Characterization of regions in hsMAD1 needed for binding hsMAD2. A polymorphic change in an hsMAD1 leucine zipper affects MAD1-MAD2 interaction and spindle checkpoint function.

In eukaryotes, the mitotic spindle assembly checkpoint provides a monitor for the fidelity of chromosomal segregation. In this context, the mitotic arrest deficiency protein 2 (MAD2) censors chromosomal mis-segregation by monitoring microtubule attachment/tension, a role that requires its attachment to kinetochores. Studies in yeast have shown that binding of MAD1 to MAD2 is important for the checkpoint function of the latter. The interactions between human MAD1 (hsMAD1) and human MAD2 (hsMAD2) have, however, remained poorly characterized. Here we report that two leucine zipper domains (amino acids 501-522 and 557-571) in hsMAD1 are required for its contact with hsMAD2. Interestingly, in several cancer cell lines, we noted the frequent presence of a coding single nucleotide Arg to His polymorphism at codon 558 located within the second leucine zipper of hsMAD1. We found that hsMAD1H558 is less proficient than hsMAD1R558 in binding hsMAD2 and in enforcing mitotic arrest. We also document a first example of loss-of-heterozygosity for a spindle checkpoint gene (at the hsMAD1 558 locus) in a human breast cancer. Based on our findings, it is possible that hsMAD1H558 could be an at-risk polymorphism that contributes to attenuated spindle checkpoint function in human cells.

In eukaryotes, the mitotic spindle assembly checkpoint provides a monitor for the fidelity of chromosomal segregation. In this context, the mitotic arrest deficiency protein 2 (MAD2) censors chromosomal mis-segregation by monitoring microtubule attachment/tension, a role that requires its attachment to kinetochores. Studies in yeast have shown that binding of MAD1 to MAD2 is important for the checkpoint function of the latter. The interactions between human MAD1 (hs-MAD1) and human MAD2 (hsMAD2) have, however, remained poorly characterized. Here we report that two leucine zipper domains (amino acids 501-522 and 557-571) in hsMAD1 are required for its contact with hsMAD2. Interestingly, in several cancer cell lines, we noted the frequent presence of a coding single nucleotide Arg to His polymorphism at codon 558 located within the second leucine zipper of hsMAD1. We found that hsMAD1H558 is less proficient than hsMAD1R558 in binding hsMAD2 and in enforcing mitotic arrest. We also document a first example of loss-of-heterozygosity for a spindle checkpoint gene (at the hsMAD1 558 locus) in a human breast cancer. Based on our findings, it is possible that hsMAD1H558 could be an at-risk polymorphism that contributes to attenuated spindle checkpoint function in human cells.
To maintain genome integrity, precise partitioning of chromosomes from one mother cell to two daughter cells occurs during mitosis. If spindle damage or segregation mistakes happen, a mitotic metaphase-to-anaphase checkpoint that halts propagation of error is activated (1,2). Accordingly, cells arrest in metaphase until corrections are achieved and impartial allocation of chromosomes can be ensured. Loss of spindle checkpoint is one of several processes that could result in aneuploidy. Because deviation from euploidy is seen in 70 -80% of all types of human cancers (3), aneuploidy could be an important contributor to transformation.
Genetics studies in yeast and mammals have implicated at least 7 genes (4,5) in mitotic spindle checkpoint (MSC) 1 func-tion, BUB (budding uninhibited by benomyl) 1-3 and MAD (mitotic arrest deficiency) 1-3, and Mps1 (monopolar spindle; see Refs. 6 and 7). The MAD BUB, and Mps1 proteins form complexes that regulate orderly chromosomal segregation and nuclear division (1,2). Although details of how this complex works remain to be fully understood, it is generally agreed that the distal effector protein of the metaphase-anaphase checkpoint is MAD2, which has been found to bind CDC20 and to inhibit the function of the anaphase-promoting complex (8 -10).
The high frequency of aneuploidy in cancers has prompted the thought that loss-of-function mutations in human BUB/ MAD proteins might occur rather commonly. Intriguingly, empirical data have largely been incongruent with this hypothesis (11). Indeed, in a survey of 21 lung cancer cell lines and 25 primary lung cancers, Takahashi et al. (12) concluded there was a lack of any obviously inactivating mutation in hsMAD2. Similarly, Cahill et al. (13) found only limited changes in BUB1 from a series of 19 colorectal cancers (i.e. two total changes, a single missense mutation, and a 197-bp internal deletion). Currently a definition of simple recessive mutation(s) that prevalently explains aneuploidy remains elusive.
