Direct interactions of mitotic arrest deficient 1 (MAD1) domains with each other and MAD2 conformers are required for mitotic checkpoint signaling

As a sensitive signaling system, the mitotic checkpoint ensures faithful chromosome segregation by delaying anaphase onset even when a single kinetochore is unattached to mitotic spindle microtubules. The key signal amplification reaction for the checkpoint is the conformational conversion of “open” mitotic arrest deficient 2 (O-MAD2) into “closed” MAD2 (C-MAD2). The reaction has been suggested to be catalyzed by an unusual catalyst, a MAD1:C-MAD2 tetramer, but how the catalysis is executed and regulated remains elusive. Here, we report that in addition to the well-characterized middle region of MAD1 containing the MAD2-interaction motif (MIM), both N- and C-terminal domains (NTD and CTD) of MAD1 also contribute to mitotic checkpoint signaling. Unlike the MIM, which stably associated only with C-MAD2, the NTD and CTD in MAD1 surprisingly bound both O- and C-MAD2, suggesting that these two domains interact with both substrates and products of the O-to-C conversion. MAD1NTD and MAD1CTD also interacted with each other and with the MPS1 protein kinase, which phosphorylated both NTD and CTD. This phosphorylation decreased the NTD:CTD interaction and also CTD's interaction with MPS1. Of note, mutating the phosphorylation sites in the MAD1CTD, including Thr-716, compromised MAD2 binding and the checkpoint responses. We further noted that Ser-610 and Tyr-634 also contribute to the mitotic checkpoint signaling. Our results have uncovered that the MAD1NTD and MAD1CTD directly interact with each other and with MAD2 conformers and are regulated by MPS1 kinase, providing critical insights into mitotic checkpoint signaling.

The mitotic checkpoint is a crucial signal transduction pathway that contributes to faithful chromosome segregation (1)(2)(3)(4). A single unattached kinetochore delays anaphase onset, underscoring the importance of signal amplification for the mitotic checkpoint (5). The conversion of mitotic arrest deficient 2 (MAD2) 2

domain from open (O-MAD2) to closed (C-MAD2)
conformation is a well-recognized signal amplification mechanism for the mitotic checkpoint (6,7). O-MAD2 is the predominant conformer in interphase cells (8,9). During prometaphase, intracellular C-MAD2 concentration is increased to promote formation of the mitotic checkpoint complex (MCC), which binds and inhibits the anaphase promoting complex/ cyclosome (APC/C) (1)(2)(3)(4). In the current model, the MAD2 O-C conversion (i.e. O to C conversion) is catalyzed by a 2:2 MAD1:C-MAD2 tetramer localized at unattached kinetochores. Cytoplasmic O-MAD2 then heterodimerizes with the C-MAD2 moiety in the catalyst and morphs into C-MAD2. The reaction mechanism for the conversion is still unclear, but it may involve some intermediate folding states (I-MAD2) (6,7,10).
Here we report our results targeting the above two questions. While our manuscript was prepared, two reports were published indicating that MPS1 phosphorylates MAD1 to enhance MAD2 O-C conversion and MCC assembly (29,30). Our

MAD1 domains in the mitotic checkpoint
results support the importance of MAD1 Thr-716 phosphorylation, but we have also uncovered other protein-protein interactions between MAD1, MAD2, and MPS1 and the phosphorylation-dependent regulation of some of the interactions. Our work highlights the coordination of different MAD1 domains in efficient mitotic checkpoint signaling and provides further mechanistic insights into the MAD2 O-C conversion reaction.

