Structural and functional insights into the roles of the Mms21 subunit of the Smc5/6 complex

Summary

The Smc5/6 complex is an evolutionarily conserved chromosomal ATPase required for cell growth and DNA repair. Its Mms21 subunit supports both functions by docking to the arm region of Smc5 and providing SUMO ligase activity. Here we report the crystal structure of Mms21 in complex with the Smc5 arm. Our structure revealed two distinct structural and functional domains of the Smc5-bound Mms21: its N-terminal half is dedicated to Smc5 binding by forming a helix bundle with a coiled-coil structure of Smc5; its C-terminal half includes the SUMO ligase domain, which adopts a new type of RING E3 structure. Mutagenesis and structural analyses showed that the Mms21-Smc5 interface is required for cell growth and resistance to DNA damage, while the unique Mms21 RING domain confers specificity to the SUMO E2–E3 interaction. Through structure-based dissection of Mms21 functions, our studies establish a framework for understanding its roles in the Smc5/6 complex.

Introduction

The evolutionarily conserved structural maintenance of chromosome (SMC) proteins are chromosomal ATPases important for many aspects of chromosomal organization and dynamics (Nasmyth and Haering, 2005; Hirano, 2006; Onn et al., 2008). They are modular proteins containing head, arm, and hinge regions (Fig. 1A). The head region is composed of the N- and C-terminal globular domains of the protein and possesses ATPase activity. The arm region is predicted to form extended coiled-coil structures comprised of two intertwined helixes. Between the two helixes is a hinge region that mediates the interactions between two SMC proteins. The structures of the head and hinge regions, but not the arm region, have been determined (Lowe et al., 2001; Haering et al., 2002; Haering et al., 2004; Lammens et al., 2004; Woo et al., 2009). In eukaryotic cells, six SMC proteins function as three pairs and each pair binds to a unique set of proteins to form a multi-subunit complex. These complexes include cohesin, condensin, and the Smc5/6 complex (Nasmyth and Haering, 2005; Hirano, 2006; Onn et al., 2008). Cohesin is composed of Smc1/3 and two non-SMC proteins, whereas condensin contains Smc2/4 and three non-SMC proteins. The Smc5/6 complex is the most complex with eight subunits: Smc5/Smc6 and six non-SMC elements, namely Nse1, Mms21 (Nse2), and Nse3–6 (Zhao and Blobel, 2005; Sergeant et al., 2005; Pebernard et al., 2006; Taylor et al., 2008).

Figure 1. The overall structure of Mms21 in complex with the arm region of Smc5.

(A) Schematic representation of SMC structure using Smc5 as an example. The five domains of Smc5 are depicted on the top with the coiled-coil domain present in the crystal structure marked gray. The hypothetical structure of Smc5 is illustrated at the bottom.

(B) Overall view of the Mms21-Smc5 structure. The coiled-coil regions of Smc5 are shown in blue and are composed of two helixes, H1 and H2. The corresponding amino acids at the ends of each helix are indicated. Mms21 is shown in gold and the zinc ion as a green sphere. The N- and C-termini of Mms21 are indicated.

(C) The arm region of Smc5 exhibits a moderate curvature. The Cα structural representation of the arm region of Smc5 that binds to Mms21 is shown with annotation of its length and bending.

These three SMC complexes are all essential for cell growth and play distinct roles in chromosomal metabolism. Cohesin tethers sister chromatids to achieve chromatid cohesion whereas condensin tethers different regions of the same chromatid to achieve chromosome compaction. The Smc5/6 complex is implicated in recombinational repair and in the maintenance of complex genomic loci such as rDNA and telomeres, though its precise roles in these processes are not well understood (Murray and Carr, 2008; De Piccoli et al., 2009). The different functions performed by the three SMC complexes likely stem from their divergent subunit compositions and structures. Recent studies revealed several differences between the Smc5/6 complex and cohesin or condensin. The most unique feature of the Smc5/6 complex is that its Mms21 subunit possesses SUMO (small ubiquitin like modifier) ligase (or E3) activity and binds to the arm region of Smc5 (Zhao and Blobel, 2005; Andrews et al., 2005; Potts and Yu, 2005; Sergeant et al., 2005; Duan et al., 2009). This is in contrast with cohesin and condensin, in which all the non-SMC subunits bind to the head regions of the SMC proteins and do not possess enzymatic activity. How this unique subunit functions within the Smc5/6 complex is the focus of this study.

The sumoylation function of Mms21 contributes to the roles of the Smc5/6 complex in telomere and rDNA maintenance as well as in recombinational repair, with some relevant targets having been identified (Zhao and Blobel, 2005; Ampatzidou et al., 2006; Branzei et al., 2006; Potts and Yu, 2007; Pebernard et al., 2008). The SUMO E3 activity of Mms21 relies on an SP-like (SPL) RING domain (Zhao and Blobel, 2005; Andrews et al., 2005; Potts and Yu, 2005), the sequence of which is reminiscent of the SP-RING (Siz/PIAS-RING) domain of another class of SUMO E3, the PIAS (protein inhibitors of activated STAT) family (Hochstrasser, 2001). Both SPL- and SP-RING domains exhibit moderate sequence homologies to that of the RING domain found in ubiquitin E3s. The SUMO E3 RING domains, like those of ubiquitin E3s, are believed to bind the corresponding conjugating (or E2) enzymes (Geiss-Friedlander and Melchior, 2007). While the structures of the ubiquitin E3 RING are known (Eisenhaber et al., 2007), that of the SUMO E3 RING (both SP- and SPL-RING) remain unknown. It is thus unclear whether SUMO E3 RING domains form structures similar to ubiquitin RING domains; it is also unclear how they are able to specifically recognize SUMO but not ubiquitin E2s. Such specificity bears important consequences, as sumoylation and ubiquitination affect target proteins very differently. For example, ubiquitination often marks protein for degradation, whereas sumoylation alters protein interactions, localizations and activities (Geiss-Friedlander and Melchior, 2007). Structural studies of SP- and SPL-RING domains are crucial for revealing the molecular mechanisms by which they can contribute to sumoylation but not ubiquitination.

