The AAA+ ATPase TRIP13 remodels HORMA domains through N‐terminal engagement and unfolding

Reconstitution of a TRIP13‐substrate complex

To better understand the molecular basis for MAD2 conformational conversion by TRIP13, we first sought to reconstitute the TRIP13 hexamer and its complex with p31^comet and MAD2. Purified human TRIP13 adopts a mixture of oligomeric states in solution including monomers, dimers, and trimers, forming hexamers only in the presence of ATP or a non‐hydrolyzable analog, ATP‐γS (Appendix Fig S1A). In keeping with its expected ability to bind but not hydrolyze ATP, a Walker B motif mutant (E253Q) forms stable hexamers in the presence of either ATP or ATP‐γS (Appendix Fig S1C). We mixed Homo sapiens TRIP13^E253Q·ATP, p31^comet, and MAD2, and purified a stable complex of the three proteins (Fig 1B and C). Analysis of the complex by size‐exclusion chromatography/multi‐angle light scattering (SEC‐MALS) reveals a 6:1:1 stoichiometry in the TRIP13:p31^comet:MAD2 complex, as we observed previously with an equivalent complex of M. musculus proteins (Ye et al, 2015). Thus, a TRIP13 hexamer interacts strongly with a single p31^comet:MAD2 substrate complex.

Structure of TRIP13

We crystallized full‐length H. sapiens TRIP13^E253Q and determined its structure in the absence of nucleotides to a resolution of 3.02 Å and in the presence of ATP to a resolution of 2.54 Å (Fig 1D–F and Table EV1). Crystallization trials with the TRIP13^E253Q:p31^comet:MAD2 complex yielded only crystals of TRIP13^E253Q alone, precluding high‐resolution analysis of this complex (see Materials and Methods). The TRIP13 monomer closely resembles Caenorhabditis elegans PCH‐2 (Ye et al, 2015), with a small N‐terminal domain (NTD; residues 1–100) putatively involved in substrate recognition, followed by a AAA+ ATPase region with two subdomains, large and small, that together form the ATP‐binding site (Fig EV1A–C). The large AAA domain contains two loops that extend into the hexamer pore, referred to as pore loop 1 (residues ~213–227) and pore loop 2 (residues ~260–272) (Fig EV1B). Pore loop 1 contains a highly conserved tryptophan, W221, which we previously showed is critical for MAD2 conformational conversion and proper coupling of ATP hydrolysis to substrate engagement (Ye et al, 2015). In the ATP‐bound structure of TRIP13^E253Q, we observe electron density for bound nucleotide in a conserved cleft between the large and small AAA subdomains. In our structure, the γ‐phosphate of ATP is surrounded by residues K185/T186 of the Walker A motif, D252/Q253 (E253 in wild‐type TRIP13) of the Walker B motif, and the Sensor I residue N300 (Fig EV1D). The Sensor II arginine, R386, is positioned close to the ATP α‐phosphate, but not within hydrogen‐bonding distance. The arginine‐finger residue R312 of the neighboring subunit is oriented away from ATP, with its guanidinium group ~17 Å from the ATP γ‐phosphate. Despite the presence of Mg²⁺ in the crystallization solution, we do not observe a bound Mg²⁺ ion in the active site.

In both the presence and absence of ATP, crystals of TRIP13^E253Q form in space group P6₅ with one copy of TRIP13 per asymmetric unit. Adjacent TRIP13 protomers pack equivalently to C. elegans PCH‐2 and related ATPases (Fig EV2A and B), but form a helical filament instead of a closed hexamer (Fig 1E). Similar helical filaments with P6₅ crystallographic symmetry have been observed for several related AAA+ protein remodelers that nonetheless function as hexamers in solution, including Vps4 (Scott et al, 2005; Inoue et al, 2008; Monroe et al, 2014; Caillat et al, 2015), fidgetin/spastin (Roll‐Mecak & Vale, 2008; Taylor et al, 2012; Peng et al, 2013), and others. In our previous structure of a closed hexamer of C. elegans PCH‐2, four of the six subunits adopt the “closed” conformation seen in our TRIP13 structure, in which the large and small AAA domains are closely juxtaposed, while two subunits adopt an “open” conformation with the small AAA subdomain rotated ~70° away from the large subdomain (Ye et al, 2015). By overlaying two sets of adjacent TRIP13 protomers (assembled from crystallographic symmetry in the P6₅ filament) onto the two “half‐hexamers” of C. elegans PCH‐2, we could roughly model a closed TRIP13 hexamer (Figs 1F and EV2C).

