Research Article

Characterization of a DNA exit gate in the human cohesin ring

See allHide authors and affiliations

Science  21 Nov 2014:
Vol. 346, Issue 6212, pp. 968-972
DOI: 10.1126/science.1256904


Chromosome segregation depends on sister chromatid cohesion mediated by cohesin. The cohesin subunits Smc1, Smc3, and Scc1 form tripartite rings that are thought to open at distinct sites to allow entry and exit of DNA. However, direct evidence for the existence of open forms of cohesin is lacking. We found that cohesin’s proposed DNA exit gate is formed by interactions between Scc1 and the coiled-coil region of Smc3. Mutation of this interface abolished cohesin’s ability to stably associate with chromatin and to mediate cohesion. Electron microscopy revealed that weakening of the Smc3-Scc1 interface resulted in opening of cohesin rings, as did proteolytic cleavage of Scc1. These open forms may resemble intermediate states of cohesin normally generated by the release factor Wapl and the protease separase, respectively.

A cohesin ring around two DNA strands

Holding together homologous sister chromosome pairs is a vital requirement during cell division and DNA repair. A special complex, called cohesin, forms a ring made of three different proteins and functions to hold together the two sister DNA strands. Gligoris et al. and Huis in 't Veld et al. identified a specific protein-protein interface within the cohesin ring that forms a DNA exit gate. Mutations in this interface prevented cohesion between sister chromatids. Thus, the cohesin ring must indeed encircle the two DNA strands to hold them together.

Science, this issue p. 963, p. 968

To ensure proper chromosome segregation during cell division, sister chromatids must be held together from the onset of DNA replication until the metaphase-to-anaphase transition. This sister chromatid cohesion is mediated by cohesin, a protein complex also implicated in chromatin organization, gene regulation, and DNA repair. Cohesin malfunction can result in genetic disorders such as Cornelia de Lange syndrome, and it has been found to correlate with age-related female infertility and tumorigenesis (1). The core of human cohesin is composed of the subunits Smc1, Smc3, Scc1 (also known as Rad21 or Mcd1), and SA1 or SA2. Because Smc1 and Smc3 contain long coiled coils, dimerize through their hinge domain at one end, and are bridged by the kleisin Scc1 at their nucleotide-binding domains (NBDs), the cohesin core complex has a ring-like appearance (24). According to the embrace model, cohesin uses this characteristic subunit topology to physically entrap DNA (5). The loading of cohesin onto DNA depends on the Scc2-Scc4 complex and adenosine triphosphate hydrolysis (68). It has been proposed that this reaction involves a transient opening of the Smc1-Smc3 hinge, because artificial stabilization of this interface abolishes cohesin’s loading onto DNA (9, 10). Cohesin complexes that mediate sister chromatid cohesion must dissociate from DNA to allow chromosome segregation at the metaphase-to-anaphase transition. This removal involves cleavage of cohesin’s Scc1 subunit by separase (11, 12). Cohesin is also released from DNA in a proteolysis-independent manner that occurs throughout the cell cycle and depends on the cohesin-associated protein Wapl (1315). An intramolecular fusion of Scc1 and Smc3 renders cohesin refractory to this activity, which suggests that Wapl acts on the Scc1-Smc3 interface (10, 16, 17). Cohesin rings are thus suggested to be opened at distinct interfaces to allow entry and exit of DNA.

Reconstitution and visualization of the human cohesin complex

To obtain insight into how cohesin rings might open to entrap and release DNA, we reconstituted human cohesin complexes and used low-angle Pt/C rotary shadowing followed by transmission electron microscopy (henceforth rotary shadowing EM) to visualize them. Pure and stoichiometric Smc1-Smc3 heterodimers (283 kD) and cohesin complexes containing Smc1, Smc3, Scc1, and SA1 (500 kD) were obtained from baculovirus-infected insect cells (Fig. 1A). Smc1-Smc3 dimers appeared as a “pair of cherries” with a combined coiled-coil length of 98 ± 10 nm (n = 578). The additional presence of Scc1 and SA1 resulted in closed ring-like structures. SA1 was observed at varying positions, possibly reflecting the flexibility of its unstructured binding partner Scc1 (Fig. 1B). Whereas cohesin purified from HeLa cells was reported to contain a single kink in one of the coiled coils (3), we observed complexes with none, one, or two irregularities in the coiled coils.

