfer between the Uba2 and Ubc9 catalytic cysteines. To gain insights into details of E1-E2 interactions in the Sumo pathway, we determined crystal structures of the C-terminal ubiquitin fold domain (ufd) from yeast Uba2 (Uba2ufd), alone and in complex with Ubc9. The overall structures of both yeast Uba2ufd and Ubc9 superimpose well on their individual human counterparts, suggesting conservation of fundamental features of Sumo conjugation. Docking the Uba2ufd-Ubc9 and prior full-length human Uba2 structures allows generation of models for steps in Sumo transfer from Uba2 to Ubc9, and supports the notion that Uba2 undergoes remarkable conformational changes during the reaction. Comparisons to previous structures from the NEDD8 cascade demonstrate that UBL cascades generally utilize some parallel E1-E2 interaction surfaces. In addition, the structure of the Uba2ufd-Ubc9 complex reveals interactions unique to Sumo E1 and E2. Comparison with a previous Ubc9-E3 complex structure demonstrates overlap between Uba2 and E3 binding sites on Ubc9, indicating that loading with Sumo and E3-catalyzed transfer to substrates are strictly separate steps. The results suggest mechanisms establishing specificity and order in Sumo conjugation cascades.


Introduction
Post-translational modification by ubiquitin-like proteins (UBLs) is a major mechanism for regulating eukaryotic protein functions. UBLs generally become covalently attached to specific targets through a series of molecular ''handoffs'' involving multienzyme cascades consisting of an E1 activating enzyme, an E2 conjugating enzyme, and an E3 ligase (reviewed in [1]). Thus, it is of great interest to understand how E1, E2, and E3 enzymes interact with each other for UBL transfer.
Like other ubiquitin-like proteins (UBLs), the Small ubiquitinrelated modifier (Sumo) proteins become covalently ligated to targets (reviewed in [2,3,4]). Attachment of Sumo family UBLs are known to alter target functions such as protein-protein interactions, protein-DNA interactions, and subcellular localization (reviewed in [5]). As such, Sumo regulates many important processes, such as signaling, transcription, DNA repair and other stress responses, the cell cycle, and apoptosis [2,3,4]. Indeed, the budding yeast Smt3 protein (for simplification referred to as Sumo hereafter) regulates chromosome segregation, formation of the septin ring, and many other aspects of cell division [6,7,8].
Sumo family members are ligated to proteins via specific E1, E2, and E3 enzymes. The Sumo-specific E1 enzyme [the heterodimeric complex between Uba2 and Aos1 (yeast; termed Sae1-Uba2 in mammals)] initiates the process by first catalyzing adenylation of the Sumo C-terminus, which next becomes linked by a thioester bond to Uba2's catalytic cysteine [9,10,11,12]. A transthiolation reaction ensues during which Sumo is transferred from Uba2 to the catalytic cysteine of the dedicated Sumo E2 conjugating enzyme, Ubc9 [13,14,15,16]. Ultimately, either with or without facilitation by a Sumo-pathway-specific E3, Sumo is transferred from the Ubc9 catalytic cysteine to a target lysine ( [17] and references therein). In some cases, repeated cycles of Sumo transfer lead to generation of polySumo chains on targets (for review, see [18]). In recent years, structural studies have provided details for many aspects of Sumo E1-E2-E3 conjugation cascades. These include structural understandings of how human Sumo is recognized and activated by Sae1-Uba2 [19,20], how Ubc9 recognizes targets [17,21,22,23], and how two distinctive E3s function in Sumo ligation [24,25,26]. Nonetheless, many fundamental aspects of the Sumo cascade remain incompletely understood.
For example, how does the Sumo E1 bind Ubc9? The structure of the human heterodimeric Sae1-Uba2 complex that comprises the Sumo E1 [19] displays conserved E1 domains [27,28]: a heterodimeric ''adenylation'' domain comprised of portions of both Sae1 and Uba2, a domain of Uba2 containing the E1 catalytic cysteine, and a C-terminal ubiquitin-fold domain (ufd) from Uba2 that structurally resembles ubiquitin and Sumo [29]. For human Uba2, the catalytic cysteine domain has been shown to make weak interactions with Ubc9 [30]. In addition, for both yeast and human Uba2, the ufd has been implicated in Ubc9 recruitment.
