DNA Repair and Global Sumoylation Are Regulated by Distinct Ubc9 Noncovalent Complexes

Abstract

Global sumoylation, SUMO chain formation, and genome stabilization are all outputs generated by a limited repertoire of enzymes. Mechanisms driving selectivity for each of these processes are largely uncharacterized. Here, through crystallographic analyses we show that the SUMO E2 Ubc9 forms a noncovalent complex with a SUMO-like domain of Rad60 (SLD2). Ubc9:SLD2 and Ubc9:SUMO noncovalent complexes are structurally analogous, suggesting that differential recruitment of Ubc9 by SUMO or Rad60 provides a novel means for such selectivity. Indeed, deconvoluting Ubc9 function by disrupting either the Ubc9:SLD2 or Ubc9:SUMO noncovalent complex reveals distinct roles in facilitating sumoylation. Ubc9:SLD2 acts in the Nse2 SUMO E3 ligase-dependent pathway for DNA repair, whereas Ubc9:SUMO instead promotes global sumoylation and chain formation, via the Pli1 E3 SUMO ligase. Moreover, this Pli1-dependent SUMO chain formation causes the genome instability phenotypes of SUMO-targeted ubiquitin ligase (STUbL) mutants. Overall, we determine that, unexpectedly, Ubc9 noncovalent partner choice dictates the role of sumoylation in distinct cellular pathways.

INTRODUCTION

Conjugation of the small ubiquitin-like modifier (SUMO) to target proteins regulates many diverse processes related to genome stability and cellular growth (18–20, 31, 32, 37). SUMO is covalently attached to target proteins by a cascade that includes an E1 activating enzyme complex, a single E2 conjugating enzyme, and a limited number of E3 ligases (20). The ubiquitin modification system has a similar enzymatic cascade, but in stark contrast to the SUMO pathway, it has multiple E2s and numerous E3 ligases that provide a clear basis for selectivity (20). For example, in the fission yeast Schizosaccharomyces pombe the SUMO pathway includes a single E2 called Ubc9 (Hus5) and two known SUMO E3 ligases, Pli1 and Nse2 (51). Although these two E3 ligases are responsible for sumoylating largely distinct targets, how substrate specificity is generated is poorly characterized.

Division of labor between Pli1 and Nse2 is underscored by the disparate phenotypes of cells lacking either ligase. Cells lacking Pli1 exhibit greatly reduced levels of global SUMO conjugates, heterochromatin silencing defects, and altered telomere length but are insensitive to genotoxins (38, 51, 56). Conversely, Nse2 SUMO E3 ligase-deficient cells lack the major Pli1 mutant phenotypes and are hypersensitive to genotoxic stress (51). Nse2 (Mms21 of the budding yeast Saccharomyces cerevisiae) is part of the essential Smc5/6 complex that plays critical roles in DNA repair and suppressing aberrant recombination (3, 6, 11, 12, 14, 36). A phenotypic consequence of Smc5/6, Nse2/Mms21, or Ubc9 dysfunction is the accumulation of unresolved toxic recombination-dependent structures at damaged replication forks (6, 12).

Interestingly, an additional factor called Rad60 (budding yeast Esc2) that physically interacts with the Smc5/6 complex was found to coact in this suppression of aberrant recombination (5, 12, 27, 28). Rad60 defines an intriguing family of proteins conserved from yeast to humans in that they contain two SUMO-like domains (SLDs) at their C termini (5, 30). Notably, the Rad60 C-terminal SLD interacts directly with Ubc9 (39), raising the possibility that Rad60 could selectively facilitate Nse2-dependent sumoylation via recruitment of Ubc9.

Similar to ubiquitin, SUMO can form chains via covalent modification of internal lysine residues (20, 49). Structure-based in vitro studies have suggested a role for a noncovalent complex formed by Ubc9 and SUMO (Ubc9:SUMO) in SUMO polymerization; however, in vivo evidence for such a function is lacking (9, 15, 22, 48). It is also unclear whether the observed binding of Rad60 SLD2 to Ubc9 (39) antagonizes SUMO chain formation by disrupting the Ubc9:SUMO complex (42). In addition, whether the major SUMO E3 ligase Pli1 and its homologs Siz1/Siz2 of budding yeast and PIAS (protein inhibitor of activated STAT)-type ligases of higher eukaryotes (51) play a key role in forming SUMO chains remains unanswered.

The in vivo function or significance of SUMO chains is not clear. However, the discovery of a family of SUMO-targeted ubiquitin ligases (STUbLs) whose activity is stimulated by, or directed toward, SUMO chains (17, 37) has shed some light on potential roles of chains. For example, the tumor suppressor promyelocytic leukemia protein (PML) and its oncogenic fusion protein PML-retinoic acid receptor α (RARA) are degraded in a STUbL-dependent manner following treatment of cells with the therapeutic agent arsenic trioxide (23, 47, 52). This STUbL-dependent effect is particularly important in the treatment of acute promyelocytic leukemia, wherein degradation of PML-RARA releases the differentiation blockade and thereby induces remission (see references 17 and 37). STUbLs likely have other SUMO chain-related functions, as fission yeast lack a PML homolog but contain a STUbL. Dysfunction of fission yeast STUbL causes genome instability and genotoxin sensitivity due to disruption of SUMO pathway homeostasis through as yet undetermined mechanisms (37–39).

Here, we define specific mechanisms that drive the observed outputs of the SUMO pathway that include global sumoylation, SUMO chain formation, and genome stabilization. We determined the cocrystal structure of the noncovalent Ubc9:Rad60 SLD2 complex, which shows that SLD2 occupies the same noncovalent interface on Ubc9 as SUMO. Furthermore, our structure of SLD2 from the human Rad60 homologue, NFATC2IP (nuclear factor of activated T cells, cytoplasmic-2 interacting protein [NIP45]) indicates that this interaction is evolutionarily conserved in NFATC2IP SLD2. Strikingly, mutations designed to specifically abrogate the Ubc9:SLD2 complex or the Ubc9:SUMO interface reveal distinct roles for these two Ubc9 complexes. The Ubc9:SLD2 complex promotes Nse2-dependent sumoylation processes, whereas Ubc9:SUMO facilitates Pli1-dependent sumoylation and chain formation. Consistently, a Ubc9 mutation abolishing the noncovalent interactions with SUMO or Rad60 phenocopies cells lacking both E3 ligases. Overall, our structure-based in vivo analysis coupled with biochemical and genetic interrogations have provided key new insights on the impact of Ubc9 noncovalent partner choice in regulating distinct sumoylation processes.

MATERIALS AND METHODS

Crystallizations, data collection, and analyses.