Fig. 1 Cleavage of Scc1 opens cohesin rings and releases cohesin from chromatin before and after DNA replication.

(A) Silver staining of purified recombinant human cohesin dimers and tetramers after SDS–polyacrylamide gel electrophoresis (PAGE). Scc1PreScission is efficiently cleaved by PreScission protease. (B) Representative micrographs of cohesin dimers, tetramers, and cleaved tetramers after low-angle Pt/C rotary shadowing. Scc1 bridges the heads of Smc1 and Smc3 to form ring-like structures that rearrange upon Scc1 cleavage. Scale bar, 50 nm. (C) Distance between the heads of Smc1 and Smc3 for dimers (n = 274), tetramers (n = 304), and cleaved tetramers (n = 168). (D) Cohesin tetramers with Scc1GFP-TEV adopt an open conformation after cleavage by TEV protease. Scale bar, 50 nm. (E) Xenopus sperm chromatin, purified human cohesin, and interphase Xenopus egg extract were incubated for 75 min. TEV protease was then added for 15 min to cleave Scc1GFP-TEV. Chromatin-bound material was analyzed by immunoblotting. Human and Xenopus Scc1 can be distinguished by the GFP-induced mobility shift. Mcm5 accumulates on chromatin when DNA replication is inhibited by p27Kip1, whereas sororin binds to cohesin on replicated DNA.

Cleavage of Scc1 opens cohesin rings and releases cohesin from chromatin

It is well established that cleavage of Scc1 by separase at the metaphase-to-anaphase transition, as well as artificially induced cleavage of Scc1 by other proteases, leads to cohesin’s release from DNA (4, 11). However, the structural consequences of Scc1 cleavage have not been analyzed. We therefore generated recombinant cohesin with a triple tobacco etch virus (TEV) or human rhinovirus (PreScission) protease recognition site near the C-terminal end of Scc1. Rotary shadowing EM showed that cleaved cohesin tetramers resembled Smc1-Smc3 dimers with respect to flexibility (Fig. 1, B to D, and fig. S1). This confirms, as had been assumed, that Scc1 provides the physical connection between the NBDs of Smc1 and Smc3 and that Scc1 cleavage opens the cohesin ring.

We next used Xenopus egg extract to test whether reconstituted cohesin could be loaded onto chromatin. Binding of endogenous Xenopus cohesin is abolished by the addition of geminin, an inhibitor of prereplication complex assembly, but not by the addition of the Cdk inhibitor p27Kip1, which prevents DNA replication (18, 19). Purified human cohesin was loaded in a similar manner and was released from chromatin upon TEV protease–induced proteolysis of Scc1. When DNA replication was inhibited, Scc1 cleavage still released cohesin from chromatin (Fig. 1E and figs. S2 and S3). Cohesin’s ring structure is thus essential for the association with DNA before and after replication.

A cohesin proximity and interaction map

We used chemical cross-linking and mass spectrometry to generate a lysine proximity map of cohesin and probe its topology. In a series of experiments using cohesin core complexes, a complex of Wapl and its binding partner Pds5B and cohesin bound to Pds5B-Wapl, 51 unique intermolecular and 185 unique intramolecular cross-links were obtained (Fig. 2, fig. S4, and tables S1 and S2). We generated a distance map using high-resolution information from a murine Smc1-Smc3 hinge domain crystal structure (20). Of the 21 lysine-lysine pairs that were detected within this region, 19 bridged amino acids fewer than 24 Å apart. This distance reflects spatial constraints imposed by the cross-linker and thus validates our proximity map (Fig. 2B and fig. S4).

Fig. 2 A cohesin proximity map generated by cross-linking followed by mass spectrometry.

