Spindle Checkpoint Protein Xmad1 Recruits Xmad2 to Unattached Kinetochores

Demembranated frog sperm nuclei, as well as cytostatic factor (CSF)- arrested and cycling egg extracts were prepared as described (Murray, 1991), except that dejellied eggs were activated by incubation with the calcium ionophore A23187 at 0.5 μg/ml for 4–5 min at room temperature. After activation, eggs were incubated at room temperature for another 15–20 min to ensure complete inactivation of CSF activity. Activation of the spindle checkpoint in CSF-arrested extracts was performed as described previously (Minshull et al., 1994; Chen et al., 1996). To examine the effect of recombinant 6H-Xmad2 (Chen et al., 1996), extracts were incubated on ice for 1 h with the recombinant protein at concentrations indicated in the figures before the addition of any nuclei or nocodazole.

The maintenance of the Xenopus fibroblast cell line XTC and human HeLa cells as well as the preparation the cell lysates were performed as described (Chen et al., 1996).

Proteins were transferred to nitrocellulose membrane after SDS-PAGE. The membrane was first preblocked with PBS (2.7 mM potassium chloride, 137 mM sodium chloride, 1.5 mM potassium phosphate, 4.3 mM sodium phosphate, pH 7.2) containing 2% BSA and 0.2% Tween-20, and then probed with affinity-purified anti-Xmad1 (1 μg/ml) or with anti-Xmad2 (0.5 μg/ml) antibodies in the same buffer for 1 h at room temperature. The blot was washed three times with PBST (PBS + 0.2% Tween-20), incubated with peroxidase-conjugated anti–rabbit antibody (Amersham Corp., Arlington Heights, IL) for 30 min in PBST, washed three times with PBST, and then developed by enhanced chemiluminescene detection (ECL; Amersham Corp.).

Xmad1 was purified from Xmad2 immunoprecipitates. To immunoprecipitate Xmad2, Affi-Prep protein A support beads (Bio-Rad Laboratories, Hercules, CA) were washed twice with 1 ml of extract buffer (XB; 10 mM Hepes, pH 7.8, 50 mM sucrose, 100 mM potassium chloride, 10 mM magnesium chloride, 1 mM calcium chloride) containing 10 μg/ml each of leupeptin, pepstatin, and chymostatin (LPC). The beads were then incubated with 5 μg of affinity-purified anti-Xmad2 antibodies (Chen et al., 1996) or with control IgG (Harlow and Lane, 1988) at 4°C for 30 min. After washing the beads three times with 1 ml of XB plus LPC, 500 μl of CSF- arrested egg extracts were added to the antibody-coated beads and gently mixed on a mixer at 4°C for 1 h. The Xmad2 protein complex was isolated by centrifugation at 16,000 g for 1 min, and the immunoprecipitates were washed five times with 1 ml of XB plus LPC and 1% Triton X-100 or 0.1% SDS. Proteins bound to the antibodies were solubilized in SDS sample buffer, resolved by SDS-PAGE, and then visualized by Coomassie blue staining. The 85-kD protein that was specifically isolated in the Xmad2 immunoprecipitates was then subject to mass spectrometry analysis to determine its sequence.

The protein band was excised from the gel and in-gel digested with trypsin (unmodified, sequencing grade; Boehringer Mannheim Corp., Indianapolis, IN) as described (Shevchenko et al., 1996). Digestion buffer contained 50% (vol/vol) H₂¹⁸O. H₂¹⁸O was purchased from Cambridge Isotope Laboratories (Andover, MA) and was additionally purified by microdistillation using a sealed glass microdistillator constructed in house. The unseparated pool of tryptic peptides recovered from the gel matrix was sequenced on a experimental prototype of a hybrid quadrupole Time-Of-Flight tandem mass spectrometer developed by PE Sciex (Ontario, Canada) and University of Manitoba (Winnipeg, Canada) (Shevchenko et al., 1997). The mass spectrometer was equipped with a nanoelectrospray ion source developed in house (Wilm and Mann, 1996).

