DNA activates the Nse2/Mms21 SUMO E3 ligase in the Smc5/6 complex

Abstract

Modification of chromosomal proteins by conjugation to SUMO is a key step to cope with DNA damage and to maintain the integrity of the genome. The recruitment of SUMO E3 ligases to chromatin may represent one layer of control on protein sumoylation. However, we currently do not understand how cells upregulate the activity of E3 ligases on chromatin. Here we show that the Nse2 SUMO E3 in the Smc5/6 complex, a critical player during recombinational DNA repair, is directly stimulated by binding to DNA. Activation of sumoylation requires the electrostatic interaction between DNA and a positively charged patch in the ARM domain of Smc5, which acts as a DNA sensor that subsequently promotes a stimulatory activation of the E3 activity in Nse2. Specific disruption of the interaction between the ARM of Smc5 and DNA sensitizes cells to DNA damage, indicating that this mechanism contributes to DNA repair. These results reveal a mechanism to enhance a SUMO E3 ligase activity by direct DNA binding and to restrict sumoylation in the vicinity of those Smc5/6‐Nse2 molecules engaged on DNA.

Introduction

Genome integrity is under constant surveillance. External insults, the metabolism of the cell, and most chromosome transactions can damage the genome, resulting in deleterious mutations or cell death. Cells have therefore developed DNA repair and DNA damage tolerance mechanisms, which often require post‐translational modifications to recruit and activate repair factors. Post‐translational modification by conjugation to the small ubiquitin‐like modifier SUMO actively participates in maintaining the integrity of the genome: SUMO plays pivotal roles during chromosome replication and segregation and in virtually all DNA repair mechanisms (Bergink & Jentsch, 2009). In fact, the extensive sumoylation of many DNA repair and replication factors constitutes an integral part of the DNA damage response (Cremona et al, 2012).

SUMOylation involves the formation of an isopeptide bond between the ε‐amino group of a lysine residue and the C‐terminus of the SUMO protein. This reaction requires the participation of an E1 activating enzyme (Uba2/Aos1) that passes the SUMO proteins to an E2 conjugating enzyme (Ubc9). Once charged with SUMO, Ubc9 transfers SUMO to the target protein. This can occur directly, by recognition of a SUMO consensus motif (Bernier‐Villamor et al, 2002; Yunus & Lima, 2006), but most often requires the participation of an E3 SUMO ligase enzyme. Budding yeast codes for three mitotic SUMO E3 ligases, including two members of the PIAS family (Siz1 and Siz2) and the Siz/PIAS‐related Nse2 protein (Johnson & Gupta, 2001; Zhao & Blobel, 2005). Since the E1 and E2 enzymes in the SUMO pathway lack known DNA binding domains, modification of chromosome‐bound proteins seems to rely on recruitment of SUMO E3 ligases to chromatin (Ulrich, 2014). This can occur by direct binding of the E3 ligase to DNA, by interaction of the E3 with chromosome‐associated proteins, or by granting access of the E3 to DNA lesions via previous phosphorylation and ubiquitination events at damaged sites. For example, an N‐terminal SAP domain localize Siz2 on DNA, where it promotes sumoylation of homologous recombination factors (Psakhye & Jentsch, 2012); additionally, Siz2 binds ssDNA through interaction with the ssDNA binding replication protein A (RPA; Chung & Zhao, 2015). The Nse2 SUMO ligase, a subunit of the Smc5/6 complex, has also been shown to play key roles in the maintenance of chromosome integrity (Zhao & Blobel, 2005; Ampatzidou et al, 2006; Branzei et al, 2006; Potts et al, 2006; Pebernard et al, 2008b; Behlke‐Steinert et al, 2009; Chavez et al, 2010; Bermúdez‐López et al, 2015). As Nse2 lacks DNA binding domains, its DNA repair functions require its stable docking onto the Smc5 protein (Duan et al, 2009; Bermúdez‐López et al, 2015), and the subsequent association of the Smc5/6 complex with damaged sites (De Piccoli et al, 2006; Lindroos et al, 2006; Tapia‐Alveal & O'Connell, 2011; Bustard et al, 2012).

Structural Maintenance of Chromosomes (SMC) complexes are topologically closed molecules formed by the heterodimerization of two elongated SMC subunits and by a distinct number of associated non‐SMC elements (Uhlmann, 2016). SMC proteins contain three different domains: an ATPase head structurally related to that of ABC transporters (hereafter named “HEAD”), an extended coiled coil region (“ARM”), and a heterodimerization or hinge domain (“HINGE”). While hinge heterodimerization closes the molecule at one end, a kleisin subunit connects the two ATPase heads at the other end, defining an inner compartment delimited mainly by two long coiled coils. Each SMC complex has specific and essential roles: Cohesin maintains connections between sister chromatids, condensin compacts chromosomes, and Smc5/6 promotes chromosome disjunction. Despite these seemingly disparate functions, all SMC complexes share a common property, which is to organize chromosomes by topologically embracing DNA inside their ring‐shaped structure. This function requires the ATPase activity of the head domains, which regulates the entry and release of DNA fibers inside the SMC ring. In prokaryotes, the ATPase‐dependent conformational changes in the head domain can affect the architecture of the coiled coil domains, altering the ability of SMC molecules to embrace DNA (Bürmann et al, 2017).

Once loaded onto chromatin, Smc5/6 participates in critical chromosome transactions during DNA replication and repair. The Smc5 and Smc6 subunits can bind strongly to DNA in vitro through several binding regions located in the hinge, the head and the arm regions (Roy & D'Amours, 2011; Roy et al, 2011, 2015; Alt et al, 2017). In vivo, the affinity of the SMC core for DNA is regulated by the other six non‐SMC subunits (Nse proteins). Nse4, the kleisin subunit in the Smc5/6 complex, associates with Nse1 and Nse3 to constitute a stable subcomplex. The Nse1‐Nse3 pair contains a patch of positively charged residues that acts as a DNA‐binding surface and mediates Smc5/6 loading onto chromatin (Zabrady et al, 2016). Although no specific function has been attributed to the Nse5 and Nse6 subunits, their alleged functional homologues in vertebrates (SLF1 and SLF2) have been proposed to promote recruitment of the Smc5/6 complex to interstrand cross‐links (Räschle et al, 2015).

The Nse2 SUMO ligase has an undefined role in sister chromatid recombination and chromosome disjunction by promoting the sumoylation of several targets including subunits in cohesin (Potts et al, 2006; Almedawar et al, 2012; McAleenan et al, 2012), Smc5/6 (Bermúdez‐López et al, 2015) and STR complexes (Bermúdez‐López et al, 2016; Bonner et al, 2016). To reach its substrates, the E3 domain must first dock onto the central part of the ARM region of Smc5 through a long N‐terminal helical domain (Duan et al, 2009). While docking of Nse2 onto Smc5 through the N‐terminal domain is essential for viability (Duan et al, 2009; Bermúdez‐López et al, 2015), the C‐terminal RING domain, which codes for the E3 SUMO ligase activity, is dispensable; however, mutations in the SUMO ligase domain render cells sensitive to DNA damage, highlighting its relevance in genome maintenance.

It is currently unclear whether all Smc5/6‐Nse2 molecules are SUMO‐active, potentially targeting soluble proteins, or whether their activity is restrained to those Smc5/6‐Nse2 molecules directly engaged on DNA, promoting the modification of chromosome‐associated targets only. Here we show that DNA stimulates the SUMO ligase activity of Nse2 via a direct and non‐specific interaction with a positively charged patch in the ARM region of Smc5. This interaction probably induces a conformational change in the Nse2 molecule that ultimately enhances its SUMO E3 ligase activity, thereby promoting DNA damage repair. Overall, our findings define a new mechanism to activate a SUMO ligase on site, which does not merely rely on recruitment of the E3 to DNA, but on local upregulation of its activity after loading onto chromatin.

Results

DNA binding enhances the SUMO conjugation activity of Nse2

Despite most known SUMO‐targets of Nse2 are chromosomal proteins (Andrews et al, 2005; Zhao & Blobel, 2005; Potts et al, 2006; Potts & Yu, 2007; Pebernard et al, 2008a; Almedawar et al, 2012; McAleenan et al, 2012; Yong‐Gonzales et al, 2012; Albuquerque et al, 2013), it is currently unknown whether chromatin loading of Smc5/6 molecules modulates its E3 ligase activity. We therefore tested whether the SUMO conjugation activity of Nse2 could be directly affected by the presence of DNA. In vitro assays using recombinant full‐length Smc5 in complex with Nse2 show a substantial increase in SUMO conjugation in the presence of single‐stranded DNA (Figs 1A and EV1). The SUMO E3 ligase activity of Nse2 is strikingly enhanced by the presence of ssDNA, as observed after 30 min in multiple turnover reactions. The C‐terminal kleisin domain of Nse4 (cNse4—from Ile246 to Asp402) was used as a model substrate in our in vitro reactions, although SUMO conjugation can also be observed internally on lysine residues of Smc5 (Figs 1A and EV1). These results suggest that DNA binds to the Smc5‐Nse2 heterodimer and stimulates the E3 SUMO ligase activity of Nse2. It is important to note that in addition to target selectivity, E3 ligases also optimize catalysis, by placing the functional groups in an optimal orientation for Ub/Ubl transfer between the E2‐thioester and the target substrate (Reverter & Lima, 2005; Deshaies & Joazeiro, 2009; Plechanovová et al, 2012; Scott et al, 2014; Buetow et al, 2015; Streich & Lima, 2016).

