BUBR1 and Closed MAD2 (C-MAD2) Interact Directly to Assemble a Functional Mitotic Checkpoint Complex

The mitotic checkpoint maintains genomic stability by ensuring that chromosomes are accurately segregated during mitosis. When the checkpoint is activated, the mitotic checkpoint complex (MCC), assembled from BUBR1, BUB3, CDC20, and MAD2, directly binds and inhibits the anaphase-promoting complex/cyclosome (APC/C) until all chromosomes are properly attached and aligned. The mechanisms underlying MCC assembly and MCC-APC/C interaction are not well characterized. Here, we show that a novel interaction between BUBR1 and closed MAD2 (C-MAD2) is essential for MCC-mediated inhibition of APC/C. Intriguingly, Arg133 and Gln134 in C-MAD2 are required for BUBR1 interaction. The same residues are also critical for MAD2 dimerization and MAD2 binding to p31comet, a mitotic checkpoint silencing protein. Along with previously characterized BUBR1-CDC20 and C-MAD2-CDC20 interactions, our results underscore the integrity of the MCC for its activity and suggest the fundamental importance of the MAD2 αC helix in modulating mitotic checkpoint activation and silencing.

The mitotic checkpoint (or spindle assembly checkpoint) prevents premature sister chromatid segregation by delaying the metaphase-to-anaphase transition in response to defective microtubule attachment at kinetochores. This failsafe mechanism is specified by a group of evolutionarily conserved proteins that includes MAD1, MAD2, BUBR1(MAD3), BUB1, BUB3, and MPS1 (1). The target of the mitotic checkpoint is anaphase-promoting complex/cyclosome (APC/C), 3 the multisubunit E3 ubiquitin ligase that promotes anaphase onset through ubiquitylating securin and cyclin B for degradation (2). In APC/C CDC20 , CDC20 is the activator subunit in the APC/C holoenzyme that also plays major roles in substrate recruitment (3).
Kinetochores lacking proper microtubule attachment or tension emit a diffusible "wait anaphase" signal that has been shown to depend primarily on MAD2 (1). MAD2 can assume two topologically distinct states called open (O) or closed (C) conformers (4,5). The two conformers exhibit distinct structures at the N and C termini, particularly in the arrangement and position of two C-terminal ␤-strands and connecting loops. A closed "safety belt"-like loop is present only in C-MAD2 where either MAD1 or CDC20 can bind to form "liganded" C-MAD2. Although the predominant conformer in cells is O-MAD2, during mitosis the MAD1⅐C-MAD2 complex localized at unattached kinetochores catalytically converts the cytosolic pool of O-MAD2 into C-MAD2 (4,5). C-MAD2 released from kinetochores can form CDC20⅐C-MAD2 complexes that may further amplify the production of C-MAD2 by converting additional molecules of O-MAD2 (6).
A MAD2 dimerization domain, mainly involving residues at its ␣C helix (122-142 residues in human MAD2), is fundamental in mediating O3 C-MAD2 conversion. Purified O-MAD2 spontaneously converts into C-MAD2 at a very slow rate (4,5); therefore recombinant MAD2 contains a mixture of O-MAD2 monomer, O:C heterodimer, and C:C homodimer (7). Acceleration of the O3 C conversion has been achieved in vitro with purified liganded C-MAD2, in the form of MAD1⅐C-MAD2 or CDC20⅐C-MAD2 complexes. The conversion depends on transient O:C heterodimerization that requires Arg 133 and Gln 134 residues (4 -7). Liganded C-MAD2 seems incapable of forming C:C homodimers due to steric clashes at the MAD2 dimerization interface (7). Interestingly, p31 comet , a negative regulator of the mitotic checkpoint, was shown to exploit the dimerization interface to block O3 C-MAD2 conversion during mitosis (8,9). Through structural mimicry, p31 comet binds to C-MAD2 at the ␣C helix thus preventing access of O-MAD2 for heterodimerization (10).
