Submolecular video-imaging of the Smc5/6 complex topologically bound to DNA

Abstract

Structural maintenance of chromosomes (SMC) complexes, including cohesin and condensin, are ring-shaped ATPases that organize chromosome architecture by holding DNA segments together. However, the dynamic mechanisms underlying their DNA binding and processing remain unclear. Using high-speed atomic force microscopy, we directly visualized the eukaryotic Smc5/6 complex on DNA at submolecular resolution. Smc5/6 adopted an “I”-shaped conformation with closely aligned SMC arms and transitioned to an “O”-shaped open ring upon ATP binding. ATP-bound Smc5/6 stably associated with DNA through its ATPase head domains, whereas ATP hydrolysis promoted DNA entrapment within the SMC compartment and repositioning near the hinge. Smc5/6 also tethered two DNA segments together and stabilized twisted DNA structures, leading to DNA compaction. Our findings provide a visual demonstration of how an SMC complex employs its ring architecture to facilitate distinct DNA binding modes, deepening our understanding of the ATP-dependent dynamic transactions of Smc5/6 during DNA binding.

SMC complexes organize higher-order genome architecture through their unique DNA-binding activities. Using HS-AFM the authors reveal the submolecular dynamics of Smc5/6 on DNA, providing mechanistic insight into this ATP-dependent interaction.

Introduction

Members of the structural maintenance of chromosomes (SMC) proteins, including eukaryotic cohesin (Smc1/3), condensin (Smc2/4) and the Smc5/6 complex, play fundamental roles in chromosome organization and segregation^1–4. SMC complexes are thought to fold genomic DNA by mediating physical interactions between two distinct chromosomal segments. SMC complexes are characterized by their ring-like architecture, which is composed of a pair of SMC ATPase subunits and a kleisin subunit that further interacts with various regulatory subunits. Studies on purified SMC complexes have revealed diverse DNA transactions driven by ATP binding and hydrolysis. Cohesin, in particular, is well known for its ability to topologically entrap DNA within its ring, which underpins the establishment of sister chromatid cohesion^5,6. In addition, all eukaryotic SMC complexes, including Smc5/6, have been shown to extrude DNA loops in a continuous manner^7–10. This unique function may underlie the role of cohesin in shaping interphase genome architecture and of condensin in driving chromosome compaction during mitosis. DNA loop extrusion by prokaryotic SMC complexes has also recently been reported¹¹. Moreover, cohesin and condensin have been shown to tether two separate DNA strands, offering an alternative molecular mechanism for establishing DNA–DNA interactions^12–14. However, it remains unclear how SMC complexes bind to DNA and exploit their ring-like architecture to carry out their functions, particularly from a dynamic structural perspective.

Smc5/6 is essential for accurate chromosome segregation during both mitosis and meiosis^2–4. This is primarily due to its role in eliminating harmful DNA structures that can hinder DNA replication, including repair intermediates of DNA damage through homologous recombination, stalled replication forks, DNA–RNA hybrids (R-loops), DNA supercoils, and DNA intertwined between sister chromatids^15–19. Purified Smc5/6 binds directly to DNA and stabilizes DNA supercoils in an ATP-dependent manner^20,21. Smc5/6 has been shown to exhibit a higher affinity for single-stranded and double-stranded DNA junctions formed during DNA replication and recombinational repair, suggesting that it can recognize higher-order DNA structures^22,23. In addition, like cohesin and condensin, Smc5/6 is capable of topological DNA entrapment²⁴. These unique DNA interactions appear to be directly linked to the functions of Smc5/6 in a wide range of chromosomal processes. Therefore, it is crucial to elucidate the structural dynamics of Smc5/6 on DNA in order to gain further molecular insights into how these complexes enable various DNA interactions.

Recent studies using single-molecule fluorescence imaging have revealed the molecular dynamics of Smc5/6 but lacked the structural details^10,19,22,23. Cryo-electron microscopy studies have provided atomic-level structures of Smc5/6, albeit with limited dynamic information^25–28. High-speed atomic force microscopy (HS-AFM) has rapidly developed as a unique approach to visualize the nanodynamics of biomolecules in an active state in real space under physiological buffer conditions^29,30. This approach has revealed the molecular behavior of SMC proteins including cohesin, condensin and the SMC-related MRN DNA repair complex^31–33. By contrast, the dynamics of DNA-bound SMC proteins have not been adequately addressed, as the previous studies have primarily used a conventional dry AFM or samples strongly adsorbed electrostatically to charged AFM stages^32,34,35.

Here we visualized the dynamics of Saccharomyces cerevisiae Smc5/6 topologically bound to DNA using HS-AFM combined with a mica-supported lipid bilayer system, which has been successfully applied to investigating the intrinsic nature of proteins bound to scaffold molecules such as DNA and cytoskeletons^36–38. Our observations provide deeper mechanistic insights into the ATP-driven functions of Smc5/6.

Results

The Smc5/6 core complex adopts both I- and O-shaped conformations

The Smc5/6 complex consists of eight subunits, including a pair of SMC subunits (Smc5 and Smc6) and six non-SMC element (NSE) subunits that form a stable dimer via hinge domains. The other ends of the extended SMC coiled-coil arms are ATPase heads that engage upon ATP binding. The Nse4 kleisin subunit bridges the two SMC heads through asymmetric interactions, forming a closed ring topology. Nse1 and Nse3 stably interact with Nse4, while Nse2 associates with Smc5 at the center of its coiled-coil arm. The core of Smc5/6 consists of these six subunits (hereafter referred to as the Smc5/6 hexamer) (Fig. 1a). Further interaction with the Nse5–Nse6 subcomplex, which regulates topological DNA loading of the core complex, results in the formation of an octameric complex (hereafter referred to as the Smc5/6 octamer).

Fig. 1. Structural evaluations of the budding yeast Smc5/6 hexamer complex on mica.

a Schematic of the Smc5/6 hexamer. b Representative overview HS-AFM images of Smc5/6 hexamer molecules on a mica imaging stage. Wild-type (WT) hexamers without ATP (b1), with ATP (b2), and the EQ mutant molecules with ATP (b3). b1–b3: Scan area, 250 × 250 nm² at 200 × 100 pixels². Imaging rate, 1.5 s per frame. The experiments were independently repeated three times with similar results. The molecules labeled “mobile” were identified as the I-form in adjacent frames but appear blurred here because all molecules are in motion. c Enlarged views of representative I-form and O-form wild-type hexamers (from b2). d Abundance ratios of different conformations calculated from individual molecules. The Data are shown as means ± SEM. obtained from three independent experiments.

We purified the budding yeast Saccharomyces cerevisiae Smc5/6 complex as both octameric and hexameric forms, which exhibited DNA-stimulated ATPase activity as previously reported (Supplementary Fig. 1)^39–41. These purified complexes were subjected to HS-AFM observation in solution on freshly cleaved bare mica, a common imaging stage for supporting biomolecules. A buffer solution containing a high concentration of salt (300 mM KOAc) was used to minimize the disruption of Smc5/6 during HS-AFM observation. We first observed several hexamer molecules simultaneously within the same viewing window (Fig. 1b). This allowed us to visualize two types of Smc5/6 hexamers; an arm-open, ring shape (O-form) and an arm-closed, extended shape (I-form) (Fig. 1c). We quantified the occurrence frequencies of these two structures based on the conformations observed within the initial ten frames after the molecules appeared in the AFM images. The I- and O-form were classified based on whether the mica surface was exposed between the two SMC arms, as described in Supplementary Fig. 2. More than 80% of Smc5/6 hexamers were in the I-form in the absence of ATP, with a slight increase in the O-form molecules (~28%) in the presence of ATP (Fig. 1d).

We also observed the EQ mutant hexamer, which is capable of ATP binding but defective in hydrolysis (Supplementary Fig. 1b,d)^40,41. Approximately 60% of the EQ mutant complex displayed the O-form in the presence of ATP, whereas most of the molecules remained in the I-form without ATP (Fig. 1b,d). This suggests that the engagement of the ATPase heads, mediated by ATP binding, induces the conformational change to O-form of the Smc5/6 core complex. In addition to hinge and head regions, an additional globular mass was observed at the middle of one of the SMC arms, corresponding to Nse2, which is known to bind the Smc5 coiled-coil region⁴². This characteristic domain thus allowed us to distinguish Smc5 from Smc6 (Fig. 1c). In contrast to the Smc5/6 hexamer, our attempt to visualize the octamer molecules failed because Nse5–Nse6 appeared to become dissociated from the Smc5/6 hexamer core complex upon adsorption onto mica (Supplementary Fig. 3).

As DNA stimulates the ATPase activity of Smc5/6 (Supplementary Fig 1)^40,41, we wondered if DNA also influences its conformation. Therefore, we performed the HS-AFM imaging in a buffer containing an excess of short DNA molecules. Although Smc5/6 hexamer bound to 20-bp DNA as judged by an electrophoretic mobility shift assay, we observed only a slight increase in the O-form population under HS-AFM (Supplementary Fig. 4). In addition, as the large amount of short DNA adsorbed onto the mica surface, we cannot exclude the possibility that nearby adsorbed DNA indirectly influenced the observed Smc5/6 conformations.