Architecture and dynamics of a TRIP13:p31^comet:MAD2 complex

We next investigated the overall architecture and dynamics of the TRIP13^E253Q:p31^comet:MAD2 complex using XLMS and hydrogen–deuterium exchange mass spectrometry (HDX‐MS). XLMS uses a bifunctional crosslinker to identify pairs of lysine side chains that are within 15–30 Å in a native complex (Herzog et al, 2012), and the mapped crosslinks can be considered as a set of distance restraints for assembling a rough physical model of the complex. After incubation of the TRIP13^E253Q:p31^comet:MAD2 complex with the crosslinker BS³ (Appendix Fig S2A), we detected crosslinks between TRIP13 and both p31^comet and MAD2 (Fig 2A, and Tables EV2 and EV3). In related AAA+ ATPases including NSF and p97/Cdc48, the NTD is responsible for substrate recognition, and consistent with this, we found that three lysine residues in the TRIP13 NTD (K34, K35, and K44) crosslinked to p31^comet (Fig 2A and D). These residues formed crosslinks with two p31^comet residues in a disordered loop (K100 and K110) and one, K229, whose mutation to alanine has been demonstrated to disrupt TRIP13 binding both in vitro (Ye et al, 2015) and in cells (Ma & Poon, 2016) (Fig 2A and C). Thus, our crosslinking data are consistent with prior findings and indicate that p31^comet is recognized directly by the TRIP13 NTD.

While p31^comet crosslinked exclusively to the TRIP13 NTD, MAD2 formed crosslinks with both the TRIP13 NTD and three residues in pore loop 1 (referred to hereafter simply as the “pore loop”): K220, K227, and K231 (Fig 2A, C, and D). The two MAD2 residues that crosslink to the TRIP13 pore loop, K111 and K122, are located on the same face of MAD2 (Fig 2C), suggesting that this surface is positioned close to the TRIP13 hexamer pore in the ternary complex. Given the known structure of the p31^comet:MAD2 complex (Yang et al, 2007) and our model of the TRIP13 hexamer, we could use these data to roughly model a ternary TRIP13:p31^comet:MAD2 complex (Fig 2F). In this model, initial recognition of p31^comet by the TRIP13 NTD is followed by a rotation of the NTD to enable MAD2 to interact with the TRIP13 pore loops. While our crystal structure shows the TRIP13 NTDs docked against the outer face of the AAA+ domains, the NTDs of related ATPases like NSF are highly mobile and can swing “upward” upon substrate binding to enable engagement by the AAA+ domain pore loops (Zhao et al, 2015). Like NSF, it is possible that multiple TRIP13 NTDs swing upward to engage MAD2 once it is bound; this could explain the direct crosslinks we observe between the TRIP13 NTD and other surfaces of MAD2 (Fig 2A and D).

We next performed HDX‐MS, which provides information on solvent accessibility and dynamics by measuring exchange of main‐chain amide hydrogen atoms with solvent deuterium atoms (Hoofnagle et al, 2003; Skinner et al, 2012). We compared the TRIP13^E253Q:p31^comet:MAD2 complex to each individual protein, as well as binary TRIP13^E253Q:p31^comet and p31^comet:MAD2 complexes (Figs 2B–E and EV3, and Appendix Fig S2). While we were unable to examine changes in MAD2 between different states due to poor sequence coverage (data not shown), we detected changes in both TRIP13^E253Q and p31^comet upon complex formation that provide information about how these two proteins interact. We previously showed that p31^comet residues P228/K229 and the loop encompassing residues 157–162 constitute a conserved TRIP13‐binding site (Ye et al, 2015). Our HDX‐MS data show that p31^comet residues 156–165 are significantly protected from hydrogen–deuterium exchange upon TRIP13 binding, and even more protected in the ternary TRIP13:p31^comet:MAD2 complex (Fig 2B). Peptides encompassing the established MAD2‐binding site of p31^comet were protected upon binding of MAD2, and were also further protected in the ternary TRIP13^E253Q:p31^comet:MAD2 complex (not shown).

