DNA segment capture by Smc5/6 holocomplexes

Three distinct structural maintenance of chromosomes (SMC) complexes facilitate chromosome folding and segregation in eukaryotes, presumably by DNA loop extrusion. How SMCs interact with DNA to extrude loops is not well understood. Among the SMC complexes, Smc5/6 has dedicated roles in DNA repair and preventing a buildup of aberrant DNA junctions. In the present study, we describe the reconstitution of ATP-dependent DNA loading by yeast Smc5/6 rings. Loading strictly requires the Nse5/6 subcomplex which opens the kleisin neck gate. We show that plasmid molecules are topologically entrapped in the kleisin and two SMC subcompartments, but not in the full SMC compartment. This is explained by the SMC compartment holding a looped DNA segment and by kleisin locking it in place when passing between the two flanks of the loop for neck-gate closure. Related segment capture events may provide the power stroke in subsequent DNA extrusion steps, possibly also in other SMC complexes, thus providing a unifying principle for DNA loading and extrusion.

Three distinct structural maintenance of chromosomes (SMC) complexes facilitate chromosome folding and segregation in eukaryotes, presumably by DNA loop extrusion. How SMCs interact with DNA to extrude loops is not well understood. Among the SMC complexes, Smc5/6 has dedicated roles in DNA repair and preventing a buildup of aberrant DNA junctions. In the present study, we describe the reconstitution of ATP-dependent DNA loading by yeast Smc5/6 rings. Loading strictly requires the Nse5/6 subcomplex which opens the kleisin neck gate. We show that plasmid molecules are topologically entrapped in the kleisin and two SMC subcompartments, but not in the full SMC compartment. This is explained by the SMC compartment holding a looped DNA segment and by kleisin locking it in place when passing between the two flanks of the loop for neck-gate closure. Related segment capture events may provide the power stroke in subsequent DNA extrusion steps, possibly also in other SMC complexes, thus providing a unifying principle for DNA loading and extrusion.
In eukaryotes, three distinct SMC complexes (cohesin, condensin and Smc5/6) share essential tasks in the maintenance of chromosome structure and the faithful transmission of genetic information during nuclear division (reviewed in ref. 1). Some of these ATP-powered, DNA-folding machines are known to form large DNA loops by loop extrusion [2][3][4] . This is thought to allow condensin to compact chromatids in mitosis and cohesin to participate in gene regulation, DNA repair and recombination in interphase. However, loop extrusion or merely DNA translocation may well be a conserved feature of all pro-and eukaryotic relatives. Smc5/6 has indeed recently been shown to extrude DNA loops or translocate along DNA in vitro 5 , but the cellular functions and relevance of these activities remain unclear. DNA entrapment has long been considered a basic and essential feature of SMC function 1 . It provides a simple explanation for how SMC complexes stay in cis (on a given DNA molecule) over time, maintain directionality of translocation and readily bypass obstacles up to a few tens of nanometers in size that they encounter frequently on the chromosomal translocation track. However, the requirement of DNA entrapment and the exact nature of the DNA interaction remain contested, with all possibilities being considered in the recent literature (Fig. 1a): topological DNA entrapment, DNA loop entrapment (also denoted as pseudo-topological entrapment) and exclusively external DNA association. The last scenario was bolstered by the perceived bypass of very large obstacles by purified cohesin and condensin in single-molecule imaging experiments 6 . Cryo-electron microscopy (cryo-EM) studies have not yet been able to adequately address this question and in particular little is known about how Smc5/6 associates with DNA. Resolving DNA topology will be vital to refute and refine models for how the SMC ATPase powers DNA translocation and loop extrusion 7,8 .
Complete loss of Smc5/6 function leads to severe chromosome segregation defects in mitosis and meiosis [9][10][11][12][13] . Toxic DNA structures such as unresolved recombination intermediates, DNA intertwinings or incompletely replicated chromosomal regions prevent proper chromosome segregation in the absence of Smc5/6 (refs. [14][15][16]. DNA translocation by Smc5/6 may be required to detect and eliminate such DNA junctions. The Smc5/6 holocomplex comprises eight subunits: Smc5/6 and Nse1-6 in Saccharyomyces cerevisiae (Extended Data Fig. 1a). All of them cause lethality when removed by gene disruption 10 . Smc5/6 and Nse4 together form the conserved ATPase core with an elongated shape that is characteristic for all SMC complexes [17][18][19][20][21][22] . Smc5 and Smc6 Article https://doi.org/10.1038/s41594-023-00956-2 state', two molecules of ATP are sandwiched by residues of the Walker A and B motifs of the Smc5 head and the signature motif of the Smc6 head and vice versa; the engaged heads keep the two head-proximal arms at a distance, thus opening an SMC compartment 36,42,43 ; (2) in the 'juxtaposed state', the heads are ATP disengaged and the Smc5 and Smc6 arms coalign, yielding a closed SMC compartment 22 ; and (3) the 'inhibited state' relies on the intercalation of the loader Nse5/6 between the arms and heads of Smc5/ 6 (refs. 20,21), thus preventing head engagement and ATP hydrolysis. The inhibition is, in turn, overcome by the binding of a suitable DNA substrate 21 . A head-DNA interface on top of the engaged heads was described for several SMC complexes 38,[44][45][46][47][48][49][50] . Such a DNA-clamping state has recently also been described for Smc5/6 by cryo-EM 43 .
In the present study, we demonstrate that Smc5/6 holocomplexes efficiently entrap DNA molecules. Using the reconstituted DNA-loading reaction, we elucidate the position and topology of DNA in the Smc5/6 complex and unambiguously identify the neck gate as the entry gate for plasmid DNA.

