Main

Saccharomyces cerevisiae separase contains 1,630 amino acid residues and the N-terminal α-helical region covers approximately 1,160 residues. We obtained good quality crystals for the complex of S. cerevisiae separase (residues 51–1630; Fig. 1a, Extended Data Figs 1, 2, 3) and securin (residues 258–373; Extended Data Fig. 4). The N-terminal segment of securin contains the KEN-box and D-box motifs, which are important for the destruction of securin but not for the interaction with separase20. The structure was determined and refined at 3.0 Å resolution (Extended Data Table 1). Under the same crystallization condition, we observed another crystal form and collected a dataset to 3.7 Å resolution. We readily solved the structure of this crystal form and did not observe any notable structural differences.

Figure 1: Crystal structure of the yeast separase–securin complex.
figure 1

a, Domain organization of S. cerevisiae separase. The domains are labelled and colour-coded. b, Overall structure of the yeast separase–securin complex. The separase domains are coloured as in a, and the securin SIS is in magenta. The side chain of the catalytic Cys1531 residue of separase is shown as a sphere model. c, Overall structure of the complex viewed after a 50° rotation around the vertical axis. Two of the phosphorylation sites in the securin SIS (Ser277 and Ser292)29,30 are indicated by spheres and labelled. The structure figures were produced using PyMOL (http://www.pymol.org).

PowerPoint slide

Full size image

We produced crystals of another complex of S. cerevisiae separase (residues 71–1630) and securin (residues 258–373), and determined its structure at 2.6 Å resolution. The overall structure of this complex is essentially the same as that at 3.0 Å resolution (Extended Data Fig. 5). One difference is that residues 75–79 at the N-terminal end of separase form an anti-parallel β-sheet with the equivalent residues of another separase molecule in the crystal (Extended Data Fig. 5), which is an artefact owing to the longer truncation of separase in this complex.

The overall structure of the separase–securin complex has a highly elongated shape, with dimensions of 65 × 70 × 165 Å (Fig. 1b, c). The catalytic domain (CD) is located at one end of the complex, far away from the N terminus of separase at the other end. This elongated shape is generally consistent with that observed for the human separase–securin complex by electron microscopy18, but distinct from a closed form reported for the Caenorhabditis elegans complex19.

The structure shows that the N-terminal helical region (residues 53–1163) of separase can be divided into four domains (I–IV) (Fig. 1a–c). Their structures are generally like HEAT repeats, with extensions that connect each domain to the next one. Domain I is folded back onto domain II such that the loops at one end of each domain are facing each other (Fig. 2a), with a buried surface area of 1,100 Å2 for each domain, suggesting that this may be a stable association between them. Two consecutive helices (the fourth and fifth helices) in domain I are parallel to each other (Fig. 2a, Extended Data Fig. 6). The first 50 residues of separase are poorly conserved (Extended Data Fig. 1). Earlier studies showed that the first 155 residues of separase are important for function12. These residues include helices α1 to α5 of domain I. Their deletion would disrupt the structure of this domain and its interaction with domain II (Fig. 2a), and helix α3 is in the binding site for the C-terminal helix of securin (see below), thereby explaining their functional importance. Human separase is stimulated by DNA of greater than 100 base pairs21, and this stimulation is also likely to be mediated by the helical region.

Figure 2: Structures of the domains in separase.
figure 2

a, Structure of domains I and II of separase. The directions of helices 4 and 5 in domain I are indicated by the red arrows. b, Structure of domain III of separase. The helices are coloured from blue at the N terminus to red at the C terminus. c, Structure of domains IV, SD and CD of separase, and overlay of the structure of the SD–CD of C. thermophilum separase (grey)17. The active site is indicated by the red asterisk. The red arrowhead indicates the region in which the β4A–β4B hairpin has a different conformation and is partly disordered in C. thermophilum separase, and the purple box highlights the region in which domain IV provides an extra strand (labelled 6) to the β-sheet of the SD.