In TRIP13^E253Q, two regions were strongly protected upon binding either p31^comet or the p31^comet:MAD2 complex: residues 61–74 and 95–112 in the protein's NTD, and several peptides in the TRIP13 pore loops. In our structure of TRIP13, residues 61–74 reside in the NTD adjacent to residues K34, K35, and K44 (Fig 2E), supporting a role for this surface as the p31^comet binding site. TRIP13 residues 95–112 constitute the NTD‐ATPase domain linker, and protection of this region upon p31^comet binding (Fig EV3) suggests either decreased mobility or direct protection from solvent upon substrate binding. The TRIP13 95–112 peptide also shows significant bimodality in its HD exchange profile in all samples (Fig EV3 and Appendix Fig S2C). A bimodal HD exchange profile suggests that this region can adopt at least two differently solvent‐accessible conformations in solution, supporting the idea of TRIP13 NTD mobility.

The TRIP13 pore loops were also significantly protected from HD exchange upon association with either p31^comet or the p31^comet:MAD2 complex (Figs 2B and EV3). Two peptides from this loop (residues 218–228 and 222–232) showed both significant bimodality in their HD exchange profile and significant protection upon substrate binding, even in the presence of p31^comet alone. These data suggest that the pore loops can adopt multiple conformations in solution that differ in their solvent accessibility and that p31^comet, despite its inability to engage and activate TRIP13's ATPase activity (Ye et al, 2015), nonetheless affects the structure of TRIP13 upon binding. We propose that p31^comet binding may “prime” TRIP13 for proper MAD2 engagement and remodeling, subtly affecting the overall TRIP13 hexamer structure, and imparting significant additional order onto the pore loops in particular.

TRIP13 pore loops directly engage the disordered MAD2 N‐terminus

Our XLMS data show that MAD2 residues K111 and K122 are positioned near the TRIP13 pore loops in the ternary complex. MAD2 possesses a short 10‐ to 15‐amino acid disordered region at its N‐terminus that extends from the same surface as K111 and K122 in the C‐MAD2 conformation (Fig 2C). We therefore tested whether the TRIP13 pore loops directly interact with the MAD2 N‐terminus using a cysteine crosslinking assay similar to one previously described for the AAA+ ATPase Vps4 (Fig 3A) (Yang et al, 2015). We engineered MAD2 and TRIP13 constructs lacking solvent‐exposed cysteines (MAD2^ΔCys and TRIP13^ΔCys), and re‐introduced single cysteines in the TRIP13 pore loop (W221C) and in the MAD2 N‐terminus (S6C), αA helix (A17C, S25C, or S40C), or at native internal cysteine sites (C79, C106) (Fig 3B). We mixed TRIP13^W221C and TRIP13^ΔCys at a 1:8 molar ratio in the presence of ATP or the non‐hydrolyzable analog ATP‐γS to generate TRIP13 hexamers with ≤ 1 pore loop cysteine and then reduced these cysteines with DTT. We then activated cysteines in MAD2 using DTNB (Ellman's reagent), incubated with p31^comet and pre‐assembled TRIP13 hexamers, and then detected crosslinks by non‐reducing SDS–PAGE (Fig 3C). We found that MAD2^S6C, with a cysteine in the disordered N‐terminus, robustly formed crosslinks with TRIP13 in the presence of either ATP or ATP‐γS, indicating that the MAD2 N‐terminus is directly engaged by the TRIP13 pore loops. MAD2 mutants with internal cysteines all showed crosslink formation in the presence of ATP, but not ATP‐γS, indicating that ATP hydrolysis is required to processively unfold MAD2 (Fig 3C). In all cases, MAD2‐TRIP13 crosslinks depended on cysteines in both proteins, required p31^comet, and disappeared upon the addition of reducing agent (Fig 3D and Appendix Fig S3). These data show that TRIP13 interacts with the disordered MAD2 N‐terminus upon initial association with p31^comet:MAD2, and can then unfold at least 106 residues of MAD2 (205 residues total) in the presence of ATP.

