Structural Basis for a Safety-Belt Mechanism That Anchors Condensin to Chromosomes

1In preparation for cell divisions, eukaryotic chromosomes undergo large-scale conformational changes to form rod-shaped structures that enable their successful segregation into the daughter cells. Multisubunit protein complexes named condensins have been recognized as the major molecular machines that coordinate these changes in genome architecture. Condensins furthermore fulfill pivotal roles in many other aspects of nuclear function, including the regulation of gene expression, during interphase.

2Like other members of the structural maintenance of chromosomes (SMC) family of protein complexes, condensins are characterized by a large, ring-shaped architecture. The condensin ring is formed by heterodimerization of its Smc2 and Smc4 subunits via globular “hinge” domains, which are located at one end of ∼40-nm-long intramolecular antiparallel coiled coils, and the connection of ATP-binding cassette (ABC)-transporter-like ATPase domains at the other end of the coils by the Brn1Cnd2, NCAPH/H2 kleisin subunit (Figure 1A). The central region of the kleisin recruits to the condensin ring the Ycs4Cnd1, NCAPD2/D3 and Ycg1Cnd3, NCAPG/G2 subunits, which contain tandem repeats of amphipathic α helices referred to as HEAT (huntingtin, elongation factor 3, protein phosphatase 2A, Tor1 kinase) motifs.

3Condensin rings, like the related cohesin and prokaryotic SMC complexes, are thought to encircle chromosomal DNA. This topological mode of DNA binding might form the basis for the creation of large chromatin loops, their maintenance, or both. Recent polymer dynamics simulations demonstrated that loop extrusion by condensin can, at least in principle, recapitulate the formation of cylindrical mitotic chromosomes in silico and produce structures that are consistent with electron micrographs and chromosome-conformation-capture contact maps of mitotic chromosomes. Yet, it is difficult to imagine how mere topological entrapment of chromatin fibers within ring-shaped protein complexes could conceivably result in the creation of loops of several kilobase pairs in size or achieve the active compaction of DNA substrates observed in magnetic tweezers experiments.

4It hence seems inevitable that condensin needs to make direct contact with DNA. In vitro DNA-binding experiments suggest that the Smc2–Smc4 dimerization hinge interface is able to bind to short, single-stranded, but surprisingly not to double-stranded (ds), DNA molecules. In contrast, a “non-SMC” subcomplex composed of the central region of the kleisin and HEAT-repeat subunits binds double-stranded, but not single-stranded, DNA. The non-SMC subcomplex plays an integral role in the association of condensin with chromosomes, since chromosomal localization of complexes that lack either HEAT-repeat subunit is largely restricted to the axes of chromosomes assembled in Xenopus laevis egg extract and complexes without the Ycg1 HEAT-repeat subunit fail to associate with mitotic chromosomes in budding yeast and human cells. Nevertheless, the mechanistic basis for the loading of condensin complexes onto chromosomes and the role of the HEAT-repeat subunits in this process have remained unknown.

5Here, we unveil a direct DNA interaction site in the non-SMC subcomplex, which is formed by the Ycg1 HEAT-repeat and Brn1 kleisin subunits. Co-crystal structures of Ycg1–Brn1 with and without DNA duplexes at near-atomic resolution reveal a conserved, positively charged groove. DNA bound within the groove is locked into place by its entrapment by a peptide loop of the kleisin subunit. We demonstrate the contributions of groove and kleisin loop for condensin binding to DNA in vitro and to mitotic chromosomes in vivo, reveal its function in the DNA-dependent stimulation of the Smc2–Smc4 ATPase activity, and identify a regulatory mechanism for loop closure in human cells. Our findings suggest a “safety-belt” mechanism that enables condensin to stably bind chromosomes independent of DNA sequence and provides a possibility for condensin to form chromatin loops.

6Even though the non-SMC subcomplex of condensin binds dsDNA, none of its proteins contain conventional DNA-binding motifs. To narrow down the position of the DNA interaction site, we expressed and purified Chaetomium thermophilum (Ct) Ycs4, Ycg1, and Brn1 subunits individually and tested their binding to a 35-bp DNA substrate in an electrophoretic mobility shift assay (EMSA; Figure 1B). None of the individual subunits reproduced the distinct DNA upshift that we had observed with the Ct Ycs4–Ycg1–Brn1 non-SMC subcomplex (Figure S1A). Since Ycs4- and Ycg1-HEAT-repeat subunits do not stably interact with each other directly, DNA binding might require complex formation between the Brn1 kleisin and either, or possibly both, of the HEAT-repeat subunits. The weak DNA binding by the purified Ct Ycs4 protein was, however, reduced—rather than enhanced—by addition of a purified fragment of Ct Brn1 that binds to both HEAT-repeat subunits (Ct Brn1336–714; Figure 1B). A copurified complex of Ct Ycs4 and the region of Ct Brn1 that binds specifically to this HEAT-repeat subunit (Ct Ycs4–Brn1225–512) similarly failed to shift DNA efficiently (Figure S1B).

7In contrast, addition of Ct Brn1336–714 to Ct Ycg1 resulted in a distinct DNA upshift (Figure 1B). We observed a similarly effective DNA upshift with a copurified subcomplex of Ycg1 and its interacting region of Brn1 (Ct Ycg1–Brn1515–634; Figure S1B), which equaled the upshift measured for the Ct Ycs4–Ycg1–Brn1 non-SMC complex (Figure S1A). Quantitation of the DNA interaction of the Ct Ycg1–Brn1515–634 subcomplex by isothermal titration calorimetry (ITC; Figure 1C) revealed a dissociation constant Kd of 1.7 ± 0.2 μM, which is identical to the affinity we had previously measured for the Ycs4–Ycg1–Brn1 ternary complex (Kd = 1.7 ± 0.2 μM, ). DNA binding required the presence of the first 16 of 19 HEAT-repeat motifs of Ycg1 (Ct Ycg124–883; Figure S1C) and the C-terminal two-thirds of the interacting Brn1 region (Ct Brn1549–634; Figure S1D). We conclude that the main DNA-binding activity of the non-SMC subcomplex resides within the Ycg1-HEAT-repeat subunit and the region of the Brn1 kleisin subunit it interacts with.

8To gain detailed insight into the nature of the DNA binding site, we solved crystal structures of the Ycg1–Brn1 subcomplex of Saccharomyces cerevisiae (Sc) and of the homologous Cnd3–Cnd2 subcomplex of Schizosaccharomyces pombe (Sp) to 2.8 Å and 2.6 Å resolution, respectively (Table S1). Both structures are strikingly similar and reveal a harp-shaped conformation of the HEAT-repeat subunit, with the kleisin subunit winding along the concave surface of the HEAT-repeat solenoid (Figures 2A, 2B, and S2A). The peculiar curvature of Ycg1 and Cnd3 has its cause in three discontinuities in the directionality of the solenoid formed by the 19 HEAT-repeat motifs (Figures S2B and S3A). Several residues that have previously been found to destabilize Cnd3 are involved in contacts between the repeat helices (Figure S2C), which emphasizes the importance of the HEAT-repeat solenoid conformation.