ATP-dependent conformational changes of the Smc5/6 core complex

The I-form observed in our HS-AFM analyses is reminiscent of the cryo-electron microscopy (cryo-EM) structure of the budding yeast Smc5/6 hexamer in the absence of ATP (Fig. 2a,b)²⁵. Accordingly, the pseudo-AFM image generated from the cryo-EM structure by AFM image simulation (Fig. 2c, wild-type, apo state, see Methods) was virtually identical to the experimentally observed Smc5/6 hexamer using HS-AFM (Fig. 2a), with similar domain-to-domain distances, including hinge to Nse2 and Nse2 to ATPase heads (15 nm and 23 nm for the experiment, 16 nm and 23 nm for the pseudo-AFM image, respectively). In contrast to the apo state (I-form), the previously published cryo-EM structure of the ATP-bound state, in which the ATPase heads and Nse1/Nse3 clamp DNA, was confined to the ATPase head region²⁶. This prompted us to construct an atomic model of the O-form to simulate the structural transition from the apo state (I-form) to the ATP-bound state (O-form), guided by our HS-AFM observations of the O-form molecules (Fig. 2d) and previously published cryo-EM structures of both states^25,26.

Fig. 2. Molecular prediction of ATP-dependent conformational change of Smc5/6 hexamer.

a A representative HS-AFM zoomed image of the I-form Smc5/6 hexamer (wild-type without ATP). Scan area, 100 × 100 nm² at 160 × 80 pixels², cropped to 73 × 56 nm². Imaging rate, 0.5 s per frame. The experiments were independently repeated three times with similar results. b, Atomic model of the budding yeast Smc5/6 hexamer in the apo state from its cryo-EM structure (EMD-13895, PDB-7QCD) (related to Supplementary Movie 1). c A pseudo-AFM image of the Smc5/6 hexamer in the apo state using the atomic model from b. In panels a and c, below the images, the corresponding line profiles of the molecular height across the two red arrowheads are presented. d A representative HS-AFM zoomed image of the O-form Smc5/6 hexamer (EQ mutant with ATP). Scan area, 100 × 100 nm² at 160 × 80 pixels², cropped to 55 × 55 nm². Imaging rate, 0.5 s per frame. The experiments were independently repeated three times with similar results. e A predicted atomic model of Smc5/6 hexamer in the ATP-bound state (see Supplementary Fig. 5 and Methods) (related to Supplementary Movie 1). f A pseudo-AFM image of the Smc5/6 hexamer in the ATP-bound state using the atomic model from e. In panels d and f, below the images, the corresponding line profiles of the molecular height across the two red and blue arrowheads are presented, respectively.

Both the I- and O-form molecules observed by HS-AFM exhibited consistently extended SMC arms, leading us to speculate that the SMC arms can be considered as rigid chains. Therefore, we constructed a coarse-grained model based on a freely-jointed chain approach⁴³, using the published cryo-EM structure of the apo-state Smc5/6 (Supplementary Fig. 5, Supplementary Table 1, 2)²⁵. The SMC arms are composed of long antiparallel α-helices with several discontinuities (joints). Our HS-AFM observations of the O-form molecules showed that both SMC arms bent at obtuse angles near the middle. Consistent with this interpretation, Smc6 contains two joints where the two antiparallel peptide strands are disordered at the same positions in the middle of its arm. Smc5 also contains a similar joint, located near Nse2 on the hinge side. Therefore, we placed one and two flexible points on Smc5 (joint 3) and Smc6 (joint 1 and 2) arms, respectively (Fig. 2e). We also placed two flexible points at the bases of the arms, near the hinge domains. A harmonic oscillator model-type force was applied between the ATPase heads to recapitulate the ATP-dependent head engagement.

Next, we connected these flexible points with rigid chains to generate a simple framework. The I-form framework was converted into the O-form framework by applying virtual external pulling forces (Fig. 2e, Supplementary Fig. 5, Supplementary Movie 1), guided by the ATP-bound cryo-EM structure²⁶. The conversion to the O-form was only observed when the forces were applied to joint 1 and 3 in opposite directions (i.e., pulling apart). Finally, we generated the pseudo-AFM image based on the simulated O-form atomic model (Fig. 2f, see Methods). This modeling produced an image essentially identical to the experimentally observed Smc5/6 O-form. Together, our molecular simulation and HS-AFM analysis strongly suggest that Smc5/6 undergoes dynamic structural transitions between the I-form and O-form during ATP hydrolysis cycles.

To directly track the ATP-dependent conformational changes of the Smc5/6 hexamer, we further scanned these molecules over time within smaller scanning ranges. While the Smc5/6 hexamer primarily remained in the I-form without ATP, it exhibited transitions between I-form and O-form in the presence of ATP (Fig. 3a, Supplementary Movie 2). To quantify these conformational changes, we measured the arm–arm distances between Smc5 and Smc6 at the center of their arms in each time frame (Fig. 3b,c). The histograms derived from 10 molecules of Smc5/6 hexamers showed a large single peak in the absence of ATP, whereas two distinct peaks were observed in the presence of ATP. Gaussian fitting of the histograms was performed to calculate the mean arm–arm distances in each condition: 5 nm (I-form) in the absence of ATP, and 8.5 (I-form) and 19 nm (O-form) in the presence of ATP. Like the wild-type, the EQ mutant exhibited two distinct peaks in the histogram in the presence of ATP. However, in contrast to the wild-type, the majority (75%) of the EQ mutants were assigned to species with larger arm–arm distances with a mean distance of 18 nm (O-form) (Fig. 3b,c, Supplementary Movie 2). Note that we only observed subtle differences in head–hinge distances between the I- and O-forms (Supplementary Fig. 6). These results provide further support for the conclusion that ATP binding and subsequent hydrolysis regulate the conformational changes of the Smc5/6 core complex.

Fig. 3. Real-time tracking of structural dynamics of Smc5/6 on mica.

a Successive HS-AFM images of wild-type hexamers without ATP (a1), with ATP (a2), the EQ mutant hexamers with ATP (a3) on a mica imaging stage (related to Supplementary Movie 2). a1 and a3: Scan area, 100 × 100 nm² at 160 × 80 pixels², cropped to 65 × 60 nm². Imaging rate, 0.5 s per frame. a2: Scan area, 150 × 150 nm² at 160 × 80 pixels², cropped to 75 × 63 nm². Imaging rate, 0.7 s per frame. b Real-time traces of arm–arm distance between Smc5 and Smc6, analyzed from the molecule shown in panel a. c Distributions of arm-to-arm distances extracted from successive HS-AFM images under each condition. The total number of frames and analyzed molecules are indicated in each chart. The values in brackets next to each component label represent the fitted mean values of the Gaussian distributions. d Successive HS-AFM images of the pentamer lacking Nse2 (related to Supplementary Movie 3). Representative I/O- and B-form molecules are shown. Scan area, 120 × 120 nm² at 200 × 100 pixels², cropped to 62 × 65 nm². Imaging rate, 0.63 s per frame. e As in c, but the analyses were done with the I/O- and B-forms of the pentamer lacking Nse2. f Quantification of SMC arm bending angles. The red and blue histograms represent the probability density distributions of the bending angles of the Smc5 and Smc6 arms, respectively (see also Methods). The mean angle values and their standard deviations are shown as μ and σ above each peak.

Observation of Smc5/6 lacking Nse2

Although the Smc5/6 hexamer adopted two distinct shapes (I- and O-forms), its SMC arms remained extended and never folded back. This characteristic differs from cohesin and condensin, whose arms can sharply fold back at the “elbow”, displaying a ‘B’-form structure^32,33. Unlike cohesin and condensin, the Smc5/6 complex contains Nse2, which stably associates with the Smc5 arm⁴³, suggesting that Nse2 constrains the bending of the Smc5 arm.

To investigate the effect of Nse2 on Smc5/6 conformations, we purified Smc5/6 lacking Nse2 (the pentamer, Supplementary Fig. 7a). Like the WT complex, the majority of the pentamer molecules displayed the I- and O-forms (Fig. 3d, Supplementary Fig. 7b). In addition, although at low frequency (3%), molecules with the two arms bent toward each other were also observed, which we defined as the B-form (Fig. 3d and Supplementary Fig. 7b). Note that this B-form conformation had not been detected in the presence of Nse2. We further scanned individual pentamer molecules over time to measure their arm–am distances (Fig. 3e). Although the I- and O-form molecules frequently transitioned between the two conformations (i.e., I-to-O or O-to-I transitions), the B-form tended to remain in the bent conformation (Supplementary Movie 3). The histogram generated from the I/O-form molecules revealed two peaks corresponding to the I- and O-forms at 7 nm and 17 nm, respectively. By contrast, the B-form exhibited larger arm–arm distances than the I/O-forms, with a distinct peak at 33.5 nm. These results suggest that Nse2 constrains the conformation of Smc5/6. We note that the B-form is likely to represent a structurally unstable state because of its low frequency and distinct conformational dynamics.