PowerPoint slide

Full size image

The helical hairpins of domain III are arranged in a tight, right-handed super-helix (Fig. 2b). This domain is located in the middle of the structure and has contacts with domain II at one end (buried surface area of approximately 900 Å2) and domain IV at the other (1,100 Å2) (Fig. 1b). It also has a direct contact with a segment immediately following the catalytic Cys1531 residue in the CD (strands β4A, β4B and the connecting loop; Fig. 2c, Extended Data Figs 3, 6). The overall arrangement of the helices in this domain has similarity to that in several other structures based on a DaliLite search22, and the closest homologues include the TPR domain of the G-protein signalling modulator LGN23,24,25 and others (Extended Data Fig. 7).

Domain IV has extensive interactions with the domain immediately before the CD (substrate-binding domain (SD), Fig. 1a; see next paragraph), covering one face of its β-sheet (buried surface area of approximately 1,900 Å2 for each domain) and providing one extra strand to this β-sheet (Fig. 2c). It also interacts with two of the surface helices of the CD. Therefore, domain IV could be considered as forming a module together with the SD–CD.

An α + β domain is located just before the CD in separases18,26. The backbone fold of this domain has no similarity to proteases such as caspases17. The domain is tightly associated with the CD and contributes to substrate binding. We will refer to it as the substrate-binding domain (SD) (Fig. 1a). Other names for the SD and CD include PPD (pseudo-protease domain) and APD (active protease domain), respectively17. The SD contains a five-stranded (β1–β5) mostly anti-parallel β-sheet, and there is a large insertion (residues 1200–1380) between the third and fourth strands, including a hairpin of two long helices (αA–αB) and a four-helical bundle (αC–αF) (Fig. 2c). The backbone fold of this β-sheet is RNase H-like, with close homologues including part of the catalytic PIWI domain of Argonaute27 and others (Extended Data Fig. 7).

The overall structure of the SD–CD of S. cerevisiae separase is similar to that of the Chaetomium thermophilum separase17 (Fig. 2c), with a root mean squared (r.m.s.) distance of 1.5 Å for their equivalent Cα atoms and amino acid sequence identity of 39%, suggesting that the binding of securin did not cause a large overall conformational change in these two domains. The β-sheets of the two domains are juxtaposed in the structure, although the two neighbouring strands at the interface (β5 of SD and β6 of CD) are at an angle of nearly 90°. The αA–αB hairpin from SD covers a portion of the active site in CD and helps to mediate substrate binding.

Securin is primarily in an extended conformation, with only two short helices, and runs along the entire length of separase (Figs. 1b, 3a). We will refer to this segment of securin, residues 258–373, as the separase interaction segment (SIS). The backbone direction of SIS is anti-parallel to that of separase, such that the N-terminal region of SIS interacts with the C-terminal region of separase. The securin SIS has contacts with every domain of separase, being positioned in prominent grooves in its surface (Fig. 3a). Approximately 4,600 Å2 of the surface area of securin is buried in the interface with separase (3,900 Å2 surface area burial), indicating the tight association and the stability of this complex.

Figure 3: Interactions between separase and securin.
figure 3

a, Molecular surface of separase, coloured by the domains. The active site of separase is indicated by the red asterisk. The view is as in Fig. 1c. b, Omit Fo − Fc electron density at 3.0 Å resolution for residues 258–269 of securin, contoured at 2σ. c, Interactions between residues 258–265 of securin SIS (magenta) with the active site of separase. The side chains of residues in the interface are shown as stick models and labelled. The bound position of a substrate-mimic inhibitor to C. thermophilum separase is shown in grey17. d, Close-up of the active site region showing the differences between the bound position of securin (magenta) and that of the inhibitor to C. thermophilum separase (grey)17. e, Interactions between residues 271–288 of securin SIS (magenta) with domain III (light blue) of separase.