For canonical UBL pathways, E2s generally appear to bind an E1 ufd [29]. Indeed, crystal structures from the pathway of another UBL, NEDD8, showed previously that the NEDD8 E1's ufd binds NEDD8 E2s, either Ubc12 (also known as UBE2M) or UBE2F, via the E2 N-terminal helix and b1b2-loop [31,32,33]. These E1 ufd -E2 interactions appear to be conserved across the Sumo, NEDD8, ISG15 and ubiquitin pathways [29]. For example, replacing the UBA6 or UBA7 ufds with that of UBA1 was sufficient to swap E2 specificities, and substituting the E2 UbcH7's N-terminal helix and b1b2-loop regions with that of the E2 UbcH8 swaps E1 specificity [34,35]. In terms of the Sumo pathway, the human Uba2 ufd was also shown to bind Ubc9 and to play an essential role in human Sae1-Sae2-mediated Sumo transfer to Ubc9 [19,36]. The ufd of yeast Uba2 is essential for viability [19]. And from the E2 side, budding yeast Ubc9's Nterminal helix and b1b2-loop were also shown to be important for binding to Uba2 and formation of a Ubc9,Sumo complex [37]. Notably, the Ubc9 b1b2-loop is significantly extended relative to other E2s, and has long been recognized as a unique structural feature of Ubc9 [38,39]. Nonetheless, there is no high-resolution data for E1 ufd -E2 interactions other than those for the NEDD8 pathway, despite their importance for the Sumo cascade in particular and UBL transfer in general. Therefore, to obtain a more detailed understanding of Uba2's ufd interactions with Ubc9 and to obtain broad insights into how E1-E2 specificity is established in general, we performed structural analysis of the Uba2 ufd -Ubc9 complex from S. cerevisiae.

Results and Discussion
Structure of an isolated ubiquitin-fold domain from Uba2 (Uba2 ufd ) To characterize a domain from yeast Uba2 that binds Ubc9, we determined the crystal structure of the isolated yeast Uba2 ubiquitin-fold domain (ufd), corresponding to residues 439-563, at 1.6 Å resolution (Table 1, Figure S1, Figure 1). For the most part, residues 440-551 were clearly visible in the electron density with all secondary structures well-defined and for consistency numbered here according to their counterparts in the prior structures of full-length human Uba2 [19]. Only a loop encompassing residues 531-540 displayed weaker electron density precluding building of several side-chains and resulting in high Bfactors, presumably due to greater flexibility.
Like ubiquitin and Sumo, the Uba2 ufd adopts a modified bgrasp fold. Uba2 ufd consists of a twisted 5-stranded antiparallel bsheet on one side, one 3-turn a-helix on the opposite side of the sheet, and two peripheral short helices. The b-sheet and connecting loops form one V-shaped surface. A second V-shaped surface comes from the edge of b-strand 23, the following loop, and a-helix37. The two V-like structures are adjacent to each other, and together form a W-shaped surface ( Figure S2).

Structure of a Uba2 ufd -Ubc9 complex
In order to understand E1-E2 interactions in a Sumo pathway, we determined the crystal structure of a Uba2 ufd -Ubc9 complex from yeast. Phases were obtained by molecular replacement, using the prior structure of yeast Ubc9 and a polyalanine model of human Uba2 residues 449-546 as searchmodels [19,40]. There are two molecules per asymmetric unit, which superimpose with 0.3 Å root mean square deviation (rmsd) over all alpha carbons, so only one version of the complex is discussed here. The structure of the Uba2 ufd is nearly identical in the complex as on its own (0.47 Å rmsd). The structure of Ubc9 superimposes well with the prior structure of free Ubc9 (0.55 Å rmsd), with the most notable difference being slight variation in the position of the b1b2-loop. The Uba2 ufd -Ubc9 complex adopts an overall oblong elliptical shape, with the W-shaped surface from Uba2 ufd binding the Nterminal a-helix1 and b1b2-loop from Ubc9 ( Figure 1, Figure  S2). This interaction is distal -,30 Å -to the Ubc9 catalytic cysteine. The Uba2 ufd -Ubc9 interface buries 1557 Å 2 of surface area [41].