TRIP13 requires the disordered MAD2 N‐terminus for conformational conversion

We have previously shown that TRIP13 and p31^comet can together convert C‐MAD2 to O‐MAD2 (Ye et al, 2015), explaining their ability to disassemble MCC and inactivate the SAC (Teichner et al, 2011; Westhorpe et al, 2011; Eytan et al, 2014; Wang et al, 2014; Miniowitz‐Shemtov et al, 2015). To test whether TRIP13 engagement of the MAD2 N‐terminus is required for conformational conversion, we designed a series of N‐terminal truncations of MAD2 (Fig 4A). In structures of C‐MAD2, the first residue of helix αA is L13 (Luo et al, 2002; Sironi et al, 2002); in O‐MAD2, residues 11–15 are in the β1 strand, and αA begins at residue A17 (Luo et al, 2000; Mapelli et al, 2007). We designed truncations of M. musculus MAD2 starting at residues M1 (full‐length; FL), S6 (ΔN5), I11 (ΔN10), and S16 (ΔN15), and successfully purified all but ΔN10 (see Materials and Methods). All MAD2 constructs also contained the R133A mutation, which disrupts MAD2 homodimer formation and enables separation of O‐MAD2 and C‐MAD2 populations by anion‐exchange chromatography (see, e.g., Appendix Fig S7) (Sironi et al, 2001; Luo et al, 2004). When expressed in Escherichia coli, FL and ΔN5‐MAD2 adopt both O‐MAD2 and C‐MAD2 conformations, while ΔN15‐MAD2 adopts solely the C‐MAD2 conformation as previously reported (Mapelli et al, 2007), likely because this construct lacks the β1 strand needed to stabilize the O‐MAD2 conformation. Both ΔN5‐MAD2 and ΔN15‐MAD2 strongly bind a closure motif‐containing CDC20^127‐147 peptide (Fig 4B) [again, as previously reported for ΔN15‐MAD2; (Mapelli et al, 2006)] and bind p31^comet (Fig 4C), and equivalent H. sapiens MAD2 truncations behave identically (Appendix Fig S4B and C). Thus, removing up to 15 residues from the MAD2 N‐terminus does not affect the protein's ability to bind a closure motif peptide and adopt the closed conformation in vitro.

We previously showed that p31^comet and MAD2 together stimulate the ATPase activity of M. musculus TRIP13 and that mutation of the conserved pore loop tryptophan residue W221 to alanine uncouples ATPase stimulation from substrate binding (Ye et al, 2015). We next tested the ability of our MAD2 ΔN mutants to stimulate TRIP13 and found that, while FL‐MAD2 stimulated TRIP13's ATPase activity roughly threefold in the presence of p31^comet, neither ΔN5‐MAD2 nor ΔN15‐MAD2 stimulated TRIP13 (Fig 4D). Equivalent assays with human proteins again showed identical results (Appendix Fig S4D). These data indicate that TRIP13's ATPase activity is tightly coupled to engagement of the MAD2 N‐terminal tail by the TRIP13 pore loops. Finally, we directly tested for TRIP13‐mediated conformational conversion of MAD2 using anion‐exchange chromatography as previously described (Ye et al, 2015). We incubated purified C‐MAD2 FL, ΔN5, and ΔN15 with p31^comet, TRIP13, and ATP and then separated the reaction products on a Mono‐Q ion‐exchange column. While FL‐MAD2 was robustly converted to O‐MAD2, we detected no conversion of either ΔN5‐MAD2 or ΔN15‐MAD2 (Fig 4E). Thus, the disordered MAD2 N‐terminus is required to stimulate the ATPase activity of TRIP13, and for TRIP13‐mediated MAD2 conformational conversion.

The spindle assembly checkpoint is compromised in cells containing ΔN‐MAD2

RNAi knockdown of TRIP13 in cells, or depletion of TRIP13 from cell extracts, has been shown to cause defects in MCC disassembly, compromising inactivation of the SAC and delaying anaphase onset (Wang et al, 2014). Conversely, deleting TRIP13 entirely in cells compromises activation of the SAC, likely by eliminating the pool of O‐MAD2 necessary for recruitment to kinetochores and assembly into the MCC (Ma & Poon, 2016). As ΔN‐MAD2 cannot be properly engaged by TRIP13 in vitro, its effects may mimic that of TRIP13 knockdown or deletion. To explore this idea, we inhibited endogenous MAD2 expression in DLD1 cells by RNAi targeting the 3′ UTR and then rescued with either full‐length or ΔN‐MAD2 constructs (Fig 5A and B). While ΔN5‐ and ΔN10‐MAD2 expressed at near‐wild‐type levels in these cells (Appendix Fig S5), ΔN15‐MAD2 showed poor expression and was not examined further (not shown). We first examined the O‐MAD2/C‐MAD2 ratio in these cells. Similar to prior findings in HeLa cells (Luo et al, 2004), we found that FL‐MAD2 overwhelmingly adopted the open conformation in unsynchronized DLD1 cells (Fig 5C). In contrast, ΔN5‐ and ΔN10‐MAD2 were ~50% and nearly 100% closed, respectively (Fig 5C), in close agreement with the phenotype of TRIP13‐deleted cells (Ma & Poon, 2016). These data support the idea that truncating the MAD2 N‐terminus compromises TRIP13‐mediated conformational conversion, and suggest that SAC activation may be compromised in these cells as well.