9A comparison with known crystal structures of the cohesin-associated HEAT-repeat subunits SA2, Pds5, and Scc2 shows that all four proteins have similarly curved shapes. Yet, the relative positions and orientations of the individual HEAT motifs vary notably between Ycg1 and the cohesin HEAT-repeat subunits (Figure S4). It is hence conceivable that these subunits have evolved from a common ancestral protein to fulfill distinct functions in cohesin and condensin complexes (see Discussion).

10The Brn1 kleisin subunit makes extended contacts with the inner surface of the Ycg1 harp, burying a total surface area of 3,754 Å2. Several conserved aromatic residues within a Brn1 segment that contains a short α-helical stretch (α2) are involved in hydrophobic interactions with HEAT motifs 1 and 2 of the N-terminal “shoulder” of the harp (Figures 2B and 2C). The next ∼20 residues form a well-ordered linker that contains a short helix (α3) and follows the inward-facing ridge of HEAT motifs 3–6 of the N-terminal lobe of Ycg1 before crossing over to the C-terminal lobe of Ycg1. Through multiple hydrophobic, electrostatic, and hydrogen bond interactions with the HEAT-repeat subunit, Brn1 forms a rigid “bridge” between the two Ycg1 lobes (Figures 2B and 2D). The consecutive segment that follows contains another short helix (α4) and a conserved coiled region that packs through hydrophobic interactions and hydrogen bonds onto Ycg1 HEAT motifs 13–15 (Figure 2D).

11One of the two molecules in the asymmetric unit of the Ycg1–Brn1 and Cnd3–Cnd2 crystals shows additional electron density for a segment that corresponds to a long helix (α1) in the N-terminal region of the kleisin subunit (Figures 2A and 2B). This helix and the succeeding well-ordered turn stack onto HEAT motifs 16 and 17 and onto the α4 segment of the kleisin via several highly conserved aromatic residues in a manner that resembles a “latch” that has been inserted into a “buckle” (Figures 2B and 2E). Through the interaction of latch and buckle regions, the kleisin subunit adopts a loop configuration (Figure 2B). As we demonstrate below, this loop has important implications for DNA binding.

12The absence of density for the latch segment in half of the complexes suggests that the α2 to α4 segments of the kleisin subunit are sufficient for complex formation with the HEAT-repeat subunit. This notion is consistent with the finding that mutation of patches of conserved hydrophobic residues within the Brn1 α2 (Sc Brn1461–464) or α4 (Sc Brn1498–504 or Sc Brn1512–514) segments, which make major contacts with the HEAT repeats, disrupt binding to Ycg1 (Figure S2D; ). Mutation or deletion of conserved patches that contribute to the stacking of the N-terminal α1 segment of Brn1 (Sc Brn1393–394 or Sc Brn1402–410) has, in contrast, little effect on the overall affinity to Ycg1, while nevertheless rendering condensin non-functional. Whereas the latch segment of the kleisin subunit might not be essential for complex formation with the HEAT-repeat subunits, it is essential for condensin function.

13Mapping of the electrostatic surface potential onto the Sc Ycg1–Brn1 structure revealed that the wide groove between the two shoulders of the Ycg1 harp and the Brn1 bridge displays a high degree of positive surface charge (Figure 3A). To test whether this groove is involved in DNA binding, we designed charge-reversal mutations for patches of highly conserved basic groove residues located within the N- or C-terminal lobes of Ycg1 (YC1 or YC2) or at either end of the disordered linker of Brn1 (BC1 or BC2; Figures 3A, 3B, S3A, and S3B). Mutation of any of these four patches strongly reduced the affinity of Ct Ycg1–Brn1 subcomplexes for DNA in vitro (Figures 3C) without affecting protein complex formation (Figure S5A). Combination of the patch mutations in Ycg1 (YC1/2) or Brn1 (BC1/2) even completely abolished any measurable DNA interaction of Ct Ycg1–Brn1 or Ct Ycs4–Ycg1–Brn1 subcomplexes (Figures 3C and S5B). These findings confirm that the direct protein-DNA binding site in condensin’s non-SMC subcomplex is formed by Ycg1–Brn1, with little or no contribution from the other HEAT-repeat subunit. Mutation of Brn1 BC1/2 residues was even sufficient to abolish DNA binding by condensin holocomplexes, which, like complexes lacking the Ycg1 subunit, failed to shift DNA in vitro (Figure S5C). We conclude that the Ycg1–Brn1 groove forms an essential protein-DNA interaction site in condensin complexes.

14If this conclusion were correct, condensin complexes with mutations in the DNA binding groove should no longer be able to associate with chromosomes in vivo. Indeed, mutation of both positively charged patches in Brn1 (BC1/2) substantially reduced the levels of condensin on chromosomes when measured by chromatin immunoprecipitation followed by quantitative PCR (ChIP-qPCR) at two independent binding sites in the budding yeast genome (Figure 3D), although the mutations did not affect the assembly of condensin holocomplexes (Figure S5D). As expected, cells expressing the BC1/2 double mutant version as their only source of Brn1 failed to divide, and cells expressing only BC1 or BC2 single mutant versions displayed proliferation defects (Figure 3E) despite producing the mutant Brn1 proteins at wild-type levels (Figure S5E).

15Finally, we tested whether DNA binding by the HEAT-repeat and kleisin subunits is a conserved feature in the two distinct condensin complexes found in most metazoan organisms. We introduced BC1/2 mutations into the NCAPH or NCAPH2 kleisin subunits of Homo sapiens (Hs) condensin I or II, respectively (Figure S3B). Whereas wild-type versions of NCAPH or NCAPH2 fused to enhanced green fluorescent protein (EGFP) distinctively stained the axes of mitotic chromosomes in live cells, mutant versions of either kleisin subunit remained exclusively cytoplasmic (Figure 3F). We conclude that DNA binding by the Ycg1–Brn1 groove is essential for the stable association of condensin complexes with chromosomes and consequently condensin function.

16To unambiguously determine the position and orientation of DNA bound to the Ycg1–Brn1 subcomplex, we solved the co-crystal structure of Sc Ycg1–Brn1 with 18-bp dsDNA to 2.98 Å resolution (Figure 4A and Table S1). As expected from the surface charge distribution and the results of the mutation experiments, we detected additional electron density corresponding to a DNA double helix within the Ycg1–Brn1 groove (Figure S5F). In the crystal, DNA duplexes pack via base-stacking interactions, with one copy of Ycg1–Brn1 contacting two-thirds of one DNA molecule and one-third of a symmetry-related DNA molecule (Figures 4A and S5F). The DNA helix is unwound by ∼65 degrees over 11 bp and bent by 7 degrees at one end, which results in an offset in the end-to-end stacking arrangement. This conformation is most likely necessary to compensate for the fact that the 18-bp DNA is slightly too long for the 1.5 helical turns required to produce the 21-screw axis observed in the crystal and is not a consequence of binding to Ycg1–Brn1. Several conserved basic residues of both protein subunits, including those we had mutated (Sc Ycg1K70, K71, Sc Ycg1K849, Sc Brn1K409, R411, and Sc Brn1K456 corresponding to Ct Ycg1K100, K101, Ct Ycg1K917, Ct Brn1R539, R541, and Ct Brn1K559), interact with the phosphate backbone of both DNA strands (Figures 4B and 4C). Contacts with both backbones of the duplex and the absence of interactions with the nucleobases explain the lack of sequence specificity in DNA binding by the non-SMC subcomplex and preference for double- over single-stranded DNA.