To further investigate the mechanical effect of Nse2 on arm conformation, we quantified the bending angles of both SMC arms in the presence or absence of Nse2 (Fig. 3f and Supplementary Fig. 8). For this analysis, we focused on I/O-form molecules, as these were also predominantly observed in the absence of Nse2. Arm angles were defined as follows: when the head and hinge domains were positioned downward and upward, respectively, bending toward the left was assigned a negative angle, whereas bending toward the right was assigned a positive angle (see the left schematic in Fig. 3f). In the presence of Nse2 (hexamer), the Smc5 arm showed a peak at negative angles below 0°, whereas the Smc6 arm exhibited a broader distribution. The standard deviation of the Smc6 angles was approximately two-fold greater than that of Smc5 (Fig. 3f, WT + ATP). In the absence of Nse2 (the pentamer), the Smc5 and Smc6 arms could not be distinguished. Nonetheless, the standard deviations of both arms increased approximately two-fold compared with that of Smc5 in the presence of Nse2 (hexamer) (Fig. 3f, pentamer+ATP). These results suggest that Nse2 contributes to maintaining the Smc5 arm at a defined angle, whereas the Smc6 arm is more flexible.

Observation of the Smc5/6 complex bound to DNA

To visualize DNA-bound Smc5/6 by HS-AFM, we first reconstituted the DNA loading of the Smc5/6 using previously established assays (Fig. 4a)⁶. Since Nse5–Nse6 has been reported to be essential for the efficient topological DNA loading of Smc5/6²⁴, we used the Smc5/6 octamer for the reconstitution of DNA loading. Following incubation with circular DNA, Smc5/6 was retrieved by immunoprecipitation through a 3×Pk tag on Smc6, and the bound DNA was analysed by agarose gel electrophoresis after a high-salt wash. DNA recovery was observed when Smc5/6 was incubated in the presence of ATP, but not with ADP, cytidine triphosphate (CTP), or in the absence of nucleotides (Fig. 4b). We also found that DNA loading depended on Nse5–Nse6, as efficient DNA recovery was not observed with the Smc5/6 hexamer alone, whereas the addition of separately purified Nse5–Nse6 restored DNA loading in the presence of ATP (Supplementary Fig. 9a). Single-site restriction digestion of the captured DNA released it from the bead-bound Smc5/6, confirming the topological nature of Smc5/6 loading in the presence of ATP (Supplementary Fig. 9b). Similar results were observed with the EQ mutant Smc5/6 octamer, indicating that efficient ATP hydrolysis is not rate-limiting for topological Smc5/6 loading (Supplementary Fig. 9b,c). Consistent with this conclusion, the efficient DNA recovery with Smc5/6 was also observed with Smc5/6 in the presence of the non-hydrolysable ATP analog ADP-AlF_x (Fig. 4b), which has been reported to promote topological DNA loading of cohesin^44,45. Although another analog, ATP-γ-S, did not support DNA loading of Smc5/6, ATP-γ-S appears to be an imperfect ATP mimic for Smc5/6, similar to cohesin⁶.

Fig. 4. Visualization of DNA-bound Smc5/6 complex using the supported lipid bilayer stage.

a Schematic of DNA loading assay. b Representative agarose gel analysis of the experiment as depicted in a. The graph presents the quantification of Smc5/6-bound DNA (obtained from three independent experiments; means ± SD). c ATP-independent DNA binding of Smc5/6 was analyzed using an electrophoretic mobility shift assay (EMSA). The experiments were independently repeated three times with similar results. d Schematic of HS-AFM imaging using a positively charged lipid membrane as the imaging stage. e Overview scans of the DNA-bound wild-type Smc5/6 molecule, displaying the characteristic SMC ring structure (related to Supplementary Movie 4). Scan area, 350 × 280 nm² at 200 × 80 pixels², cropped to 260 × 206 nm². Imaging rate: 1 s per frame. The experiments were independently repeated three times with similar results. f Successive HS-AFM images of the interaction between two EQ mutants bound to a DNA molecule. The cyan arrows indicate the direction of molecular movement as the molecules interact with each other, moving closer together and further apart. Scan area, 220 × 176 nm² at 200 × 80 pixels², cropped to 138 × 93 nm². Imaging rate: 0.6 s per frame. The blue asterisks indicate an aggregated structure of the lipid. g Quantification of Smc5/6 molecules bound to DNA. The Data are shown as means ± SEM. obtained from three independent experiments. NC: nicked circular DNA, CC: covalently closed circular DNA.

To further assess DNA-binding activity, we performed an electrophoretic mobility shift assay using covalently closed, supercoiled DNA and nicked (relaxed) circular DNA (Fig. 4c). This assay clearly indicated a distinct DNA shift in the presence of the Smc5/6 octamer and ATP. A weaker but still noticeable DNA shift was also observed in the presence of ADP, suggesting ATP-independent, non-topological DNA binding mediated by the DNA-binding motifs of the Smc5/6 complex⁴⁶. Additionally, Smc5/6 exhibited a relatively higher affinity for supercoiled DNA than for relaxed DNA. These results are consistent with previous biochemical studies^20,24.

Having established ATP-dependent Smc5/6 loading in bulk reactions, we next sought to visualize individual Smc5/6 molecules bound to DNA. After loading Smc5/6 octamer onto circular DNA (1895 bp, ~644 nm) in the presence of ATP, the reaction mixture was applied onto the mica in a buffer containing potassium ions. However, under these conditions, multiple Smc5/6 molecules stacked on the DNA, as both DNA and Smc5/6 were adsorbed on the mica surface, making it difficult to track the dynamics of individual Smc5/6 on DNA (Supplementary Fig. 9d).

To overcome this difficulty, we adapted our previously developed supported lipid bilayer system, which features a weakly positively charged surface that adsorbs DNA while repelling proteins (Fig. 4d)^37,38. This approach successfully visualized both wild-type and EQ mutant Smc5/6 loaded onto DNA, revealing their characteristic ‘SMC’ shape along with Nse2 subdomain and freely dynamic behavior (Fig. 4e,f, Supplementary Fig. 10, Supplementary Movie 4). In most cases, a single Smc5/6 molecule bound to one DNA molecule, although in some instances, multiple Smc5/6 bound to the same DNA molecule (Fig. 4f,g, Supplementary Fig. 10, Supplementary Movie 5). We also observed that Smc5/6 bound to both supercoiled and relaxed DNA, with a slight preference for supercoiled DNA, consistent with the bulk biochemical experiments shown in Fig. 4c. In contrast to ATP, DNA-bound Smc5/6 was rarely detected using ADP (Fig. 4g). We also did not observe Smc5/6 bound to DNA when the hexamer complex lacking Nse5–Nse6 was used (Fig. 4g).

Given that topological Smc5/6 loading strictly depends on ATP and Nse5–Nse6 (Fig. 4b, Supplementary Fig. 9a–c)^20,24, the Smc5/6 molecules observed by HS-AFM likely entrapped DNA within the ring compartment. Consistent with this notion, Smc5/6 randomly slid along the DNA during scanning (Supplementary Note 1). Although Smc5/6 exhibited ATP-independent, non-topological DNA binding in the bulk assay (Fig. 4c), this electrostatic DNA binding appeared to be weak and was therefore rarely observed under HS-AFM on the lipid stage. If this were the case, Smc5/6 would be unable to bind to pre-adsorbed DNA on the lipid stage through electrostatic contact, resulting in a failure to establish subsequent topological entrapment. Consequently, we did not observe a Smc5/6 molecule bound to the DNA when the octamer complex was added after DNA had been adsorbed onto the lipid stage (Supplementary Fig. 11). This result further supports the conclusion that the current experimental setting visualized topologically loaded Smc5/6 on DNA. Together, these results demonstrate that the current HS-AFM setup enables visualization of both the subdomain-level structure and the dynamics of Smc5/6 on DNA.

Smc5/6 interacts with DNA through both the ATPase heads and hinge regions

Like other SMC complexes, the ring architecture of Smc5/6 is composed of Smc5, Smc6 and the Nse4 kleisin subunit. The ring is further divided into two compartments, the SMC and the kleisin compartments, by ATPase head engagement (Fig. 5a). Upon topological DNA entrapment, Smc5/6 can potentially sequester DNA within either the SMC or kleisin compartment²⁴. We observed two distinct types of DNA-bound Smc5/6 molecules, one of which could be further categorized into two subclasses (see below). These types were classified by inspection of multiple consecutive frames. One type was associated with DNA at the hinge domain, exhibiting either the O-form or I-form structures (referred to hereafter as Hinge-O/I; Fig. 5b, Supplementary Movie 6). Approximately 50% of the wild-type Smc5/6 molecules were classified as Hinge-O/I (Fig. 5e). The other type bound to DNA at the ATPase heads, predominantly adopting the I-form structure (hereafter referred to as Head-I; Fig. 5c, Supplementary Movie 6). We noticed that, unlike the observations on mica (Fig. 2a), Nse2 was not detected in the DNA-bound, Head-I molecules. During real-time tracking, we occasionally observed transitions from the Head-I to the O-forms with the appearance of the Nse2 subunit (hereafter referred to as Head-O, Supplementary Fig. 12a, Fig. 5d). This suggests that DNA-bound Smc5/6 in the Head-I state was adsorbed onto the lipid membrane surface in a different orientation, probably with Nse2 facing down, compared to the observations on mica. Supporting this notion, Nse2 was not visible in the pseudo-AFM image when Nse2 was facing downwards in the apo-state atomic model (I-form) (Supplementary Fig. 12b,c). Note that we cannot rule out the possibility that the observed ‘I-form’ configuration is derived from the ‘O-form’ configuration that is perpendicular to the AFM stage, albeit that such a state is predicted to be obscure due to molecular swing during scanning. Nonetheless, these results suggest that the Head-O/I configurations sequester DNA within the kleisin compartment, whereas the Hinge-O/I configurations encircle DNA within the SMC compartment.

