Crystal structure of the cohesin loader Scc2 and insight into cohesinopathy

Significance

The ring-shaped cohesin traps chromosomes inside its ring and regulates chromosome segregation during mitosis and transcription during interphase. The sister chromatid cohesion 2 protein (Scc2) opens the cohesin ring and loads it onto chromosomes. Mutations of cohesin subunits and regulators perturb transcription and cause human developmental diseases called cohesinopathy. Scc2 is the most frequently mutated cohesin regulator in cohesinopathy. In this study, we report the crystal structure of a fungal Scc2 protein, which represents a high-resolution snapshot of the cohesin loader. We have identified a set of Scc2 mutations in cohesinopathy that disrupt the binding of Scc2 to the kleisin subunit of cohesin. Our results provide critical insight into cohesin loading and cohesinopathy.

Abstract

The ring-shaped cohesin complex topologically entraps chromosomes and regulates chromosome segregation, transcription, and DNA repair. The cohesin core consists of the structural maintenance of chromosomes 1 and 3 (Smc1–Smc3) heterodimeric ATPase, the kleisin subunit sister chromatid cohesion 1 (Scc1) that links the two ATPase heads, and the Scc1-bound adaptor protein Scc3. The sister chromatid cohesion 2 and 4 (Scc2–Scc4) complex loads cohesin onto chromosomes. Mutations of cohesin and its regulators, including Scc2, cause human developmental diseases termed cohesinopathy. Here, we report the crystal structure of Chaetomium thermophilum (Ct) Scc2 and examine its interaction with cohesin. Similar to Scc3 and another Scc1-interacting cohesin regulator, precocious dissociation of sisters 5 (Pds5), Scc2 consists mostly of helical repeats that fold into a hook-shaped structure. Scc2 binds to Scc1 through an N-terminal region of Scc1 that overlaps with its Pds5-binding region. Many cohesinopathy mutations target conserved residues in Scc2 and diminish Ct Scc2 binding to Ct Scc1. Pds5 binding to Scc1 weakens the Scc2–Scc1 interaction. Our study defines a functionally important interaction between the kleisin subunit of cohesin and the hook of Scc2. Through competing with Scc2 for Scc1 binding, Pds5 might contribute to the release of Scc2 from loaded cohesin, freeing Scc2 for additional rounds of loading.

Cohesin consists of four core subunits: structural maintenance of chromosomes 1 and 3 (Smc1 and Smc3), and sister chromatid cohesion 1 and 3 (Scc1 and Scc3). (1–5). Smc1 and Smc3 are related ATPases that heterodimerize through their hinge domains. The C-terminal winged helix domain and the N-terminal helical domain (NHD) of Scc1 bind to the Smc1 ATPase head and a coiled-coil segment adjacent to the Smc3 ATPase head, respectively, forming a tripartite ring (6, 7). Scc3 contains Huntingtin-elongation factor 3-protein phosphatase 2A-TOR1 (HEAT) repeats, binds to the middle region of Scc1, and provides a landing pad for several cohesin regulators (8, 9). Cohesin can topologically trap DNA inside its ring (10), thereby regulating many facets of chromosome biology. During interphase, cohesin mediates the formation of chromosome loops that impact transcription (11). During S phase, cohesin physically links replicated chromosomes to establish sister-chromatid cohesion (12, 13), which is a prerequisite for faithful chromosome segregation during mitosis. Finally, cohesin contributes to homology-directed repair of DNA breaks, in part, through holding the sister chromatid in proximity of the breaks and presenting it as the repair template (14).

For cohesin to accomplish these diverse tasks, its association with chromosomes has to be tightly controlled both spatially and temporally. Cohesin dynamics on chromosomes are regulated by a set of accessory proteins, including the Scc2–Scc4 loading complex and the releasing complex comprising Pds5 and winged apart-like protein (Wapl) (2, 4, 5). Scc2–Scc4 promotes the loading of cohesin onto chromosomes in telophase and G1 (15). Pds5–Wapl can release cohesin from chromosomes by opening the Smc3–Scc1 interface (6, 7, 16, 17). Both cohesin loading and release by these complexes require the ATPase activity of cohesin (17–21). Therefore, the Scc2–Scc4 and Pds5–Wapl complexes harness the energy of ATP hydrolysis to open the cohesin ring, allowing cohesin to entrap DNA dynamically. The acetyltransferase Eco1 (Esco1/2 in vertebrates) acetylates the Smc3 ATPase head and stabilizes cohesin on chromosomes (22–25), possibly through attenuating the ATPase activity of cohesin (19). In vertebrates, Smc3 acetylation enables the binding of sororin to cohesin and Pds5 (26). Sororin alters the molecular interactions among cohesin, Pds5, and Wapl, and inhibits the releasing activity of the Pds5–Wapl complex (26, 27).