The MCC, composed of BUBR1, BUB3, MAD2, and CDC20, was isolated biochemically from HeLa cells as a factor that can bind and potently inhibit mitotic APC/C (Ͼ3000-fold higher activity over recombinant MAD2) (11). The MCC is evolutionarily conserved as homologous complexes have also been identified in Saccharomyces cerevisiae, Schizosaccharomyces pombe, Xenopus laevis, and mouse (12)(13)(14). How MCC subunits interact to form the complex and how MCC binds and inhibits APC/C are not fully comprehended (15)(16)(17). The current model on MCC composition proposes that MAD2 and the cell cycle-independent BUBR1⅐BUB3 subcomplex do not interact directly, but are brought together with CDC20 as a bridging subunit (18). Once bound to APC/C, MCC efficiently prevents APC/C from binding and ubiquitylating substrates (19). The inhibition of APC/C is thought to arise mainly from the two KEN boxes at the N terminus of BUBR1. Although the KEN box is a well characterized degron in many APC/C substrates, BUBR1 may utilize KEN boxes to inhibit APC/C acting as a pseudosubstrate (17,18,20,21). Early in vitro studies showed that high concentrations of both MAD2 and BUBR1 can inhibit the APC/C by binding to and sequestering CDC20 from the APC/C core subunits, and the formation of CDC20⅐C-MAD2 complex is still accepted as the terminal step of the mitotic checkpoint signal transduction (22)(23)(24). It is unclear how MAD2 and BUBR1 coordinate in the MCC to impart the potent APC/C inhibition activity for the whole complex.

EXPERIMENTAL PROCEDURES
DNA Constructs-Human full-length BUBR1, MAD2, and CDC20 cDNAs were amplified from a prostate cDNA library (Invitrogen) or freshly prepared reverse transcribed cDNAs provided by Dr. Douglas Leaman (University of Toledo). Fulllength cDNAs and fragments were usually cloned into pENTR-D/TOPO first and then subcloned into various destination vectors using the Gateway recombination reactions (Invitrogen). Point mutations were generated using the QuikChange Site-Directed Mutagenesis kit (Stratagene). All constructs were confirmed by DNA sequencing. The features of different protein fragments or mutants are described in supplemental Table S1.
Cell Culture, Synchronization, and Transfection-HeLaM, a subline of HeLa, was maintained in DMEM with 10% fetal bovine serum at 37°C in 5% CO 2 . To block cells in prometaphase, HeLaM cells were treated with 2.5 mM thymidine for 24 h and then directly released into medium containing 60 ng/ml nocodazole for 12 h. DNA transfection was performed using polyethylenimine (25) at a DNA:polyethylenimine ratio of 1:3 or FuGENE 6 (Roche).
Cell Lysates, Immunoblotting, Immunoprecipitation, and GST Pulldown-Cells were lysed in cell lysis buffer (1 ϫ PBS, 10% glycerol, 0.5% Nonidet P-40) supplemented with protease inhibitors (Protease Inhibitor Mixture set III, EDTA-free; Calbiochem) and phosphatase inhibitors (100 mM NaF, 1 mM Na 3 VO 4 , 60 mM ␤-glycerophosphate, 100 nM Microcystin LR). The protein concentration of the lysates was measured using the BCA Protein Assay kit (Pierce). Immunoblotting was used to probe specific proteins in the cell lysates, immunoprecipitates, or in vitro binding assays. In some experiments the blots were scanned, and the intensities of bands of interest were quantified using Kodak Molecular Imaging software. For immunoprecipitation, 200 -300 g of lysates were incubated with appropriate antibodies (0.5-1 g) at 4°C for 1 h and then mixed with protein A-agarose beads (RepliGen) for another 1 h. Immune complexes were washed four times with cell lysis buffer containing 250 mM NaCl and then subjected to SDS-PAGE separation. For GST pulldown experiments, cell lysates were added directly to glutathione-agarose (Pierce) and incubated for 1.5 h at 4°C.