Fig. 5. Classification of Smc5/6 molecules bound to DNA.

a Schematics of the apo, ADP-bound, and ATP-bound Smc5/6 complex. The SMC ring is divided into the SMC and kleisin compartments. Real-time imaging of three types of DNA-bound Smc5/6 molecules showing Hinge-O/I (b), Head-I (c), and Head-O (d) states (related to Supplementary Movie 6). The cyan arrows indicated the direction of the molecular movement. b: Scan area, 150 × 150 nm² at 160 × 80 pixels², cropped to 130 × 68 nm². c Scan area, 150 × 120 nm² at 160 × 64 pixels², cropped to 128 × 71 nm². d Scan area, 220 × 176 nm² at 200 × 80 pixels², cropped to 155 × 78 nm². Imaging rates: 0.37 s per frame (b), 0.34 s per frame (c), 0.6 s per frame (d). e Abundance ratios of DNA-bound Smc5/6 conformations. The Data are shown as means ± SEM obtained from three independent experiments. f Distribution of Smc5/6 binding angles, analyzed from a total of 252, 55, 166, and 450 frames of successive HS-AFM images for EQ Head-O, WT Head-I, EQ Head-I, and WT Hinge-O/I, respectively (n = 3 molecules per condition).

In contrast to the wild-type, the majority (~95%) of the EQ mutant bound to DNA at the ATPase heads, with the Hinge-O/I configuration being rarely observed. Furthermore, approximately 70% of the EQ mutant displayed the Head-O structure while bound to DNA (Fig. 5d,e, Supplementary Movie 6). Although Smc5/6 in the Head-O state appears to entrap DNA within the kleisin compartment, this binding mode could also arise without topological entrapment if Smc5/6 passes DNA through disengaged ATPase heads and clamps it between the ATPase heads and Nse1/Nse3, as demonstrated in the cryo-EM structure of the DNA-bound Smc5/6 hexamer²⁶. If this were the case, pre-engagement of the ATPase heads would be expected to reduce DNA binding by preventing such clamping. To test this possibility, we pre-incubated the EQ mutant with ATP to promote ATPase head engagement before initiating Smc5/6 loading by adding DNA. However, similar levels of DNA recovery were observed regardless of the order of ATP addition in the bulk loading assay (Supplementary Fig. 13a). Similarly, pre-incubation of the EQ mutant with ATP did not alter the efficiency of DNA binding or the frequency of the Head-O molecules observed using HS-AFM (Supplementary Fig. 13b,c), supporting the conclusion of DNA entrapment within the kleisin compartment in the Head-O state. Given that the EQ mutant, which stabilizes ATPase head engagement, primarily sequesters DNA at the ATPase head region, ATP hydrolysis appears to be essential for establishing the Hinge-O/I DNA-binding state.

During HS-AFM observation, the Smc5/6 complex in the Hinge-O/I configurations on DNA often exhibited swaying motion, whereas the head-bound molecules remained relatively stable. To quantify these differences in mobility, we measured the binding angles of Smc5/6 relative to DNA at each scanning time point (Fig. 5f, Supplementary Fig. 14a,b; see Methods). The histograms of the binding angles showed that the Head-I molecules were distributed in a single peak, bound at an average of 73° (76° for the EQ mutant). Similarly, the Head-O molecules derived from the EQ mutant showed a single peak distribution, but with a peak binding angle of 104°, suggesting that Head-O and Head-I are similarly tilted relative to 90° but in opposite directions. In contrast, Hinge-O/I from the wild-type complex exhibited a broad distribution of binding angles, which could be fitted to multiple Gaussian peaks at 55, 90, and 130°. These results suggest that Smc5/6 stably grasps the DNA strand at the ATPase head region, while DNA binding at the hinge domain is less rigid, allowing Hinge-O/I molecules to exhibit greater flexibility. Additionally, we measured the angles of the DNA curvature at Smc5/6 binding sites (Supplementary Fig. 14c). The binding sites of Hinge-O/I molecules exhibited larger curvature angles (median at 110°) compared to those of Head-O/I molecules (less than 50°). These differences indicate that the Hinge-O/I preferentially associates with curved regions of DNA, whereas the head-binding Smc5/6 molecules bind to less curved regions, reflecting distinct DNA binding states.

Smc5/6 modulates DNA structure

Single-molecule fluorescence imaging and biophysical studies have demonstrated that Smc5/6 compacts DNA strands in an ATP-dependent manner^10,20,21. As shown in Fig. 6a, we noticed that the nicked circular DNA substrate, which typically appears as an extended circle strand under HS-AFM (–Smc5/6, Fig. 6a1), tended to become highly deformed upon binding of Smc5/6. We classified the Smc5/6-bound DNA molecules into the following three types: Smc5/6 bound to one DNA segment (Bound 1, Fig. 6a2), Smc5/6 interacting with two DNA segments (Bound 2, Fig. 6a3), and Smc5/6 binding at twisted regions of the DNA (Twisted, Fig. 6a4). For the “Bound” species, we primarily observed “Bound 1” (15 out of 17 molecules), with fewer Smc5/6-bound DNA molecules classified as “Bound 2”. Subsequent real-time tracking revealed that the “Bound 2” state was transient, with Smc5/6 tending to bridge then release two segments of DNA (Supplementary Fig. 15).

Fig. 6. DNA structures modulated by Smc5/6 tethering activity.

a Representative HS-AFM images of Smc5/6 bound to nicked, relaxed circular DNA substrates on a lipid membrane. The four binding conformations are shown: –Smc5/6 (a1), bound1 (a2), bound2 (tethered) (a3), and twisted (a4). a1: Scan area, 350 × 280 nm² at 200 × 80 pixels², cropped to 280 × 252 nm². a2: Scan area, 450 × 360 nm² at 240 × 96 pixels², cropped to 315 × 245 nm². a3: Scan area, 350 × 280 nm² at 200 × 80 pixels², cropped to 301 × 252 nm². a4: Scan area, 350 × 280 nm² at 200 × 80 pixels², cropped to 283 × 224 nm². Imaging rates: 1 s per frame (a1, a4), 1.5 s per frame (a2), 0.9 s per frame (a3). b Quantification of the radius of gyration of Smc5/6-bound DNA. c Same as b except that Smc5/6-bound DNAs were further classified according to the number of Smc5/6 molecules. Schematic (d) and representative agarose gel image (e) of the DNA–DNA tethering assay. The graph shows the quantification of captured free DNA (from three independent experiments; means ± SD). f A schematic model depicting the ATP-dependent transactions of Smc5/6 on DNA, including topological loading, diffusive movement and DNA-DNA tethering.

In the “Twisted” DNA species, Smc5/6 appeared to interact with more than one DNA segment. To quantify the degree of compaction, we calculated the radius of gyration (R_g) of the “Bound” and “Twisted” DNA molecules. As shown in Fig. 6b, the mean R_g of both the “Bound” ( ~ 75 nm) and “Twisted” ( ~ 55 nm) species were smaller than that of the naked DNA substrate ( ~ 80 nm), with a higher degree of DNA compaction for the “Twisted” species. We note that the presence of Smc5/6 introduced a small measurement error in R_g, estimated to be approximately 2–3% (Supplementary Fig. 16). Interestingly, we also found that multiple Smc5/6 molecules tended to bind to the “Twisted” species more frequently than the “Bound” species (Fig. 6c). These results suggest that Smc5/6 mediates DNA compaction through DNA twisting, particularly when multiple Smc5/6 molecules are loaded onto DNA. Notably, similar DNA compaction was observed with the EQ mutant (Fig. 6c), indicating that this compaction depends on ATP binding rather than ATP hydrolysis. This property contrasts with DNA loop extrusion, which has been reported to require continuous ATP hydrolysis by SMC complexes^7–10.

As Smc5/6 was frequently observed in contact with two DNA strands during our HS-AFM analyses, we speculated that DNA compaction might be facilitated by tethering two independent DNA strands. Notably, 80% (33 out of 38) of the “Twisted” species, which were bound by more than two Smc5/6 molecules, stably maintained DNA tethering throughout multiple HS-AFM frames (~100 frames). Among them, we also observed Smc5/6 moving coordinately with the two tethered DNA segments (Supplementary Fig. 17 and Supplementary Movie 7).