(A) Intermolecular cross-links (n = 51) are shown in a Circos diagram. Gray boxes illustrate the NBDs and hinge domains of Smc1 and Smc3, the α-helical regions in Scc1, and the α-helical repeats of SA1, Wapl, and Pds5B. The three dashes indicate FGF motifs in the N-terminal part of Wapl. (B) The Cα-Cα distance of all lysines (gray, n = 561) in the Smc1-Smc3 hinge domain was compared to the experimentally observed lysine-lysine cross-links in this region (blue, n = 21). (C) Schematic representation of the cohesin proximity map.

The majority of inter- and intramolecular cross-links in Smc1 and Smc3 were found in their coiled coils, likely as a consequence of solvent accessibility. Most cross-links in Pds5B and Wapl were also found within their unstructured and solvent-accessible regions. Consistently, regions in Pds5B that did not contain cross-linked lysines were also resistant to proteolysis by trypsin (fig. S4). For Scc1, only six intermolecular and two intramolecular cross-links were obtained, despite its predicted lack of secondary structure. Residues in the middle of Scc1 (Lys323, Lys335, and Lys406) formed cross-links to SA1, Pds5B, and Wapl. This confirms Scc1’s central region as a docking point for these proteins (21, 22). Binding studies using recombinant proteins showed that the HEAT repeats of Pds5B and, as previously reported (21, 23), the unstructured N-terminal half of Wapl associate with cohesin through Scc1 and SA1, respectively (fig. S5).

Scc1 binds the coiled coil of Smc3 to close the cohesin ring

Two residues that reside near the end of Scc1’s N-terminal helical domain (NHD), Lys72 and Lys86, form cross-links to Lys185 and Lys188 in Smc3, respectively. These residues are located at the onset of Smc3’s coiled coil, which suggests that Scc1’s NHD binds Smc3 through a mechanism that differs from the interaction between Scc1’s C-terminal winged-helix domain (WHD) and Smc1’s NBD (24). An α helix in ScpA has been shown to interact with the onset of the Smc coiled coil in the bacterial Bacillus subtilis condensin complex (25) (Fig. 3A). Our cross-linking proximity map suggested analogous contacts in human cohesin—a notion further supported by the conserved amphiphilic nature of the NHD in Scc1 and condensin’s kleisin CapH (Fig. 3B) and by the observation that point mutations in the NHD of yeast Scc1 reduced binding to Smc3 and compromised viability (26). To test the importance of this amphiphilicity, we generated cohesin complexes containing four point mutations in Scc1 (Fig. 3C). Rotary shadowing EM showed that most of these mutated complexes adopted an open conformation reminiscent of the cohesin complexes in which Scc1 was proteolytically cleaved (Fig. 3, D to F). Cohesin’s proposed exit gate is thus formed by hydrophobic interactions between Scc1’s NHD and the onset of Smc3’s coiled coil, and this interaction is essential to keep cohesin rings closed.

Fig. 3 Scc1 binds the coiled coil of Smc3 to close the cohesin ring.

(A) Cartoon of cohesin tetramers and structure of the prokaryotic Smc (red) and ScpA (green) interaction (PDB ID 3ZGX). (B) Sequence alignment of the second part of the NHD in ScpA (prokaryotic condensin), Scc1 (cohesin), and CapH (condensin) from human (hs), fly (dm), budding yeast (sc), fission yeast (sp), and B. subtilis (bs). Predicted α helices (green boxes) and the introduced PreScission recognition site are shown. (C) Top-down helical wheel representation of wild-type and mutant Scc1NHD. Hydrophobic (orange) and hydrophilic (blue) amino acids are indicated. (D and E) Tetramers with mutated Scc1 were visualized by rotary shadowing EM. Scale bars, 50 nm. (F) Distance between the heads of Smc1 and Smc3 in mutant complexes (n = 283). The dashed circle indicates the distribution of wild-type tetramers (see Fig. 1C). (G and H) Tetramers containing wild-type (wt) or mutant (mut) Scc1PreScission were bound to anti-FLAG beads through Smc3FLAG and exposed to PreScission protease. Supernatant (dissociated) and peptide eluates after extensive washing (bound) were analyzed by silver staining (top panel) and immunoblotting (lower panels) after SDS-PAGE. Upon cleavage, mutant Scc1N and SA1 dissociated from Smc1-Smc3.