We next examined mitotic timing in cells expressing ΔN‐MAD2. RNAi knockdown of MAD2 causes a significant shortening of the time between nuclear envelope breakdown (NEBD) and anaphase onset, indicating that the SAC cannot be activated in the absence of MAD2 (Fig 5D). SAC activation is rescued by expression of FL‐MAD2, but rescue with ΔN5‐ and ΔN10‐MAD2 had a more variable effect. In agreement with our ion‐exchange chromatography results showing that nearly all ΔN10‐MAD2 is in the C‐MAD2 state, these cells showed a SAC activation defect nearly as strong as that observed in the MAD2 RNAi‐alone cells. ΔN5‐MAD2, in contrast, showed wide variability in mitotic timing, with many cells showing short‐to‐normal timing, but a small percentage showing extremely delayed anaphase onset (Fig 5D). This indicates that while many cells expressing ΔN5‐MAD2 cannot activate the SAC, some may retain enough O‐MAD2 to assemble MCC and activate the SAC. Because ΔN5‐MAD2 is refractory to TRIP13‐mediated disassembly, SAC inactivation could therefore be compromised in these cells. Finally, we examined whether ΔN‐MAD2 could support SAC activation in the presence of the microtubule poison nocodazole. Nocodazole was added 1 h prior to the start of time‐lapse filming, and the percentage of mitotic cells at time zero that remained arrested in mitosis was followed for several hours. While RNAi knockdown of MAD2 caused a complete loss of SAC activation, as indicated by a quick exit from mitosis by all cells, full‐length MAD2 fully rescued SAC activation and nocodazole‐mediated mitotic arrest (Fig 5E). As in our previous assay, ΔN5‐ and ΔN10‐MAD2 had an intermediate effect, with ~60% and ~40% of cells, respectively, showing robust SAC activation (Fig 5E).

N‐terminal truncation of HORMAD1 compromises its removal from the meiotic chromosome axis

Prior to its identification as a key component of the spindle assembly checkpoint, TRIP13 was implicated in meiotic chromosome axis remodeling, specifically the removal of MAD2‐related HORMAD1 and HORMAD2 from the chromosome axis upon synaptonemal complex assembly (Wojtasz et al, 2009). While a direct interaction between these proteins and TRIP13 has not yet been demonstrated, we sought to test whether truncating their disordered N‐termini would compromise their removal from meiotic chromosomes. We designed truncations of M. musculus HORMAD1 roughly equivalent to ΔN5‐MAD2 (ΔN5‐HORMAD1) and ΔN10‐MAD2 (ΔN20‐HORMAD1; Fig 6A) and then introduced vectors encoding these proteins into testes of HORMAD1 ^−/− mice. We examined chromosome localization of eGFP‐tagged HORMAD1 constructs in late zygotene/early pachytene, when HORMAD1/2 are being removed from regions of homolog pairs that have undergone synapsis. We found that full‐length HORMAD1‐eGFP localizes to chromosome axes early in meiotic prophase, and is then efficiently removed upon homolog synapsis, as judged by a lack of co‐localization with the synaptonemal complex protein SYCP1 (Fig 6B). Truncation of the HORMAD1 N‐terminus did not affect initial localization of the protein to chromosome axes, but caused a progressive defect in synapsis‐linked removal: While full‐length HORMAD1‐eGFP was never co‐localized with SYCP1, ΔN5‐HORMAD1‐eGFP showed at least partial co‐localization with SYCP1 in 32% of cells, and ΔN20‐HORMAD1 showed co‐localization with SYCP1 in 88% of cells (Fig 6C and D). We found that bulky N‐terminal tags on HORMAD1 also significantly disrupt its removal from the chromosome axis (Appendix Fig S6); these tags may disrupt recognition and engagement by TRIP13. Overall, these results strongly support the idea that TRIP13 mediates removal of HORMAD1 from meiotic chromosomes directly, using a HORMA domain‐remodeling mechanism similar to the one we outlined for MAD2.

