Solution structure of a yeast ubiquitin-like protein Smt3: The role of structurally less defined sequences in protein–protein recognitions

Abstract

Smt3 belongs to a growing family of ubiquitin-related proteins involved in posttranslational protein modification. Independent studies demonstrate an essential function of Smt3 in the regulation of nucleocytoplasmic transport, and suggest a role in cell-cycle regulation. Here we report the high-resolution NMR structure of yeast Smt3 in the complex free form. Our comparison of the Smt3 NMR structure with the Smt3 crystal structure in complex with the C-Terminal Ulp1 protease domain revealed large structural differences in the binding surface, which is also involved in the Smt3-Ubc-9 interaction detected by NMR. The structural differences in the region indicate the important functions of conserved residues in less structurally defined sequences.

Posttranslational modifications are important means to regulate activities and functions of many proteins in cells. The best characterized modification that involves covalent attachment of one protein to another is the ubiquitin conjugation system. Smt3 belongs to a growing family of ubiquitin-like proteins isolated from divergent eukaryotic organisms (Meluh et al. 1995; Saitoh et al. 1997^; Hochstrasser 2000). Smt3 was originally isolated as a high-copy suppressor of a mutation in MIF2, the gene of a centromere binding protein in S. cerevisiae (Frauke 2000). Smt3 regulates functions of many proteins through linking covalently to its targets. The single copy Smt3 gene is essential for S. cerevisiae viability and is shown to play roles in cell-cycle regulation and chromosome segregation (Jiang and Koltin 1996; Li and Hochstrasser 1999; Takahashi et al. 1999; Tanaka et al. 1999).

Previous structural studies of human Smt3 homolog Sumo-1 indicated that Sumo-1 and ubiquitin are highly homologous in their three-dimensional structures (Bayer et al. 1998; Jin et al. 2001). Previous biological studies showed that both proteins undergo similar pathways in the conjugation processes to their targets (Hochstrasser 2000). However, Sumo-1/Smt3 family and ubiquitin family members encode rather different fates for many of their shared target proteins. In order to control the fates of their targets precisely, Sumo-1/Smt3 and ubiquitin employ two different sets of enzymes in their conjugation processes. Since Sumo-1 and ubiquitin are structurally homologous, one critical question is how Sumo-1 family and ubiquitin family members are distinguished by their corresponding modification enzymes.

The X-ray structure of Ulp1, an E1 enzyme in the Smt3 modification pathway, in covalent link with Smt3 was solved recently (Mossessova and Lima 2000). The structure provides a starting point to understand how the modification enzymes recognize Smt3. The structure of the complex identifies the interaction surfaces between the two proteins. Many structural studies have indicated that protein-protein interactions induce changes in the structures of proteins. A comparison of structures of Smt3 complex-free and Smt3 in Ulp1 complex will provide important information on how the recognition is supported by structural changes in Smt3.

We present here the NMR structure of Smt3 in complex-free form and its interaction with the human Ubc9 protein. The structural comparison of Smt3 in free and Ulp1 complex forms indicates important structural changes in Smt3 as a result of the protein-protein interaction. The results also show that Ubc9 and Ulp1 partially share recognition surfaces on Smt3. The results indicate that the structure of Smt3 has to adjust differently when Smt3 interacts with different modification enzymes.

Results

Resonance assignment

A¹H-¹⁵N HSQC spectrum of ¹⁵N-labeled Smt3 at pH 6.5 and 27°C with corresponding residue assignments indicated is shown in Figure 1 ▶. In this spectrum, 92 expected correlation peaks for the backbone amide resonances were assigned. The spectrum also shows several additional peaks from asparagine, glutamine, and arginine side chain resonances (peaks connected by solid line, unassigned). Backbone assignments were made primarily on the basis of CBCA(CO)NH and HNCA spectra, with the use of HBHA(CO)NH and TOCSY-¹⁵N-HSQC spectra to resolve ambiguities. Sequential assignments were further verified with the analysis of ¹⁵N NOESY-HSQC spectra. Aliphatic side chains were assigned using TOCSY-¹⁵N-HSQC, HCCH-COSY, and HCCH-TOCSY data. Aromatic side chain assignments were deduced from 2D ¹H-NOESY, and ¹³C-HSQC.

Fig. 1.