Figure 1. Stimulation of the SUMO E3 ligase activity of the Smc5‐Nse2 complex upon binding to DNA .

1. Time‐course SUMO conjugation reaction in the presence or absence of ssDNA (virion ϕx174) using recombinant full‐length Smc5‐Nse2 complex. The substrate utilized was the C‐terminal kleisin domain of Nse4 (cNse4). Reaction was run at 30°C and stopped at indicated minutes by adding SDS‐loading buffer. Labels indicate the bands in the SDS–PAGE of the proteins in the reaction mixture (N‐S2, cNse4‐SUMO2; N‐2S2, cNse4‐2SUMO2; and N‐3S2, cNse4‐3SUMO2).
2. Left, SYPRO‐stained and Western blot (anti‐SUMO2) of the SUMO conjugation reaction by Smc5‐Nse2 complex in the presence of ssDNA (virion ϕx174) and dsDNA (pET‐DUET‐1) at either 1 or 10 nM. Reactions were run at 30°C and stopped at 60 min by adding SDS‐loading buffer. Right, bar diagram comparison of the relative SUMO conjugated cNse4 substrate in the presence of either ssDNA (virion ϕx174) or dsDNA (pET‐DUET‐1) at indicated concentrations. Straight line shows the basal cNse4‐SUMO conjugation in the absence of DNA. Data values are mean ± s.e.m. and n = 3 technical replicates.
3. Left, Western blot of the SUMO conjugation reaction in the presence of 50b oligonucleotide and the 5 kb virion ϕx174 at the indicated concentrations. Reactions were run at 30°C and stopped after 60 min by adding SDS‐loading buffer. Right, bar diagram comparison of the relative SUMO conjugated cNse4 substrate in the presence of 25, 34, or 50 bases oligonucleotides at indicated concentrations. Straight line shows the basal cNse4‐SUMO conjugation in the absence of DNA. Data values are mean ± s.e.m. and n = 3 technical replicates. Bar diagrams calculation was generated using ImageJ software (Schneider et al, 2012).

Figure EV1. Enhancement of the SUMO E3 ligase activity of Smc5‐Nse2 upon binding to DNA .

1. Time‐course conjugation reaction of SUMO1 (left) or Smt3 (right) in the presence or absence of ssDNA (virion ϕx174) at 8 nM using full‐length Smc5‐Nse2 complex. The substrate utilized was the C‐terminal kleisin domain of Nse4 (cNse4). Reactions using human or yeast E1 and E2 enzymes, for SUMO1 and Smt3, respectively, were run at 30°C and stopped at indicated minutes by adding SDS‐loading buffer. Labels indicate the bands in the SDS–PAGE of the proteins in the reaction mixture (N‐S1, cNse4‐SUMO1; N‐2S1, cNse4‐2SUMO1; N‐Smt3, cNse4‐Smt3; N‐2Smt3, cNse4‐2Smt3).
2. SYPRO‐stained SDS–PAGE of one of the triplicate SUMO conjugation reactions used to calculate the ssDNA vs. dsDNA plot in Fig 1B. The reactions were run for 60 min at 30°C and stopped by adding SDS‐loading buffer.
3. SYPRO‐stained SDS–PAGE of one of the triplicate SUMO conjugation reactions used to calculate the oligonucleotide plot in Fig 1C. The reactions were run for 60 min at 30°C and stopped by adding SDS‐loading buffer.

Most of our in vitro assays utilize single‐stranded DNA (ssDNA, 5 kb virion ϕx174), which seems to produce a higher increase in SUMO conjugation than a double‐stranded DNA plasmid of a similar length (Fig 1B), although both types of DNA molecules can produce a substantial increment in SUMO conjugation (Figs 1B and EV1). Single‐stranded DNA has been reported to bind Smc5 and Smc6 molecules with higher affinity than double‐stranded DNA through multiple binding sites, including parts of the coiled coil ARM domain (Roy et al, 2011, 2015). Additionally, shorter ssDNA molecules, such as random oligonucleotides of 20, 34, and 50 nucleotides, can also enhance SUMO conjugation when used at higher concentrations (μM) than the virion plasmid (nM) in a dose‐dependent manner (Figs 1C and EV1). Comparative activity assays using different types of ssDNA molecules at similar nucleotide concentration reveal an equal stimulation of the SUMO conjugation for either oligonucleotides of 50b or long ssDNA molecules (5 kb virion ϕx174 ssDNA; Fig 1C). Small oligonucleotides (25b) can also enhance SUMO conjugation when used at higher concentrations (Fig 1C). These results suggest a non‐specific dose‐dependent binding of DNA to the Smc5/6 complex resulting in a stimulation of the E3 ligase activity of Nse2.

A minimal Smc5 ARM domain is sufficient for upregulation of DNA‐dependent SUMO conjugation

Nse2 has weaker SUMO conjugation activity in the absence of Smc5, with no significant increase in the presence of ssDNA (Fig 2B and E, and Appendix Fig S1). This observation indicates that the Smc5 protein might directly interact with DNA to enhance sumoylation. Smc5 has been reported to bind DNA through multiple binding sites (Roy et al, 2011, 2015). Therefore, we mapped the regions in Smc5 responsible for sensing DNA that could enhance SUMO conjugation. We prepared two truncations of Smc5 in complex with Nse2: one lacking the dimerization HINGE domain (ΔHinge/Smc5), and another lacking the ATPase HEAD domain (ΔHead/Smc5). In both cases, the coiled coil domain of Smc5 included the Nse2‐binding region (Fig 2A). Full‐length and Smc5 truncations displayed a comparable and remarkable increase in SUMO conjugation in the presence of DNA (Fig 2B and Appendix Fig S1). The ΔHinge/Smc5 construct displayed a lower fold increase in sumoylation, most probably due to its already higher activity than wild‐type Smc5, even in the absence of DNA (Fig 2B and Appendix Fig S1). We speculate that the HINGE domain may exert an inhibitory role on Nse2‐dependent sumoylation. As cNse4 does not interact with either ΔHead/Smc5 or Arm/Smc5 (Appendix Fig S2), these assays also indicate that the enhancement of the SUMO E3 ligase activity can occur either on external substrates (such as cNse4) or on internal lysines residues of Smc5 (Fig 2C and D, and Appendix Fig S1). The only common region in all Smc5 constructs corresponds to the ARM coiled coil region that docks Nse2 (from Asp302 to Thr366, and from Arg737 to Gln813), indicating that it might contain a minimal DNA binding region (DNA sensor) that promotes the stimulatory effect on the E3 ligase activity of Nse2.

Figure 2. Enhancement of the SUMO E3 ligase activity upon DNA binding by Smc5‐Nse2 truncation complexes.

1. Schematic representation of the domain composition of the heterodimeric full‐length Smc5‐Nse2 complex.
2. Bar diagram representation of the relative SUMO conjugation activity of Nse2, full‐length Smc5‐Nse2, ΔHinge/Smc5‐Nse2, ΔHead/Smc5‐Nse2, and Arm/Smc5‐Nse2 truncation constructs (schematic representation above). Orange bars indicate the presence of ssDNA (virion ϕx174), and red bars indicate absence of ssDNA. Reaction rates were performed at least in three different independent experiments (see Fig EV2). Data values are mean ± s.e.m.; and n = 3 technical replicates. Significance was measured by a two‐tailed unpaired t‐test relative to wild‐type. **P < 0.01.
3. SYPRO‐stained (left) and Western blot (right) time‐course SUMO conjugation reaction using and ΔHead/Smc5‐Nse2 truncation construct in the presence of ssDNA. T7‐tagged Smc5 and E1 were immunodetected by an anti‐T7 antibody. Reactions were run at 30°C and stopped at indicated times by adding SDS‐loading buffer.
4. SYPRO‐stained (left) and Western blot (right) time‐course SUMO conjugation reaction using and ΔHinge/Smc5‐Nse2 truncation construct in the presence of ssDNA. T7‐tagged Smc5 and E1 were immunodetected by an anti‐T7 antibody. Reactions were run at 30°C and stopped at indicated times by adding SDS‐loading buffer.
5. Western blot of the time‐course SUMO conjugation reaction in the presence of ssDNA (virion ϕx174) using either Nse2 (left) or Arm/Smc5‐Nse2 complex (right). The reactions were run in the presence or absence of cNse4 external substrate. Reaction was run at 30°C and stopped at indicated minutes by adding SDS‐loading buffer (N‐S2, cNse4‐SUMO2; N‐2S2, cNse4‐2SUMO2; N‐3S2, cNse4‐3SUMO2; and pS2, poly‐SUMO2).

