Identification of a new small ubiquitin-like modifier (SUMO)-interacting motif in the E3 ligase PIASy

Small ubiquitin-like modifier (SUMO) conjugation is a reversible post-translational modification process implicated in the regulation of gene transcription, DNA repair, and cell cycle. SUMOylation depends on the sequential activities of E1 activating, E2 conjugating, and E3 ligating enzymes. SUMO E3 ligases enhance transfer of SUMO from the charged E2 enzyme to the substrate. We have previously identified PIASy, a member of the Siz/protein inhibitor of activated STAT (PIAS) RING family of SUMO E3 ligases, as essential for mitotic chromosomal SUMOylation in frog egg extracts and demonstrated that it can mediate effective SUMOylation. To address how PIASy catalyzes SUMOylation, we examined various truncations of PIASy for their ability to mediate SUMOylation. Using NMR chemical shift mapping and mutagenesis, we identified a new SUMO-interacting motif (SIM) in PIASy. The new SIM and the currently known SIM are both located at the C terminus of PIASy, and both are required for the full ligase activity of PIASy. Our results provide novel insights into the mechanism of PIASy-mediated SUMOylation. PIASy adds to the growing list of SUMO E3 ligases containing multiple SIMs that play important roles in the E3 ligase activity.

Most known SUMO E3 ligases possess one or more SUMOinteracting motifs (SIMs) that mediate non-covalent interactions with SUMO (2,3,(5)(6)(7)(8)(9)13). SIMs are required for the E3 ligase activity and play an important role in positioning the donor SUMO from the E2-SUMO complex for efficient SUMO discharge and catalysis. SIMs consist of a core sequence of three or four hydrophobic aliphatic residues (isoleucine, leucine, and valine) with an acidic or polar residue at position 2 or 3, forming a triad or tetrad core sequence of (V/I)X(V/I)(V/I) or (V/I)(V/ I)X(V/I/L) (where X is usually Asp, Glu, or Ser) (6, 14 -16). The hydrophobic core of SIM is often flanked by a stretch of acidic amino acids that promote electrostatic SUMO-SIM interactions (14,17). Structural studies of SUMO-SIM complexes have demonstrated that SIM peptides adopt a parallel or antiparallel ␤-strand conformation upon binding to SUMO in a hydrophobic groove formed by the ␤2-strand and ␣1-helix (5,14,17,18).
The PIAS and Siz families of SUMO E3 ligases (2, 4, 10, 13) contain several conserved regions (see Fig. 1A) such as an N-terminal SAP (scaffold-associating region/Acinus/PIAS) domain involved in binding AT-rich DNA, a PINIT domain important for substrate recognition and subcellular localization, an SP-RING needed for interaction with the E2 Ubc9, and a C-terminal SIM needed for binding SUMO (2,3,13). Currently, PIASy is known to contain one SIM at the C-terminal region (4,13). We report here the identification of a second SIM in PIASy, positioned between the SP-RING and the previously known SIM. NMR spectroscopy and mutational analysis revealed the significance of the new SIM in PIASy-mediated SUMO conjugation.