PowerPoint slide

Full size image

Most importantly, residues 258–269 of the securin SIS, with good electron density (Fig. 3b), are located in the active site of separase, interacting with the SD and CD (Fig. 3a, c). Therefore, one mechanism for securin to inhibit the protease activity of separase is through blocking substrate access to the active site. Residues Ile259–Glu260–Ile261 of securin mimic the P5–P3 residues of the substrate17 (Fig. 3c). Notably, the side chain of Ile259 has hydrophobic interactions with Trp1249 and Trp1250 located near the tip of the αA–αB helical hairpin of SD, and that of Glu260 is positioned near the N terminus of helix αE in CD. However, the P1 Arg residue of the substrate is replaced with Pro263 in securin. It assumes a different conformation and its main-chain carbonyl is 6 Å away from the thiol group of the catalytic Cys1531 residue (Fig. 3c, d). The position of Pro263 also clashes with that of His2083 in the C. thermophilum CD17, the second member of the catalytic machinery, and the conformation of the equivalent His1505 in the securin complex is probably an inactive one (Fig. 3d). These differences ensure that securin is not cleaved by separase even though it is located in the active site. Earlier studies showed that mutations of residues of securin equivalent to the P2, P1 and P1′ residues could turn it into a substrate14,17 (Extended Data Fig. 8).

Residues 264–269 of securin have interactions with the CD, and the P′ side of the substrate may be recognized by this region of separase as well. In contrast to the sequence conservation of the P side of the substrate, the P′ side is much less conserved (Extended Data Fig. 8). A P1′ Arg residue, found in several fungal species but not in Schizosaccharomyces pombe or C. thermophilum, may mediate the degradation of this fragment by the N-end rule pathway28. The structure shows that a groove in the active site region can accommodate the P′ residues, and the β4A–β4B hairpin following the catalytic Cys1531 may contribute to substrate binding here (Extended Data Fig. 8; see below). Residues 264–269 of securin are in contact with the rim of this groove rather than being located at its bottom, possibly owing to the shift in the position of the Pro263 residue relative to that of the P1 Arg of the substrate (Fig. 3c).

Following the active site region, residues 271–288 of the securin SIS are in close contact with domain III of separase (Fig. 3e), and residues 290–296 are located in a deep groove at the interface between domains III and II (Fig. 3a, Extended Data Fig. 6). These interactions are mostly hydrophobic in nature, and also involve several aromatic (Phe and Tyr) residues. Residues 306–316 of the securin SIS are placed on the side of domain II of separase (Fig. 1c). Finally, residues 365–373 at the C terminus of securin form a short helix (Extended Data Fig. 4) and are located in a surface depression in domain I (Fig. 1b, Extended Data Fig. 6). Residues 299–305 and 317–362 of securin are disordered in the current structure.

The securin SIS contains three potential phosphorylation sites29,30 (Fig. 1c): Ser277 (Fig. 3e), Ser292 (Extended Data Fig. 6) and Thr304. The S277A/S292A/T304A triple mutant had weaker interaction with separase and reduced nuclear location of separase, whereas the single and double mutants had no growth phenotype30. pSer277 should have favourable interactions with Arg1130 in domain IV of separase (Fig. 3e). The side chain of Ser292 is hydrogen-bonded to the main-chain amides of residues 294 and 295 in the SIS, and Thr304 is in a disordered region in the current structure, suggesting possible conformational rearrangements of Ser292 and Thr304 after phosphorylation.

Separase residues in contact with securin are generally well conserved (Extended Data Figs 1, 2, 3). The securin SIS is poorly conserved among species, but the residues in the major contact sites with separase are better conserved (Extended Data Fig. 4), suggesting that the securin homologues may share a common mechanism of interacting with and inhibiting separase. The observed binding mode of the SIS indicates that securin probably helps to stabilize the overall structure of separase, as suggested by previous biochemical studies12,13,14,15. It might also be possible that once securin is removed by proteasome degradation, the structure of separase alone could undergo a conformational change to assume a different arrangement of its domains, especially for domains I–III. The β4A–β4B segment is a loop (L4) in the structure of the C. thermophilum SD–CD (Extended Data Figs 6, 8), and residues in this loop are important for catalytic activity17. This segment is stabilized by contacts with domain III (Extended Data Figs 6, 8), and therefore a change in the position of domain III might be necessary for catalysis. This would represent another, indirect mechanism for securin to inhibit the catalytic activity of separase.

