Structural analysis of a complex between small ubiquitin-like modifier 1 (SUMO1) and the ZZ domain of CREB-binding protein (CBP/p300) reveals a new interaction surface on SUMO

We have recently discovered that the ZZ zinc finger domain represents a novel SUMO binding motif. In this study we identify the binding epitopes in the ZZ domain of CBP and SUMO1 using NMR spectroscopy. The binding site on SUMO1 represents a unique epitope for SUMO interaction spatially opposite to that observed for canonical SUMO Interaction Motifs (SIMs). HADDOCK docking simulations using chemical shift perturbations and residual dipolar couplings was employed to obtain a structural model for the ZZ domain-SUMO1 complex. ITC experiments supports this model by showing that the mutation of key residues in the binding site abolishes binding and that SUMO1 can simultaneously and non-cooperatively bind both the ZZ domain and a canonical SIM motif. The binding dynamics of SUMO1 was further characterized using 15 N CPMG relaxation dispersions, which

SUMO (small ubiquitin-like modifier) is a protein structurally homologous to ubiquitin that functions as a post-translational modifier involved in diverse cellular processes such as chromatin remodeling, ubiquitin E3 ligation, autophagy, cytoskeletal scaffolding, and DNA repair. Higher eukaryotes encode at least three different functional SUMO isoforms, in the human genome designated as SUMO1-3, which can conjugate to a number of target proteins (1)(2)(3). SUMO modified target proteins are recognized by SUMO interacting motifs (SIMs), the minimal consensus sequence, which is ψ-K-X-E, where ψ consists of a large hydrophobic amino acid and where X can be any amino acid (4,5). In addition to this canonical type of SIM, closely related variants such as inverted SIMs and phosphorylation-dependent SIMs have also been recently described (6)(7)(8). The binding site for the canonical SIM is located in a groove between the α-helix and β-sheet in SUMO, where the SIM motif can bind either in a parallel or antiparallel fashion (6). We have recently shown that the ubiquitin ligase HERC2 can bind SUMO1 through its zinc-finger ZZ domain (9), which therefore represents a new class of SUMO binding motifs. Isothermal titration calorimetry (ITC) showed that the ZZ domain binds to SUMO1 with M affinity in the same range as the canonical SIMs (10). The ZZ domain consists of two anti-parallel β-sheets and a short α-helix coordinating two zinc ions as revealed by the first published ZZ domain structure originating from CBP (CREB-binding protein/p300) (11). There are approximately 20 ZZ domains (11,12) identified in the human genome, but a general biological function has not been assigned.
Here we present the complex between SUMO1 and the ZZ domain from CBP, where we map the interaction interface using NMR spectroscopy. The structure of this particular ZZ domain has previously been determined in solution by NMR and thus has favorable properties for structural studies (11). Using 15 N HSQC spectra, we could identify the interaction surfaces on the ZZ domain and SUMO1, where the residues affected by the binding were broadened beyond detection. On SUMO1 this constituted a unique binding epitope that subsequently could be used to model the protein-protein complex using HADDOCK docking simulations (13). Furthermore, in order to address the protein dynamics between different states of SUMO1 we performed CPMG relaxation dispersion experiments to map the micro-to millisecond dynamics of SUMO1, where we discovered that apo-SUMO1 experiences intrinsic conformational exchange. The conformational exchange was quenched in SIM-bound SUMO1, whereas the intrinsic conformational exchange in ZZ domainbound SUMO1 was largely unperturbed.

EXPERIMENTAL PROCEDURES
Molecular biology -DNA sequences for the ZZ domain from CBP (UniProtKB = Q92793) and SUMO1 (UniProtKB = P63165) were synthesized by Geneart and sub-cloned into a pNIC28-BsaI vector using ligation independent cloning (14). The resulting expression constructs contained a His-tag capture sequence and a tobacco etch virus (TEV) protease cleavage site at the N terminus preceding the expressed protein of interest. In order to obtain SUMO1 protein that behaves as monomers in solution without the need of the addition of an reducing agent, the single cysteine residue in SUMO1, Cys52, was mutated to serine using Quikchange site-directed mutagenesis (15) and transformed into MACH-1 cells. The SUMO1 mutant was shown to have the same structural and binding properties as the wildtype protein (data not shown). The C52S mutant was used in all subsequent experiments and is hereafter referred to as SUMO1. The sequences for the expression constructs were verified by sequencing. Finally, in order to test the importance of residues in the identified binding epitopes, one mutant was created for the ZZ domain, denoted ZZmut, corresponding to residues N36A, K38A and S39A, while two mutants were designed for SUMO1 corresponding to residues H75A, K78A and E83A denoted SUMO1mut1 and K16A, H43A, H75A, K78A and E83A denoted SUMO1mut2 respectively.