Details of Uba2 ufd -Ubc9 interactions
The Uba2 ufd -Ubc9 interface is interdigitated, with the two Vshaped grooves from Uba2 binding alternating regions of Ubc9, as in a layered sandwich. The ''V'' corresponding to the concave Uba2 ufd b-sheet holds the Ubc9 N-terminal a-helix1 along its entire length (Figure 2A). Ubc9's Leu6 is loosely surrounded by a hydrophobic surface from Uba2's Leu478, Leu485, and Leu511 from three distinct strands. The remainder of the interaction is largely ionic (Figure 2A). The Ubc9 N-terminus forms a hydrogen bond with Uba2's Gln483. Ubc9's Gln10 makes a hydrogen bond with Uba2's Ser476 and Glu515. Ubc9's Arg13 makes salt bridges with Uba2's Asp488 and Asp490, and along with Ubc9's Lys14 interacts with Uba2's Tyr489. Ubc9's Lys14 and Arg17 interact with Uba2's Asp474 and Asp490, respectively. At the C-terminus of Ubc9's a-helix1, Lys18 forms a hydrogen bond with the backbone oxygen from Uba2's Gly517.
The structure helps explain the deleterious effects of previously reported mutants in yeast Ubc9, which showed that residue substitutions in place of Lys14, or a combination of Lys14, Arg17, and Lys18 from the N-terminal helix, or deleting a portion of the b1b2-loop, hinders binding to full-length Aos1-Uba2 [37]. We further tested the effects of mutating additional key interface residues. Consistent with the structure, individual Ala substitutions in place of Ubc9's Arg13, Met36, or Leu38 hinders Aos1-Uba2mediated generation of a Ubc9,Sumo complex ( Figure 2B, Figure S3). Similar results are observed for mutating the key acidic patch residues Asp488 and 490 from Uba2 ( Figure 2B, Figure S3).

Conservation of Ubc9 and Uba2 interacting residues across species
Several structural features and individual amino acids involved in the Uba2 ufd -Ubc9 interaction are conserved among Ubc9s and Uba2s from yeast to humans. Yeast and human Ubc9 share 56% sequence identity ( Figure 1D). Accordingly, the Uba2-binding side-chains from the Ubc9 a-helix1 are 100% conserved as basic residues. Also, the b1b2-loop structure is conserved among Ubc9s, but is uniquely extended relative to E2s for other UBLs. Here, Asp33 is conserved as acidic, and Met36 and Leu38 are conserved as hydrophobics.
The sequence similarity is reflected by extensive structural similarity between yeast and human Ubc9 (0.53 Å Ca rmsd, previously noted for other structures in [40]). While the many Ubc9 structures, from different organisms and in different complexes, do generally superimpose well, the primary site of very minor differences is in the orientation of the b1b2-loop. It is possible that limited flexibility of this unique Ubc9 structure may play a role in binding to the Uba2 ufd.