Two-dimensional ¹H-¹⁵N HSQC spectrum at 600 MHz of Smt3 obtained at 20°C and pH 6.5. Backbone amide resonances are labeled with the one-letter amino acid code and residue number.

Secondary structure

The pattern of short and medium-range NOEs and ³J_HNHα coupling constants (Fig. 2 ▶) served to identify the secondary structural elements of Smt3. The pattern of d_NN(i, i+2), d_αN(i, i+3) and d_αN(i, i+4) NOE connectivities with ³J_HNHα coupling constants smaller than 6 Hz indicated the presence of two α-helices (Leu45-Gln56 α1 and Thr77-Asp82 α2). Characteristic long-range backbone NOE connectivities (i-j > 4) and large coupling constants (>9.0 Hz) were used to define β-strands. Based on this standard, five β-strands (Ile24-Lys30 β1, Glu34-Lys 40 β2, Leu63-Tyr 67 β3, Ile70-Ile72 β4, and Asp87-Glu94 β5) were identified with β1 to β2, β3 to β4, and β4 to β5 being antiparallel and β1 to β5 being parallel. Similar to Sumo-1, the data indicated that the first 23 residues at the N-terminus were disordered. The location and the extent of the secondary structural elements were further supported by the analysis of ¹³C^α, and ¹H^α chemical shift index (CSI) method (Wishart et al. 1995a; Wishart and Sykes 1994). The consensus results of the chemical shift index of ¹³C^α, and ¹H^α in Smt-3 (Fig. 3 ▶) were in good agreement with the secondary structural elements derived from other data.

Fig. 2.

Summary of sequential and medium-range NOE patterns and ³J_Hα-NH coupling constants for Smt3. The relative intensity of sequential NOEs is indicated by the height of the bars, and horizontal lines indicate the observation of medium-range NOEs between residue pairs. Also indicated are ³J_Hα-NH values of < 6 Hz (□) and > 8.0 Hz (▪). Secondary structural elements are indicated as arrows (β-strand) and rectangles (α-helix).

Fig. 3.

Chemical shift index derived from ¹H^α and ¹³C^α chemical shifts of Smt3. An index of either −1 or +1 indicates a shift deviation from the random coil value of either greater than 0.1 ppm (¹H^α) to lower frequency or 0.7 ppm (¹³C^α) to higher frequency. Chemical shifts within the variance of the random coil values defined above are indicated by 0.

Tertiary structure of Smt3

The 3D structure was calculated using the software program DYANA 1.5. A total of 987 interproton restraints and 133 dihedral angle constraints were used in the final round of the structural calculation. Due to the disordered nature, the N-terminal 23 residues were excluded in the structural calculation. Distributions of constraints by range and by sequence position are summarized in Table 1 and Figure 4A ▶. At the final stage, 100 conformers were calculated; of these, 20 DYANA conformers with the lowest target function were selected to represent the structure (Fig. 5A ▶). The coordinates have been deposited in the RCSB Protein Data Bank with accession code 1L2N. For the defined region of Smt3 (Glu21-Ile96), the RMSDs to the mean structure were 0.37 Å for backbone atoms and 1.00 Å for all nonhydrogen atoms. The RMSD values for each residue are plotted in Figure 4B ▶. In the calculated structures, 67.8% of the residues were in the most favored region on Ramachandran plot using the software program PROCHECK (Laskowski et al. 1996), 31.0% in the additionally allowed region, 1.2% in the generously allowed region, and 0% in the disallowed region. The structural statistics are summarized in Table 1.

Table 1.

Structural statistics for the best 20 DYANA Smt3 conformers

(A) NMR-derived restraints
Total interproton restraints	1146
Intraresidue (|i-j|=0)	246
Sequential (|i-j|=1)	285
Medium-range (l < |i-j| <5)	190
Long-range (|i-j|>4)	266
Hydrogen bonds	26
Dihedral angles (ϕ,ξ)	133
(B) Residual violationsa
DYANA target function (Å2)	1.64 ± 0.18
Upper limit
Sum (Å)	6.4 ± 0.6
Maximum (Å)	0.37 ± 0.06
van der Waals
Sum (Å)	6.5 ± 0.4
Maximum (Å)	0.24 ± 0.05
(C) Average RMSD to mean structure (Å)
Backbone atoms N, Cα, C` (Å)	0.53 ± 0.12
All heavy atoms (Å)	1.41 ± 0.19
(D) Ramachandran plot (% residues)
Residues in most favored regions	67.8
Residues in additional allowed regions	31.0
Residues in generous allowed regions	1.2
Residues in disallowed regions	0

^a In part B, values are mean (standard deviation).

