The Cdc20-binding Phe Box of the Spindle Checkpoint Protein BubR1 Maintains the Mitotic Checkpoint Complex During Mitosis*

Background: The spindle checkpoint protein BubR1 inhibits the anaphase-promoting complex through binding to Cdc20. Results: We identify a new Cdc20-binding motif within BubR1 termed the Phe box. Conclusion: The Phe box maintains steady-state levels of BubR1-containing checkpoint complexes in human cells. Significance: Our study provides key insights into the homeostatic mechanisms of a key spindle checkpoint complex. The spindle checkpoint ensures accurate chromosome segregation by monitoring kinetochore-microtubule attachment. Unattached or tensionless kinetochores activate the checkpoint and enhance the production of the mitotic checkpoint complex (MCC) consisting of BubR1, Bub3, Mad2, and Cdc20. MCC is a critical checkpoint inhibitor of the anaphase-promoting complex/cyclosome, a ubiquitin ligase required for anaphase onset. The N-terminal region of BubR1 binds to both Cdc20 and Mad2, thus nucleating MCC formation. The middle region of human BubR1 (BubR1M) also interacts with Cdc20, but the nature and function of this interaction are not understood. Here we identify two critical motifs within BubR1M that contribute to Cdc20 binding and anaphase-promoting complex/cyclosome inhibition: a destruction box (D box) and a phenylalanine-containing motif termed the Phe box. A BubR1 mutant lacking these motifs is defective in MCC maintenance in mitotic human cells but is capable of supporting spindle-checkpoint function. Thus, the BubR1M-Cdc20 interaction indirectly contributes to MCC homeostasis. Its apparent dispensability in the spindle checkpoint might be due to functional duality or redundant, competing mechanisms.

The spindle checkpoint ensures accurate chromosome segregation by monitoring kinetochore-microtubule attachment. Unattached or tensionless kinetochores activate the checkpoint and enhance the production of the mitotic checkpoint complex (MCC) consisting of BubR1, Bub3, Mad2, and Cdc20. MCC is a critical checkpoint inhibitor of the anaphase-promoting complex/cyclosome, a ubiquitin ligase required for anaphase onset. The N-terminal region of BubR1 binds to both Cdc20 and Mad2, thus nucleating MCC formation. The middle region of human BubR1 (BubR1M) also interacts with Cdc20, but the nature and function of this interaction are not understood. Here we identify two critical motifs within BubR1M that contribute to Cdc20 binding and anaphase-promoting complex/cyclosome inhibition: a destruction box (D box) and a phenylalanine-containing motif termed the Phe box. A BubR1 mutant lacking these motifs is defective in MCC maintenance in mitotic human cells but is capable of supporting spindle-checkpoint function. Thus, the BubR1M-Cdc20 interaction indirectly contributes to MCC homeostasis. Its apparent dispensability in the spindle checkpoint might be due to functional duality or redundant, competing mechanisms.
Accurate chromosome segregation relies on the correct attachment of chromosomes to the mitotic spindle. During mitosis, the two opposing kinetochores of a pair of sister chromatids must attach to microtubules emanating from the two opposite spindle poles. This type of bipolar attachment, termed bi-orientation, ensures that the separated sister chromatids are segregated evenly to daughter cells (1)(2)(3). Errors in chromosome attachment can lead to chromosome missegregation and aneuploidy (4).
The spindle checkpoint monitors and promotes proper chromosome attachment during mitosis (3,5,6). Unattached or improperly attached kinetochores generate an inhibitory signal that prevents progression into anaphase by inhibiting the anaphase-promoting complex/cyclosome (APC/C) 5 (6,7). APC/C is a multisubunit E3 ubiquitin ligase whose activity is required for anaphase onset and mitotic exit (8). When bound to its mitotic activator Cdc20 (9), APC/C ubiquitinates multiple substrates, including securin and cyclin B1, targeting them for degradation by the proteasome. Degradation of securin and cyclin B1 triggers sister-chromatid separation and mitotic exit. By inhibiting APC/C Cdc20 in response to inappropriately attached kinetochores, the spindle checkpoint prevents chromosome segregation and mitotic exit until all chromosomes achieve bi-orientation.
MCC is thought to inhibit APC/C by at least two mechanisms (6,14). First, through acting as a pseudo-substrate, it competitively blocks substrate recruitment by APC/C. APC/C substrates are recognized through multiple, short degradation motifs called degrons, such as the destruction box (D box) and the KEN box (8). The KEN box is recognized by Cdc20 or its homolog Cdh1 (15)(16)(17), whereas the D box binds at the interface between Cdc20 (or Cdh1) and the core APC/C subunit APC10 (18,19). The MCC component BubR1 (or its yeast ortholog Mad3) contains two KEN boxes, which are required for checkpoint function and compete with KEN box-containing substrates for Cdc20 binding (20 -23). Second, MCC alters the mode of Cdc20 binding to APC/C: Cdc20 on its own and Cdc20 as a part of MCC bind to different locations on APC/C (17,24,25), suggesting that MCC might anchor Cdc20 to a site that is not compatible for catalysis.
Biochemical and structural studies of MCC have revealed intricate interactions among Mad2, Cdc20, and BubR1. Using a seat belt-like structural element, Mad2 in its active, closed conformation traps the Mad2-interacting motif of Cdc20 in a topological embrace (26 -28). The N-terminal region of BubR1 (BubR1N) contains two KEN boxes and interacts directly with Cdc20 through the first KEN box (KEN1) (16,23). Cdc20bound Mad2 also makes direct contact with Mad3 (and quite possibly BubR1N) (17,29), buttressing the weak Cdc20-KEN1 interaction. Thus, through simultaneously engaging both Cdc20 and Mad2, BubR1N nucleates MCC assembly, with KEN1 playing a pivotal role in this process. The second KEN box of BubR1 (KEN2) is also required for spindle checkpoint function in human cells and blocks substrate binding by APC/C (23), although the mechanism by which KEN2 achieves this task has not been established. Even more mysteriously, the middle region of BubR1 (BubR1M) has been shown to bind to Cdc20 in the absence of Mad2 (11,30). The function of the BubR1M-Cdc20 interaction in the spindle checkpoint is unknown.