17Comparison of free and DNA-bound structures shows that Ycg1 and Brn1 undergo fairly minor conformational changes upon DNA binding. In addition to a slight shift of the N- and C-terminal shoulders of Ycg1, the most notable difference upon DNA binding is the reorientation toward the double helix of residues within the α1 latch segment of Brn1, including two positively charged residues of the BC1 patch (Sc Brn1K409 and Sc Brn1R411; Figure 4D). The change in the conformation of the latch segment might either help to stabilize DNA binding or alter the affinity between the latch and the buckle. The latter hypothesis is consistent with the formation of hydrophobic contacts between a valine residue located between the two charged residues (Sc Brn1V410) and the buckle residues (Sc Ycg1F808 and Sc Brn1W408) in the DNA-bound structure (Figure 4D).

18Whereas several basic residues of Ycg1 and Brn1 directly contact the DNA phosphate backbone, several other residues are oriented toward the DNA but are too far away to make direct electrostatic interactions (Figures 4B and 4C). As a result, the Ycg1–Brn1 groove might allow some degree of flexibility in accommodating the DNA double helix. This notion is further supported by a second crystal structure of Ycg1–Brn1 in complex with the same 18-bp DNA substrate, which crystallized with slightly different unit cell dimensions (Table S1). Compared to the first DNA co-structure, the C-terminal Ycg1 shoulder is slightly translocated (Figure 4E), and as a consequence, the contacts between this part of Ycg1–Brn1 and DNA differ considerably from the other crystal form (Figure 4C). The high elasticity in the conformation of the two Ycg1 shoulders and the low affinity measured for short DNA substrates in vitro (Figure 1C) raises the question of how condensin can nevertheless use the Ycg1–Brn1 groove to achieve a stable interaction with chromosomal DNA in vivo.

19Even in the co-crystal structures of Sc Ycg1–Brn1 with DNA, 43 residues of the linker that connects the α1 and α2 segments of Brn1 remained disordered. The trajectories of the visible residues that flank the linker, however, indicate that the path of the connecting peptide leads over the bound DNA helix (Figure 4A). Since the length of this linker is considerably shorter in most other species (Figure S3B), we shortened the linker of Sc Brn1 by 27 residues to match the linker length of the human NCAPH kleisin subunit and co-crystallized the new construct with DNA (Table S1). The structure of this construct very closely resembles the previous Sc Ycg1–Brn1–DNA structures (Figure S5G). Although the shortened linker also remained disordered in this structure, the only conceivable way to span the distance of ∼36 Å between the two visible ends with the short linker is by running the connecting peptide over the DNA helix. As a consequence, the Brn1 kleisin subunit encircles the DNA double helix within the loop that is formed by the association of its N-terminal latch and C-terminal buckle segments (Figure 4A).

20If this were indeed the case, then covalent connection between the contact points of the loop would create a circularized Brn1 peptide that should continue to entrap topologically restrained circular (but not linear) DNA even after protein denaturation (Figure 5A). We first engineered three pairs of cysteine residues into the Brn1 latch and buckle segments (Ct Brn1E514C, R629C, Ct Brn1L521C, S611C, and Ct Brn1D525C, S610C, which correspond to Sc Brn1S384, S524, Sc Brn1M391, T506, and Sc Brn1D395, S505) at positions that should be crosslinked with the thiol-reactive compound dibromobimane (bBBr; Figure S6A). In addition, we designed three pairs of cysteine residues in Brn1 that should be too far apart for crosslinking by bBBr (Ct Brn1E514C, Q613C, Ct Brn1A527C, R629C, and Ct Brn1S568C, R629C, which correspond to Sc Brn1S384, R508, Sc Brn1N397, S524, and Sc Brn1K465, S524). As expected, only the former three cysteine combinations produced a crosslink between the two ends of the Brn1 loop, which was detectable by SDS-PAGE after site-specific cleavage of the Brn1 linker region with human rhinovirus (HRV)-3C protease (Figure S6B). The crosslinking experiments hence confirm that, even in soluble Ycg1–Brn1 complexes, latch and buckle regions associate with each other as revealed by the crystal structures.

21We then incubated Ycg1–Brn1 subcomplexes with circular DNA substrates before the addition of bBBr, denatured the proteins at 65°C with 1% SDS, and detected DNA by ethidium bromide (EtBr) staining after agarose gel electrophoresis. Addition of bBBr crosslinker, but not of DMSO solvent only, resulted in a marked reduction of the mobility of the DNA for the samples that contained crosslinkable Brn1 cysteine combinations, but not for samples without Brn1 cysteine pairs or with cysteine combinations that were too far apart to be crosslinked (Figure S6B). We selected one crosslinkable cysteine combination (Ct Brn1E514C, R629C) and repeated the experiment with two additional thiol-reactive compounds, bismaleimidoethane (BMOE) and dithiobismaleimidoethane (DTME), on linear and circular DNA substrates. Whereas crosslinking of Ct Ycg1–Brn1E514C, R629C with BMOE or DTME, like bBBr, produced an upshift of the circular DNA, none of the crosslinkers produced an upshift of the linear DNA substrate (Figure 5B). These results suggest that the DNA upshift is due to the entrapment of the circular DNA species within the chemically circularized Brn1 loop, in contrast to linear DNA, which presumably can slide out of the Brn1 peptide loop.

22Importantly, chemical cleavage of the covalent linkage between the Brn1 cysteine residues by reduction of the disulfide linker of DTME with dithiothreitol (DTT), as well as site-specific proteolytic cleavage of the Brn1 loop peptide backbone with HRV-3C protease after BMOE crosslinking, fully reverted the upshift of circular DNA substrates (Figures 5B, 5C, and S6C). Furthermore, addition of BMOE created a crosslink of the Brn1 loop and resulted in the formation of an SDS-resistant complex with circular DNA substrates even in the context of the Ct Ycs4–Ycg1–Brn1 ternary complex (Figure S6D). These experiments provide proof that the DNA bound within the Ycg1–Brn1 groove must be enclosed by the Brn1 loop.

23Since latch and buckle segments are able to engage with each other even in the absence of DNA (Figure 2), the latch would need to disengage at least temporarily from the buckle to allow the entry of DNA into the Brn1 kleisin loop. If this were the case, locking latch and buckle segments by covalent crosslinking should prevent any subsequent DNA entrapment by the kleisin loop. We repeated the kleisin loop crosslinking experiment but, this time, crosslinked the Ct Brn1E514C, R629C cysteine pair before adding DNA. This indeed reduced the affinity of the Ct Ycs4–Ycg1–Brn1 ternary complex for a circular DNA substrate (Figure S7A). DNA was still upshifted at higher protein concentrations, either because crosslinking had been incomplete or because the long flexible kleisin loop was able to assume an orientation that still allowed access to the Ycg1–Brn1 DNA binding groove even when its two ends had been covalently linked.

24These findings suggest that the latch-buckle interaction is essential for condensin’s association with DNA. Notably, many residues of the Brn1 latch and buckle segments are remarkably conserved among condensin kleisin subunits (Figures 6A and S3B). To test the functional importance of the latch-buckle interaction, we shortened Brn1 to create a version that lacks the α1 latch segment (ΔN: Ct Brn1539–634) or mutated conserved hydrophobic residues within this region to negatively charged residues (ΦD: Ct Brn1L521D, F524D, W532D, W538D). Both modified versions of Brn1 still formed a complex with Ycg1 (Figure S7B), which confirms that the latch segment is not essential for the stable interaction between the two subunits.