MAD1 N-terminal and C-terminal domains (NTD and CTD) are required for efficient mitotic checkpoint signaling
In studying the MAD2 O-C conversion, earlier work has detailed the conformational changes of MAD2 (6,32). We reasoned that better characterization of the MAD1:C-MAD2 catalyst would provide further mechanistic insights into the conversion reaction and hence the signal amplification step of the mitotic checkpoint. We noted that even though MAD1 is commonly depicted as a rigid coiled-coil protein, parts of its NTD and CTD have been shown or predicted to remain disordered or adopt other structures ( Fig. 1a and Fig. S1) (11,33,34). We first investigated the possible contribution of MAD1 NTD and MAD1 CTD to the mitotic checkpoint using a "separation of function" system developed by Maldonado and Kapoor (26). In this system, an mCherry-Mis12-MAD1 fusion construct was exploited to examine catalytic efficiency of the MAD1:C-MAD2 catalyst without concerns over the kinetochore targeting aspect of its regulation (26) (Fig. 1b). Although endogenous MAD1 and MAD2 disappeared from metaphase kinetochores which presumably were occupied by spindle microtubules, expression of wild-type MAD1 (MAD1 WT ) fused with constitutive kinetochore protein Mis12 retained MAD1 at the metaphase plate and recruited GFP-MAD2 L13A to these metaphase kinetochores (26,28) (Fig. 1c and Fig. S2). MAD2 L13A is a MAD2 mutant locked in C conformation (13,32). The persistence of MAD1 and MAD2 at metaphase attached kinetochores was sufficient to trigger a Ͼ12-h mitotic arrest in HeLa cells (26,28) (Fig. 1d). The arrest was dependent on C-MAD2 binding to MAD1, as cells expressing the fusion with MAD2-bindingdeficient MAD1 AA mutant (K541A, L543A in MIM) finished mitosis within ϳ60 min on average ( Fig. 1d and Fig. S2) (26). Note no GFP-MAD2 L13A was localized at metaphase kinetochores containing mCherry-Mis12-MAD1 AA , although GFP-MAD2 L13A did appear at the last few unattached kinetochores, most likely because of presence of endogenous MAD1 there (Fig. S2, compare the second and third columns). Furthermore, co-expression of MAD2 ⌬C10 , an O-conformer locked mutant of MAD2 (6,7), abolished the mitotic arrest in MAD1 WT transfected cells (data not shown), corroborating that the arrest was because of O-C conversion-dependent checkpoint responses (28,35,36).
We then examined the association of different mCherry-Mis12-MAD1 fusion proteins with endogenous MAD1 and MAD2 by Mis12 immunoprecipitation. Little MAD2 was found to associate with MAD1 AA as predicted; steady-state levels of MAD2 binding to other MAD1 mutants were also slightly reduced (Fig. 1e). The reduction of MAD2 binding became more obvious for some mutants when FLAG immunoprecipitation was performed using cells transfected with FLAG-tagged MAD1 constructs (Fig. 1f). In addition, MAD1 ⌬CTD and MAD1 ⌬NTD did not interact with endogenous MAD1, indicating a dimerization defect (Fig. 1, e and f; note that mCherry-Mis12-MAD1 ⌬NTD runs at the same position as endogenous MAD1. More on MAD1 dimerization in the last section of "Results"). Taken together, the results shown in Fig. 1 have confirmed the important role of MAD1 MIM , but also revealed that both NTD and CTD of MAD1 are required for an efficient mitotic checkpoint.

MAD1 NTD and MAD1 CTD bind to both O-MAD2 and C-MAD2
We hypothesized that the NTD and CTD of MAD1 facilitate mitotic checkpoint responses by enhancing MAD2 O-C conversion. Based on the analogy to an isomerase which at least transiently interacts with its substrate and product, we prepared recombinant proteins and examined potential interactions between GST-tagged MAD1 NTD or MAD1 CTD with untagged O-or C-MAD2, supplied as MAD2 ⌬C10 or MAD2 L13A conformation-locked mutants, respectively (6, 7, 36) (Fig. 2a). When proteins were used at roughly endogenous concentrations (see "Experimental Procedures"), GST-pulldown results showed that NTD and CTD bound to both conformers of MAD2 (Fig. 2b, lanes 1, 3 and 5, 7). Under similar conditions, the MAD1 MIM only bound to C-MAD2, just as reported previously (compare MAD2 in Fig. 2b, lanes 2 and 6) (11,38). The conformational status of the O-and C-MAD2 mutants was further verified, because only MAD2 L13A bound to GST-CDC20 (111-138) or GST-BUBR1  , also as reported before (36,38) (Fig. S3). The novel interactions were not mediated by tags, as GST alone did not pull down any MAD2 (Fig. 2b,  lanes 4 and 8). Importantly, GFP-MAD2 ⌬C10 was found to be recruited to centromeres in interphase cells expressing mCherry-Mis12-MAD1 fusions, supporting the idea that the interaction between O-MAD2 and MAD1, although surprising, could happen in cells ( We attempted to further define the regions on MAD1 NTD or MAD1 CTD responsible for association with MAD2. Several MAD1 NTD or MAD1 CTD truncations produced either insoluble or heavily degraded proteins (data not shown). However, testing with MAD1 NTD truncations that we were able to purify, including MAD1  , MAD1 (327-423) , and MAD1 (327-488) , revealed dramatically reduced MAD2-binding capability of these fragments (Fig. 2d). The globular subdomain (638 -718 residues) of MAD1 CTD retained approximately half the MAD2-binding capacity of MAD1 CTD whereas the coiledcoil subdomain in the CTD showed only residual binding (Fig. 2e). These combined results suggest that the integrity of MAD1 NTD or MAD1 CTD is crucial for binding to O-MAD2 or C-MAD2.