To probe the function of Cdc20 binding by BubR1M in the spindle checkpoint, we first defined the mechanism by which BubR1M interacted with Cdc20. We found that human BubR1M contains two Cdc20-binding motifs: a D box and a novel Cdc20-binding motif that we termed the Phe box. The Phe box is conserved in vertebrate BubR1 proteins and is also present in Bub1 and cyclin A2. The D and Phe boxes of BubR1M bind cooperatively to Cdc20 and inhibit APC/C Cdc20 in vitro. The D box of BubR1M binds to both human Cdc20 and Cdh1. In contrast, the Phe box selectively binds to human Cdc20, but not to Cdh1, thus explaining the inability of BubR1 to inhibit APC/C Cdh1 in vitro. Strikingly, a BubR1 mutant with the Phe and D boxes deleted is deficient in MCC maintenance in human cells, indicating a contribution for BubR1M in MCC homeostasis. Consistent with previous reports (30,31), this mutant is functional in supporting the spindle checkpoint in human cells. In fact, it is slightly more active in delaying mitotic exit when cells are challenged with low concentrations of taxol. The fact that BubR1M is not required for the spindle checkpoint suggests a possible functional duality of the BubR1M-Cdc20 interaction or the existence of compensatory APC/C inhibitory mechanisms or both.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-The full-length Cdc20, Cdc20 WD40 (residues 161-477), and Mad2 L13A proteins were purified as previously described (16,32). The coding regions of various fragments of the human BubR1 middle region (BubR1M) were cloned into the pGEX-6p vector (GE Healthcare). The coding sequence of miniBubR1 gene was commercially synthesized (Genescript) and cloned into the pGEX-6p vector. The amino acid sequence of miniBubR1 is: MGDEWELSKENVQPLRQGRIMSTLQGALAQESGGAST-AELSKPTVQPWIAPPMPRAKENELQAGPWNTGRSGGG-GGSVPFSIFDEFLLSEKKNKSPPADPPRVLAQRRPLAVLKT-SESITSNEDVSPDVCDEFTGIEPLSEDAIITGFRNVTICPNP-EDTCDFARAARFVSTPFHELE. All constructs were verified by DNA sequencing. The pGEX-6p-BubR1 plasmids were transformed into the Escherichia coli strain BL21. These GST-BubR1 proteins that included a PreScission protease cleavage site between GST and BubR1 were purified with glutathione-Sepharose 4B beads (GE Healthcare) and used in GST pulldown assays. For other assays, the fusion proteins were cleaved with the PreScission protease to remove the GST moiety. The cleaved BubR1 proteins were further purified with a Superdex 200 gel filtration column (GE Healthcare) in a buffer containing 25 mM Tris (pH 8.0), 150 mM NaCl, and 1 mM DTT. The Phe box peptide of BubR1 (with the sequence of GPSVPFSIFDEFL-LSEKKNKSPPA) was chemically synthesized.
Protein Binding Assays-For binding between human BubR1M and Cdc20 (or Cdh1), purified GST-BubR1M proteins were bound to glutathione-Sepharose 4B beads (GE Healthcare). Full-length Myc-Cdc20, truncation mutants of Myc-Cdc20, and full-length Myc-Cdh1 were in vitro translated in reticulocyte lysate containing [ 35 S]methionine and incubated with the GST-BubR1M beads. The beads were washed four times with TBS containing 0.05% Tween. The proteins bound to beads were eluted by boiling in SDS loading buffer and separated on SDS-PAGE. The 35 S signal was detected and quantified using a Fuji phosphorimager. Glutathione-Sepharose beads bound to GST were used as controls. The relative binding intensity was defined as the binding intensity of each sample (subtracting background binding by GST beads) divided by 20% input.
Microscale Thermophoresis-Cdc20 WD40 was labeled with a fluorophore on cysteines using the blue-maleimide protein labeling kit (Nanotemper, Munich, Germany), following the manufacturer's instructions. The labeled protein was exchanged into the experimental buffer containing 20 mM Tris (pH 9.0) and 150 mM NaCl. Serial dilutions of BubR1 506 -583 and the Phe box peptide were made by 15 successive 1:1 dilutions of the highest titrant concentration (1 mM and 200 M, respectively) into the experimental buffer. These solutions were each diluted 1:1 with a solution of 400 nM labeled Cdc20 WD40. Thus, the final concentration of the labeled Cdc20 WD40 was 200 nM in all samples, and the highest concentrations of BubR1 506 -583 and the Phe box peptide were 100 and 500 M, respectively. The samples were placed in glass capillary tubes for ϳ30 min before the final measurements were taken in a Monolith NT.115 device (Nanotemper). For the Cdc20 WD40-Phe box peptide titration, the LED (illumination) power of the instrument was set to 90%, and the IR laser power was set to 80%. For the Cdc20 WD40-BubR1 506 -583 titration, these values were set to 50 and 80%, respectively. Once the LED was turned on, fluorescence was monitored as a function of time. There was a 5-s waiting period before the IR laser was ignited. The IR laser remained on for 30 s, followed by a 5-s monitoring of the recovery period. The resulting fluorescence time traces were analyzed to generate the isotherms. The mean fluorescence from a 0.5-s time span just after the "temperature jump" phase of the traces (F c ) was compared with that from a 2-s span just prior to the extinguishment of the IR laser (F h ) as follows.
The isotherm is thus a comparison of the concentration of the ligand versus these F n values. Subsequent analysis was performed as previously described (33). Figures from the thermophoretic experiments were generated in GUSSI. Ubiquitination Assays-APC/C ubiquitination assays were performed as previously described (34).
Mammalian Cell Culture and Transfection-HeLa TetON cells were grown in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% bovine serum (Invitrogen). Plasmid and siRNA transfections were performed using Effectene (Qiagen) and Lipofectamine RNAiMAX (Invitrogen), respectively, according to instructions from the manufacturers. The siRNAs used in this study were synthesized by Dharmacon and were used at a final concentration of 5 nM. The siRNAs had the following sequences: siBubR1, CAAGAUGGCUGUAUU-GUUU; and siLuc, UCAUUCCGGAUACUGCGAU.