25We then compared binding of wild-type and mutant Ct Ycg1–Brn1 subcomplexes to short linear or long circular DNA substrates by EMSA (Figures 6B and 6C). All Ct Ycg1–Brn1 subcomplexes bound to the linear 35-bp DNA substrate with similar affinities (non-shifted DNA depleted at 800 nM protein; Figure 6B). This suggests that the kleisin loop is unable to stabilize binding to short linear DNA fragments, which might easily slide out of the loop. Wild-type Ct Ycg1–Brn1 subcomplexes shifted the 2.1-kb circular DNA at much lower protein concentrations (non-shifted DNA depleted at 21 nM; Figure 6C). The low micromolar value we measured by ITC for a short 25-bp duplex (Figure 1C) might hence considerably underestimate the affinity of Ycg1–Brn1 to DNA of physiologically relevant lengths, which can be topologically restrained by the kleisin loop. Importantly, shortening or mutation of the latch segment greatly reduced the affinity of Ct Ycg1–Brn1 subcomplexes for 2.1-kb circular DNA (non-shifted DNA depleted at 83 nM; Figure 6C). The Ct Ycs4–Ycg1–Brn1 ternary complex likewise bound a circular 2.1-kb DNA with higher affinity when compared to a linear DNA of the same size (Figure S7C). Mutation of the Brn1 latch region similarly increased the amounts of ternary complexes (Figure S7D) or even condensin holocomplexes (Figure S7E) required to shift the 2.1-kb circular DNA substrate, although the effect on DNA binding was less severe than mutation of the Brn1 BC1/2 patches.

26Kleisin loop closure is furthermore important for chromosome association and function of condensin complexes in vivo, since the latch mutant version of Brn1 (ΦD: Sc Brn1M391D, F394D, W402D, W408D) failed to support proliferation in budding yeast (Figure 6D) despite being expressed at wild-type levels (Figure S7F). Mutation of the analogous residues in the human NCAPH kleisin subunit to either charged (ΦD: Hs NCAPHY428D, F431D, W439D, W445D) or small hydrophobic side chains (ΦA: Hs NCAPHY428A, F431A, W439A, W445A) dramatically reduced condensin association with mitotic chromosomes in cultured cells (Figure 6E). We conclude that while DNA seems to be able to access the basic Ycg1–Brn1 groove even without being encircled by the kleisin loop, only latch-mediated entrapment increases the affinity sufficiently to allow a stable interaction with chromosomes.

27Addition of DNA had been reported to stimulate the ATPase activity of condensin holocomplexes, but not of Smc2–Smc4 dimers alone. This implies that the DNA binding site in the Ycg1–Brn1 subcomplex might be able to act as a sensor that triggers activation of the Smc2–Smc4 ATPase activity. Whereas the presence of a 6.4-kb circular DNA enhanced the ATPase activity of Sc condensin holocomplexes containing wild-type Brn1 more than 5-fold, it had little effect on the ATPase rates measured for complexes containing Sc Brn1BC1/2 or Sc Brn1ΦD mutant versions (Figure 6F). Brn1 loop closure around the DNA double helix is hence essential to activate ATP hydrolysis by the Smc2–Smc4 ATPase domains.

28Although SMC protein complexes have been identified as the principal molecular machines that determine the three-dimensional organization of eukaryotic genomes, it had remained unclear how the complexes interact with their DNA substrates. Our structures reveal the formation of a composite DNA-binding groove by the HEAT-repeat and kleisin subunits and the entrapment of the DNA double helix by the kleisin peptide loop. This unconventional mode of DNA binding differs entirely from previously observed interactions with nucleic acids of other proteins that contain HEAT-repeat motifs. The facts that DNA contacts are made exclusively with the phosphate backbone and are conformationally adaptable provide the HEAT-repeat and kleisin subunits of condensin with the means to bind DNA independent of sequence or specific local structure, which contrasts sharply with the conventional concept of protein-DNA recognition. This binding mode might target condensin to chromosomal positions where the double helix is freely accessible—for example, at transcriptionally active genes and other nucleosome-depleted regions.

29The absolute requirement for the Ycg1–Brn1 DNA-binding groove for the stable association of condensin with chromosomes in yeast and human cells suggests that it might play a direct role in loading condensin onto chromosomes and might thereby enable condensin to bypass the requirement for specific chromosomal loader proteins. Cohesin complexes, in contrast, rely on the function of a separate Scc2Nipbl–Scc4Mau2 dimer for their loading onto chromosomes. Since the architecture of the Scc2 HEAT-repeat solenoid very closely resembles the shape of the Ycg1 harp (Figure S4), it is tempting to speculate that both proteins might bind DNA in a similar manner.

30The conformational flexibility in the contacts with DNA revealed by two Ycg1–Brn1–DNA crystal forms and the low binding affinity to short linear DNA fragments suggest that the positively charged groove alone can only weakly interact with DNA duplexes. Rapid dissociation of DNAs of physiologically relevant lengths bound in such a manner is only prevented by their entrapment within the kleisin peptide loop. The kleisin thereby acts analogous to a safety belt, akin to the way the safety belt of the spindle checkpoint protein Mad2 prevents dissociation of its protein ligands. In this model, belt closure by association of latch and buckle regions anchors condensin to chromosomal DNA, which is consistent with the finding that mutation of conserved hydrophobic residues that are involved in the latch-buckle contact strongly reduces condensin binding to mitotic chromosomes. Since the safety belt cannot stabilize binding to short linear DNA substrates, it is conceivable that, even when entrapped by the kleisin loop, DNA double helices might be able to slide through the Ycg1–Brn1 groove, at least until they encounter a physical obstacle—for example, a nucleosome.

31A corollary is that the safety belt needs to transiently open for DNA to enter the groove. Opening presumably occurs by dissociation of the kleisin latch, which would then expose the conserved positively charged BC1 and BC2 patches in the Brn1 linker region to assist in the capture of DNA double helices (Figure 7A). After DNA entry into the binding groove, association of Brn1 latch and buckle regions fastens the safety belt and results in stable DNA entrapment (Figure 7B). Notably, the α1 helix of Hs NCAPH contains a target motif for cyclin-dependent kinases (CDKs) and is phosphorylated during mitosis. Mutation of the single serine residue within this motif, which is conserved among vertebrate species, to a non-phosphorylatable alanine residue almost completely abolishes condensin’s association with mitotic chromosomes (Figure 6E). The phosphorylated serine residue might interact with the conserved neighboring arginine and histidine residues of the buckle region (Figure 6A), which, in combination with the conformational changes of latch residues that we observe upon DNA binding (Figure 4D), presumably helps to lock the safety belt shut. Since we could not detect any sequence similarity between the α1 latch and loop regions of condensin and cohesin kleisin subunits, in contrast to sequence and structural similarities between the α2 to α4 segments (Figure S4), it is conceivable that the safety-belt mechanism might be specific to condensin.