Therefore, we produced this Smc5 coiled coil ARM region (named Arm/Smc5) in complex with Nse2, based on the published crystal structure of Nse2‐Smc5 (PDB 3HTK; Duan et al, 2009). Activity assays in the presence or absence of cNse4 also display a striking comparable enhancement in SUMO conjugation upon DNA binding to Arm/Smc5‐Nse2, similarly to the other Smc5 long truncation constructs (Fig 2B and E). These results indicate that the DNA binding patch involved in the enhancement of the E3 ligase might be restricted to this ARM/Smc5 region in contact with Nse2.

A positive‐patch region of Smc5 ARM domain interacts with DNA

Similarly to the full‐length Smc5‐Nse2 complex, different ssDNA molecules can also stimulate SUMO conjugation in the Arm/Smc5‐Nse2 construct in a dose‐dependent manner (Fig 3A). Additionally, single turnover reactions using this minimal Arm/Smc5‐Nse2 construct display a strong enhancement of the E2‐thioester SUMO discharge in the presence of a 50b ssDNA oligonucleotide, indicating the role of the DNA binding promoting the isopeptidic bond formation by the stimulation of the E3 ligase (Fig 3B).

Figure 3. A positive‐patch region on the surface of Smc5 ARM domain interacts with DNA .

1. Western blot of the SUMO conjugation reaction in the presence of different oligonucleotides using the minimal Smc5/ARM‐Nse2 construct. Reactions were run at 30°C and stopped after 60 min by adding SDS‐loading buffer. 25b (μM), 34b (μM), and 50b (μM), stands for a 25, 34, and 50 bases oligonucleotides, respectively, and the indicated concentration is in μM units. 5 kb (nM) stands for the virion ϕx174, and the indicated concentration is in nM units (N‐S1, cNse4‐SUMO1; N‐2S1, cNse4‐2SUMO1; and pS1, poly‐SUMO1) (*overexposed chemiluminescent signal).
2. Left, Ubc9‐thioester formation in the presence of E1, E2 enzymes, Alexa488‐SUMO1, and ATP. Right, single turnover reaction of the SUMO conjugation reaction in the presence or absence of ssDNA (50b) using Arm/Smc5‐Nse2 as E3. Samples were run in the presence (below) or absence of β‐mercaptoethanol (above).
3. Ribbon representation of the complex between the ARM domain of Smc5 (yellow and orange) and Nse2 (pink) (PDB 3HTK; Duan et al, 2009). Lysine residues forming the positive‐charged patch in the surface of the coiled coil Smc5 ARM are labeled and shown in stick representation (blue). Zinc atom in the Nse2 RING domain is depicted as a yellow sphere.
4. Bar diagram representation of the SUMO conjugation rates of activity assays of Arm/Smc5‐Nse2 KE mutants in the presence (orange bars) or absence (red bars) of ssDNA (virion ϕx174), relative to wild type (set to 1). Reaction rates were performed at least in three different independent experiments. Data values are mean ± s.e.m. and n = 3 technical replicates. Significance was measured by a two‐tailed unpaired t‐test relative to wild‐type. *P < 0.05, **P < 0.01, ***P < 0.001.
5. DNA binding properties of wild‐type, K337E/K344E/K764E and K743E/K745E Arm/Smc5‐Nse2 mutants, were determined by electrophoretic mobility shift assays (EMSA) saturation experiments. Protein complexes were incubated for 30 min at 30°C before loading the agarose gel electrophoresis. Numbers above gel indicate the molar ratio (×10³) of protein over ssDNA (virion ϕx174) in each lane.

Our next goal was to uncover the DNA binding regions on the surface of the minimal Smc5 ARM domain that can stimulate the activity of Nse2. Interestingly, the crystal structure of the Smc5‐Nse2 complex revealed positive‐charged patch regions in the Smc5 coiled coil surface of Arm/Smc5‐Nse2 that could fulfill non‐specific interaction to DNA (Fig 3C). To check this electrostatic interaction, we produced several Arm/Smc5‐Nse2 constructs with different combinations of lysine to glutamic acid mutants to counter‐charge binding to phosphate groups of DNA.

All tested Arm/Smc5‐Nse2 KE mutants show a comparable activity in the absence of DNA (Figs 3D and EV2), indicating in all cases that the mutagenesis has not compromised either the structure or the catalytic properties of the enzyme. However, in the presence of DNA, all single point, double, and K337E/K344E/K764E mutants reduce SUMO conjugation at different levels, reaching an almost complete loss of enhancement in the K743E/K745E mutant (Fig 3D). Interestingly, the most relevant lysine residues locate in the region next to the RING domain, reducing the effect as they move away from that region. We attribute this decrease in the SUMO E3 ligase activity of Nse2 to an electrostatic perturbation in DNA binding. Also, in contrast to the KE mutants, the SUMO conjugation enhancement of the K743R/K745R double mutant is similar to the wild‐type form (Fig 3D), confirming the role of the electrostatic charge of this interface in the DNA binding. Additionally, electrophoretic mobility shift assays (EMSA) using two different Arm/Smc5‐Nse2 KE mutants, K337E/K344E/K764E and K743E/K745E, showed a significant reduction in the DNA binding in comparison with the wild‐type form, confirming the perturbation of the binding between DNA and the Arm/Smc5‐Nse2 complex (Fig 3E). These EMSA experiments were conducted at higher DNA:protein ratios in comparison with the in vitro activity assays, denoting the unspecific electrostatic binding between ARM/Smc5 and the DNA molecule.

Figure EV2. Mutagenesis analysis of the SUMO E3 ligase activity of the Smc5‐Nse2 constructs upon binding to DNA .

1. SYPRO‐stained (left) and Western blot (right) of the time‐course reaction of SUMO conjugation in the presence or absence of ssDNA (virion ϕx174) at 8 nM, using either wild‐type Arm/Smc5‐Nse2 or K337E/K344E/K764E mutant. The reactions were run at 30°C with in the presence of the C‐terminal kleisin domain of Nse4 as a substrate. (N‐S2, cNse4‐SUMO2; N‐2S2, cNse4‐2SUMO2; N‐3S2, cNse4‐3SUMO2; and pS2, poly‐SUMO2).
2. SYPRO‐stained SDS–PAGE of SUMO conjugation reactions of wild‐type Arm/Smc5‐Nse2 at indicated NaCl concentrations in the presence or absence of ssDNA (virion ϕx174) at 8 nM (N‐S2, cNse4‐SUMO2; N‐2S2, cNse4‐2SUMO2; N‐3S2, cNse4‐3SUMO2; and pS2, poly‐SUMO2).
3. SYPRO‐stained SDS–PAGE of SUMO conjugation reactions of wild‐type and 7KE mutant of Δhinge/Smc5‐Nse2 and Δhead/Smc5‐Nse2 in the presence or absence of 50b ssDNA.

DNA binding to Smc5‐Nse2 triggers a conformational change

The enhancement of the SUMO conjugation upon DNA binding could derive from the structural modification in the Nse2 E3 ligase to stimulate its enzymatic activity. To test this idea, we used circular dichroism spectroscopy, which measures the differential absorption of the circularly polarized light. In the far ultraviolet region, this variation arises mainly from changes in the secondary structure elements and is highly sensitive to conformational changes of proteins (Kelly et al, 2005).

The circular dichroism analysis of the Arm/Smc5‐Nse2 complex displays the spectra of a well‐folded α‐helical rich protein, with two characteristic ellipticity minimals at 210 and 222 nm, respectively (Fig 4 and Appendix Fig S3). The structural integrity of the complex was confirmed by temperature denaturation after incubation at 100°C, which resulted in a total loss of the circular dichroism signal (Fig 4A). Interestingly, increasing concentrations of DNA (ϕx174) produced a dose‐dependent change of the circular dichroism spectra, which might indicate a DNA‐induced structural change. Interestingly, the variation of ellipticity displayed by the wild‐type spectra was reduced significantly at different levels when four different types of Arm/Smc5‐Nse2 KE mutants were used (Fig 4A and Appendix Fig S3). In all cases, the change in the molar ellipticity for the Arm/Smc5‐Nse2 KE mutants did not reach the saturation levels displayed by the wild‐type form under similar experimental conditions, probably indicating a loss of affinity between DNA molecules and the KE mutants (Fig 4A). In summary, all our circular dichroism experiments are indicative of a structural change in the secondary structure elements of Arm/Smc5‐Nse2 upon DNA binding.