SUMO ligase activity of C-terminal truncations of PIASy
We have previously demonstrated the crucial role of PIASy in SUMO-2/3 modification of mitotic substrates TopoII␣ and PARP1 (10 -12). Our initial deletion analysis of PIASy has shown that both N-and C-terminal regions of PIASy are essential for its activity (10). Although we demonstrated that the N terminus is required for localization of PIASy at the mitotic centromeres (19), the role of the C terminus of PIASy is unclear. To evaluate the role of the C-terminal domain of PIASy in SUMOylation, we tested C-terminal deletions of PIASy (Fig.  1A) in the reconstituted in vitro SUMOylation assay (11). At physiological concentration of the SUMOylation enzymes, PIASy N414 (a fragment lacking the 87 C-terminal residues) was inefficient in SUMOylating both TopoII␣ and PARP1 (Fig.  1, B and C). The extent of SUMO conjugation using N414 was similar to the SUMOylation observed in the absence of PIASy (depicted by comparable higher shifted bands with N414 and no PIASy for both TopoII␣ and PARP1; Fig. 1, B and C, last two lanes). Next, we examined a longer PIASy construct, N454, that lacked the currently known SIM (here after referred as original SIM) but retained all the components previously reported as sufficient for the ligase activity of Siz1, the PIAS homolog in yeast (4,20). Surprisingly, N454 lacking the original SIM was capable of SUMOylating both the mitotic substrates (Fig. 1, B and C) albeit at a lower efficiency compared with full-length PIASy (WT). Consistent with the previous observation in Siz1 (4), inclusion of the original SIM and the flanking acidic domain (PIASy N480) further enhanced the SUMO conjugation (Fig. 1, B and C).
Furthermore, results of a fluorescence polarization binding assay using SUMO-3 labeled with fluorescein maleimide and PIASy(287-454) showed an apparent K d of 81 M (Fig. 2E). The value obtained is consistent with the observed intermediate exchange NMR behavior and previously reported dissociation constants for other SUMO-SIM interactions, which are in the 1-120 M range (6,9,14,18,22).
We also titrated a longer construct containing the original SIM, PIASy(287-501), into 15 N-labeled SUMO-3 (Fig. 3A). As expected, incorporation of the original SIM and the acidic domain resulted in more pronounced chemical shift perturbations when compared with results of the NMR titrations using the shorter PIASy(287-454) (Fig. 3B). Stepwise addition of PIASy(287-501) caused a concentration-dependent decrease in intensities of SUMO-3 peaks. Some SUMO-3 residues showed changes in peak positions in addition to reduced peak intensities (e.g. Arg-35, Lys-41, Tyr-46, and Asn-67), and several peaks broadened beyond detection (Fig. 3B). Although most SUMO-3 residues showed some degree of line broadening from the beginning of the titration, which is not uncommon with the formation of a high molecular mass complex (SUMO-3, 13 kDa; PIASy(287-501), 28 kDa) slowing down the tumbling rate (23), specific residues were more affected than the others as illustrated by a peak intensity ratio graph (Fig. 3C). The SUMO-3 residues strongly affected by PIASy(287-501) were identical to the residues affected by PIASy(287-454) and mapped to the ␤2-strand and ␣1-helix, which form the SIM-binding surface on the SUMO (Fig. 3D).

Identification of a new SIM in PIASy
NMR titrations of 15 N-labeled SUMO-3 with PIASy(287-454), which lacked the original SIM (Fig. 2), also perturbed the same SUMO-3 residues that a SIM peptide would, suggesting that an additional SIM may be present within the PIASy(287-454) fragment. Furthermore, a SIM prediction web server (24) suggested that there is a high probability that the PIASy(287-454) fragment contains a SIM (herein termed as new SIM) within residues 417-435 (Fig. 4A). The new SIM harbors a conserved hydrophobic core with the sequence "ILVL" from amphibian to mammalian PIASy and "VVDI" in fish PIASy and is preceded by a few acidic and serine residues, which are potential sites for phosphorylation (Fig. 4A). These features of the new SIM are similar to the characteristic known SIM sequences that bind SUMO in antiparallel orientation (9,14,15,22). However, unlike known SIMs, the new SIM lacks a long stretch of acidic residues. (The acidic stretch in PIAS proteins varies from 5 to 7 residues, but the acidic stretch in PIASy is exceptionally long with 17 residues following the known SIM.) To confirm the role of the new SIM in SUMO binding, we subcloned and expressed a 45-residue peptide encompassing the PIASy region containing the new SIM fragment (PIASy(409 -454) WT; Fig. 4A). We also mutated the three critical ILV hydrophobic residues of the new SIM into triple alanines (PIASy(409 -454) mut; Fig. 4A). Both WT and mut PIASy(409 -454) showed greater stability and solubility in comparison with the longer PIASy fragments used for NMR that contained the SP-RING and were thus ideal for NMR studies. NMR titrations using 15 N-labeled SUMO-3 with either WT or mut PIASy(409 -454) were used to test the role of the ILV hydrophobic residues in the new SIM in binding SUMO-3. The WT fragment resulted in concentration-dependent changes in peak positions of SUMO-3 resonances (Fig. 4, B and C), indicating PIASy(409 -454) WT interaction with SUMO-3 at the fast exchange NMR time scale. In contrast, NMR titrations with the mut fragment barely induced any discernible chemical shift perturbations on SUMO-3 (Fig. 4, D and E), indicating that mutating the ILV residues of the new SIM into triple alanines interfered with SUMO-3 interaction. Analysis of the weighted chemical shift deviations (CSDs; Fig. 4F) identified that the strongly affected SUMO-3 residues upon binding of PIASy(409 -454) WT are clustered near the known SIM-binding groove of SUMO-3 (Fig. 4G). These results highlight the importance of the ILV hydrophobic residues of the new SIM in PIASy-SUMO recognition.
We also titrated 15 N-labeled SUMO-3 with a longer construct, PIASy(409 -501), which contains both the new SIM and the original SIM plus the acidic domain (Fig. 5A). Similar to the earlier titrations, identical SIM-binding residues of SUMO-3 were most strongly perturbed upon the addition of PIASy(409 -501) (Fig. 5, B, C, and D). Comparison of the results of NMR titrations of SUMO-3 with PIASy(409 -501) (where both SIMs are present) and PIASy(409 -454) (where only the new SIM is present) indicated that the presence of both SIMs conferred a higher magnitude of change in the chemical shift perturbations of specific SUMO-3 resonances even at a lower ligand concentration. Important SUMO-3 SIM-binding residues (such as Gln-30, Phe-31, Leu-42, Ala-45, and Tyr-46) showed severe peak intensity reduction starting at 1:0.25 molar ratios and complete peak broadening at a 1:0.5 molar ratio (Fig. 5, B and C), indicating binding in the intermediate exchange NMR time scale.