Crystallization experiments on human NFATC2IP SLD2 (residues 345 to 419) buffered in 50 mM Tris, pH 7.5, 300 mM NaCl, 0.5 mM EDTA, and 1 mM β-mercaptoethanol produced diffraction-quality crystals with 1.7 M ammonium sulfate and 100 mM bicine, pH 9.0, grown at 20°C. Initial crystallization conditions for Ubc9:SLD2 complex buffered in 20 mM Tris, pH 8.0, 150 mM NaCl, 0.5 mM EDTA, and 1 mM β-mercaptoethanol were discovered with the Hauptman-Woodward high-throughput robotic crystallization screen (25). Initial conditions were optimized by hand to 20% polyethylene glycol (PEG) 8K, 200 mM imidazole malate buffer, pH 5.8, 600 mM LiNO₃, and 1 mM dithiothreitol (DTT) precipitant, using vapor diffusion crystallization experiments that produced diffraction-quality crystals at 20°C. For NFATC2IP SLD2, flash-frozen single crystals in mother liquor were used for X-ray diffraction data collection at 100 K; Ubc9:SLD2 crystals were supplemented with 20% (wt/vol) ethylene glycol cryoprotectant in the mother liquor before being frozen, and X-ray diffraction data were collected at 100 K. The NFATC2IP SLD2 data set was collected from a single crystal that diffracted to 1.6 Å, while a 1.9-Å Ubc9:SLD2 data set was collected on a single crystal. Both datasets were collected on an ADSC Q315 detector at the SIBYLS beamline 12.3.1 (Lawrence Berkeley National Laboratory, California). The NFATC2IP SLD2 data set and the Ubc9:SLD2 data set were both indexed, merged, and scaled using the HKL2000 package (35). NFATC2IP SLD2 crystals belonged to the P12₁1 space group, having cell dimensions of 32.5 Å, 32.4 Å, and 32.9 Å and 90°, 96.7°, and 90°; Ubc9:SLD2 crystals belonged to the P6 space group, having cell dimensions of 115.2 Å, 115.2 Å, and 34.8 Å and 90.0°, 90.0°, and 120.0°. Successful crystallographic phasing for NFATC2IP SLD2 and for Ubc9:SLD2 structures was completed by a molecular replacement (MR) search on a computer cluster containing 288 dual-processor servers with a total of 576 3.4-GHz Intel XEON-EMT central processing units (CPUs), using Phaser (26) of the CCP4 package. For NFATC2IP SLD2, SUMO-1 (Protein Data Bank [PDB] code 2UYZ) provided phasing, while for Ubc9:SLD2 the previously determined SLD2 and the human Ubc9 structures were used as the MR models. In NFATC2IP SLD2 one molecule is present in the asymmetric unit of the crystal, and an initial NFATC2IP SLD2 model was built into σA-weighted 2F_o−F_c and F_o−F_c (where F_o and F_c are the observed and calculated structure factor amplitudes, respectively) maps using Coot (16) and was followed by cycles of refinement using PHENIX (1), with 5% of the reflections not included in the refinement to monitor R_free. Validation of the structural model within PHENIX included MolProbity (13) to judge quality of the final models and a Ramachandran plot determining that 100% of residues are in the allowed region. For structural statistics, see Table 1. In the Ubc9:SLD2 complex one molecule is also present in the asymmetric unit of the crystal, and an initial Ubc9:SLD2 model was built into σA-weighted 2F_o−F_c and F_o−F_c maps using Coot and was followed by cycles of refinement using PHENIX, with 10% of the reflections not included so as to monitor R_free. A Ramachandran plot of the final 1.9-Å Ubc9:SLD2 complex determined that 97.8% of residues are in the allowed region, and 2.2% of residues are in the additionally allowed region. Structural statistics are given in Table 1. The Ubc9:SLD2 structural superimpositions of NFATC2IP, SLD2, and S. pombe Ubc9 were conducted using Sequoia (50).

Table 1.

X-ray diffraction data collection and refinement statistics^a

Parameter	Ubc9:Rad60 SLD2 complex	NFATC2IP SLD2
Data collection
Wavelength (Å)	0.98	1.13
Space group	P6	P1211
Unit cell parameters
a, b, c (Å)	115.2, 115.2, 34.8	32.5, 32.4, 32.9
α, β, γ (o)	90.0, 90.0, 120.0	90.0, 96.7, 90.0
Resolution (Å)	50–1.9	50–1.6
Rsym	4.9 (39.6)	4.3 (11.3)
Completeness (%)	99.1 (98.7)	97.1 (78.6)
I/σI	30 (3.4)	40.6 (7.4)
Redundancy	3.4 (3.1)	3.4 (2.3)
Refinement statistics
Resolution (Å)	33.3–1.9 (1.97–1.9)	32.7–1.6 (1.83–1.6)
No. of reflections	20,031 (1,763)	8,866 (2,781)
No. of atoms
Protein	2,059	766
Water	193	108
Avg B factor (Å)
Protein	30.4	17.4
Water	37.2	28.6
Rwork (%)	19.9	17.0
Rfree (%)	23.9	21.4
RMSD
Bond length (Å)	0.007	0.005
Bond angle (o)	1.2	1.0
Ramachandran plot
Allowed (%)	97.8	100
Additionally allowed (%)	2.2	0

^aData were collected on single crystals, and values in parentheses represent the high-resolution shell.

S. pombe strain construction, general yeast techniques, and drug treatments.

Standard yeast methods were performed as described previously (29). Strains used in this study are listed in Table 2. Drugs were obtained from Sigma-Aldrich. SUMO with a D81R mutation (SUMO^D81R) and with a K14R K30R double mutation (SUMO^K14/30R) was created using QuikChange (Stratagene). These constructs were then integrated via gene replacement into SUMO::ura4⁺. Transformants were counter-selected using 5-fluoroorotic acid, and stable clones were sequenced to confirm the presence of the SUMO mutations.

Table 2.