Mutations of cohesin subunits and accessory proteins cause human developmental diseases termed cohesinopathy, including Cornelia de Lange syndrome (CdLS) (28, 29). About 60% of CdLS cases involve mutations of Scc2 [also called Nipped B-like protein (NIPBL)] (30–32). Thus, Scc2 is crucial for normal human development. Although both Scc2 and Scc4 are required for cohesin loading in vivo, Scc2 alone has been reported to enhance the topological loading of recombinant fission yeast cohesin onto circular DNA in vitro (18). The crystal structure of Scc4 bound to the N-terminal segment of Scc2 has been determined (33, 34). Low-resolution structures of the Scc2–Scc4 complex have also been determined by EM (33, 34). Despite the progress, the mechanism by which the Scc2–Scc4 complex promotes cohesin loading is not understood. In particular, the Scc2–Scc4 complex has been reported to interact with multiple subunits of cohesin (18), but which interaction is functionally important is unclear.

To gain more insight into Scc2 function, we have determined the crystal structure of Scc2 from the thermophilic fungus, Chaetomium thermophilum (Ct). Scc2 consists almost entirely of helical repeats that fold into a molecular hook. The overall structure of Scc2 is similar to two other Scc1-binding HEAT repeat proteins, Scc3 and Pds5, which also adopt highly bent structures (8, 9, 27, 35, 36). Like Scc3 and Pds5, Scc2 interacts with Scc1. Mutation of conserved Scc2 residues targeted by missense cohesinopathy mutations diminishes Scc1 binding, establishing the functional relevance of the observed Scc2–Scc1 interaction. Pds5 competes with Scc2 for binding to Scc1. Through competing with Scc2 for Scc1 binding, Pds5 may help to release Scc2 from the loaded and acetylated cohesin, allowing Scc2 to perform additional rounds of loading.

Results

Crystal Structure of Ct Scc2.

Scc2 contains an N-terminal disordered region that binds to Scc4 and a C-terminal HEAT repeat domain (Fig. 1A). Structures of Scc4 in complex with Scc2N reveal that Scc2N wraps around the tetratricopeptide repeat (TPR) domain of Scc4, forming extensive interactions (33, 34). Scc4 is required for cohesin loading in vivo likely through stabilizing Scc2 and through targeting Scc2 to defined chromosome loci (33), but is dispensable for cohesin loading in an in vitro reconstituted assay (18). Thus, the C-terminal HEAT repeat domain of Scc2 is critical for cohesin loading.

Fig. 1.

Crystal structure of Ct Scc2. (A) Schematic drawing of the Ct Scc2–Scc4 complex. (B) Cartoon drawing of the crystal structure of Ct Scc2_385–1,840 in two orientations. The helical insert domain (HID) is colored gray. The rest of the protein is colored blue. The positions of the HEAT repeats are indicated. All structure figures are made with PyMOL (www.pymol.org). (C) Cartoon drawing of the crystal structure of the human [Homo sapiens (Hs)] SA2–Scc1 complex, in the same orientation and scale as Scc2 on the left in B. SA2 and Scc1 are colored blue and magenta, respectively. (D) Cartoon drawing of the crystal structure of Lachancea thermotolerans (Lt) Pds5 with the bound Scc1 shown in sticks, in the same orientation and scale as Scc2 on the left in B.

We coexpressed Ct Scc2 and Ct Scc4 in insect cells with the baculoviral expression system and obtained the full-length Scc2–Scc4 complex. Consistent with previous reports (33, 34), Scc4 stabilized the full-length Scc2 protein. In the absence of Scc4, Scc2 underwent extensive proteolysis presumably at its N-terminal region. Because we could not crystalize the Scc2–Scc4 complex, we made a series of truncation mutants of Scc2 and obtained diffracting crystals of Scc2_385–1,840, which contained the entire HEAT repeat domain of Scc2. The crystal structure of Scc2 was then determined to 2.8-Å resolution with the single-wavelength anomalous diffraction (SAD) method (Table S1).

Table S1.

Data processing and refinement statistics

Crystals	Se-Met	Native
Data collection
Space group	P1 (remerged to P21)	P21
Cell dimensions
a, b, c; Å	66.80, 88.80, 160.30	66.73, 88.78, 160.30
	67.13, 89.24, 161.22 (in P21)
α, β, γ; °	89.95, 93.47, 90.56	90.00, 93.17, 90.00
	90.00, 93.45, 90.00 (in P21)
Resolution, Å	50–3.20 (3.26–3.20)*	36.2–2.80 (2.85–2.80)
	50–3.20 (3.26–3.20)
Rsym or Rmerge†	17.6 (NA)	13.5 (33.5)
	18.3 (NA) (in P21)
CC1/2	1.000 (0.494)	1.000 (0.642)
	1.000 (0.693) (in P21)
<I>/<σI>	10.6 (1.5)	17.4 (1.7)
	14.8 (2.3) (in P21)
Completeness, %	99.2 (98.7)	99.9 (99.8)
	100.0 (100.0) (in P21)
Multiplicity of observations	5.8 (5.3)	5.2 (4.8)
	11.2 (10.2) (in P21)
No. of unique reflections	60,615 (2,983)	46,160/2,301
	31,317 (1,536) (in P21)
Refinement
Resolution, Å		36.24–2.80 (2.87–2.80)
No. of reflections (work/free)		44,294 (2,041)/2,004 (100)
Rwork/Rfree		22.8 (32.7)/26.7 (36.4)
No. atoms
Protein		10,314
Ions		13
Rmsd
Bond lengths, Å		0.003
Bond angles, °		0.488
Completeness, %		95.5 (65.0)
Ramachandran plot
Favored, %		93.2
Outliers, %		0.5
Allowed, %		6.3

CC_1/2, Pearson correlation coefficient; <I>/<σ_I>, the empirical signal-to-noise ratio; NA, not applicable.