32The discovery of a DNA-binding activity in the Ycg1–Brn1 subcomplex does not rule out additional roles of these subunits. For example, the HEAT-repeat domains might promote the interaction between individual condensin complexes in a way as has been proposed for the interaction between cohesin complexes or might drive the self-assembly of condensin complexes into a hydrogel-like network. DNA binding by Ycg1–Brn1 also does not exclude the existence of additional DNA binding sites in the condensin holocomplex. It is, for example, conceivable that Ycs4, even though it does not contribute significantly to the DNA-binding activity of condensin complexes in vitro, could do so in vivo, possibly by acting as a sensor or modulator of the DNA-bound state. Other regions, including the Smc2–Smc4 hinge or ATPase head domains, might form additional DNA interaction sites in the condensin holocomplex. These sites might only become accessible for DNA binding once the DNA double helix has been enclosed by the Brn1 kleisin loop and the Smc2–Smc4 ATPase has been activated.

33We note that the presence of direct protein-DNA binding sites in condensin is not mutually exclusive with the topological entrapment of DNA within the Smc2–Smc4–Brn1 ring structure. Instead, it could function to anchor condensin complexes to chromosomes for their subsequent entrapment. In this scenario, activation of the Smc2–Smc4 ATPase activity, once DNA has been bound in the groove and the safety belt has been closed, results in a conformational change that opens the Smc2–Smc4–Brn1 ring for DNA entry (Figure 7C). Future studies will be needed to reveal whether DNA remains bound to the Ycg1–Brn1 groove after transport of the chromatin fiber into the ring. If so, holding on to the original binding site while condensin actively translocates away from this site (Figure 7D; ) would provide an elegant solution for how condensin complexes can generate the large-sized chromatin loops that are thought to constitute the basic organizational principle of higher-order chromosome architecture.

34Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Christian H. Haering (christian.haering@embl.de).

35Female HeLa Kyoto H2B-mCherry cells were cultivated in DMEM (Life Technologies) containing 10% FBS (Life Technologies), 1% PenStrep (Invitrogen), and 1% glutamine (Invitrogen) at 37°C, 5% CO2.

36Saccharomyces cerevisiae strains are derived of W303. Genotypes of strains C4237, C4239, C4257, C4259, C4261, C4491, C4493, C4516, C4724, C4895 and C5037are listed in the Key Resources Table.

37Proteins for crystallography and biochemistry were expressed in Escherichia coli Rosetta (DE3) pLysS cells (Merck, Cat#70954) pre-cultured at 37°C and then shifted to 18°C for induction in 2 × TY or Terrific Broth (TB) medium.

38Expression of Ct Ycg1, Ct Ycs4, Ct Ycg1–Brn1, or Ct Ycs4–Brn1 and the crystallization constructs Sc Ycg1–Brn1 or Sp Cnd3–Cnd2 was induced for 18 hr from pET-MCN vectors in Escherichia coli Rosetta (DE3) pLysS (Merck) grown at 18°C in Terrific Broth (TB) medium (for Sc Ycg1–Brn1) or 2 × TY medium (for all other constructs). Cells were lysed by sonication at 4°C in lysis buffer (50 mM TRIS–HCl pH 7.5, 500 mM NaCl, 20 mM imidazole, 5 mM β-mercaptoethanol containing cOmplete protease inhibitor cocktail tablets without EDTA (cOm–EDTA, Roche)). The lysate was cleared by centrifugation at 45,000 × gmax and loaded onto Ni-Sepharose 6FF (GE Healthcare). After washing with 30-40 column volumes (CV) lysis buffer, proteins were eluted in 5-7 CV elution buffer (lysis buffer plus 300 mM imidazole). The eluate was dialyzed overnight in dialysis buffer (25 mM TRIS–HCl pH 7.5, 300 mM NaCl, 1 mM DTT) at 4°C, diluted with low-salt buffer (25 mM TRIS–HCl pH 7.5, 100 mM NaCl, 1 mM DTT) to a final salt concentration of 150 mM NaCl and loaded onto a 6 mL RESOURCE Q (GE Healthcare) anion exchange column pre-equilibrated with low-salt buffer. After washing with 3-5 CV low-salt buffer, proteins were eluted by increasing NaCl concentrations to 1 M in a linear gradient of 60 mL. Peak fractions were pooled and loaded onto a Superdex 200 26/60 column (GE Healthcare) equilibrated in SEC-buffer (25 mM TRIS–HCl pH 7.5, 500 mM NaCl, 1 mM DTT). Peak fractions were pooled and concentrated by ultrafiltration (Vivaspin 30,000 MWCO, Sartorius).

39Ct Brn1336-714 was expressed as a N-terminal GST fusion construct from a pGEX 6P-1 as described above. Cells were lysed at 4°C by sonication in lysis buffer (50 mM TRIS–HCl pH 7.5, 500 mM NaCl, 2 mM DTT containing cOm–EDTA). The lysate was cleared by centrifugation at 45,000 × g max and loaded onto Glutathione Sepharose 4B beads (GE Healthcare). The GST-fusion protein was eluted from the beads with lysis buffer containing 10 mM L-glutathione. The eluate was dialyzed and purified over RESOURCE Q as described above. Peak fractions were pooled, concentrated by ultrafiltration (Vivaspin 10,000 MWCO) and loaded on a Superdex 200 10/30 column (GE Healthcare) equilibrated in SEC buffer (25 mM TRIS–HCl pH 7.5, 500 mM NaCl, 1 mM DTT). Peak fractions were pooled and concentrated by ultrafiltration (Vivaspin 10,000 MWCO).

40Trimeric Ct Ycg1–Ycs4–Brn1 complexes were prepared by mixing purified Ct Ycs4 at 1.1-fold molar excess with purified Ct Ycg1–Brn1225-634 subcomplexes and 30 min incubation on ice. An equimolar complex was separated via 6 mL RESOURCE Q as described above. Peak fractions were pooled and concentrated by ultrafiltration (Vivaspin 30,000 MWCO).

41Condensin holocomplexes (Sc Smc2, Smc4-StrepII3, Brn1-His12-HA3, Ycg1, Ycs4; Sc Smc2, Smc4-StrepII3, Brn1-His12-HA3, Ycs4; Sc Smc2, Smc4-StrepII3, Brn1BC1/2-His12-HA3, Ycg1, Ycs4; Sc Smc2, Smc4-StrepII3, Brn1ΦD-His12-HA3, Ycg1, Ycs4 and Sc Smc2Q147L, Smc4Q302L-StrepII3, Brn1-His12-HA3, Ycg1, Ycs4) were expressed in Sc (strains C4491, C4493, C4516, C5037 and C4724) and purified as described previously. Expression was induced with 2% (w/v) galactose in synthetic complete (SC) –TRP –URA medium for 14-18 hr at 30°C. Cells were harvested and lysed by cryogenic grinding (SPEX Sample Prep Freezer/Mill 6970). Lysates were thawed in lysis buffer (50 mM TRIS–HCl pH 7.5, 200 mM NaCl, 20 mM imidazole, 5 mM β-mercaptoethanol, 1 mM MgCl2, 5% (v/v) glycerol) containing cOm–EDTA and cleared by centrifugation at 45,000xgmax. The cleared lysate was loaded on a 5 mL HisTrap HP column (GE Healthcare), washed with lysis buffer containing 500 mM NaCl and lysis buffer containing 40 mM imidazole and eluted in elution buffer (50 mM TRIS–HCl pH 7.5, 200 mM NaCl, 200 mM imidazole, 5 mM β-mercaptoethanol, 1 mM MgCl2, 5% (v/v) glycerol). The eluate was supplemented with 1.5 mM EDTA, 0.01% (v/v) Tween20, 0.1 mM PMSF and incubated with 1 mL Strep-Tactin Superflow high capacity beads (IBA) at 4°C for 16 hr. Beads were washed in SB-wash buffer (50 mM TRIS–HCl pH 7.5, 500 mM NaCl, 1 mM DTT, 1 mM MgCl2, 5% (v/v) glycerol, 0.01% (v/v) Tween20). For complexes used for ATPase assays, beads were washed additionally with SB-ATP-wash buffer (50 mM TRIS–HCl pH 7.5, 200 mM NaCl, 50 mM KCl, 1 mM ATP, 1 mM DTT, 10 mM MgCl2, 5% (v/v) glycerol, 0.01% (v/v) Tween20). The protein was eluted in SB-elution buffer (50 mM TRIS–HCl pH 7.5, 200 mM NaCl, 1 mM DTT, 1 mM MgCl2, 5% (v/v) glycerol, 10 mM d-Desthiobiotin). Eluate fractions were concentrated by ultrafiltration (Vivaspin 30,000 MWCO) and loaded on a Superose 6 10/30 GL column (GE Healthcare) equilibrated in Sup6 buffer (50 mM TRIS–HCl pH 7.5, 200 mM NaCl, 1 mM DTT, 1 mM MgCl2, 5% (v/v) glycerol). Peak fractions were pooled and concentrated by ultrafiltration (Vivaspin 30,000 MWCO).