Protein expression and purification -Plasmids were transformed into Escherichia coli (E. coli) BL21 (DE3) Rosetta cells prior to protein expression. Protein expression was started from overnight cultures grown in Terrific Broth (TB) media supplemented with 50 g/ml kanamycin and 30 g/ml chloramphenicol. SUMO1 and the SUMO1 mutants were expressed in TB media with 50 g/ml kanamycin and 30 g/ml chloramphenicol, grown to OD600 ~ 0.6-0.7 at 37ºC, induced with 0.5 mM isopropyl-βthiogalactopyranoside (IPTG) and grown overnight at 18ºC. The ZZ domain and the ZZ mutant were expressed in TB media as above with the addition of 150 M ZnCl2 according to a previously published protocol (11). For the NMR experiments, SUMO1 and the ZZ domain were labeled with 15 N and 13 C using minimal media (16). SUMO1 was grown in minimal medium to OD600~0.6-0.7, induced with 0.5 mM IPTG, and grown overnight at 18ºC. The ZZ domain was grown in minimal media with 150 M ZnCl2 until reaching an OD600~1.2-1.5, induced with 1.0 mM IPTG and grown overnight at 18ºC. Cells were harvested by centrifugation at 6000 RPM for 10 min, resuspended in 100 mM Tris-HCl pH 8.0, 300 mM NaCl, 10 mM imidazole with benzonase and Complete-EDTA free tablets (Roche Life Sciences), homogenized using a French press cell disruptor and centrifuged at 11000 g for 30 min. Purification of the proteins was performed using immobilized metal ion affinity chromatography (IMAC) on an ÄKTA express system (GE Healthcare) at 4C with the HiTrap chelating columns (17) equilibrated with 50 mM Tris-HCl pH 8.0, 300 mM NaCl, 10 mM imidazole, and eluted using a linear gradient of 10-500 mM imidazole. Protein identity was verified using mass spectrometry (MS). The eluted fractions with protein were collected, dialyzed (Molecular weight (MW) cut-off 3500 Da) against 50 mM Tris-HCl pH 7.5, and processed with TEV protease in order to remove the His-tag. The TEV protease treatment retains the amino acids pair Ser-Met at the Nterminal of the expressed proteins. We have therefore designated the first two residues as S0 and M1 for both the ZZ domain and SUMO1. The final amino acid sequences for the two proteins are therefore; ZZ domain: SM1QDRFVYTCNECKHHVETRWHCTVCEDY DLCINCYNTKSHAHKMVKWGLGLDD53, and SUMO1: SM1SDQEAKPSTEDLGDKKEGEYIKLKVIGQ DSSEIHFKVKMTTHLKKLKESYSQRQGVPM NSLRFLFEGQRIADNHTPKELGMEEEDVIEV YQEQTGGHSTV101. In the final purification step, the proteins were loaded onto an ion exchange MonoQ column (ÄKTA, GE Healthcare) equilibrated with 50 mM Tris pH 7.5 and eluted with a linear gradient of 0-0.4 mM NaCl. The purity was verified by SDS-PAGE and electrospray ionization mass spectrometry (ESI-MS).
Peptide synthesis -The SIM-peptide from PIASX (6) hereinafter denoted SIMPX, corresponding to the following sequence; KVDVIDLTIESSSDEEEDPPAKRQM, was synthesized by Biosyntan Gmbh (Berlin, Germany). Purity and molecular mass was confirmed by MALDI-TOF MS.
Protein characterization -The secondary structure of the ZZ domains (ZZ wildtype and ZZmut) and the SUMO1 variants (SUMO1, SUMO1mut1, and SUMO1mut2) was verified by circular dichroism (CD) spectroscopy using 0.5-1.0 mg/ml protein in 20 mM MOPS buffer at pH 6.8.
Isothermal Titration Calorimetry (ITC) experiments -ITC experiments were performed using a VP-ITC200 instrument (GE Healthcare). The samples were extensively dialyzed against 20 mM MOPS pH 6.8. For the experiments addressing the binding properties of the ZZ domains (ZZ wildtype and ZZmut) and the SUMO variants (SUMO1, SUMO1mut1, and SUMO1mut2), 0.5 mM SUMO1 or SUMO1 mutant (SUMO1mut1 or SUMO1mut2) was titrated into 0.04 mM ZZ wildtype or ZZ mutant. The concentrations were determined from the respective absorbance at 280 nm. In the experiments investigating the SUMObinding of the ZZ domain and the SIM peptide SIMPX, 0.5 mM SIMPX was titrated into the ITC sample cell containing 0.05 mM SUMO1. Finally, 0.5 mM SIMPX was titrated into a sample containing the pre-formed ZZ domain-SUMO1 complex. All experiments were performed at 10 ºC and run until saturation was achieved. The data were fitted using a model describing one binding site using the software provided by the manufacturer (18).