Consistent with results obtained for yeast cohesin (4), Scc1 cleavage fragments (Scc1N and Scc1C) and SA1 remained stably bound to Smc1-Smc3 after Scc1 proteolysis (Fig. 3G and fig. S6). If mutations in Scc1’s NHD disturb its interaction with Smc3, Scc1N is predicted to dissociate from cohesin upon Scc1 cleavage. When tested, most Scc1N and SA1 did indeed dissociate, indicating that the Scc1 exit gate mutant bound the Smc1-Smc3 dimer through its C-terminal WHD before cleavage. Despite their similar appearance after rotary shadowing EM, this experiment thus confirmed that complexes with mutated Scc1 or cleaved Scc1 are topologically distinct (Fig. 3, G and H). Note that our in vitro results show that Scc1 cleavage fragments can remain associated with cohesin in the absence of Smc3 acetylation, unlike a previous proposal (27) (fig. S6).

The Scc1-Smc3 interaction is essential for a stable chromatin interaction

To test the in vivo effects of a disrupted Smc3-Scc1 interaction, we generated stably transfected HeLa cell lines in which wild-type and exit gate mutant Scc1 tagged with green fluorescent protein (Scc1GFP) could be expressed at near-endogenous levels. Wild-type and mutant Scc1GFP were incorporated into cohesin complexes and could be purified from nuclease-digested chromatin (fig. S7). Mutations in Scc1’s NHD did thus not disrupt cohesin’s ability to interact with chromatin, despite their profound effect on cohesin’s structure. Consistently, these mutations did reduce but not abolish cohesin’s binding to sperm chromatin in Xenopus egg extract (fig. S8). Because exit gate mutants, as well as complexes in which Scc1 was proteolytically cleaved, are active adenosine triphosphatases, the reduced recruitment to chromatin is not caused by a reduction of cohesin’s enzymatic activity (fig. S9).

To further investigate how the cohesin exit gate mutant binds to chromatin, we used the GFP tag on Scc1 to measure its fluorescence recovery after photobleaching (FRAP). For wild-type Scc1GFP, reduced GFP levels were still detectable several minutes after bleaching, indicating a chromatin residence time of 7 to 9 min. In contrast, the residence time of mutant Scc1GFP was in the range of 1 min (Fig. 4, A to F). This chromatin interaction was sensitive to depletion of the Scc2-Scc4 complex by RNA interference, indicating that the loading of the exit gate mutant still depends on the cohesin loading machinery (fig. S10). Cohesin complexes that contain mutant Scc1 can thus interact with chromatin, but not in a stable manner. This residual chromatin association could reflect cohesin recruitment to the loading complex, or it could indicate that these mutants can still close their exit gates transiently.

Fig. 4 The interaction between Smc3 and Scc1 is essential for a stable chromatin interaction.

(A) G1-phase cells expressing comparable amounts of wild-type (wt) or mutant (mut) Scc1GFP were analyzed by FRAP. Circle and arrows indicate the bleached area. Scale bar, 10 μm. (B) FACS profiles of G1- and G2-phase synchronized cells used for FRAP analysis. (C and D) Average FRAP for 12 cells per condition. (E and F) Fitted curves were used to calculate the percentage of cohesin that is chromatin-bound and the residence time of chromatin-bound cohesin. (G and H) Cells were treated with control or Wapl siRNA and synchronized in G1. Cells expressing comparable amounts of wild-type or mutant Scc1GFP were analyzed by FRAP. The average FRAP for 12 cells per condition is shown. (I and J) Fitted curves were used to calculate the percentage of cohesin that is chromatin-bound and the residence time of chromatin-bound cohesin. Error bars represent SEM.