A novel interface in MAD2 is employed for its association with NTD or CTD of MAD1
A MAD2 molecule has two well-characterized interfaces for protein-protein interactions: the "safety belt" characteristic of C-MAD2 conformation to which MAD1 MIM and CDC20 bind (11, 38 -40), and the dimerization domain (primarily ␣C helix) that allows MAD2 to form O:C or C:C dimers or interact with p31 comet or BUBR1 (13,32,36,41,42). NTD and CTD of MAD1 bound to both MAD2 ⌬C10 and MAD2 L13A , suggesting no discrimination against either MAD2 conformation (Fig. 2b). To further support this notion, wild-type MAD2, which exists as a mixture of both O and C conformers (8), bound to all three MAD1 fragments, whereas another MAD2 mutant MAD2 S195D , predominantly in O conformation (43), bound to MAD1 NTD or MAD1 CTD but not MAD1 MIM (Fig. 3a, compare the left and middle panels). On the other hand, MAD2 LARQ (L13A/R133E/Q134A), a dimerization defective C-MAD2 mutant (36), associated with MAD1 NTD or MAD1 CTD at similar levels as MAD2 L13A (Fig. 3b). MAD2 ⌬N10 , which might mimic an intermediate conformation during O to C MAD2 conversion (10), bound to all three MAD1 fragments (Fig.  3c). The above results suggest that the safety belt (or C conformation), the dimerization domain (at least the RQ mutant), or the N-terminal 10 amino acids are not essential for MAD2 to interact with MAD1 NTD or MAD1 CTD . Therefore, a novel interface on MAD2, possibly shared by both O and C conformers, is employed for interactions with MAD1 NTD and MAD1 CTD . GST-NTD tends to be more labile for degradation. The asterisk marks expected size of GST-NTD. b, GST-tagged MAD1 NTD , MIM, CTD, or GST alone were incubated with either MAD2 L13A or MAD2 ⌬C10 , and GST pulldown assays were followed by Western blotting of GST and MAD2. The final concentrations of GST or GST-tagged MAD1 fragments were 60 nM, whereas those of MAD2 mutants were 230 nM. Similar results have been replicated for at least five times by two investigators, using different batches of protein preparations. c, immunofluorescence of two interphase cells transfected with GFP-MAD2 ⌬C10 . Note that the GFP signals are recruited to centromeres only in the cell co-expressing mCherry-Mis12-MAD1 ⌬CTD . Centromeres are stained with anti-CENP-I antibody. Scale bar, 10 m. d, GST-tagged MAD1 N-terminal truncations were incubated with MAD2 ⌬C10 or MAD2 L13A followed by GST pulldown assays. The result is representative of four replicates. e, GST-tagged MAD1 CTD coiled-coil subdomain (597-638 residues) or globular subdomain (638 -718 residues) were incubated with MAD2 ⌬C10 or MAD2 L13A followed by GST pulldown and Western blotting. The result is representative of three replicates by two investigators based on different batches of proteins.

MAD1 domains in the mitotic checkpoint MPS1 kinase phosphorylates MAD1 NTD and MAD1 CTD
It is known that the kinase activity of MPS1 is essential for maintaining mitotic arrest even when MAD1 constitutively localizes at kinetochores as a fusion with mCherry-Mis12 (25)(26)(27)(28). We hypothesized that MPS1 might phosphorylate MAD1 or MAD2 to regulate the efficiency of the MAD1:C-MAD2 catalyst. In vitro kinase assays found that MPS1 indeed phosphorylated MAD1 NTD and MAD1 CTD (Fig. 4a, compare lanes 3 and 5 in both the Coomassie Blue stain panel and the autoradiography panel). The specificity of the kinase assay was validated by the fact that reversine, a previously characterized MPS1 inhibitor (44), reduced in vitro phosphorylation of an artificial MPS1 substrate myelin basic protein (Fig. 4a, compare lanes 1 and 2 in the right panel). GST alone, GST-MAD1 MIM , and MAD2 conformers (MAD2 L13A and MAD2 ⌬C10 ) were not good substrates for MPS1 under the experimental condition (Fig. 4a, lanes 4  and 6-8). Interestingly, recombinant MPS1 kinase binds to MAD1 NTD or MAD1 CTD but not MIM in the absence of ATP (Fig. 4b, note the presence of MPS1 signal in lanes 1 and 3 but not in lanes 2 and 4).
We then tested whether MPS1 phosphorylation affects the interactions between MAD1 CTD and MAD2 conformers. We focused on CTD because it was not only easily purified without apparent degradation but also showed functional significance (see below). We did notice that GST-MPS1 but not GST alone could also bind to MAD2 L13A and MAD2 ⌬C10 (Fig. 4c). In vitro incubation of MPS1, CTD, and MAD2 led to phosphorylation of CTD, as evidenced by the mobility shift detected by anti-His 6 antibody, and positive signals of phospho-Thr antibody (Fig. 4, d and e, lane 1). However, the association of either MAD2 L13A or MAD2 ⌬C10 with CTD did not show obvious changes when compared with reactions in the absence of ATP or in the presence of MPS1 inhibitor reversine (Fig. 4, d and e, compare MAD2 signals in lanes 1-3). The interactions between phosphorylated CTD and MPS1, nevertheless, became weaker (Fig.  4, d and e, compare MPS1 signals in lanes 1-3). ATP, reversine, and DMSO (solvent control) did not affect the interactions between CTD and MAD2 conformers (Fig. 4, d and e, lanes  4 -6).