In addition to loop extrusion, cohesin and condensin have also been shown to tether two separate DNA strands in an ATP-dependent manner^12–14. To test whether Smc5/6 mediates DNA–DNA tethering, a linear DNA fragment biotinylated at both ends was immobilized on streptavidin-conjugated magnetic beads (Fig. 6d). These DNA-coated beads were then incubated with circular plasmid DNA (free DNA) and Smc5/6 in the presence of ATP. After a high-salt wash, the captured free circular DNA was analyzed by agarose gel electrophoresis. Approximately 25% of the input free DNA was recovered in the presence of ATP (Fig. 6e, WT). In contrast, recovery of free DNA was less than 3% when ATP was replaced by ADP or omitted. These results indicate that Smc5/6 mediates DNA–DNA tethering in an ATP-dependent manner. Efficient recovery of free DNA was also observed with the EQ mutant in the presence of ATP (Fig. 6e, EQ) or with WT in the presence of ADP–AlFx (Supplementary Fig. 18a), indicating that the observed DNA–DNA tethering does not require efficient ATP hydrolysis, consistent with the DNA compaction observed by HS-AFM.

To investigate whether the observed DNA tethering was mediated by topological entrapment, we digested free DNA at one site using a restriction enzyme after recovery of the DNA tethering products (Supplementary Fig. 18b). This released linearized free DNA into the supernatant (Supplementary Fig. 18c,d). Similarly, single digestion of bead-attached DNA released intact free DNA from the beads and simultaneously dissociated Smc5/6 from the DNA (Supplementary Fig. 18c,d). These bulk biochemical experiments suggest that Smc5/6 mediates DNA–DNA tethering via topological DNA entrapment.

Discussion

In this study, we visualized the structural dynamics of budding yeast Smc5/6 complexes at submolecular resolution using HS-AFM under a physiological buffer condition. Using the Smc5/6 hexamer, we primarily observed two types of extended molecules, the I- and O-forms, depending on their ATP binding state. These structural variations are distinct from those of cohesin and condensin, which display sharply folded conformations by bending the SMC arms at their elbows in a manner not seen with Smc5/6^{32,33,47–51}. Our data suggest a potential role for Nse2 in promoting the extended conformation of Smc5/6 by preventing acute bending at the Smc5 elbow. The low frequency of the B-form observed in the absence of Nse2 under HS-AFM conditions may result from weak adsorption of Smc5/6 molecules onto the mica surface, thereby restricting more dynamic three-dimensional conformational transitions. Consistent with this notion, recent electron microscopy observations have shown that Smc5/6 adopts a shorter, more compact conformation in the absence of Nse2²⁸. As Nse2 is essential for cell viability⁵², the observed structural constraint of Smc5/6 may be related to its chromosomal functions, probably through its DNA-binding strategies observed in this study.

By optimizing the experimental conditions, namely through the use of the supported lipid bilayer system^37,38 which minimizes the surface adsorption of Smc5/6, we were able to visualize the molecular dynamics of Smc5/6 bound to DNA with unprecedented spatiotemporal resolution. This result revealed the existence of three distinct types of DNA-bound Smc5/6. Both the Head-O (EQ mutant) and Head-I (wild-type) molecules bind to DNA at the ATPase head region, whereas the Hinge-O/I molecules, observed mainly with wild-type Smc5/6, hold DNA within the SMC compartment.

We speculate that Smc5/6 initially entraps DNA within the kleisin compartment and subsequently transfers it to the SMC compartment upon ATP hydrolysis (Fig. 6f, Supplementary Fig. 19). The Head-O molecule, frequently observed with the EQ mutant, represents the state immediately after Smc5/6 topologically entraps DNA (Fig. 6f, Topological loading). A recent biochemical study using protein crosslinking experiments has shown that Smc5/6 topologically entraps DNA through transient opening of the Smc6–Nse4 interface (N-gate)²⁴. In the non-ATP binding apo state, Nse5–Nse6 mediates the N-gate opening by inhibiting the interaction between Nse4 and Smc6^24,27,28. The reclosure of the N-gate is facilitated by the ATP-dependent engagement of the SMC ATPase heads and DNA binding, leading to the topological capture of DNA within the kleisin compartment^24,28. Subsequent ATP hydrolysis destabilizes and converts the O-form to the I-form, which is commonly observed with the wild-type complex (Fig. 6f, Head-I).

How does Smc5/6 ultimately deliver DNA to the SMC compartment? A recent cryo-EM study of fission yeast cohesin suggests that it initially entraps DNA via the N-gate by forming the DNA-gripping intermediate⁴⁹. In this state, DNA is also positioned at the top of the ATPase heads, along with the cohesin loader. A similar intermediate, termed the segment-capture state, has been proposed for Smc5/6, in which a small DNA loop is topologically entrapped within the Nse4 kleisin compartment, while being held by the SMC compartment²⁴ (Fig. 6f, segment-capture). We speculate that the segment-capture state is relatively unstable and predisposed to transition into the Hinge-O/I or Head-O/I state. ATP hydrolysis induces disengagement of the ATPase heads, allowing the DNA segment located above the head to pass through, thereby forming the Hinge-O/I state (Fig. 6f, DNA within the SMC compartment). Conversely, slippage of a small DNA loop from the SMC compartment without the head disengagement leads to the formation of the Head-O/I states. Recent cryo-EM studies of the prokaryotic SMC complexes have revealed the DNA-holding state, where DNA is trapped within the kleisin compartment but not within the SMC compartment^53,54. The Head-O/I Smc5/6 molecules may represent a similar DNA-holding mode. Alternatively, the Head-O state, which is frequently seen in the EQ mutant, could represent an altered segment-capture state. In the Head-O form of the EQ mutant, most of the small DNA loop has already slipped out, yet the complex still retains DNA clamped between the ATPase heads and Nse1/Nse3. Our bulk loading assay showed that the EQ mutant retained DNA more efficiently than the wild-type complex after DNA linearisation (Supplementary Fig. 9c), reflecting the distinct nature of the Head-O state. Once DNA is entrapped, Smc5/6 exhibits diffusive motion along the DNA, a property that may facilitate its translocation along chromatin from its initial loading site (Fig. 6f, Diffusive movement).

Although our HS-AFM observations directly visualize the dynamic behavior of DNA-bound Smc5/6 and the dependence on Nse5–Nse6 provides additional support, our interpretation of topological loading is based on convergent but indirect evidence rather than direct biochemical proof of DNA entrapment within a closed Smc5/6 ring. Nevertheless, HS-AFM complements these biochemical approaches by providing insight into the structural dynamics of reconstituted DNA-bound Smc5/6 without requiring covalent ring closure.

In addition to topological DNA entrapment, Smc5/6 has been shown to form DNA loop through extrusion¹⁰. Cohesin and condensin have been proposed to mediate DNA loop extrusion via bending and stretching motions of their SMC arms at the elbows^32,33,55,56. In contrast, Nse2, which restricts the Smc5/6 arm bending in our HS-AFM observations, has paradoxically been reported to essential for Smc5/6-mediated loop extrusion¹⁰. In addition, Smc5/6 facilitates loop extrusion via dimer formation¹⁰. These observations suggest that Smc5/6 may employ a distinct mechanism of DNA loop extrusion compared to cohesin and condensin. Whether and how the Nse2-regulated conformation of Smc5/6 is linked to both DNA loop extrusion and topological DNA entrapment remains an important topic for further investigation.

Our HS-AFM observations also revealed that Smc5/6-bound DNA molecules, especially those with multiple Smc5/6 molecules, tended to be twisted, leading to DNA compaction. One possible mechanism for this observed DNA compaction is DNA loop extrusion by Smc5/6, which has been reported to require continuous ATP hydrolysis¹⁰. However, we observed a similar level of DNA compaction with the EQ mutant, suggesting that efficient ATP hydrolysis is not required for the compaction observed in this study. Furthermore, our biochemical experiments also detected physical tethering of two independent DNA molecules in an ATP-binding-dependent, but not hydrolysis-dependent manner. Although the structural resolution of our HS-AFM images was insufficient to determine how Smc5/6 clusters establish a stable compacted DNA state, the combined HS-AFM and biochemical results suggest that Smc5/6 promotes DNA compaction through a DNA–DNA tethering mechanism (Fig. 6f, DNA–DNA tethering).

Previous force measurement studies using magnetic tweezers have shown that Smc5/6 mediates more efficient DNA compaction when two DNA strands are aligned in close proximity²⁰. Thus, Smc5/6 may associate more stably with DNA in the tethered state. Interestingly, cohesin and condensin in their monomeric form have been shown to capture a second DNA molecule by topological entrapment^12–14, suggesting that DNA–DNA tethering activity is a conserved biochemical property among SMC complexes, in addition to DNA loop extrusion. The mode by which Smc5/6 molecules associate with DNA at the twisted region remains unclear due to the limited structural resolution in HS-AFM. In the future, it will be of great interest to investigate whether, and how, Smc5/6 employs the DNA-binding strategies we observed to recognize and manipulate higher-order and intrinsically harmful DNA structures that arise during the processes of DNA repair and transcription.