Mechanistically, binding of the spindle effector protein, hs-MAD2, to kinetochores is a critical step in mitotic checkpoint function. During mitosis, unattached kinetochores attract MAD2 to produce a diffusible "wait anaphase" signal (14,15) which is then silenced by spindle attachment. We previously identified and cloned an intracellular binding partner of hs-MAD2, human MAD1 (16), whose intracellular biological activity remains largely uncharacterized. Earlier it was thought that oligomerized hsMAD2 was wholly sufficient to produce cell cycle arrest; however, recent data (17) suggest that binding of hsMAD2 to hsMAD1 rather than oligomerization is the critical step required for spindle checkpoint function. Because Mad2 mutations in aneuploid tumors are exceedingly rare (12,18,19) and because binding of MAD2 to MAD1 is required for checkpoint function (17), we wondered whether frequent loss of MAD2-associated wait anaphase function might be because of changes in its binding protein, MAD1. Here we show that human MAD1 dominantly dictates MAD2 migration into kinetochore-proximal nuclear punctates through a process that requires MAD1-MAD2 binding. We also show that binding of full-length MAD2 by full MAD1 requires two leucine zipper domains (amino acids 501-522 and 557-571) in hsMAD1. One of these two leucine zipper domains is disrupted by an Arg-558 to His-coding single nucleotide polymorphism in mad1 allele frequently found in many human cancer cell lines. biotics (100 units/ml penicillin-G, 100 g/ml streptomycin) and 10% fetal bovine serum. hsMAD1 expression vector was constructed by in-frame insertion of full-length MAD1 open reading frames with either 558R or 558H codon into pcDNA4HisMax vector (Invitrogen). Deletions in hsMAD1 and hsMAD2 were by PCR-based mutagenesis. All mutations were confirmed by sequencing. Fragments were inserted in-frame into pEGFP-C1 or pDs-Red-C1 vector (CLONTECH). Transient transfection of HeLa and Hct116 cells was performed using LipofectAMINE reagent (Invitrogen) according to the manufacturer's protocol.
Sequence and Mutational Analysis-Sequence information of hs-MAD1 is based on NM003550 in which we corrected a 1-nucleotideinsertion sequencing error by us previously reported under GenBank TM accession number U33822. PCR amplification for hsMAD1 cDNA or genomic DNA was performed with GC-RICH PCR system (Roche Molecular Biochemicals). The entire coding sequences of hsMAD1 were amplified from tumor cDNA libraries (matched cDNA pairs from tumor and normal tissues: CLONTECH) and cDNA generated by cells to cDNA (Ambion) from cancer cell lines, using the following primers 5Ј-ATATATGAATTCCATGGAAGACCTGGGGGAAAACACC-3Ј and 5Ј-AAATTTGGATCCGATTTTATTTCACAAGGTGAGGAAC-3Ј, digested with EcoRI and BamHI, and cloned into pUC vector for sequencing. hsMAD1 exon17 was amplified from genomic DNA using the following primers: 5Ј-GTGTGAGAATTCCTGCAGGGTGACTATGACCAG-3Ј and 5Ј-GAGTCTGGATCCCTGCCACCTCCTTGGACGATGGCAGAC-3Ј. Where indicated, genomic PCR products were also digested with AccII for analysis.
Immunofluorescence and Confocal Microscopy-Asynchronous HeLa cells were fixed with methanol, stained with anti-hsMAD1 monospecific antibody and anti-kinetochore human serum ANA-C (Sigma), washed, and visualized by incubation with anti-mouse IgG fluorescein isothiocyanate (Sigma) and anti-human IgG TRITC (Sigma). HeLa cells transfected with GFP-hsMAD1 were fixed with methanol, stained with ANA-C, washed, and visualized by staining with anti-human IgG TRITC. Live HeLa cells transfected with GFP-hsMAD1, GFP-hsMAD2, or RFP-hsMAD2 were visualized by confocal microscopy Axiovert 135M (Carl Zeiss).
Mitotic Indices-Cells were cotransfected with EGFP-C1 plasmid and pcDNA4HisMax-558R or -558H or control plasmid. 24 h after transfection, cells were treated with 0.2 g/ml nocodazole. Cells were fixed with 1% formaldehyde, 0.2% glutaraldehyde and stained with 10 g/ml Hoechst 33258 (Sigma). To measure mitotic indices, at least 300 cells with green signal were counted per assay at the indicated times after nocodazole treatment. In Fig. 9, mitotic indices were measure in Hoechst 33258-stained cells without transfection.