NMR experiments and resonance assignments -Protein samples were dialyzed against 20 mM MOPS pH 6.8 and concentrated to 0.3-0.5 mM using Vivaspin concentrators with a MW cut-off of 3500 Da. D2O was added to a concentration of 7 % (v/v) for the spectrometer lock and NaN3 was added to a concentration of 0.02 % (w/v) to prevent bacterial growth in the samples. The backbone resonance assignments for SUMO1 and the ZZ domain were based on published assignments (11,19) in combination with standard 3D triple-resonance experiments for the backbone assignments (HNCA (20), HNCOCA (21), HNCACB (22) and CBCACONH (23)). Spectra were processed with NMRpipe (24), applying zerofilling and linear prediction in the indirect dimensions and solvent filter and polynomial baseline correction in the direct dimension. CCPNMR (25) was used for visualization of spectra and resonance connectivity analysis. Resonance assignment was achieved for 98% of the backbone amides in the ZZ domain and 86% in SUMO1. The SIMPX-bound SUMO1 was assigned by performing a titration of SIMPX into a SUMO1 sample and by following the chemical shift changes upon complex formation. The samples for measuring RDCs were prepared by adding the PF1 phage (ASLA Biotech) to a final concentration of 10 mg/ml in the NMR samples.
RDC measurements and molecular modeling -Residual dipolar couplings (RDCs) were measured using interleaved IPAP-experiments (26) at 25ºC with spectral widths of 8000 Hz ( 1 H) and 1835 Hz ( 15 N), sampled over 1024 and 256 points respectively. RDCs were measured for each component of the ZZ domain-SUMO1 complex using two separate samples, in which one protein was labeled with 15 N/ 13 C and the other unlabeled. In total, 91 RDCs were collected, 44 for SUMO1 and 47 for the ZZ domain.
Formation of the ZZ-SUMO1 and SIMPX-SUMO1 complexes -In order to confirm binding and establish complexes with 1:1 stoichiometry, the NMR samples were prepared by ITC titrations of unlabeled ZZ domain into 15 N/ 13 C labeled SUMO1, and vice versa. The final NMR samples of the ZZ domain-SUMO1 complex contains 4% unbound SUMO1, as calculated from the Kd and protein concentrations. The complex between 15 N/ 13 C labeled SUMO1 and SIMPX was formed in a similar way using ITC titrations resulting in NMR samples of the SIMPX-SUMO1 complex that contains 2% unbound SUMO1. 15 N HSQC experiments were subsequently run on the complexes and the resulting spectra were compared with those for unbound proteins.
Structure calculations -Structural refinements of SUMO1 and the ZZ domain was done using Xplor-NIH (27). For the ZZ domain, NOE restraints, dihedral and hydrogen bond restraints were adopted from the published structure (Protein Data Bank code 1TOT) (11) and imported from the Biological Magnetic Resonance Bank (BMRB) and converted from Amber (28) to Xplor-NIH format using in-house scripts. Zinc ions were included into the structure calculations by adding distance and angle restraints in a similar way as for the published structure of the ZZ domain of CBP (11). Structure calculations were initiated from an elongated protein structure using a simulated annealing protocol in which NOE, dihedral, hydrogen bond, and RDC restraints were applied as well as restraints for the zinc coordination. Structure calculations were performed in an iterative fashion, where NOE, dihedral angle, and RDC restraints were iteratively pruned from the calculations in order to obtain a converged molecular structure ensemble. Structure analysis were done using the protein structure validation software suite (PSVS) (29) and PALES (30). Residues in the N-and C-terminal (1-5, 49-53) were considered flexible based on observed peak intensities in the 15 N-HSQC, and were excluded from the structure calculations. In the case of the NMR structure of SUMO1 (Protein Data Bank code 1A5R), the restraints are not available. Synthetic NOE distance restraints were therefore created by collecting all 1 H-1 H distances in the first model of the structural ensemble of SUMO1 excluding all restraints corresponding to distances longer than 6.5Å or shorter than 2.8Å. The structure calculation protocol was otherwise the same as for the ZZ domain, where NOE distance restraints and RDCs were iteratively pruned. The weight of RDC restraints in the structure calculations was increased as compared to the calculations of the structure of the ZZ domain due to the use of synthetic NOE restraints. Residues in the N-and C-terminal were considered flexible (1-20, 95-103) based on R20 values from the CPMG relaxation dispersion experiments and were excluded from the structure calculations. The structure analysis were performed using the PSVS suite (29) and PALES (30).
Docking simulations -Docking simulations were performed using HADDOCK (13) and CNS (31) following published protocols (32). Chemical shift changes upon binding were used as ambiguous interaction restraints (AIR) in combination with RDCs measured for the ZZ domain-SUMO1 complex. The docking simulations involved active restraints for those residues experiencing chemical exchange, while neighboring residues were treated as passive. Residues with a solvent-accessible surface-area (SASA) < 20% were filtered out. The active residues in the ZZ domain were N36, K38 and A41, while the passive residues were E11, C12, W20, V24, C31, I32, N33, Y35, T37, S39 and H42. The active residues in SUMO1 were H43, K46, M82, E83 and E85, while the passive residues were K23, K25, T42, K45, G81, E84 and V87. RDCs belonging to residues in the flexible N-and Cterminals were not included as restraints in the docking simulations. The protonation state of the two histidine residues (H40 and H42), coordinating the zinc ion in the metal site of the ZZ domain, was set to match the metal coordination pattern (11). Docking simulations were run using the standard HADDOCK protocol with 1000 initial docking simulations, followed by 200 refinement simulations, and subsequently 200 final refinement simulations including explicit water molecules. RDCs were used as intervector projection angle restraints (33) in the first 2 simulation steps and as direct restraints in the final water refinement procedure.