Methods

Cell culture media

Culture media used for protein expressions are listed below:

YP-raffinose medium:	1% yeast extract, 2% peptone, 1.5% raffinose, supplemented with 100 µg mL–1 adenine
Terrific-broth medium:	1.2% tryptone, 2.4% yeast extract, 0.4% glycerol, 17 mM NH2PO4 and 72 mM K2HPO4

Protein purification buffers

Buffer used for protein purifications are listed below:

Cell lysis buffer:	40 mM Tris-HCl (pH 7.5), 0.5 mM tris(2-carboxyethyl)phosphine (TCEP), 10% (v/v) glycerol, 0.3 M NaCl
R buffer:	20 mM Tris-HCl (pH 7.5), 0.5 mM TCEP, 10% (v/v) glycerol
R1 buffer:	20 mM Tris-HCl (pH 7.5), 0.5 mM TCEP, 10% (v/v) glycerol, 0.4 M NaCl, 0.005% (w/v) Tween 20
R2 buffer:	20 mM Tris-HCl (pH 7.5), 0.5 mM TCEP, 10% (v/v) glycerol, 0.4 M NaCl
R3 buffer:	20 mM Tris-HCl (pH 7.5), 0.5 mM TCEP, 10% (v/v) glycerol, 0.4 M NaCl and 0.01% (w/v) Tween 20
R4 buffer:	20 mM Tris-HCl (pH 7.5), 0.5 mM TCEP, 10% (v/v) glycerol, 0.3 M NaCl, 0.005% (w/v) Tween 20

Purification of the budding yeast Smc5/6 complex

Complementary DNAs (cDNA) encoding the budding yeast Smc5, Smc6, Nse1, Nse2, Nse3, Nse4, Nse5, and Nse6 were amplified by PCR from Saccharomyces cerevisiae W303. The resultant cDNAs were cloned into yeast shuttle vectors under the control of the GAL1 or GAL10 promoter. The Smc6-encoding cDNA, fused to 3× Pk epitope tag and a protein A tag at the C-terminus, separated by a PreScission protease recognition sequence, along with the SMC5 cDNA, was cloned into the YIplac211 (URA3) vector. The cDNA encoding Nse1, fused to a 7× histidine tag at the N-terminus, and the NSE3 cDNA were cloned into the YIplac128 (LEU2) vector. The Nse4-encoding cDNA, fused to a 7× histidine tag at the C-terminus, and the NSE2 cDNA were cloned into the YIplac204 (TRP1) vector. The Nse6-encoding cDNA, fused to a 7× histidine tag at the N-terminus, and the NSE5 cDNA were cloned into the pRSII402 (ADE2) vector. The linearized DNA constructs were sequentially integrated into the budding yeast strain (MATa, ade2-1, trp1-1, can1-100, leu2,3,112, his3-11,15, ura3-52, pep4Δ::HIS3MX) at the URA3, LEU2, TRP1 and ADE2 loci, respectively.

The resultant Smc5/6 octamer (comprising Smc5, Smc6, Nse1, Nse2, Nse3, Nse4, Nse5, and Nse6) expression yeast was grown in 7.5 L (1.5 L × 5 cultures) of YP-raffinose media at 30 °C to a cell density of 1–3 × 10⁷ per mL. Galactose was then added to the cell culture at a final concentration of 2% and the cells were grown further at 30 °C for 2 h to induce protein expression. The cells were collected by centrifugation and washed once with ice-chilled cell lysis buffer and then resuspended in one quarter volume of the same buffer containing 0.5 mM phenylmethylsulphonyl fluoride (PMSF). The cell suspension was dropped directly into liquid nitrogen for freezing and then ground in a freezer mill (SPEX CertiPrep 6850).

All subsequent procedures were carried out at 4 °C. The resultant cell powder was thawed at 4 °C and resuspended for 30 min in 3 volumes of cell lysis buffer containing 0.005% (w/v) Tween 20, 0.5 mM PMSF and cOmplete protease inhibitor cocktail (Merck). After clarification by centrifugation at 28,000 × g for 45 min, the cell lysate was loaded onto 1.5 mL of human IgG agarose (Sigma) at 0.75 mL min^–1 using a peristaltic pump. The beads were washed with 20 mL of R1 buffer containing 0.5 mM PMSF, then 80 mL of R2 buffer. The beads were resuspended in the same buffer with 1 mM TCEP, 10 µg mL^–1 RNase A and 5 U mL^–1 of PreScission protease and incubated overnight. The eluate was passed through 0.2 mL of glutathione Sepharose (Cytiva) to remove the PreScission protease. This fraction was then diluted with R buffer to a final concentration of 0.2 M NaCl and applied to a heparin HP column (1 mL, Cytiva). The column was developed with a linear gradient of 0.2 to 1 M KCl in R buffer. Peak fractions (~0.35 M KCl) were pooled and concentrated by ultrafiltration (Amicon Ultra, 30 K, Millipore), then loaded onto a Superose 6 Increase 10/300 GL gel filtration column (24 mL, Cytiva), which was equilibrated in R buffer containing 0.3 M KCl. Peak fractions were concentrated by ultrafiltration, aliquoted, snap-frozen in liquid nitrogen, and stored at –80 °C. Smc5/6 hexamers (Smc5, Smc6, Nse1, Nse2, Nse3, and Nse4) and EQ mutants (Smc5 E1015Q and Smc6 E1048Q) were expressed and purified using the same procedure. Smc5/6 pentamer (identical to the hexamer but lacking Nse2) was purified using the same procedure but omitting the gel filtration procedure.

Purification of the budding yeast Nse5–Nse6 heterodimer

The Nse5–Nse6 complex was expressed and purified in Escherichia coli. as described in ref. ⁴¹ with the following modification. Complementary DNAs (cDNA) encoding the budding yeast Nse5 and Nse6 were amplified by PCR from Saccharomyces cerevisiae W303. The Nse5-encoding cDNA fused to a PreScission protease recognition sequence and a 2× Strep tag at the C-terminus and the NSE6 cDNA, both of which were connected with the ribosome binding sequence, were cloned into the E. coli protein expression vector pET21a under the control of a T7 promoter. The resultant Nse5–Nse6 expression plasmid was introduced into the E. coli Rosetta 2 (DE3). The transformants were grown in 3 L (1.5 L ×2 cultures) of Terrific-broth medium containing 100 µg mL^–1 ampicillin and 17 µg mL^–1 chloramphenicol at 37 °C to an optical density of 0.6 at 600 nm. Cell cultures were cooled to 20 °C and further cultivated for 16 h after the addition of isopropyl β-D-1-thiogalactopyranoside (IPTG) to a final concentration of 0.25 mM to induce protein expression.

All subsequent procedures were carried out at 4 °C. Cells were collected by centrifugation and washed once with ice-chilled R2 buffer. The cell pellets were resuspended in 5 volumes of the same buffer supplemented with 1 mM PMSF and 1 mM 4-(2-aminoethyl)benzenesulphonyl fluoride hydrochloride (AEBSF). After the addition of one tenth volume of 10× BugBuster solution (Merck), the cells were disrupted by sonication. After clarification by centrifugation at 28,000 × g for 45 min, the cell lysate was passed through a 0.22 µm filter and then loaded onto 2 mL of Strep-Tactin agarose (IBA) at 0.75 mL min^–1 using a peristaltic pump.

The beads were washed sequentially with 40 mL of R3 buffer, 20 mL of R4 buffer containing 1 mM ATP and 5 mM MgCl₂ and then 60 mL of R4 buffer. The beads were further washed with 5 mL of R4 buffer containing 2.5 mM TCEP and then resuspended in 1.5 mL of the same buffer containing 5 U mL^–1 of PreScission protease and incubated overnight. The eluate was passed through 0.2 mL of glutathione-Sepharose (Cytiva) to remove the PreScission protease. This fraction was then diluted with R buffer to a final NaCl concentration of 0.15 M and applied to a heparin HP column (1 mL, Cytiva), which was developed with a linear gradient of 0.1–1 M NaCl in R buffer. Peak fractions (eluting around ~0.3 M NaCl) were pooled and concentrated by ultrafiltration (Amicon Ultra, 30 K, Millipore), then applied to a Superdex 200 Increase 10/300 GL column (24 mL, Cytiva), equilibrated in R buffer containing 0.2 M NaCl. Peak fractions were again concentrated by ultrafiltration, and aliquots were snap-frozen in liquid nitrogen and stored at –80 °C.

DNA substrates and antibody

The 1895 bp plasmid DNA, pYST400, used for HS-AFM analysis was generated by ligation of the annealed oligonucleotides (5ʹ-TATGCTGAGGTACGATATCTTA-3ʹ and its complementary oligo DNA) with the digested DNA fragment of pBluescript KSII (+) with SspI and PvuII. Covalently closed circular plasmid DNA (cccDNA) of pYST400 (1895 bp), pBluescript KSII (+) (3.0 kbp) and pSKsxAS (4.3 kbp)¹² and pKSII-E2 (7.8 kbp)¹² was prepared by the plasmid purification kit (Maxi prep. Kit, Qiagen). Relaxed circular DNA (rcDNA) was prepared by treating cccDNA with E. coli topoisomerase I (NEB). Nicked circular DNA (ncDNA) of pYST400 or pBluescript KSII (+) was prepared using Nt.BbvCI or Nb.BssSI, respectively. To prepare immobilized, closed topology DNA, the 3.0 kbp linear DNA was first prepared by PCR amplification with a pair of 5′ biotinylated DNA primers (759 and 760 bp) using pKSII-E2 as the template. The resultant linear DNA was mixed and immobilized on streptavidin conjugated magnetic beads (Thermo), as described in ref. ¹². Short DNA substrates used for electrophoretic mobility shift assays were prepared by heat annealing using the oligonucleotide pairs DB20-F (5′-TATTTACTTTATGTTCATTT-3′) and DB20-R (5′-AAATGAACATAAAGTAAATA-3′) for the 20-bp substrate, and DB30-F (5′-TCCTTATACTTATTTACTTTATGTTCATTT-3′) and DB30-R (5′-AAATGAACATAAAGTAAATAAGTATAAGGA-3′) for the 30-bp substrate.