42Selenomethionine-labeled Sp Cnd2-Cnd3 was expressed applying methionine pathway inhibition and purified as described above.

43Crystals of selenomethionine-labeled and native Sp Cnd3–Cnd2 (Table S1) were grown at 7°C by hanging-drop vapor diffusion. Volumes of 1 μL protein (7-12 mg/mL in 10 mM TRIS–HCl 7.5, 100 mM NaCl, 1 mM DTT) was mixed with 1 μL crystallization solution 1 (18% (w/v) Sokalan CP 42 (Molecular Dimensions), 0.1 M BIS–TRIS pH 5.8, 0.1 M Li acetate) for labeled or with 1 μL crystallization solution 2 (3%–4% (w/v) PEG 4,000, 0.1 M Na citrate pH 5.2-5.4, 0.2 M Na acetate) for native protein. For the selenomethionine-labeled crystals, drops were pre-equilibrated for one day at 7°C before streak seeding with previously grown and crushed crystal clusters. Crystals were harvested after 7-9 days and cryo-protected by addition of crystallization solution containing 25% (v/v) glycerol and 25% (w/v) Sokalan CP 42 (labeled) or 30% (v/v) PEG 400 (native) before flash freezing in liquid nitrogen. Single-wavelength anomalous dispersion data were collected at a wavelength of 0.979 Å (peak) and native data at 0.977 Å at beamline ID29, European Synchrotron Radiation Facility (ESRF, Grenoble, France). Data were processed with XDS and POINTLESS and scaled and merged with AIMLESS of the CCP4 suite.

44Sc Ycg1–Brn1 crystals (Table S1) were grown by hanging-drop vapor diffusion after mixing 1 μL protein (8 mg/mL in 10 mM TRIS–HCl pH 7.5, 300 mM NaCl, 1 mM DTT) and 1 μL crystallization solution (3% (w/v) PEG 4,000, 0.1 M Na citrate pH 5.5, 0.2 M Na acetate) at 20°C before streak seeding with previously grown and crushed crystal clusters. Crystals were harvested 15 days after setup and cryo-protected by addition of 35% (v/v) PEG 400 before flash freezing in liquid nitrogen. The dataset was collected at a wavelength of 1.000 Å at beamline ID29, ESRF. Data were processed as described above with the exception of using SCALA instead of AIMLESS.

45Sc Ycg1–Brn1–DNA (I, II and short kleisin loop, Table S1) crystals were grown by sitting drop vapor diffusion after mixing 100 nL sample and 100 nL crystallization solution in an MRC 2-well plate (Hampton Research). dsDNA was prepared by annealing a palindromic 18 bp HPLC-purified DNA oligo (IDT, GATGTGTAGCTACACATC, modified from ) in annealing buffer (10 mM TRIS–HCl pH 7.5, 130 mM NaCl, 1 mM DTT) in a temperature gradient of 0.1°C/s from 95°C to 4°C. The crystallization sample contained 1.2-fold molar excess of DNA over protein. For Sc Ycg1–Brn1-DNA (I), protein (5 mg/mL in 10 mM TRIS–HCl pH 7.5, 250 mM NaCl, 1 mM DTT) and DNA were mixed with crystallization buffer (4% (w/v) PEG 4,000, 0.05 M Na citrate pH 6.6, 5 mM MgSO4) and crystals were harvested after 15 days at 7°C. For Sc Ycg1–Brn1–DNA (II), protein (5 mg/mL in 10 mM TRIS–HCl pH 7.5, 250 mM NaCl, 1 mM DTT) and DNA were mixed with crystallization buffer (5% (w/v) PEG 4,000, 0.05 M Na cacodylate pH 6.8, 10 mM BaCl2) and crystals were harvested after 9 days at 7°C. For Sc Ycg1–Brn1short kleisin loop–DNA, protein (6 mg/mL in 10 mM TRIS–HCl pH 7.5, 250 mM NaCl, 1 mM DTT) and DNA were mixed with crystallization buffer (4% (w/v) PEG 4,000, 0.05 M Na citrate pH 5.6, 5 mM MgSO4) and crystals were harvested after 12 days at 7°C. Crystals were flash frozen in liquid nitrogen after addition of 2 μL crystallization buffer containing 37.5% (v/v) PEG 200. Datasets of Sc Ycg1–Brn1–DNA (I and II) were collected at a wavelength of 0.873 Å at beamline ID23-2, ESRF. The dataset of Sc Ycg1–Brn1short kleisin loop–DNA was collected at a wavelength of 0.979 Å at beamline ID29, ESRF. All datasets were processed as described above using SCALA.

46Single anomalous dispersion data for Sp Cnd3–Cnd2, merged from 4 independent datasets, were used to locate 26 selenium sites with SHELX followed by site refinement, phasing, and density modification. An initial model was built using Phenix AutoBuild and manual adjustment in Coot. The higher resolution, native Cnd3–Cnd2 dataset was solved by molecular replacement searching with the initial Cnd3–Cnd2 SeMet model using Phenix Phaser-MR. The model was built by iterative rounds of manual adjustments with Coot and of restrained refinements with phenix.refine.

47The Sc Ycg1–Brn1 structure was solved by molecular replacement with an adapted Sp Cnd3–Cnd2 as search model using Phenix Phaser-MR. An initial model was built using Phenix AutoBuild and manual adjustments with Coot. The structure was further improved in iterative rounds of manual correction with Coot and restrained refinements with phenix.refine.

48The Sc Ycg1–Brn1–DNA (I) structure was solved by molecular replacement searching with the Sc Ycg1–Brn1 model using Phenix Phaser-MR. The model was built in iterative rounds using density modified maps of Phenix AutoBuild and manual adjustments in Coot. The structure was finalized in iterative rounds of manual adjustments with Coot and restrained refinements with phenix.refine. Sc Ycg1–Brn1–DNA (II) and Sc Ycg1–Brn1short kleisin loop–DNA models were solved by molecular replacement searching with Sc Ycg1–Brn1–DNA (I) model using Phenix Phaser-MR. The models were built and refined as described for Sc Ycg1–Brn1–DNA (I).