Figure 4. Binding of DNA induces distinct conformational changes in Arm/Smc5‐Nse2 wild type and mutants.

1. Left, circular dichroism (CD) analysis of purified wild‐typE Arm/Smc5‐Nse2. Black line, wild‐type Arm/Smc5‐Nse2 without bound ligand (no 5 kb ssDNA); dashed black line, denatured protein; blue line, ligand/Smc5‐Nse2; numbers in graphs (dotted black lines), molar ratio of ligand/Smc5‐Nse2. Right, conformational changes induced by 5 kb circular ssDNA quantified by ligand titration until signal change in mean residue ellipticity (MRE) at 222 nm achieved saturation. Blue squares Arm/Smc5‐Nse2 (wild‐type); gray triangles Arm/Smc5‐Nse2 (K333E/K344E); green circles Arm/Smc5‐Nse2 (K764E); pink triangles Arm/Smc5‐Nse2 (K333E/K344E/K764E); red diamonds Arm/Smc5‐Nse2 (K743E/K745E).
2. Left, tryptophan intrinsic fluorescence of wild‐type Arm/Smc5‐Nse2. Black line in each panel, native Arm/Smc5‐Nse2 without bound DNA; blue line, 5 kb circular ssDNA/Smc5‐Nse2; pink line, 50b linear ssDNA/Smc5‐Nse2; and green line, 25b linear ssDNA/Smc5‐Nse2. Right, conformational changes of wild‐type Arm/Smc5‐Nse2 induced by different types of DNA quantified by changes in center of mass (CM, redshift) of fluorescence spectra upon titration of the ligand until signal change achieved saturation. Blue circles, 5 kb circular ssDNA/Smc5‐Nse2; pink diamonds and 50b linear ssDNA/Smc5‐Nse2. Inset, titration curves of 5 kb circular ssDNA/Smc5‐Nse2 reactions containing 0 or 100 mM NaCl (solid and hollow symbols). Below, table of the dissociation constants of the curves.
3. Degree of conformational changes in CM induced by 60b linear ssDNA. Blue bar, Arm/Smc5‐Nse2 (wild‐type); gray bar, Arm/Smc5‐Nse2 (K333E/K344E); green bar, Arm/Smc5‐Nse2 (K764E); pink bar, Arm/Smc5‐Nse2 (K333E/K344E/K764E); red bar, Arm/Smc5‐Nse2 (K743E/K745E). Reactions were performed at least in three different independent experiments. Data values are mean ± s.d. and n = 3 technical replicates. Significance was measured by a two‐tailed unpaired t‐test relative to wild‐type. ****P < 0.0001.

Taking advantage of the lack of tryptophan residues in the ARM region of Smc5, we could follow structural changes in Nse2 by Trp‐intrinsic fluorescence, by measuring a redshift of fluorescence emission upon DNA binding. The two tryptophan residues of Nse2, Trp109 and Trp154, are located in opposite ends of the Nse2 structure and buried in the helical interface with the ARM coiled coil structure (Fig EV3). Binding of different types of DNA molecules, such as circular ssDNA (5 kb) and small ssDNA molecules (25 and 50b), produce a similar significant redshift of the Trp‐fluorescence emission, indicative of a conformational change. In all cases, a dose‐dependent curve of the fluorescence emission is observed in titration experiments using increasing concentrations of DNA (Fig 4B). Interestingly, the equilibrium constants of the saturation curves between the 50b oligonucleotide and long ssDNA are quite similar (2.7‐fold increase). These results are in agreement with our in vitro activity assays for the different types of DNA molecules.

Figure EV3. Conformational changes in Arm/Smc5‐Nse2 followed by tryptophan intrinsic fluorescence (raw data).

1. Ribbon representation of the complex between the ARM domain of Smc5 (yellow and orange) and Nse2 (magenta) (PDB 3HTK). Tryptophan residues in Nse2 are labeled and shown in stick representation (blue). Zinc atom in the Nse2 RING domain is depicted as a yellow sphere.
2. Intrinsic Trp‐emission spectra of wild‐type Arm/Smc5‐Nse2; K333E/K344E; K764E; K333E/K344E/K764E; and K743E/K745E mutants. Colored thick lines, native Arm/Smc5‐Nse2 without bound ligand (no 60b linear ssDNA); black lines, emission spectra after addition of DNA in a concentration able to induce the half of the maximal transition determined by titration experiments.
3. Values of center of mass before and after DNA addition. These data were used to calculate the ΔCM presented in Fig 4C.

Moreover, four different Arm/Smc5‐Nse2 KE mutants can reduce significantly the Trp‐fluorescence emission when compared to the wild‐type form under similar experimental conditions (Figs 4C and EV3), probably indicating a loss of affinity between the DNA molecule and the complex. Interestingly, NaCl‐concentration dependence could be observed in either the Trp‐fluorescence emission signal (Fig 4B, inset), as well as in the SUMO conjugation assays (Fig EV2). All our results suggest that the electrostatic interaction between the DNA and the ARM domain (DNA sensor), probably through the interaction between the negatively charged phosphate groups and positively charged lysine residues, produces a structural change that results in an enhancement of the SUMO conjugation activity of the Nse2 E3 ligase. Notably, a highly negatively charged small polymer such as enoxaparin (a low molecular weight heparin; Lima & de Prat‐Gay, 1997) can also stimulate the SUMO E3 ligase activity of Nse2/Smc5 complex in our in vitro conjugation assays (Fig EV4), mimicking the non‐specific charged‐based interaction of the DNA molecules with the DNA sensor of Smc5.

Figure EV4. Enhancement of the SUMO E3 ligase activity of Smc5‐Nse2 upon binding to enoxaparin.

1. Time‐course of the SUMO2 conjugation in the presence of increased concentration of enoxaparin using recombinant Full‐length/Smc5‐Nse2 or ARM/Smc5‐Nse2 complex. The substrate utilized was the C‐terminal kleisin domain of Nse4 (cNse4). Reaction was run at 30°C and stopped at indicated minutes by adding SDS‐loading buffer. Φ stands for the 5 kb virion ssDNA (5 nM) used as positive control. Labels indicate the bands in the SDS–PAGE of the proteins in the reaction mixture. N‐S2, cNse4‐SUMO2; N‐2S2, cNse4‐2SUMO2; and pS2, polySUMOylation.
2. Western blot of the samples presented in panel (A). SUMOylated proteins were immunodetected by an anti‐SUMO2 antibody

Compromising DNA binding to Smc5 sensitizes yeast cells to DNA damage

To test the relevance of the DNA sensor in vivo, we generated different SMC5 expression plasmids containing KE mutations. We first focused our attention on K743 and K745 in coiled coil 2, as they seem to more strongly modulate the DNA‐dependent activation of Nse2 in vitro (Fig 3D). We introduced an smc5‐K743,745E expressing plasmid, as well as one expressing the smc5‐K743,745R allele (which does not change the positive charge of the residues), into a conditional smc5 mutant strain. Yeast growing spot analysis revealed that both mutants support the viability of the conditional allele under non‐permissive conditions, indicating that the overall function of the Smc5/6 complex is not substantially affected by the presence of the KE or KR mutations (Fig EV5). However, we noticed that while the K743,745E mutant rendered cells moderately sensitive to MMS, the K743,745R mutant did not. This suggests that the DNA damage sensitivity stems from a change to a negatively charged residue, rather than mutation of lysines. Despite K764 also seems to be important for DNA‐dependent sumoylation in vitro (Fig 3D), the sensitivity was not substantially increased in the triple smc5‐K743,745,764E mutant (smc5‐3KE) relative to the double smc5‐K743,734E mutant (Fig EV5). On the other hand, smc5‐K337,344E double mutant cells were not sensitive to DNA damage. Moreover, a quadruple smc5‐K337,K344,K354,K355E mutant (smc5‐4KE), containing two other KE mutations close to the RING domain of Nse2 (Fig 3C), did not substantially affect the DNA damage sensitivity. We next reasoned that, in the context of an integer Smc5/6 complex, positively charged residues in the Smc5 ARM region may act redundantly for interaction with DNA and DNA‐dependent activation of Nse2. We therefore generated an allele combining both sets of mutations in the ARM domain of Smc5, hereafter referred to as smc5‐7KE (K337,K344,K354,K355,K743,K745,K764E). Interestingly, growth of the smc5‐7KE expressing cells was severely impaired in the presence of MMS (Fig EV5).