A conserved hydrogen‐bond network stabilizes the C‐MAD2 conformation

While our crosslinking data show that TRIP13 can unfold at least 106 residues of MAD2, our earlier measurements of ATP usage by TRIP13 suggested that on average, conformational conversion of MAD2 is accomplished using only 8–10 ATP molecules (Ye et al, 2015), likely not enough to mediate complete MAD2 unfolding. Examination of the structure of C‐MAD2 reveals that the C‐terminal safety belt β‐hairpin (strands β8′ and β8″) interacts directly with the N‐terminal αA helix (residues ~13–35) through a network of buried hydrogen bonds (Figs 7A and EV5A and B), suggesting that disruption of this network by limited N‐terminal unfolding could mediate MAD2 conformational conversion. Supporting this idea, prior work showed that mutation of MAD2 safety belt residues T188 or H191, which interact directly with S16 on the αA helix, causes MAD2 to preferentially adopt the open conformation (Yang et al, 2008). All three residues are highly conserved in MAD2 orthologs, and an equivalent buried hydrogen‐bond network is also conserved in the meiotic HORMAD proteins (Figs 7B and EV5C and D) (Kim et al, 2014).

To test the idea that disruption of the buried hydrogen‐bond network linking MAD2's N‐terminus to the safety belt promotes the open conformation, we mutated MAD2 S16 to alanine or lysine and examined the equilibrium open/closed ratio of these mutants by ion‐exchange chromatography. We purified O‐MAD2 (which in wild‐type MAD2 is stable at 4°C) for each variant and then incubated each protein for 24 h at 37°C. With wild‐type MAD2, this treatment results in near‐complete spontaneous conversion to the C‐MAD2 form (Fig 7C). Replacement of S16 with a bulky lysine (S16K) probably sterically inhibits docking of the β8′‐β8″ hairpin, and this mutant adopted exclusively the open conformation, even after 24 h at 37°C (Fig 7C). The S16A mutant, by contrast, adopted both open and closed conformations when purified from E. coli (Appendix Fig S7), but purified O‐MAD2^S16A showed no conversion to the closed conformation after 24 h at 37°C (Fig 7C). Thus, disrupting the conserved hydrogen‐bond network linking helix αA to the β8′‐β8″ hairpin favors the open conformation of MAD2, likely by destabilizing the closed conformation (Fig 7D). These data suggest that TRIP13 could act by unfolding a limited stretch of the MAD2 N‐terminus, perhaps as few as 15–20 residues, which would disrupt the αA‐β8′‐β8″ hydrogen‐bond network and mediate a closed‐to‐open conformational conversion without complete MAD2 unfolding.

Cloning and protein purification

All proteins were cloned into pET‐based vectors with N‐terminal His₆ tags cleavable by TEV protease or, in the case of MAD2 variants for biochemical assays, the SUMO isopeptidase Ulp1. Proteins were expressed in E. coli strain Rosetta2(DE3)pLysS (EMD Millipore) except for TRIP13, which was expressed in E. coli LOBSTR (Kerafast, Inc.) (Andersen et al, 2013). Proteins were purified by Ni²⁺ affinity (Ni‐NTA; Qiagen) and ion‐exchange (HiTrap SP HP or HiTrap Q HP; GE Healthcare) chromatography, His₆ tags were cleaved by incubation with TEV protease (Kapust et al, 2001) at 4°C overnight, and His₆‐SUMO tags were cleaved by incubation with S. cerevisiae Ulp1 at 4°C overnight. Proteins were concentrated and stored at 4°C for crystallization, or at −80°C for biochemical assays.