Anti-Pk antibody (anti-V6; MCA1360GA, Bio-Rad) was used for immunoprecipitation and western blotting. For each immunoprecipitation reaction, 2 μL of anti-Pk antibody (1 mg/mL) was incubated with 10 μL of Protein A-conjugated magnetic beads (Thermo Fisher Scientific). For western blotting, the antibody was used at a 1:2,000 dilution.

Buffers used for the biochemical and HS-AFM experiments

Buffer used for the biochemical and HS-AFM experiments in this study are listed below:

ATPase reaction buffer:	25 mM HEPES-KOH (pH 7.5), 1 mM TCEP, 5% (v/v) glycerol, 75 mM potassium acetate (KOAc), 30 mM KCl, 1 mM magnesium acetate (Mg(OAc)2), 0.01% (w/v) NP-40
SDB buffer	20 mM Tris-HCl (pH 7.5), 1 mM TCEP, 15% glycerol, 100 mM KCl, 1 mM Mg(OAc)2, 0.05% (w/v) NP-40
SL buffer:	25 mM HEPES-KOH (pH 7.5), 1 mM TCEP, 5% (v/v) glycerol, 75 mM KOAc, 30 mM KCl, 1 mM Mg(OAc)2, 0.05% (w/v) NP-40
IP500 buffer:	25 mM HEPES-KOH (pH 7.5), 500 mM NaCl, 0.25% (w/v) NP-40, 1 mM ethylenediaminetetraacetic acid (EDTA)
IP100 buffer:	25 mM HEPES-KOH (pH 7.5), 100 mM NaCl and 0.1% (w/v) NP-40
Elution buffer:	10 mM Tris-HCl (pH 7.5), 0.5% (w/v) sodium dodecyl sulfate (SDS), 1 mM EDTA, 1 mg mL–1 proteinase K
RE buffer:	25 mM Tris-HCl (pH 7.5), 1 mM TCEP, 10 mM MgCl2, 100 mM NaCl and 0.1% (w/v) NP-40
DC500 buffer:	25 mM HEPES-KOH (pH 7.5), 500 mM NaCl, 1 mM EDTA, 0.05% (w/v) NP-40
DC100 buffer:	25 mM HEPES-KOH (pH 7.5), 100 mM NaCl, 0.05% (w/v) NP-40
AFM-A buffer:	25 mM HEPES-KOH (pH 7.5), 1 mM TCEP, 10% (v/v) glycerol, 300 mM KOAc, 1 mM Mg(OAc)2
AFM-B buffer:	25 mM HEPES-KOH (pH 7.5), 1 mM TCEP, 5% (v/v) glycerol, 75 mM KOAc, 1 mM Mg(OAc)2
AFM-C buffer:	25 mM HEPES-KOH (pH 7.5), 1 mM TCEP, 10% (v/v) glycerol, 30 mM KOAc, 1 mM Mg(OAc)2

ATPase assay

Purified budding yeast Smc5/6 (100 nM) was mixed with pBluescript KSII rcDNA (16.6 nM) on ice in 20 µL of ATPase reaction buffer. The reaction was initiated by adding ATP (0.5 mM) and incubated at 30 °C. 4 µL of samples were then taken at each time-point and placed on ice to stop the reaction. Generation of inorganic phosphate on ATP hydrolysis was measured using a commercially available phosphate-detection kit (BioAssay Systems).

Electrophoretic mobility shift assay

Smc5/6 was mixed with both cccDNA and ncDNA from pBluescript KSII (0.8 nM) at increasing concentrations in 15 µL of SL buffer, containing 5 mM ATP or ADP, and incubated at 30 °C for 15 min. Samples were analyzed by 0.8% agarose gel electrophoresis in TAE (pH 7.5) buffer at room temperature (~23 °C) for 90 min at 3.6 V cm^–1, followed by Sybr Gold staining. Gel images were captured using an Image Quant LAS-400 mini gel documentation system (Fujifilm). In experiments involving short DNA, Smc5/6 was incubated with 20- or 30-bp DNA in 10 µL of SDB buffer containing 5 mM ATP. Samples were analyzed as described above, except that a 2% agarose gel was used.

Smc5/6 loading assay

Smc5/6 was mixed on ice in 15 µL of SL buffer, either in the absence or presence of 5 mM ATP, ADP, CTP, ATP-γ-S or 5 mM ADP supplemented with 1 mM AlCl₃ and 10 mM NaF. The reaction was initiated by the addition of ncDNA of pBluescript KSII (3.3 nM) and incubated at 30 °C for 30 min. The reaction was terminated by adding ice-chilled IP500 buffer. Anti-Pk (V5, Bio-Rad), bound to protein A-conjugated magnetic beads (ThermoFisher), was added to the reaction mixture and rocked at 4 °C overnight (~15 h). The beads were washed three times with 750 µL of IP500 buffer and then once with IP100 buffer. The washed beads were resuspended in elution buffer and incubated at 37 °C for 20 min. The recovered DNA was separated by 1% agarose gel electrophoresis in TAE buffer (pH 8.2) at room temperature (~23 °C) for 60 min at 3.6 V cm^–1, followed by Sybr Gold staining. Gel images were captured using an Image Quant LAS-400 mini gel documentation system (Fujifilm), and band intensities were quantified using MultiGauge software (Fujifilm).

In the experiment involving linearisation of Smc5/6-bound DNA, the loading reaction and subsequent immunoprecipitation were carried out as described above using cccDNA of pBluescript KSII (+). The resultant beads were further washed with RE buffer. The washed beads were incubated with PstI (20 U, Takara) in 10 µL of RE buffer at 7 °C for 20 min. The NaCl concentration was adjusted to 500 mM in 15 µL, and incubated for an additional 15 min on ice. DNA in both the supernatant and bead fractions was analyzed as described above.

DNA–DNA tethering assay

The concentrations indicated are final concentrations in the total reaction volume (15 µL). Smc5/6 (100 nM) was initially mixed on ice in 10 µL of SL buffer containing the indicated adenosine derivatives (5 mM). The reaction was initiated by adding 5 µL of a mixture of rcDNA (pSKsxAS, 4.3 kbp, 1.67 nM) and the immobilized, closed-topology DNA beads (3.0 kbp, 1.16 nM) in SL buffer. The mixture was incubated at 30 °C for 30 min with occasional agitation. The beads were then resuspended in 150 µL of DC500 buffer and further incubated at 30 °C for 3 min with occasional agitation. After washing the beads once with DC500 buffer and then once with DC100 buffer, the recovered beads were suspended in 15 µL of elution solution and incubated at 37 °C for 20 min. The recovered DNA was separated by 1% agarose gel electrophoresis in TAE buffer (pH 8.2) at room temperature (~23 °C) for 70 min at 3.6 V cm^–1, and stained with Sybr Gold.

In the experiment including linearisation of captured DNA by Smc5/6, the DNA tethering reaction and subsequent high-salt wash were performed as described above. The resultant beads were further washed with RE buffer, then resuspended and further incubated at 7 °C for 20 min in 10 µL of the same buffer containing PstI (20 U, Takara) to digest rcDNA (pSKsxAS) or PflMI (20 U, NEB) to digest immobilized DNA at a single locus, respectively. The NaCl concentration was adjusted to 500 mM in 15 µL and incubated for an additional 15 min on ice. DNA in the supernatant and beads fractions was analyzed as described above, using elution solution without proteinase K. Smc5/6 in both fractions were also analyzed by SDS-PAGE followed by western blotting using an anti-Pk antibody to detect Smc6.

High-speed AFM setup

As described in our previous studies³⁸, we employed a lab-built high-speed AFM equipped with a commercially available cantilever specialized for bio-imaging (Olympus: BL-AC10DS-A2). The resonance frequency, spring constant, and the quality factor of the cantilever in a buffer solution were approximately 500 kHz, 0.1 N m^–1, and 1.5, respectively. An amorphous carbon tip, approximately 500 nm in length, was fabricated on the cantilever bird’s beak by electron beam deposition. The free oscillation amplitude of the cantilever was a few nm_p-0, with the setpoint amplitude set to around 90% of the free amplitude.

In our previous setup, we have employed full-cycle digital lock-in detection using a field-programmable gate array to measure the cantilever amplitude. To increase the detection bandwidth, we implemented a half-cycle detector⁵⁷, which enabled imaging without damaging molecules even during high-speed scanning. We used control software developed in Visual Studio for HS-AFM control⁵⁸. To quantitatively measure the molecular size and DNA length, we implemented a function to compensate for the nonlinearity and tip-position dependence of the XY piezo scanner⁵⁹.

High-speed AFM imaging

For observations of Smc5/6 using a mica imaging stage, proteins were incubated in AFM-A buffer in the absence or presence of 5 mM ATP or ADP on ice for 5 min. Then, 1 µL of the protein mixture was deposited on a freshly cleaved mica surface and incubated for 5 min before washing away unbound proteins with the same buffer. HS-AFM observations were performed in the same solution. Although approximately 80–90% of the observed molecules were imaged in an intact state, the remaining molecules showed dissociation of either the head or hinge domains at the start of observation or within the first few frames. Only molecules that remained intact throughout imaging for several tens of frames were included in the analysis shown in Figs. 1, 3.