Functional characterization of MPS1 phosphorylation sites on MAD1
Realizing some limitations of in vitro binding experiments especially with protein fragments, we decided to directly test the potential effects of MPS1 phosphorylation on MAD1 in cells. Through mass spectrometry we identified eight in vitro MPS1 phosphorylation sites on MAD1 NTD and MAD1 CTD . The sites are Thr-8, Ser-22, Ser-62, Thr-323, Ser-598, Ser-610, Figure 3. A novel interface in MAD2 is employed for its association with NTD or CTD of MAD1. a, GST-pulldown assays using GST-tagged MAD1 fragments after incubation with His 6 -tagged wild type MAD2 (MAD2 WT ) or MAD2 S195D mutant. b, GST-pulldown assays using GST-tagged MAD1 fragments with MAD2 L13A and MAD2 LARQ . The asterisk marks expected size of GST-NTD. The result is representative of four replicates. c, GST-pulldown assays using GST-tagged MAD1 fragments with MAD2 ⌬N10 .

MAD1 domains in the mitotic checkpoint
Thr-624, and Thr-716 (Fig. 5a). To test the importance of these sites in cells, a mCherry-Mis12-MAD1 8A mutant with all the sites mutated into alanine was expressed and found to be defective in maintaining the mitotic arrest (229 Ϯ 23 min mitotic duration) (Fig. 5b). Mutating all four NTD sites into alanine showed no effect, but alanine mutants at the four CTD sites (MAD1 CTD4A ) shortened the mitotic duration comparable with MAD1 8A (Fig. 5b). Further tests found that little CTD 4A is phosphorylated when compared with wild-type CTD (CTD WT ) in vitro, supporting that the CTD four sites, Ser-598, Ser-610, Thr-624 and Thr-716, constituted primary phosphorylation sites of MPS1 kinase under our in vitro experimental conditions (Fig. S4a).
Next, a mCherry-Mis12-MAD1 CTD4E phosphomimic mutant was prepared and surprisingly found to be also defective in The asterisks indicate the expected sizes of corresponding proteins. The arrow indicates GST-MPS1. Reversine (Rev) was used in lane 2 to validate the kinase specificity. b, GST-pulldown assays using GST or GST-tagged MAD1 fragments (ϳ80 nM final concentration) after incubation with His-tagged MPS1 (ϳ40 nM). The binding reactions did not contain ATP. The result is representative of at least three replicates by two investigators. c, GST-pulldown assays using GST or GST-tagged MPS1 (ϳ60 nM) after incubation with untagged MAD2 ⌬C10 or MAD2 L13A (both at 230 nM final concentrations). The binding reactions did not contain ATP. The result is representative of at least three replicates by two investigators. d and e, immunoprecipitation using anti-His 6 antibody (1 g) to detect His-tagged MAD1 CTD interactions with untagged MAD2 L13A (d) or MAD2 ⌬C10) (e) after incubation in the presence/absence of GST-MPS1 kinase, ATP, or MPS1 inhibitor reversine. All recombinant proteins in the incubation reactions were used at final concentrations equivalent to their endogenous levels. The results have been replicated five times.

MAD1 domains in the mitotic checkpoint
maintaining the mitotic arrest (Fig. S4b). We reasoned that, for a functional mitotic checkpoint in cells, one or more of the four residues identified through in vitro kinase assays could not tolerate mutations. Both phosphoresistant and phosphomimic mutants at individual sites were then prepared. The mutants at Ser-598 or Ser-624 did not show obvious defects in mitotic checkpoint responses when fused with mCherry-Mis12. However, the mutants at Ser-610 and Thr-716 showed significant differences in mitotic arrest durations when compared with the MAD1 WT fusion (Fig. 5c). Furthermore, the T716E mutant maintained the checkpoint longer than the T716A mutant (441 Ϯ 38 versus 242 Ϯ 48 min, p Ͻ 0.05, Student's t test), indicating that Thr-716 is likely the residue activated by MPS1 for mitotic checkpoint signaling. However, the mCherry-Mis12 fusion of MAD1 T716E mutant could not maintain mitotic arrest when transfected cells were challenged by reversine (Fig. S4c), suggesting either the phosphomimic mutant is imperfect or MPS1 has additional key substrates required for a functional mitotic checkpoint.
To further understand the imaging results with different CTD mutants, we examined whether these mutants affected the interaction of MAD1 CTD with MAD2. Both CTD 4A and CTD 4E showed defects in binding to O-MAD2 and C-MAD2 (Fig. S4d). Furthermore, CTD T716A bound less O-MAD2 and C-MAD2 as compared with CTD WT or CTD T716E (Fig. 5d). Similarly, the functional defective MAD1 Y634E mutant also bound to less O-MAD2 and C-MAD2 than MAD1 Y634F (Fig.  S4e). These results correlate mitotic arrest durations with the capability of MAD1 CTD mutants to bind to MAD2 (Figs. 1d and 5 and Fig. S4).