^*Values in parentheses correspond to the last resolution shell.

^†$R_{\text{merge}}={\displaystyle \sum _h{\displaystyle \sum _j\left|{\text{I}}_{hj}-\langle {\text{I}}_h\rangle \right|}}/{\displaystyle \sum _h{\displaystyle \sum _j{\text{I}}_{hj}}},$ where I_hj is the intensity of observation j of the reflection h.

Scc2 consists of 24 HEAT repeats and a helical insert domain located between repeats H7 and H8 (Fig. 1B and Fig. S1). Scc2 is shaped like a handled hook, with H1–H7 forming the handle and H8–H24 forming the hook. This highly bent architecture is also observed in the low-resolution EM structures of the budding and fission yeast Scc2 proteins (33, 34). Thus, the curvature of Scc2 is a defined, conserved feature of the cohesin loader.

Fig. S1.

Sequence alignment of Scc2 proteins from various species. The conserved residues are shaded yellow. The secondary structure elements of Ct Scc2 are shown on top. Hs Scc2 residues mutated in CdLS are indicated by green dots. Ct Scc2 residues required for Scc1 binding are indicated by open red circles. Ct, chaetomium thermophilum; Hs, Homo sapiens; Sc, Saccharomyces cerevisiae; Sp, Schizosaccharomyces pombe.

The highly bent structure of Scc2 is reminiscent of the structures of Scc3 [stromal antigen 1 or 2 (SA1/2) in vertebrates] and Pds5 (8, 9, 27, 35, 36) (Fig. 1 C and D). In fact, a search through the Dali server identified the structure of SA2 bound to Scc1 to be most closely related to the structure of SA2 bound to Scc2. Both Scc3 and Pds5 have highly bent structures, which are shaped like a dragon and a plier lever, respectively. In the cases of Scc2 and Scc3, the curvatures in their structures are formed by highly distorted HEAT repeats whose two helices deviate substantially from their antiparallel arrangement. In the case of human Pds5, the curvature is further stabilized by a tightly bound cofactor, inositol hexakisphosphate (27).

Both Scc3 and Pds5 are Scc1-binding proteins, and they bind to different regions of Scc1 (8, 9, 27, 35–37). Scc3 binds to the middle region of Scc1, whereas Pds5 binds to a region immediately adjacent to the NHD. The structure of the human SA2–Scc1 complex reveals that a 70-residue fragment of Scc1 binds across a large segment of SA2, forming extensive interactions (8) (Fig. 1C). Mutagenesis results have shown that the interfaces between the three C-terminal helices of Scc1 (αB–αD) and the bottom of the U-shaped body of the SA2 dragon contribute much of the binding energy to the SA2–Scc1 interaction (8). Likewise, the structures of the fungal Pds5–Scc1 complexes reveal that Scc1 binds at the C-terminal jaw of the Pds5 plier lever (35, 36) (Fig. 1D). The fact that Scc2 shares a similar, curved shape with Scc3 and Pds5 raises the intriguing possibility that Scc2 might use its C-terminal hook to interact with Scc1.

Scc1 Binding by Scc2.

To identify with which cohesin subunit Scc2 interacts, we obtained ³⁵S-labeled Smc1–Smc3 heterodimer, Smc1, Smc3, Scc3, and Scc1 through in vitro translation, and tested their binding to GST-Scc2 (Fig. 2A). GST-Scc2 only bound to Scc1, but not to Smc1, Smc3, or Scc3. Deletion mutagenesis mapped the Scc2-binding region of Scc1 to an N-terminal region encompassing residues 126–230 (Fig. 2B and Fig. S2). As determined by isothermal titration calorimetry, purified recombinant Scc2_385–1,840 and Scc1_126–230 bound to each other with a dissociation constant (K_d) of 20.4 nM (Fig. 2 C and D). Thus, Scc2 can directly bind to the N-terminal region of Scc1 with high affinity. Ct Scc1 fragments that contain the NHD cannot be expressed in bacteria or insect cells, preventing us from quantitatively measuring a potential contribution of that domain to Scc2 binding.

Fig. 2.

Ct Scc2 binds to the N-terminal region of Ct Scc1. (A) GST or GST-Scc2_385–1,840 proteins were immobilized on Glutathione Sepharose beads. Beads were incubated with the indicated ³⁵S-labeled cohesin subunits. The input and bound proteins were separated by SDS/PAGE gels, which were stained with Coomassie (Bottom) and analyzed with a phosphorimager (Top). (B) Domains and motifs of Ct Scc1. SCS, separase cleavage site; WHD, winged helix domain. (C) Coomassie-stained gel of purified recombinant Scc2_385–1,840 and Scc1_126–230. (D) Isothermal titration calorimetry analysis of the binding between Scc2_385–1,840 and Scc1_126–230.

Fig. S2.