The sequences between Uba2 ufds are generally less conserved ( Figure 1E, Figure 3A, Figure 3B). Despite only 17% sequence identity, a search of the Protein Data Bank using DALI identifies the ufd region from a prior structure of full-length human Uba2 (also called Sae2) as having the highest degree of structural similarity with the yeast Uba2 ufd (Ca rmsd 1.85 Å ) [19,20,42]. The main differences are localized to two regions, which are distal to the Ubc9-binding site. The first difference is the short a-helix36a, which corresponds to an insertion in the sequence of yeast Uba2 between b-strand21 and a-helix 36 ( Figure 1E, Figure 3A, Figure 3B). The second is at the C-terminal portion of the domain where the sequences are most divergent: the loop preceding b-strand24, the loop between b-strands24 and 25, and the region following b-strand25 that is disordered in human Uba2 but in the yeast Uba2 ufd is ordered and forms a more extended structure. Nonetheless, the striking similarity of the overall structures of yeast and human Uba2 ufds supports the notion that the structurally observed yeast Uba2 ufd -Ubc9 interactions will be preserved across species. Indeed, some key interface residues are relatively conserved, including Ser476 (Ser or Asp across species), Arg484 (charged or polar), Leu485 (Leu, Ile, or Val across species), and Asp488 and Asp490 (Ser or Asp, and Asp or Glu Residues in additional allowed regions (%) 6.3 9.7 Residues in disallowed regions (%) 0.9 0 Highest resolution shell is shown in parenthesis. R work = g|F o -F c |/gF o . R free is the cross-validation of R-factor, with .5% of the total reflections omitted during model refinement. doi:10.1371/journal.pone.0015805.t001 across species, respectively) ( Figure 1E). Furthermore, human Ubc9 was shown to cause chemical shift perturbations in the corresponding region of human Uba2 ufd [36].
Comparison of Uba2 ufd -Ubc9 with prior E1ufd structures from other UBL pathways To gain insights into similarities and differences among E1 ubiquitin-fold domains, we compared the yeast Uba2 ufd -Ubc9 crystal structure with the corresponding regions of crystal structures of yeast Uba1 (E1 for ubiquitin) and human Uba3 (E1 for NEDD8) ( Figure 3A, Figure 3C, Figure 3D, Figure S4). The yeast Uba2 ufd displays overall structural similarity to the ufds from Uba1 and Uba3 (2.4 and 1.6 Å rmsd, respectively), including the W-shaped surface. Furthermore, structural comparison of Uba2 ufd -Ubc9 with prior structures of Uba3 in complex with NEDD8 E2s (Ubc12 or UBE2F) reveal that the Sumo and NEDD8 pathways use parallel E1 ufd and E2 surfaces to mediate interactions ( Figure 3E, Figure 3F).
Despite their overall common mode of binding, an intriguing difference is that the relative positions of the E1 ufd and E2 catalytic cysteines are considerably offset between Uba2 ufd -Ubc9 and Uba3 ufd -Ubc12 or Uba3 ufd -UBE2F ( Figure 3E, Figure 3F). The detailed interactions between Uba2 and Ubc9 described above (Figure 2A) also differ substantially from corresponding contacts in the NEDD8 pathway [31,33]. Furthermore, global structural differences between Uba2 and other E1 ufds, and between Ubc9 and other E2s help explain why Ubc9 does not bind E1s for other UBLs and vice-versa. In the Uba1 ufd the region corresponding to the Uba2 ufd 's short a-helix37 is extended by a complete turn, and the preceding loop in Uba2 ufd corresponds to a 2-turn helix in Uba1 ( Figure 3C) [28]. This latter helix is even longer in Uba3's ufd ( Figure 3D) [27,31]. These Uba1 and UBA3 helical insertions would clash with the Uba2-binding basic residues from Ubc9's a-helix1 and extended b1b2 loop. Thus, several structural differences between E1 ufd s, and corresponding distinctions between their cognate E2s, may help establish specificity of E1-E2 interactions.
To address the possibility of whether Sumo thioester-linked to the catalytic Cys177 of one Uba2 can be transferred to Ubc9 bound to another Uba2, we examined generation of the Ubc9,Sumo thioester-linked conjugate upon mixing two defective mutant versions of the Sumo E1. In one mutant, the Uba2 catalytic Cys177 was mutated to Ala to prevent formation of a Uba2,Sumo intermediate, but the ufd was wild-type. The other retained ability to form a thioester linkage with Sumo at Cys177, but harbored Ala mutations in the Ubc9-binding ufd residues Asp488 and Asp490. On their own, each mutant is defective for generating the thioester-linked Ubc9,Sumo conjugate. Mixing the two E1s does not rescue the other's defect ( Figure 4B).