Fig. 4.

(A) Number of NOE restraints per residue for NMR structure of Smt3. For each residue, the numbers of intraresidue, sequential, medium-range, and long-range NOEs are shown with white, light gray, dark gray, and black bars, respectively. (B) Plots of backbone heavy atom (solid line) and all nonhydrogen atom (dotted) average RMSD values for each residue for the family of 20 structures relative to the mean structure.

Fig. 5.

NMR structure of Smt3. (A) Stereo superposition of 20 selected conformers with the lowest target functions from the final DYANA calculations. (B) A ribbon diagram showing the NMR-derived tertiary structure used in this study. The average structure was generated and analyzed by MolMol.

The structure of Smt3 consists of two α-helices and one β-sheet. The β-sheet consists of three antiparallel β-strands and one parallel β-strand (see the ribbon diagram in Fig. 5B ▶). A short sequence (Ile70 to Ile72), which was identified as β4 in the secondary structure analysis on the basis of NOE connectivities and J-coupling constants, was not detected as a β-strand by the program MolMol (Koradi et al. 1996). This putative β-strand tilts away from the plane defined by the other four β-strands. Therefore, this three-residue sequence may not form an ideal β-strand in the current solution environment. Overall, the fold of Smt3 is highly similar to those of Sumo-1 and ubiquitin. Helix α1 (Leu45 to Gln56) is rotated approximately 45° relative to the first two β-strands (β1 and β2), and this arrangement represents the typical ubiquitin-like conformation. Like Sumo-1, α1 helix is strongly amphipathic in nature, with hydrophobic residues pointing inward and hydrophilic residues pointing into the solvent. Relevant contacts at the helix-sheet interface in Smt3 occur between hydrophobic side-chain Leu26, Ile35, Phe37, Phe65, Ile89 of the β-strands, and Leu45, Leu48, Phe52, Lys54, and Gln56, of the helix α1. These hydrophobic residues are conserved in ubiquitin-like proteins and appear to be essential to maintain the ubiquitin superfold. Helix α2 of Smt3 consists of approximately six residues. Long-range NOEs were observed from the N-terminal residues Pro44 and Leu45 of α1 to the N-terminus of α2 (Thr77) and the preceding residue Gln76, making both helices into a conformation perpendicular to each other, which is also a typical ubiquitin fold.

Human Ubc-9-Smt3 interaction

Smt3 and ubiquitin homologs are both highly sequence-conserved within their subfamilies. However, the proteins from the two subfamilies show obvious differences on the surface residues. The conserved surface residues on each subfamily may provide distinct recognition markers within their families. A previous study indicated that human Sumo-1 and its E2 modification enzyme Ubc-9 interacted in solution under NMR condition (Liu et al. 1999). An HSQC was run on ¹⁵N-enriched Smt3 in the presence of regular Ubc-9. Compared to the HSQC of Smt3 in solution alone, the resonances of many residues of Smt3 in the presence of Ubc-9 are also severely perturbed (Fig. 6 ▶). These residues are mainly located in the loop (G31, S32, S33) between strand 1 and strand 2, β-strands 4 and 5, and the residues in between (from L63 to H92). These perturbed residues show a high degree of homology between Sumo-1 and Smt3, even the residues on the less defined loops. A ¹⁵N edit NOESY-HSQC experiment performed on partially deuterated Smt3 in the presence of regular UBC-9 failed to collect sufficient NOE signals to assign the perturbed resonances and to detect intermolecular NOE signals between the two proteins. Therefore, the interaction is relatively weak, and the complex and the free proteins are in the mode of fast-exchange on the time scale of miniseconds under the current NMR condition.

Fig. 6.

Superposition of ¹H-¹⁵N HSQC spectra of Smt3, free and in complex with Ubc9. The Smt3/Ubc9 ratio in the complex is approximately 1:1. The cross-peaks of free Smt3 are shown in red and those of Smt3 in the complex are shown in black. Only peaks that were affected by the complex formation are labeled.