Though important for some functions of the Smc5/6 complex, the E3 activity of Mms21 is non-essential, because mutants lacking this domain grow at a reasonable rate (Zhao and Blobel, 2005; Andrews et al., 2005). In contrast, lacking the entire Mms21 protein, like lacking any other subunit of the Smc5/6 complex, is lethal (Zhao and Blobel, 2005; Andrews et al., 2005; Sergeant et al., 2005). Thus, Mms21 must possess additional functions that render it essential for cell growth. Currently, these functions are not known. Considering its unique position on the arm region of Smc5, it is possible that Mms21 carries out these additional functions via this interaction.

In this study, we seek to understand how Mms21 achieves its multiple roles within the Smc5/6 complex using a combination of structural, biochemical and genetic approaches. First, we solved the atomic structure of the budding yeast Mms21 in complex with the associated Smc5 arm region. This structure revealed that the C-terminal SUMO E3 domain of Mms21 forms a new type of RING structure that is different from the RING or U-box structure of ubiquitin E3s. Based on this structure, we identified two properties that allow its interaction with SUMO, but not ubiquitin, E2s. In addition, we found that the N-terminal half of Mms21 is devoted to the interaction with the arm region of Smc5. Structure-guided mutagenesis studies showed that disruption of this interaction led to cell death and DNA repair defects. These results provide new mechanistic insights into the functions of the SPL-RING domain of SUMO E3s and also into the additional roles of Mms21 in the Smc5/6 complex.

Results

The region of Smc5 associated with Mms21 forms a moderately bent coiled-coil structure

To gain mechanistic insights into the functions of Mms21, we determined its crystal structure in conjunction with the associated Smc5 regions. Using limited protease treatment, we defined the Mms21-interacting region to a portion of the arm region of Smc5 that is comprised of the following two sequences: Asp302 to Leu369 and Lys733 to Gln813 (Fig. 1A) and (X.D., X.L., X.Z., H.Y., unpublished data). These two regions of Smc5 were linked by a four-amino-acid linker peptide (Gly-Ser-Gly-Ser) and the resulting chimeric protein formed a stable complex with the full-length Mms21. The phase of the structure of this complex was solved at 3.9Å by the Se-Met single-wavelength anomalous dispersion (SAD) method. Refinement against the native data was achieved at a resolution of 2.3Å (Table 1).

Table 1.

Crystallographic data statistics

Data Statistics	Native	SeMet
Source	APS 19BM	APS 24ID
Wavelength (Å)	0.97899	0.97916
Resolution (Å)	50-2.3	50-3.9
Space group	C2221	C2221
Unit cell (Å) a, b, c	47.465, 91.574, 249.215	47.138, 91.566, 250.583
Number of observations	105,443	29,652
Unique reflections	21,519	4,942
Redundancy	4.9 (1.9)	6.0 (3.9)
Completeness (%)	88.0 (77.3)	93.4 (85.6)
Mean I/σI	19.1 (1.6)	32.8 (5.6)
Rmerge	0.09 (0.41)	0.07 (0.19)
Refinement Statistics
Resolution Limits (Å)	50.0-2.3
Number of reflections (working/test)	19,067/1,036
Rwork/Rfree	25.6% / 27.0%
Average B factor	59.0
Rms deviations
Bond lengths (Å)	0.007
Bond angles (°)	1.266
Number of atoms
Protein	water	Zn
3105	117	1
Ramachandran plot
Most favored	Additionally allowed	Generously allowed	Disallowed region
85.9%	13.8%	0.3%	0.0%

^*Numbers in parentheses are for the high-resolution bin.

Rsym = |I-<I>/I, where I is the intensity of an individual measurement and <I> is the average intensity from multiple observations.

Rwork = |Fobs|-k|Fobs|.

Rfree = Rwork against 5% of the data removed prior to refinement.

In the structure of the Mms21-Smc5 complex, the two segments of Smc5 are present as two long anti-parallel helixes (H1 and H2) that wrap together for half a round forming a coiled-coil structure (Fig. 1B). The well-refined region of the H1 helix contains 60 residues (Lys304 to Arg363) while that of H2 contains 73 residues (Asp739 to Lys811). A few residues at each end of the helixes as well as the linker between the two are not visible in the electron density map.

The core interface between the two helixes is predominantly hydrophobic. The well-complemented region of these two helixes extends for about 80Å. As the entire arm region of Smc5 is predicted to contain residues 129 to 495 and residues 660 to 944 (see methods), which is about six times as long as the coiled-coil region determined in the structure, the length of the Smc5 arm is estimated to be approximately 480Å (48 nm). This prediction is consistent with the 50 nm cohesin and condensin fibers observed in electron microscope studies (Haering et al., 2002; Anderson et al., 2002). Note that the lengths of the SMC proteins should be largely determined by their coiled-coil regions, as the head and hinge regions are globular structures with diameters of no more than a few nanometers (Haering et al., 2002; Haering et al., 2004). Another interesting feature of the Smc5 coiled-coil region is that it exhibits a moderate curvature of 165° (Fig. 1C). In summary, this Smc5 structure provides the first glimpse into the arm regions of SMC proteins, and the bending may be an important determinant in shaping the overall structure of the Smc5/6 complex.