Crystallization and structure determination

For crystallization, TRIP13^E253Q was concentrated to 15 mg/ml and exchanged into a buffer containing 25 mM Tris–HCl pH 7.5, 0.2 M NaCl, 5 mM MgCl₂, and 1 mM TCEP. For crystallization without ATP, TRIP13^E253Q was mixed 1:1 in hanging drop format with well solution containing 0.1 M Tris–HCl pH 8.5, 0.8 M LiCl, and 4% PEG 3350. Crystals (~90 × 90 × 50 μm) were cryoprotected by addition of 20% glycerol and flash‐frozen in liquid nitrogen. For ATP‐bound TRIP13^E253Q, 1 mM ATP was added to the protein solution; then protein was mixed 1:1 in hanging drop format with well solution containing 0.1 M Tris–HCl pH 8.5, 0.2 M KCl, 10% pentaerythritol propoxylate. Crystals (~100 × 100 × 60 μm) were cryoprotected by addition of 10% glycerol and 12% pentaerythritol propoxylate (22% total) and flash‐frozen in liquid nitrogen.

X‐ray diffraction datasets for TRIP13^E253Q apo and TRIP13^E253Q·ATP were collected on NE‐CAT beamline 24ID‐E at the Advanced Photon Source (Argonne National Laboratory, Argonne IL; see Appendix for support statement). Phasing was performed by molecular replacement with PHASER (McCoy et al, 2007), using a structure of C. elegans PCH‐2 (PDB ID 4XGU) (Ye et al, 2015) with most side chains removed. Initial models were iteratively rebuilt in COOT (Emsley et al, 2010) and refined using phenix.refine (Adams et al, 2010) with a consistent free‐R set across multiple structures (Table EV1).

TRIP13^E253Q:p31^comet:MAD2 complex assembly and crosslinking mass spectrometry

For assembly of the TRIP13^E253Q:p31^comet:MAD2 complex, proteins were mixed in a final molar ratio of 2:1:1 (three copies of p31^comet:MAD2 per TRIP13^E253Q hexamer) in a buffer containing 20 mM HEPES pH 7.5, 0.3 M NaCl, 5 mM MgCl₂, 5% glycerol, and 1 mM ATP and then incubated for 60 min at 4°C. Complexes were injected onto a Superdex 200 Increase 10/300 GL size‐exclusion column (GE Life Sciences) and fractions representing the TRIP13^E253Q hexamer plus associated p31^comet and MAD2 were pooled and concentrated to 1 mg/ml. Crosslinking was performed using isotopically coded D₀/D₁₂ BS³ (bis‐sulfosuccinimidylsuberate; Creative Molecules) for 30 min at 30°C, followed by quenching in 100 mM NH₄HCO₃. Crosslinked proteins were digested with Lys‐C and trypsin, purified by reversed‐phase chromatography, and analyzed by tandem mass spectrometry (Orbitrap Elite, Thermo Scientific) (Herzog et al, 2012). Fragment ion spectra were searched and crosslinks identified by the dedicated software program xQuest (Rinner et al, 2008; Walzthoeni et al, 2015).

Hydrogen–deuterium exchange mass spectrometry

HD exchange experiments were conducted with a Waters Synapt G2S system; 5 μl of samples containing 10 μM individual proteins or complexes in exchange buffer (300 mM NaCl, 20 mM Tris pH 7.5, 10% glycerol, 1 mM DTT) was mixed with 55 μl of the same buffer made with D₂O for several deuteration times (0, 1, 2, 5, 10 min) at 15°C. The exchange was quenched for 2 min at 1°C with an equal volume of quench buffer (3 M guanidine HCl, 0.1% formic acid). Proteins were cleaved with pepsin and separated by reverse‐phase chromatography and then directed into a Waters SYNAPT G2s quadrupole time‐of‐flight (qTOF) mass spectrometer. Peptides were identified using PLGS version 2.5 (Waters, Inc.), and relative deuterium uptake was calculated using DynamX version 2.0 (Waters Corp.).

ATPase assays

ATPase reactions contained assay buffer (25 mM Tris–HCl at pH 7.5, 200 mM NaCl, 10 mM MgCl₂, 1 mM DTT, 5% glycerol) plus 250 μM ATP, and were incubated 30 min at 37°C. Reactions were stopped and triplicate samples were assayed using the ADP‐Glo kinase assay kit (Promega) in 384‐well microplates using a TECAN (Mannedorf, Switzerland) Infinite M1000 spectrophotometer. Most assays were performed with 100 or 200 nM TRIP13 (monomer concentration); below this concentration ATPase activity was highly inconsistent, potentially due to dissociation of the TRIP13 hexamer. For most assays, p31^comet and MAD2 were added in excess, at 2 and 5 μM, respectively. Data were converted to K _cat using a standard curve for ADP, and fit using PRISM version 7 (GraphPad Software, La Jolla, CA, USA).