S. pombe strains used in this study

Strain no.	Descriptiona
NBY36	cds1::ura4+
NBY303	rad60-4::kanMx6
NBY780	h+
NBY1457	pmt3::ura4+
NBY1493	pli1::kanMx6
NBY1540	nse2-SA::kanMx6 h−
NBY1717	pREPKZ-Rad60-SLD2 (aa 332-406)::LEU2+
NBY1740	rhp51::ura4+ top1::kanMx6 pli1::hphMx6
NBY1848	rhp51::ura4+ top1::kanMx6
NBY1982	pREPKZ-Rad60-SLD2-E380R (aa 332-406)::LEU2+
NBY2027	rhp51::ura4+
NBY2047	rad60E380R::kanMx6 h+
NBY2217,	ubc9;K14E-TAP::kanMx6 h+
NBY2220	ubc9P21L-TAP::kanMx6 h+
NBY2225	ubc9F24S-TAP::kanMx6 h+
NBY2471	slx8I230T::hphMx6.
NBY2475	slx8I230T::hphMx6 pli1::kanMx6
NBY2927	ulp2::kanMx6
NBY3010	pmt3D81R
NBY3020	pmt3::ura4+ pREP41-EGFP-Pmt3::LEU2+
NBY3022	pmt3::ura4+ pREP41-EGFP-Pmt3D81R::LEU2+
NBY3025	pmt3::ura4+ pREP41-EGFP-Pmt3allR::LEU2+
NBY3048	pli1::kanMx6 pmt3D81R
NBY3049	slx8I230T::kanMx6 pmt3D81R
NBY3078	slx8I230T::kanMx6 pmt3::ura4+ pREP41-EGFP-Pmt3::LEU2+
NBY3079	slx8I230T::kanMx6 pmt3::ura4+ pREP41-EGFP-Pmt3D81R::LEU2+
NBY3102	slx8I230T::kanMx6 pmt3::ura4+ pREP41-EGFP-Pmt3allR::LEU2+
NBY3137	rad60E380R::kanMx6 pmt3::ura4+ pREP41-EGFP-Pmt3::LEU2+
NBY3138	rad60E380R::kanMx6 pmt3::ura4+ pREP41-EGFP-Pmt3D81R::LEU2+
NBY3140	nse2-SA::kanMx6 pmt3::ura4+ pREP41-EGFP-Pmt3::LEU2+
NBY3141	nse2-SA::kanMx6 pmt3::ura4+ pREP41-EGFP-Pmt3D81R::LEU2+
NBY3157	pli1::kanMx6 pmt3::ura4+ pREP41-EGFP-Pmt3D81R::LEU2+
NBY3203	ubc9S127T pREP41::LEU2+
NBY3204	ubc9S127T pREP41-Ubc9::LEU2+
NBY3205	ubc9S127T pREP41-Ubc9H20D::LEU2+
NBY3232	pmt3K14R/K30R
NBY3272	pmt3K14R/K30R slx8I230T::kanMx6
NBY3273	pmt3K14R/K30R nse2-SA::kanMx6
NBY3289	pmt3K14RK30R rad60E380R::kanMx6
NBY3315	pmt3K14R/K30R ulp2::kanMx6

^aUnless otherwise indicated, strains are ura4-D18 leu1-32. EGFP, enhanced green fluorescent protein.

In vitro pulldown experiments.

In vitro pulldowns have been described previously (39). Pulldowns were performed in 50 mM Tris, pH 8.0, 1 mM EDTA, 150 mM NaCl, 10% glycerol, and 0.2% Nonident P-40. His₆-SUMO^D81R and glutathione S-transferase-tagged Ubc9 with the mutation H20D (GST-Ubc9^H20D) were created using QuikChange. For the in vitro competition assay, a premixed suspension of His₆-SLD2 and His₆-SUMO were added to GST-Ubc9 prebound to glutathione (GSH)-Sepharose (GE Healthcare). A 5% input sample was taken at the time of mixing the recombinant proteins. A Li-Cor Odyssey scanner was used to quantify the Coomassie-stained proteins.

In vivo Western analysis of total levels of sumoylated proteins.

Western blot analysis of total levels of sumoylation in S. pombe has been previously described (38). Samples were resolved on 4 to 20% (wt/vol) Tris-glycine gradient gels (Invitrogen). The polyclonal SUMO antibody used in this study was raised against a rabbit as a His₆-SUMO fusion, which was expressed/purified from Escherichia coli BL21(DE3).

Bacterially based sumoylation assay of S. pombe core sumoylation proteins.

The S. pombe E1 (Rad31 and Fub2), E2 (Ubc9), and either a conjugatable or nonconjugatable SUMO (SUMO-gg or SUMO-Δgg, respectively, for SUMO with or without the GG residues required for conjugation, respectively) were cloned into a single pRSFDuet-1 vector (Novagen) essentially as described previously (34). Conjugatable SUMO^D81R-gg was created using QuikChange. For the bacterial sumoylation assay, the plasmid described above was transformed into BL21(DE3). Kanamycin-resistant colonies were then used to inoculate 1 ml of ZYM-5052 autoinduction medium (34, 45). The next day, a bacteria culture equivalent to an optical density at 600 nm (OD₆₀₀) of 1.0 was harvested and lysed in 50 μl of sample loading buffer supplemented with 10% β-mercaptoethanol. Five microliters of the lysates was resolved on 4 to 20% (wt/vol) Tris-glycine gradient gels and immunoblotted with antiserum to SUMO (1:10,000) with Li-Cor DyLight 800-conjugated anti-rabbit IgG (1:20,000). The Western blot was scanned using a Li-Cor Odyssey scanner.

Mass spectrometry identification of lysine residues for SUMO chain formation.

SUMO attachment sites were identified as previously described (53, 54). Briefly, purified SUMO conjugates were reduced, alkylated, and digested by the sequential addition of Lys-C and trypsin proteases. Digested peptide mixtures were then fractionated online using ion exchange and reverse-phase chromatography and eluted directly into a linear trap quadrupole (LTQ)-Orbitrap, where tandem mass spectra were acquired. Collected spectra were analyzed using the SEQUEST and DTASelect algorithms. Modified peptides were identified using differential modification searches that considered a +227.1270-Da mass shift on lysines corresponding to the LGG peptide that remains covalently linked to SUMO-modified lysines after tryptic digestion. False-positive rates were estimated using a decoy database approach, and modified spectra with a false-positive rate of less than 5% were considered potential SUMO-modified peptides.

Protein structure accession numbers.

The coordinates for NFATC2IP SLD2 and Ubc9:SLD2 complex have been deposited in the Protein Data Bank under accession codes 3RD2 and 3RCZ, respectively.

RESULTS

Crystal structure of the Ubc9:SLD2 complex.

To experimentally define the basis for the Ubc9 interaction with Rad60, we determined the crystal structure of the S. pombe Ubc9:Rad60 SLD2 complex to 1.9 Å (Fig. 1A) . This structure was refined (Table 1) to 19.9% R_work and 23.9% R_free, providing accurate and detailed information on the interaction and interface. High-resolution electron density maps allowed definition of most Rad60 and Ubc9 residues, except the SLD2 C-terminal residue and the first two and last two residues of Ubc9, all of which are likely disordered and away from the interaction interface. Interestingly, little conformational change is induced in SLD2 upon binding to Ubc9, with structural superimposition of complex and apo forms of this protein domain having a root mean square deviation (RMSD) of 0.63 Å.

Fig. 1.