Eight different peptide sequences were obtained from mass spectrometry (see Fig. 5). Degenerate oligonucleotides were made corresponding to the sequences EDQAAQ, TPTEH, AQEQDAA, EDNTTV, NDSEDNT, and DAAI/LVK (mass spectrometry can not differentiate I from L). All oligonucleotides were made in both orientations, and used in PCR reactions in all possible pairwise combinations (10 pmol/μl each) using a Xenopus egg cDNA library (20 ng/μl, a gift of J. Minshull, formerly at University of California, San Francisco) as templates. The 550-bp PCR product generated from a reaction containing oligonucleotides for the peptides EDQAAQ and AQEQDAA was successfully cloned into the plasmid pBLUESCRIPT KS(−) (Stratagene, La Jolla, CA). The PCR product was sequenced to verify its authenticity, and was used to generate a probe by PCR using digoxigenin-labeled nucleotides (Boehringer Mannheim Corp.). The digoxigenin-labeled probe was then used to screen a Xenopus ovary library (a gift of J. Minshull). The phage plaques were lifted with nylon filters (Hybond-N; Amersham Corp.) as described (Sambrook et al., 1989). Phage DNA was denatured by autoclaving the filters for 1 min. Filters were washed briefly in 0.15 M sodium phosphate, pH 7.0, with 2% SDS, and then preblocked for 30 min at 65°C in hybridization buffer (0.5 M sodium phosphate, pH 7.0, 1% BSA, 7% SDS, 1 mM EDTA). Filters were then probed with the digoxigenin-labeled probe at 50 ng/ml in hybridization buffer at 65°C for 16 h. After hybridization, the filters were washed three times with 0.15 M sodium phosphate, 2% SDS at 65°C for 10 min each, three times with TTBS (25 mM Tris, pH 7.5, 0.5 M NaCl, 0.1% Tween 20) at room temperature for 5 min each, and once with TTBS containing 4% nonfat dry milk. The filters were then probed for 1 h with alkaline phosphatase-labeled anti-digoxigenin antibodies (Boehringer Mannheim Corp.) at 1:5,000 dilution in TTBS plus 4% milk. After washing the filters three times with TTBS for 10 min each and once with alkaline phosphatase color development buffer (100 mM sodium chloride, 5 mM magnesium chloride, 100 mM Tris-Cl, pH 9.5), the filters were then developed with 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium (BCIP/NBT) as described (Sambrook et al., 1989). The positive clones were picked and used for secondary screening with the same procedures. Three clones of 1, 1.4, and 1.6 kb were isolated and the largest one was sequenced using Sequenase Version 2.0 (United States Biochemical Corp., Cleveland, OH).

The full-length Xmad1 coding sequence was tagged with six histidines at the NH₂-terminus, and expressed in Escherichia coli using the Qiaexpress vector pQE9 (QIAGEN Inc., Valencia, CA). Protein expression was induced with 0.1 mM isopropyl β-d-thiogalactopyranoside (IPTG) for 4 h at 37°C. Bacteria were collected by centrifugation, resuspended, and then sonicated in sonication buffer (50 mM sodium phosphate, pH 8.0, 300 mM NaCl, 0.5% Triton X-100, 1 mg/ml lysozyme, 1 mM PMSF, 10 μg/ml LPC). The inclusion bodies containing Xmad1 were pelleted by centrifugation at 8.5 krpm for 30 min at 4°C. After extensive washing with high salt buffer (10 mM Tris, pH 7.2, 100 mM sodium phosphate, pH 7.2, 0.5 M NaCl), RIPA buffer (10 mM Tris, pH 7.2, 150 mM NaCl, 1% Triton X-100, 0.1% SDS, 1 mM sodium orthovanadate), and water, the inclusion bodies were then solubilized in SDS sample buffer and resolved by SDS-PAGE. The gel slice containing Xmad1 band was diced, and the protein was eluted into a buffer of 10 mM Hepes, pH 7.2, and 1 mM DTT. The purified protein was used to raise antisera in rabbits and mice (Berkeley Antibody Co., Richmond, CA), and to affinity purify antibodies. Affinity purification of anti-Xmad1 antibodies was performed as described (Chen et al., 1992). Control IgG was purified from preimmune serum over a protein A column made from Affi-Prep Protein A (156-0006; Bio-Rad Laboratories) as described (Chen et al., 1992). Purified antibodies were dialyzed and concentrated in Antibody Dilution Buffer (10 mM sodium phosphate, pH 7.4, 100 mM potassium chloride, 1 mM magnesium chloride, 50% glycerol).