Thus, it seems likely that Uba2 undergoes a conformational change to bring the Uba2 and Ubc9 catalytic cysteines together for Sumo transfer. Previous structural studies of the NEDD8 E1, which showed dramatic ufd reorientation, may provide some insights into such a conformational change. Superimposing Uba2 ufd -Ubc9 on the Uba3 ufd in a structure of the NEDD8 E1 complexed with two molecules of NEDD8 and a catalyticallyinactive E2 (Ubc12), which represents an intermediate en route to UBL transfer from E1 to E2, shows how ufd rotation might allow Ubc9 to be reoriented toward Uba2's catalytic cysteine [32]. However, even upon rotation of the Uba2 ufd to the position observed in the NEDD8 E1, a significant E1-to-E2 cysteine-tocysteine gap remains ( Figure 4C). It is possible that Uba2's ufd undergoes an even greater rotation during Sumo transfer. Another excellent possibility for how the gap between the Uba2 and Ubc9 catalytic cysteines could be closed comes from the recent landmark finding that the Uba2 catalytic cysteine domain can undergo dramatic reorientation and remodeling [20]. It seems likely that the proposed combination of ufd and catalytic cysteine domain rotations bring the Uba2 and Ubc9 cysteines together.
Sumo also binds noncovalently to Ubc9, via the ''backside'' distal from the E2 catalytic cysteine [47,48,49]. This interaction is conserved from yeast to humans [47,48,49]. Although the function of this noncovalent Ubc9-Sumo interaction is not well understood, mutant versions of Ubc9 where this interaction is disrupted cannot support yeast viability [47], and are impaired at generating long poly-Sumo chains [48,49]. Previous biochemical data showed that the noncovalent Sumo interaction competes with Ubc9's interaction with Uba2 [37,47]. Indeed, superposition of Ubc9 from our complex with Uba2 ufd with prior structures of complexes with Sumo reveal mutually exclusive binding to Ubc9's a-helix1, particularly Arg13 and Arg17 ( Figure 5C). Thus, the Uba2 ufd recognizes multipurpose binding sites on Ubc9, which may serve to order association with many different partners in the Sumo cascade.
In a related vein, human Ubc9 is distinct from yeast Ubc9 in becoming modified on Lys14 via an isopeptide bond to human Sumo, forming a covalent complex referred to as Ubc9*Sumo [50]. Ubc9*Sumo can bind E1 and form a thioester-linked complex to another Sumo at its active site, which subsequently can be transferred to targets. Sumoylation of Ubc9 shifts its E3-independent target specificity [50]. For the substrate Sp100, it seems Sp100's Sumo-interacting motif (SIM) is recruited to the Sumo linked to Ubc9's Lys14. Comparison of the crystal structure of human Ubc9*Sumo with that of yeast Uba2 ufd -Ubc9 reveals that Uba2 overlaps the position of Sumo in the covalent complex ( Figure 5D). How then might human Ubc9*Sumo bind Uba2 to become charged with Sumo? Given the high structural similarity between human and yeast Ubc9 [40] and Uba2 ufd (Figure 3), and prior NMR and mutational data consistent with conservation of our structurally-observed Uba2 ufd -Ubc9 interactions in their human counterparts [19,36,51], it seems likely that human Uba2 and Ubc9 interact as in our structure. A plausible model would be that there is some flexibility between Ubc9 and Sumo in Ubc9*Sumo such that Sumo's b-grasp domain moves out of the way when Uba2 binds. This is consistent with ,4-fold more human Aos1-Uba2 being required to fully bind Ubc9*Sumo in comparison to Ubc9 in a qualitative interaction study [50]. Additional interactions might stabilize Uba2-Ubc9 complex formation during the transthiolation reaction. Indeed, although quantitative enzyme kinetic comparison was not performed, under some conditions Ubc9 and Ubc9*Sumo show equal formation of thioester-linked intermediates with Sumo [50]. Future studies will be required to understand the dynamics of interactions between Sumo, Ubc9, Uba2, and E3s in transfer cascades.