Ubc9:SLD2 complex and NFAT2CIP SLD2 structure. (A) The 1.9-Å resolution crystal structure of Ubc9:SLD2 complex reveals that SLD2 (green) binds to the α1-β1 noncovalent interaction site of Ubc9 (cyan), instead of the active site centered on reside Cys93. (B) SLD2 FEG motif Glu380, 3₁₀-helix-β5 residues Glu396, Asp399, and Gln400, and the Arg339 side chain form hydrogen bond interactions (dashed black lines) with Ubc9 residues. (C) The upper panels shows the Rad60 SLD2 sequence, containing the Ubc9-interacting FEG motif, aligned to budding yeast Smt3p, human SUMO-1, and human NFATC2IP SLD2. Secondary structure of Rad60 SLD2 is depicted above the alignment. Red highlights amino acid side chains that form hydrogen bonds or salt bridges to Ubc9, green highlights residues having significant hydrophobic interactions in the Ubc9 interface, and NFATC2IP residues marked in blue show conserved residues likely having direct interactions with Ubc9. The lower panel shows fission yeast, budding yeast, mouse, and human Ubc9 structure-based sequence alignment, with fission yeast Ubc9 secondary structure depicted above the alignment. Red highlights side chains forming hydrogen bonds or salt bridges to SUMO/SLD2 residues at the noncovalent interface, green highlights residues having hydrophobic interactions at this interface, and bold marks the Asp-His-Pro-Phe-Gly-Phe motif that forms a key component of Ubc9's interaction with SUMO/SLD2. H. sapiens, Homo sapiens; M. musculus, Mus musculus. (D) The 1.6-Å crystal structure of NFATC2IP SLD2 reveals a β-GRASP fold composed of five β-strands, an α-helix, and a 3₁₀-helix. The ellipse highlights the region in SUMO between β-strand 2 and α-helix 1 observed to bind SIM sequences. (E) Electrostatic surface diagram for NFATC2IP SLD2 in the same orientation as in panel D, with the ellipse marking the surface of the β-strand 2 α-helix 1 region that shows a net negative charge and lack of a clear binding cleft, distinct from the electropositive SIM-binding pocket in SUMO. (F) Structural superimposition of the Ubc9:SLD2 complex (cyan, Ubc9; green, SLD2) with NFATC2IP SLD2 (orange), showing close structural similarity, despite low sequence identities. (G) Close-up view of structural superimposition shown in panel F revealing a well-conserved, noncovalent Ubc9-binding face in NFATC2IP SLD2, with well-conserved residues depicted as sticks. These residues include Asp394 of β3–β4-loop FDG motif and the carboxylate residues in the 3₁₀-helix β-strand 5 region, including residues Glu410 and Asp413, in addition to Glu416 that is suitably orientated for interactions with Ubc9.

The Ubc9:SLD2 interaction occurs through direct hydrogen bonds and indirect water-mediated interactions, with six ordered waters in the interface. The Glu380 side chain of a Rad60 Phe-Glu-Gly motif hydrogen bonds to Ubc9 Arg13. The SLD2 3₁₀-helix-β-5 residues Glu396 and Asp399 form hydrogen bonds with Ubc9 Arg17. The Gln400 side chain of the SLD2 β-strand 1 hydrogen bonds to the Ubc9 His20 backbone amine and carbonyl groups. Adjacent to this is the SLD2 β-1 Arg339 that interacts with the His20 main chain carbonyl and the Asp19 side chain carboxylate of Ubc9 (Fig. 1B). The Ubc9 Phe22 side chain, the first Phe of a conserved Asp-His-Pro-Phe-Gly-Phe motif containing His20, interacts with a hydrophobic region created on one side by the SLD2 β-1/β-2 loop and aliphatic side chain atoms of Arg339 and Ser341 in the loop. The aliphatic side chain atoms of residues Ser402 and Val404 from SLD2 β-5 provide the other side of the pocket. The total buried surface area of this Ubc9:SLD2 interface is 597.8 Ų. Notably, comparison of the SLD2 complex structure to the human and budding yeast noncovalent Ubc9:SUMO complex structures (9, 15, 22) reveals that the buried surface areas are similar. In the human SUMO-1 and mouse Ubc9 complex (22), the buried surface area is 613.1 Ų; it is 560.4 Ų in the human SUMO-1 and human Ubc9 complex (9) and 529.0 Ų and 526.7 Ų in the two complexes present in crystallographic asymmetric unit of budding yeast Smt3p-Ubc9p (15). The SLD2 amino acid side chains directly interacting with Ubc9 are well conserved with those in SUMO that were observed to directly interact in the noncovalent Ubc9 interface (Fig. 1C), and SLD2 and SUMO bind to Ubc9 in similar orientations in these noncovalent complexes.

NFATC2IP SLD2 crystal structure.

Ubc9 shows considerable sequence and structural conservation between species. S. pombe Ubc9 shares 66% sequence identity and a 0.6-Å RMSD of the backbone α-carbon atoms to human Ubc9 (PDB 2PE6), 66% identity and a 0.48-Å RMSD to mouse Ubc9 (PDB 2UYZ), and 61% identity and a 0.84-Å RMSD to budding yeast Ubc9 (PDB 2EKE). This sequence conservation suggests strong evolutionary pressure for conservation of the surface features required for protein interaction, including binding of Rad60 family members.

Therefore, to define whether the Ubc9 interface is structurally conserved in the human NFATC2IP protein, we crystallized and solved the structure of the homologous human SLD2. Our 1.6-Å crystal structure of human NFATC2IP SLD2 (Fig. 1D), residues 345 to 419, refined to 17.0% R_work and 21.4% R_free (Table 1). Electron density was observed for all of the NFATC2IP SLD2 residues, including the N-terminal His₆ tag. Interestingly, the region at β-strand 2/α-helix 1, which has a net negative charge in the NFATC2IP SLD2 structure, resembles SLD2 (Fig. 1E). This is distinct from the positively charged cleft that is formed between β-strand 2 and α-helix 1 in SUMO-1, SUMO-2, and SUMO-3 paralogues and functions to noncovalently bind hydrophobic SUMO-interacting motifs (SIMs) of partner proteins. NFATC2IP SLD2 shares only 20% sequence identity with an S. pombe SLD2 homologue but superimposes to this structure with 1.79-Å RMSD of the backbone α-carbon atoms of our SLD2 structure in complex with Ubc9 (Fig. 1F). NFATC2IP SLD2 also has 28% identity and a 1.54-Å RMSD to human SUMO-1 (PDB 2PE6) and 35% identity and a 1.57-Å RMSD to human SUMO-2 (PDB 1WM3). The noncovalent Ubc9 interaction interface observed in our S. pombe Ubc9:SLD2 complex structure and in Ubc9:SUMO complexes from budding yeast and human proteins is conserved in the human NFATC2IP SLD2 structure. This includes residues Asp394 of an FDG motif in the β-3–β-4 loop and 3₁₀-helix–β-5 loop residues Glu410 and Asp413 (Fig. 1G).

Phenotypes caused by Ubc9 mutations at the noncovalent interface.

Based on our Ubc9:SLD2 structure, we predicted that a fission yeast Ubc9^H20D mutation would disrupt the noncovalent interface (Fig. 1). By analogy, an H20D mutation in human Ubc9 uncouples its in vitro interaction with SUMO-1, without significantly impacting Ubc9-SUMO thioester formation (22). We tested in vitro interaction between His₆-SLD2 and recombinant wild-type GST-Ubc9 or GST-Ubc9^H20D by GSH-Sepharose pulldown and Western analysis, revealing that GST-Ubc9^H20D abolishes the interaction (Fig. 2A). To test the effect of Ubc9^H20D mutation on in vivo Ubc9 function, we expressed either Ubc9^H20D or a wild-type control in cells hypomorphic for Ubc9 (ubc9-3). Unlike wild-type Ubc9, Ubc9^H20D failed to suppress either temperature sensitivity or sensitivity to the replication-stalling agent hydroxyurea (HU) of the ubc9-3 mutant (Fig. 2B). To extend this analysis, we randomly generated a series of Ubc9 mutants and identified the mutations by sequencing. Mutations that compromise Ubc9 function mapped to anticipated regions encompassing the hydrophobic and catalytic cores (data not shown) and K14E at the E1 binding site, which caused HU sensitivity (Fig. 2C) (24). In addition, we identified two mutations, P21L and F24S, at the noncovalent Ubc9:SLD2 interface that also result in HU sensitivity (Fig. 2C). P21 and F24 side chains stack together in order to project the side chain of F22 outwards from the surface of the protein. The extended F22 side chain interacts with a shallow hydrophobic pocket on the surface of Rad60 SLD2 that is formed in part by hydrophobic regions of the Rad60 Arg339 and Rad60 Ser341 side chains (Fig. 1B). The HU sensitivities of the Ubc9^P21L and Ubc9^F24S mutants might therefore reflect a weakened interface between Rad60 and Ubc9. It should be noted that F24 is also part of the Ubc9 hydrophobic core, which likely accounts for the temperature sensitivity of the Ubc9^F24S mutant.