Discussion

Both Smt3 and Sumo-1 belong to a subfamily of ubiquitin-like proteins. Smt3 is the yeast functional and structural homolog of human Sumo-1. The two proteins show more than 50% amino acid sequence identity. The three-dimensional structures of Sumo-1 were determined by homonuclear NMR (Bayer et al. 1998) and heteronuclear NMR spectroscopy (Jin et al. 2001). A schematic representation of the C^α atom coordinates of NMR structure of Smt3 overlaid with the structure of Sumo-1 from the heteronuclear NMR data is shown in Figure 7A ▶. Apparently the two structures are highly homologous. The pair-wise backbone (residue Thr22-Gln95) C^α RMSD is 3.20 Å between the average NMR structure of Smt3 and the average structure of Sumo-1.

Fig. 7.

A stereo diagram of the backbone C^α atom coordinates of NMR solution structure of Smt3 (current structure, red) overlaid with (A) that of SUMO-1 structure (blue) obtained from heteronuclear NMR. (B) That of the X-ray crystal structure of Smt3 (green) in complex with the C-Terminal Ulp1 protease domain.

Since yeast Smt3 also interacts with human Ubc-9, it is important to compare the structures of the two surfaces involved in Ubc-9 interaction. The binding interactions between Ubc-9 and Sumo-1 caused chemical shift perturbation in the binding interfaces of both Ubc-9 and Sumo-1. The data have indicated that the interface on Sumo-1 is defined by residues 26, 32, 64–71, and 81–91 (Liu et al. 1999) containing β strands 1, 3, 4, 5, helix α2, and the sequence between β3 and β5 (Fig. 8A ▶). Chemical shift perturbation was also used to map the Ubc-9 binding interface on Smt3. Residues 26, 28, 31–33, 37, 63–72, 77, and 84–92 have the most significant changes in chemical shifts upon complex formation with Ubc-9 (Fig. 8A ▶). These residues are also located in β strands 1, 3, 5, and the loop between β1 and β2 and the sequence between β3 and β5. Many of the perturbed residues are highly conserved in Smt3 subfamily members. Among the surface residues that are affected by complex formation, two negatively charged residues, Glu84 and Glu90 and a positively charged residue, Arg71 are identical between Smt3 and Sumo-1 and among many other Smt3 family members. In addition, three negatively charged surface residues, Asp68, Glu85, and Asp87, and several exposed hydrophobic residues, that is, Leu66, Phe67, Met83, and Val88, are either highly conserved or moderately conserved in the family members. Compared to the Sumo-1-Ubc-9 interface, Smt3 shows a similar binding surface in the Smt3-Ubc-9 complex. This surface is located on the hydrophilic side of the β-sheet and is involved in the loop between β-strands 1 and 2 and the sequence between β-strand 3 and β-strand 5. If the β-sheet and helix α1 of Smt3 and Sumo-1 are aligned, the remaining sequences between the two proteins show large RMSD differences (Fig. 8B ▶), an indication that these sequences may have to adjust conformations in the Ubc-9 complexes.

Fig. 8.

(A) Alignment of the sequences of Sumo-1 and Smt3. Shown are residues with large chemical shift changes in Sumo-1 (indicated above the alignment by a vertical line topped by #), Smt3 (vertical line topped by * and both (vertical line) when they contact Ubc-9. Ulp-1 contact residues in Smt3 are indicated with ↑ on the bottom of the alignment. Plots of backbone C^α RMSD values for each residue when the core secondary structures are aligned between NMR structures of Smt3 (current structure) and Sumo-1 (B), and between the NMR structure of Smt3 (current structure) and the crystal structure of Smt3 in the Ulp-1 complex (C).

Our structural comparison between Smt3 in the Ulp1 complex and in the complex-free form also indicates large structural perturbations in the sequences that participated in the Ulp1 interaction. A schematic representation of the C^α atom coordinates of the NMR structure of Smt3 overlaid with the X-ray crystal structure of Smt3 in complex with the C-terminal Ulp1 protease domain is shown in Figure 7B ▶. It is obvious that the residues from 60 to 74 and the C-terminus of the protein are perturbed dramatically after the alignment of the core secondary structures of two Smt3 structures (Fig. 8C ▶). One obvious difference is that strand 5 in Smt3 is longer in the complex than in the free protein. The Ulp1-Smt3 interaction apparently also modifies the local structure of Smt3. Many of these Ulp1 recognition residues reside in less defined conformations, and are able to adopt alternative conformations. Furthermore, many of these residues are involved in both Ubc-9 and Ulp1 interactions. Since Ubc-9 and Ulp1 represent two rather different structural motifs, it is likely that the less structured loop sequences in Smt3 can adopt different conformations in response to the divergent modification enzymes.