The N-terminal region of Mms21 forms extensive interactions with Smc5

Mms21 in this complex forms a bipartite structure. Its N-terminal domain (NTD) contributes to Smc5 binding, whereas its C-terminal domain (CTD) contains a variant RING structure and has no contact with Smc5 (Fig. 1B). The detailed structure of the Mms21 CTD will be discussed later in the text. The core of the Mms21 NTD is composed of two elongated helixes (Pro16 to Ser52 and Glu60 to Ala100) wrapped around the coiled-coil regions of Smc5 (Fig. 1B and Fig 2A). The first helix is divided into two short alpha helixes (α1 and α2) and a six-amino acid non-typical helix stretch (T2; Fig. 2A); the second helix (α3) displays a standard helical structure.

Figure 2. The structure of Mms21 and the interface between Mms21 and Smc5.

(A) The secondary structure of budding yeast Mms21 and the alignment of Mms21 orthologs from different species. Structure-based sequence alignment of Mms21 orthologs from S. cerevisiae (sc), S. pombe (sp), Zebrafish, Xenopus, mouse, rat, and human are shown. Conserved residues are indicated in red and identical residues are highlighted. The amino-acid numbers and secondary structures of the budding yeast Mms21 are displayed above the sequence.

(B)–(E) Electrostatic surface representations of Smc5 and Mms21 illustrate charge and shape complementarities. The electrostatic surface representations of the arm region of Smc5 (B) and Mms21 (D) are shown. In addition, each representation is shown with the ribbon image of its binding partner (C and E). (D) and (E) are viewed from the sides opposite those in (B) and (C).

The interface between the Mms21 NTD and Smc5 is extensive, resulting in the burial of 2360Ų of surface area from Mms21 and 2426Ų from Smc5 (Figs. 2B–2E). As a result, the NTD of Mms21 and the coiled-coil region of Smc5 form a well-structured, stable helix bundle. Nineteen hydrogen bonds and four salt bridges contribute greatly to the interaction between these two proteins (Figs. 2B–2E). Moreover, most of the hydrophobic residues are located in the interior of the four-helix bundle, thereby providing additional van der Waals interactions (Figs. 2B–2E).

We also examined the thermodynamic properties of the Mms21-Smc5 interaction. Isothermal titration calorimetry (ITC) measurements revealed relatively large negative enthalpies (between −18.21 and −22.0 kcal/mol) and positive entropies (between 10.81 and 15.20 kcal/mol) for this interaction at three temperatures (10°C, 15°C and 20°C, Table S1). The favorable enthalpies indicate that this interaction is energetically driven by exothermic enthalpy and that the hydrophilic interactions, including hydrogen bonds and salt bridges, contribute to complex formation. The unfavorable entropies suggest a potential conformational change during complex formation. In addition, the enthalpy of the interaction displayed linear temperature dependence with a heat capacity change of −307 cal/mol·K (Table S1 and Supp Fig. 1), suggesting significant hydrophobic interactions within the complex. Taken together, the thermodynamic properties of the Mms21-Smc5 interaction are consistent with the structural results and confirm the contributions from hydrogen bonds, salt bridges, and hydrophobic interactions.

Four segments of the N-terminal region of Mms21 are required for interaction with Smc5

To understand the functions of the Mms21 NTD, we identified four segments in the region that make the largest contributions to Smc5 binding based on the aforementioned structural information. These include: the T1 region, which is the most N-terminal region of Mms21 wrapped around Smc5; the α2 and α3 helixes, which contact Smc5 in a parallel fashion; and the T2 region, which is composed of six residues located between α1 and α2, and closely interacts with Smc5 (Figs. 3A–3B). We constructed deletion mutants that lack one or more of these regions and examined their interaction with the coiled-coil region of Smc5 in vitro. We found that Mms21 mutant proteins exhibited progressively weakened Smc5 binding as more of these regions were deleted (Figs. 3B–3C), suggesting that all of these regions contribute to the Mms21-Smc5 interaction.

Figure 3. The regions and residues important for Mms21-Smc5 interaction.

(A) Illustration of Mms21 residues important for Smc5 interaction. Each residue is indicated by its side chain in the ribbon/electrostatic surface representations of the Mms21-Smc5 structure. Residues mutated in M1 are marked red; those mutated in M2–M5 are marked as green, magenta, cyan and blue, respectively.

(B) Deletion and mutation constructs of Mms21. The schematic map of Mms21 is shown with the annotation of Smc5-interacting segments (T1, T2, α2, and α3), along with the mutations made in these segments (M1–M5) and with the RING domain. The representations of the four deletion constructs are drawn at the bottom.

(C) Deletion mutants of Mms21 decreased the interaction with the arm region of Smc5. Wild-type as well as each of the deletion mutant Mms21 proteins were expressed as His6-tagged proteins in E. coli and purified by Ni-NTA resins. These proteins were tested for their ability to pull-down the purified arm region of Smc5 (Smc5-arm-GST) as described in the text and methods. The eluate was examined by SDS-PAGE, and Coomassie blue-stained gel images are shown with the labels of the corresponding Mms21 constructs on top. The Smc5-arm-GST protein was not pulled down by the Ni-NTA resins (E). Smc5-arm-GST protein and various Mms21 proteins are denoted by arrows. Quantification of several experiments yielded the relative ratios between Smc5 and Mms21 proteins in the eluate; the average ratio and standard deviation are graphed in the lower panel.

(E) Point mutations of Mms21 decreased the interaction with the arm region of Smc5. Same as (C), except point mutant Mms21 proteins were used.