The complex structure also provides only the second detailed view of E1 ufd -E2 interactions for any UBL, and demonstrates a common overall mode of E2 recruitment by E1 ufds: a ufd's ''W'' (i.e., double-''V'') -shaped surface engages an E2's N-terminal helix and b1b2-loop. Specificity within the Sumo pathway is established by many unique favorable interactions between Uba2 and Ubc9. Furthermore, Ubc9 is distinct among E2s as having an extended rigid b1b2-loop. This is recognized by a conserved Cterminal ''V'' from Uba2 ufd , comprised of Uba2 elements spanning from b-strand23 through a-helix37. The Uba2 ufd Cterminal ''V'' is minimized relative to the corresponding regions in ubiquitin and NEDD8 E1s [27,28], which would be unable to accommodate the Ubc9 b1b2-loop. Structural comparison to Ubc9-E3 structures/models, and to Ubc9-Sumo, demonstrates how E1-E2 interactions are mutually exclusive with forming E2-E3 or noncovalent E2-UBL complexes in Sumo cascades.
Finally, docking the Uba2 ufd -Ubc9 structure on prior structures of full-length Uba2 indicate that the Uba2 ufd would need to undergo significant rotation so that the Uba2 and Ubc9 cysteines could face each other. This is also a common feature of previous structures from the NEDD8 cascade [27,31,32,52]. Furthermore, based on conformational changes observed for Uba2 at earlier steps in the Sumo activation cycle, it seems that the Uba2 cysteine domain may also undergo conformational changes to meet the Ubc9 cysteine for Sumo transfer [20]. Along these lines, it is noteworthy that unique features to Uba2s and Ubc9s also establish a distinct orientation of the E2 catalytic cysteine relative to the E1's E2-binding ufd ( Figure 3E, Figure 3F). It will be interesting in the future to understand how UBL-specific E1 conformational changes accommodate such distinct geometries during E1-to-E2 UBL transfer.

Cloning, protein expression and purification
Expression constructs were made by standard PCR/ligation procedures, with sequences verified by automated sequencing procedures. Yeast Sumo, Smt3, is referred to as Sumo in the text. For biochemical studies, Smt3/Sumo residues 13-97 were expressed from pGEX-2TK (GE), harboring an N-terminal site for phosphorylation by Protein Kinase A (PKA), and was purified as previously described for other UBLs [33].
Uba2 ufd (residues 439-563) was expressed from pGEX-4T1 (GE). BL21(DE3) harboring the expression construct were grown in LB with ampicillin (100 mg/mL) at 37uC until reaching an OD 600 of ,1.0. Expression was induced by the addition of IPTG to a final concentration of 0.6 mM, followed by overnight culturing at room temperature. Cells were resuspended in 50 mM Tris-HCl, 0.2 M NaCl, 5 mM DTT, pH 8.0 supplemented with 2.5 mM with PMSF, and lysed by sonication on ice. GST-Uba2 ufd was purified by glutathione-affinity chromatography, treated with a 1:100 ratio of thrombin during overnight dialysis into 25 mM Tris-HCl, 0.2 M NaCl, 5 mM DTT, pH 8.0 at 4uC overnight. GST and any remaining GST-Uba2 ufd fusion protein were removed by glutathione-affinity. For co-crystallization, Uba2 ufd was purified by gel filtration chromatography using a SD75 column (GE) in 25 mM Tris-HCl, 0.15 M NaCl, 5 mM DTT, pH 7.6, concentrated to 22 mg/ml (Bio-Rad Protein Assay), aliquotted, flash-frozen in liquid nitrogen, and stored at 280uC until further use. For crystallization on its own, Uba2 ufd prepared