MPS1 phosphorylation of MAD1 modulates interdomain and intermolecular interactions
Because MIM, CTD, and NTD of MAD1 are all required for an effective mitotic checkpoint (Fig. 1), we next investigated whether there is coordination among these domains. The presumed non-coiled-coil domains along the length of MAD1 might provide certain bends and turns to the molecule, and an earlier model has suggested that MAD1 CTD folds back to the proximity of the catalytic core consisting of MIM and associated C-MAD2 (11). We found that NTD directly interacted with CTD (Fig. 6, a, lane 1 and b, lane 2). CTD interacted with itself (Fig. 6a, lane 3), agreeing with published crystal structure (PDB: 4DZO) and acting as a positive control in this binding assay (33). MIM does not directly interact with CTD or NTD (Fig. 6, a, lane 2 and b, lane 1). Addition of MPS1 kinase reduced the interactions between MAD1 NTD and MAD1 CTD (Fig. 6c). The effect depended on the MPS1 kinase activity as omitting ATP from the reactions or adding MPS1 inhibitors (reversine or AZ3146) reversed the MPS1 effect on NTD:CTD interaction (Fig. 6c, compare His signals in lane 2 and those in lanes [3][4][5]. MAD1 dimerization has been observed in MAD1 MIM : MAD2 and MAD1 CTD crystal structures (PDB: 1GO4 and 4DZO, respectively) (11,33). However, whether MAD1 dimerization is essential for its activity to promote MAD2 O-C conversion is unclear. We noticed that in Fig. 1, e and f, MAD1 truncations missing either NTD or CTD did not co-immunoprecipitate endogenous MAD1 as well as other fusions. Recombinant GST-CTD 4A fragment also bound less His-CTD WT and even less His-CTD 4A (Fig. 6d, compare lanes 1-3). These results suggest a possible role of CTD in MAD1 dimerization or oligomerization (see "Discussion").

Discussion
MAD2 O-C conversion is a key signal amplification step for the mitotic checkpoint. The current model suggests that the conversion is catalyzed by an unusual catalyst: the MAD1:C-MAD2 complex localized at unattached kinetochores. How the catalysis is achieved is still unclear. Together with two very recent publications (29,30), our results support the roles of MAD1 CTD and MPS1 kinase in promoting the MAD2 O-C conversion. In particular, our work agrees with the critical role

MAD1 domains in the mitotic checkpoint
of MAD1 Thr-716, which is phosphorylated by MPS1, but also suggests that Ser-610 and Tyr-634 are potentially key residues for regulating MAD2 O-C conversion (Figs. 1 and 5). Ser-610 is phosphorylated by MPS1 in vitro (Fig. 4), and Tyr-634 was reported to be phosphorylated in vivo, although we have not confirmed this yet in mitotic cells (37). We have also uncovered additional protein-protein interactions between MPS1, MAD1, and MAD2 (Figs. 2-4 and 6). Of particular interest, we found that MAD1 NTD and MAD1 CTD interact with each other and both bind to O-MAD2 and C-MAD2. These results have been integrated into an updated model for the MAD1:C-MAD2 catalyst that promotes MAD2 O-C conversion (Fig. 7). Many mechanistic details remain to be filled, so we hereby discuss the implications of our results.

The functions of MAD1 NTD and MAD1 CTD in MAD2 O-C conversion
The crystal structure of MAD1 MIM in complex with C-MAD2 has solidified the now classical model of the MAD1: C-MAD2 catalyst as a 2:2 heterotetramer (11) (Fig. S1). Each of the two "liganded" C-MAD2s tightly wraps around one MIM monomer through its safety belt loop (11,18,38). C-MAD2 then utilizes its dimerization domain to recruit O-MAD2 and converts the latter into C-MAD2, resulting in signal amplification for the mitotic checkpoint (6,7,12). The conversion may go through multiple intermediate states (I-MAD2) (6, 7, 10, 13, 32).
Although much of MAD1 has been usually simplified as rather stiff coiled coils, the known structures of MAD1 MIM and MAD1 CTD and careful analyses of MAD1 NTD showed that mul-