Figure EV5. Counter‐charge mutation in various lysine residues of the positively charge patch in Smc5 lead to increased MMS sensitivity.

Growth test analysis of yeast strains carrying an endogenous SMC5 gene under the GAL promoter. Each strain expresses the indicated SMC5 allele form a centromeric plasmid; 10‐fold serial dilutions of the liquid cultures were spotted in solid media containing the indicated concentration of MMS, and pictures were taken after 48 h.

The sensitivity of KE mutant cells to MMS resembles that of Nse2 RING mutants (Andrews et al, 2005; Zhao & Blobel, 2005), suggesting that mutation of lysines in the DNA sensor of Smc5 alters the ability of yeast cells to activate the Nse2 SUMO ligase. The nse2‐CH allele carries the C200A and H202A RING‐disrupting mutations and renders cells sensitive to genotoxic stress (Branzei et al, 2006). To directly compare the sensitivity of both type of mutants (smc5‐KE and nse2‐CH), we integrated the KE alleles into the endogenous SMC5 locus (see Fig 5A for location of mutated lysines). As shown in Fig 5B, and in agreement with results from ectopically expressed Smc5 (Fig EV5), smc5‐K743,745E and smc5‐3KE cells display growth defects in the presence of MMS. The smc5‐7KE cells exhibit stronger MMS sensitivity, to levels similar to nse2‐CH mutant cells. In contrast, smc5‐K743,745R and smc5‐7KR mutant cells, which maintain the positive charge of the mutated residues, are not sensitive to MMS (Fig 5C). This indicates that the MMS sensitivity in smc5‐KE cells stems from counter‐charge mutations, rather than loss of lysine residues per se. We propose that lysines located in the ARM domain of Smc5 synergistically collaborate in the repair of MMS‐induced DNA damage.

Figure 5. A positively charged patch in Smc5 is required for DNA repair in vivo .

1. Ribbon representation of the complex between the ARM domain of Smc5 (yellow and orange) and Nse2 (red) (PDB 3HTK) (Duan et al, 2009), showing positions of mutated lysine residues covered with black asterisks (3KE: K743E/K745E/K764E; 4KE: K337E/K344E/K354E/K355E; and 7KE: K337E/K344E/K354E/K355E/K743E/K745E/K764E).
2. Growth test analysis of wild‐type, nse2‐CH, and the indicated smc5‐KE mutants; 10‐fold serial dilutions of the liquid cultures were spotted in YPD and pictures taken after 48 h.
3. Same as in (B) but using the nse2‐CH, smc5‐3KE, smc5‐K743,745E, smc5‐K743,745R, smc5‐7KE, and smc5‐7KR mutants.

A DNA sensor in Smc5 participates in sumoylation in vivo

In accordance with the DNA damage sensitivity assays on plates, pull‐down of 6xHis‐Flag‐tagged SUMO (HF‐SUMO) shows that Smc5 sumoylation is diminished in smc5‐3KE and smc5‐K743,745E mutants, while it is not affected in smc5‐4KE cells (Fig 6A). The reduced sumoylation in smc5‐K743,745E cells is not due to mutation of lysine residues, as the smc5‐K743,745R mutant shows wild‐type levels of Smc5‐SUMO (Fig 6A). Quantification of sumoylated species from SUMO pull‐downs shows that sumoylation significantly drops to about 60% in smc5‐3KE mutants (relative to wild‐type Smc5; Fig 6C). The smc5‐K743,745E mutation, but not the smc5‐K743,745R mutant, also affects sumoylation of another Nse2 target, the Sgs1 protein (a homologue of the Bloom's and Werner's syndrome genes and a member of the STR complex; Bermúdez‐López et al, 2016). This finding indicates that the DNA sensor is also required to increase sumoylation of protein targets outside the Smc5/6 complex (Appendix Fig S4).

Figure 6. A DNA sensor in Smc5 participates in sumoylation in vivo .

1. Protein extracts from exponentially growing wild‐type, smc5‐3KE, smc5‐K743,745E, smc5‐4KE, and smc5‐K743,745R cells were prepared under denaturing conditions; 6xHis‐Flag‐tagged SUMO (HF‐SUMO) was pulled down from protein extracts to purify sumoylated species and analyzed by Western blot. A strain with no HF‐tag was used as control; arrow points to unmodified form of the protein, vertical bar to sumoylated forms.
2. Left, same as in (A), but using wild‐type and two independent clones of smc5‐7KE; right, same as in (A), but using wild‐type, smc5‐7KE, and smc5‐7KR mutant cells.
3. Quantification of Smc5 sumoylated species in pull‐downs from the indicated smc5‐KE mutants, relative to wild‐type controls from at least three independent experiments. Boxes, 25–75% data range; whiskers, total data range; black bar, median; gray cross, mean. ****P < 0.0001; ***P < 0.001; **P < 0.01; *P < 0.05. 1‐way ANOVA, n = number of samples analyzed.
4. Co‐immunoprecipitation analysis of the Smc5‐Nse2 interactions. NSE2‐6HA cells were transformed with centromeric plasmids expressing the indicated SMC5 alleles, and protein extracts from exponentially growing cells subjected to anti‐HA immunoprecipitation. A strain with no HA tag was used as control.
5. Same as in (D), but using SMC6‐6HA cells.
6. Quantification of Smc5‐6HA signal from immunofluorescence on chromosome spreads prepared from exponentially growing cultures of the indicated genotypes. The mean value on wild‐type spreads was arbitrarily set to 1; each dot represents one nucleus; red line, median. One‐way ANOVA, n = number of nuclei analyzed. ****P < 0.0001.

Source data are available online for this figure.

On the other hand, the smc5‐7KE allele had a stronger effect on Smc5 sumoylation, which further dropped to about one‐third of wild‐type levels (Fig 6B and C). This effect does not simply depend on loss of lysine residues, as the smc5‐7KR mutant protein displayed higher levels of sumoylation than smc5‐7KE (Fig 6B, right panel). However, we also noted that sumoylation of the smc5‐7KR protein was lower than its wild‐type counterpart. We speculate that the partial reduction in the sumoylation levels of the smc5‐7KR protein might be due to loss of potential SUMO acceptor sites. To directly compare sumoylation levels in KE and nse2 mutant cells, we used an nse2ΔC mutant strain, carrying a deletion in the C‐terminal RING domain. As shown in Appendix Fig S4, SUMO pull‐down analysis indicates a low and similar level of sumoylation for both smc5‐7KE and nse2ΔC cells, relative to wild‐type cells, and a more modest reduction in the smc5‐3KE strain.

As KE mutations lay close to the Nse2 binding domain in Smc5, it is possible that the diminished sumoylation in smc5‐KE mutants is due to reduced Smc5‐Nse2 association. However, co‐immunoprecipitation experiments indicate that the smc5‐3KE and smc5‐7KE proteins bind efficiently to the Nse2 SUMO ligase (Fig 6D). On the other hand, while the smc5‐3KE mutant protein shows wild‐type levels of association with Smc6, smc5‐7KE competes less efficiently for binding to Smc6 (Fig 6E). This observation suggests that the smc5‐7KE allele might compromise the stability of the Smc5/6 complex.

Finally, to analyze whether the sumoylation defects of KE mutants stem from deficient recruitment to chromatin, we tested chromosomal association of Smc5‐6HA by immunofluorescence on chromosome spreads. As shown in Fig 6F, there are no significant changes in chromatin binding between wild‐type and mutant Smc5‐3KE proteins. These data suggest that DNA sensing by the positively charged patch in the coiled coil of Smc5 occurs after loading of the Smc5/6 complex onto chromatin. In contrast, the extended smc5‐7KE DNA sensor mutant shows significantly reduced chromatin association, relative to wild‐type Smc5 (Fig 6F). We speculate that this decrease might be due to the combined action of, among others, defective interaction with DNA in the ARM domain, loss of putative acceptor lysines in Smc5, and compromised stability of the Smc5/6 complex.

Overall, we conclude that K743 and K745 represent a minimal positively charged patch in the Smc5 molecule, which acts as a DNA sensor in yeast, able to interact with DNA and to promote the activity of the Nse2 SUMO ligase thus ensuing repair of MMS‐induced DNA damage.