Using the structural information, we further identified residues within these regions that are in closest proximity to Smc5 and examined their importance in Smc5 binding by testing the effects of alanine replacement at these residues. In the T1 region, four amino acids (Pro9, Val12, Leu14, and His15) were found to contact Smc5 (Fig. 3D). His15 forms a hydrogen bond and a salt bridge with Asn791 of Smc5, whereas the three hydrophobic amino acids contribute to multiple van der Waals contacts with Smc5. A mutant protein, Mms21-M1, was made by replacing these four residues with alanines.

Similarly, two to four residues in each of the other Smc5-contacting regions were identified and mutated to alanines, resulting in four additional Mms21 mutants (Mms21-M2 through -M5). Mms21-M2 contains mutations at two hydrophobic amino acids (Leu25 and Leu30) in the T2 region (Fig. 3D). Both leucine residues provide van der Waals contacts with Smc5. Mms21-M3 contains three mutations (Ile33Ala, Tyr34Ala and Cys37Ala) at the N-terminus of the α2 helix, resulting in the loss of multiple van der Waals contacts with Smc5 (Fig. 3D). Mms21-M4 contains three mutations (Gln40Ala, Thr44Ala, and Gln47Ala) at the C-terminus of the α2 helix, resulting in the loss of seven hydrogen bonds and multiple van der Waals contacts with Smc5 (Fig. 3D). Lastly, Mms21-M5 contains four mutations in the α3 helix (Tyr76Ala, Glu79Ala, Ser80Ala, and Phe83Ala), resulting in the loss of one salt bridge, one hydrogen bond, and multiple van der Waals contacts (Fig. 3D).

The five mutated Mms21 proteins (-M1 through -M5) as well as the wild-type protein were expressed as His6-tagged proteins from E. coli and purified on nickel-nitrilotriacetic acid resins. These proteins were tested for their ability to pull down the arm region of Smc5. As shown in Figure 3E, mutations at these residues led to different degrees of reduction in Smc5 binding. Quantification of several trials of the pull-down experiments showed that the Mms21-M2 protein resulted in the most severe defect in Smc5 binding, followed by -M3, and then -M5. Mms21-M1 and -M4 led to moderate yet significant defects in Smc5 binding. Thus, we conclude that all the aforementioned residues contribute to Smc5 binding, though their relative contributions are different. The residues mutated in M2 are the most important, followed by those mutated in M3 and M5 and finally by those mutated in M1 and M4.

Mutations in Mms21 that affect Smc5 interaction lead to growth defects and DNA damage sensitivity

Next we investigated the biological significance of the Mms21-Smc5 interaction by testing whether Mms21 mutations that weaken the interaction with Smc5 can result in cell growth defects. First we replaced the wild-type MMS21 with each of the four aforementioned deletion constructs. This was performed in diploid yeast cells to provide one copy of wild-type MMS21 in order to avoid potential lethality during strain construction. The resulting strains were sporulated to generate haploid progeny containing only the mms21 deletions. We observed that these haploid mutants either grew slowly (ΔP16 and ΔL30) or were inviable (ΔS53 and ΔQ103); with larger deletions resulting in increased phenotypic severity (Fig. 4A). This result is consistent with the notion that interaction with Smc5 is necessary for the essential function(s) of Mms21. Since deletion constructs can perturb protein folding, we also performed similar tests using point mutants defective in Smc5 binding.

Figure 4. mms21 mutations which decreased Smc5 interaction result in cell death and DNA damage sensitivity.

(A–C) Deletions and point mutations of Mms21 diminishing their interaction with Smc5 result in cell death or growth defects. Representative tetrads from diploid strains heterozygous for the indicated mms21 deletions or mutations are shown. The spore clones containing mms21 deletions or mutations are indicated by diamonds and squares, respectively.

(D) Sensitivities of mms21 mutants toward replication blocking agents MMS, UV, and HU. Wild-type and mutant mms21 strains as indicated were examined by comparing their growth on YPD plates with or without the toxins at the indicated concentrations.

(E) The mms21 point mutations do not decrease Mms21 protein levels. Cell lysates were made from cultures growing at 23°C (left panel) or 30°C (right panel). Wild-type and mutant Mms21 proteins were tagged with HA and their levels examined by protein blotting using anti-HA antibody (top). Equal loading was shown by examining the same blot using anti-PGK antibody (bottom).

(F) Mms21-M1, -M3 and -M5 proteins affect Smc5 binding in vivo. Cells contain Smc5-TAF and HA-tagged wild-type (WT) or mutant Mms21 proteins as indicated. Smc5 and associated proteins were isolated using anti-ProA antibody and examined by Western blot (left panel). The relative amounts of Mms21 proteins that co-purified with Smc5 were quantified from several blots; the average ratio and standard deviation are presented on the right panel.

Using an approach similar to the one described above, mms21-M1 to M5 were integrated at the MMS21 chromosomal locus. We found that mms21-M1 and -M2 both led to cell death at 30°C, the normal growth temperature for yeast, but could give rise to medium and tiny sized spore clones, respectively, at 23°C (Figs. 4B and 4C). In addition, mms21-M3 resulted in slow growth, while mms21-M4 and -M5 appeared to grow as wild-type (Fig. 4B). We further examined mms21-M4 and -M5 for their sensitivity to DNA damaging agents, as the Smc5/6 complex is required to overcome replication blocks caused by hydroxyurea (HU), methyl methanesulfonate (MMS) and ultraviolet radiation (UV) (Murray and Carr, 2008). As shown in Figure 4D, mms21-M5, but not -M4, cells were sensitive to these replication blocking agents, suggesting that mms21-M5, but not -M4, affects the complex’s repair function. The growth or repair defects exhibited by mms21 mutants were not due to insufficient protein levels, as the mutant proteins were expressed at or above wild-type levels (Fig. 4E). Taken together, our results suggest that mms21 mutations affecting Smc5 binding impair the essential and repair functions of the Smc5/6 complex.