NMR relaxation dispersion experiments -NMR 15 N CPMG relaxation dispersion were measured using CT-CPMG relaxation dispersion experiments (34,35) at static magnetic field strengths of 14.1 and 18.9 T at 10ºC. The temperature was calibrated using a methanol sample before each series of experiments. The constant relaxation time was 40 ms. The experiments utilized the phase cycle proposed by Yip and Zuiderweg to suppress artefacts due to off-resonance effects (36). The effective relaxation rate (R2eff) at a given CPMG refocusing frequency was determined from 2 data points, as described (37). Relaxation dispersions were sampled using 18-24 R2eff values. The spectral widths were 8000 Hz ( 1 H) and 1835 Hz ( 15 N) at 14.1 T and 10666 Hz ( 1 H) and 2447 Hz ( 15 N) at 18.9 T, with 512 and 64 points collected in direct and indirect dimensions, respectively. Spectra were processed using NMRpipe (24). The processing protocol involved a solvent filter, cosine-squared apodization functions, zero filling to twice the number of increments in all dimensions, and a polynomial baseline correction in the 1 H dimension. Peak intensities were measured as the integral over 5×3 points ( 1 H/ 15 N) centered on the peak maximum. The signal-to-noise ratio (S/N) was estimated by calculating the standard deviation of 200 samples of integrated 5×3-point windows in empty regions of each spectrum. Errors in the extracted R2eff were estimated from the S/N using error propagation. The relaxation dispersion data were analyzed using CPMGfit v2.3 (38). Relaxation dispersion curves were fitted to the Carver-Richards two-state exchange model (39,40): in which: and = kex 2 - 2 , = -2kex(1-2pm), kex = k1 + k-1 is the sum of the forward and reverse rate constants, is the chemical shift difference between the exchanging conformations, R20 is the average limiting value of the relaxation rate constant for processes other than chemical exchange, pm is the relative population of the minor (less populated) conformational state, which is related to the major conformational state pM as pm = 1 -pM, and cp = 1/ncp is the spacing between refocusing pulses in the CPMG train. The statistical significance of each fit was assessed by also fitting the data to a constant R20 value (i.e. modeling a flat dispersion profile, indicating the absence of exchange), and the F-test was used to discriminate between models by rejecting the null hypothesis that the model with more parameters does not provide a significantly better fit than the simpler model at the level p < 0.001. Errors in the fitted parameters were estimated from 1000 synthetic data sets created using Monte-Carlo simulations (41,42).

RESULTS
To define the molecular details in the interaction between the ZZ domain and SUMO1, we employed a combination of chemical shift mapping, RDCs, docking simulations, and protein engineering. In addition, experiments addressing the dynamic properties of the interactions between SUMO1 and SIMs were performed to advance our understanding of SUMO recognition.
Chemical shift mapping of the binding interface between SUMO1 and the ZZ domain -The binding interface on the ZZ domain and SUMO1 was mapped using binding-induced chemical shift differences measured in 15 N-HSQC spectra, where the binding was verified by ITC. A comparison of these 15 N HSQC spectra shows that 6 residues in the labeled SUMO1 are broadened beyond detection upon the addition of the ZZ domain due to intermediate chemical exchange between the free and bound states (Fig 1A). In a similar fashion, 5 residues in the ZZ domain are broadened beyond detection upon adding SUMO1 (Fig. 1B). The affected residues in SUMO1 are located in a contiguous region of the structure: L24 is situated in a -strand, and M82, E83 and E85 are found in an adjacent loop, while H43 and K46 are located in the N-terminal part of the -helix and preceding loop (Fig. 1C). This binding site is spatially distinct from that observed for classical SUMO interacting motifs (SIMs), which is located on the opposite side of the protein (4-6). The perturbed residues in the ZZ domain are located in the -helix and the following loop (Fig. 1D). The side chains of C34 and H40 are part of the metal coordination for both of the Zn 2+ ions present in the ZZ domain, indicating that the presence of bound Zn 2+ ions  (11)) with the structure software PALES (30), with Q-values of 0.93 and 0.90, respectively (43). The initial agreement between the RDCs back-calculated from the published structures and the experimental RDCs is low ( Fig. 2A, B), most likely