The structure of the separase–securin complex provides a molecular basis for the large amount of experimental data on this complex. This structure is also supported by our biochemical observations. The complex from the fungus Zygosaccharomyces rouxii was used for these studies, and its separase and the SIS of securin are closely related to those from S. cerevisiae (Extended Data Figs 1, 2, 3, 4). Zygosaccharomyces rouxii securin contains only 282 residues owing to a shorter N-terminal segment (Fig. 4a). We found that separase was produced in the insoluble fraction in insect cells unless an appropriate securin was co-expressed with it (Fig. 4b, Extended Data Fig. 9), which was probably due to the instability/misfolding of separase. On the basis of this observation, we assessed the importance of various segments of securin for producing soluble separase, which would reflect both its binding and chaperone functions. The N-terminal segment of securin, up to residue 166 (equivalent to residue 258 of S. cerevisiae securin, Extended Data Fig. 4), could not help produce soluble separase, and deleting these 165 N-terminal residues had little effect on the production of soluble separase (Fig. 4b, c). Additional experiments showed that deleting residue 231 to the C terminus of securin led to a relatively small reduction in the solubility of separase (Fig. 4b, c). These data indicate that residues 166–230 of Z. rouxii securin, equivalent to residues 258–316 of S. cerevisiae securin, are required for interactions with separase, in agreement with our structural information.

Figure 4: Biochemical characterizations of the interactions between separase and securin.
figure 4

a, Domain organizations of S. cerevisiae and Z. rouxii securin. The KEN motif and the destruction box (D-box) are also indicated. b, Z. rouxii separase (with an N-terminal His tag) was co-expressed with various segments of Z. rouxii securin. The eluates from nickel columns were separated by SDS gel electrophoresis. The positions of separase and securin are indicated by the black and red arrowheads, respectively. WT, full-length Z. rouxii separase; WT-CS, full-length Z. rouxii separase C1497S mutant; WT′, Z. rouxii separase with an internal deletion of residues 952–1010, corresponding to a segment in domain IV of the S. cerevisiae separase structure. For gel source data, see Supplementary Fig. 1. c, Summary of the expression results in b. The solubility levels of separase for various securin segments are indicated.

PowerPoint slide

Full size image

We also observed that the segment covering residues 189 to the C terminus of Z. rouxii securin was able to produce a small amount of soluble separase (Fig. 4b, c). This segment would be missing the residues located in the active site of separase, suggesting that binding to the active site is not essential for complex formation with separase. Conversely, gel-filtration studies of this complex showed that it migrated in the void volume, indicating that it was mostly aggregated. Therefore, even though this segment of securin could bind separase, it might not be sufficient for its chaperone function to promote the correct separase folding and/or conformation.

The segment covering residues 166–188 of Z. rouxii securin was not able to help produce soluble separase, indicating that binding to the active site region of separase alone is not sufficient to produce a stable complex and that the N-terminal helical region of separase has a crucial role in the regulation by securin. In addition, residues 189–250 of securin produced a much lower amount of soluble separase compared to that for residues 189–282 of securin, indicating an appreciable role for the C-terminal residues of securin.

Overall, our studies have produced the first crystal structure information on the separase–securin complex and defined the molecular basis for the tight regulation of separase by securin. The structure provides a foundation for understanding the biochemical and biological functions of this complex from S. cerevisiae. While the N-terminal helical region is poorly conserved among separases at the amino acid sequence level, they are better conserved at the secondary structure level (Extended Data Figs 1, 2). Therefore, it might be possible that the separase–securin complexes from other organisms could assume a similar overall architecture as observed here, which is supported by the similar shape for the human separase–securin complex18. The information obtained from this structure may also be useful for deciphering the molecular mechanisms of other separase–securin complexes.

Methods

No statistical methods were used to predetermine sample size. The experiments were not randomized, and investigators were not blinded to allocation during experiments and outcome assessment.