Immunofluorescent staining of Xmad1 and Xmad2 in XTC cells was performed as described (Chen et al., 1996). Assembly of metaphase chromosomes in egg extracts was performed with an in vitro metaphase to anaphase transition assay as described (Chen and Murray, 1997; Shamu and Murray, 1992). In brief, 20 μl of CSF-arrested extracts were induced to enter interphase by adding calcium chloride, and sperm nuclei (1,500 per μl of extract) were allowed to replicate in these interphase extracts for 80–90 min at 23°C. One half volume (10 μl) of CSF-arrested extracts was then added to drive the extracts into mitosis, and the reactions were incubated for another 80-90 min to establish metaphase arrest. For experiments shown in Fig. 3, 6H-Xmad2 protein was supplemented at 0.1 μg/μl after the addition of CSF-arrested extracts. After metaphase arrest was achieved, calcium chloride was added to inactivate the CSF activity and to induce anaphase. To examine the nuclear and spindle morphology, 1 μl of each sample was removed and fixed in Fix solution (1.1% formaldehyde, 48% glycerol, 1 mg/ml Hoechst 33258 made in MMR [100 mM sodium chloride, 2 mM potassium chloride, 1 mM magnesium sulfate, 2 mM calcium chloride, 5 mM Hepes, 0.1 mM EDTA]). Specimens were viewed with a Nikon Microphot-FXA microscope equipped with a 60×, 1.4 NA oil immersion objective, and photographed with Kodak Tri-X pan 400 film.

To study the effect of antibodies on Xmad1 and Xmad2 localization as shown in Fig. 9, anti-Xmad1 (30 ng/μl of extracts) or anti-Xmad2 (8 ng/μl of extracts) antibodies were added immediately after the addition of CSF-arrested extracts in the in vitro metaphase to anaphase assay. After metaphase arrest was established, the mitotic spindle was disrupted by adding nocodazole to 10 μg/ml. (To avoid inactivation of experimental extracts by excess DMSO, a 10 mg/ml stock of nocodazole made in DMSO was first diluted 50-fold in a separate aliquot of extract before being added to the samples.) After 10 min of incubation, reactions were diluted with 9 vol of XB plus 0.5% Triton X-100, and incubated at room temperature for another 10 min. Samples were then layered over 5 ml of 30% glycerol cushion made in XB plus 0.5% Triton X-100 with a coverslip placed at the bottom, and the chromosomes were collected onto the coverslip by centrifugation at 10 krpm in a HB-4 rotor (Sorvall, Newtown, CT) for 10 min at 4°C. Chromosomes were fixed in PBS containing 1% formaldehyde for 10 min. Immunofluorescent staining of Xmad1 and Xmad2 was performed as described for cultured cells (Chen et al., 1996). For Figs. 8 and 9, images were collected using a charge-coupled device (CCD) camera (KAF1400 chip, 5 MHz controller; Princeton Laboratories, Inc., Princeton, NJ) attached to a fluorescence microscope (model BX50; Olympus America, Lake Success, NY). Images were collected and processed with the Metamorph Imaging System (version 3.0; Universal Imaging Corporation, West Chester, PA) and converted to Photoshop format (Adobe Systems Inc., Mountain View, CA).

For immunodepletion of Xmad1 or Xmad2, 15 μl of Affi-Prep protein A beads were washed twice in 1 ml of XB plus LPC. The beads were then mixed with affinity-purified anti-Xmad1 (12 μg) or anti-Xmad2 (3 μg) antibodies on a rocker for 30 min at 4°C. After washing the beads twice with 1 ml of XB plus LPC, as much residual buffer was removed as possible to prevent dilution of the extracts. The beads were further rinsed quickly with 60 μl of egg extracts, spun at 16,000 g in a microcentrifuge for 1 min at 4°C, and supernatants were removed. 60 μl of extracts were then incubated with the antibody-coated beads for 1 h on a rocker at 4°C. At the end of the incubation, the beads were spun at 16,000 g in a microcentrifuge for 1 min at 4°C. The supernatants were removed and subjected to another centrifugation to avoid any contamination with the beads. This procedure removed >95% of Xmad1 or Xmad2 protein from the extracts. For mock depletion, the beads were prepared in the same manner, except that control IgG was used.