For anti-Myc IP, cells were lysed in the lysis buffer (40 mM Tris-HCl, pH 7.5, 105 mM KCl, 1 mM MgCl 2 , 0.05% Nonidet P-40) freshly supplemented with 1ϫ SigmaFast protease inhibitors (Sigma), 0.5 M okadaic acid (LC Labs), 10 M cytochalasin B (Sigma), 10 units/ml TurboNuclease (Accelagen), and 1.5 mM DTT. Lysates were incubated with anti-Myc beads at 4°C for at least 2 h. The beads were washed with the high salt buffer (lysis buffer with 200 mM KCl and 0.2% Nonidet P-40) twice and with the lysis buffer three times. Bound proteins were eluted with the SDS loading buffer. For total cell lysates, cells were directly lysed with SDS loading buffer. Lysates and IP were separated on SDS-PAGE, transferred to nitrocellulose membranes (Bio-Rad), and blotted with the desired antibodies. The blots were scanned with an Odyssey Infrared Imaging System (LI-COR).
Flow Cytometry-HeLa cells were transfected with the indicated plasmids. At 8 h after plasmid transfection, they were transfected with siRNAs. At 24 h after siRNA transfection, taxol or nocodazole were added at the indicated concentrations, and samples were collected another 15 h later. Cells were washed with PBS and fixed in cold 70% ethanol. After fixation, cells were washed with PBS, incubated for at least 3 h with the MPM2 antibody (1:400 dilution in PBS containing 0.2% Triton X-100 and 3% BSA). Cells were washed with PBS and incubated for 30 min with the anti-mouse Alexa 488 secondary antibody (1:200 dilution in PBS containing 0.2% Triton X-100 and 3% BSA). Cells were washed again and resuspended in PBS containing 200 g/ml DNase-free RNase A (Qiagen) and 2 g/ml propidium iodide (Sigma). Samples were scanned with a FAC-Scalibur flow cytometer (BD Biosciences), and the data were processed with the FlowJo software.
Live Cell Microscopy-HeLa cells were transfected with the BubR1 plasmids together with a plasmid encoding eGFP in 8-well chambered coverslips (LabTek) and were transfected again with siRNAs at 8 h after plasmid transfection. At 40 h after siRNA transfection, taxol was added to a final concentration of 30 nM. Cells were then imaged with a DeltaVision system (GE Healthcare) fitted with a humidified environmental chamber at 37°C and 5% CO 2 . Differential interference contrast and eGFP images were acquired with a 40ϫ objective at 10-min intervals for 30 -36 h. The images were manually analyzed to determine the mitotic durations. Only cells expressing eGFP were included in the quantification. The t 50 is defined as the time at which 50% of the mitotic cells have undergone adaptation or apoptosis.
Immunofluorescence-HeLa cells were cultured and transfected with the desired siRNAs in 4-well chambered slides. At 42 h after transfection, nocodazole (40 nM) was added for 30 min to depolymerize microtubules. Cells were then prefixed with ice-cold 0.5% formaldehyde in PBS for 10 min on ice. The cells were briefly washed with PBS and pre-extracted in ice-cold 0.2% Triton X-100 in PBS for 5 min. Cells were postfixed with ice-cold 4% formaldehyde in PBS for 15 min. After being washed twice with PBS, cells were permeabilized with 0.2% Triton X-100 in PBS for 20 min at room temperature and incubated with mouse anti-Cdc20 (1:100), rabbit anti-BubR1 (1 g/ml), and CREST antibodies (1:1000) diluted in PBS containing 3% BSA and 0.2% Triton X-100 for 3 h at room temperature. Cells were then washed three times with 0.2% Triton X-100 in PBS and incubated with the appropriate secondary antibodies diluted in PBS with 3% BSA and 0.2% Triton X-100 for 3 h at room temperature. Cells were washed three times with 0.2% Triton X-100 in PBS and once with PBS. The slides were mounted using ProLong Gold with DAPI (Invitrogen). Images were acquired with a DeltaVision system (GE Healthcare), fitted with a 100ϫ objective. The images were deconvolved using the provided algorithm with the "conservative" setting, and projected by sum projection. Projections were quantified in ImageJ. For each cell, a mask was created using the CREST signal. This mask was then applied to the Cdc20 and BubR1 channels to measure the total intensity. Normalized intensities were defined as the ratios between the Cdc20 (or BubR1) intensities and those of CREST.

RESULTS
Human BubR1M Contains Two Cdc20-binding Motifs-Initial studies on the BubR1-Cdc20 interaction suggested the existence of multiple, independent Cdc20-binding regions (11). In addition to the N-terminal region, the middle region of BubR1 (BubR1M) containing residues 526 -700 was implicated in Cdc20 binding. This middle Cdc20-binding region of BubR1 was later further narrowed down to a fragment containing residues 490 -560 (30). We set out to further define the Cdc20binding elements in this region. We purified various BubR1M fragments as GST fusion proteins and tested their binding to in vitro translated 35 S-labeled Cdc20. As expected, a BubR1M fragment containing residues 526 -613 (BubR1 526 -613 ) bound to Cdc20 (Fig. 1A). It also bound to Cdc20 lacking the N-terminal 60 or 157 residues (Cdc20 ⌬N60 or ⌬N157). A smaller, 21-residue BubR1 fragment (BubR1 526 -546 ) also bound Cdc20 but less efficiently than BubR1 526 -613 did. These results suggest that at least two motifs of BubR1M might cooperatively bind to the C-terminal WD40 domain of Cdc20.
Sequence alignment of BubR1 proteins from different vertebrate species revealed two conserved motifs (Fig. 1B). The first motif has the consensus of FXIFDE (where X indicates any residue) and will be hereafter referred to as the Phe box because it contains two phenylalanines. The second motif is a D box, with the well established consensus RXXL. Intriguingly, two other Cdc20-binding proteins, Bub1 and cyclin A2, each contain a motif that conforms to the Phe box consensus, suggesting that the Phe box might be a widespread Cdc20-binding motif (Fig.  1C).