Discussion

Many aspects of chromosome replication, repair, and segregation require post‐translational modification by SUMO to maintain the integrity of the genome (Bergink & Jentsch, 2009). A plethora of SUMO‐targeted chromosome‐associated proteins, many of them involved in nucleic acid metabolism, has been identified in recent proteomic screens (Tatham et al, 2011; Lamoliatte et al, 2014; Cubeñas‐Potts et al, 2015). Preferential sumoylation of chromosome‐associated proteins raises the mechanistic question of what promotes “on‐site sumoylation” (Ulrich, 2014; Sarangi & Zhao, 2015). It has been proposed that specific sumoylation of chromosome‐bound targets requires either the recruitment of the SUMO E3 ligase to chromatin or DNA‐dependent structural changes in the substrate to make it amenable to modification (Ulrich, 2014; Sarangi & Zhao, 2015). Here, we show a novel regulatory mechanism that can account for “on‐site” sumoylation, based on the stimulation of the Nse2 SUMO ligase in the Smc5/6 complex by DNA. Our proposed mechanism differs from a few other cases in which DNA binding affects SUMO conjugation, even in the absence of a E3 ligase, i.e., enhanced affinity of PARP‐1 for E2‐SUMO thioester upon DNA binding (Zilio et al, 2013); stimulation of SUMO conjugation of PCNA when loaded to DNA (Parker et al, 2008). In our model, the stimulation requires non‐specific binding of DNA to a positively charged patch in the coiled coil of Smc5 (DNA sensor), which subsequently mediates a conformational change in the SUMO ligase. This mechanism is ideal to promote sumoylation of Smc5/6‐dependent targets on site and to elude the modification of targets not directly engaged on chromatin.

Although it is formally unproven that modification of Smc5/6‐Nse2 targets occurs on chromatin, circumstantial evidence suggests that it is highly probable. For example, Smc5/6 binds to damaged loci such as double‐strand breaks (De Piccoli et al, 2006; Lindroos et al, 2006) and stalled replication forks (Bustard et al, 2012), but also to ribosomal DNAs repeats, telomeres, and centromeres (Torres‐Rosell et al, 2005; Lindroos et al, 2006; Pebernard et al, 2008b; Jeppsson et al, 2014); not surprisingly, most Smc5/6‐Nse2 targets are enriched at these sites. It is worth noting that DSBs, telomeres, and stalled replication forks are characterized by the accumulation of ssDNA. Smc5‐Smc6 heterodimeric molecules have a higher affinity for ssDNA than for dsDNA (Roy et al, 2015), suggesting that their localization to these sites, and the consequent upregulation of sumoylation, may be partly driven by the intrinsic affinity of the SMC core for ssDNA. Although both single‐ and double‐stranded DNA molecules are capable of potentiating sumoylation, our observations indicate that ssDNA has a somewhat stronger effect than dsDNA. Consequently, we speculate that the enhancement of the E3 ligase activity by DNA might be partially responsible for the upregulation of protein sumoylation upon DNA damage, as the latter normally induces the accumulation of ssDNA. Of note, Smc5‐Smc6 heterodimers have a higher affinity for ssDNA than the Rad51 recombinase (Roy et al, 2015), suggesting that they might even outcompete the later for binding to single‐stranded DNA fibers. It is also worth noting that the Nse1‐Nse3 subcomplex displays higher affinity for dsDNA than for ssDNA in vitro (Zabrady et al, 2016). Hence, further experiments will be required to assess whether ssDNA plays a preferential role in sumoylation of Smc5/6‐Nse2 targets.

One of the most striking effects observed in this study is the absence of DNA‐dependent effects on Nse2 ligase activity (Fig 2E and Appendix Fig S1). This finding indicates that docking to the Smc5/6 complex has a positive influence on Nse2‐dependent sumoylation. We currently cannot discard that other regions in the Smc5/6 complex, different from the DNA sensor described here, participate and/or modulate the effect of DNA on the E3 ligase activity. Still, we have shown that the ARM region in the Smc5 molecule is sufficient to confer the DNA dependency in vitro. Interestingly, the analysis of Smc5 truncations indicates that other domains in the molecule also modulate the E3 ligase, although in a completely different manner: Removal of the HINGE domain in Smc5 leads to a higher sumoylation level, suggesting that the hinge physically obstructs sumoylation (Fig 2). Thus, one attractive possibility is that the association of DNA with the hinge (Alt et al, 2017) primes the subsequent DNA‐dependent activation of the E3 ligase. This suggests a two‐step model for DNA association with Smc5, first through the HINGE domain and, after removal of the hindrance, with the DNA sensor in the ARM domain.

We have analyzed two different substrates to study the DNA‐dependent activation of the Nse2 SUMO ligase in vitro: (i) internal lysines of the Smc5 protein, which have already been reported to be intrinsically sumoylated by Nse2 (Zhao & Blobel, 2005); and (ii) the C‐terminal kleisin domain of Nse4 (cNse4), which binds to the ATPase HEAD domain of Smc5 and does not interact with either the ARM or the HINGE domains (Appendix Fig S2). Nse4 sumoylation proceeds in trans (in opposition to E3 auto‐sumoylation) and can thus be considered an external substrate (Pichler et al, 2017); this is particularly evident under conditions where Nse4 does not interact with either Smc5 or Nse2 (e.g., in ΔHead/Smc5 or Arm/Smc5 truncations; Fig 2). Sumoylation of Smc5 and cNse4 is equally enhanced by binding to different types of DNA (ssDNA and dsDNA) in a dose‐dependent manner. The proximity of the DNA sensor in Smc5 to the RING domain of Nse2 should facilitate the enhancement of the SUMO E3 ligase activity. In accordance, Trp‐intrinsic fluorescence and circular dichroism indicate structural remodeling of the Smc5‐Nse2 complex upon DNA binding. Interestingly, our DNA titration by circular dichroism is analogous to the CpG DNA binding to the Toll‐like Receptor 9 ectodomain (Latz et al, 2007), in both cases resulting in a similar DNA dose‐dependent modification of the CD spectra.

Crystallographic studies have proven the role of the RING domain of E3 ligases in the recruitment of the charged E2‐conjugating enzyme, whereby a simultaneous interaction with SUMO (or ubiquitin) and the E2 enzyme would enhance the catalytic reaction. Thus, it is feasible to conceive a structural reorientation of the RING domain in Nse2 that would facilitate either the interaction with the E2‐conjugating enzyme or the transfer to the substrate, as proposed here. Structural data on the rearrangement of the Smc5‐Nse2 structure in the presence of DNA will be extremely useful to prove it.

The phenotypes displayed by the KE point mutants in the Smc5 DNA sensor patch are reminiscent of mutants in the RING domain of Nse2, affecting cell growth in the presence of genotoxic agents. We propose that the K743 and K745 residues constitute a minimal DNA sensor involved in activation of the SUMO ligase. Although other lysines in the Smc5‐ARM domain might have a synergistic effect on DNA binding in vivo, the smc5‐7KE mutant might disturb natural lysine acceptor sites. Additionally, it is possible that the extended positively charged patch participates in other functions not directly related to sumoylation, including Smc5/6 complex stability; in accordance with this notion, the smc5‐7KE mutant protein shows reduced chromatin binding and reduced association with Smc6 (Fig 6E and F).

We propose that both KE and RING mutants compromise the enzymatic activity of the Nse2 SUMO E3 ligase, either by preventing the interaction with the E2 enzyme (in the case RING domain mutants) or by perturbing the association of DNA with the Smc5 DNA sensor (in DNA sensor patch mutants). It is worth noting that the KE DNA sensor mutants only disrupt regulation of the E3 by DNA and do not affect the basal activity of the E3 ligase (Fig 3D). We propose that the moderate DNA damage sensitivity of smc5‐K743,745E and smc5‐3KE cells corresponds to an Smc5/6 complex that has impaired DNA‐dependent enhancement of sumoylation but carries a functional E3 ligase; the higher MMS sensitivity of nse2 mutants would correlate to a situation of severely impaired sumoylation. We predict that most chromosome‐bound Nse2 targets (Potts et al, 2006; Almedawar et al, 2012; McAleenan et al, 2012; Bermúdez‐López et al, 2015, 2016; Bonner et al, 2016), including cohesin, the Sgs1‐Top3‐Rmi1 complex, and the Smc5/6 complex itself, will be regulated by this mechanism. Thus, we propose that a reduced sub‐optimal level of sumoylation in Nse2 targets accounts for the MMS sensitivity of smc5‐KE mutant cells.