The new SIM is crucial in PIASy-mediated SUMOylation
To evaluate the significance of the new and original SIMs for the E3 ligase activity of PIASy, we mutated the three essential core residues of the new and original SIMs into alanines (Fig.  6A). Three full-length PIASy constructs with multiple alanine mutations were introduced in the original SIM (construct ⌬O-SIM; Fig. 6A), the new SIM (construct ⌬N-SIM), and both new and original SIMs (construct ⌬W-SIM). The mutants were examined for their ability to facilitate SUMOylation of TopoII␣ and PARP1 in the in vitro SUMOylation assay (Fig. 6, B and C). Mutation of the hydrophobic core of original SIM (⌬O-SIM) did not interfere with the E3 ligase role of PIASy and exhibited high SUMOylation levels close to the wild-type PIASy for both TopoII␣ and PARP1 (Fig. 6, B, C, D, and E). However, mutation of the new SIM (⌬N-SIM) resulted in less efficient SUMOylation of TopoII␣ (Fig. 6, B and D) and significantly low SUMOylation of PARP1 compared with the WT PIASy (Fig. 6, C and E). Furthermore, mutation of both the original and new SIMs (⌬W-SIM) significantly reduced SUMOylation of TopoII␣ and nearly abolished the E3 ligase activity of PIASy on PARP1 (Fig.  6, D and E). Overall, both TopoII␣ and PARP1 showed a similar trend in SUMOylation, displaying a reproducible reduction in SUMOylation with the ⌬N-SIM and the ⌬W-SIM PIASy mutants in comparison with the WT PIASy. However, the observed effects were more distinct for PARP1 compared with TopoII␣ (largely due to the ability of TopoII␣ to undergo SUMOylation even in the absence of PIASy). Our results suggest that the new SIM plays an important role in PIASy-mediated SUMOylation, and both SIMs together are needed for the full ligase activity of PIASy.