49All structures were refined with hydrogens (‘riding’ model) and validated using MolProbity (Table S1).

50Structures were visualized with PyMOL (Schrödinger, LLC). Surface conservation graphics were created using the ConSurf server using a specified multi-sequence alignment (see below). The electrostatic surface potential graph was created with APBS. Structural alignments with cohesin-associated HEAT-repeat subunits were performed with the SUPERPOSE program of CCP4.

51The 6-FAM labeled 35-bp dsDNA was prepared by annealing two complementary HPLC-purified DNA oligos (IDT, 5′-6-FAM-CCTATAGTGAGTCGTTCGATATTACAATTCACTGG-3′; 5′-CCAGTGAATTGTAATATCGAACGACTCACTATAGG-3′) in annealing buffer (10 mM HEPES–KOH pH 7.5, 125 mM NaCl, 5 mM MgCl2) at a concentration of 20 μM in a temperature gradient of 0.1°C/s from 95°C to 4°C. The EMSA reaction with 35-bp dsDNA or 2.1-kb circular, supercoiled plasmid was prepared with a constant DNA concentration of 0.2 μM or 10 nM, respectively, and various concentrations of purified protein in binding buffer (10 mM HEPES–KOH pH 7.5, 125 mM NaCl, 5 mM MgCl2, 1 mM DTT). After 30 min incubation and addition of 3.5% (v/v) glycerol, free DNA and DNA–protein complexes were resolved by electrophoresis for 6-14 hr at 6-2 V/cm, respectively, on 1.8% (w/v) for short dsDNA or 0.8% (w/v) for 2.1-kb circular plasmid DNA TAE-agarose gels at 4°C. Gels with 6-FAM labeled short dsDNA were detected directly, gels with 2.1-kb plasmid DNA were stained with ethidium bromide prior to detection on a Typhoon FLA 9,500 scanner (GE Healthcare) with excitation at 473 nm with LPB (510LP) or at 532 nm with LPG (575LP) filter setting, respectively.

52Reactions with 2.1-kb supercoiled circular or linear DNA substrates were prepared at 14.4 nM (20 ng/μL) concentrations and 10 μM wild-type or cysteine-containing Ct Ycg1–Brn1515-634 or 2.5 μM wild-type or cysteine-containing Ct Ycs4–Ycg1–Brn1225-634. DNA and protein samples were re-buffered into crosslinking buffer (1 × PBS, 5 mM MgCl2) prior to mixing and incubated for 30 min on ice. 10% (v/v) DMSO, 30 μM dibromobimane (bBBr in DMSO, Sigma-Aldrich), 30 μM disthiobismaleimidoethane (DTME in DMSO, ThermoFisher Scientific) or 30 μM bismaleimidoethane (BMOE in DMSO, ThermoFisher Scientific) were added to the reaction followed by incubation for 10 min on ice. The crosslinking reaction was stopped by addition of 5 mM DTT for 5 min on ice for bBBr and BMOE. DTT was either not added or added to 55.5 mM for the DTME crosslinking reaction. Brn1 cleavage at the HRV-3C site was induced by addition of 1 μg HRV-3C protease and incubation for 25 min at 7°C. Proteins were denatured by addition of SDS to 1% (w/v) and incubation for 10 min at 65°C. Samples were resolved by electrophoresis for 14 hr at 2 V/cm on 0.8% (w/v) TAE agarose gels at 4°C, followed by detection of DNA by ethidium bromide staining.

53Crosslinking experiments of cysteine containing Ct Ycg1-Brn1515-634 with and without HRV-3C site followed by SDS-PAGE and Coomassie staining were performed as described above for the upshift experiments of circular DNA with the exception of incubating 10 μM protein with 20 μM annealed, palindromic 22-bp dsDNA (IDT, 5′-GATTCGTGTAGCTACACGAATC-3′, ) in all reactions. Crosslinking experiments of cysteine containing Ct Ycs4–Ycg1–Brn1225-634 with BMOE prior EMSA and incubation with 2.1-kb supercoiled-circular DNA was done as described above without 22-bp dsDNA incubation before crosslinking.

54Reactions (10 μL) were set up with 0.5 μM condensin holocomplex, with or without 24 nM relaxed circular 6.4-kb plasmid DNA in ATPase buffer (40 mM TRIS–HCl pH 7.5, 125 mM NaCl, 10% (v/v) glycerol, 5 mM MgCl2, 5 mM ATP, 1 mM DTT and 33 nM [α32P]-ATP; Hartmann Analytic). Relaxed circular DNA was prepared by treating the negatively supercoiled plasmid DNA with E. coli topoisomerase I (NEB). After treatment with topo I, relaxed DNA was purified by phenol-chloroform extraction and ethanol precipitation. ATP hydrolysis reactions were incubated at RT (∼25°C) and were initiated by addition of ATP. A volume of 0.8 μL of the reaction mix was spotted onto PEI cellulose F TLC plates (Merck) every 3 min for a total duration of 15 min. The reaction products were resolved on TLC plates using 0.5 M LiCl and 1 M formic acid solution and detected by exposing the TLC plates to a phosphorimager screen and analysis on a Typhoon FLA 9,500 scanner (GE Healthcare). ATP hydrolysis rates were calculated from the ADP/ATP ratios from time points in the linear range of the reaction.

55The 25-bp dsDNA substrate was prepared by annealing two complementary HPLC-purified DNA oligonucleotides (IDT, 5′-CCTATAGTGAGTCACAATTCACTGG-3′; 5′- CCAGTGAATTGTGACTCACTATAGG-3′) in annealing buffer (10 mM HEPES–KOH pH 7.5, 125 mM NaCl, 5 mM MgCl2) at a concentration of 20 μM using a temperature gradient of 0.1°C/s from 95°C to 4°C. The DNA was then re-buffered into ITC buffer (25 mM TRIS–HCl pH 7.5, 200 mM NaCl, 2 mM MgCl2, 1 mM BME) by multiple rounds of concentration and dilution using ultrafiltration (Vivaspin 3,000 MWCO). Ct Ycg1–Brn1 was dialyzed to ITC buffer. Ct Brn1-Ycg1 was injected with a concentration of 250 μM into 10 μM 25-bp dsDNA at 25°C.

56ITC measurements were performed on a MicroCal iTC200 microcalorimeter (GE Healthcare). The ITC data were corrected for the dilution heat and fitted with the MicroCal Origin software package applying one set of binding sites model. Standard deviation values of the fit were calculated from the original data points.