The severity of the phenotype of mms21-M2 to M5 cells correlates well with the degree of impairment in Smc5 binding in vitro. The moderate defect in Smc5 binding exhibited by the M4 mutation (85% of wild-type level, Fig. 3E) appears insufficient to cause an obvious phenotype when other interacting interfaces remain intact. The more pronounced impairment in Smc5 binding exhibited by the other three mutations (M2, 19% of wild-type; M3, 48% of wild-type; and M5, 61% of wild-type; Fig. 3E) resulted in cellular defects ranging from cell death (M2), poor growth (M3), to normal growth but with DNA damage sensitivity (M5). Thus, for mms21-M2 to M5, weaker Smc5 binding correlates with a more severe phenotype. The mms21-M1 mutant is an exception to this correlation, as it exhibited a moderate defect in Smc5 binding in vitro (84% of wild-type, Fig. 3E), yet led to cell death at 30°C (Fig. 4B). To understand this discrepancy, we examined the interaction between mms21-M1 and Smc5 in vivo. As shown in Figure 4F, mms21-M1 led to a large reduction in Smc5 binding (5% of wild-type). This reduction was more pronounced than mms21-M3 and -M5, which exhibit 30% and 66% of wild-type level binding, respectively. This observation provides an explanation for the strong phenotype exhibited by mms21-M1. The different in vitro and in vivo binding of mms21-M1 suggests that the residues affected by M1 (the T1 region) have additional roles in Smc5 binding in vivo.

The SPL-RING structure has both similarities and differences with the RING and U-box domains of ubiquitin E3 ligases

The core of Mms21 CTD contains the SPL-RING domain, which forms a distinct type of RING structure. Similar to the RING structure of ubiquitin E3s, this region of Mms21 contains three short β strands (β1– β3), an alpha-helical domain (α7), and two loops (T7 or Loop 1, and T8 or Loop2; Fig. 1B, Fig 2A, and Fig 5A). Strikingly, unlike the RING structure of ubiquitin E3s, in which each loop features a zinc ion stabilized by four cysteines and/or histidines, Mms21 contains only one such loop (Loop2) while its other loop (Loop 1) does not contain any zinc ion (Fig. 1B and Fig 5A). The zinc ion in Mms21 is located between Loop2 and beta sheets β1 and β2. It is coordinated by three cysteines (Cys200, Cys221, and Cys226) and His202 (Fig. 5B). The Loop 1 (Leu182 to Tyr191) stays close to the β2 and α7 regions (Figs. 5A and 5C). It does not contain a zinc ion and is stabilized by the hydrogen bonds formed between Cys184 on Loop 1 and Asp205 on β2, as well as multiple van der Waals contacts among Cys184, Pro185 and Ile186 from Loop 1 and Phe204 and Asp205 from β2 (Fig. 5C). The three cysteines and one histidine required to incorporate the zinc ion in Loop2, as well as the five residues important for stabilizing Loop 1, are highly conserved among Mms21 homologs and PIAS family proteins (Figs. 5D and 5E), suggesting that SPL-RING is a common structure for Mms21 proteins and is likely shared by the SP-RING type SUMO ligases.

Figure 5. Mms21 SPL-RING structure is different from RING and U-box structures.

(A) The structure of the SPL-RING domain in Mms21. This structure contains two loops (Loops 1 and 2), one short helix (α7), and three short β sheets (β1-3). The zinc ion close to Loop 2 is indicated by a green sphere.

(B) A close-up view of the zinc ion interacting with His202 and three cysteines (Cys200, Cys221, and Cys226) to stabilize Loop 2.

(C) A close-up view of the residues critical for the maintenance of Loop 1 structure. Residues critical for Loop 1 structure are indicated.

(D) The alignment of SPL-RING sequences of Mms21 proteins from budding yeast (sc), fission yeast (sp) and humans, with the RING sequences of c-Cbl and Ring1b, and with the U-box sequence of CHIP. The blue boxes indicate the three cysteines and one histidine surrounding the zinc ion near Loop 2s in Mms21, c-Cbl and Ring1b. Magenta boxes indicate the five residues required for the Loop 1 structure of Mms21. Note that these residues are conserved among Mms21 homologs, but not in c-Cbl Ring1B or CHIP. Red boxes mark the four cysteines surrounding the zinc ion in Loop 1 of c-Cbl and Ring1b.

(E) The alignment of the SPL-RING sequence of Mms21 and the SP-RING sequence of the PIAS proteins. The blue and magenta boxes are as described in (D).

(F) The superimposition of SPL-RING with RING and U-box structures. The SPL-RING of Mms21 (gold) is superimposed with the RING structure of c-Cbl (silver, left), the RING structure of Ring1b (blue, middle) and the U-box structure of CHIP (salmon, right).