Topological DNA entrapment by the Smc5/6 ring
Salt-stable DNA binding by the octameric yeast Smc5/6 holocomplex depends on a circular DNA substrate, being indicative of a topological component of the Smc5/6 DNA association 21 . In the present study, we delineate the position of DNA in the Smc5/6 complex by creating covalently closed DNA compartments followed by protein-DNA coisolation proteins harbor the canonical 'hinge' domains for heterotypical dimerization connected via ~35-nm-long, antiparallel, coiled-coil 'arms' to globular ABC-type 'head' domains with highly conserved motifs for ATP binding and hydrolysis [23][24][25] . The kleisin subunit Nse4 bridges the head of the κ-SMC protein Smc5 to the head-proximal coiled coil ('neck') of the ν-SMC protein Smc6. This creates the SK ring structure shown in other SMC complexes to be, in principle, capable of entrapping DNA [26][27][28][29][30] . Nse4 serves as an attachment point for two kite proteins (the Nse1 and Nse3 subunits), also found in the prokaryotic SMC complexes 31 , whereas the Smc5/6-specific subunit Nse2 attaches to the coiled-coil arm of Smc5 (ref. 32). Finally, a stable subcomplex is formed by the Nse5 and Nse6 subunits 19,21 . It contacts the Smc5/6 hexamer via multiple interfaces, including one on the arms and one on the heads. Single-molecule tracking and biochemical studies suggest a function of Nse5/6 in chromosome loading 21,33 . In S. pombe (but not in S. cerevisiae), disruption of nse5 and nse6 genes cause less severe phenotypes when compared with deletion of other Smc5/6 subunits, indicating that the hexamer can partially function without Nse5/6 in fission yeast 34 . Curiously, loop extrusion by S. cerevisiae Smc5/6 does not depend on the loader Nse5/6 and is actually inhibited by it, possibly implying that Smc5/6 holocomplexes have functions beyond loop extrusion 5 . A dedicated loader appears to be involved in viral restriction by Smc5/6 in human cells 35 .
We then separated protein-DNA catenanes from free DNA species by agarose gel electrophoresis (schematics in Extended Data Fig. 3a) 53 . Crosslinking at the three SK ring interfaces resulted in a characteristic laddering of a small (1.8-kb) plasmid at an elevated Smc5/6 concentration (400 nM) (Fig. 1c), presumably due to the reduced mobility of DNA molecules associated with increasing numbers of Smc5/6 complexes. The laddering was fully dependent on ATP and disappeared when plasmid DNA was linearized by a restriction enzyme before gel electrophoresis, as expected for a topological interaction (Fig. 1c). This loading reaction reached a steady state relatively quickly (within minutes; see below). Moreover, three protein preparations lacking one of the six engineered cysteines showed little or no DNA laddering, with the residual gel shift probably explained by low levels of off-target crosslinking (Extended Data Fig. 2c).
The laddering was virtually abolished in the absence of the loader Nse5/6 ( Fig. 1d) and also clearly reduced by ATP-hydrolysis-deficient mutations in both Smc5 and Smc6 ('EQ/EQ'), indicating that ATP hydrolysis promotes more robust DNA entrapment. Based on similar experiments using protein gel electrophoresis, we estimate that a large fraction of Smc5/6 entraps plasmid DNA under our experimental conditions (Extended Data Fig. 2d). We conclude that topological DNA loading by the Smc5/6 holocomplex is a robust and efficient reaction.