Discussion
Results of our NMR titrations and in vitro SUMOylation assays revealed a new SIM in PIASy that was found to be essential for efficient PIASy-dependent SUMO-2/3 conjugation. In contrast to the original SIM of PIASy (or the known C-terminal SIM of other PIAS family members), the new SIM contains an uninterrupted core of four hydrophobic residues and lacks the conventional extended acidic residue stretch (Fig. 4). Previous studies of SUMO-SIM interaction have identified the importance of the hydrophobic core of SIM in SUMO binding (6,14,16). Consistent with these findings, our NMR results showed that hydrophobic core residues of the new SIM are essential in binding to SUMO-3, and mutation of these critical residues abolishes binding (Fig. 4). Others have shown the role of the acidic stretch surrounding the hydrophobic core of SIM in SUMO paralog binding specificity (14). Although presence of the negative charges in SIM has been shown to contribute more toward SUMO-1 binding, SIMs lacking the acidic stretch preferentially bind SUMO-2 (7,14). The absence of the acidic stretch in the new SIM, together with the essential role of the new SIM in PIASy-mediated SUMOylation (Fig. 6), provides a likely explanation for the preference of PIASy in conjugation of SUMO-2/3 over SUMO-1 to substrates TopoII␣ and PARP1 (10 -12).
To address the question of why there are two SIMs in PIASy, the following observations are instructive. First, in addition to the covalent interaction between the E2 conjugating enzyme Ubc9 and SUMO to form the E2-SUMO thioester complex, the backside of Ubc9 also recruits a second molecule of SUMO through non-covalent interactions to form the E2-SUMO complex (25,26). Second, the E3 ligase ZNF451 exploits these interactions to preferentially interact with two molecules of SUMO, the donor SUMO from E2-SUMO D and the backside E2-SUMO, via its dual SIMs (8,9). In ZNF451, the first SIM plus the inter-SIM region positions the E2-SUMO D thioester complex for productive catalysis, whereas the second SIM acts as an anchor to bind to the backside SUMO to facilitate poly-SU-MOylation (8,9). Finally, although PIAS proteins use the SP-RING domain to directly bind to E2 Ubc9, others have shown that the (original) SIM of PIAS proteins also interacts with the backside E2-SUMO to form a non-covalent ternary complex (27). Combining our discovery of the new SIM in PIASy with the above observations (8,9,(25)(26)(27), we propose a working hypothesis that, analogously to ZNF451, the dual SIMs of PIASy may also interact simultaneously with the donor SUMO (E2-SUMO D ) and the backside SUMO (E2-SUMO) to mediate efficient SUMOylation.
Our results can be interpreted in line with this working hypothesis. Our NMR titrations with a PIASy construct containing both the new and original SIMs showed improved SUMO-binding affinity (Fig. 5) compared with a shorter construct consisting only of the new SIM (Fig. 4). Even with small amounts of PIASy(409 -501) (containing dual SIMs), specific SUMO-3 SIM-interacting residues including Gln-30, Phe-31, Leu-42, Ala-45, and Tyr-46, were in intermediate exchange (where peaks were broadened) (Fig. 5); however, with PIASy(409 -454) (containing the new SIM alone), these resi- . Gray and red lines depict mean and one standard deviation from the mean (1), respectively. D, the SUMO-3 residues affected upon complex formation with PIASy(409 -501) are highlighted (red, residues displaying significant peak intensity reduction; orange, residues with significant change in the peak position). Error bars represent S.D.
dues were in the fast exchange NMR time scale (Fig. 4), suggesting relatively weaker binding compared with PIASy(409 -501). Because both SIMs individually bind to the same surface on SUMO-3 through classical SUMO-SIM interactions (27), the stronger NMR shift perturbations observed with PIASy(409 -501) are probably reflective of a synergistic effect occurring upon the binding of different SUMO-3 molecules to the tandem SIM.
Additionally, our in vitro SUMOylation assays confirmed that PIASy stimulated SUMO conjugation in a SIM-dependent manner and that both SIMs are important for effective ligase activity. Although mutation of the new SIM (⌬N-SIM) rendered SUMOylation slow and inefficient, probably implying its role in the donor SUMO (E2-SUMO D ) positioning and chain initiation, mutation of both SIMs together (⌬W-SIM; Fig. 6) or deletion of the dual SIM region (N414; Fig. 1) nearly eliminated the PIASy E3 ligase activity. In concordance, the presence of both SIMs together resulted in the robust ligase activity of PIASy (N480; Fig. 1). Our results suggest a hypothesis where the SP-RING domain and the new SIM of PIASy may configure the E2-SUMO D thioester complex, and the original SIM may engage the backside SUMO for efficient catalysis and SUMO conjugation by PIASy. To summarize, we have identified a new SIM in PIASy that is critical for the efficient ligase activity of PIASy and distinct from other members of the PIAS family that contain a single SIM at their C terminus.