A Frequent Single Nucleotide Polymorphism Affects hsMAD1
Function-The rarity of Mad2 mutations in aneuploid human tumors led us to ask whether loss of hsMad2-dependent spindle checkpoint function might instead emanate from changes in hsMad1. To investigate this possibility, we surveyed human cancer cells for alterations in Mad1. We directly sequenced Mad1 open reading frames from 17 solid tumors and 6 HTLV-1-transformed cells. From these 23 samples, several sporadic and inconsistent point changes in Mad1 were noted (Fig. 1a). However, one frequent (Ͼ50% occurrence) and consistent finding was a change in exon 17 of hsMad1. Here, a coding single nucleotide g to a polymorphism at codon 558 ( Fig. 1b) replacing an Arg (arginine) with an His (histidine) was found in 9 of 17 solid tumors and 3 of 6 HTLV-transformed cells.
Previously, Cahill and colleagues (13) had raised the example that a single amino acid substitution in BUB1 was sufficient to alter mitotic spindle checkpoint. Thus, we considered whether the Arg to His substitution at codon 558 in hsMAD1 might also influence function. MAD1 has a long coiled-coil structure that is highly conserved in Drosophila, Xenopus, rodents, and primates. Residue 558 lies within a leucine zipper region that was suggested to be important for the proteinprotein binding of hsMAD1 (20). Computer modeling implicates the Arg to His change at position 558 to produce a small but distinct perturbation in the coiled-coil of hsMAD1 (Fig. 2a). We wondered whether this might commensurately affect protein-protein binding between hsMAD1 and hsMAD2 (16) thereby disturbing MAD2-dependent function. We constructed two otherwise isogenic hsMAD1 expression vectors that respectively encoded for either MAD1R558 or MAD1H558. Both vectors were tagged identically at their N termini with consecutive HisG and Xpress epitopes (Fig. 2b). The two vectors were separately introduced into cells. Transfected cells were found to express equally the two exogenously introduced MAD1 proteins (Fig. 2b, lanes 3 and 4, anti-HisG) as well as to have equal amounts of cell endogenous MAD2 protein (Fig. 2b, lanes 3 and 4, anti-MAD2). We then immunoprecipitated whole cell lysates with anti-Xpress (Fig. 2b, lanes 1 and 2), and we measured the amount of immunoprecipitated Xpress-tagged MAD1 and co-precipitated MAD2 by Western blotting with anti-HisG (Fig. 2b, lanes 1 and 2, top) and anti-MAD2 (Fig. 2b, lanes 1 and 2, bottom), respectively. Normalizing for recoveries, we found that consistent with the predicted disturbance in coiled-coil sequence MAD1H558 was 4.5-fold worse than MAD1R558 in binding MAD2.
We next asked whether the MAD1R558 to MAD1H558 change influences intracellular mitotic checkpoint. To address this, we transiently transfected spindle checkpoint intact Hct116 cells separately with vectors expressing either MAD1R558 or MAD1H558. The transfected cells were then challenged with microtubule-disrupting agent, nocodazole, to induce checkpoint activity. Competency of mitotic arrest was assessed. To minimize the potential for cellular adaptation and resulting mitotic slippage (21), mitotic indices were monitored within 6 h of exposure to nocodazole. In this rapid single round cell cycle measurement, we consistently found MAD1H558 to reduce the efficiency of nocodazole-induced mitotic arrest ( Loss of checkpoint activity of hsMAD1H558. a, comparison of predicted coiled-coil structures between hsMAD1R558 and -H558. The y axis is the probability of forming coiled-coils. The plots were produced by the PAIR-COIL program with MTIDK matrix at the COILS website (www.ch.embnet.org/software/COLIS_form.html). b, schematic representation of hsMAD1 expression vectors based on pcDNA4 HisMax plasmid (top). Co-immunoprecipitation of hsMAD2 with either hsMAD1R558 or hsMAD1H558 was performed with HeLa cells transfected with either pcDNA4HisMax-558R (lane 1) or pcDNA4HisMax-558H (lane 2). Immunoprecipitation was with anti-Xpress. Western blottings were with anti-HisG or anti-MAD2. c, mitotic indices of Hct116 cells overexpressing either hsMAD1R558 or hsMAD1H558. Cells were co-transfected with a green fluorescent protein (GFP) expressing plasmid, pEGFP-C1 plus pcDNA4HisMax vector (pcDNA4), or pcDNA4HisMax-558R (pcDNA4 -558R), or pcDNA4HisMax-558H (pcDNA4 -558H). After 6 h of nocodazole treatment, GFP-positive cells were scored. Values are averages from three independent assays; error bars represent S.E. minority (ϳ25%; see pcDNA4 values, Fig. 2c) of these asynchronously cultured cells was poised to enter metaphase during the 6-h assay, if this effect were translated to a homogeneous population of metaphase-synchronized cells then single-round mitotic arrest differences between MAD1H558 versus MAD1R558 would range between 32 and 40%.