because the structures have not been refined using RDC restraints. In order to improve the correlation between experimental and theoretical RDCs, the published protein structures were therefore first refined versus the experimental RDCs. Given that the changes in chemical shifts are minor upon forming the protein-protein complex, we do not expect any major conformational changes upon formation of the ZZ domain-SUMO1 complex. Hence, as an initial approximation, we assumed that the structure of the complex could be modeled using the published structures and the RDCs measured for the complex. Structural restraints have been published for the ZZ domain (Protein Data Bank code 1TOT) and these were converted to Xplor-NIH format using a structure refinement protocol from the program suite (44). RDCs measured for the protein-protein complex as well as NOE distance restraints, dihedral angle restraints, and hydrogen bond restraints were used for the refinement. Prior to refinement, the protein was modeled to correspond to the human ZZ domain by adding a C-terminal Asp and mutating Thr41 to Ala41, followed by energy minimization using Xplor-NIH. Structural refinement was done using the first model of the published structure ensemble as the input structure. Residues in the N-and Cterminal (1)(2)(3)(4)(5)(49)(50)(51)(52)(53) were considered flexible based on observed peak intensities in the 15 N-HSQC. Refinement was performed in an iterative manner until the resulting structure shows no NOE violations > 0.5 Å. Since there are no published distance restraints available for SUMO1, we employed a different approach for the RDC refinement of this structure. Synthetic NOE distance restraints were created by calculating all 1 H-1 H distances in the first model of the published protein structure (Protein Data Bank code 1A5R), filtering out distances larger than 6.5 Å and shorter than 2.8 Å. In addition, the protein structure was prepared for refinement by changing Cys52 to Ser52. Refinement was done iteratively using the same protocol as for the ZZ domain. Residues in the N-and C-terminal were considered flexible (1-20, 95-103) based on CPMG relaxation dispersion experiments and excluded from the calculations. The refinement was performed iteratively until the resulting structure had no NOE violations > 0.5 Å. For each iteration, NOE violations > 0.5 Å and distances > 6.5 Å were filtered out. Q-values after refinement were 0.30 and 0.26 for the ZZ domain and SUMO1, respectively (Fig 2C,D) (43). The refined protein structures for the ZZ domain and SUMO1 are highly similar to the original structures after RDC refinement with RMSD values for C of 0.29 Å and 0.68 Å for the ZZ domain (residues 6-48) and SUMO1 (residues 21-94), respectively. The main structural differences in terms of C RMSDs for the ZZ domain are observed for residues 10-18 in -strand 1 and in the SUMO1binding loop 36-46. The main differences in C RMSDs for SUMO1 are seen for residues 33-43 in -strand 2 and for residues 70-87 in the ZZ domain binding loop.
Structural model for the ZZ domain-SUMO1 complex -After final refinement in water, the docked structures were clustered using the standard HADDOCK algorithm, where 115 structures were clustered into one ensemble (Fig 3A, B). Nine additional clusters of structure ensembles were found, each containing 10 structures or fewer. The RMSD of all residues in the interacting protein interface is 1.6 ± 0.6 Å, whereas the RMSD between all backbone atoms in the docked structure was 1.6 ± 0.6 Å. The RMSD between the protein interface and backbone atoms was 1.3 ± 0.4 Å and 2.2 ± 0.6 Å for SUMO1 and the ZZ domain, respectively (Fig. 3C, D). The HADDOCK score for the final structure ensemble was -43 ± 18, and the Q-value calculated for the complex was 0. 36 showing that the experimental RDCs demonstrate a good agreement with the calculated RDCs for the modeled protein complex (Fig. 3E) (43). The magnitude and orientation of the RDC alignment tensor for the ZZ domain and SUMO1 and the ZZ domain-SUMO1 complex was calculated using Module (45), where the magnitudes and rhombocities are shown in Table 1. Using Module, the RDC alignment tensor was fitted separately for the ZZ domain and SUMO1 in the protein complex indicating highly similar tensor orientations and magnitudes ( Fig 3F). The observed structures for the ZZ domain and SUMO1 in the protein complex are nearly identical to the individually refined structures of the uncomplexed proteins with C RMSDs of 0.001 Å for both the ZZ domain and SUMO1. By extending the