To test whether the Phe and D boxes mediated the binding of BubR1 526 -613 to Cdc20, we generated point mutants targeting key residues in their consensus sequences (Fig. 1D). Mutation of the Phe box (mPhe) markedly reduced the binding of BubR1 526 -613 to either the full-length Cdc20 or Cdc20 ⌬N157 (Fig. 1E). Mutation of the D box (mD) also weakened the binding between BubR1 and Cdc20, albeit to a lesser degree. These results confirm the involvement of both the Phe and D boxes in mediating BubR1M binding to the WD40 repeat domain of Cdc20.
We next measured the binding affinity between BubR1 506 -583 (which had fewer proteolytic fragments) and the WD40 domain of Cdc20 (Cdc20 WD40; residues 161-477), using microscale thermophoresis (Fig. 1F). BubR1 506 -583 (which contained both the Phe and D boxes) bound to Cdc20 WD40 with a dissocia-   Both Phe and D Boxes of BubR1 Contribute to APC/C Cdc20 Inhibition in Vitro-Having established the involvement of Phe and D boxes in mediating the binding of BubR1M to Cdc20, we next asked whether these motifs were required for BubR1-mediated inhibition of APC/C Cdc20 . Full-length BubR1-Bub3 complexes containing wild type BubR1 or the indicated mutants were assayed for their ability to inhibit APC/C Cdc20 in vitro. As expected, wild type BubR1-Bub3 inhibited the APC/ C Cdc20 activity toward cyclin B1 ( Fig. 2A). Mutations of either the KEN1 or KEN2 box did not affect this inhibition. In contrast, a BubR1 mutant with residues 525-561 (which encom-pass both the Phe and D boxes) deleted, termed ⌬PheD, reduced the ability of BubR1 to inhibit APC/C Cdc20 . These results are consistent with previous observations that, in the absence of Mad2, the middle region of BubR1 is critical for its APC/C inhibitory activity in vitro (11). The dispensability of the KEN boxes in BubR1-mediated inhibition of APC/C (in the absence of Mad2) is consistent with the fact that APC/C Cdc20catalyzed ubiquitination of cyclin B1 relies on the D box of cyclin B1. The D box of BubR1M likely directly competes with the D box of cyclin B1 for Cdc20 binding, whereas the Phe box provides the necessary affinity.
To simplify and extend the analysis, we constructed a miniaturized version of BubR1, termed miniBubR1, which contained only the four identified Cdc20-binding motifs of BubR1: KEN1, KEN2, Phe, and D boxes (Fig. 2B). As expected, the full-length BubR1-Bub3 complex inhibited the activity of APC/C Cdc20 toward cyclin B1 (Fig. 2B)  was capable of inhibiting APC/C Cdc20 , albeit with lower potency (Fig. 2C). For example, miniBubR1 at 400 nM achieved a level of APC/C inhibition comparable to that of the full-length BubR1-Bub3 at 50 nM. Thus, miniBubR1 has substantial APC/C inhibitory activity but cannot fully recapitulate the APC/C inhibition by BubR1-Bub3. Other BubR1-Bub3 domains or elements contribute to APC/C Cdc20 inhibition, either by providing additional interaction surfaces or by correctly positioning the multiple Cdc20-binding motifs. We then introduced mutations into each of the four Cdc20binding motifs in miniBubR1 and assayed the ability of these mutants to inhibit APC/C Cdc20 . To better reveal the potential involvement of these motifs, we used a concentration of mini-BubR1 (200 nM) that produced only partial APC/C inhibition. Consistent with the results observed using full-length BubR1-Bub3, mutations of either the Phe or the D box reduced the APC/C inhibitory activity of miniBubR1 (Fig. 2D). In contrast, mutations of either the KEN1 or the KEN2 box did not affect the ability of miniBubR1 to inhibit APC/C Cdc20 . Therefore, the Phe and D boxes are critical for APC/C inhibition by BubR1 in vitro.
Because the MCC contains also Mad2, we next tested whether the Phe and D boxes are required for APC/C Cdc20 activity in the presence of all MCC components. Addition of BubR1-Bub3 and Mad2 individually at low concentrations was not sufficient to inhibit APC/C Cdc20 , but addition of BubR1-Bub3 and Mad2 together resulted in synergistic APC/C Cdc20 inhibition (Fig. 2E). In contrast to the results observed when BubR1-Bub3 was added alone, mutation of the KEN1 or KEN2 box decreased the synergistic APC/C inhibition in the presence of Mad2. Deletion of the Phe and D boxes did not decrease the ability of BubR1-Bub3 to synergize with Mad2. These results indicate that the BubR1M region has the ability to inhibit APC/ C Cdc20 in the absence of Mad2. However, when Mad2 is present, APC/C Cdc20 inhibition by BubR1 relies more on the KEN1 and KEN2 boxes. These results suggest the intriguing possibility that BubR1 binds and inhibits APC/C Cdc20 using different modes, depending on whether Mad2 is present or not.
The Phe Box Does Not Interact with Human Cdh1-Both APC/C activators, Cdc20 and Cdh1, contain highly homologous C-terminal WD40 repeat domains that form similar ␤ propeller structures to directly interact with KEN and D boxes. Surprisingly, in the absence of Mad2, human BubR1 alone is capable of inhibiting APC/C Cdc20 in vitro but does not inhibit APC/C Cdh1 efficiently (11). We wondered whether the Phe box had binding specificity toward Cdc20. Consistent with data in Fig. 1, the BubR1 526 -583 and BubR1 526 -613 fragments that contained both Phe and D boxes bound to Cdc20 efficiently (Fig.  3A). Mutation of either box reduced Cdc20 binding. The smaller BubR1 526 -546 fragment containing only the Phe box retained substantial Cdc20 binding. In contrast, this Phe boxcontaining fragment failed to interact with Cdh1 (Fig. 3B). The BubR1 526 -583 and BubR1 526 -613 fragments containing both Phe and D boxes also bound less efficiently to Cdh1 than to Cdc20. This weaker Cdh1 binding by BubR1M was completely dependent on its D box, as mutation of the D box abolished the interaction between BubR1M and Cdh1 (Fig. 3B). Taken together, these results indicate that the Phe box in BubR1 preferably binds to human Cdc20. The cooperative binding of both Phe and D boxes thus enables BubR1 to bind to Cdc20 with higher affinity. Cdh1, on the other hand, can only engage the D box of BubR1 and thus binds with weaker affinity.