Comparison of the SPL-RING domain with structures in PDB using the DALI server (Holm and Sander, 1993) revealed that it has similarities with RING structures of several ubiquitin E3s, such as the Ring1b/Bmi1 dimer of the Polycomb complex and c-Cbl (Zheng et al., 2000; Buchwald et al., 2006). The RING domains of Ring1b and c-Cbl can be superimposed nicely on SPL-RING with a Root Mean Square Deviation (RMSD) of 1.3 and 1.9 Å, respectively (Fig. 5F). The Loop2 structures of the three proteins share the greatest similarities; the histidine and three cysteines that interact directly with the zinc ion are conserved (Figs. 5D and 5F). The Loop 1 regions are different in SPL-RING and the ubiquitin E3 RINGs. In Ring1b and c-Cbl, Loop 1 contains a zinc ion that is stabilized by four cysteines (Fig. 5D). In Mms21, this loop does not contain zinc and stays closer to the short helix (α7) and β2, such that they form one hydrogen bond and multiple van der Waals contacts. In addition to RING structures, SPL-RING also exhibits similarities with the U-box domain found in ubiquitin E3s such as CHIP (Zhang et al., 2005; Xu et al., 2006; Xu et al., 2008). SPL-RING and the U-box domain of CHIP can be superimposed with an RMSD of 1.5 Å (Fig. 5F). Differences between these two structures are found in both Loop 1 and Loop 2. In contrast to the SPL-RING or RING domains, neither Loop 1 nor Loop 2 of the U-box domain contains zinc; instead, each loop is stabilized by one set of hydrogen bonds (Figs. 5D and 5F). In summary, SPL-RING is similar to the RING and U-box domains, but displays unique properties that can provide a structural basis for the functional specificities exhibited by SUMO and ubiquitin E3s.

Mms21 can recognize SUMO E2, but not ubiquitin E2s

The RING domains of both ubiquitin and SUMO E3s are responsible for interacting with their corresponding E2s. However, the mechanisms by which these two types of E3 specifically recognize the correct E2 are not clear. The structures of the SUMO E2, Ubc9, and those of the ubiquitin E2s, such as UbcH7 and Ubc13, are similar, with an RMSD of 1.8Å and 1.6Å, respectively. The structural similarities between SUMO and ubiquitin E2s and E3s allow us to use the determined structure of the c-Cbl-UbcH7 and CHIP-Ubc13 complexes to generate a superimposed model for the Mms21 RING-Ubc9 complex.

Comparing the superimposed structure of Mms21-Ubc9 with those of the c-Cbl-UbcH7 and CHIP-Ubc13 complexes, two differences are observed. First, the most important residues for ubiquitin E2 and E3 recognition are Phe63 in UbcH7 and Met68 in Ubc13, two relatively large hydrophobic amino acids (Zheng et al., 2000; Zhang et al., 2005); Figs. 6A and 6B). These residues protrude into a groove of the RING in c-Cbl or U-box in CHIP to make multiple van der Waals contacts. However, the residue occupying this position in Ubc9 is Ser70 (Figs. 6A and 6B). This change ablates the van der Waals interactions with c-Cbl or CHIP, thus providing one explanation for the inability of Ubc9 to recognize ubiquitin E3s. Moreover, Ser70 is conserved in all Ubc9 homologs (Supp. Fig. 2) and makes several new contacts with Mms21 SPL-RING based on modeling. We mutated Ser70 to alanine and found that this mutation greatly reduced sumoylation activity in vitro (Fig. 6C). This is consistent with the notion that Ser70 of Ubc9 is important for its function. Second, the MSΦI/LΦ (with Φ indicating hydrophobic residues) sequence at the extreme N-terminal region of Ubc9 and residues in the Mms21 Loop 1 are in close proximity and can contact each other (Fig. 6D and Supp. Fig. 2). This feature is not observed for the Mms21-UbcH7 or the Mms21-Ubc13 pair in the superimposed models (Fig. 6D). Compared to Ubc9, the N-terminus of UbcH7 lacks two residues and its first three residues are disordered in the crystal structure, while the N-terminus of Ubc13 projects away from the RING structure (Supp. Fig. 2 and Fig. 6D). Thus, while the N-terminal helix of Ubc9 can interact with Mms21 Loop 1, a similar interaction is unlikely to form between UbcH7 or Ubc13 and Mms21. Moreover, the N-terminal five amino acids are conserved among Ubc9 homologs (Supp. Fig. 2), further indicating their importance. To test this idea, we examined a Ubc9 mutant that lacks these residues in the in vitro sumoylation assay and found that it exhibited a greatly reduced activity (Fig. 6C). Thus, this region is required for Ubc9 function. We also examined two residues, I186 and T187, on Mms21 Loop 1, which are predicted to contact Ubc9 (Fig. 6D). We found that both I186A and T187A mutations greatly decreased Mms21 activity in vitro (Fig. 6E), indicating their importance for Mms21 function.

Figure 6. Comparison of the E2–E3 interactions for sumoylation and ubiquitination.

(A–B) The superimposition of Ubc9 with UbcH7 (A) and with Ubc13 (B). Shown are the region of UbcH7 (cyan) that contacts c-Cbl (gray), the region of Ubc13 (cyan) that contacts CHIP (gray), and the corresponding region of Ubc9 (green). The Phe63 residue in UbcH7 and Met68 in Ubc13, which protrude into a groove in the RING domain of c-Cbl or in the U-box domain of CHIP, respectively, are indicated. The corresponding residue Ser70 in Ubc9 is labeled.

(C) Removal of the five conserved N-terminal amino acids or mutating Ser70 greatly diminishes Ubc9 function. In vitro sumoylation assays using purified full-length Ubc9 (FL), a mutation lacking the five conserved N-terminal amino acids (ΔN), and a serine 70 to alanine mutation (S70A) were examined by anti-SUMO blot. Note that all reactions contain the same amount of SUMO, SUMO E1, Mms21 and full-length or mutant Ubc9 proteins.