DNA entrapment in the kleisin compartment
We next wondered where DNA might be located in the Smc5/6 complex. We previously showed that, under the conditions promoting salt-stable DNA binding and DNA entrapment ( Fig. 1c-d), the complex exhibits head engagement as measured by cysteine crosslinking 21 . Head engagement produces an SMC ('S') and a kleisin ('K') compartment Article https://doi.org/10.1038/s41594-023-00956-2 (see schemes in Fig. 2a). DNA must be in at least one of the two compartments. We first combined the cysteine pair at ATP-engaged heads with cysteines at both SMC-Nse4 interfaces to generate a covalently closed K compartment, which yielded robust DNA entrapment again in an ATP-and loader-dependent manner (Fig. 2b, top). The efficiency of entrapment in the K compartment appeared somewhat reduced when compared with the SK ring, but this is probably explained by the reduced efficiency of engaged-head crosslinking (~25%) 21 . A series of samples crosslinked at different time points after mixing showed that entrapment was detectable after 30 s and was saturated within a few minutes in both the SK ring and the K compartment (Extended Data Fig. 2e), supporting the notion that the two types of entrapment emerge from the same biochemical reaction and possibly correspond to the same state. We conclude that the K compartment of the ATP-engaged Smc5/6 holocomplex is occupied by DNA.

DNA segment capture in the SMC compartment
Next, we combined head and hinge cysteines to create a covalently closed S compartment (Fig. 2a). Contrary to the K compartment, no loader-dependent DNA entrapment was observed for the S compartment ( Fig. 2b, 43 . We generated cysteine pairs to crosslink Nse3 to both Smc5 and Smc6 (Extended Data Fig. 4) and combined them with hinge cysteines to create an upper SMC compartment (S up ) and with the head cysteines to generate a lower SMC compartment (S lo ) (Fig. 2a). Intriguingly, the cysteine combinations for the S up and the S lo compartments both resulted in robust DNA laddering (Fig. 2b, lower panels), demonstrating that a DNA segment is efficiently captured in the SMC compartment and held as a loop with (at least) one DNA passage in the S lo compartment and another one (or more) in the S up compartment (Fig. 2c).

The loader Nse5/6 opens the neck gate in Smc5/6
In the course of the above experiments, we noticed that the crosslinking of the cysteine pair at the Smc6-Nse4 interface, designated the neck gate, is strongly sensitive to ligands. Although crosslinking was robust (~90%) with or without ATP and DNA in the absence of the loader (Fig. 3, lanes 2-5), it was strongly hampered when the loader Nse5/6 was added, especially when plasmid DNA was present (Fig. 3, lanes 7 and 8). Such a behavior was not observed for cysteine pairs at the hinge 21 , along the coiled-coil arms (Extended Data Fig. 5) or at the Smc5-Nse4 interface (Extended Data Fig. 6a). Of note, addition of the loader led to some off-target crosslinking of a native cysteine in Nse5 with Smc5 (G1054C) (Extended Data Fig. 6b). More importantly, however, neck-gate crosslinking in the octamer was partially restored by addition of ATP alone (Fig. 3, lane 9) and virtually fully restored by addition of both ATP and DNA (lane 10). Similar crosslinking efficiencies were observed with complexes harboring all cysteines for SK ring crosslinking (Extended Data Fig. 6c). Experiments performed with complexes carrying mutations of active site residues showed that this stimulatory effect of DNA on neck-gate closure requires ATP engagement of Smc5/6 heads (Extended Data Fig. 6d, lanes 9 and 10) but not ATP hydrolysis (Extended Data Fig. 6e, lanes 9 and 10). Altogether, the presented results thus suggest that binding of the Smc5/6 hexamer to the loader Nse5/6 specifically detaches one of the three ring interfaces.
DNA clamping, that is, ATP engagement of SMC heads, together with DNA binding at the head-DNA interface, appears to promote gate closure. It is interesting that fusion of Nse4 to Smc6, but not to Smc5, is lethal in S. cerevisiae, supporting the notion that opening of the neck gate is essential for Smc5/6 function (Extended Data Fig. 6f) 18 . The neck gate is largely closed in the presence of ATP and DNA, implying that, once loaded onto DNA, Smc5/6 holocomplexes form a ring that encompasses the DNA double helix.