Fig. 2.

Phenotypes caused by Ubc9 mutations at the noncovalent interface. (A) Western analysis following in vitro GSH-Sepharose pulldowns of either bacterially expressed GST-Ubc9 or GST-Ubc9^H20D incubated with His₆-SLD2. Immunoblots are shown using antiserum for hexahistidine (His₆) or GST. (B) Serial dilutions of the ubc9-3 mutant, plated on medium lacking thiamine to induce protein expression from the indicated plasmids. Cells were either untreated (No Drug) at the indicated temperatures or treated with HU at 32°C. pVector denotes an empty vector. (C) Serial dilutions of the indicated strains, which were plated on nonselective medium and either untreated (No Drug) at the indicated temperatures or treated with HU at 32°C. α, anti.

In vivo effect of abrogating the noncovalent Ubc9:SUMO complex.

Based on structural homology, Ubc9^H20D is also likely to disrupt the fission yeast Ubc9:SUMO complex in addition to Ubc9:SLD2. We confirmed this hypothesis through in vitro binding assays of recombinant SUMO and Ubc9/Ubc9^H20D (Fig. 3A). Therefore, to test the specific contribution of noncovalent SUMO-binding to Ubc9 function, we made a SUMO^D81R mutant that disrupts only the Ubc9:SUMO complex. We previously determined that Rad60^E380R was unable to bind Ubc9 (39), and our cocrystal structure shows that E380 forms a hydrogen bond to R13 of Ubc9 (Fig. 1B). The SUMO^D81R mutation is analogous to Rad60^E380R as it does not detectably interact with GST-Ubc9 in vitro (Fig. 3B). We also assayed SUMO^D81R functionality in a bacteria-based sumoylation system. In this system SUMO, the E1 heterodimer Rad31/Fub2, and E2 Ubc9 proteins are coexpressed at high levels, and SUMO conjugation activity can be assessed by Western analysis of the total bacterial lysates. The positive control of wild-type SUMO forms a ladder of conjugates that correspond to the molecular weight of SUMO species ranging from di-SUMO to high-order poly-SUMO conjugates (Fig. 3C), which are absent when a nonconjugatable form of SUMO is expressed (Fig. 3C, SUMO-Δgg). SUMO^D81R is conjugation proficient as it produced a sumoylation pattern indistinguishable from that of the wild type (Fig. 3C).

Fig. 3.

In vivo effect of abrogating the noncovalent Ubc9:SUMO complex. (A) Western analysis following in vitro GSH-Sepharose pulldowns of either bacterially expressed GST-Ubc9 or GST-Ubc9^H20D incubated with His₆-SUMO. Immunoblots are shown using antiserum for His₆ or GST. (B) Western analysis following in vitro GSH-Sepharose pulldowns of bacterially expressed GST-Ubc9 incubated with either His₆-SUMO or His₆-SUMO^D81R. Immunoblots are shown using antiserum for His₆ or GST. (C) Western analysis of a bacteria-based sumoylation assay. Total lysates of BL21(DE3) strains expressing S. pombe E1, E2, and either conjugatable SUMO (SUMO-gg), nonconjugatable SUMO (SUMO-Δgg), or conjugatable SUMO^D81R (SUMO^D81R-gg) are shown. Immunoblotting was performed using antiserum for SUMO. Coomassie staining is shown as a loading control. (D) Serial dilutions of the indicated strains plated on medium and either untreated (No Drug), treated with the indicated concentrations of genotoxins, or UV irradiated. (E) Western analysis of total lysates of the indicated strains, immunoblotted using antiserum for SUMO or tubulin (loading control). (F) Western analysis of total lysates of the indicated strains immunoblotted using antisera for SUMO, Cdc2, and GST (loading controls). Strains containing pRad60-SLD2 and pRad60-SLD2^E380R were cultured in medium lacking thiamine to induce overexpression (OP) of these ectopic vectors. (G) In vitro competition assay of SLD2 and SUMO binding to Ubc9. Proteins were detected following SDS-PAGE separation by Coomassie staining and then quantified. Numbers shown below gels indicate the ratio of bound SLD2 to SUMO (SLD2/SUMO). (H) Western analysis of lysates from the indicated strains, immunoblotted with antisera to SUMO and tubulin (loading control). Cells were cultured either with or without hydroxyurea (15 mM) treatment for 4 h.

To analyze the in vivo function of Ubc9:SUMO, we generated a fission yeast strain in which the genomic copy of SUMO was replaced with SUMO^D81R. Strikingly, compared to SUMOΔ, SUMO^D81R cells exhibited nearly wild-type growth rates and sensitivity to genotoxins (Fig. 3D). Therefore, in contrast to Ubc9^H20D, specific abrogation of the Ubc9:SUMO complex does not impact the roles of sumoylation in the DNA repair response. We next analyzed total SUMO conjugates in SUMO^D81R cells by Western blotting for comparison with those in reference strains. Notably, in both SUMO^D81R and pli1Δ cells, the wild-type pattern of SUMO conjugates was undetectable (Fig. 3E). In contrast, bulk SUMO conjugates were readily detected in cells lacking Nse2 SUMO ligase activity (nse2-SA) or the Ubc9:SLD2 complex (Fig. 3E, rad60^E380R). We also observed that overexpression of wild-type SLD2 but not the SLD2^E380R mutant, likely through competitive inhibition of the Ubc9:SUMO complex, strongly depleted endogenous SUMO conjugates (Fig. 3F). Thus, both Pli1 and the noncovalent Ubc9:SUMO complex are critical for bulk sumoylation but not for modifying key targets in the DNA repair response.