Hct116 is a diploid human colon cancer that expresses large amounts of endogenous MAD1 protein (Fig. 3a, lane 1). Because of its high level of MAD1, the true potency of exogenously introduced MAD1H558 could be partially masked in Hct116. We had previously mapped hsMad1 to human chromosome 7 (22). Because monosomy 7 (i.e. a Mad1 haplotype) is rather common in several types of human cancers, we sought to verify whether in settings of reduced MAD1 expression (e.g. monosomy 7 cells) the mitotic arrest differences between MAD1H558 and MAD1R558 would be enhanced. In two cell lines KG-1 (Fig. 3a, lane 2), an acute myelogenous leukemia with monosomy 7, and LS411N (Fig. 3a, lane 3), a human Dukes' type B cecum carcinoma with monosomy 7, were directly verified to have low amounts of endogenous MAD1. In these two cells, we repeated the comparison between MAD1R558 and MAD1H558. Consistent with expectation and with the above Hct116 results, we observed arrest efficiency differences of 30 -40% between MAD1H558 and MAD1R558 (Fig. 3b). These results suggest that an hsMad1h558 genotype superimposed on an otherwise hsMad1 haplotype (i.e. monosomy 7) potentially conveys a worse cancer prognosis than an hsMad1 haplotype alone.
hsMAD1 Locates hsMAD2 into Nuclear Bodies-The above functional observations prompted us to consider how MAD1H558 might mechanistically influence intracellular spindle checkpoint function. Previously, in vitro results from Xenopus egg extracts have suggested that soluble Xmad1 might recruit Xmad2 to kinetochores (23). However, potential biological differences between meiotic Xenopus soluble egg extracts and mitotic human cells question whether convergent or divergent rules govern MAD1, MAD2, and kinetochore interaction in the latter setting (24). For example, whereas CENP-E is required in Xenopus egg extracts for Xmad2 attachment to and signaling from kinetochores (24), depletion of CENP-E in HeLa cells paradoxically conferred chronic hsMAD2 association to and signaling from human kinetochores (25). Further confusing the understanding of MAD1/MAD2 activity in human cells, several different immunostained profiles of hsMAD1 during various stages of the cell cycle have been reported (16,26). To understand better the MAD1H558 phenotype, we created chimeric green fluorescent (GFP)-hsMAD1 to analyze protein localization as well as intracellular interactions between MAD1 and MAD2.
Asynchronously cultured HeLa cells were transfected with a plasmid vector expressing GFP-hsMAD1. Cells were then examined for green fluorescence and also immunostained with anti-kinetochore antibody, ANA-C (Fig. 4a). ANA-C stained the interphase nuclei to produce an array of small dots consistent with centromeres/pre-kinetochores (Fig. 4a, middle panel). The GFP-hsMAD1 protein also presented constitutively in the interphase nuclei as bright green, slightly larger, nuclear punctates (Fig. 4a, left), some of which overlaid proximally with a subset of ANA-C stained dots (Fig. 4a, right).