DNA clamping is required for efficient gate closure
The above observations suggested that DNA binding plays a role in neck-gate closure. We next determined whether the head-DNA interface is indeed crucial for topological DNA entrapment or important only subsequently, for example, for the conversion of a putative 'DNA-holding' intermediate into the DNA segment-capture state (see below). Based on sequence conservation and structure comparison (Extended Data Fig. 7a-c), we identified three positively charged residues on Smc5 (K89, R139, R143) and another three on Smc6 (R135, K200, K201) as putative DNA-binding residues (Fig. 4a). To test whether these residues are important for DNA entrapment, we mutated them by alanine and glutamate substitution (charge removal and reversal, respectively) in isolation or in combination. Mutant alleles were tested for functional complementation of a respective deletion mutant in yeast using plasmid shuffling 55 . Smc5 alleles harboring double or triple glutamate substitutions resulted in a strong growth phenotype, whereas alanine mutations were apparently tolerated well (Fig. 4b).
The smc6 gene was somewhat more sensitive to the mutagenesis with the triple alanine mutant also exhibiting clear growth retardation (Fig. 4c). A selection of residues based on a recent cryo-EM map 43 resulted in similar outcomes (Extended Data Fig. 7d,e). The deficiencies of the single, double and triple alanine mutants in smc6 were aggravated when combined with the smc5(3A) allele, suggesting that these residues have partially overlapping functions (Extended Data Fig. 7f) as previously observed for bacterial Smc 38 . The sextuple alanine mutant strain (smc5(3A), smc6(3A) or 3A3A) displayed a null-like phenotype, supporting the notion that the head-DNA interface is crucial for an essential function of Smc5/6. Selecting one partially and one fully defective mutant (smc6(3A) and 3A3A, respectively) we tested for DNA loading of Smc5/6 when DNA binding at the heads is compromised. First, we measured ATP-dependent, salt-stable DNA binding by pull-down assays as described previously 21  only poorly recovered plasmid DNA, whereas the 3A3A variant was completely unable to do so, together strongly suggesting that DNA binding is crucial for Nse5/6-mediated DNA loading (Fig. 4d). Next, we combined the Smc6(3A) and the 3A3A mutations with the cysteine variants for SK crosslinking and DNA entrapment. Using the standard buffer conditions (150 mM NaCl), both mutants failed to support any DNA entrapment (Fig. 4e, top). When the reaction buffer was adjusted to mimic the conditions used for salt-stable DNA binding (150 mM KOAc; a 'milder' salt with larger ions), the Smc6(3A) variant supported residual DNA entrapment, whereas the 3A3A variant did not (Fig. 4e,  bottom). These results demonstrate that the head-DNA binding interface is crucial for topological DNA entrapment by Smc5/6. Of note, the mutant complexes showed slightly reduced ATPase activity (Extended Data Fig. 2b). However, the defect in DNA entrapment is not explained by the lack of ATP hydrolysis, because the EQ/EQ complex quite efficiently entrapped DNA despite failing to support any noticeable ATPase activity ( Fig. 4e and Extended Data Fig. 2b). Moreover, the mutant complexes produced ring crosslinking albeit at somewhat reduced efficiency, in particular the 3A3A mutant, an effect that was more pronounced in the stringent salt buffer (Extended Data Fig. 6g). The latter is probably explained by defects in neck-gate closure, because the complex lacking DNA-binding residues on Smc5 and Smc6 heads also failed to display DNA-stimulated neck-gate closure (Extended Data Fig. 6h). DNA binding at the heads might thus serve (at least) two related purposes during DNA entrapment: recruiting DNA for entrapment and promoting gate closure.