Our data indicate that the Ulp1 interaction surface and the Ubc-9 interaction surface of Smt3 are partially overlapped. Before Smt3 can interact with Ubc-9, the noncovalent interaction between Smt3 and Ulp1 has to be abolished. In the Smt3 conjugation process, the covalent bond between Ulp1 and Smt3 must be disrupted to form a new covalent bond between Ubc-9 and Smt3. It is not clear whether this bond transfer requires noncovalent interaction between Smt3 and Ulp1 to break first and then to form Ubc-9-Smt3 noncovalent complex before the covalent bond transfer can take place, or the covalent bond transfer goes in prior to the swap of the noncovalent interactions. To clarify the transfer process, a systematic study of binding constants among Smt3, Ulp1, and UBC-9 in both the covalent linked and noncovalently linked forms is necessary. Nevertheless, the affinity between Smt3 and Ubc-9 apparently helps the dissociation of the Smt3-Ulp1 complex. It is obvious that highly conserved surface residues in Sumo-1/Smt3 family members play important roles in the correct recognition of their modification enzymes.

Materials and methods

Expression and purification of yeast Smt3

The gene encoding Smt3 protein was a generous gift from Dr. M. Hochstrasser (Department of Biochemistry and Molecular Biology, University of Chicago, Chicago, IL). The protein was produced in E. coli BL21 (DE3) (Novagen, Madison, WI) to take advantage of the cell strain's protease deficiency. Uniformly labeled Smt3 was obtained by growing cells in isotopically enriched minimal media containing 0.6% Na₂PO₄, 0.3% KH₂PO₄, 0.15% NaCl, 1mM MgCl₂, 0.1mM CaCl₂, and 0.5% Basal Medium Eagle Vitamin Solution (Gibco BRL, Gaithersburg, MD). (¹⁵NH₄)₂SO₄(1g/ L) and unlabeled glucose(4g/L) were used for ¹⁵N-labeling or 1.5g/L (¹⁵NH₄)₂SO₄ and 2 g/L of ¹³C-enriched glucose for double labeling in H₂O. The multiple histidine-tagged protein used in this study was purified under a standard procedure using Ni-NTA resin under denaturing conditions to better remove proteases (QIAGEN manual 1999, third ed.). Finally Smt3 was renatured by dialysis against a phosphate buffer (50 mM Na₂HPO₄, pH 6.5, and 100 mM NaCl). In the NMR studies, protein concentrations were 0.7 and 1 mM as measured by the BioRad protein assay using bovine serum albumin as the standard (BioRad Laboratories, Hercules, CA).

NMR spectroscopy and data processing

All NMR measurements were performed in either 10% D₂O/90% H₂O or 100% D₂O at 27°C on a Bruker DRX600. Standard homonuclear 2D and heteronuclear 2D and 3D experiments were acquired for the backbone and sidechain and NOE constraint assignments and J³ J-coupling constants (Cavanagh et al. 1996). Data were processed and analyzed using the commercial software program SYBYL (Tripos, MO). All ¹H dimensions were referenced to internal 2,2-dimethyl-2-silapentane-5-sulfonate (DSS), and ¹³C and ¹⁵N were indirectly referenced to DSS (Wishart et al. 1995b).