In addition to the DNA‐dependent enhancement of the SUMO E3 ligase activity, we cannot underestimate the contribution played by the DNA molecule in bringing Smc5/6 molecules in close proximity to its substrates. In fact, we have observed a non‐specific binding of the C‐terminal kleisin domain of Nse4 (cNse4) to DNA in EMSA experiments (Appendix Fig S2), although this interaction is only observed at protein:DNA ratios much higher than those able to induce the structural rearrangement and activation of the ligase. Nevertheless, a scaffolding function of DNA cannot account for all the effects observed in vitro as: (i) Short DNA molecules should substantially reduce the probability of enzyme‐substrate co‐binding to the same DNA molecule, a prediction that is not observed when comparing 5 kb‐ and 50b‐long ssDNAs (Fig 1C); (ii) both sumoylation and conformational changes can be promoted by enoxaparin, a small negatively charged molecule, pointing to electrostatic interactions as activators of sumoylation (Fig EV4); and (iii) our in vivo data indicate that Smc5/6 complexes loaded on chromatin are less active when mutated in their DNA sensor (smc5‐3KE mutant in Fig 6A and F). Moreover, ATPase mutant Smc5/6 complexes show diminished SUMO E3 activity. The dependence on the ATPase activity is probably twofold: Binding to ATP promotes a conformational change in the Smc5‐Nse2 molecule that stimulates its SUMO ligase activity (Bermúdez‐López et al, 2015); additionally, binding to ATP regulates the association of the Smc5/6 holocomplex with DNA (Kanno et al, 2015) what, according to the results presented here, further enhances its SUMO ligase activity. Therefore, we speculate that the Smc5/6 complex can be first activated by ATP‐dependent remodeling of the molecule and loading onto DNA; this would facilitate contacts between DNA and the DNA sensor patch in Smc5, which would subsequently activate the SUMO E3 ligase activity of Nse2 (Fig 7). This mechanism should help to confine the activity of a critical E3 ligase toward substrates pre‐loaded on chromatin and bound to the same stretch of DNA. It is worth noting that the ATPase activity in bacterial SMC complexes may regulate the dynamic association with DNA by altering the conformation of the coiled coil domains (Bürmann et al, 2017). As the Nse2 SUMO ligase docks to the coiled coil domain of Smc5, it is possible that the ATPase in the Smc5/6 complex might not only alter the juxtaposition of the coiled coils (Alt et al, 2017; Bürmann et al, 2017), but also promote the concomitant activation of Nse2 after entrapment of DNA inside the Smc5/6 ring.

Figure 7. Model for ssDNA‐dependent activation of the SUMO E3 ligase Nse2.

Smc5/6 complex first uses its ATPase activity to associate with DNA. After loading onto DNA, binding of the positively charged patch (DNA sensor) in the Smc5 ARM domain further activates the Nse2 SUMO ligase. Nse1, Nse3, Nse5, and Nse6 subunits of the Smc5/6 complex are not shown in the figure.

In summary, we have revealed a novel mechanism regulating a SUMO E3 ligase activity both by localization, occurring only upon association with DNA, and by the structural rearrangement of the E3 ligase. Given the prominent role played by ubiquitin and ubiquitin‐like ligases in genome integrity, it will be very interesting to know if other E3s operating on chromosome‐associated proteins employ analogous mechanisms.

Materials and Methods

DNA substrates

Oligonucleotides sequences were randomly selected: 25b (ACCGCGCGCTTATTC AACAATGTTG), 34b (GACAGGATCCATGTCTAGTACAGTAATATCAAGG), 50b (TGCCATATTGACAAGACGGCAAAGATGTCCTAGCAATCCATTGGTGATCA), and 60b (CGCGGTCGACGGTTACCCATACGATGTTCCTGACTATGCGGCCTTGAACGATAATCCTAT). Circular single‐strand DNA, ΦX174 (5 kb), was purchased from New England Biolabs. pET28a (5 kbp) and pET‐DUET‐1 (5 kbp) circular double‐stranded DNA were purchased from Novagen.

Plasmids construction for overexpression in bacterial cells

For biochemical assays, open reading frame of full‐length SMC5 from Saccharomyces cerevisiae S288c was cloned into pET28a vector. COIL/PCOILS (Alva et al, 2016) was used to predict the boundaries of Smc5 Head, Arm, and Hinge domains. The three constructs generated by PCR amplification ΔHinge/SMC5 (Met1‐Thr366 and Arg737‐Asp1093 linked by a two amino acid linker peptide Gly‐Thr), ΔHead/SMC5 (Lys310‐Glu815), and Arm/SMC5 (Asp302‐Thr366‐Gly‐Thr‐Arg737‐Gln813) were then individually cloned into pET28a vector. By the use of the mutagenesis protocol, the following mutations were introduced into Arm‐SMC5 expression vector: (K333E), (K344E), (K757E), (K764E), (K770E), (K777E), (K337E/K344E), (K743E/K745E), (K337E/K344E/K764E), and (K743R/K745R). We also introduced 7KE (K337E/K344E/K354E/K355E/K743E/K745E/K764E) mutations into ΔHinge/SMC5 and ΔHead/SMC5. To allow co‐expression with SMC5 constructs, the ORF of full‐length budding yeast NSE2(MMS21) was cloned into pET15b vector.

Protein expression and purification

All Full‐length/SMC5, ΔHinge/SMC5, ΔHead/SMC5, and Arm/Smc5 recombinant proteins containing N‐terminal His6‐tag were co‐expressed with His6‐tag‐Nse2 in Escherichia coli Rosetta 2(DE3) cells (Novagen). Alternatively, His6‐tag‐Nse2 was expressed alone using the same cell system. Bacterial cultures were grown at 37°C to OD₆₀₀ = 0.6, before 0.5 M IPTG addition. Cultures were then incubated for 4 h at 30°C (Full‐length/Scm5‐Nse2, and Nse2) or for 16 h at 20°C (ΔHinge/Scm5‐Nse2, ΔHead/Scm5‐Nse2, and Arm/Scm5‐Nse2 wild type and mutants) and harvested by centrifugation.

Cell pellets were equilibrated in Lysis Buffer (20% sucrose, 50 mM Tris pH 8.0, 1 mM 2‐mercaptoethanol, 350 mM NaCl, 20 mM imidazole, 0.1% IGEPAL), and cells were disrupted by sonication. Cell debris was removed by centrifugation (40,000 g). Hexa‐histidine tagged proteins were purified by metal affinity chromatography using Chelating Sepharose Fast Flow resin (GE Healthcare) and eluted with 20 mM Tris pH 8.0, 350 mM NaCl, 1 mM 2‐mercaptoethanol, and 250 mM imidazole. Fractions containing the Nse2, Full‐length/Smc5‐Nse2, ΔHinge‐Smc5, ΔHead‐Smc5, or Arm/Smc5‐Nse2 (wild‐type and mutants) were further purified by gel filtration (Superdex 200 HiLoad; GE Healthcare) followed by ion‐exchange chromatography (Resource Q, or S for ΔHead‐Smc5; GE Healthcare).

Smc5‐Nse2 SUMOylation reactions

SUMOylation reactions of Nse2 (monomer), or Full‐length/Smc5‐Nse2, ΔHinge/Smc5‐Nse2, ΔHead/Smc5‐Nse2, and Arm/Smc5‐Nse2 (heterodimers) were performed in a reaction mix containing 40 mM HEPES (pH 7.5), 10 mM MgCl₂, 0.2% Tween‐20, 50 mM NaCl, 4 mM dithiothreitol, 2 mM ATP, 32 μM SUMO1 or 2, 2 μM cNse4 (C‐terminal kleisin domain, Ile246‐Asp402), 300 nM Sae1‐Sae2 (E1), 200 nM Ubc9 (E2), and the E3 (0.8 μM Nse2 alone, 0.38 μM Full‐length/Smc5‐Nse2, 0.48 ΔHead/Smc5‐Nse2, 0.42 μM ΔHinge/Smc5‐Nse2, or 1.25 μM Arm/Smc5‐Nse2). The same reactions were also performed in parallel with the addition of DNA substrates (5 kb circular ssDNA, 5 kb circular dsDNA‐, 60b‐, 50b‐, 34b‐, and 20b‐oligonucleotides) or enoxaparin at indicated concentrations. Samples were taken in different time intervals and stopped with SDS‐Sample loading buffer (0.25 M Tris–HCl buffer pH 6.8, 10% (w/v) SDS, 30% (v/v) glycerol, 0.7 M 2‐mercaptoethanol, and 0.05% bromophenol blue). For dose‐dependent experiments, increased DNA amounts (see Fig 1) were added to reaction tubes and aliquots were taken after 60 min and stopped with SDS‐Sample loading buffer. Products were verified by SDS–PAGE and visualized after SYPRO Ruby (Invitrogen) staining or by Western blotting with anti‐SUMO1 or anti‐SUMO2 (Sigma‐Aldrich), or anti‐T7 (Novagen). Single turnover experiment was performed as described (Yunus & Lima, 2005) with minor changes. The E2‐thioester was formed in a reaction mix that includes 20 mM HEPES (pH 7.5), 50 mM NaCl, 10 mM MgCl₂, 0.1% Tween‐20, 400 nM DTT, 100 nM Sae1‐Sae2 (E1), 1 μM Ubc9 (E2), and 500 nM of (Alexa488)‐labeled SUMO1. The reaction was initiated by the addition of 1 μM ATP and was incubated at 30°C for up to 15 min. Then, the reaction was quenched with the addition of 5 mM EDTA. To follow the thioester transfer mediated by the E3, we added to the reactions 1.25 μM Arm/Smc5‐Nse2 (E3) and 2 μM of cNse4 (substrate) in the absence or presence of 0.8 μM of 50b oligonucleotide. Products of the time‐course reactions were analyzed by SDS–PAGE and visualized by Alexa488 fluorescence emission in a Molecular Imager Versadoc MP4000 System (Bio‐Rad).