Structural studies of Cdc20 and Cdh1 have revealed similar APC/C degron-binding surfaces in these activators (15)(16)(17). They each contain a single, highly conserved KEN box-binding site on the top face of the ␤ propeller formed by their WD40 domain and a D box-binding site at the side of the ␤ propeller (Fig. 3C). Despite extensive efforts, we could not determine the crystal structure of human Cdc20 bound to the Phe box of BubR1. The electron density maps of crystals grown with Cdc20 bound to the Phe box peptide did not contain interpretable density belonging to the peptide.
The structure of the budding yeast Cdh1 bound to its inhibitor Acm1 revealed the details of how Cdh1 interacts with Acm1 (15). Interestingly, the N-terminal region of the Cdh1binding region of Acm1 (residues 61-66) has the sequence of FMLYEE, which conforms to the consensus of the Phe box. This motif binds to the bottom face of WD40 propeller of yeast Cdh1 (Fig. 3, C and D). This region in Acm1 was named the A motif, although the previous analysis focused on the two glutamates, and the potential importance of the aromatic and hydrophobic residues preceding them was not shown (35). We will henceforth refer to the FMLYEE motif in Acm1 as the Phe box. While this manuscript was under review, a study identified a motif similar to the A motif and Phe box in yeast Clb5, termed the ABBA motif, and showed that mutation of this motif delays Clb5 degradation (36). Thus, the Phe box is present in both APC/C inhibitors and substrates.
Superposition of the yeast Cdh1 and human Cdc20 structures underscores the overall similarity between their ␤ propellers (Fig. 3C). A closer inspection of the Phe box-binding site revealed that many residues in yeast Cdh1 critical for Phe box binding are conserved in human Cdc20 (Fig. 3D). For example, the two hydrophobic residues Phe-61 and Leu-63 in the Phe box form extensive hydrophobic interactions with Ile-311, His-316, His-355, and Val-371 of yeast Cdh1. These residues correspond to Ile-235, Tyr-240, Tyr-279, and Val-295 in human Cdc20, respectively, which are compatible for hydrophobic interactions with Phe-528 and Ile-530 of the BubR1 Phe box. Surprisingly, these yeast Cdh1 residues are not conserved in human Cdh1 and correspond to Ser-238, Leu-243, Glu-282, and Thr-300 in human Cdh1. Similarly, the two acidic residues Glu-65 and Glu-66 of the Phe box in Acm1 exhibit favorable electrostatic interactions with Lys-333, Lys-335, and Arg-338 in yeast Cdh1, which correspond to Gln-257, Lys-259, and Arg-262 in human Cdc20. Two of these three basic or polar residues are again not conserved in human Cdh1.
Our structural modeling thus suggests the intriguing possibility that the BubR1 Phe box might bind to human Cdc20 in a manner similar to the binding between yeast Cdh1 and the Phe box/A motif of Acm1. This binding mode provides a straightforward explanation for the inability of human Cdh1 to recognize the Phe box, because it lacks many of the putative Phe box-binding residues. Consistent with this hypothesis, three of the five mutations targeting predicted Phe box-binding residues, including K259E, R262S, and Y279E, greatly reduced Cdc20 binding to the Phe box of BubR1 (Fig. 3E).

BubR1 Phe and D Boxes Contribute to MCC Homeostasis in
Mitotic Human Cells-Our results so far established the roles for the Phe and D boxes of BubR1 in Cdc20 binding in vitro. We next tested whether these motifs also mediated the binding of BubR1 to Cdc20 in human cells. We constructed a BubR1 mutant with residues 525-561 (which encompass both the Phe and D boxes) deleted, termed ⌬PheD. As important controls, we also included the BubR1 mutants with its KEN boxes mutated (mKEN1 and mKEN2) in our experiments. HeLa cells were transfected with Myc-BubR1 WT, mKEN1, mKEN2, mD, mPhe, or ⌬PheD and arrested in mitosis with the microtubuledepolymerizing drug nocodazole. Lysates of mitotic cells col- lected through shake-off were subjected to immunoprecipitation with anti-Myc antibody beads. The lysates and the anti-Myc immunoprecipitates were analyzed with quantitative Western blot. Myc-BubR1 WT efficiently pulled down Bub3, Cdc20, and Mad2, consistent with MCC formation (Fig. 4, A  and B). Myc-BubR1 mKEN1 completely lost its binding to Cdc20 or Mad2, whereas mKEN2 still retained some binding (Fig. 4A). These results were consistent with published reports showing that the KEN1 box, but not the KEN2 box, of BubR1 was required for MCC formation in human cells (23). Strikingly, despite retaining proper Bub3 binding, Myc-BubR1 ⌬PheD bound to Cdc20 and Mad2 much less efficiently, at 70 and 30% of the WT levels (Fig. 4A). Mutations of either the Phe or D box also reduced Cdc30 and Mad2 binding, indicating that both motifs are involved in maintaining MCC levels. One possibility is that BubR1 recruits Cdc20 to kinetochores through the Phe and D boxes. In fact, a role for the BubR1 middle region in Cdc20 kinetochore targeting was recently reported (37). We first tested whether BubR1 was required for the kinetochore localization of Cdc20. The kinetochore localization of Cdc20 was slightly elevated in cells depleted of BubR1 (Fig. 4, B and C). Thus, BubR1 is not required for the kinetochore targeting of Cdc20 in our system. Whether this discrepancy is due to technical differences in experimental design or reflects the existence of multiple Cdc20 targeting mechanisms differentially favored in specific cell lines remains to be determined.