comparisons between complexed and uncomplexed proteins to include all atoms in the assessments, also incorporating all surface accessible sidechains, the RMSDs are 2.32 Å and 2.29 Å for the ZZ domain and SUMO1 respectively. As expected, the binding interface for the docked protein-complex corresponds to the surface residues experiencing chemical shift perturbations upon binding. The binding surface on the ZZ domain consists of the residues spanning from the C-terminal end of the -helix to the loop connecting the -helix to β-strand 1. The binding surface on SUMO1 consists of residues belonging to β-strand 1, the N-terminal part of the α-helix and the preceding loop as well as the long loop connecting β-strands 3 and 4. The binding surface on SUMO1 consists of a convex groove on the protein surface. This interaction surface represents a completely new binding epitope for SUMO1, previously not observed in the Protein Data Bank, which is clearly separated in space from known interaction motifs. The previously known interaction motifs for SUMO1 are the SUMO-SIM motif (4-6) the Ubc9 motif and the SENP/RanGAP motif (46,47). None of these interaction motifs are overlapping with the ZZ domain-SUMO1 binding site, suggesting that the ZZ domain can bind to SUMO1 in the presence of other interaction motifs. In order to verify the binding mode observed in the ZZ domain-SUMO1 complex, we created one mutant of the ZZ domain and two mutants of SUMO1, where we mutated residues in the binding interface to alanines. The residues selected for alanine substitution where either residues with charged side chains or with side chains with the potential to form hydrogen bonds, thus removing important electrostatic interactions in the binding interface. We find that the ZZ domain mutant still has the ability to bind SUMO1, albeit with significantly lower enthalpy of binding as compared to the wildtype (Fig. 4A, B and Table 2). The two mutants of SUMO1, did however, completely abolish binding to the ZZ domains ( Fig.  5A-D), clearly indicating that these residues are important for the complex formation and further supporting the molecular model derived from the NMR data. To compare the SUMO binding characteristics of the ZZ domain with a canonical SIM motif, we performed ITC experiments between SUMO1 and a SIM peptide denoted SIMPX from PIASx (6). As expected, we could show that SIMPX binds to SUMO1 with a similar affinity as the ZZ domain ( Fig. 4C and Table 2). To investigate whether SIMPX and the ZZ domain can bind to SUMO simultaneously, ITC experiments between SIMPX and a pre-formed complex between the ZZ domain and SUMO1 were carried out. These experiments revealed that SIMPX binds with a similar affinity and enthalpy of binding to SUMO1 as it does in the absence of the ZZ domain, indicating that SIMPX can bind SUMO1 independent of the ZZ domain in a non-cooperative manner (Fig. 4 C, D and Table 2).
CPMG relaxations dispersion experiments - 15 N CPMG relaxation dispersions were measured for apo, SIMPX-bound, and ZZ domain-bound SUMO1 at static magnetic field strengths of 14.1 and 18.9 T (Supplemental Table 1). Apo-SUMO1 showed 52 backbone amides experiencing significant conformational exchange (p<0.01), as exemplified in Fig. 6A. The exchanging residues are highlighted in Fig. 6D, showing that large parts of the core domain (residues 20-95) are undergoing exchange. Each residue was initially fitted to residue-specific exchange parameters, kex and pm. Residues in the core domain with a kex < 2000 s -1 were fitted to a global exchange process, resulting in kex = 1185 ± 91 s -1 and pM = 0.987 ± 0.002. 15 N CPMG relaxation dispersions for the SIMPXbound SUMO1 revealed 30 residues experiencing exchange. (Fig. 6B, E). Residues were selected for a global fit based on the same criteria as for apo-SUMO1, and were fitted to a global exchange rate kex = 479 ± 30 s -1 and major population pM = 0.942 ± 0.01 (Table 3). 15 N CPMG relaxation dispersion experiments on ZZ domain-bound SUMO1 showed 48 residues with significant exchange (Fig. 6C, F). Again, residues in the core domain of SUMO1 were selected for the global fit, giving a global exchange rate kex = 1418 ± 203 s -1 and major population pM = 0.978 ± 0.019 (Table 3). In comparison to apo-SUMO1, the ZZ domain bound SUMO1 has a larger number of residues in the flexible N-terminal exhibiting significant dispersions. Residues belonging to the N-terminal part (2-5, 7-8, 10, 13-16 and 18-19) were fitted to a global exchange rate kex= 1924 ± 145 s -1 and a major population pM = 0.990 ± 0.002.