Mature eggs from the frog Xenopus laevis are naturally arrested at metaphase of meiosis II by CSF (Sagata et al., 1989), and are released from this arrest by a transient increase in the cytoplasmic calcium concentration that is triggered by fertilization. Extracts made from unfertilized eggs maintain the CSF arrest and can be induced to enter anaphase by adding calcium to inactivate CSF activity, leading to degradation of cyclin B and inactivation of Cdc2. When the spindle checkpoint is activated in these extracts, by adding a high density of sperm nuclei and nocodazole, the mitotic arrest cannot be overcome by calcium addition (Minshull et al., 1994). We have previously shown that adding a neutralizing anti-Xmad2 antibody overcomes the spindle checkpoint (Chen et al., 1996), and we expected that immunodepletion of Xmad2 would also inactivate the checkpoint. Fig. 1,B shows that egg extracts depleted of Xmad2 were not able to activate the checkpoint; despite the presence of a high concentration of sperm nuclei and nocodazole, calcium addition inactivated the histone H1 kinase activity of Cdc2 and led to the formation of interphase nuclei (data not shown), indicating that Xmad2 and/or its associated proteins are required for the checkpoint. If Xmad2 was the only checkpoint component removed from the egg extracts during immunodepletion, adding recombinant Xmad2 should restore the checkpoint. When recombinant Xmad2 was added back to depleted extracts to restore the level of Xmad2 to normal levels, egg extracts failed to maintain a mitotic arrest (Fig. 1,B), suggesting that some additional checkpoint components were also removed during Xmad2 immunodepletion, or that recombinant Xmad2 protein was not fully functional. When recombinant Xmad2 protein was added to higher levels, it prevented exit from mitosis in the presence of nocodazole (Fig. 1,B). In addition, it also blocked the inactivation of H1 kinase activity in the absence of nocodazole, a condition where the checkpoint is not normally active (Fig. 1,C). This effect of recombinant Xmad2 is dose-dependent. At low concentrations, Xmad2 partially maintained H1 kinase activity. At 100 μg/ml Xmad2, ∼17-fold the endogenous level of Xmad2, inactivation of H1 kinase was completely inhibited (Fig. 1 C) and the added sperm nuclei remained condensed (data not shown), indicating that the extracts remained arrested in mitosis. These results suggest that recombinant Xmad2 protein is functional and at high levels induces mitotic arrest. The finding that overexpression of Mad2 in fission yeast arrests cells in mitosis (He et al., 1997; Kim et al., 1998) is consistent with our results and argues that the effects of high levels of Xmad2 are due to an increased concentration of the protein rather than an aberrant activity of the recombinant protein.

The inability of calcium addition to trigger mitotic exit in CSF-arrested extracts in the presence of excess Xmad2 is similar to the phenomenon seen after activation of the spindle checkpoint. We examined the effect of excess Xmad2 in various conditions that normally do not activate the checkpoint. When extracts were incubated with a low density of nuclei or with nocodazole alone or in the absence of both nuclei and nocodazole, excess Xmad2 prevented calcium from inactivating H1 kinase activity (Fig. 2,A). The persistently high level of H1 kinase activity correlated with a high level of cyclin B (Fig. 2,A), suggesting that Xmad2 inhibits Cdc2 inactivation by preventing cyclin B degradation. These results show that excess Xmad2 prevents mitotic exit under conditions where the spindle checkpoint is normally not active. The effect of excess Xmad2 was also observed in cycling extracts prepared from activated eggs, which lack CSF activity (Fig. 2 B), demonstrating that excess Xmad2 has the general effect of preventing mitotic exit rather than simply inhibiting the ability of calcium to overcome the metaphase arrest induced by CSF.

At the onset of anaphase, ubiquitin-mediated protein degradation triggers the destruction of cyclin B (Glotzer et al., 1991; Hershko et al., 1991; King et al., 1995; Sudakin et al., 1995) and proteins involved in sister chromatid cohesion (Holloway et al., 1993; Surana et al., 1993; Cohen-Fix et al., 1996; Funabiki et al., 1996). Because sister chromatid separation is independent of cyclin B degradation and inactivation of Cdc2 (Holloway et al., 1993), we used egg extracts to test whether excess Xmad2 inhibits degradation of the proteins involved in sister chromatid cohesion as well as of cyclin B. Extracts can be induced to replicate the DNA of sperm nuclei and then to enter a metaphase arrest that can be released by adding calcium (Shamu and Murray, 1992). In control extracts, the sister chromatids started to separate 10–15 min after calcium addition, and chromosomes decondensed at 40–60 min as extracts entered interphase (Fig. 3). The presence of excess Xmad2 prevented calcium from inducing anaphase (Fig. 3). The intact mitotic spindle and alignment of chromosomes at the metaphase plate indicated that excess Xmad2 did not disrupt spindle structure or interfere with kinetochore attachment to the spindle. Thus excess Xmad2 prevents three hallmarks of the metaphase to anaphase transition: Cdc2 inactivation, cyclin B degradation, and sister chromatid separation. This mitotic arrest is independent of any chromosome component or microtubule perturbation (Fig. 2 A), suggesting that excess Xmad2 constitutively activates the spindle checkpoint.