APC/C Activation Assay Using Concentrated Mitotic Extracts-The extracts were prepared following Braunstein et al. (28) with minor modifications. Nocodazole-arrested HeLaM cells were harvested by pipetting, washed with ice-cold PBS, and resuspended in 75% of pellet volume of hypotonic buffer (20 mM Hepes-NaOH, pH 7.6, 5 mM KCl, 1 mM DTT) containing protease inhibitors. After repeated freeze-thawing, the cell lysates were centrifuged at 16,000 ϫ g for 1 h. The supernatants were collected, supplemented with glycerol to 10% (v/v), aliquoted, and stored in liquid nitrogen. The protein concentration of the extracts was 15-20 mg/ml. To assay for APC/C activity, 20-l reaction mixtures contained 10 l of concentrated mitotic extract, 2 l of 10ϫ degradation mixture (100 mM Tris-HCl, pH 7.6, 50 mM MgCl 2 , 10 mM DTT, 10 mg/ml ubiquitin, 100 mM phosphocreatine, 5 mM ATP, 0.1 mg/ml UbcH10), and 1 l of 20ϫ creatine phosphokinase (1 mg/ml). Recombinant proteins were added in some experiments. Reactions were incubated at 30°C, and 3-l samples were withdrawn at various times, and then rapidly quenched with SDS-PAGE sample buffer. Degradation of cyclin B and securin was followed by immunoblotting.
Protein Expression-GST-tagged BUBR1 (1-371) and CDC20 (111-138) , His 6 -tagged UbcH10, and wild-type and mutant MAD2 were expressed in Escherichia coli BL21-CodonPlus (DE3) RIPL (Stratagene) at 37°C or 25°C. His 6tagged CDC20 and GST-tagged full-length BUBR1 were expressed in Sf9 cells following the manufacturer's instructions (Invitrogen). Proteins were purified using glutathione-agarose (Pierce) or Probond nickel beads (Invitrogen). The His tag was removed from MAD2 by tobacco etch virus protease cleavage. Concentrations of recombinant proteins were determined by comparing the target band with BSA standards on Coomassie Blue-stained gels.
In Vitro O3 C-MAD2 Conversion-This was carried out based on Simonetta et al. (6) and illustrated in supplemental Fig. S3. 5 l of GSH-agarose beads were coated with 25 l of GST-CDC20 (111-138) (0.4 mg/ml) for 2 h at 4°C. The beads were then washed four times with cell lysis buffer containing 250 mM NaCl. The GST-CDC20 (111-138) -coated or mocktreated beads (5 l) were then incubated with 20 l of untagged MAD2 WT (ϳ0.1 mg/ml) for 24 h at 25°C. During the incuba-tion, the preexistent C-MAD2 in MAD2 WT is expected to be captured by GST-CDC20 (111-138) and then acts as a catalyst to convert more O-MAD2 into C-MAD2.
In Vitro Binding Assay-Recombinant proteins were incubated at 37°C in 20-l reactions in binding buffer (10 mM Tris-HCl, pH 8.0, 150 mM NaCl, 2 mM MgCl 2 , 0.5% Nonidet P-40, 5% glycerol, 100 g/ml BSA) for 1 h. The reactions were then mixed with 5 l of glutathione-agarose and gently rocked at 4°C for 1 h. Beads were washed four times with cell lysis buffer containing 250 mM NaCl before SDS-PAGE.
Yeast Two-hybrid Assay-Yeast two-hybrid was carried out as described (29). Coding sequences of two testing proteins were fused with either LexA DNA binding domain or B42 activation domain and used as bait and prey, respectively. The bait and prey DNA constructs were transformed into the yeast strain SKY48. Approximately 1 ϫ 10 5 confirmed transformants were plated on a galactose-containing leucine dropout reporter plate. Colony growth, indicating positive protein-protein interactions, was followed for 4 days.