DNA passage through the neck gate
We wondered whether DNA might indeed enter the Smc5/6 ring by passing through the neck gate that opens on association with the loader Nse5/6 and closes on DNA encounter. To test this hypothesis directly, we created an Smc6-Nse4 fusion protein. The linker peptide covalently connects the carboxy terminus of the Smc6 polypeptide to the amino terminus of Nse4 and includes a recognition sequence for cleavage by the 3C protease (Extended Data Fig. 8a). This construct is nonfunctional in vivo (Extended Data Fig. 6f; as previously reported for a related construct 18 ). The reconstituted Smc5/6 complex harboring this fusion protein hydrolyzed ATP normally (Extended Data Fig. 2b). We next combined the Smc6-Nse4 fusion protein with cysteines for DNA entrapment experiments (see schematics in Extended Data Fig. 8b). When combining cysteines at the hinge and the Smc5-Nse4 interface with the Smc6-Nse4 fusion protein, the ring species failed to produce any gel shift, demonstrating that DNA entrapment is not possible when Smc6 is linked to Nse4, as expected if the neck gate indeed serves as the DNA entry gate (Fig. 5a, lanes 5-7 and Extended Data Fig. 8b). Curiously, however, we recovered DNA laddering when combining the Smc6-Nse4 fusion with only the neck-gate crosslink albeit with an altered laddering pattern and distinctively smaller steps, probably  owing to the smaller size of the protein ring species formed with Smc6 and Nse4 protein only (Fig. 5a, lanes 8-10 and Extended Data Fig. 8b).
Combining the Smc6-Nse4 fusion with all three cysteine pairs yielded DNA laddering with larger as well as smaller laddering steps, presumably due to partial crosslinking (Fig. 5a, lanes 11-13 and Extended Data Fig. 8b). To reconcile these observations, we suggest that the linker may be long enough to embrace the incoming DNA, thus not blocking the loading reaction itself (Fig. 5b). This would lead to pseudo-topological entrapment of a DNA loop in an expanded 'SK-plus-linker' compartment (see schematics of closed and denatured compartments in Fig. 5a), thus explaining the absence of laddering when the Smc6-Nse4 interface was closed only by the peptide linker but not by the crosslink. Cleavage of the linker peptide by 3C protease before (Extended Data Fig. 8c) or after (Fig. 5c) the loading reaction eliminated the entrapment by the linker compartment and also removed the smaller-sized steps observed with the fusion protein plus all three cysteine pairs. Also consistent with the linker wrapping around DNA, we find that the Smc6-Nse4 fusion complex supported normal or near-normal, salt-stable DNA binding regardless of whether the peptide linker was left intact or cleaved, whereas a control complex with an open kleisin did not (Extended Data Fig. 8d). Taken together, these results strongly suggest that the engineered peptide linker indeed embraces the incoming DNA and identifies the neck gate as the DNA entry gate.
Our experiments imply that all plasmid DNA loading happens via the neck gate. However, a recent study showed that, in addition to the neck gate, the hinge serves as DNA entry gate for topological entrapment by cohesin in vitro 53 . Related experiments performed with the covalently linked Smc5/6 hinge domains (using a SpyTag-SpyCatcher pair; Extended Data Fig. 8e) did not display any noticeable DNA entrapment defects (Fig. 5d), confirming that passage via the hinge does not substantially contribute to DNA entrapment in Smc5/6 and highlighting intriguing differences in DNA loading between these SMC complexes.

The Smc5/6 ring entraps plasmid DNA in vivo
Finally, we wanted to investigate whether DNA entrapment by the Smc5/6 complex also occurs in vivo. We introduced the cysteine pairs for SK ring crosslinking into the nse4, smc5 and smc6 genes in budding yeast by allelic replacement. After BMOE crosslinking in cells, immunoprecipitation and southern blotting analysis, as developed to detect cohesin DNA coentrapment 27 , we could clearly demonstrate the entrapment of a CEN plasmid by Smc5/6, similar to cohesin but without leading to cohesion of plasmid DNA (Fig. 6). A control strain lacking one of the six cysteines for crosslinking of Smc5/6 did not show DNA entrapment.

DNA entrapment by SMC complexes
Many of the diverse SMC functions in genome maintenance and chromosome organization are thought to arise directly from a DNA motor activity 1 . Revealing the DNA topology underlying loop extrusion and translocation is a prerequisite for a basic understanding of SMC activity. We show in the present study that purified preparations of Smc5/6 holocomplexes quickly and effectively entrap circular DNA substrates in a topological manner. This reaction is strictly ATP and loader dependent, and DNA entrapment is also detectable in vivo. We reveal the DNA entry gate and characterize the trajectory of DNA into the Smc5/6 ring, showing that DNA loading involves a looped DNA substrate and eventually leads to the capture of a DNA segment by Smc5/6.
Although the notion of DNA entrapment was challenged again recently by single-molecule imaging, which apparently showed the bypass of very large obstacles on DNA by translocating cohesin and condensin 6 , we argue based on our findings and previous reports that DNA entrapment must not be discounted. Earlier studies measuring salt-stable DNA binding are in support of DNA entrapment by cohesin, condensin and Smc5/6, but they were short of revealing the    Experiments with entire bacterial chromosomes have provided strong evidence for DNA entrapment in the SK ring (and the K but not the S compartment) of Smc-ScpAB and MukBEF but did not reveal the mode of entrapment due to the complexity of the DNA substrate (including branches on the replicating chromosome) 28,41 . Cohesin SK rings are known to support DNA entrapment in vivo at least in the context of sister chromatid cohesion 27 . Single-molecule imaging experiments with cohesin, however, suggest that loop extrusion might not require cohesin ring opening, implying entrapment of a DNA loop or external DNA association rather than topological DNA entrapment 2 . Entrapment of a DNA loop has recently also been proposed for the SK ring of yeast condensin based on coisolation experiments after site-specific crosslinking 50 .
Our data clearly demonstrate the topological DNA entrapment by Smc5/6 in vivo and elucidate its reliance on Nse5/6 in vitro. Nse5/6 has previously been implicated in chromosome loading in vivo by single-molecule tracking in fission yeast 33 . Curiously, Nse5/6 is dispensable for DNA loop extrusion by Smc5/6 in vitro and actually hinders it 5 . To reconcile these observations, one may invoke Nse5/6 converting dynamic loop-extruding Smc5/6 complexes into more stably bound DNA-entrapping complexes. The latter may (or may not) support DNA translocation 54 . The functions of all these states and activities remain to be discerned.