The failure of exogenous Xmad2 at normal concentration to restore the checkpoint activity in Xmad2-depleted extracts (Fig. 1,B) suggested that other checkpoint components bound to Xmad2 and were removed during the immunodepletion. To test this possibility, we examined the anti-Xmad2 immunoprecipitates for Xmad2-associated proteins. Immunoprecipitates of Xmad2 from CSF-arrested extracts contained an 85-kD protein (p85) that was absent in immunoprecipitates made with a control antibody (Fig. 4,A). This protein was still associated with Xmad2 when the immunoprecipitates were washed with a buffer containing 0.1% SDS (Fig. 4 A), indicating that it formed a tight complex with Xmad2.

We sequenced p85 by mass spectrometry. To simplify the readout of the peptide sequences the protein was in-gel digested in a buffer containing H₂¹⁶O and H₂¹⁸O in equal proportion. The digestion resulted in characteristic isotopic labeling of COOH-terminal carboxyl groups of tryptic peptides (Fig. 4,B, a). The unseparated protein digest was then analyzed by nanoelectrospray tandem mass spectrometry using a novel hybrid Quadrupole Time-Of-Flight instrument. Eight peptide ions originating from p85 were fragmented in the mass spectrometer (Fig. 4,B, b). Seven tryptic peptides were sequenced completely and one peptide was sequenced partially, covering altogether 90 amino acid residues (Fig. 5,A). We used this sequence information to isolate Xenopus cDNA clones that encoded p85. The largest cDNA clone we isolated had an open reading frame of 718 amino acids and contained exact matches to six of the eight peptide sequences (Fig. 5,A). Two of the peptide sequences contained mismatches at residues Gln³²³Ser³²⁴ and Gly⁶⁵⁶Glu⁶⁵⁷. Because these residues are conserved in a recently isolated human Mad1 homologue (Jin et al., 1998) that shares 57.9% identity with p85, these mismatches likely reflect errors in mass spectrometry resulting from low intensity of the particular fragment ions. The sequence of Xmad1 predicts that much of the protein forms a coiled coil. In budding yeast we found that Mad2 tightly associated with Mad1 (Chen, R.-H., K. Hardwick, and A. Murray, unpublished result), which is also predicted to be a coiled-coil protein. Although sequence alignment showed that Mad1 and p85 shared only 19% identity, the sequences of p85 and Mad1 gave similar coiled-coil plots (Fig. 5,B), indicating that they are alike in structure. The human Mad1 homologue also associated with the human Mad2 (Jin et al., 1998) and strikingly resembled p85 in coiled-coil plots (Fig. 5,B). Based on these characteristics, we predict that p85 is the Xenopus homologue of Mad1, and refer to it as Xmad1. A search in the data base also revealed that a Schizosaccharomyces pombe Mad1 homologue (these sequence data are available from GenBank/EMBL/DDBJ under accession No. Z95620) shared 23.9% identity with Xmad1 (Fig. 5 A).

To investigate the function of Xmad1, we generated antiserum against full-length recombinant Xmad1. The affinity-purified antibody specifically recognized an 85-kD protein in frog egg extracts (Fig. 6,A), which was also present in Xmad2 immunoprecipitates (Fig. 6,A), consistent with the idea that this protein is the Xmad2-binding protein that we originally identified. The antibody also recognized an 84-kD protein in the frog cultured cell line XTC (Fig. 6,A), indicating that the somatic Xmad1 may be a different isoform or modified differently from the protein in oocytes. In HeLa cells, the antibody recognized an 86-kD Mad2-binding protein (Fig. 6,A), indicating the presence of a Xmad1 homologue in human cells. The anti-Xmad1 antibodies were used to examine the abundance of Xmad1 throughout the cell cycle in egg extracts. Like Xmad2, Xmad1 level remained unchanged when the CSF-arrested extract was induced to exit mitosis by calcium addition (Fig. 6,B). The protein level also stayed constant in an extract that went through two cell cycles (Fig. 6 C). These results demonstrate that Xmad1 and Xmad2 levels are not regulated during the cell cycle, including at the metaphase to anaphase transition, and that the proteins are not subject to proteolysis at specific cell cycle stages. These findings are similar to those seen in budding yeast, in which Mad1 (Hardwick and Murray, 1995) and Mad2 (Chen, R.-H., unpublished results) levels are constant throughout the cell cycle.