To further probe the mechanism by which BubR1M contributes to MCC homeostasis, we transfected HeLa cells with various Myc-BubR1 plasmids, arrested them in mitosis with nocodazole or the microtubule-stabilizing drug taxol, and treated the cells with the proteasome inhibitor MG132 for 1 h before collecting the mitotic cells through shake-off. The lysates and the anti-Myc immunoprecipitates were again analyzed with quantitative Western blot. Cdc20 degradation by the proteasome is an important pathway for MCC disassembly (38 -40). MG132 blocks this pathway and hampers MCC disassembly. Under these conditions, mutation on BubR1 KEN1 still abolished Mad2 binding and greatly reduced Cdc20 binding to BubR1 (Fig. 5), consistent with the notion that KEN1 is directly required for MCC formation. In contrast, the amounts of Cdc20 and Mad2 bound to BubR1 ⌬PheD were similar to those bound to BubR1 WT in cells treated with nocodazole and MG132 and were slightly reduced in cells arrested in taxol and MG132. These results indicate that the Phe and D boxes of BubR1 are not strictly required for MCC formation. They might contribute to MCC homeostasis through interfering with proteasome-dependent MCC disassembly.
The Phe and D Boxes of BubR1 Are Dispensable for the Spindle Checkpoint in Human Cells-The Phe and D boxes of BubR1 are required for the ability of BubR1 to inhibit APC/ C Cdc20 in vitro, in the absence of Mad2. They are involved in maintaining the normal levels of MCC in mitotic HeLa cells. We next tested whether these motifs were required for the spindle checkpoint in human cells. HeLa cells were depleted of the endogenous BubR1 by transfection with a BubR1 siRNA (siBubR1) and were simultaneously transfected with siBubR1resistant transgenes encoding Myc-BubR1 WT or mutants targeting various motifs. The cells were incubated with high concentrations of nocodazole (500 nM) or taxol (200 nM) for 15 h, stained with the MPM2 antibody (which recognizes mitotic phospho-epitopes), and analyzed with flow cytometry. As expected, depletion of BubR1 greatly reduced the mitotic index of cells treated with nocodazole or taxol (Fig. 6, A and B). Ectopic expression of Myc-BubR1 WT, but not mKEN1 or mKEN2, restored the mitotic arrest in cells depleted of the endogenous BubR1, consistent with the known requirement for both KEN boxes in the spindle checkpoint. Myc-BubR1 mPhe, mD, and ⌬PheD were fully functional in restoring nocodazoleor taxol-triggered mitotic arrest. All Myc-BubR1 proteins were expressed at levels comparable with that of the endogenous BubR1 (Fig. 6C). Thus, although the Phe and D boxes are important for Cdc20 binding and APC/C Cdc20 inhibition in vitro and for MCC homeostasis in human cells, they are appar-ently dispensable for the mitotic arrest triggered by high concentrations of spindle poisons. Our observation is in fact consistent with previous reports showing that deletions of the middle region of BubR1 do not produce strong checkpoint defects (30,31). Recent studies have shown that the length of mitotic arrest and presumably the strength of the spindle checkpoint can be tuned by the number of unattached kinetochores or the severity of spindle damage (41)(42)(43). We wondered whether the Phe and D boxes of BubR1 were required for the less robust mitotic arrest triggered by low concentrations of taxol (Fig. 6, D and E). The mitotic index of control HeLa cells treated with 30 nM taxol (ϳ30%) was indeed lower than that of cells treated with 200 nM taxol (ϳ45%). Surprisingly, in the presence of 30 nM taxol, cells expressing Myc-BubR1 mPhe, mD, or ⌬PheD all exhibited a small but statistically significant increase in mitotic index, as compared with cells expressing Myc-BubR1 WT (Fig. 6, D and  E). This result suggests that when cells are challenged with low degrees of spindle damage, the Phe and D boxes of BubR1 might even attenuate the strength of the spindle checkpoint.
To further confirm this surprising result, we carefully measured the duration of mitotic arrest in HeLa cells expressing BubR1 WT, mD, mPhe, and ⌬PheD in the presence of 30 nM taxol with live cell imaging. Under these conditions, controldepleted cells were arrested in mitosis for several hours (Fig.  7A). Most cells then underwent mitotic slippage or adaptation (Fig. 7B). This outcome was in stark contrast to cells treated with high concentrations of taxol, virtually all of which underwent apoptosis following the mitotic arrest (44). This difference was again consistent with the notion that low concentrations of  taxol induce a weaker spindle checkpoint response. As expected, cells transfected with siBubR1 and a control vector quickly exited mitosis and underwent adaptation, indicative of a gross checkpoint defect (Fig. 7, A and B). Expression of Myc-BubR1 WT fully rescued the checkpoint defect caused by siBubR1. Consistent with the flow cytometry data, cells expressing Myc-BubR1 ⌬PheD remained in mitosis slightly longer than cells expressing Myc-BubR1 WT did (⌬t 50 ϭ 4.1 Ϯ 2.5 h) (Fig. 7A). Myc-BubR1 mD had a smaller effect, whereas mPhe was comparable with WT. Myc-BubR1 WT and mutants were expressed at comparable levels (Fig. 7C). Thus, despite its positive contribution to MCC levels, the middle region of BubR1 is dispensable for the spindle checkpoint. It appears to attenuate the strength of the checkpoint when cells experience low grade spindle damage.

DISCUSSION
BubR1 is an evolutionarily conserved spindle checkpoint protein and a critical component of MCC (the BubR1-Bub3-Cdc20-Mad2 complex), which is a potent checkpoint inhibitor of APC/C. Two regions of human BubR1, the N-terminal region (BubR1N) and the middle region (BubR1M), mediate Cdc20 binding and APC/C inhibition. BubR1N shares sequence similarity with its yeast homolog Mad3 and, on its own, binds to Cdc20 weakly and does not inhibit APC/C (23). In the presence of Mad2, however, BubR1N (or Mad3) simultaneously engages Cdc20 with its first KEN box and contacts the Cdc20-bound Mad2, thus nucleating the formation of MCC (17,23,29). Thus, the spindle checkpoint function of BubR1N is well established. It collaborates with Mad2 to bind to Cdc20 cooperatively and is critical for MCC assembly and APC/C inhibition. In contrast, BubR1M binds to Cdc20 and inhibits APC/C in the absence of Mad2 in vitro (11), but deletions of BubR1M do not produce spindle checkpoint defects in mammalian cells (30,31). The nature and function of the BubR1M-Cdc20 interaction have thus remained murky. In this study, we show that BubR1M contains two Cdc20-binding motifs conserved in vertebrates: a D box and a Phe box. These two tandem motifs bind cooperatively to Cdc20 and are required for APC/C inhibition in vitro. They also contribute to the maintenance of MCC levels in human cells during checkpoint-mediated mitotic arrest.