RESULTS
C-MAD2 Delays APC/C Activation in a Chromosome-free Mitotic Cell Extract-It is widely assumed C-MAD2 amplified by unattached kinetochores constitutes a major "wait anaphase" signal that targets CDC20 to inhibit APC/C activity. To understand the effector formation step of the mitotic checkpoint mechanistically, we used a chromosome-free mitotic extract that partially recapitulates mitotic checkpoint inhibition of the APC/C (11,28). The extract retained sufficient checkpoint activity to inhibit APC/C for about 15 min before endogenous cyclin B1 and securin were abruptly degraded (Fig. 1A, control). To bypass the signal amplification step that normally requires O3 C-MAD2 conversion, the MAD2 L13A mutant, which is known to be locked in the C-conformation (10,30), was purified and added directly into the extract (see supplemental Table S1 for features of protein fragments/mutants used in this study). The addition of MAD2 L13A , at 2-and 20-fold excess over endogenous MAD2, extended APC/C inhibition by ϳ15 and 30 min, respectively (Fig. 1, A and B). Importantly, the extended APC/C inhibition correlated with prolonged MCC-APC/C association, as indicated by higher levels of BUBR1 and MAD2 in the CDC27 (an APC/C subunit) immunoprecipitates from 20ϫ MAD2 L13A supplemented extract, particularly at 30 min (Fig. 1, C and D). In the presence of 20ϫ MAD2 L13A , cyclin B1, and securin, degradation was observed after 60 min (Fig. 1, A and B), coinciding with Ͼ50% of

JOURNAL OF BIOLOGICAL CHEMISTRY 21175
BUBR1 dissociating from the APC/C (Fig. 1, C and D). Because the MAD2 level in the CDC27 immunoprecipitates was still high at 60 min, the result suggested that simultaneous binding of both BUBR1 and MAD2 is necessary for efficient APC/C inhibition. Delay of APC/C activation was also observed in extracts prepared from GST-MAD2 L13A -transfected mitotic cells compared with control extracts (Fig. 1E). The delay in activation was specific to C-MAD2 because addition of excess O-MAD2 (MAD2 ⌬C10 ) did not alter APC/C activation kinetics whereas wild-type (MAD2 WT ), a mixture of O-and C-MAD2, displayed a modest effect (Fig. 1F).

C-MAD2 with R133E/Q134A Mutations in the ␣C Helix Compromises MCC-APC/C Interaction and APC/C Inhibition-
Due to the importance of O:C-MAD2 heterodimerization in the production and amplification of C-MAD2 wait anaphase signals, a dimerization-defective MAD2 (e.g. R133E/Q134A) is unable to support the mitotic checkpoint (31). However, if MAD2 dimerization is only required for O3 C-MAD2 conversion, addition of a dimerization-defective C-MAD2 (L13A/ R133E/Q134A, MAD2 LARQ ) to the chromosome-free mitotic extract should delay APC/C activation as effectively as MAD2 L13A . Surprisingly, addition of up to 100-fold molar excess of MAD2 LARQ did not delay APC/C activation ( Fig. 2A). Similar observations were made when another C-MAD2 mutant (L13Q, MAD2 LQ ) (30) and its dimerization-defective counterpart (L13Q/R133E/Q134A, MAD2 LQRQ ) were tested in the extracts (supplemental Fig. S1A). The dimerization defect of MAD2 LARQ was confirmed in the immunoprecipitation experiments shown in Fig. 2B as it failed to bind endogenous MAD2. Interestingly, the ability of MAD2 LARQ to bind CDC20 was not compromised in vivo (Fig. 2B) or in vitro (see Fig. 4A; see also supplemental Fig. S1B for MAD2 LQRQ ). Nevertheless, MAD2 LARQ co-immunoprecipitated only half the levels of BUBR1, BUB3, and CDC27 compared with MAD2 L13A (Fig.  2B). Consistently, fractionation of mitotic lysates using sizeexclusion chromatography demonstrated that MAD2 L13A expression in HeLa cells more effectively promoted MCC-APC/C interaction than MAD2 LARQ , as evidenced by higher relative enrichment of CDC27, BUBR1, CDC20, HA-tagged and endogenous MAD2 in the fractions corresponding to the MCC⅐APC/C supracomplex (8ϳ9 ml eluates, Fig. 2, C and D). Together, these results indicate that Arg 133 and Gln 134 residues in the ␣C helix of C-MAD2 are critical for MCC formation and APC/C inhibition in addition to their involvement in O3 C-MAD2 conversion.