We used the anti-Xmad1 antibodies to determine whether Xmad1 is involved in the spindle checkpoint in frog egg extracts. When an extract was preincubated with anti-Xmad1 antibodies, it failed to establish mitotic arrest in the presence of a high concentration of sperm nuclei and nocodazole, indicating that the antibody interfered with the spindle checkpoint (Fig. 7,A). This effect was specific to Xmad1, because antibody preblocked with recombinant Xmad1 had no effect on the checkpoint (Fig. 7,A). When added to egg extracts after the checkpoint was first activated, the antibody still abolished the checkpoint function (Fig. 7,B). These results show that, like Xmad2 (Fig. 7 B; Chen et al., 1996), Xmad1 is critical for both establishing and maintaining the spindle checkpoint.

Immunofluorescent staining of Xmad1 in XTC cells revealed that Xmad1 and Xmad2 localized to the nuclear envelope and the nucleus during interphase (Fig. 8,A). At prophase, Xmad1 dissociated from the nuclear envelope before all of the Xmad2 did, and a fraction of both Xmad1 and Xmad2 localized to kinetochores (Fig. 8,B). The kinetochore staining of both anti-Xmad1 and anti-Xmad2 antibodies persisted at prometaphase (Fig. 8,C), and the staining was absent from the kinetochores at metaphase and anaphase (Fig. 8, D and E). In cells arrested at mitosis by nocodazole, both anti-Xmad1 and anti-Xmad2 antibodies stained the kinetochores (Fig. 8 F). The colocalization of Xmad1 and Xmad2 likely reflects the binding of a Xmad1/ Xmad2 complex to the kinetochores. These results suggest that Xmad1 is part of the complex that senses unattached kinetochores and activates the spindle checkpoint.

We also investigated the association of Xmad1 and Xmad2 with the kinetochores of chromosomes assembled in egg extracts. We first allowed frog egg extracts to assemble metaphase spindles and then added nocodazole to the extracts to induce microtubule depolymerization. When chromosomes were isolated during the metaphase arrest, neither Xmad1 nor Xmad2 was found on the kinetochores (Fig. 9,A), whereas anti-Xmad1 and anti-Xmad2 antibodies both stained paired dots on chromosomes that had been isolated after nocodazole addition (Fig. 9,B). The punctate staining colocalized with staining from anti– CENP-E antibodies (data not shown), indicating that it represented kinetochore staining. These results are consistent with the in vivo data that both Xmad1 and Xmad2 bind kinetochores that are not associated with microtubules, and dissociate after microtubules attach to kinetochores. We next asked whether the anti-Xmad1 and anti-Xmad2 antibodies that block the checkpoint in egg extracts have an effect on the localization of these proteins. When antibodies made against full-length Xmad2 (Chen et al., 1996) were added before nocodazole, both Xmad1 and Xmad2 still associated with kinetochores (Fig. 9,D), showing that our anti-Xmad2 antibodies did not prevent Xmad1 or Xmad2 from binding to kinetochores. This result demonstrates that binding of Xmad1 and Xmad2 to kinetochores is not sufficient to activate the checkpoint, and suggests that Xmad2 interacts with proteins other than Xmad1 at kinetochores and that our anti-Xmad2 antibodies blocked this interaction. On the other hand, adding anti-Xmad1 antibodies kept both Xmad1 and Xmad2 from localizing to kinetochores (Fig. 9 C), suggesting that anti-Xmad1 antibodies abolish the checkpoint by preventing Xmad1 from binding to unattached kinetochores. Even when Xmad1 and Xmad2 were first allowed to bind to unattached kinetochores, the addition of anti-Xmad1 antibodies displaced both proteins from these sites (data not shown). These results indicate that Xmad1 is essential for kinetochore localization of Xmad2 and that Xmad1 must reside on kinetochores to maintain the checkpoint.

We next investigated the dependence of Xmad2 function on Xmad1. If the major role of Xmad1 were to recruit Xmad2 to unattached kinetochores, Xmad1 would be dispensable when the checkpoint is constitutively activated by a high level of Xmad2. We tested this hypothesis by immunodepleting Xmad1 from egg extracts. When >95% of Xmad1 was removed (Fig. 10,A), the extracts were unable to activate the checkpoint in response to microtubule depolymerization at high concentrations of sperm nuclei (Fig. 10,B), even though 60% of Xmad2 remained in these extracts (Fig. 10,A). However, addition of excess recombinant Xmad2 could still trigger mitotic arrest in Xmad1-depleted extracts (Fig. 10,B). Similarly, when >95% of Xmad1 was removed by Xmad2-immunodepletion (Fig. 10,A), recombinant Xmad2 also induced a mitotic arrest (Fig. 10 B). In addition, the dose response curve to Xmad2 in extracts depleted for Xmad1 was the same as that in mock-depleted extracts (data not shown), consistent with the notion that the effect of excess Xmad2 is independent of Xmad1. Taken together, these observations suggest that Xmad1 recruits Xmad2 to the kinetochore but is not involved in signaling downstream of Xmad2.