The loader opens the neck gate for DNA entry
Strict topological entrapment (as previously documented for cohesin and in the present paper for Smc5/6) requires the passage of one annular particle through an opening in another. In the present study, we unequivocally demonstrate that the neck gate serves as the major and probably the only entry gate for plasmid DNA in Smc5/6. This conclusion is based on an Smc6-Nse4 fusion protein, in which the linker does not block DNA loading but entraps the incoming DNA in an artificial linker compartment. The identity of the entry gate is furthermore supported by its efficient opening on contact of the Smc5/6 hexamer with the loader Nse5/6 and by the lethality caused by the relevant fusion in vivo in budding yeast 18 . Other possible gates do not seem to contribute to DNA entrapment in the reconstituted reaction (as judged by the absence of DNA entrapment by complexes harboring the Smc6-Nse4 fusion protein (as well as cysteines for crosslinking the hinge and Smc5-Nse4)). The situation appears more complicated in other SMC complexes. The neck gate is known to support DNA exit in cohesin [59][60][61][62][63] and has also been implicated in DNA entry in cohesin 48,58 as well as condensin 64 . A recent paper demonstrated that cohesin DNA entry can be supported by the hinge and the neck gate in vitro, however, with entry only via the hinge being dependent on the loader protein Scc2 (ref. 53), possibly suggesting that the hinge is the main or physiological gate for DNA entry, in contrast to what we observed in the present study for Smc5/6. Moreover, sister chromatid cohesion can be built by cohesin complexes harboring an Smc3-Scc1 'neck-gate' fusion protein. This fusion protein also supports DNA entrapment by cohesin in vivo (when combined with cysteines for crosslinking of the other ring interfaces, unlike observed with Smc5/6 in Fig. 5) 27,65,66 , again implying fundamental differences with Smc5/6. A recent study using fusion proteins, in contrast to earlier reports, suggested that condensin loading does not require an entry gate, at least for DNA loop extrusion 50 .
We find that the association with the loader destabilizes the contact between Smc6 and Nse4. Knowing that the loader intercalates between Smc5 and Smc6 heads 20,21 and that the Nse4 middle part is rather short, we propose that steric constraints prevent Nse4 from being attached to both SMC proteins simultaneously when the loader is also bound, resulting in the detachment of the least stable interface. DNA binding might reverse the effect of the loader, by evicting it from between the SMC heads and enabling productive head engagement (and ATP hydrolysis) 21 as well as ring re-closure by Smc6-Nse4 association. Assuming that the presence of DNA keeps the neck gate shut also during subsequent ATP hydrolysis cycles, this would result in a stable DNA association possibly facilitating DNA translocation over extended periods of time. The neck-gate interface also appears labile in purified preparations of cohesin and condensin (sub)complexes with the binding partners detaching from one another on ATP-head engagement 64,67 (or incubation with cohesin unloading factors 68 ). A similar reaction has been proposed for fission yeast Smc5/6 based on yeast two-hybrid experiments 69 . Neck-gate opening on ATP binding and head engagement, however, diametrically contrasts with our finding on budding yeast Smc5/6, where ATP-dependent head engagement leads to closure rather than opening of the gate (Fig. 7). Whether these mechanistic differences have biological consequences will be important to work out.