similarly, except that gel filtration was performed in 50 mM Tris-HCl, 0.2 M NaCl, 5 mM DTT, pH 7.6, and Uba2 ufd was concentrated to 40 mg/ml. Ubc9 was expressed from pGEX4T3 (GE) [37] in BL21 (DE3) Gold cells. After growth at 37uC until the OD 660 reached ,0.8, expression was induced by the addition of 1 mL of 0.6 M Figure 5. Implications of Uba2 ufd -Ubc9 structure for Sumo cascades. (A) Comparison of E1-E2 and E2-E3 interactions in Sumo cascades. Structure of S. cerevisiae Uba2 ufd -Ubc9 (magenta, cyan), with Ubc9 superimposed on the previous structure of human Ubc9 (cyan) bound to RanBP2/ Nup358 (olive) from the complex with RanGAP-1 (not shown) and Sumo-1 (not shown) [24]. Uba2 ufd and a portion of RanBP2/Nup358 are shown with a semi-transparent surface to highlight the overlapping regions. (B) Structure of S. cerevisiae Uba2 ufd -Ubc9 (magenta, cyan), with Ubc9 superimposed on a model for S. cerevisiae Ubc9-Siz1 (olive), based on previous structures of Siz1 and c-Cbl-UbcH7 [25,46]. Briefly, the RING domain c-Cbl and UbcH7 were docked onto the SP-RING domain of Siz1, and then Ubc9 was modeled in place of UbcH7. Uba2 ufd and Siz1 are shown with a semi-transparent surface to highlight the overlapping regions. (C) Comparison of E1-E2 and noncovalent E2-UBL interactions in Sumo cascades. Structure of S. cerevisiae Uba2 ufd -Ubc9 (magenta, cyan) superimposed in previous structure of noncovalent Ubc9-Sumo (cyan, yellow) complex from S. cerevisiae [47]. Uba2 ufd and Sumo are shown with a semi-transparent surface to highlight the overlapping regions. (D) Structure of S. cerevisiae Uba2 ufd -Ubc9 (magenta, cyan), with Ubc9 superimposed on human Ubc9*Sumo (cyan*yellow) [50]. Uba2 ufd and Sumo are shown with a semi-transparent surface to highlight the overlapping regions. doi:10.1371/journal.pone.0015805.g005 IPTG/liter of culture, and induction was carried out overnight at 16uC. GST-Ubc9 was purified by glutathione affinity. For biochemical studies, GST-Ubc9 bound to glutathione sepharose was treated with thrombin overnight at 4uC to release wild-type and mutant versions of Ubc9. Wild-type and mutant versions of Ubc9 were concentrated in a final buffer of 50 mM Tris-HCl pH 7.6, 200 mM NaCl, 1 mM DTT, aliquotted, flash-frozen, and stored until use. For crystallography, GST-Ubc9 was treated with a 1:30 ratio of thrombin overnight at 16uC. Ubc9 was purified to homogeneity by cation exchange with a homemade column packed with ResourceS resin (GE) using a NaCl gradient in 20 mM HEPES, 1 mM DTT, pH 7.0, concentrated to 27 mg/ml, aliquotted, flash-frozen in liquid nitrogen, and stored at 280uC until further use.
For biochemical assays, the expression construct for Sumo E1 (the heterodimeric Aos1-Uba2 complex) was described previously [37]. BL21(DE3) Gold cells harboring the expression construct were grown in LB with ampicillin (100 mg/mL) at 37uC until reaching an OD 600 of ,0.8. After cooling to 16uC for 45 minutes, expression was induced by the addition of 1 mL of 0.2 M IPTG/ liter of culture, shaking overnight at 16uC. GST-Aos1-Uba2 was purified by glutathione affinity chromatography, and eluted protein was cleaved overnight with thrombin at 4uC. Aos1-Uba2 was further purified by anion exchange chromatography, concentrated in a final buffer of 50 mM Tris-HCl pH 7.6, ,290 mM NaCl, 1 mM DTT, aliquotted, flash-frozen in liquid nitrogen, and stored at 280uC until further use.