To observe DNA-bound Smc5/6, the Smc5/6 octamer (50 nM) was first incubated with 1895 bp circular DNA (5.3 nM, pYST400) in AFM-B buffer containing 5 mM ATP at 30 °C for 2 h. For HS-AFM observation of DNA-bound Smc5/6 using a mica imaging stage, 1 μL of the DNA loading reaction was deposited on a freshly cleaved mica surface and incubated for 5 min. After rinsing with AFM-C buffer containing 5 mM ATP, HS-AFM observation was performed in the same buffer condition.

For observations of DNA-bound Smc5/6 using a mica-supported lipid bilayer imaging stage, 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC; Avanti Polar Lipids) and 1,2-dipalmitoyl-3-trimethylammonium-propane (DPTAP; Avanti Polar Lipids), dissolved in chloroform, were mixed at a weight ratio of 9:1. The chloroform was then evaporated to form a lipid film, which was subsequently hydrated in water to prepare liposomes. This preparation was stored in a freezer and, immediately before the HS-AFM experiment, thawed and dissolved in a 10 mM MgCl₂ solution at a final concentration of 0.2 mg mL^–1. The mixture was sonicated for 1 min using an ultrasonic bath sonicator (AS ONE, AUC-06L) to produce small unilamellar vesicles. 2 μL of the sonicated liposome solution was deposited onto a freshly cleaved mica surface, incubated for 5 min, and rinsed with excess Milli-Q water (>100 μL) and 20 μL of AFM-B buffer. 1 μL of the DNA loading reaction was deposited onto the lipid membrane imaging stage and incubated for 5 min. HS-AFM imaging was performed in AFM-B buffer containing 5 mM of the indicated adenosine derivatives, after rinsing with the same buffer.

HS-AFM data analysis

All HS-AFM images and movies are presented after being processed with a Gaussian-kernel filter (standard deviation ~1 nm) and color scaling adjustments. The individual HS-AFM datasets consist of multiple frames; however, the offset in the Z direction varies for each frame due to scanner drift. Conventionally, color scaling has been performed based on the maximum and minimum height values. However, this method results in an unstable color scale across frames due to the influence of noise and dust adsorption/desorption. To standardize the color scale across frames, we developed a method called PeakRef. In this approach, pixels are sorted based on their height values within each frame, and the substrate height is obtained by identifying the height value that occurs most frequently. Color scaling is then applied using the obtained substrate height, ensuring consistent color scaling across all frames. In Figs. 2a, d, 4e, 6a, a frame averaging method with a triangular window over the previous and subsequent frames was also applied to reduce noise and enhance the static structures.

To calculate the elbow angle of Smc5/6, three control points were manually placed along the ridgeline of the molecular arms (Fig. 3f, Supplementary Fig. 8), and the angle between the two straight lines (vectors of a and b) connecting these points was calculated using a trigonometric equation expressed by

$$\theta ={\cos }^{-1}\left(\frac{\mathbf{a}\cdot \mathbf{b}}{\parallel \mathbf{a}\parallel \parallel \mathbf{b}\parallel }\right).$$ 1

To evaluate the curvature distribution of DNA (Fig. 5, Supplementary Fig. 14c), the curvature angle was calculated at each point of the DNA. To determine the pseudo-curvature angle of a smooth line, the point of interest on the curve was locally approximated as two straight lines within the range of ±30 nm, a value slightly shorter than the persistence length of DNA (~50 nm), and the angle between these straight lines was calculated.

To quantify the compactness of DNA (Fig. 6b,c), we calculated the radius of gyration (R_g), which is expressed as

$$R_{\mathrm{g}}=\sqrt{\frac{1}{N}{\sum }_{i=1}^N{\left({\mathbf{r}}_i-{\mathbf{r}}_{\mathrm{CM}}\right)}^2},$$ 2

where r_CM represents the coordination vectors of the center of mass, and N denotes the total number of points. Initially, several tens of control points were manually placed along the DNA contour. These points were then interpolated using a spline method to increase the total number of points to 800. Given that the total length of the DNA was ~644 nm, this corresponds to a model of beads strung together at intervals of ~0.8 nm. For the calculation of R_g, we used data acquired within the first ten frames to minimize the effect of cantilever scanning.

Prediction of the ATP-dependent conformational change of Smc5/6

We performed a molecular simulation using a variant of the freely-jointed chain (FJC) model⁴², with data assimilation from our HS-AFM results (Fig. 2) and the published cryo-EM structures of Smc5/6, to predict the molecular trajectory of ATP-dependent conformational changes of the Smc5/6 hexamer (Supplementary Fig. 5, Supplementary Movie 1).

First, a coarse-grained (CG) model was constructed based on the all-atom structural model of the apo state Smc5/6 hexamer (I-form) derived from the cryo-EM structure (PDB-7QCD)²⁵. In this model, CG-particles, denoted as ‘p’, were connected by chains of rigid segments (Supplementary Fig. 5). Here, p_n represents the n-th CG-particle, corresponding to the Cα of amino acids identified by their residue sequence numbers, as listed in Supplementary Table 1. The 10 CG particles were placed at the following amino-acid residues: E187 (p1) and T200 (p2) on the Smc5 ATPase head, D739 (p3) on the Smc5 arm, Q647 (p4) and N618 (p5) on the Smc5 head, N503 (p6) on the Smc6 hinge, V761 (p7) and I805 (p8) on the Smc6 arm, and T1035 (p9) and S229 (p10) on the Smc6 ATPase. These 10 CG-particles were connected by 9 segments with fixed distances, which were maintained according to the initial structure from PDB-7QCD to avoid changes during the simulation. The 9 segments were further classified into 6 parts, denoted as ‘S’ (Supplementary Table 2): S1 (p1-p2-p3), S2 (p3-p4), S3 (p4-p5-p6), S4 (p6-p7), S5 (p7-p8) and S6 (p8-p9-p10). The angles between CG-particles within S1, S3, and S6 were fixed at their initial values, while the angles between the parts were allowed to vary arbitrarily.

This CG framework model was converted into the O-form by pulling p3 and p7 in opposite directions. During this process, a harmonic oscillator model-type force, which mimics ATP-dependent head engagement, was applied between p1 and p10 to maintain p1–p10 distance at 1.2 nm. Finally, an all-atom model of the O-form was generated by projecting the all-atom model of the I-form onto the coordination of the CG model of the O-form for each part.

Pseudo-AFM images were generated using UMEX Simulator, where the tip-convolution algorithm inspired by that used in the SPM simulator^60,61. An ideal symmetrical hemisphere with a tip radius of 5 nm, terminating in a cone-shaped tip with a half-angle of 10 degrees, was assumed. The tip was scanned by tracing the topmost surface of the molecular structure, and the tip trajectory was reconstructed into a pseudo-AFM image. Molecules were treated as rigid, and their stiffness, fluctuations, and scanning feedback errors were not considered.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

K.U. and Y.M. conceived the study. K.U., Y.K., N.K. and Y.M. designed the experiments. K.U. developed HS-AFM control software and an amplitude detector with half-cycle detection; performed HS-AFM imaging and simulations; and analyzed the data. Y.K. contributed to optimization of HS-AFM observations of Smc5/6. Y.K. and Y.M. purified Smc5/6 proteins, prepared DNA substrates and performed bulk biochemical experiments. N.K and Y.M. supervised the study. K.U. and Y.M. wrote the manuscript with input from all coauthors.

Peer review

Peer review information

Nature Communications thanks Niccolò Banterle and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by PRESTO, Japan Science and Technology Agency (JST) (JPMJPR20E3 and JPMJPR23J2 to K.U., JPMJPR19KB to Y.M.); CREST, JST (JPMJCR1762 to N.K.); KAKENHI, Japan Society for the Promotion of Science (19K15409 and 21K04849 to K.U., 26119003 and 20H00327 and 24H00402 to N.K., 23K23814 and 24H02282 to Y.M.); the Takeda Science Foundation (to YM); and the World Premier International Research Center Initiative (WPI), MEXT.

Data availability

The minimum dataset necessary to interpret and verify the findings of this study, and to support further analysis and extension of the work, is provided in the main text, Supplementary Information, Supplementary Movies, and Source Data file. The individual numerical values underlying the quantitative graphs from the HS-AFM analysis and bulk biochemical experiments, together with the raw gel images, are provided in the Source Data file. The structural model used in this study is available under the following accession codes: 7QCD (CryoEM structure of the I-form Smc5/6-holocomplex). The converted O-form Smc5/6 PDB file is available on Zenodo at [https://doi.org/10.5281/zenodo.21505466]. Source data are provided with this paper.

Code availability

The source code used to convert I-form Smc5/6 to O-form is available on Zenodo at https://doi.org/10.5281/zenodo.21505466. The custom software and executable files used in this study are available at GitHub: https://github.com/HS-AFM-developer/HS-AFM-software/tree/main. The algorithm is described in detail in ref. ⁶².

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77122-x.