Ct Scc2 binds to the N-terminal region of Ct Scc1. (A and B) GST or GST-Scc2_385–1,840 proteins were immobilized on Glutathione Sepharose beads. Beads were incubated with the indicated ³⁵S-labeled Scc1 fragments. The input and bound proteins were separated by SDS/PAGE gels, which were stained with Coomassie (Left) and analyzed with a phosphorimager (Right).

Our Scc1-binding results are consistent with an earlier report that identified residues 145–152 in the fission yeast Rad21 (Scc1 ortholog) as a binding element of Scc2–Scc4 in binding reactions on peptide arrays (18). The same study also reported numerous interactions between Scc2–Scc4 and peptides from other cohesin subunits, including Smc1, Smc3, and Scc3. In contrast, we did not observe detectable interactions between Scc2 and these other cohesin subunits. The underlying reasons for this discrepancy are unknown at present. It is possible that some isolated peptides used in the previous study are not properly folded and can develop nonphysiological interactions with Scc2–Scc4. In any case, our binding assays clearly establish an interaction between Scc2 and the N-terminal region of Scc1. We stress that our binding assays do not rule out possible, functional interactions between Scc2 and other cohesin subunits.

Cohesinopathy Mutations Diminish the Scc2–Scc1 Interaction.

The majority of CdLS cases are caused by mutations in one of the two alleles of NIPBL (encoding human Scc2), including missense, nonsense, and deletions, suggesting that these Scc2 mutations are loss-of-function mutations and Scc2 haploinsufficiency is the underlying cause for disease (32). Human Scc2 is a much larger protein, with its C-terminal HEAT repeat domain sharing high sequence similarity with Scc2 proteins from other species, including Ct Scc2 (Fig. S1). Mapping of the Scc2 residues conserved from yeast to man onto the structure of Ct Scc2 reveals that these conserved residues cluster around two patches (I and II) in the hook of Scc2 (Fig. 3A). Interestingly, many of these conserved residues are targeted by missense CdLS mutations in human Scc2 (Fig. 3 B and C and Fig. S3), suggesting that these residues are functionally important. We note that not all residues corresponding to those residues mutated in CdLS are surface-exposed. Mutations of buried residues may disrupt the folding and stability of Scc2 locally or globally.

Fig. 3.

Cohesinopathy mutations target conserved residues in Scc2. (A) Surface drawing of Ct Scc2 with conserved residues colored red and nonconserved residues mutated in CdLS colored yellow. Zoomed-in views of the cartoon diagram of Scc2, with residues in patches I (B) and II (C) shown in sticks. The color scheme is the same as in A.

Fig. S3.

Cohesinopathy mutations target conserved residues in Scc2. Stereo views of the cartoon diagram of Scc2, with residues in patches I (A) and II (B) shown in sticks. The color scheme is the same as in Fig. 3A.

We then created a panel of Ct Scc2 mutants targeting residues in patches I and II, based on the corresponding CdLS missense mutations in human Scc2, and expressed them as GST fusion proteins. Strikingly, the vast majority (16 of 19) of these GST-Ct Scc2 mutants were deficient in binding to the N-terminal fragment of Ct Scc1, albeit to varying degrees (Fig. 4 A and B). Our mutagenesis results establish the specificity of the Scc2–Scc1 interaction, and suggest that the hook of Scc2 might contact the N-terminal region of Scc1. The Scc2–Scc1 interaction is likely conserved in humans and other species. A deficient Scc2–Scc1 interaction might be an underlying cause of cohesinopathy, although this hypothesis remains to be formally tested.

Fig. 4.

Cohesinopathy mutations in Scc2 diminish Scc1 binding. (A) GST-Scc2_385–1,840 wild type (WT) and the indicated mutant proteins were immobilized on Glutathione Sepharose beads. Beads were incubated with ³⁵S-labeled Scc1_126–230. The bound proteins were separated by SDS/PAGE gels, which were stained with Coomassie (Bottom) and analyzed with a phosphorimager (Top). (B) Quantification of the relative Scc1 intensities in A. Mean ± SD, n = 3 independent experiments. Mutants with less than 50% of the WT activity are shown as red bars. (C) Surface drawing of Scc2 colored by the electrostatic potential (blue, positive; red, negative; white, neutral) in two orientations.

We again emphasize that not all mutated residues are likely to make direct contact with Scc1. For example, A1367 and L1373 are buried in a hydrophobic core of Scc2 and cannot form direct contact with Scc1. Instead, the A1367F and L1373P mutations may perturb the conformation of this segment of Scc2, indirectly affecting the binding of Scc1. Likewise, R1053 and R1081 form favorable electrostatic interactions with D1084, whereas R1090 and R1120 contact D1123, along the spine of Scc2. These residues may not form direct contact with Scc1 either. Their mutations may alter the local conformation of HEAT repeats H9, H10, and H11, thus impacting Scc1 binding.

The fission yeast Scc2 has DNA-binding activity, and the Scc2–Scc4 complex stimulates the ATPase activity of the fission yeast cohesin only in the presence of DNA (18). An inspection of the electrostatic potential of Ct Scc2 reveals no large, contiguous, positively charged surfaces that could serve as DNA-binding sites (Fig. 4C). The largest positively charged patch is the conserved patch I in the middle region of Scc2. We have shown, however, that mutations of several basic residues in this patch, including R1081, R1090, K1091, R1092, and R1120, diminish Scc1 binding. Because the N-terminal Scc2-binding region of Scc1 (residues 126–230) is highly negatively charged and has a pI of 3.73, this positively charged site is likely involved in Scc1 binding, as opposed to DNA binding. The C-terminal region of Ct Scc2 (residues 1,888–1,946; not included in our crystallization construct) contains a long stretch of basic residues. It will be interesting to test whether this C-terminal basic region of Scc2 contributes to DNA binding by Scc2, although this region is poorly conserved.