Role of the D box of BubR1M in APC/C inhibition-The D box with the consensus RXXLXXXXN is a well established APC/C degron found in many APC/C substrates and inhibitors and interacts with both Cdc20 and Cdh1. The D box of cyclin B1 is critical for its ubiquitination by APC/C (45). The fact that BubR1M contains a D box required for APC/C inhibition strongly suggests that, in the absence of Mad2, BubR1M inhibits APC/C through acting as a pseudo-substrate to competitively block D box-dependent recruitment of cyclin B1. Why BubR1M is not itself ubiquitinated is unclear, but it may possibly be because the BubR1M D box lacks the conserved asparagine at the end. This type of shorter D boxes might only bind to Cdc20 but might not effectively engage the D box co-receptor APC10. Additionally, the full-length BubR1 is protected against APC/C-dependent ubiquitination by post-translational modifications, such as acetylation (46). Interestingly, the budding yeast Mad3 also contains a D box in its C-terminal region ( 429 RKAL 432 ), which is critical for its checkpoint function (20). Mutation of this D box reduces, but does not abolish, Cdc20 binding. It will be interesting to test whether, in the absence of Mad2, the D box of Mad3 suffices to inhibit ubiquitination of D box substrates by APC/C Cdc20 in vitro.
Role of the Phe Box in Cdc20 Binding-The Phe box is related to the A motif previously identified in Acm1 (35) and is also present in other Cdc20-binding proteins, including human Bub1 and cyclin A2 (Fig. 1C). Structural modeling (with the human Cdc20 and yeast Cdh1-Acm1 structures as templates) and mutagenesis identified the Phe box-binding site in human Cdc20 (Fig. 3). Intriguingly, yeast Acm1 has been shown to bind specifically to yeast Cdh1, but not to yeast Cdc20 (15). Several of the key yeast Cdh1 residues that interact with the Phe box are not conserved in yeast Cdc20, suggesting that the Acm1 specificity for yeast Cdh1 is at least partially dependent on binding to the Phe box. As expected from the fact that yeast Cdc20 does not recognize the Phe box, Mad3 does not appear to have a Phe box. On the other hand, our results show that the BubR1 Phe box mediates binding to human Cdc20, but not to human Cdh1. Although it is unclear whether this switch in binding specificity has functional implications, this phenomenon can be explained by structural comparisons. Many of Phe box-binding residues are not conserved in human Cdh1, explaining the specificity of the Phe box in binding to human Cdc20.
It will be interesting to test whether the putative Phe boxes of Bub1 and cyclin A2 are also involved in their interaction with Cdc20. Intriguingly, the Phe box in Bub1 is located only two residues N-terminal to its first KEN box (KEN1) and ϳ100 residues N-terminal to the second KEN box (KEN2) (47). The two KEN boxes of Bub1 have been previously shown to bind cooperatively to Cdc20 (48). Because Cdc20 only has one KEN boxbinding site, this cooperativity remained unexplained. In light of the fact that a Phe box in Bub1 lies immediately adjacent to KEN1, it is possible that the Bub1 Phe box and KEN2 bind cooperatively to Cdc20. The effects of KEN1 mutations on Bub1 binding to Cdc20 might be due to collateral damage to the conjoined Phe box.
Different Modes of APC/C Inhibition by BubR1 in the Presence or Absence of Mad2-Despite the ability of BubR1 to directly bind Cdc20 and inhibit APC/C in vitro, it is clearly established that the full MCC is a more potent inhibitor than either Mad2 or BubR1 alone. Interestingly, our results indicate that the BubR1M region is critical for APC/C inhibition in the absence of Mad2, but this region is largely dispensable for the inhibitory synergy observed in the presence of Mad2. These in vitro observations could explain the lack of checkpoint defects in human cells expressing BubR1 ⌬PheD, because these cells contain endogenous Mad2. Finally, a recent study showed that MCC could inhibit a second Cdc20 molecule already bound to APC/C (49). This trans-inhibitory activity of MCC requires the KEN2 box and a newly discovered D box adjacent to KEN2 in the N-terminal region of BubR1. Therefore, BubR1 contains five Cdc20-binding motifs: two KEN boxes, two D boxes, and a Phe box. These motifs can in theory engage three Cdc20 molecules simultaneously. How these elements collaborate to inhibit APC/C Cdc20 clearly requires further investigation.
Mechanisms by Which D and Phe Boxes Contribute to MCC Homeostasis-The Phe and D boxes are not only involved in Cdc20 binding and APC/C inhibition in vitro but also contribute to maintaining high level MCC in human cells. As compared with the wild type, a BubR1 mutant with a small region encompassing the Phe and D boxes deleted is less efficient in binding Cdc20 and Mad2 in human cells arrested in mitosis by nocodazole. Interestingly, blocking the activity of the proteasome largely bypasses the need for these motifs in maintaining MCC homeostasis. Thus, the Phe and D boxes contribute to MCC homeostasis, possibly by slowing the proteasome-dependent MCC disassembly.
The structure of the fission yeast MCC suggests that the KEN1-containing, N-terminal region of Mad3 (and by analogy BubR1N) is critical for MCC formation by contacting both Cdc20 and Mad2 (17). BubR1M is unlikely to directly contact Cdc20 in the context of MCC. The role of the BubR1M-Cdc20 interaction in maintaining MCC is probably indirect.