Comparison of exchanging states in SUMO1 -Upon addition of SIMPX, most of the backbone residues in SUMO1 experiences a significant change in chemical environment as seen from a comparison of the crosspeak positions in the 15 N HSQC for SIMPX-bound and apo-SUMO1 (Fig.  7A). Residues not affected by the binding of SIMPX are located in the N-and C-terminal parts of SUMO1, whereas most backbone residues in the core region of SUMO1 experience a shift in chemical environment upon addition of SIMPX (Fig. 7B, C). However, only a subset of the residues affected by the binding of the SIM-peptide is within 5 Å of SIMPX, as gauged from the SIM-SUMO1 structure (Protein Data Bank code 2ASQ). This result indicates that the binding of SIMPX induces a conformational change in SUMO1. A closer look at the residues exhibiting conformational exchange in apo-SUMO1 reveals that these are primarily located in or close to the SIM-binding site. In summary, all residues within 5 Å of SIMPX, with the exception of residues 39, 43, 48, and 53, exhibit conformational exchange in apo-SUMO1. Upon addition of SIMPX to SUMO1, the exchange is quenched for many residues, but 20 residues show exchange in both apo-and SIMPX-bound SUMO1.Of these, 6 residues are within 5 Å of SIMPX. A small number of residues in the SIMPX-SUMO1 complex experiences chemical exchange not detected for apo-SUMO1, and of these, only one residue is within 5Å of the SIM binding site. The remainders of these residues are located in the flexible N-terminal, or in the hydrophobic core of SUMO1, suggesting a change in the hydrophobic packing upon binding. A comparison of residues that show exchange in the apo-and ZZ domain-bound states of SUMO1 reveals that these groups of residues are similar. In addition, a set of residues shows exchange in the ZZ domain-SUMO1 complex that is not observed in apo-SUMO.Most of these residues are located in the flexible N-terminal region, which is involved in binding the ZZ domain and therefore flexible in the apo state. Residues that show exchange for apo-SUMO1 but not for the ZZ domain-SUMO1 complex are mainly located in the region indicated to bind the ZZ domain based on the chemical shift perturbations observed for SUMO1. The 15 N chemical shift changes between the major and minor state was determined from global fits of the CPMG relaxation dispersions for the various states of SUMO1. The chemical shift changes extracted from the global fit of the apo-SUMO1 CPMG relaxation dispersion experiments does not show a direct one-to-one correspondence with the 15 N chemical shift changes observed between the SIMPX-bound and apo-SUMO1 (Fig. 8A) with an RMSD of 0.96 ppm. The data suggest that apo-SUMO1 is exchanging between a major ground state and a high-energy state that involves other conformations than those of the SIMPX-bound state. In contrast, the chemical shift change between the minor and major state in the SIMPX-SUMO1 complex correlates well with the 15 N chemical shift change measured from the 15 N HSQCs of the apoand SIMPX-bound SUMO1 with an RMSD of 0.08 ppm (Fig. 8B). These data imply that SUMO1 is exchanging between free and bound states in the SIMPX-SUMO1 sample.
Furthermore, the comparison of the globally fitted chemical shift changes from the CPMG relaxation dispersions between the minor and major state for apo-and ZZ domain-bound SUMO1 yields a good correlation with an RMSD of 0.14 ppm (Fig. 8C). Notably, the chemical exchange rates and the populations of ZZ domain-bound SUMO1 and apo-SUMO1 are the same within errors and with a similar set of residues showing exchange broadening in both states. These observations might suggest that the binding of the ZZ domain to SUMO1 has no major effect on the intrinsic conformational exchange of SUMO1. A direct comparison of the chemical shift change between the ZZ domain bound SUMO1 and apo-SUMO1 is not feasible due to the lack of 15 N cross peaks caused by exchange broadening upon formation of the ZZ domain-SUMO1 complex. An alternative interpretation of the data is that the CPMG dispersions observed for the ZZ domainbound SUMO1 sample involves exchange between free and bound states; in this scenario, the good match between the chemical shifts of the minor state of apo-SUMO1 and the ZZ domain-bound SUMO1 indicates that SUMO1 binds the ZZdomain by conformational selection. The latter interpretation is further supported by the expectation, based on the binding affinities from ITC and the NMR sample conditions, that both the SIMPX-and ZZ-domain SUMO1 samples contain a minor population of the peptide-free state, which is in good agreement with the minor state population determined from the CPMG dispersions. The on-and off-rates of the various states was calculated based on a global fit of the populations and exchange rates. For the SIMPX-SUMO1 complex, the koff is 28 ± 2 s -1 and kon is 1.6 ± 0.1 x10 6 M -1 s -1 . The calculated value of kon indicates that the binding of SIMPX to SUMO1 is essentially diffusion controlled. In comparison, the off-rate for the ZZ domain-SUMO1 complex is 31 ± 5 s -1 , as calculated for the core residues, while the results for the N-terminal residues directly involved in binding of the ZZ-domain peptide yield koff = 19 ± 2 s -1 . We speculate that the release of the ZZ-domain peptide might involve a two-step process, with slightly different life times of the interactions with the Nterminal segment and the core residues. The overall slower off-rate for the ZZ-domain is in keeping with the slightly higher affinity of the ZZ domain-SUMO1 complex than SIMPX-SUMO1. The offrates observed for the different bound states of SUMO1 are in the same regime as the intrinsic conformational exchange of apo-SUMO1, suggesting that these processes might be linked.