Pds5 Competes with Scc2 for Binding to Scc1.

Because Pds5 and Scc2 bind to overlapping N-terminal regions in Scc1 (Fig. 2B), we tested whether they competed for binding to a fragment of Scc1 (residues 51–230) that encompasses both binding regions. As expected, GST-Scc2 bound efficiently to this larger Scc1 fragment (Fig. 5A). Addition of increasing amounts of recombinant Ct Pds5 diminished the binding of Scc1 to GST-Scc2 in a dose-dependent manner. Ct Pds5 itself did not bind to GST-Scc2 beads. Thus, Pds5 and Scc2 indeed compete for binding to Scc1. They cannot simultaneously bind to Scc1 to form a ternary complex.

Fig. 5.

Pds5 competes with Scc2 for Scc1 binding. (A) GST or GST-Scc2_385–1,840 proteins were immobilized on Glutathione Sepharose beads. Beads were incubated with ³⁵S-labeled Scc1_51–230 in the absence or presence of increasing amounts of Pds5. Input and bound proteins were separated by SDS/PAGE gels, which were stained with Coomassie (Bottom) and analyzed with a phosphorimager (Top). The relative binding intensities of Scc1 are quantified and indicated below the autoradiograph. (B) Model for Scc2-dependent cohesin loading onto DNA, highlighting the importance of the Scc2–Scc1 interaction. The Scc2–cohesin complex (mediated, in part, by the Scc2–Scc1 interaction) might be the functional entity that promotes transcription. The competition between Pds5 and Scc2 for Scc1 binding might trigger the release of Scc2 from the loaded and acetylated cohesin.

Pds5 inhibits Scc2-dependent topological loading of the fission yeast cohesin onto DNA in a reconstituted cohesin-loading assay in vitro (19). Our demonstration of a direct competition between Pds5 and Scc2 for Scc1 binding provides a straightforward explanation for the inhibitory effect of Pds5 in cohesin loading. Taken together, these findings support a role of the Scc2–Scc1 interaction in cohesin loading. Unfortunately, we cannot test this possibility directly, because we have so far failed to reconstitute Scc2-dependent topological loading of recombinant Ct cohesin onto DNA.

Discussion

Mechanism of Scc2-Dependent Cohesin Loading.

The cohesin ring can topologically entrap DNA. For DNA to enter this ring, one or more of the three interfaces or gates—the hinge interface, the Smc3–Scc1N interface, and the Smc1–Scc1C interface—need to be opened (38) (Fig. 5B). Artificial tethering of the Smc1 and Smc3 hinges prevents cohesin loading in the budding yeast, whereas fusing Smc3 to the N terminus of Scc1 and tethering of the Smc1–Scc1C interface still allows functional cohesin loading (39). These results suggest that cohesin loading involves the opening of the hinge interface. Cohesin loading requires the ATPase activity of cohesin. It is unclear how ATP hydrolysis by the ATPase heads can open the hinge interface that is tens of nanometers away. An alternative model posits that Scc2 and DNA form a transient complex with cohesin and stimulate ATP hydrolysis by cohesin, opening the Smc3–Scc1N interface to allow DNA entry (19). Because the Smc3–Scc1N interface is adjacent to the ATPase heads, it is easier to envision how ATP hydrolysis might open this interface. On the other hand, this model cannot easily explain the finding that the Smc3–Scc1 fusion protein supports functional cohesion. Thus, the identity of the DNA entry gate(s) of cohesin remains to be established.

In this study, we have shown that Scc2 interacts with the N-terminal region of Scc1, but not other cohesin subunits. Several lines of evidence suggest that the Scc2–Scc1 interaction is specific and might be functionally important for cohesin loading. First, Scc2 has an overall fold and shape that are similar to two other Scc1-binding proteins, Scc3 and Pds5. Second, cohesinopathy mutations targeting conserved residues in Scc2 disrupt the Scc2–Scc1 interaction. Finally, Pds5, which is known to block Scc2-dependent cohesin loading in vitro, blocks the Scc2–Scc1 interaction. We speculate that this Scc2–Scc1 interaction, along with DNA-bridged interactions between Scc2 and other cohesin subunits, might stabilize the ATP-bound, heads-engaged conformation of cohesin, thereby promoting ATP hydrolysis and cohesin loading through the yet-to-be-determined entry gate(s) (Fig. 5B). Alternatively, Scc2 rigidifies the flexible segments of Scc1 and allows the ATPase-driven head disengagement to disrupt the Smc3–Scc1N or Smc1–Scc1C interface. We do not know the biochemical activity of the handle of Scc2, which harbors a highly conserved motif (residues 656–671 in Ct Scc2), but suspect that this handle might contact subunit interfaces in intact cohesin to promote the opening of the cohesin ring.