57Immunoprecipitation of endogenous condensin complexes from yeast was performed as described previously. Yeast strains C4237, C4239 and C4261 were grown at 30°C in 2 l YPAD to OD600 = 1, harvested by centrifugation and lysed by cryogenic grinding (SPEX Sample Prep Freezer/Mill 6970) in lysis buffer (50 mM TRIS–HCl pH 8.0, 100 mM NaCl, 2.5 mM MgCl2, 0.25% (v/v) Triton X-100, 1 mM DTT, 1 mM PMSF) containing 2 × cOm–EDTA. The lysate was cleared by centrifugation at 20,400 × gmax and incubated with 100 μL protein A coupled Dynabeads (ThermoFisher Scientific) that were previously bound to anti-PK (V5) tag (Abd Serotec, MCA1360) antibody for 2 hr at 4°C. Beads were washed with IP buffer (50 mM TRIS–HCl pH 8.0, 100 mM NaCl, 1 mM DTT, 5 mM EDTA, 0.25% (v/v) Triton X-100) and eluted in 20 μL 2 × SDS loading buffer (100 mM TRIS–HCl pH 6.8, 4% (w/v) SDS, 20% (v/v) glycerol, 0.2% (w/v) bromophenol blue, 0.2 M DTT) by boiling at 95°C for 5 min prior to SDS-PAGE and Coomassie staining or western blotting with antibodies against the PK (V5) tag (Abd Serotec, MCA1360), Sc Ycg1 or α-tubulin (TAT1).

58ChIP-qPCR experiments were performed as described previously. Yeast strains C4237, C4239 and C4261 were grown in 42 mL YPAD at 30°C to OD600 = 0.6 and fixed with 4.7 mL fixation buffer (9.5 mM TRIS–HCl pH 8.0, 19 mM NaCl, 0.095 mM EGTA, 3% (v/v) formaldehyde, 0.19 mM EDTA) for 30 min at 16°C. Fixation was stopped by addition of glycine to 125 mM (final concentration), followed by washing steps in PBS and PIPES buffer (100 mM PIPES–KOH pH 8.3). Cells were lysed by spheroplasting with 0.5 mg/mL zymolase T-100 (AMS Biotechnology) in HEMS buffer (100 mM HEPES–KOH pH 7.5, 1 mM EGTA, 1 mM MgSO4, 1.2 M Sorbitol, 1 mM PMSF containing cOm–EDTA), followed by resuspension of cells in 1.5 mL lysis buffer (50 mM HEPES–KOH pH 7.5, 140 mM NaCl, 1 mM EDTA, 1% (v/v) Triton X-100, 0.1% (w/v) sodium deoxycholate, 1 mM PMSF containing cOm–EDTA). Chromatin was sheared by sonication to a length of ∼500 bp using a Bioruptor UCD-200 (Diagenode) for 9 min, 30 s on, 60 s off settings (‘high level’).

59Lysate was cleared by centrifugation at 16,800 × gmax and pre-cleared with 50 μL protein A dynabeads (ThermoFisher Scientific) for 90 min at 4°C. 10% of the cleared lysate was used to check sonication, 12% was kept on ice as input sample. 2 μg anti-PK (V5) tag antibody (Abd Serotec MCA1360) was added to the remaining lysate and samples were incubated at 4°C for 16 hr before addition of 100 μL protein A dynabeads (ThermoFisher Scientific) for another 4 hr at 4°C. Beads were washed with lysis buffer, wash buffer (10 mM TRIS–HCl pH 8.0, 0.25 M LiCl, 0.5% (w/v) sodium deoxycholate, 1 mM EDTA, 1 mM PMSF containing cOm–EDTA) and TE buffer (10 mM TRIS–HCl pH 8.0, 1 mM EDTA containing cOm–EDTA). Samples were eluted in 320 μL TES buffer (50 mM TRIS–HCl pH 8.0, 10 mM EDTA, 1% (w/v) SDS) at 65°C for 8 hr. After addition of 30 μg RNaseA (Roche) for 90 min at 37°C and 200 μg Proteinase K (Roche) for 90 min at 65°C, DNA was purified via a spin column (QIAGEN) and eluted in 50 μL EB buffer.

60qPCR reactions were set up for 5 μL of 1:5 and 1:25 dilutions for immunoprecipitated samples and 1:5, 1:50, 1:500 and 1:5,000 dilutions for input samples with SYBR green PCR Master mix (Applied Biosystems) and 5 μM qPCR primers (5′-TGGTGTGGAAGTCCTAATATCG-3′ and 5′-TGCATGATCAAAAGGCTCAA-3′ for CEN4 and 5′- TTTCTGCCTTTTTCGGTGAC-3′ and 5′- TGGCATGGATTTCCCTTTAG-3′ for rDNA) on an Applied Biosystems 7,500 Fast Real-Time PCR System. Data were calculated from two independent experiments with two qPCR runs each.

61Female HeLa Kyoto H2B-mCherry cells were transiently transfected with pC1 Flag-EGFP-NCAPH or Flag-EGFP-NCAPH2 as described previously, with the following modifications: Transfections were performed in Lab-Tek dishes (ThermoFisher Scientific) by addition of transfection mix (400 ng plasmid DNA and 1.5 μL FuGene (Promega) in 50 μL Opti-MEM medium (Life Technologies)) at 50% confluency (1 × 105 cells). After 12-16 hr, the medium was replaced with high-glucose DMEM (Life Technologies) containing 10% FBS (Life Technologies), 1% PenStrep (Invitrogen), and 1% glutamine (Invitrogen). 5-12 hr before imaging, fresh medium supplemented with 250 ng/mL nocodazole was added. Cells were washed once and imaged 36 hr after transfection in medium without phenol red containing 250 ng/mL nocodazole. Image acquisition was performed on a Zeiss LSM 780 microscope in 16-bit mode and four lines averaging with a Plan-Apochromat 63 × /1.40 oil DIC M27 objective at 37°C and 5% CO2. Excitation and emission wavelengths were 488 nm and 520-560 nm or 561 nm and 580-650 nm for EGFP or mCherry, respectively. The cell line was authenticated and tested for mycoplasma contamination.

62Images were analyzed with Fiji. First, background was subtracted using the rolling ball algorithm. Chromatin regions were segmented based on the mCherry fluorescence signal and the whole cell was segmented based on the bright field image. Cytoplasmic regions were selected after subtracting the areas of chromatin from the whole cell regions. Mean fluorescence intensities of EGFP images were measured for chromatin and cytoplasmic regions. Data were calculated from two independent experiments.

63Brn1 and Ycg1 sequences were aligned from 35 divergent species (10 plants, 10 fungi, 10 animals, 5 protists; Table S2) with MAFFT using the Smith-Waterman local algorithm (L-INS-i) with default settings and gaps were manually optimized. Conserved residues were highlighted using the ClustalW color code.

64Statistical details of experiments can be found in the figure legends or Method Details section.

65Crystal structures have been deposited in the Protein Data Bank (https://www.rcsb.org/pdb) under ID codes PBD: 5OQR (Sp Cnd3–Cnd2), PBD: 5OQQ (Sc Ycg1–Brn1), PBD: 5OQP (Sc Ycg1–Brn1-DNA (I)), PBD: 5OQO (Sc Ycg1–Brn1–DNA (II)), and PBD: 5OQN (Sc Ycg1–Brn1short kleisin loop–DNA).

66M.K. purified proteins and conducted biochemical assays. M.K. and M.H. crystallized the proteins, collected data, and solved the structures. F.M. carried out imaging experiments with human cells. M.K. and J.M. conducted functional assays in yeast. J.M. performed ChIP-qPCR experiments. S.B. provided purified condensin holocomplexes and performed ATPase assays. V.R. performed ITC experiments. M.K., M.H., and C.H.H. designed the study. C.H.H. supervised the work. M.K, M.H., and C.H.H. wrote the manuscript with input from all authors.