Protein expression and purification

Separase and securin were co-expressed in insect cell system using Multibac technology (Geneva Biotech)31. Separase and securin gene fragments were cloned into the pFL dual plasmid through its PH and p10 promoters, respectively. The construct used for crystallization contained residues 51–1630 or 71–1630 of S. cerevisiae separase and residues 258–373 of securin. A hexa-histidine tag was added to the N terminus of separase, whereas securin did not carry a tag. Bacmids were generated by transforming constructs to DH10EMBacY competent cells (Geneva Biotech) and isolated by Bac-to-Bac protocol (ThermoFisher). Bacmids were further transfected into Sf9 cells using Cellfectin II (ThermoFisher) following the manufacturer’s instructions to produce P1 baculoviruses. P2 viruses were obtained by amplifying P1 viruses in Sf9 once and were used for large-scale protein expression.

High5 cells were grown in Fernbach flasks (PYREX) in ESF 921 medium (Expression Systems) by shaking at 120 r.p.m. at 27 °C until the density reaches 2 × 106 cells ml−1. Cells were infected by P2 virus and collected 48 h after infection by centrifugation at 500g. The harvested cells were resuspended in lysis buffer containing 50 mM phosphate, pH 7.6, 500 mM NaCl, 5% (v/v) glycerol, 20 mM imidazole and protease inhibitor cocktail (Sigma) and lysed by sonication. Cell lysates were centrifuged at 24,000g for 40 min at 4 °C before incubating with nickel beads (Qiagen). After 1 h, beads were transferred to a gravity flow column (Biorad) and washed extensively with lysis buffer. Protein was eluted with a buffer containing 50 mM phosphate (pH 7.6), 500 mM NaCl, 5% (v/v) glycerol and 500 mM imidazole. Protein eluate was further purified by gel filtration using Sephacryl S-300 column (GE Healthcare) equilibrated in a buffer containing 5 mM HEPES, pH 7.6, 400 mM NaCl and 2 mM DTT. The protein samples were concentrated to 10 mg ml−1 and stored at −80 °C.

Protein crystallization

Crystals were grown using the sitting-drop method at 20 °C by mixing 1.0 μl protein solution at 3 mg ml−1 concentration with 1.0 μl well solution containing 0.25 M Na/K-phosphate (pH 6.0) and 14% (w/v) polyethylene glycol 3,350. Crystals appeared in 3 days and were transferred to the well solution plus 30% (v/v) glycerol as cryoprotectant before being flash-frozen in liquid nitrogen for data collection at 100 K. For phasing, a gold derivative was obtained by soaking crystals in mother liquor with 2 mM KAu(CN)2 for 5 h and transferring to cryoprotectant for 2 min. The crystals belong to space group P3221 with one separase–securin complex in the asymmetric unit.

Data collection and structure determination

X-ray diffraction datasets up to 3.0 Å resolution were collected at beamline 24-ID-E of the Advanced Photon Source (APS). The data were processed with the HKL package32. The structure was solved using the single isomorphous replacement with anomalous scattering method with the gold derivative with AutoSharp33, which located 10 gold sites. The atomic model was built manually with Coot34 and refined with PHENIX35. The crystallographic information is summarized in Extended Data Table 1.

From the same crystallization condition, crystals belonging to space group C2 were also obtained, with unit cell parameters of a = 238.4 Å, b = 89.5 Å, c = 119.0 Å and β = 105.2°. There is one separase–securin complex in the asymmetric unit. The best diffraction dataset that we could collect on this crystal form extended to 3.7 Å resolution. The structure could be readily solved with the molecular replacement method using the structure in P3221 as the model, and no significant structural differences were observed.

X-ray diffraction datasets for the crystal containing residues 71–1630 of separase were collected at beamline 24-ID-C of the Advanced Photon Source (APS). The structure was refined at 2.6 Å resolution, and the crystallographic information is summarized in Extended Data Table 1.

For the 2.6 Å structure, 94.8% of the residues are in the favoured region of the Ramachandran plot. For the 3.0 Å structure, 96.9% of the residues are in the favoured region of the Ramachandran plot. No residues are in the disallowed region in either structure.

Mutagenesis studies

Site-specific and deletion mutations were introduced with the QuikChange kit (Agilent) and sequenced for confirmation. The mutants were expressed in insect cells following the same protocol as that for the wild-type protein.

Data availability

The atomic coordinates and the diffraction datasets have been deposited at the Protein Data Bank with accession codes 5U1S and 5U1T. All other data are available from the corresponding author upon reasonable request.