MAD1 domains in the mitotic checkpoint
tiple MAD1 segments may adopt alternative structures or remain disordered (11,33) (Fig. 1 and Fig. S1). It is likely that liganded C-MAD2 binding to the MIM region of MAD1 still constitutes the catalytic core and represents the MAD2 stably associated with MAD1 in cells (Fig. 1e), but earlier work by others has already lent strong support to possible MAD1 CTD involvement in the mitotic checkpoint responses (12-18, 29, 30). At least part of the MAD1 NTD contributes to MAD1 nuclear pore or kinetochore localization and interactions with other proteins including Ndc80, Plk1, Nek2A, Tpr, Cep57, and CENP-E (20 -23, 33, 45-49), and the regions spanning 400 -500 or 420 -485 could possibly affect the mitotic checkpoint signaling (14,29). Nevertheless, the possibility that MAD1 NTD and MAD1 CTD directly impact the MAD2 O-C conversion was not thoroughly studied until this work.
Our results suggest that a full-length MAD1 molecule may employ NTD and CTD to at least transiently interact with the substrates (O-MAD2), the products (C-MAD2), and even the intermediate states (I-MAD2, represented by MAD2 ⌬N10 in Fig. 3c) (50) of the MAD2 O-C conversion reaction. The interaction with O-MAD2 might increase local concentration of substrates, driving the conversion reaction. Similarly, enrichment of I-MAD2 by MAD1 NTD or MAD1 CTD could facilitate its interaction with CDC20, coupling the MAD2 O-C conversion reaction with the energy favorable assembly of the MCC (29,30). It should be noticed that O-MAD2 conversion to C-MAD2 can certainly be accomplished in the absence of CDC20 (12,13). Following the law of mass action, the NTD or CTD-retained C-MAD2 could also drive its interaction with BUBR1 as a step of MCC assembly (36). These possible scenarios involving MAD1 NTD and MAD1 CTD for C-MAD2 production and MCC assembly may underlie the compromised mitotic checkpoint responses when cells were transfected with mCherry-Mis12-MAD1 ⌬NTD or -MAD1 ⌬CTD (Fig. 1). Consistent with this idea, our data indicated that MAD1 NTD and MAD1 CTD bind to a new interface on MAD2. Such a binding mode is postulated to be advantageous, as the I-MAD2 or C-MAD2 products anchored on MAD1 could have two other protein-protein interaction interfaces, the safety belt (for CDC20) and the dimerization domain (for BUBR1), readily available to assemble the MCC. In addition, the MAD2 interactions with NTD or CTD must be weak or transient in cells, as mCherry-Mis12-MAD1 ⌬NTD and MAD1 ⌬CTD do not show significant reduction in co-immunoprecipitated MAD2 whereas FLAG-MAD1 NTD only shows faint MAD2-binding (Fig. 1, e and f). Such transient interactions are likely also beneficial for stepwise relay mechanisms to facilitate MAD2 O-C conversion and MCC assembly.
In the MAD1 MIM :C-MAD2 tetramer structure (11,32), the dimerization interfaces along the two C-MAD2 molecules to recruit O-MAD2 face toward opposite directions. It is therefore possible that endogenous full-length MAD1, when bound by C-MAD2 at MIM to form a catalyst, contains more than one catalytic center. It remains to be seen whether MAD1 NTD and MAD1 CTD asymmetrically affect the two presumable MAD1 MIM :C-MAD2 catalytic centers or directly contribute to the catalysis through stabilizing some transient I-MAD2 states such as those with unfolded N or C terminus (10,13). In addition, MAD1 CTD and MAD1 NTD are important for MAD1 olig-omerization (Figs. 1, e and f, and 6d). We noticed an earlier publication suggesting that Ser-214 phosphorylation by ATM affects MAD1 dimerization (51). Whether MAD1 dimerization regulates MAD2 conversion is an interesting topic for future exploration.

The impact of mitotic kinases on MAD2 O-C conversion
Whatever the catalytic mechanism the MAD1:C-MAD2 complexes employ to promote MAD2 O-C conversion, the catalysis is likely initiated or enhanced by mitotic checkpoint kinases whose activities have been shown important for the checkpoint, including MPS1, BUB1, and Aurora B (25,26,28,31). These kinases not only help enrich the MAD1:C-MAD2 catalysts to unattached kinetochores, but also directly impact their catalytic efficiency. In this work we have focused on the potential effect of MPS1 on increasing the catalytic efficiency of the MAD1:C-MAD2 complex.
We have demonstrated that MPS1 interacts with and phosphorylates MAD1 NTD and MAD1 CTD (Fig. 4). Phosphoresistant mutants at in vitro MPS1 phosphorylation sites in MAD1 CTD , especially the T716A mutation, compromised the mitotic checkpoint responses in cells and showed reduced interaction with MAD2 in vitro (Fig. 5). No increase in MAD2 interactions with MAD1 CTD was detected either using the phosphomimic T716E mutant or after direct in vitro phosphorylation by MPS1 (Figs. 4 and 5). We also expected to see differential bindings of O-MAD2 and C-MAD2 to MAD1 NTD or MAD1 CTD , especially after MPS1 phosphorylation, but do not have experimental evidence to support the idea yet (Fig. 4). We cannot exclude the possibility that in cells and in the context of full-length MAD1, CTD phosphorylation shows differential binding toward either conformer of MAD2. Synergistic effect with phosphorylation by other kinases such as BUB1 and Aurora B may also be necessary to see such effects. Phosphorylation by MPS1 did reduce its own interaction with CTD and the interaction between NTD and CTD (Figs. 4 and 6).
We propose that the NTD:CTD interaction occurs in interphase cells and represents an inactive state of MAD1 even though its MIM associates with C-MAD2 in a cell cycle-independent manner (18,24). The interaction between MPS1 and MAD1 might contribute to the MAD1 recruitment to kinetochores. Once concentrated at unattached kinetochores, MPS1 kinase becomes activated and phosphorylates BUB1 to stably anchor MAD1 (30,31,57). Activated MPS1 also phosphorylates MAD1 CTD , most likely at Thr-716, relieving itself and the NTD from CTD. The now exposed and phosphorylated CTD could bind to O-MAD2 and C-MAD2 and also facilitate the