(D) The superimposition of Mms21 SPL-RING domain with c-Cbl and CHIP in complex with the N-terminal regions of corresponding E2s. The SPL-RING domain of Mms21 (gold) and the RING of c-Cbl (gray) are shown with the N-terminal region of Ubc9 (green) in the left panel and with that of UbcH7 (cyan) in the middle panel. The SPL-RING domain of Mms21 (gold) and the U-box of CHIP (gray) are shown with the N-terminal region of Ubc13 (blue) in the right panel. Note that the N-terminal region of Ubc9, but not that of UbcH7 or Ubc13, forms helix turns which interact with four residues on Mms21 Loop 1.

(E) I186A and T187A mutations of Mms21 greatly reduced sumoylation function. Experiments were carried out as in (C).

Discussion

An important feature distinguishing the Smc5/6 complex from cohesin and condensin is its Mms21 subunit, which possesses SUMO ligase activity and uniquely binds to the arm of Smc5 (Zhao and Blobel, 2005; Potts and Yu, 2005; Andrews et al., 2005; Sergeant et al., 2005; Duan et al., 2009). Its sumoylation activity has been shown to be important for processes such as recombinational repair and telomere maintenance (Zhao and Blobel, 2005; Ampatzidou et al., 2006; Branzei et al., 2006; Potts and Yu, 2007; Pebernard et al., 2008). Moreover, genetic studies suggest that Mms21 must have additional functions that contribute to cell growth and possibly to DNA repair. We suspect that these functions may be related to its unique positioning within the complex.

To understand how Mms21 carries out its multiple functions, we determined the structure of the Mms21-Smc5 complex, which contains full-length Mms21 and the arm region of Smc5 that binds to Mms21. This structure revealed that Mms21 has a bipartite fold. Its N-terminal half forms a helical bundle with the arm region of Smc5. Examination of this region delineated four segments of Mms21 important for Smc5 binding. Our mutagenesis studies suggested that these regions provide functions required for growth and DNA repair. The C-terminal region of Mms21 formed an SPL-RING structure distinct from those found in ubiquitin E3s. The SPL-RING contained one zinc ion in Loop 2 but none in Loop 1. Rather, Loop 1 was stabilized by five evolutionarily conserved residues. Modeling studies showed that residues in Loop 1 form interactions with the extreme N-terminal region of the SUMO E2 but not with ubiquitin E2s. These features, as well as others (see below), likely contribute to the specificities seen in E2–E3 recognition in the sumoylation and ubiquitination pathways.

The interaction between the coiled-coil region of Smc5 and the N-terminal domain of Mms21

The structure of the Mms21-Smc5 complex revealed that N-terminal half of Mms21 is dedicated to Smc5 interaction. The interface between Mms21 and Smc5 is extensive and involves multiple types of interactions, including hydrogen bonds, salt bridges, and van der Waals contacts. These features are consistent with results from thermodynamic studies. Alignment of Smc5 proteins from different species showed that the residues involved in Mms21 interactions are largely conserved (Supp. Fig. 3), indicating that Mms21-Smc5 interactions in other species are likely to be similar to what we have shown here. The extensive interactions in the form of helix bundle reveal an interesting and somewhat unexpected mode of protein binding involving the coiled-coil region of SMC proteins, as previous models suggest that loops can form in this region and mediate the binding to other proteins (Beasley et al., 2002; Milutinovich et al., 2007). Our work reveals for the first time the molecular details of how coiled-coil regions of SMC proteins can interact with other proteins.

To understand the biological importance of the Mms21-Smc5 interaction, we dissected and examined different segments of the Mms21 NTD. Structure-guided mutagenesis studies showed that for M2–M5, more severe phenotypes were seen with mutated proteins that bound more weakly to Smc5 in vitro. For the M1 mutant, which bound Smc5 quite well in vitro, yet led to inviability at 30°C and growth defects at 23°C, the severe phenotype can be explained by its poor Smc5 binding in vivo. The difference between in vitro and in vivo Smc5 binding for mms21-M1 suggests that the region containing M1 mutations (the T1 region) plays additional roles in Smc5 interaction in the presence of other subunits or factors inside the cell. Collectively, our results show that the interaction with Smc5 is fundamental the essential roles of Mms21. Several possibilities can be envisioned for how the Mms21-Smc5 interaction affects the complex's functions. For example, by interacting with Smc5, Mms21 may facilitate Smc5 adopting a conformation needed for interacting with other subunits. Our recent work showed that Smc5 is the only subunit that interacts with all the other subunits in the Smc5/6 complex (Duan et al., 2009). It is reasonable to image that forming multiple interactions simultaneously demands a specific conformation, the adoption of which could be facilitated by Mms21. Moreover, Mms21 binding may result in a bending in the Smc5 arm that may be required for DNA tethering or for the dynamics of the complex. Future work will be needed to investigate each of these possibilities in detail.

The structural basis for the SUMO E3 function of Mms21

Another important finding of this work is the revelation that the Mms21 protein forms a RING-like structure. As the first SPL-RING structure determined for SUMO ligases, this structure reveals both similarities and differences with ubiquitin E3 RING and U-box structures. While the overall form of the three structures is similar, the major difference is how their two loops are structured. The two loops in the ubiquitin E3 RING are both maintained by zinc ions, whereas those in the U-box domain are devoid of zinc. In contrast to both, only one loop of SPL-RING contains zinc and the other is stabilized by five conserved residues. The conservation of these residues among Mms21 homologs as well as in PIAS proteins suggests that this is a common structure in these SUMO E3s.