C-MAD2 Interacts Directly with BUBR1-To delineate further the role of C-MAD2 in MCC assembly, interactions between MCC subunits were examined using purified recom-binant proteins at levels comparable with estimated intracellular concentrations (24) (Fig. 3). We initially focused on BUBR1  in these experiments as this segment of BUBR1 has been suggested to be responsible for most, if not all, of BUBR1 checkpoint activity (17,18). GST-BUBR1 (1-371) pulled down His 6 -CDC20 (Fig. 3A, lane 2), indicating direct interaction between BUBR1 and CDC20 as reported before (22). Surprisingly, GST-BUBR1  also formed binary interactions with MAD2 WT and MAD2 L13A , even though they have long been postulated to interact indirectly through mutual associations with CDC20 (18, 24) (Fig. 3A, lanes 3 and 7). Approximately 10-fold more MAD2 L13A bound to BUBR1 (1-371) than MAD2 WT . When incubated together, CDC20 and MAD2 both associated with BUBR1   (Fig. 3A, lanes 4 -6 and 8 -10). When recombinant full-length BUBR1 was examined, it was able to bind ϳ10-fold higher amounts of MAD2 L13A compared with BUBR1   (Fig. 3B). Consistent with the in vitro binding assays, GST-BUBR1 (1-371) and endogenous BUBR1 both  (24). GST pulldowns were washed and probed alongside inputs with the indicated antibodies. B, comparison of full-length BUBR1 and BUBR1  in binding to MAD2 L13A is shown. C, HeLa cells co-transfected with GST-MAD2 W75A and either GFP or GFP-BUBR1  were arrested in mitosis with nocodazole. The lysates and GST pulldowns were probed for CDC20, MAD1, GST-MAD2, GFP, and GFP-BUBR1  . D, yeast two-hybrid assay was performed. The yeast strain SKY48 harboring different combinations of baits and preys were tested for growth on galactose-containing leucine-dropout plates. The arrowheads indicate the combinations that produced colonies. E, recombinant MAD2 WT was preincubated with GST-CDC20 (111-138) immobilized on GSH-agarose beads (ϩ) or beads alone (Ϫ) for 24 h. After incubation, GSH-agarose beads were removed by centrifugation, and the resulting supernatants were transferred to new binding reactions containing GST-BUBR1  . GST pulldowns and inputs for the new reactions were probed with the indicated antibodies.
We performed three additional experiments to verify direct BUBR1-MAD2 interaction. A MAD2 W75A mutant was reported to adopt C-conformation preferentially but is incapable of binding CDC20 due to direct involvement of Trp 75 in the CDC20-C-MAD2 interaction (7). When expressed in mitotic HeLa cells, the GST-tagged MAD2 W75A mutant bound to endogenous MAD1, suggesting adoption of C-conformation (Fig. 3C). MAD2 W75A also indeed failed to interact with endogenous CDC20; however, the mutant associated with co-transfected GFP-BUBR1   (Fig.  3C). This in vivo result supports the concept that MAD2 can interact with BUBR1 independently of CDC20. Further confirmation of the interaction was conducted using yeast two-hybrid assays (Fig. 3D). When grown on leucine dropout reporter plates, colonies appeared for the following [bait ϩ prey] combinations:  ϩ MAD2 L13A ] as well as the positive control [MAD1 ϩ MAD2 L13A ] (Fig. 3D, arrowheads), whereas vector controls and [MAD2 ⌬C10 ϩ BUBR1  ] yielded no colonies.