DNA segment capture for loading as well as translocation
We propose that DNA loading involves DNA segment capture as an essential intermediate. We find that head-DNA binding is required for gate closure as well as for DNA entrapment. The head-DNA contacts conceivably hold a bent DNA segment in place during loading and further DNA contacts guide the K subunit around DNA to ensure entrapment of a DNA double helix in the SK ring (Fig. 2c). Of note, a model proposed for cohesin DNA loading via the neck gate resembles our notion of DNA segment capture 48 . However, DNA loading via the cohesin neck gate has since been shown to work regardless of the absence or presence of the loading-clamping subunit Scc2, thus challenging this notion 53 .
Why would DNA loading of SMC complexes involve such an intricate structure rather than rely on entrapping a single DNA double helix? DNA segment capture has previously been proposed to explain DNA translocation and DNA loop extrusion by bacterial Smc-ScpAB and other SMC complexes 36,54 (Fig. 7). We envisage that the segment-capture states during loading and translocation/loop extrusion are structurally related. The notion of DNA segment capture mediating DNA loading as well as translocation explains how the very same ATPase core supports two distinct biochemical reactions and thus provides a unified framework for SMC ATPase functions, possibly also applicable to other SMC-related processes. The segment-capture state is consistent with observations on the bacterial SMC complexes Smc-ScpAB and Muk-BEF made by in vivo crosslinking and chromosome isolation, as well as a cryo-EM structure of the MukBEF complex encompassing two separate DNA molecules 38,41 . If true, the similarities in DNA topology are remarkable, especially when considering the large evolutionary distances across the three SMC complexes. Notably, the entrapment results were obtained with cysteines reporting distinct ATPase states (that is, predominantly the juxtaposed and the ATP-engaged state for Smc-ScpAB and Smc5/6, respectively; in the case of MukB the cysteines did not strongly discriminate between the states), indicating that DNA topology does not change after initial loading and the entry gate may remain shut during subsequent ATP hydrolysis cycles (Fig. 7).   50,54 . Future experiments will have to elucidate how exactly DNA entrapment by Smc5/6 is linked to its function and its activities as translocator and loop extruder. Although a segment capture-type mechanism explains our findings, other suitable scenarios may exist or emerge in the future.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41594-023-00956-2. Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.

Cloning of expression plasmids
Plasmids containing multiple expression cassettes for subunits of the S. cerevisiae Smc5/6 complex were cloned as described in Taschner et al. 21 . Transcription of Smc5-containing cassettes was regulated with a tac-promoter and lambda-terminator, whereas the other subunits were transcribed by T7 promoters and terminators. A list of all expression plasmids used in the present study can be found in Supplementary  Table 1.

Protein expression and purification
All proteins and protein complexes described in the present study were expressed in Escherichia coli (DE3) Rosetta transformed with either a single plasmid or a combination of two plasmids. Supplementary Table  2 lists plasmid details about all described complexes. For all purifications, 1 liter of the strain carrying the desired plasmid(s) was grown in Terrific Broth medium at 37 °C to an optical density at 600 nm (OD 600 ) of 1.0 and the culture temperature was reduced to 22 °C. Expression was then induced with isopropyl β-d-1-thiogalactopyranoside at a final concentration of 0.4 mM and allowed to proceed overnight (typically for 16 h). All Smc5/6 complexes and the Nse5/6 dimer were purified following a published procedure described in detail in Taschner et al. 21 .

Cloning of yeast plasmids
Coding sequences for wild-type or mutant Smc6, Smc5 or Nse4 were cloned between their respective endogenous upstream and downstream regulatory sequences by Golden Gate cloning into centromeric acceptor plasmids pCEN(URA) or pCEN(TRP). N-or C-terminal tags were added as indicated. For plasmids containing two loci, we combined the individual cassettes by an additional Golden Gate Assembly step. A list of all yeast plasmids used in the present study can be found in Supplementary Table 3.

Creation of yeast strains
All strains in the present study were created on the W303 background. Functional assays to investigate mutants of Smc6, Smc5 and Nse4 were carried out using plasmid shuffling 55 . To generate suitable strains, we first deleted the respective loci in a diploid strain then transformed the resulting strain with a pCEN(URA) plasmid containing the wild-type locus. Sporulation of this strain allowed us to isolate haploid offspring with the locus on the pCEN(URA) as the sole source for the respective protein. For double shuffling strains, both loci were deleted in the diploid and re-introduced on the pCEN(URA) plasmid. Haploid shuffling strains were transformed with pCEN(TRP) plasmids containing either a wild-type or a mutant version of the respective loci. A list of all yeast strains used in the present study can be found in Supplementary  Table 4.

Plasmid shuffling assays
Haploid shuffling strains containing the wild-type locus on pCEN(URA) and another wild-type or mutant locus on pCEN(TRP) were grown on plates lacking uracil and tryptophan. Single colonies were inoculated in minimal medium lacking only tryptophan and grown for 18-24 h at 30 °C. Four tenfold dilutions were then prepared in water and 2 µl of the five samples (undiluted culture and the four dilutions) were spotted on one plate lacking uracil and tryptophan, and on another containing all amino acids as well as 1 mg ml −1 of 5-fluorouracil (5-FOA) (selecting for cells that had maintained or lost the pCEN(URA) plasmid, respectively).
Plates were incubated at 30 °C and pictures were taken at suitable time points (between 36 h and 60 h) to score growth of strains containing mutant versions of the respective Smc5-6 components.