We have identified a Xenopus homologue of the budding yeast Mad1 protein and shown that it is a component of the spindle checkpoint. Xmad1 was isolated as a protein that coimmunoprecipitated with Xmad2. When Xmad2 is depleted from extracts, adding back physiological doses does not restore the checkpoint. This observation suggests that other components of the checkpoint bind to Xmad2 and are removed with it from the extract upon immunodepletion. The observation that Xmad1 coimmunoprecipitates with Xmad2 strongly suggests that the additional depleted component is Xmad1, although we cannot exclude the possibility that there are other checkpoint components bound to Xmad2. One protein that has been reported to bind to Mad2 is Cdc20 (Fang et al., 1998; Hwang et al., 1998; Kallio et al., 1998; Kim et al., 1998), a protein that is necessary for the metaphase to anaphase transition (Dawson et al., 1995; Visintin et al., 1997; Kallio et al., 1998). Since Xmad2-depleted extracts can still exit from mitosis (data not shown), they must still contain functional Cdc20.

As expected, Xmad1 shares several characteristics with Xmad2. First, Xmad1 is important for establishing and maintaining the spindle checkpoint in egg extracts. Second, Xmad1 localizes to the nuclear envelope and the nucleus during interphase, and a fraction of the protein binds to the kinetochores of chromosomes that are not attached to microtubules during prophase and prometaphase. Despite these similarities and the tight interaction between these two proteins, Xmad1 and Xmad2 appear to play distinct roles in the checkpoint. Our studies indicate that Xmad1 likely recruits Xmad2 to unattached kinetochores, where Xmad2 is converted into a form that can prevent the onset of anaphase.

Our results extend previous experiments on the function of Mad1 and Mad2 in the spindle checkpoint. Experiments in budding yeast show that Mad1 and Mad2 bind tightly to each other (Chen, R.-H., K. Hardwick, and A. Murray, unpublished results), and that the presence of Mad2 is required for the hyperphosphorylation of Mad1 that correlates with the activation of the checkpoint (Hardwick and Murray, 1995). Mad2 has not been localized in yeast cells, and Mad1 shows a punctate localization within the nucleus (Hardwick and Murray, 1995). It remains a possibility that some of the punctate staining is at kinetochores. We now demonstrate that Xmad1 binds to unattached kinetochores in both tissue culture cells and in frog egg extracts. This is the same pattern of localization for vertebrate homologues of Mad2 (Figs. 8 and 9; Chen et al., 1996; Li and Benezra, 1996), Bub1 (Taylor and McKeon, 1997), Bub3 (Taylor et al., 1998), and Mad3 (Taylor et al., 1998). The result that anti-Xmad1 antibodies block the association of both Xmad1 and Xmad2 with the kinetochores in egg extracts suggests that the binding of Xmad2 to unattached kinetochores is dependent on Xmad1. Even when Xmad1 and Xmad2 are allowed to bind to unattached kinetochores first, the subsequent addition of anti-Xmad1 antibodies removes both proteins from kinetochores (data not shown), consistent with the idea that both proteins must reside on kinetochores to maintain the checkpoint. Immunodepletion experiments showed that almost all of the Xmad1 is bound with Xmad2, whereas only 40% of Xmad2 is bound with Xmad1 (Fig. 10 A), suggesting that Xmad1 is the limiting factor in the Xmad1–Xmad2 complex.

Although the spindle checkpoint is normally triggered by unattached kinetochores, we have demonstrated that a high level of Xmad2 in frog egg extracts also inhibits the metaphase to anaphase transition under conditions where the checkpoint is not normally activated. Excess Xmad2 blocks three hallmarks of anaphase: degradation of cyclin B, inactivation of Cdc2, and sister chromatid segregation. Consistent with our result, overexpression of Mad2 homologue in fission yeast also blocks anaphase onset (He et al., 1997; Kim et al., 1998), and addition of human Mad2 protein to frog egg extracts also prevents cyclin B degradation (Li et al., 1997; Fang et al., 1998). Furthermore, deletion of as few as 10 amino acids from COOH terminus of Xmad2 abolishes the ability of excess protein to induce metaphase arrest (data not shown). Introduction of a high level of the recombinant human Mad2 of similar truncation also fails to induce a mitotic arrest in Xenopus embryos and to inhibit ubiquitination of cyclin B in vitro (Fang et al., 1998). Taken together, these results indicate that the phenotypes we observed were not due to some aberrant activity of the recombinant protein. We can rule out several possibilities for how excess Xmad2 induces metaphase arrest. The arrest cannot be dependent on binding of Xmad2 to kinetochores, since it occurs in extracts without any nuclei. In addition, it is unlikely that Xmad2 acts as a competitive inhibitor of proteolysis mediated by the anaphase-promoting complex (APC), because Xmad2 lacks any trace of the destruction boxes found in other APC substrates (Glotzer et al., 1991; review in Hershko, 1997 and Cohen-Fix and Koshland, 1997) and Xmad2 levels remain constant during the cell cycle (Fig. 6).