To further explore the potentially competitive nature of SLD2 and SUMO for Ubc9 binding, we performed an in vitro competition assay. Immobilized GST-Ubc9 was mixed with both SLD2 and SUMO and incubated for 1 h. Samples were taken every 15 min, washed, resolved by SDS-PAGE, and Coomassie stained. Quantification of Ubc9-bound SUMO and SLD2 showed that in this in vitro context, Rad60 SLD2 has apparently higher affinity for GST-Ubc9 (Fig. 3G). We previously showed that Rad60 undergoes nucleo-/cytoplasmic shuttling and is excluded from the nucleus by Cds1-dependent phosphorylation during replication arrest with HU (5). We therefore determined the effects of HU-induced Rad60 delocalization on global sumoylation. We observed that HU promotes the formation of high-molecular-weight SUMO conjugates (Fig. 3H). Importantly, however, similar SUMO species were induced by HU in both the rad60-4 and cds1Δ strains in which Rad60 is both constitutively nuclear and its Ubc9 interface is intact (Fig. 3H) (5, 40). The rad60^E380R mutant exhibits elevated SUMO conjugates before and after HU treatment, which, based on the foregoing results, is most likely the result of replicative stress in the sickly rad60^E380R strain. Therefore, the HU-induced SUMO species are largely the result of replicative stress and not reduced competition between Rad60 SLD2 and SUMO for Ubc9.

Ubc9 partner choice promotes distinct sumoylation pathways.

In nse2-SA but not pli1Δ cells, sumoylation defects cause sensitivity to genotoxins (2, 51, 56). Interestingly, rad60^E380R mutant cells are hypersensitive to the same DNA-damaging agents as nse2-SA cells (Fig. 4A). However, like pli1Δ cells, SUMO^D81R mutants are not appreciably sensitive to genotoxins (Fig. 4A). Together with their effects on global sumoylation (Fig. 3E), this raises the intriguing possibility that Ubc9:SLD2 facilitates Nse2-dependent sumoylation, whereas Ubc9:SUMO promotes Pli1-dependent sumoylation.

Fig. 4.

Ubc9 partner choice promotes distinct sumoylation pathways. (A) Serial dilutions of the indicated strains plated on medium and either untreated (No Drug) or treated with the indicated concentrations of drugs. (B) A representative tetrad dissection is shown from a cross between rad60^E380R and SUMO^D81R mutant cells. Key depicts the genotypes present, which are denoted by various shapes placed around each colony within the tetrads. (C) As described above for panel B but with nse2-SA crossed with SUMO^D81R mutant cells. (D) Serial dilutions of the indicated strains plated on medium lacking thiamine to induce expression from the indicated vectors. Plates were either drug free (No Drug) or treated with HU. wt, wild type.

Cells lacking both Pli1 and Nse2 are synthetically sick and exhibit phenotypes similar to those deleted for SUMO (55). Therefore, we tested for an analogous genetic interaction between rad60^E380R and SUMO^D81R. Whereas either single mutant grows well, the rad60^E380R SUMO^D81R double mutant is extremely synthetically sick (Fig. 4B), exhibiting a similar phenotype to the rad60^E380R pli1Δ mutant (39). In addition, an nse2-SA SUMO^D81R double mutant is synthetically sick and phenocopies an nse2-SA pli1Δ double mutant (55) (Fig. 4C). This is in stark contrast to pli1Δ SUMO^D81R double mutant cells that show normal growth, indicating an epistatic relationship between the pli1Δ and SUMO^D81R genes (data not shown). We also generated rad60^E380R and nse2-SA strains that ectopically express SUMO or SUMO^D81R. Again, in either background, SUMO^D81R caused poor growth and enhanced sensitivity to replicative stress (Fig. 4D). Overall, these genetic interactions are consistent with roles of the Ubc9:SLD2 and Ubc9:SUMO complexes in facilitating Nse2- or Pli1-dependent sumoylation, respectively.

STUbL mutant phenotypes suppressed by SUMO^D81R and Pli1 deletion.

To further support the specific role of Ubc9:SUMO in supporting Pli1-dependent sumoylation, we determined the genetic interaction of SUMO^D81R with an slx8-29 (slx8^I230T) STUbL background, which has a mutated E3 ubiquitin ligase RING domain. The temperature sensitivity of STUbL mutant cells can be suppressed by deleting Pli1, supporting a major role of STUbLs in maintaining SUMO pathway homeostasis (38). Strikingly, the temperature-sensitive growth of slx8-29 STUbL mutant cells is also suppressed in the slx8-29 SUMO^D81R double mutant to an extent similar to that seen in the pli1Δ slx8-29 strain background (Fig. 5A). Likewise, the characteristic accumulation of high-molecular-weight SUMO conjugates in slx8-29 cells is absent in the slx8-29 SUMO^D81R double mutant (38; also data not shown). These data provide compelling support for the role of Ubc9:SUMO in facilitating Pli1-dependent sumoylation, which is toxic in STUbL mutant cells.

Fig. 5.

SUMO^D81R suppresses STUbL mutant phenotypes due to a defect in SUMO chain formation. (A) Serial dilutions of the indicated strains plated on nonselective medium and grown at either permissive (25°C) or restrictive (35°C) temperature. (B) Western analysis of total lysates of the indicated strains immunoblotted with antiserum for SUMO or tubulin (loading control). Cells were cultured at the restrictive temperature (35°C) in medium lacking thiamine to induce expression from the indicated vectors. (C) As described in panel B, but the indicated strains were cultured at 32°C. (D) Serial dilutions of the indicated strains plated on medium lacking thiamine to induce expression from the vectors and either untreated at a permissive (25°C) or restrictive (35°C) temperature or HU treated at 25°C. (E) Western analysis of total lysates of the indicated strains immunoblotted with antiserum for SUMO (*, loading control). (F and G) The indicated strains were serially diluted onto plates containing no drug and either grown at either the permissive or nonpermissive temperature or plated at 32°C with the indicated concentrations of HU.

Ubc9:SUMO complex promotes in vivo SUMO chain formation.

Suppression of STUbL mutant phenotypes by both pli1Δ and SUMO^D81R raises the question of whether suppression is due to a loss of bulk mono-SUMO conjugates or to an inability to form SUMO chains. When STUbL activity is compromised, wild-type SUMO but not a SUMO with all 9 lysine residues mutated to arginine (SUMO^allR) forms abundant SUMO chains (Fig. 5B). Interestingly, upon STUbL inactivation SUMO^D81R forms a major di-SUMO species but none of the higher-molecular-weight chains observed with wild-type SUMO (Fig. 5B). Notably, the di-SUMO species observed for SUMO^D81R in vivo are dependent on the Pli1 E3 SUMO ligase (Fig. 5C). Overall, these data reveal the absolute dependency of in vivo SUMO chain formation in STUbL mutants on both the Pli1 E3 ligase and the Ubc9:SUMO noncovalent complex.