Site-specific BMOE crosslinking in vitro
Smc5/6 hexamers with or without indicated cysteines were diluted to a final concentration of 0.5 µM in ATPase buffer (10 mM Hepes-KOH, pH 7.5, 150 mM KOAc, 2 mM MgCl 2 and 20% (v:v) glycerol) in a total volume of 30 µl. In reactions containing the Nse5/6 dimer, this complex was added in a 1.25× molar excess (0.625 µM). For reactions containing ATP and/or plasmid DNA (25 kbp), these ligands were added at a final concentration of 2 mM and 5 nM, respectively. The circular or linear plasmid substrate was produced as described in Taschner et al. 21

Analysis of salt-stable DNA association
These assays were carried out as described in Taschner

Mini-chromosome entrapment assays
Entrapment of mini-chromosomes in vivo in budding yeast was performed following a procedure previously used for cohesin 66 .
Article https://doi.org/10.1038/s41594-023-00956-2 Briefly, strains expressing a crosslinkable version ('6C') of cohesin or Smc5/6 and containing a TRP mini-chromosome were grown together with control strains lacking one of the ring cysteines ('5C') overnight in medium lacking tryptophan. When the cultures reached an OD 600 of around 0.5, 40 ml (20 ODs) of the cultures was harvested by centrifugation (3,000 r.p.m., 3 min), and the pellets were washed once with 25 ml ice-cold phosphate-buffered saline (PBS) and subsequently transferred to screw-cap tubes. Pellets were then resuspended in 500 ml of ice-cold PBS; 30 µl of a 150 µM BMOE solution in dimethyl sulfoxide was added and crosslinking was allowed to proceed for 6 min on ice. The cells were then pelleted, the supernatant was discarded and cell pellets were snap-frozen in liquid nitrogen for storage at −80 °C. For lysis, 700 µl of lysis buffer (50 mM Hepes-KOH, pH 7.5, 100 mM KCl, 0.05% (v:v) Triton X-100, 0.025% (v:v) NP-40, 10 mM Na citrate and 25 mM Na sulfite), freshly supplemented with a cOmplete protease inhibitor tablet (Roche) and 1 mM phenylmethylsulfonyl fluoride, was added to the pellet together with acid-washed glass beads (425-600 µm), and the cells were opened using bead beating with a FastPrep 24 system (MPBio; 3× 1 min with 5-min breaks on ice). The lysate was clarified by centrifugation at 14,000g for 10 min at 4 °C, incubated with mouse monoclonal anti-V5 antibody (BioRad) for 90 min at 4 °C to bind the C-terminal Pk6-epitope on Smc6 and complexes were then captured with Dyna-Beads Protein-G for another 90 min at 4 °C. Beads were washed twice with 1 ml of wash buffer (10 mM Tris-HCl, pH 8.0, 250 mM LiCl, 0.5% (v:v) NP-40, 0.5% (w:v) Na deoxycholate and 1 mM EDTA) and once with 1 ml of TE (Tris and EDTA), and bound material was then eluted with 1× DNA-loading dye in TE supplemented with 1% (w:v) SDS and preheated to 65 °C. The obtained material was separated on a 0.8% (w:v) agarose gel (in 1× TAE (TE + acetic acid)) overnight at 4 °C and 1.5 V cm −1 , and DNA was then transferred to Hybond-N + membranes by southern blotting using alkaline transfer. Mini-chromosomes were detected with a probe against the TRP marker, which was labeled with [α-32 P]ATP using the Prime-It II Random Primer Labeling Kit (Agilent) according to the manufacturer's instructions. Membranes were exposed to a Phosphor Storage Screen (Fujifilm) and signals were detected with a Typhoon Scanner (Cytiva).

ATPase assays
Analysis of ATPase activity of selected Smc5/6 complexes was carried out exactly as described in Taschner et al. 21 .

Statistics and reproducibility
Determination of ATP hydrolysis rates was performed in three technical replicates. Mean and s.d. values were calculated. No data were excluded from the analyses. Qualitative assays of DNA entrapment were reproduced at least three times from independently grown cultures or purified material with comparable outcomes.

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

Data availability
The following publicly available datasets were used as a guide for site-directed mutagenesis: PDB accession nos. 6ZZ6 and 7TVE. Source data are provided with this paper.   evolutionary conservation of examined residues in Smc6. Smc5 residues show weaker conservation consistent with results of functional assays shown in Fig. 4. (d) Positively charged residues on the Smc5 head were mutated to alanine ('A') and the mutant alleles were tested for function by plasmid shuffling. As in Fig. 4b but with residues selected based on a recent cryo-EM structure of Smc5/6 (Yu et al. 43 ). (e) As in (D) but for residues on the Smc6 head. (f) Positively charged residues on Smc5 and Smc6 heads were mutated to alanines in combinations. As in Fig. 4b