Cohesin with a destabilized exit gate does not support sister chromatid cohesion

To test whether Wapl can still destabilize the transient association of the cohesin exit gate mutant with chromatin, we analyzed wild-type and mutant Scc1GFP by FRAP after depletion of Wapl by RNA interference. Whereas the residence time of wild-type complexes on chromatin doubled when Wapl was depleted, exit gate mutants were not stabilized (Fig. 4, G to J, and fig. S11). The effect of mutations at the Smc3-Scc1N interface thus resembles continuous destabilization by Wapl. Consistently, the binding dynamics of mutant Scc1GFP did not differ between G1 and G2 phase, whereas the residence time of wild-type Scc1GFP on chromatin increased after DNA replication, as previously described (28) (Fig. 4, A to F).

The marked reduction in chromatin binding and acetylation of Smc3 (fig. S7) raised the question of whether exit gate mutants could support sister chromatid cohesion. To test whether ectopically expressed Scc1 supported cohesion, we specifically depleted endogenous Scc1 by RNA interference (siRNA) (Fig. 5A and fig. S12). Under these conditions, expression of wild-type Scc1GFP rescued cohesion defects, whereas expression of mutated Scc1GFP did not (Fig. 5, B and C). Cohesin complexes that do not stably close their exit gates are thus unable to mediate sister chromatid cohesion.

Fig. 5 Cohesin with a destabilized exit gate does not support sister chromatid cohesion.

(A) Scc1 was depleted using siRNA directed against its 3′ untranslated region (UTR, endogenous) or coding DNA sequence (CDS, endogenous and exogenous). (B and C) Prometaphase chromosome spreads were analyzed after Giemsa staining. Expression of wild-type but not of mutant Scc1 restored sister chromatid cohesion in cells depleted of endogenous Scc1 (n = 150 per condition). Scale bar, 10 μm. (D) Schematic overview of cohesin ring opening and stabilization on chromatin during the cell cycle. Cohesin core components are colored as in Fig. 3H. The asterisks indicate acetylation of Smc3 Lys105 and Lys106.


It has long been postulated that cohesin forms rings that can be opened to mediate entry and exit of DNA. Here, we used electron microscopy to demonstrate the existence of such open forms, generated either by proteolytic cleavage of Scc1 (mimicking the effect of separase at the metaphase-to-anaphase transition) or by weakening the interaction between Scc1’s NHD and the coiled coil of Smc3 (mimicking the opening of cohesin’s DNA exit gate). Because the latter is thought to be achieved by Wapl, our exit gate mutant may resemble an otherwise transient intermediate in the ring opening and closure cycle (Fig. 5D). We identified residues at the outside of the solenoid-like Pds5B that reside in direct proximity to Wapl and the Smc3-Scc1 interaction interface (fig. S13), implying that Wapl and Pds5 control the exit gate through direct interactions. However, it remains to be addressed at a mechanistic level how Wapl promotes ring opening and how this is coordinated by Pds5, antagonized by sororin, and regulated by phosphorylation events.

Supplementary Materials

Materials and Methods

Figs. S1 to S13

Tables S1 and S2

References (2949)

References and Notes

  1. Acknowledgments: We thank O. Hudecz, K. Mechtler, G. Schmauss, K. Uzunova, M. Brandstetter, H. Kotisch, G. Resch, and A. Schleiffer for excellent technical support and I. Berger for an introduction to MultiBac. P.J.H. was supported by the Austrian Science Fund (W1221) and received funding from the European Community (FP7/2007-2013, no. 227764). Research in the laboratory of J.-M.P. is supported by Boehringer Ingelheim, the Austrian Science Fund (SFB-F34 and Wittgenstein award), the Austrian Research Promotion Agency (FFG Laura Bassi), the Vienna Science and Technology Fund (WWTF LS09-13), and the European Community (FP7/2007-2013, no. 241548, MitoSys).
View Abstract

Stay Connected to Science

Navigate This Article