The competition between Pds5 and Scc2 for binding to Scc1 is intriguing and possibly relevant to cohesin regulation in vivo. Smc3 acetylation can occur at low levels outside S phase and is regulated by the ATPase activity of cohesin (40, 41). Although binding of Pds5 to cohesin does not require Smc3 acetylation in human cells, the Pds5-dependent binding of sororin to cohesin requires Smc3 acetylation (27). In the budding yeast, Pds5 protects cohesin from deacetylation and turns over more rapidly on pericentric chromatin in cells lacking Eco1 (37, 42), suggesting that Pds5 might interact with acetylated Smc3 in that organism. Thus, Smc3 acetylation and subsequent binding of Pds5 (Pds5–sororin interaction in vertebrates) might actively promote the release of Scc2 from loaded cohesin, freeing it to perform additional rounds of cohesin loading (Fig. 5B). Conversely, the soluble cohesin released from chromosomes by Pds5–Wapl interaction or separase is deacetylated by Hos1 in the budding yeast and by Hdac8 in human cells (43, 44). Inactivation of Hdac8 in human cells traps cohesin in the acetylated and sororin-bound form (44), which likely also contains Pds5. Inactivation of Hos1 in the budding yeast also prevents Smc3 deacetylation during mitosis and proper cohesion in the ensuing S phase (43). In these situations, persistent Pds5 binding to acetylated cohesin is expected to prevent Scc2-dependent cohesin loading, contributing to the observed phenotypes. We speculate that cohesin deacetylation might be critical for releasing Pds5 (or Pds5–sororin interaction) from cohesin, enabling a fresh cycle of Scc2-dependent loading onto chromosomes.

Implications for Cohesinopathy.

Mutations of one of the two Scc2 alleles account for the majority of known CdLS cases (32). The expression of the remaining wild-type Scc2 allele is variable in the cells of patients who have CdLS, and is inversely correlated with the severity of disease phenotypes (45). Patients with higher Scc2 expression show milder phenotypes. This correlation is observed even in patients with CdLS who have mutations in other cohesin subunits. Thus, proper levels of Scc2 are particularly important for human development. Because the cells of patients with CdLS have no gross defects in cohesin loading or sister-chromatid cohesion, the prevailing model is that Scc2 mutations cause cohesinopathy by perturbing the transcription of developmental genes (28, 29). Although it is possible that low levels of Scc2 impair cohesin loading at specific genes, it is equally possible that the cohesin–Scc2 complex is the functional entity that regulates transcription. In support of this hypothesis, in mouse embryonic stem (ES) cells, cohesin, Scc2, and the mediator co-occupy the enhancer and promoter regions of actively transcribed genes and activate transcription through the formation of enhancer-promoter DNA loops (46). We have shown that many Scc2 missense mutations in CdLS specifically disrupt the Scc2–Scc1 interaction. In the future, it will be interesting to test whether these Scc2 mutations disrupt the cohesin–Scc2 mediator complex and the formation of the enhancer-promoter loops in human ES cells or induced pluripotent stem cells. These experiments will provide fresh insight into the molecular basis of cohesinopathy.

Materials and Methods

Protein Expression and Purification.

The cDNAs of Ct Scc2 and Pds5 were synthesized and subcloned into the pFastBacHT vector. Ct Scc2 (native and selenomethionine-labeled) and Pds5 proteins were then expressed in insect cells and purified with standard procedures. Details are provided in SI Materials and Methods.

Crystallization of Ct Scc2.

Ct Scc2_385–1,840 crystals grew within a few days after the mixing of the protein solution (5 mg/mL) with the reservoir solution containing 150 mM sodium citrate tribasic dihydrate (pH 5.2), 5 mM tris(2-carboxyethyl)phosphine hydrochloride (TCEP-HCl), and 9% (wt/vol) PEG 6000. Details are provided in SI Materials and Methods.

Data Collection and Structure Determination.

X-ray diffraction data were collected at an Advanced Photon Source beamline. HKL-3000 was used to process both the l-selenomethionine and native diffraction datasets (47). Phases were obtained through a SAD experiment. Details are provided in SI Materials and Methods. Data processing, phasing, and model refinement statistics are provided in Table S1.

Protein-Binding Assays.

GST pull-down assays and isothermal titration calorimetry were used to analyze the binding between Scc2 and cohesin subunits. Details are provided in SI Materials and Methods.

SI Materials and Methods

Protein Expression and Purification.