How then can the BubR1M-Cdc20 interaction maintain MCC levels in human cells? We propose the following speculative model (Fig. 8), which provides a plausible answer to this question. In this model, two types of Cdc20-containing checkpoint complexes exist at equilibrium: the BubR1-Bub3-Cdc20 complex (BBC) and the intact MCC comprising BubR1-Bub3-Cdc20-Mad2. BBC involves the cooperative binding of the Phe and D boxes to Cdc20, whereas MCC requires the cooperative binding of Mad2 and the KEN1 of BubR1 to Cdc20. MCC is a more potent APC/C inhibitor, but Cdc20 as a subunit of MCC is more prone to undergo autoubiquitination and proteasome- , which binds to BBC to promote MCC formation. The N-terminal region of BubR1 is critical for MCC formation, with the first KEN box (KEN1; indicated by a red bar) contacting Cdc20 directly. The Phe and D boxes (indicated by black and blue bars, respectively) might not directly contact Cdc20 in MCC. In the reverse reaction, p31 comet and the ATPase TRIP13 extract C-Mad2 from MCC to produce BBC and O-Mad2. Cdc20 in MCC is more efficiently autoubiquitinated by APC/C and is degraded through the proteasome. Because the Phe and D boxes bind to Cdc20 in BBC, extraction of Mad2 from MCC formed by BubR1 ⌬PheD will cause its complete disassembly into free BubR1-Bub3, Cdc20, and Mad2. Reassembly of these components into MCC might occur at a slower rate than that of MCC assembly through Mad2 incorporation into BBC. Rates of MCC disassembly pathways (proteasome-dependent degradation of Cdc20 and p31 comet /TRIP13-mediated Mad2 extraction) are unlikely to be affected by the Phe and D box deletion. The slower rate of reassembly in the presence of constant rates of disassembly explains why the steady-state level of MCC containing BubR1 ⌬PheD is lower. A potential explanation for why BubR1 ⌬PheD is still functional in the checkpoint is the existence of a redundant mechanism (mechanism X) of APC/C inhibition. This mechanism might compete with the Phe and D box of BubR1 for Cdc20 binding. In cells with BubR1 ⌬PheD, this mechanism might be enhanced, resulting in longer mitotic arrest under conditions of low grade spindle damage. dependent degradation (50), triggering complete MCC disassembly. With the help of p31 comet and TRIP13 (38 -40, 51-56), Mad2 can dissociate from MCC in a proteasome-independent pathway, producing BBC. Because Mad2 promotes Cdc20 autoubiquitination, Cdc20 in BBC should be more resistant to proteasome-dependent degradation. With the addition of active Mad2 (produced at unattached kinetochores), this partially disassembled MCC subcomplex can readily reassemble into the intact MCC. BubR1 lacking the Phe and D boxes cannot form BBC. The p31 comet -and TRIP13-dependent Mad2 extraction causes the complete disassembly of MCC containing this BubR1 mutant, slowing its reassembly and reducing its steady-state levels.
Mutation of KEN1 completely abolishes the BubR1-Bub3-Cdc20 interaction in nocodazole-arrested mitotic human cells (Fig. 4A), suggesting that the vast majority of Cdc20-containing complexes exist as the intact MCC. The steady-state level of BBC is very low. Nevertheless, the transient existence of this complex is critical for maintaining high levels of MCC, presumably through preventing its complete disassembly.
Functions of the Phe and D Boxes in the Spindle Checkpoint-Although deletion of the Phe and D boxes reduces MCC levels severalfold, the BubR1 ⌬PheD mutant is fully functional in the spindle checkpoint. One simple explanation for this conundrum is that the residual amount of MCC is still above the threshold needed for APC/C Cdc20 inhibition. This possibility, however, does not explain why BubR1 ⌬PheD is slightly more effectively in maintaining mitotic arrest when the checkpoint is activated by low grade spindle damage, such as that caused by low concentrations of taxol.
To explain the apparent gain of function of BubR1 ⌬PheD, we envision two additional possibilities that are not mutually exclusive. In the first possibility, the checkpoint employs an independent mechanism that collaborates with MCC to inhibit APC/C (Fig. 8). This mechanism is partially redundant to and might even compete with BubR1M for Cdc20 binding. The removal of BubR1M strengthens this independent mechanism, which sustains the checkpoint along with the decreased amount of MCC. This functional compensation might even produce longer mitotic arrest in response to low grade spindle damage, at least in HeLa cells. Along this vein, we note that Bub1 also contains a Phe box. Deletion of the Phe box in BubR1 might increase Bub1 binding to Cdc20 and enhance Bub1-mediated Cdc20 phosphorylation, which can inhibit APC/C Cdc20 in the absence of MCC (57). In the second possibility, the Phe and D boxes of BubR1 have dual, opposite functions in the spindle checkpoint. Aside from their positive roles in maintaining MCC, they might also have negative roles. For example, BubR1 binds to protein phosphatase 2A through the KARD motif (residues 664 -681) (58), which lies ϳ100 residues C-terminal to the D box in BubR1M. The close proximity of Cdc20 bound to BubR1M and protein phosphatase 2A bound to KARD might promote Cdc20 dephosphorylation and checkpoint silencing. Thus, similar to Mad2, which inhibits APC/C as part of the MCC and at the same time promotes MCC disassembly through its interaction with p31 comet , BubR1 might also have dual positive and negative functions that contribute to APC/C inhibition in the presence of an active checkpoint but promote fast MCC disassembly and APC/C activation when the checkpoint needs to be silenced. Deletion of the Phe and D boxes is expected to eliminate both their positive and putative negative functions, resulting in a largely net neutral effect on the checkpoint. Under specific conditions or in certain cell lines, elimination of the negative function by Phe and D box mutations might even outweigh the reduction in the positive function, resulting in gain of function phenotypes.
Conclusion-In summary, we have identified two functional motifs, the Phe and D boxes, in the middle region of BubR1 that mediate Cdc20 binding and APC/C inhibition. These motifs promote MCC homeostasis in human cells but are not required for the spindle checkpoint. This apparent contradiction highlights an important gap in our understanding of the spindle checkpoint and leads to several plausible and testable hypotheses, including functional duality of these motifs or the existence of redundant, competing mechanisms that inhibit APC/C.