Protein interaction assays

For MAD2‐p31^comet binding assays, equimolar amounts of proteins were incubated 60 min at 20°C, bound to Ni‐NTA resin (Qiagen), and then analyzed by SDS–PAGE. For fluorescence polarization assays, N‐terminal FITC‐labeled CDC20 closure motif peptide (residues 127–147; EAKILRLSGKPQNAPEGYQNR) was resuspended in binding buffer (25 mM Tris–HCl pH 7.5, 200 mM NaCl, 5% glycerol, 1 mM DTT, 0.1% NP‐40). Fifty microliter reactions containing 50 nM peptide plus 40 nM–40 μM MAD2 (FL or N‐terminal truncations) were incubated 30 min at room temperature, and then, fluorescence polarization was read in 384‐well plates using a TECAN Infinite M1000 PRO fluorescence plate reader. All binding curves were done in triplicate. Binding data were analyzed with GraphPad Prism v. 7 using a single‐site binding model.

Cysteine crosslinking

Cysteine crosslinking assays were performed essentially as described (Yang et al, 2015). Cysteines in MAD2 were activated by addition of 5,5′‐dithiobis‐(2‐nitrobenzoic acid) (DTNB) and then buffer‐exchanged into DTT‐ and DTNB‐free crosslinking buffer. TRIP13^W221C and TRIP13^ΔCys were mixed in a 1:8 molar ratio in the presence of 1 mM ATP and 5 mM DTT and then buffer‐exchanged to DTT‐free buffer. p31^comet was prepared as TRIP13. TRIP13 (20 μM monomers), MAD2 (5 μM), and p31^comet (5 μM) were mixed in the presence of 1 mM ATP or ATP‐γS for 1 min at 30°C, and then, reactions were quenched and analyzed by non‐reducing PAGE and Western blotting.

Ion‐exchange chromatography and MAD2 conversion assays

Examination of MAD2 conformation by ion‐exchange chromatography, and MAD2 conformational conversion assays, was carried out largely as described (Ye et al, 2015).

Time‐lapse imaging

Full‐length or N‐terminally truncated MAD2 was incorporated into parental Flp‐In TRex‐DLD‐1 parental cells that stably express mRFP‐tagged histone H2B (H2B–mRFP) (Holland et al, 2010; Han et al, 2013) using FRT/Flp‐mediated recombination (Tighe et al, 2004), and expression induced with 1 μg/ml doxycycline. siRNAs were directed against the 3′ untranslated region of Mad2, not present in Flp‐In constructs.

To determine unperturbed mitotic timing, cells were seeded onto CELLSTAR μClear 96‐well plate (Greiner bio‐one) with 50 nM of oligonucleotides using Lipofectamine RNAiMAX (Thermo Fisher); 24 h after transfection, doxycycline was added to express MAD2 for 24 h before analyzing by time‐lapse microscopy. Cells were maintained at 37°C and 5% CO₂ and images were collected using a CQ1 confocal image cytometer (Yokogawa) with a 40× objective at 10‐min time intervals. Movies were assembled and analyzed using ImageJ (National Institutes of Health) software. To measure mitotic slippage, nocodazole was added 1 h before filming.

Mouse spermatocyte protein localization

Proteins were expressed in spermatocytes by injection and electroporation of CMV‐driven constructs into testes of live juvenile CD1 mice (13dpp) according to published protocols (Shoji et al, 2005; Shibuya et al, 2014). Spermatocytes were collected 24 h after electroporation, and nuclear surface spreads were carried out as described previously (Peters et al, 1997).

Data availability

The crystal structures reported here are available at the Protein Data Bank (http://www.rcsb.org) under the accession codes 5VQ9 (TRIP13^E253Q Apo; https://doi.org/10.2210/pdb5vq9/pdb) and 5VQA (TRIP13^E253Q·ATP; https://doi.org/10.2210/pdb5vqa/pdb). Diffraction datasets for both structures are available at the SBGrid Data Bank (https://data.sbgrid.org) under the accession codes 409 (TRIP13^E253Q Apo; https://doi.org/10.15785/sbgrid/409) and 410 (TRIP13^E253Q·ATP; https://doi.org/10.15785/sbgrid/410).