The cDNA of Ct Scc2 was synthesized at GenScript. A cDNA fragment of Ct Scc2 encoding residues 385–1,840 was subcloned into the pFastBacHT vector that introduced an N-terminal His₆-tag followed by a tobacco etch virus (TEV) protease cleavage site. The Ct Scc2_385–1,840 baculovirus was made with the Bac-to-Bac system (Invitrogen). For protein expression, Hi5 insect cells (Sigma–Aldrich) were infected with the Scc2_385–1,840 baculovirus and cultured for about 50 h at 27 °C. Cells were harvested, resuspended in buffer I [20 mM Tris⋅HCl (pH 7.5), 500 mM NaCl, 20 mM imidazole, 2 mM 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride (AEBSF), and a protease inhibitor mixture], and lysed by sonication. After centrifugation, the supernatant was incubated with the Ni²⁺ Sepharose 6 Fast Flow resin (GE Healthcare) for 1 h at 4 °C. The Ni²⁺ Sepharose resin was washed with buffer II [20 mM Tris⋅HCl (pH 7.5), 1 M NaCl, and 20 mM imidazole] and buffer III [20 mM Tris⋅HCl (pH 7.5), 100 mM NaCl, and 20 mM imidazole]. The His₆-Scc2 protein was then eluted with buffer IV [20 mM Tris⋅HCl (pH 7.5), 100 mM NaCl, and 150 mM imidazole], and incubated with the TEV protease at 4 °C overnight to cleave the His₆-tag. The cleaved Scc2 protein was diluted with an equal volume of buffer V [20 mM Tris⋅HCl (pH 8.0) and 50 mM NaCl], applied onto a HiTrap Q HP column (GE Healthcare) that had been equilibrated with buffer V and eluted with a linear 0–500 mM NaCl gradient. Fractions containing the Scc2 protein were combined and further applied onto a HiLoad 16/60 Superdex 200 prep grade column (GE Healthcare) that had been equilibrated with buffer VI [20 mM Tris⋅HCl (pH 7.5), 200 mM NaCl, and 5 mM TCEP]. The purified Scc2 protein was then concentrated to 20 mg/mL using an Amicon Ultra-15 centrifugal filter unit (Millipore).

For the preparation of selenomethionine-labeled Ct Scc2_385–1,840, Hi5 cells were infected with the Scc2_385–1,840 baculovirus and incubated for 10 h at 27 °C with shaking. The infected cells were collected with centrifugation. The cell pellet was washed and resuspended in the methionine-deficient ESF921 medium (Expression Systems). The cells were further incubated with shaking for another 8 h. The cells were collected again, washed, resuspended with the ESF921 medium supplied with 150 mg/L l-selenomethionine (SeMet), and incubated for an additional 38 h. The SeMet-labeled Scc2 protein was then purified as described above.

The full-length Ct Pds5 cDNA was synthesized at GenScript and subcloned into the pFastbacHT vector. The Pds5 baculovirus was made with the Bac-to-Bac system. Hi5 insect cells were infected with the baculovirus and incubated for about 50 h at 27 °C. Ct Pds5 was purified as described (27). Purified Ct Pds5 was stored in the storage buffer [20 mM Tris⋅HCl (pH 7.5), 150 mM NaCl, and 1 mM DTT] at −80 °C.

Crystallization of Ct Scc2.

All crystallization experiments were performed at 20 °C. Initial screens were performed with a Phoenix crystallization robot (Art Robbins Instruments), using commercially available screening kits from Hampton Research, Qiagen, and Molecular Dimensions. Crystallization conditions obtained from the initial screens were optimized with the hanging drop vapor diffusion method and microseeding. The Ct Scc2_385–1,840 crystals suitable for X-ray diffraction studies grew within a few days after the protein solution (5 mg/mL) was mixed with the reservoir solution containing 150 mM sodium citrate tribasic dihydrate (pH 5.2), 5 mM TCEP-HCl, and 9% (wt/vol) PEG 6000. The SeMet-labeled Scc2 crystals were grown under the same conditions. All crystals were cryoprotected with the reservoir solution including 15% (wt/vol) ethylene glycol.

Data Collection and Structure Determination.

X-ray diffraction datasets were collected at the Advanced Photon Source (APS) beamline Sector 19-ID. HKL-3000 was used to process both the SeMet and native diffraction datasets (47). Computational corrections for absorption in the crystal, as well as corrections to the calculations of the Lorentz factor due to the minor misalignment of the goniostat, were applied (48, 49). Anisotropic diffraction was corrected to adjust the error model and to compensate for a radiation-induced increase of nonisomorphism within the crystal (50–52). The Se-Met crystal diffracted to a resolution of 3.2 Å and was slightly anisotropic, with indexing, integration, and scaling indicating P1 symmetry with values of α and γ close to 90°. The diffraction of the native crystal was more anisotropic (3.0 Å in the x direction, 2.9 Å in the z direction, and 2.5 Å in the y direction), with indexing, integration, and scaling indicating P2₁ symmetry. The symmetry discrepancy between the native and SeMet datasets prompted us to assess how much the intensities in the P1 symmetry disagreed with the P2₁ symmetry. Using HKL-3000 (“automatic corrections” option), we estimated the magnitude of the increase in systematic error caused by merging data in higher symmetry to be 3.7% of the native intensity. The anomalous signal level was estimated to be 5.0% of the native intensity. We weighted the decrease in uncertainty of the signal estimates, caused by the higher multiplicity of observations in P2₁, against the increase of systematic error in the signal estimates due to imposing higher symmetry. Based on this consideration, we decided to remerge the two SeMet datasets in P2₁ symmetry, and used P2₁ symmetry in the heavy atom search, density modification, and initial model building performed with the SeMet dataset. Data processing statistics are presented in Table S1.