Crystallization, data collection, and structure determination and refinement Initial crystallization screening was performed with a Mosquito crystallization robot (TTP Labtech) with commercial 96-well screens. After testing several conditions (chemical and temperature) and Uba2 ufd -Ubc9 complexes (from human and yeast with different ufd domain boundaries), crystals were obtained at 4uC in the Index HTS condition G2 (Hampton Research), for a 1.1:1 volume:volume mixture of Uba2 ufd at 22 mg/ml and Ubc9 at 27 mg/ml. A diffraction-quality crystal was obtained after multiple cycles of streak-seeding into 20% (w/v) PEG3350, 0.1 M Bis-Tris pH 5.5, 0.2 M Li 2 SO 4 , and harvested and flash-frozen in mother liquor supplemented with 20% (v/v) glycerol. The crystals belong to space-group P2 1 . Data to 2.3 Å resolution were obtained at the SERCAT-ID beamline, and processed with HKL2000 [53]. Initial models for the complex were obtained by performing molecular replacement in the P1 space group using PHASER [54], after searching for 4 copies of the structure of yeast Ubc9 (2GJD.pdb, Chain A) [40] and of a polyAla model of human Uba2 residues 449-546 from 1Y8Q.pdb, Chain C [19]. One of the 4 complexes from this solution was subsequently used as a search model to find two Uba2 ufd -Ubc9 complexes per asymmetric unit in P2 1 . The model was built using Coot [55] and refined using TLS parameters in REFMAC [56]. Due to low sequence similarity with the human Uba2 ufd , we wished to obtain additional data to confirm the structure in regions where side-chains were poorly visible. We thus obtained crystals for the isolated Uba2 ufd at 4uC in the Ammonium Sulfate HTS condition H8 (Qiagen). After optimization in manual trays, a crystal was obtained in 3 M (NH 4 ) 2 SO 4 , 1% (v/v) MPD, and harvested and flash-frozen in 3 M AmSO4, 1% (v/v) MPD, 25% (v/v) glycerol. Data to 1.6 Å resolution were obtained at the SERCAT-ID beamline, and processed with HKL2000 [53]. Examination of the data with Phenix.xtriage indicated the space group to be I 4 , with nearly perfect merohedral twinning and a twin operator of -k,-h,-l. The structure was obtained by molecular replacement using PHASER [54], with the Uba2 ufd from the partially-refined Uba2 ufd -Ubc9 complex structure as a search model for one molecule in the asymmetric unit. After rebuilding using Coot [55] and refinement including twin laws using Phenix [57], the isolated Uba2 ufd structure was then used as a guide for final refinement of the Uba2 ufd -Ubc9 structure. Data collection and refinement statistics are provided in Table 1. The single Ramachandran outlier in the high resolution Uba2 ufd structure corresponds to Gly517, with the conformation supported by the electron density.
Assays for formation of a Ubc9,Sumo thioester conjugate Sumo was labeled with 32 P at an N-terminal PKA site using PKA (New England Biolabs) and [c-32 P]ATP as described previously for other UBLs [33]. Reactions were performed at 18uC, which is the ambient room temperature for our dedicated room for use of radioactivity, in 10 ml reaction volumes containing 50 mM Tris-HCl (pH 7.6), 5 mM ATP, 10 mM MgCl 2 , 2 mg/ml BSA, 5 mM creatine phosphate, 0.3 U/ml creatine kinase, 0.3 U/ ml inorganic pyrophosphatase, 5 nM Aos1-Uba2, 250 nM Ubc9 or the indicated mutant, and 10 mM [c-32 P]-Sumo. Reactions were quenched at 1, 3, 10 and 30 minutes by the addition of an equal volume of 2X SDS sample buffer, resolved on 15% SDS-PAGE gels, dried, and exposed to film. In the assay containing both Aos1-Uba2 (Cys177Ala) and Aos1-Uba2 (Asp488Ala and Asp490Ala), 5 nM of both mutant E1s was used.