To further address the role of SUMO chains in STUbL mutant phenotypes, we tested whether SUMO^allR, which does not form chains (Fig. 5B), also suppresses slx8-29 phenotypes. Expression of SUMO, SUMO^D81R, or SUMO^allR as the sole source of SUMO in wild-type cells has no major effect on their sensitivity to HU or elevated temperature (Fig. 5D). In contrast, expression of wild-type SUMO in slx8-29 cells caused the anticipated hypersensitivity to both HU and high temperature (Fig. 5D). Intriguingly, expression of either SUMO^D81R or SUMO^allR suppressed both the temperature and HU sensitivity of slx8-29 cells (Fig. 5D). In addition to using ectopically expressed SUMO^allR in these studies, we wished to identify key lysine residues involved in SUMO chain propagation and construct genomic replacements for more detailed genetic analyses. To this end, we generated a form of SUMO that contains the terminating sequence RLGG and expressed it in our bacterial sumoylation assay (Fig. 3C). Following purification of abundant SUMO chain species and mass spectrometry-based analysis of tryptic digests, we identified several lysine residues that were modified with the LGG tripeptide, including the amino terminal K14 and K30 residues (Table 3). These residues were also recently identified by comparison to budding yeast SUMO chain modification sites and site-directed mutagenesis (43). Fission yeast expressing SUMO^K14/30R from the genomic locus appeared to be wild type for growth but, like SUMO^D81R and pli1Δ cells, showed reduced levels of high-molecular-weight SUMO conjugates (Fig. 5E). Inactivation of slx8-29 still caused an increase in the abundance of SUMO^K14/30R conjugates, but, importantly, these did not attain the extremely high molecular weight species observed for wild-type SUMO in the slx8-29 background. This is consistent with other lysine residues in SUMO^K14/30R being utilized for inefficient chain formation (Table 3). Despite the residual increase in SUMO conjugates detected above, SUMO^K14/30R strongly suppressed both the drug and temperature sensitivity of slx8-29 STUbL mutant cells (Fig. 5F). This result contrasts with the apparent toxicity of preventing SUMO chain formation in budding yeast STUbL mutant cells (33). However, paralleling results in budding yeast (8), SUMO^K14/30R suppresses the temperature and HU sensitivity of cells lacking the SUMO deconjugating enzyme Ulp2 (Fig. 5G). Overall, these data demonstrate that it is mainly the defect in SUMO chain formation in pli1Δ and SUMO^D81R cells that mitigates STUbL mutant phenotypes.

Table 3.

Identification of SUMO-attachment sites by mass spectrometry^a

Peptide sequence	Residue modified	Charge state	Xcorr	ΔCN	ΔPPM	No. of spectra
ESPSANISDADK*SAITPTTGDTSQQDVKPSTEHINLK	K14	3	3.5917	0.5542	−4.4	1
SAITPTTGDTSQQDVK*PSTEHINLK	K30	2	4.1200	0.6256	−2.7	12
SAITPTTGDTSQQDVK*PSTEHINLK	K30	3	3.7596	0.5353	3.3	24
VVGQDNNEVFFK*IK	K51	2	5.6006	0.5692	3.2	22
VVGQDNNEVFFK*IK	K51	3	3.8757	0.5546	1.3	8
K*TTEFSK	K54	1	1.7313	0.3521	4.5	1
TTEFSK*LMK	K60	1	1.7629	0.3435	3.6	2
TTEFSK*LMK	K60	2	3.3620	0.5519	2.6	4
LMK*IYCAR	K63	1	2.6702	0.4424	4.5	1

^aPeptides containing potential SUMO chain linkages were identified with the SEQUEST algorithm using a differential modification search that considered a potential mass shift of +221.1270 on modified lysines. Charge state refers to the charge state of the precursor ion that corresponds to the spectra identified by tandem mass spectrometry. Xcorr and ΔCN are scoring metrics from the SEQUEST algorithm that indicate how well the actual spectrum matches the theoretical spectrum with the highest Xcorr and ΔCN listed for each peptide. ΔPPM is the difference in mass between the measured peptide precursor with the highest Xcorr and its predicted peptide sequence. The number of spectra indicates the number of independent spectra matching the indicated peptide sequence. K*, SUMO-modified lysine residue.

Critical SUMO chain-independent functions for Pli1 and Ubc9:SUMO.

Cells lacking Pli1 require the homologous recombination (HR) factor Rhp51 (Rad51) for viability (Fig. 6 A) (56). We also found that pli1Δ is synthetic lethal with mus81Δ, which inactivates the heterodimeric Holliday junction resolvase Mus81-Eme1 that plays a specific role in replication restart following fork collapse (4, 41; also data not shown). Based on these specific genetic interactions, we hypothesized that topoisomerase I (Top1) may be inducing replication fork collapse in pli1Δ cells. Indeed, deletion of Top1 rescued the synthetic lethal relationship between pli1Δ and rhp51Δ or mus81Δ (Fig. 6A and data not shown). These results echo the similar genetic relationships uncovered between budding yeast SUMO ligase-deficient cells and the HR machinery, suggesting that the mechanism is evolutionarily conserved (10). Notably, we determined that like pli1Δ cells, SUMO^D81R cells also require Rhp51 for viability (Fig. 6B). Furthermore, the synthetic lethality of SUMO^D81R rhp51Δ cells is suppressed by deletion of Top1 (Fig. 6B). To test if this pli1Δ and SUMO^D81R phenotype is due to the defect in SUMO chain formation exhibited by these mutants, we crossed SUMO^K14/30R cells with rhp51Δ cells. In contrast to the pli1Δ and SUMO^D81R mutations, the SUMO^K14/30R mutation is viable in combination with rhp51Δ (Fig. 6C). Thus, the lethality of both pli1Δ and SUMO^D81R with rhp51Δ is not due to a defect in SUMO chain formation. Finally, we tested the genetic interaction between SUMO^K14/30R and nse2-SA or rad60^E380R. Again, distinct from the synthetic sickness or lethality of the SUMO^D81R mutation in combination with nse2-SA and rad60^E380R, SUMO^K14/30R had no effect on the growth or genotoxin sensitivity of either nse2-SA or rad60^E380R cells (Fig. 6D and E). Overall, these data demonstrate that the Ubc9:SUMO complex has both a SUMO chain formation role, in addition to facilitating SUMO chain-independent Pli1 functions in global sumoylation.

Fig. 6.

Critical SUMO chain-independent functions for Pli1 and Ubc9:SUMO. (A) A representative tetrad dissection is shown of an rhp51Δ pli1Δ top1Δ strain back-crossed against a wild-type strain. The key depicts the genotypes present, which are denoted by various shapes around each colony; wild-type colonies are not marked. (B) As described above, with the rhp51Δ top1Δ strain crossed against the SUMO^D81R mutant. (C) As described above, from a cross between the rhp51Δ and SUMO^K14/30R mutants. (D and E) Serial dilutions of the indicated strains plated on medium that was either drug free (No Drug) or treated with the indicated concentrations of HU or camptothecin (CPT). All cells were grown at 30°C.

DISCUSSION

Global sumoylation, SUMO chain formation, and genome stabilization are all outputs generated by a limited repertoire of enzymes. Mechanisms driving selectivity for each of these processes are largely uncharacterized. Our structure determines that SLD2 occupies the same noncovalent interface previously shown to bind SUMO (9, 15, 22, 48). Moreover, our NFATC2IP SLD2 structure indicates that this binding is evolutionarily conserved. Due to this previously undefined structural and functional duality, it is therefore necessary to consider the potentially compound effects of Ubc9 mutations at this noncovalent Rad60/SUMO binding interface. In addition, these results raise a fundamental question: How does each noncovalent Ubc9 complex contribute to the regulation of sumoylation in vivo?