Increases in the level or activity of a number of proteins other than Mad2 can also activate the checkpoint in cells that have normal spindles. Adding an active form of Ste11, a budding yeast MAPK kinase family member, arrests frog egg extracts at mitosis (Takenaka et al., 1997). In budding yeast, overexpression of Mps1, the kinase that is thought to phosphorylate Mad1, arrests cells at mitosis without spindle defects (Hardwick et al., 1996), and overexpression of Mph1, a S. pombe homologue of Mps1, also induces metaphase arrest (He et al., 1998). All these observations suggest that the activity or the level of the spindle checkpoint components must be tightly regulated in order for the metaphase to anaphase transition to progress properly.

How is Xmad2 activated by unattached kinetochores and how does it inhibit the onset of anaphase? We speculate that once Xmad1 binds to kinetochores and activates its associated Xmad2, Xmad2 is converted into a form that no longer binds Xmad1 and thus diffuses throughout the cell preventing the onset of anaphase. The free, inactive Xmad2 can then interact with and become activated by kinetochore-bound Xmad1, making the kinetochore a catalytic center for activating the checkpoint (Fig. 11). The active molecules of Xmad2 would be slowly inactivated, thus ensuring that the checkpoint would be turned off once all the kinetochores had bound to microtubules. The molecular basis of Xmad2 activation is obscure since we have been unable to detect posttranslational modifications on this protein (unpublished data). One possible form of activation is changes in the state of Xmad2, as indicated by the finding that monomers and tetramers of the recombinant human Mad2 protein differ in their ability to inhibit cyclin degradation in egg extracts (Fang et al., 1998).

The most likely target of Xmad2 is Cdc20. In budding yeast, Cdc20 binds to Mad2 and Mad3 proteins, and mutations in Cdc20 that block this binding act as dominant checkpoint-defective mutants (Hwang et al., 1998), suggesting that Cdc20 is the target of the checkpoint. Similar results have also been obtained in fission yeast (Kim et al., 1998). Cdc20 homologues also mediate the association of Mad2 with APC in vertebrates (Gorbsky et al., 1998; Fang et al., 1998). A simple model is that interacting with unattached kinetochores converts Xmad2 into a form that can interact and inhibit Cdc20. In the absence of kinetochores, Xmad2 may assume the active conformation at a very slow rate, but this small amount of active Xmad2 cannot efficiently inhibit Cdc20 function. As the level of Xmad2 is increased, the concentration of the active form rises by mass action until it is sufficient to inhibit Cdc20, thus arresting extracts in metaphase (Fig. 11). In such a scheme, kinetochore-bound molecules that convert Xmad2 into its active form would be dispensable for excess Xmad2 to arrest the cell cycle, whereas the arrest would require molecules that functioned with Xmad2 to inhibit Cdc20. Our data show that excess Xmad2 can still activate the checkpoint when >95% of Xmad1 is immunodepleted, suggesting that Xmad1 is not involved in signaling events downstream of Xmad2, and that the major role of Xmad1 in the checkpoint is to recruit Xmad2 to unattached kinetochores. Future experiments will be required to determine how kinetochores monitor their interactions with microtubules and control the binding of Xmad1, how Xmad1 binding activates Xmad2, and how the activated Xmad2 inhibits Cdc20 and any other targets of the spindle checkpoint.

We thank T. Huffaker and M. Goldberg for critically reading the manuscript; T. Matsumoto for sharing the fission yeast Mad1 sequence before its publication; and A. Hyman for bringing together the labs of A.W. Murray and M. Mann.

This work was supported by grants from National Institutes of Health (NIH), the Packard Foundation, the Markey Foundation, and the March of Dimes (to A.W. Murray), and a Start-Up Fund from Cornell University, as well as an additional NIH grant (to R.-H. Chen).

APC

anaphase-promoting complex

BUB

budding unihibited by benzimidazole

CSF

cytostatic factor

MAD

mitotic arrest-deficient