DISCUSSION
Protein modifications, such as SUMOylation, play important roles in intricate protein-protein interaction networks involving numerous intracellular processes such as the response to DNA-double strand breaks and chromatin remodeling. In the response to DNA damage, the SUMOylated target proteins are recognized by SUMO Interactions Motifs (SIMs) present in various DNA damage signal and repair proteins. Until recently, one type of SIM had been described that is characterized by a signature sequence involving a stretch of charged anionic amino-acid residues (10). The newly discovered SUMObinding ZZ domain represents a new SUMO recognition motif distinct from the classical SIMs (9). We now show that the ZZ domain interacts with a unique epitope in SUMO1 and that the complex between the ZZ domain and SUMO1 displays a different binding mode from those previously described for SUMO-mediated interactions. Mutation of residues in the interaction surface for the ZZ domain did not significantly decrease the affinity for SUMO1 even though the enthalpy of binding is reduced. From previous studies we have established that the zinccoordination is crucial for the interaction between the HERC2 ZZ domain and SUMO1 (9). By mutating the zinc-coordinating residues in the ZZ domain, the SUMO-binding capacity was lost. We therefore conclude that the presence of the coordinated zinc in the ZZ domain close to the interaction interface with SUMO1 play an important role in the ZZ domain-SUMO1 complex formation. Importantly, mutation of two epitopes in SUMO1, representing residues in the ZZ domain-SUMO1 interface, completely eliminates binding to the ZZ domain, clearly indicating the importance of these epitopes in ZZ domain recognition. The CPMG relaxation dispersion data demonstrate that apo-SUMO1 exchanges between the (major) ground state and a (minor) high-energy state, which might be linked to ligand binding. However, it is clear that the minor conformation of apo-SÚMO1 does not directly match that of the bound state in the case of SIMPX binding, indicating that this peptide does not bind by strict conformational selection, but induces additional conformational changes upon binding, or that it selects one conformer out of a manifold present in the high-energy state. Also in the case of the ZZ domain, the available data can be interpreted in two ways: either the peptide does not affect the intrinsic dynamics of apo-SUMO1 or it binds by conformational selection. Thus, the available data clearly show that the dynamic process of peptide binding to SUMO1 has markedly different signatures for the two cases of SIMPX and the ZZ domain. Taken together we have shown that the new SUMObinding ZZ domain binds to SUMO1 through interactions with a unique and spatially distinct site in SUMO1 as compared to the SIMs, and that these two kinds of recognition motifs can bind SUMO1 simultaneously in a non-cooperative fashion. Our data therefore provide new insights into the interactions between SUMO binding motifs and SUMO, and extend the repertoire of proteins recognizing SUMO to also include regulatory proteins harboring ZZ domains.     (49). SUMO1 is colored green, the ZZ domain is colored red, and the peptide corresponding to a SIM-motif colored in blue (Protein Data Bank code 2ASQ (6)). B. Model of the ZZ domain-SUMO1 complex shown in a ribbon representation using the same color scheme as in A with the two zinc ions depicted as blue spheres. The residues in SUMO1 and the ZZ domain affected by the interaction in the NMR epitope mapping experiments are indicated in the model, corresponding to the same residues shown for the individual protein models in Fig. 1 C,D. C. Per residue RMSD between interface residues and backbone Cα atoms in the modeled complex for SUMO1 residues 1-103. The location for the secondary structure elements of SUMO1 are indicated by arrows (yellow) for -strands and a cylinder for the single -helix (red). D. Per residue RMSD between interface residues and backbone Cα atoms in the modeled complex for the ZZ domain residues 1-53. The location for the secondary structure elements of the ZZ domain are indicated by arrows (yellow) for -strands and a cylinder for the helical segment (red). E. Experimental RDCs plotted versus calculated RDCs for the modeled protein complex. F. Orientations of the RDC alignment tensor for the ZZ domain and SUMO1 in the ZZ domain-SUMO1 complex, in which tensor orientations were fitted using Module (45). The ZZ domain (yellow) and SUMO1 (blue) are shown in ribbon representations.     Black corresponds to apo and red corresponds to the SIMPX-bound SUMO1. The affected residues are indicated by their respective residue number. B. Weighted 1 H, 15N chemical shift differences between the SIMPX-and the apo-state of SUMO1 plotted per backbone residue. The location for the secondary structure elements of SUMO1 are indicated by arrows (yellow) for -strands and a cylinder for the single -helix (red). C. SUMO1 is colored in green, residues with a significant weighted chemical shift differences (>0.05 PPM) are colored in red, where a subset is indicated by their respective residue number, while SIMPX is shown in blue (Protein Data Bank code 2ASQ).  15 N Chemical shift difference for backbone amides between major and minor state of apo-SUMO1 from global fits of CPMG relaxation dispersion curves plotted versus 15 N chemical shift difference between apo-and SIMPX-bound states. B. 15 N Chemical shift difference for backbone amides between the major and minor state of SIMPX-bound SUMO1 from global fits of CPMG relaxation dispersion curves plotted versus 15 N chemical shift difference between apo-and SIMPX-bound states. C. 15 N chemical shift differences for backbone amides between major and minor state from global fits of CPMG relaxation dispersion curves for apo-and ZZ domain-bound SUMO1 plotted in a covariance graph.

TABLE 1.
Summary of structural statistics for the structural refinement using RDCs of the ZZ domain and SUMO1, as well as structural statistics of the docked ZZ domain-SUMO1 complex. Structural analysis was performed using PSVS (29) and PALES (30). The magnitude and orientation of the RDC alignment tensor for the ZZ domain and SUMO1, and the ZZ domain-SUMO1 complex was calculated using Module (45). *AIR restraints indicate the ambiguous restraints used in the HADDOCK docking procedure. NMR