DISCUSSION
In this study, we presented evidence demonstrating a direct interaction between BUBR1 and C-MAD2 and showed that the interaction plays critical roles in MCC-mediated inhibition of APC/C. The results are consistent with the notion that the integ-rity of MCC is essential for its highly potent APC/C inhibitory activity compared with BUBR1 or MAD2 alone (11,19,22,24). As C-MAD2 is believed to be part of the wait anaphase signal that is generated from unattached kinetochores (1), our findings regarding the BUBR1-C-MAD2 interaction, in addition to the well characterized CDC20-C-MAD2 interaction (23), establishes a biochemical connection between a signal transducer (C-MAD2) and an effector (MCC) of the mitotic checkpoint pathway. Our results also suggest that future efforts to reconstitute MCC for functional or structural studies will likely benefit from using C-conformer locked MAD2 instead of wild-type MAD2 that exists predominantly in the O-conformation. . The R133E/Q134A mutant reduces C-MAD2 binding to BUBR1 but not CDC20. A, purified recombinant proteins were incubated at intracellular concentrations of endogenous proteins (Fig. 3 legend) for 1 h. GST pulldowns and inputs were probed with the indicated antibodies. Both GST-BUBR1  and GST-CDC20 (111-138) were detected with anti-GST antibody. B, comparison of the full-length BUBR1 binding to MAD2 L13A and MAD2 LARQ is shown. C, human BUBR1 structure and two models of MCC architecture are illustrated schematically. The two CDC20 binding domains of BUBR1 are shown as CDC20 BD1 and CDC20 BD2. One or both KEN boxes in the N-terminal region of BUBR1 (1-371 residues) are required for CDC20 binding. The GLEBS motif is required for BUBR1-BUB3 interaction. The ␣C helix of C-MAD2 is represented by a zigzag line.
The molecular architecture of a functional MCC remains to be characterized further. Two possible explanations are proposed to incorporate our results with existing models (Fig. 4C). In the simplest scheme (model on the left), CDC20 binds MAD2 at the C-terminal safety belt, stabilizing MAD2 in the C-conformation and presenting the ␣C helix of C-MAD2 for direct interaction with BUBR1. The BUBR1-C-MAD2 interaction, involving residues Arg 133 and Gln 134 of MAD2 and the N-terminal 371 residues of BUBR1, may be facilitated by BUBR1-CDC20 interactions which rely on the KEN boxes in the N-terminal region and/or the second region between residues 400 and 750 of BUBR1. Although the two CDC20 binding domains of BUBR1 have been localized to nonoverlapping BUBR1 fragments (17,18,22), whether full-length BUBR1 utilizes the two domains to bind two CDC20 molecules or different regions of a single CDC20 molecule remains to be defined. BUB3 forms a cell cycle-independent complex with BUBR1; however, the role of BUB3 in the MCC complex is unclear because the MCC isolated from S. pombe lacks detectable BUB3 (13). In the second model (model on the right in Fig. 4C), we speculate that C-MAD2 binds to BUBR1 preferentially as a homodimer, based on the fact that the R133E/Q134A mutant disrupts MAD2 dimerization and meanwhile is unable to interact efficiently with BUBR1 (Figs. 2 and 4 and supplemental Fig.  S1). However, as homodimerization was suggested to only occur between "unliganded" C-MAD2 (7), it is difficult to imagine the relationship of the C:C-MAD2 homodimer to CDC20bound C-MAD2 in a fully assembled MCC.
The proposed MCC architecture models will open avenues for future study. For example, we realize that binding of either BUBR1 or C-MAD2 to the ␣C helix of C-MAD2 will block access of O-MAD2, thus formation of the MCC will naturally prevent "chronic C-MAD2 amplification," a question raised for the model suggesting that "liganded C-MAD2" catalyzes O3 C-MAD2 conversion (6,32). Moreover, as p31 comet , the negative regulator of the checkpoint, also binds to the ␣C helix of C-MAD2 (9, 10), switching binding partners at this region of C-MAD2 seems to hold significant importance in exerting control over both activation and silencing of the mitotic checkpoint. Indeed, a recent report has suggested that p31 comet may interfere with MAD2 association with MCC, leading to MCC disassembly and checkpoint silencing (33).