The sequence-specific backbone resonance assignment of 98 amino acid residues of Smt3 was achieved through a combination of NOESY-¹⁵N-HSQC (Fesik and Zuiderweg 1990), TOCSY-¹⁵N-HSQC (Cavanagh and Rance 1992), HNCA (Yamazaki et al. 1994), HBHA(CO)NH (Grzesiek and Bax 1993), and CBCA(CO)NH (Grzesiek and Bax 1992) spectra by matching intra- and interresidue ¹³C^α, ¹H^α and/or ¹³C^β, ¹H^β chemical shifts for a given residue (associated with a ¹H-¹⁵N correlation peak) and the previous residue in the primary sequence. Complete backbone assignments were obtained except for M1. The remaining side chain resonances were assigned using a combination of HCCH-COSY (Ikura et al. 1991) and HCCH-TOCSY (Kay et al. 1993). Side chain resonances of residues from Lys 11 to Pro 20 were unassigned owing to highly overlapped cross peaks. The ring resonances of the His23, Phe36, Phe37, Phe65, and His92 were assigned by analyzing 2D ¹H-NOESY (Kumar et al. 1980), ¹³C-HSQC (Vuister and Bax 1992), and 3D NOESY-¹³C-HSQC (Palmer et al. 1991) spectra in D₂O on the basis of NOEs to the ¹H^α and ¹H^β resonances of those amino acids. More than 85% of all side chain resonances were assigned for Smt3.

Derivation of structural restraints and structural calculation

NMR distance restraints were collected from three different NOESY spectra: 3D NOESY-¹⁵N-HSQC (mixing time 150msec) for amide protons, 3D NOESY-¹³C-HSQC in D₂O (mixing time 100msec) for aliphatic protons, and 2D ¹H-NOESY in D₂O, (mixing time 100msec) for aromatic protons. ³J_HNHα coupling was measured in a 3D HNHA (Vuister and Bax 1993) spectrum. NOE restraints were grouped into four distance ranges: strong, 1.8–2.8 Å; medium, 1.8–4.0 Å; weak, 1.8–5.0 Å; and very weak, 1.8–6.0 Å for backbone-backbone NOE correlations, while an additional 0.5 Å was added to the upper bound of the NOE restraints for backbone-side chain and side chain-side chain NOEs (Clore et al. 1987). Pseudoatom corrections were applied for distances involving methyl protons, aromatic ring protons, and nonstereospecifically assigned methylene protons (Wüthrich et al. 1983). Hydrogen bond restraints were employed in areas of regular secondary structures, displaying characteristic NOE cross-peaks. Each deduced hydrogen bond was represented by two distance constraints: 1.7–2.4 Å for H^N-O and 2.7–3.4 Å for N-O. Dihedral angle restraints were derived from ³J_HNHα coupling values obtained from a 3D HNHA experiment and were included in the refinement protocol. The ϕ angles were restrained to −60 ± 30⁰ for ³J_HNHα < 6.0 Hz and d_NN > d_αN, −120 ± 50⁰ for ³J_HNHα = 8.0–9.0 Hz, and −120 ± 30⁰ for ³J_HNHα > 6.0 Hz. Further restraints for ϕ and ξ were added on the basis of the consensus chemical shift index (Wishart et al. 1995; Wishart and Sykes 1994) and NOEs patterns characteristic of secondary structure: helical residues, −60 ± 40⁰ (ϕ) and −50 ± 50⁰ (ξ); β-strand residues, −120 ± 40⁰ (ϕ) and −130 ± 50⁰ (ξ).

Structure calculations were performed with the program DYANA 1.5 (Güntert et al. 1997) using a 40,000-step energy minimization procedure. For the initial rounds of structure calculations, only sequential, intraresidual, medium-range NOEs and unambiguous long-range NOEs, and coupling constants were used. Later, all other long-range NOEs and hydrogen bonds were introduced in consecutive steps. One hundred structures were calculated in each round, and of these, the 20 structures with the lowest target functions were used to analyze restraints violation and to assign additional NOE restraints for the following round. This process was repeated until more than 90% NOE peaks in the spectra had been assigned and all violations were eliminated. In the final stage, the 20 structures with the lowest target functions were used for the structural analyses. All subsequent analyses of the structure and graphic representations of the three-dimensional structures were performed using MOLMOL (Koradi et al. 1996) and PROCHECK-NMR (Laskowski et al. 1996).

Abbreviations

• Smt3, yeast ubiquitin-like protein

• Sumo, small ubiquitin-related modifier

• 2D and 3D, two- and three-dimensional

• NMR, nuclear magnetic resonance

• HSQC, heteronuclear single quantum correlation

• NOE, nuclear Overhauser effect

• NOESY, nuclear Overhauser effect spectroscopy

• TOCSY, total correlated spectroscopy

• RMSD, root-mean-square deviation