It has been proposed that, based on the potentially competitive nature of occupancy at the Ubc9 noncovalent interface, Rad60 could antagonize SUMO chain formation (39, 42). However, if blocking Ubc9:SUMO complex formation were the sole function of SLD2, then the SUMO^D81R mutation would be expected to suppress the phenotypes of rad60^E380R cells. Instead, we see a strong negative genetic interaction between the SUMO^D81R and rad60^E380R alleles, which indicates nonoverlapping roles of each noncovalent Ubc9 complex. The latter observation agrees with our in vivo analyses that delineate roles for Ubc9:SLD2 and Ubc9:SUMO in promoting Nse2- and Pli1-dependent sumoylation, respectively.

SUMO^D81R cannot form a noncovalent complex with Ubc9 (Ubc9:SUMO), and phenotypes of SUMO^D81R cells are synonymous with those of cells with deletions of Pli1. For example, both SUMO^D81R and pli1Δ cells exhibit strong synthetic sickness with either nse2-SA or rad60^E380R. These genetic interactions are not due to defective SUMO chain formation in pli1Δ and SUMO^D81R cells as the SUMO^K14/30R mutant, which largely blocks chain formation, is not sick in either the nse2-SA or rad60^E380R background. Furthermore, both SUMO^D81R and pli1Δ mutants, but not SUMO^K14/30R, exhibit Top1-dependent lethality when combined with the rhp51Δ mutant. These genetic interactions highlight functions of Ubc9:SUMO in Pli1-mediated sumoylation that are independent of SUMO chain formation. On the other hand, analysis of SUMO chain-dependent functions of Ubc9:SUMO has provided important mechanistic insight on STUbL function. We determined that both SUMO^D81R and SUMO^K14/30R suppress phenotypes associated with STUbL dysfunction, similar to deleting Pli1. These results determine that it is solely the chain-propagating role of Pli1 and Ubc9:SUMO that is toxic to STUbL mutant cells. Overall, these data demonstrate that Ubc9:SUMO has separable functions in sumoylation, either to promote global sumoylation or SUMO chain formation (Fig. 7).

Fig. 7.

Model for the roles of Ubc9:SLD2 and Ubc9:SUMO complexes in sumoylation. The Ubc9:SUMO complex supports both Pli1-mediated global sumoylation and Pli1-mediated SUMO chain formation. The Ubc9:Rad60 SLD2 complex, through the observed interaction of Rad60 with the Smc5/6 complex, facilitates sumoylation by the Nse2 E3 SUMO ligase. Rad60 also has Ubc9 complex-independent functions as indicated. Although not supported by our data, we do not fully exclude the possibility that Rad60, by competing for Ubc9 binding, can play a minor role in modulating the function of Ubc9:SUMO. Additionally, Nse2 could also play a role in SUMO chain formation, but such species were not detectable in our analyses. See Discussion for further details.

In contrast to the SUMO^D81R mutation, specific disruption of the Ubc9:SLD2 interface by rad60^E380R causes nse2-SA-like phenotypes. Examples of such phenotypes are hypersensitivity to genotoxins, synthetic sickness in combination with either SUMO^D81R or pli1Δ, and no major defects in global sumoylation (39, 55). In addition, Rad60 is a physical interactor of Smc5/6 (5) and a multicopy suppressor of defects in the Smc5/6 complex, of which Nse2 is a core component (28, 30). Further strong support for Rad60 and paralogs functioning in Nse2-dependent sumoylation comes from analyses of recombination defects in cells hypomorphic for Nse2/Mms21, Rad60/Esc2, or Ubc9. During replicative stress, toxic Rad51-dependent recombination structures are formed in mutants of each of the foregoing factors (6, 12, 27, 44). Notably, accumulation of these toxic structures can be suppressed in both esc2Δ and smc5/6 mutants by deletion of the FANCM-related helicase Mph1 (12, 46). Together with our biochemical and genetic results, the aforementioned data provide compelling support for functional overlap between Rad60 paralogs and Smc5/6-mediated sumoylation by the Nse2 subunit. These findings are incorporated in our model for the roles of Ubc9:SLD2 and Ubc9:SUMO in sumoylation (Fig. 7).

We propose that through transient interaction with Smc5/6, Rad60 could bring Ubc9-SUMO thioester species into proximity with the Nse2 SUMO E3 ligase to facilitate target sumoylation. Similar to this proposal, our results further indicate that the key role of Ubc9:SUMO is to enhance the sumoylation reaction that is catalyzed by Pli1. Notably, we determined that Pli1 is absolutely required for in vivo SUMO chain formation, as assayed in the STUbL genetic background that enables detection of SUMO polymers (Fig. 5) (39). In the same setting, the SUMO^D81R mutant is able to form Pli1-dependent di-SUMO species but not the much-higher-order species observed with wild-type SUMO. Elegant in vitro studies have previously implicated the Ubc9:SUMO interface in the formation of SUMO chains (9, 15, 22, 48). Our data show that this mechanism does occur in vivo and, furthermore, demonstrate that in marked contrast to prior in vitro results, an E3 SUMO ligase is essential for in vivo SUMO chain formation (including di-SUMO). Moreover, since endogenous SUMO conjugates are barely detectable in SUMO^D81R, SUMO^K14/30R, and pli1Δ cells, such conjugates in wild-type cells are predominantly Pli1-dependent chains.

We envisage that the noncovalent Ubc9:SUMO complex likely facilitates Pli1-dependent sumoylation in vivo through two mechanisms. First, recognition of sumoylated proteins via the noncovalent Ubc9:SUMO interface would increase the local concentration of Ubc9-SUMO and thus, promote Pli1-dependent chain formation. Notably, the high expression levels of SUMO, E1, and E2 components in our bacterial sumoylation assay efficiently generate SUMO chains. Due to this artificially induced concentration effect, chain formation in this system does not require the noncovalent Ubc9:SUMO complex (i.e., chain formation is not blocked by SUMO^D81R) or the E3 ligase Pli1. Second, as Ubc9 is itself sumoylated (21), the noncovalent Ubc9:SUMO interface could mediate self-multimerization of the Ubc9-SUMO species, thereby forming a concentrated pool of conjugatable SUMO. Next, the SUMO-interacting motif (SIM) of Pli1 would bring this pool into proximity with Pli1 to stimulate sumoylation. The latter model is analogous to the proposed role of the UbcH5:ubiquitin complex in promoting Brca1-Bard1-dependent ubiquitination through UbcH5 multimerization (7). Both of these potential mechanisms are consistent with SUMO^D81R blocking SUMO chain formation in vivo, as processivity requires high local concentrations of conjugatable SUMO. Herein, we have delineated DNA repair and global sumoylation roles for the noncovalent Ubc9 interface. Depending on the interaction partner, Ubc9 can be recruited to facilitate sumoylation carried out by either of the two major SUMO E3 ligases of fission yeast. Ubc9:SLD2 promotes genomic stability, most likely through Nse2-dependent sumoylation, whereas Ubc9:SUMO drives Pli1-dependent sumoylation, including SUMO chain formation.