MAD1 domains in the mitotic checkpoint
MAD2 O-C conversion either directly or through increased association with CDC20 (30), as discussed in the last section.
Although deleting MAD1 NTD compromised the mitotic checkpoint when examined using the mCherry-Mis12 fusions (Fig. 1), mutating presumable MPS1 phosphorylation sites within MAD1 NTD had no impact on the mitotic checkpoint (Fig. 5), and others found no requirement for MAD1 N-terminal 400 or 420 residues for MAD2 O-C conversion or checkpoint responses (14,29). This suggests that a functional critical region lies within 420 -485 residues, which happens to contain a likely non-coiled-coil structure (Fig. 1). However, neither of the MAD1 fragments spanning 327-423 or 327-488 residues bind to MAD2 in vitro (Fig. 2), indicating the positive effect of MAD1 NTD on the mitotic checkpoint signaling needs further clarification.
Interestingly, Faesen et al. (29) reported that MAD2 can be phosphorylated by MPS1 in vitro at Ser-195. Previously, phosphorylation at this site was proposed to render MAD2 hard to convert and thus more likely to stay in O conformation (43,58). Whether this particular MAD2 phosphorylation by MPS1 represents a feedback regulatory mechanism can be further explored. In this regard, it might be useful to note that MPS1 interacts with C-MAD2 and O-MAD2 whereas MAD2 S195D binds to MAD1 at NTD or CTD but not MIM, behaving indeed like an O-MAD2 (Figs. 3a and 4c).
Although recent work has proposed how BUB1 might help the MAD2 O-C conversion and MCC assembly (29 -31), it will still be interesting to further define how other kinases such as Aurora B regulate the MAD1:C-MAD2 catalyst to increase C-MAD2 production. In addition, there have been reports that some tyrosine kinases help silence the mitotic checkpoint to drive anaphase onset (59 -63). MAD1 Y634 seems to play important roles in the mitotic checkpoint ( Fig. 1 and Fig. S4), and it was once reported to be phosphorylated in vivo (37). Further characterization of this site as well as the apparently essential Ser-610 and Thr-716 residues at the MAD1 CTD will provide more mechanistic insights into the functions of MAD1 in MAD2 O-C conversion.

DNA constructs
The MAD1 and MAD2 DNA fragments and mutants used in the work are summarized in Table S1. The pMSCV-mCherry-Mis12-MAD1 constructs with wild-type or a mutant MAD1 (K541A, L543A) were gifts from Dr. Tarun Kapoor (Rockefeller University) (26,28). Other MAD1 mutants or truncations were cloned or mutated in an intermediate pENTR2B vector (Invitrogen), cut out as a NotI-EcoRI fragment, and then used to replace the MAD1 gene in the pMSCV-mCherry-Mis12 backbone also cut with NotI and EcoRI. Mutagenesis was conducted using QuikChange Lightning Multi Site-Directed Mutagenesis Kit (Agilent). Different fragments of MAD1 were also cloned into a pENTR/D-TOPO vector using TOPO Cloning Kit (Invitrogen) and then recombined into pDEST15 (for N-terminal GST tag) or pDEST17 (for N-terminal His 6 -tag) vectors through LR reactions as instructed for the Gateway cloning system (Invitrogen). MAD2 expression constructs were pre-pared similarly as before (36), but with a tobacco etch virus (TEV) cleavage site inserted between the tags and second MAD2 codon during pENTR cloning steps. MPS1 was recombined into pDEST10 or pDEST20 and transformed into DH10BAC to prepare recombinant bacmids for Sf9 insect cell transfection. All constructs were confirmed by DNA sequencing.

Cell culture and transfection
HeLaM, a subline of HeLa (35), was maintained in DMEM with 10% fetal bovine serum at 37°C in 5% CO 2 . DNA transfection was carried out using TransIT-LT1 reagent (Mirus) following the manufacturer's instructions or using polyethyleneimine (PEI) as described before (64). Sf9 cells were grown at 27°C in SFX medium (Hyclone) in the presence of antibiotics (streptomycin/penicillin). Cellfectin (Invitrogen) was used to transfect bacmids into Sf9 cells.