Modeling studies revealed two main differences between SUMO and ubiquitin E2–E3 interactions. While Phe63 in UbcH7 and Met68 in Ubc13 are critical for recognizing ubiquitin E3s, the residue at this position in Ubc9 is Ser70, suggesting that Ubc9 is not designed to bind ubiquitin E3s. This may provide one explanation for the E2–E3 specificity. The conservation of Ser70 among Ubc9 homologs further indicates that it plays an important role. Indeed, we showed that S70A mutation greatly reduced Ubc9 activity. The second difference suggested by modeling studies is that Mms21 Loop 1 is in close proximity to the conserved MSΦI/LΦ sequence at the Ubc9 N-terminus, while the corresponding sequences in UbcH7 and Ubc13 are either missing/misorientated or stretched away. We show that a Ubc9 mutant lacking the first five amino acids and Mms21 mutants affecting two residues in Loop 1 (I186A and T187A) greatly affected the sumoylation activity. These results are consistent with the predictions from the modeling work and suggest that the two features revealed by the model are important for the sumoylation reaction.

It is worth noting that the SPL-RING structure of Mms21 stretches away from Smc5, possibly allowing for interactions with sumoylation enzymes and substrates. The region connecting the NTD and CTD of Mms21 is flexible, as suggested by the low density of some segments in this region and a high B factor, the two crystallographic features indicative of a high degree of freedom. It is possible that this region allows some relative movement between the NTD and CTD of Mms21, which may be important for the recognition of different substrates. This idea will be examined in future work.

In summary, our results have revealed a bipartite organization of Smc5-bound Mms21, with its N-terminal half extensively interacting with Smc5 and its C-terminal region forming an SPL-RING structure stretching away from Smc5. We showed that the NTD of Mms21 is required for its essential and repair functions. Its SPL-RING structure provides a molecular explanation for the specificity in recognizing SUMO, but not ubiquitin, E2s. The structural and functional understanding of this subunit within the Smc5/6 complex establishes a molecular basis from which the roles of this complex in various chromosomal processes can be further examined.

Experimental Procedures

Protein expression and purification

The budding yeast Smc5 segments containing Asp302 to Leu369 and Lys733 to Gln813 were linked by a four-amino-acid linker peptide Gly-Ser-Gly-Ser and expressed as a 6xHis-tagged protein from the pET24d vector (Novagen, Madison, WI); full-length Mms21 was expressed from pET24d in E. coli for crystal structure determination. Mms21 point mutations were generated by Quikchange (Invitrogen, Carlsbad, CA) and expressed from either pET24d or pET24a (Novagen), and Smc5 was cloned into pGEX4T3 (GE Healthcare, Piscataway, NJ) for interaction studies. All proteins were expressed in Rosetta (DE3) cells (Novagen). Protein expression and purification were performed using standard protocols. Proteins containing the 6xHis-tag were purified by Ni-NTA affinity chromatography and those containing the GST-tag were purified by glutathione-affinity chromatography.

Crystallization

The crystals of the Mms21-Smc5 complex were grown at room temperature using the hanging drop vapor diffusion method. The wells contained 100 mM Tris-HCl (pH 7.3), 18% PEG 3350 and 40 mM CaCl₂. Crystals formed overnight to 0.3 × 0.3 × 1 mm in size.

Data collection and structural determination

Diffraction data were collected at advanced photo source (APS) at beamline NE-CAT 24ID and SBC-CAT 19BM. Data were processed with HKL2000/Denzo and scalepack (Otwinowski, 1997). The native crystal was diffracted to 2.3Å and the Seleno-methional crystal was diffracted to 3.9Å. Seven seleno sites were located by SOLVE (Terwilliger and Berendzen, 1996). Refinement was performed using CNS (Brunger et al., 1998) and CCP4i (Brunger et al., 1998), and model building was carried with Coot (Emsley and Cowtan, 2004). The figures were generated with PyMOL (Delano, 2002). The final refined structure contained residues 304–363 and 739–811 of Smc5 and residues 5 to 258 of Mms21.

In vitro protein interaction and sumoylation assays

50–100 µg of purified His-tagged wild-type or mutant Mms21 proteins were added to 20 µl of Ni-NTA beads and incubated for 10 minutes. Beads were washed with 1 ml of lysis buffer to remove unbound proteins. 50–100 µg of purified GST or GST-Smc5-arm proteins were added to the beads and incubated for 30 minutes at 4°C with rotation. The beads were washed three times with 1 ml phosphate-buffered saline (PBS) buffer and eluted with 100 mM phosphate buffer containing 300 mM Imidazole and 100 mM NaCl, and the eluted samples were analyzed by SDS-PAGE. Purification of SUMO, SUMO E1 and E2 and the in vitro sumoylation assay were performed as described in Zhao and Blobel, 2005. Method for ITC measurement and thermodynamic calculation are described in the supplemental materials.

Yeast methods

The MMS21 gene was replaced with deletions or point mutations of mms21 using standard gene replacement procedures. All mutations were confirmed by DNA sequencing; primer sequences are available upon request. Wild-type and mutant alleles of mms21 were tagged with a 3HA at their C-termini; Smc5 was tagged with TAF (containing ProA; Chen et al., 2007) at its C-terminus. The tags do not affect protein function as evidenced by the normal growth and DNA damage resistance of the tagged strains. Yeast strains are listed in Table S2. Sporulation and tetrad analyses were carried out following standard methods. To detect the DNA damage sensitivity, mid-log phase YPD-grown cells were spotted in 10-fold serial dilutions (10⁵ to 10 cells) on YPD plates or plates containing MMS or HU. One spotted YPD plate was treated with UV light. Plates were incubated at 30°C for 3 days or 23°C for 4 days before pictures were taken. Co-immunoprecipitation and protein detections were carried out using standard methods as described in Zhao et al., 1998. Mms21 proteins were detected by anti-HA antibody and Smc5 proteins were detected by anti-ProA antibody.