Electrophoretic mobility shift assay

Circular 5 kb ssDNA template was used in EMSA reactions as provided by the supplier using a similar protocol reported by Roy et al (2011). For DNA binding experiments, reaction mixtures contained 10 mM HEPES pH 7.5, DNA substrate (200 ng, 10 nM), and Arm/Smc5‐Nse2 (wild‐type or mutants) or kleisin domain of Nse4 (cNse4) in a 0‐ to 4,000‐fold molar excess over DNA; see figure legends for Arm/Smc5‐Nse2‐to‐DNA molar ratios. After incubation at 30°C for 40 min, the reactions were stopped by addition of an equal volume of 1.6% low melting point (LMP) agarose containing loading buffer (0.6% glycerol, 0.005% bromophenol blue final concentrations). Mixtures were loaded on 0.5% agarose‐TAE gels, and the DNA was resolved by electrophoresis at 2.5 V/cm for 19 h at 4°C. DNA bands were visualized after staining the gel with GelRed (Biotium) and documented by Gel Doc XR System (Bio‐Rad). The experiments were performed in triplicate.

Spectroscopic measurements

Circular dichroism (CD) measurements were performed using a Jasco‐715 spectropolarimeter. Arm/Smc5‐Nse2 (wild‐type and mutants) were diluted in 4 mM HEPES pH 7.5 buffer to a final concentration of 3 μM. Circular 5 kb ssDNA was titrated at various ratios, and the contribution of buffer and DNA to the spectra was subtracted for background correction. Spectra were recorded as an average of ten scan accumulations with a scan rate of 200 nm/min, with 0.1 nm steps in a 1‐mm path length cuvette at 30°C. Raw data were converted to mean residue ellipticity.

Intrinsic fluorescence spectra were recorded using a Jasco FP‐8200 spectrofluorimeter. Tryptophan emission spectra were obtained by setting the excitation wavelength at 295 nm and collecting emission in the 310–400 nm range. These spectra were quantified as the center of spectral mass (ν) according to equation: ν = Σν _i F _i/ΣF _i where F _i stands for the fluorescence emission at a given wavelength (ν _i) and the summation is carried out over the range of appreciable values of F (Silva et al, 1986; Foguel & Silva, 1994). Arm/Smc5‐Nse2 (wild‐type and mutants) were diluted to achieve 1 μM in 20 mM HEPES pH 7.5 buffer containing 0 or 0.1 M NaCl. In the titration experiments, circular ssDNA (5 kb) was tittered up to 5 nM, and the linear ssDNA (50b) was tittered up to 5 μM. In the measurements with Arm/Smc5‐Nse2 mutants, a linear ssDNA (60b) was added to the reaction mixture to a final concentration of 0.2 μM in 20 mM HEPES pH 7.5 buffer. The temperature was maintained at 30°C. Kd were calculated with GraphPad Prism4 using ligand binding mode with triplicates (Motulsky & Christopoulos, 2003).

Yeast growth test analysis

YTR31 cells (MATa his3200 leu20 met150 trp63 ura30 GAL‐3HA‐SMC5:KAN bar1::URAca), carrying the indicated SMC5 alleles expressed from a centromeric plasmid, were inoculated in minimal dropout media containing galactose at 25°C from freshly streaked plates until the culture reached exponential phase. In Fig 5, derivates from YPM2620 (MATa his3Δ1 leu2Δ0 met15Δ0 ura3Δ0 6xHis‐FLAG‐smt3::kanMX6) were grown to exponential phase in YPD. 10‐fold serial dilutions from a culture at OD₆₀₀ = 0.5 were spotted as 3 μl drops onto YPD solid media and incubated at 30°C for 2 days. As indicated, plates were supplemented with the indicated concentrations of methyl methanesulfonate (MMS).

SUMO pull‐down

Pull‐down analysis of sumoylated proteins was performed essentially as described (Colomina et al, 2017). To purify sumoylated proteins, the budding yeast SUMO gene (SMT3) was tagged N‐terminally with a 6xHis‐Flag epitope. 100 ODs cultures were collected, and cells were mechanically broken in 6 M guanidine chloride. Protein extracts were incubated with Ni‐NTA beads in the presence of 15 mM imidazole overnight at room temperature. Beads were extensively washed with 8 M urea, and bound proteins were eluted with SDS–PAGE loading buffer. Antibodies used in Western blot analysis are anti‐HA (3F10; Roche). To integrate the KE alleles into the genome, we fused an SMC5 sequence containing the desired smc5‐KE mutation to a 6HA‐NAT selection marker. Clones were screened by anti‐HA Western blot, and the presence of the KE mutations was confirmed by sequencing.

Immunoprecipitation and Western blotting

For co‐immunoprecipitation analysis, protein extracts from a 100 OD exponentially growing culture were prepared in EBX as previously described (Almedawar et al, 2012). HA‐tagged proteins were immunoprecipitated using anti‐HA Affinity matrix (Roche).

Detection of Smc5 on chromosome spreads by immunofluorescence

Exponentially growing cultures (5 ODs) were spheroplasted as previously described (Grubb et al, 2015). After spheroplasting, 5 μl of gently resuspended spheroplasts was pipetted onto a glass slide before sequential addition of 10 μl fixative (3.4% sucrose, 4% paraformaldehyde) and 20 μl of 2% lipsol as detergent. One minute later, 20 μl of fixative was added again in a swirling motion. A pipette tip on its side was used to gently spread the nuclei and chromosome spreads were air‐dried overnight. For immunostaining, spreads were washed with PBS for 10 min in coplin jars and incubated with blocking solution (PBS, 2% milk, 5% BSA). Antibodies were incubated in blocking solution for 1 h in a humidity chamber; monoclonal rat anti‐HA (3F10, Roche) was used at 1:500 dilution to detect Smc5‐6HA, followed by a 1:1000 dilution of Alexa488‐labeled anti‐rat antibody. After air‐drying, DAPI was added in mounting media. For fluorescence microscopy, series of z‐focal plane images were collected with a DP30 monochrome camera mounted on an upright BX51 Olympus fluorescence microscope.

Analysis of SUMOylation efficiencies

Reaction rates of SUMOylation of cNse4 were quantified with ImageJ 1.49v software using the built‐in gel‐analyzer function (Schneider et al, 2012). Briefly, relative band intensities (fraction of SUMOylated protein) were calculated using a graphical method that involves generating lane profile plots, manually delineating peaks of interest, and then integrating peak areas. The calculations were performed at least in three different independent experiments (Fig EV1). Data values are mean ± s.e.m.; and n = 3 technical replicates. Significance was measured by a two‐tailed unpaired t‐test relative to wild‐type. *P < 0.05, **P < 0.01, ***P < 0.001.

Analysis of Smc5‐SUMO from pull‐downs

Bands corresponding to Smc5 sumoylated species were quantified with ImageLab (Bio‐Rad). Smc5‐SUMO signals for each smc5 mutation were made relative to the wild‐type. Four, two, and one independent clones were analyzed for smc5‐7KE, smc5‐3KE, and smc5‐4KE, respectively, in at least three independent experiments. Statistical significance was measure by one‐way ANOVA as described in the figure legend. P‐values of statistical significance are represented as ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05.

Degree of conformational changes

ΔMRE at 222 nm (%) induced by 5 kb circular ssDNA was quantified by ligand titration until signal change achieved saturation, which was assumed to displays 100% of change. Reactions for wild‐type protein were performed in three different independent experiments, and data values are mean ± s.d.

ΔCM(%) of fluorescence in the presence of 60b linear ssDNA was quantified by calculating the difference between the Center of Mass of the samples (wild type and mutants) before and after addition of a DNA concentration able to induce half of transitions seen in titration experiments using wild‐type protein (D_1/2 = 4 nm). This value was considered as 100% of change. Reactions were performed in three different independent experiments. Data values are mean ± s.d. and n = 3 technical replicates. Significance was measured by a two‐tailed unpaired t‐test relative to wild‐type. ****P < 0.0001.

Author contributions

NV, JL, and DR conducted all SUMO conjugation in vitro experiments. NV conducted the circular dichroism and fluorescence analysis. EI, NC, and JT‐R conducted the in vivo yeast experiments. DR and JT‐R analyzed the results and wrote the paper. DR and JT‐R conceived the idea for the project.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Review Process File

Source Data for Figure 6

The EMBO Journal (2018) 37: e98306