Initial phases were obtained in a SAD experiment with a SeMet crystal from data acquired at λ = 0.979 Å. The search for heavy atom positions was performed to a resolution of 5.0 Å. The 59 Se positions were identified using SHELX (53), with correlation coefficients (CC): CC_All = 43.1%, CC_Weak = 22.0%, and combined figure of merit (CFOM) = 65.1%. Relative occupancies of Se positions varied from 0.143 to 1.000. The handedness of the best solution was determined with SHELXE. The heavy atom positions were refined to 3.9 Å with MLPHARE, with the final FOM reaching 0.251 for all observations. This process resulted in 44 refined heavy atom positions. The density modification was performed with PARROT (54), and resulted in an electron density map with multiple, clearly interpretable α-helices, which we placed using the “Place Helix Here” option in “Other Modelling Tools” in Coot (55). This set of helices, which comprised about 500 alanine residues, was used as the entry model for model building with BUCCANEER (56) and refinement with REFMAC (57), run within HKL-3000. The resulting main chain model was about 90% complete (about 1,300 residues), with about 80% side chains docked into the electron density maps.

This assembly was used to perform isomorphous replacement with the native dataset using MOLREP (58), and rebuilt and refined again with BUCCANEER and REFMAC, all run within HKL-3000, to R = 26.4% and R_free = 32.7%. The resulting model was 84% complete, with 76% of side chains docked into the electron density. Positional and isotropic atomic displacement parameter refinement, as well as translation, libration, and skew atomic displacement parameter refinement, was performed to a resolution of 2.80 Å using the program PHENIX (59), with a random 4.5% of all data set aside for the R_free calculation. Phasing and model refinement statistics are provided in Table S1. The model quality was validated with Molprobity (60), and assessed to be satisfactory on the basis of the data resolution and a Molprobity score of 1.36, which corresponds to the 100th percentile of the 4,464 Protein Data Bank deposits solved at resolutions from 2.55 to 3.05 Å.

Protein-Binding Assays.

The cDNA fragment encoding Ct Scc2_385–1,840 was subcloned into pGEX6p-1. The resulting pGEX6p-1–Scc2 plasmid encoded an N-terminal GST-tagged Scc2 protein. All Ct Scc2 mutant plasmids were made with the QuikChange Kit (Stratagene). The wild-type (WT) and mutant plasmids were transformed into Escherichia coli strain BL21 (DE3). Protein expression was induced by 0.2 mM isopropyl-d-1-thiogalactopyranoside at 20 °C overnight. GST-Scc2_385–1,840 proteins were purified with the Glutathione Sepharose 4B resin (GE Healthcare), and stored at −80 °C in the storage buffer [20 mM Tris⋅HCl (pH 7.5), 150 mM NaCl, and 1 mM DTT]. The Ct Smc1, Smc3, Scc1, and Scc3 cDNAs were synthesized by GenScript and subcloned into the pCS2 vector. Various fragments of Ct Scc1 were amplified by PCR and subcloned into the pCS2 vector. These plasmids were added into a TNT Quick Coupled Transcription Translation System (Promega), and incubated at 30 °C for 90 min in the presence of [³⁵S]methionine. These ³⁵S-labeled cohesin subunits were mixed with Glutathione Sepharose beads bound to 10 μg of GST-Scc2_385–1,840 WT or mutant proteins, and incubated for 1 h at 4 °C in the binding buffer [20 mM Tris⋅HCl (pH 7.5), 150 mM NaCl, 0.1% Triton X-100, and 1% (wt/vol) dry milk]. After incubation, the beads were washed four times with the wash buffer [20 mM Tris⋅HCl (pH 7.5), 150 mM NaCl, and 0.1% Triton X-100]. The bound proteins were separated on SDS/PAGE gels, which were stained with Coomassie blue, dried, and analyzed with a phosphorimager (GE Healthcare). The intensities of ³⁵S-labeled proteins bound to beads were quantified with ImageJ (NIH).

For the Pds5 competition assay, ³⁵S-labeled Ct Scc1_51–230 was incubated with varying concentrations (400 nM, 800 nM, 1.6 μM, and 3.2 μM) of Ct Pds5 for 2 h at 4 °C in 50 μL of the binding buffer [20 mM Tris⋅HCl (pH 7.5), 150 mM NaCl, and 0.1% Triton X-100]. After incubation, the protein mixture and GST-Scc2_385–1,840 were added together to Glutathione Sepharose 4B beads. The reaction mixtures were further incubated for 1 h at 4 °C. The beads were washed four times with the binding buffer. The proteins bound to beads were separated on SDS/PAGE gels, which were stained with Coomassie blue, dried, and analyzed with a phosphorimager.

Scc1_126–230 was expressed in bacteria and purified with a combination of Ni²⁺ affinity, anion exchange, and size exclusion chromatography. Isothermal titration calorimetry (ITC) was performed with a MicroCal iTC200 (Malvern Instruments) at 20 °C. All samples were dialyzed and diluted with the ITC buffer [20 mM Tris⋅HCl (pH 7.5) and 100 mM NaCl] before the measurement. The titration consisted of 21 1.9-μL injections of Scc1_126–230 (380 μM) into a stirred reaction cell containing Scc2_385–1,840 (37.8 μM). The resulting thermogram was integrated using the program NITPIC (61). The dissociation constant (K_d) was obtained by fitting the integrated isotherm with the program SEDPHAT (62), with a correction factor being applied to the concentration of Scc1_126–230.