Live cell imaging
For determining mitotic durations, HeLa cells grown on No. 1.5 coverslip-bottomed 35-mm dishes (MatTek) were transfected with different mCherry-Mis12-MAD1 constructs. Usually 24 h after transfection, live cell imaging was started on an automated Olympus IX-81 microscope to collect phase contrast and RFP images at 15-min intervals using a 60ϫ objective lens (NA ϭ 1.42) while cells were maintained at 37°C in a heating chamber. Single-plane images were acquired for up to 13 h at multiple positions using a CoolSNAP HQ2 camera with 2 ϫ 2 binning. Student's t test was used to evaluate the statistical significance between the differences in mitotic durations after different treatment. Nuclear envelope breakdown marks the beginning and appearance of cleavage furrow the end of a mitosis. Some images were collected on a Leica TCS SP8 confocal microscope with a 63ϫ objective (NA ϭ 1.40) as z-stacks of 1.0 m.

Cell lysates, immunoblotting, and immunoprecipitation
These were performed as described before (28,64). A list of primary antibodies used in this study is summarized in Table S2.

Recombinant proteins
GST-tagged or His-tagged MAD1 fragments and His-TEVtagged MAD2 L13A or His-TEV-tagged MAD2 ⌬C10 were expressed in Escherichia coli BL21(DE3)-CodonPlus RIPL (Stratagene), normally at 16°C. His-MPS1 or GST-MPS1 was expressed in Sf9 cells after infection with recombinant baculoviruses. All expressed proteins were purified using GSHagarose or Probond nickel beads (Invitrogen). His-tagged TEV(S219P) protease (65) was used to cleave His-tag to make untagged MAD2. Peak fractions of eluted proteins were pooled, buffer-exchanged, and concentrated using Pierce protein concentrators with 10,000 molecular weight cutoff. The storage buffer is 50 mM Hepes pH7.5, 100 mM KCl, 1 mM DTT, 30% glycerol. Concentrations of recombinant proteins were determined by comparing the target band with BSA standards on Coomassie Blue-stained gels using ImageJ software (66).

MAD1 domains in the mitotic checkpoint
In vitro binding assays 4 l of 5ϫ binding buffer (100 mM Tris-HCl, pH 8.0, 750 mM NaCl, 50 mM MgCl 2 , 2.5% Nonidet P-40, 50% glycerol, 500 g/ml BSA) was mixed with recombinant GST-tagged MAD1 fragments and MAD2 mutants or His-MAD1 CTD or His-MPS1. The proteins (including related fragments or mutants) were used at roughly endogenous intracellular concentrations unless stated otherwise in the figure legends: [MAD2] ϭ 230 nM, [MPS1] ϭ 100 nM, and [MAD1] ϭ 60 nM (8,(67)(68)(69). H 2 O was added to make the final volume 20 l. The reactions were incubated at 37°C for 1 h and then mixed with 10 l GSH agarose beads and shaken at 800 rpm at 4°C for 40 min (Peqlab Thriller). The beads were pelleted and washed four times with wash buffer (20 mM phosphate buffer, pH 7.4, 250 mM NaCl, 0.5% Nonidet P-40, 10% glycerol). Then 10 l 2ϫ SDS sample buffer was added to the beads and the samples were heated at 80°C for 10 min before SDS-PAGE. The gel was transferred to PVDF membranes (Millipore) for immunoblotting. Most of the in vitro binding results have been replicated three times or more.

In vitro kinase assays
GST-tagged MPS1 kinase was purchased from Invitrogen or was purified from Sf9 cell lysates as well as His-tagged MPS1 (64). Myelin basic protein was purchased from Sigma as an artificial substrate for MPS1. In vitro kinase assays were set up similarly as described previously (28,64): 4 l of 5ϫ kinase buffer (125 mM Tris-HCl, pH 7.5, 300 mM ␤-glycerophosphate, 50 mM MgCl 2 ) was mixed with recombinant kinase, substrates, 5Ci 32 P ATP and 50 M cold ATP. In some reactions, MPS1 kinase inhibitors reversine (Calbiochem) or AZ3146 (Selleckchem) was used at 500 nM and 2 M final concentrations, respectively. DMSO concentration is kept below 0.5%. H 2 O was added to make the final volume 20 l. The reactions were incubated at 30°C for 30 min and then terminated by adding 20 l 2ϫ SDS sample buffer. Samples were subjected to SDS-PAGE followed by Coomassie Blue staining. After destaining, the SDS-PAGE gel was vacuum dried. Phosphorylation of the substrates was visualized by autoradiography. Samples for mass spectrometry were prepared using only 0.5 mM cold ATP in the kinase reactions and separated on SDS-PAGE. Phosphorylated residues on the excised bands were determined by mass spectrometry (MS Bioworks, Ann Arbor, MI).
For experiments shown in Fig. 4, d and e, in vitro kinase assays were performed by incubating GST-MPS1 and His-MAD1 CTD in the presence or absence of 0.5 mM cold ATP and 25 M reversine at 30°C for 1 h. Then untagged MAD2 L13A or MAD2 ⌬C10 were directly added into the reaction mixtures and incubated at 37 o for another hour before immunoprecipitation with 1 g anti-His 6 antibody.