Protein expression and purification
For protein expression, plasmids were freshly transformed in Escherichia coli BL21(DE3) cells and grown in 1 liter of culture medium containing kanamycin at 30 g/ml. 15 N-Labeled SUMO-3 was expressed in M9 minimal medium supplemented with 1 g of [ 15 N]ammonium chloride. Unlabeled SUMO-3 (for fluorescence polarization studies) and PIASy fragments (for NMR studies) were obtained by cell growth in LB medium. PIASy constructs containing the RING domain were also supplied 0.1 mM ZnSO 4 before and after induction. Cells were grown at 37°C until A 600 was ϳ0.7-0.8 and induced with 0.5 mM isopropyl ␤-D-thiogalactopyranoside, and cell growth was continued overnight in a 15°C shaker incubator to a final A 600 ϳ2.5 (ϳ2.1 for PIASy constructs). Cells were harvested by centrifugation (4000 rpm for 10 min), resuspended in ϳ30 ml of binding buffer (500 mM NaCl, 20 mM Tris-HCl, pH 8.0, 5 mM imidazole), and lysed by sonication. Cellular debris was removed by centrifugation (13,000 rpm for 10 min), and 600 l of 5% (v/v) polyethyleneimine was added to the supernatant to precipitate the nucleic acids. After another centrifugation (13,000 rpm for 10 min), the supernatant was loaded to a precharged Ni 2ϩ affinity column (Gold Biotechnology). The column was washed with 100 ml of binding buffer and eluted with 50 ml of elution buffer (500 mM NaCl, 20 mM Tris-HCl, pH 8.0, 250 mM imidazole). Elution fractions containing purified protein were pooled and dialyzed in NMR buffer (100 mM NaCl, 10 mM NaPO 4 , pH 6.8, 5 mM DTT). Proteins were concentrated using Amicon Ultra 3000 centrifugal filter (Millipore), and concentrations were determined by absorbance at A 280 . Recombinant proteins used in the SUMOylation assays, the E1 complex (Uba2/Aos1), E2 Ubc9, PIASy (full length and truncation and mutant constructs), PARP1, TopoII␣, and SUMO-2 were expressed and purified as described previously (10 -12, 28).

In vitro SUMOylation assay
SUMOylation reactions were performed as described previously (11,12). Briefly, 5 M SUMO-2, 2 mM ATP, 20 nM E1, 30 nM E2, 20 nM PIASy, and 500 nM T7-tagged PARP1 or TopoII␣ in reaction buffer (20 mM HEPES, pH 7.8, 100 mM NaCl, 5 mM MgCl 2 , 0.05% Tween 20, 5% glycerol, 0.5 mM DTT) were incubated at 25°C for 60 min. Although at a lower efficiency, TopoII␣ can be SUMOylated in the absence of PIASy, and this effect becomes more prominent with longer reaction times (12,19). Therefore, to better understand the contribution of SIMs of PIASy in TopoII␣ SUMOylation, TopoII␣ reactions with the SIM mutants were incubated for up to 30 min (Fig. 6B). Ali-quots were taken from the reaction at various time points, mixed with SDS-PAGE loading buffer, and boiled. Samples were resolved by SDS-PAGE and analyzed by Western blotting using HRP-conjugated anti-T7 monoclonal antibody.

Fluorescence polarization (FP)
Purified SUMO-3 and PIASy(287-454) were dialyzed into 1ϫ PBS buffer (phosphate-buffered saline, pH 7.4) overnight. The single native cysteine of SUMO-3 (residue 47) was utilized to covalently attach the fluorophore, fluorescein maleimide (FM) following the manufacturer's protocol (Invitrogen). Briefly, a 20-fold molar excess of FM prepared in dimethyl sulfoxide was added to the tris(2-carboxyethyl)phosphine-reduced protein and allowed to react in the dark at room temperature for ϳ3 h. The unbound excess dye was removed by extensive dialysis in PBS followed by passage through an Amicon Ultra 3000 centrifugal filter. Increasing concentrations of PIASy (0 -360 M) were titrated into 25 nM FM-labeled SUMO-3. Samples were incubated at room temperature for 30 min, and FP was measured with excitation and emission wavelengths of 492 and 515 nm, respectively, using a Varian Cary Eclipse fluorescence spectrophotometer. G-factor-corrected polarization values were obtained using the following formula. P ϭ ͑I VV Ϫ G ϫ I VH ͒/͑I VV ϩ G ϫ I VH ͒ (Eq. 1) Background fluorescence was subtracted from the average of five data points, and the change in millipolarization (⌬mP) was plotted against increasing concentrations of PIASy. The dissociation constant (K d ) was estimated by non-linear hyperbolic fitting of the data using GraphPad Prism 5.0.
Author contributions-K. K. and Y. A. designed, performed, and interpreted the majority of the experiments. H. P. and N. P. assisted in some of the SUMOylation experiments. R. N. D. G. helped in designing and interpreting NMR experiments. K. K. drafted the manuscript, and everyone contributed in revising the manuscript.