The fact that GFP-hsMAD1 appeared constitutively in the nucleus prompted us to query as to how in human cells MAD1 might interact with MAD2. An initial report (27) on MAD2 had found it to be largely cytoplasmic in mammalian cells. In detergent fractionation of several types of human cells, we obtained results in agreement with this cytoplasmic predominance (data not shown). Because MSC checkpoint function obliges MAD2 to localize to kinetochores, one wonders whether this protein diffuses to kinetochores purely after mitotic dissolution of the nuclear envelope or whether some active fraction of human MAD2 might be in a centromere-proximal nuclear locale prior to mitosis. To address this question in the context of GFP-hsMAD1, we constructed a red fluorescent protein (RFP)-tagged hsMAD2 protein (Fig. 4b). Asynchronous HeLa cells were then transfected with GFP-hsMAD1-alone (Fig. 4b, left column), RFP-hsMAD2-alone (Fig. 4b, middle column), or RFP-hsMAD2 ϩ GFP-hsMAD1 together (Fig. 4b, right column). Green (488 nm) and red (568 nm) signals independently captured from the same cells permitted one to distinguish MAD1 from MAD2. GFP-MAD1-alone was found to be nuclear and in punctated configurations (Fig. 4b, left column); by contrast, RFP-hsMAD2-alone was faintly whole cell to largely cytoplasmic (Fig. 4b, middle column). This profile agrees with a previously mentioned whole cell pattern for GFP-yMAD2 in yeast (28) and suggests that in the pre-mitotic phase of the cell cycle MAD2 autonomously is incapable of specific nuclear localization. Interestingly, when GFP-MAD1 and RFP-MAD2 were simultaneously expressed in the same cell, RFP-MAD2 was quantitatively driven into GFP-MAD1 nuclear dots (Fig.  4b, right column). In these interphase cells, the GFP-hsMAD1 ϩ RFP-hsMAD2-dots were indistinguishable from those of GFP-MAD1-alone (Fig. 3b, left column). This finding suggests that in human cells some portion of MAD2 can be actively localized into the nucleus by hsMAD1 prior to the dissolution of the nuclear envelope.
The Entire hsMAD2 Is Needed to Interact Functionally with hsMAD1-We found previously that all hsMAD2 truncations abolished heterodimerization of MAD1-MAD2 in vitro (16). To compare in vitro results with intracellular MAD1-MAD2 interactions, we constructed wild type and four GFP-hsMAD2 truncation mutants (Fig. 7a). Each was tested for subcellular localization in the absence (Fig. 7b, panels 1 -5) or presence (Fig. 7b, panels 6 -10) of overexpressed untagged hsMAD1. GFP-hs-MAD2 wild type (Fig. 7b, panel 1) produced green whole cell fluorescence that was dramatically converted to nuclear punctates by co-expressed hsMAD1 (Fig. 7b, panel 6). By contrast, all deletions in either N or C terminus destroyed the capacity of GFP-MAD2 to be efficiently reconfigured by hsMAD1 into nuclear dots (Fig. 7b, panels 2-5 and 7-10). Thus, consistent with our earlier in vitro binding results, intracellularly, the entire hsMAD2 protein is also required for efficient interaction with hsMAD1.
Loss of Heterozygosity at MAD1 Codon 558 in a Breast Cancer-The above results provide cell culture evidence that both leucine zipper domains in MAD1 are important for MAD1-MAD2 binding and that prior to the onset of mitosis MAD1 can convey MAD2 into nuclear punctates through direct contact. In the context of these observations, we note with interest that the MAD1H558 polymorphism impinges directly on the 557-571 leucine zipper needed for MAD1-MAD2 binding. To probe further the physiological significance of the MAD1R558/ MAD1H558 polymorphism, we next sought in vivo evidence supportive of this change being advantageous for cellular pro-  7. A requirement for intact hsMAD2 to interact with hs-MAD1. a, schematic representations of wild type hsMAD2 and four deletion mutants. b, panels 1-10 contain representative confocal images of HeLa cells transfected with various GFP-hsMAD2 expression plasmids and co-transfected with either pUC19 (panels 1-5, ϪMAD1) or with wild type untagged hsMAD1 expression plasmid (panels 6 -10, ϩMAD1).

liferation. A finding of loss of heterozygosity (LOH) at the
Mad1 558 locus would represent such in vivo evidence.
The AccII restriction enzyme cleaves the "cgcg" sequence. In our search for LOH, we noted that the Arg to His single nucleotide polymorphism at codon 558 of hsMAD1 changes a "cgcg" to "cacg," abolishing an AccII site (Fig. 8a). Accordingly, AccII restriction would be diagnostic for this 558 polymorphism. We analyzed the AccII restriction of exon 17 from hsMad1 in several human specimens; some illustrative results are shown in Fig. 8b. For instance, spindle checkpoint-defective SW480 colon cancer was homozygously h/h at position 558. On the other hand, the checkpoint intact Hct116 and HeLa cells (29) were 558r/r and 558r/h, respectively. To ask whether an Arg to His change could occur through somatic LOH, we examined eight breast tumor biopsies paired with their corresponding normal B cells. In one of eight (HCC1937) pairs, the heterozygous 558r/h genotype in the normal peripheral B cells was found to have converted to a homozygous 558h/h genotype in the matched breast tumor (Fig. 8b, lanes 4 and 5).
Expression of MAD1558H and Loss of Spindle Checkpoint in Colon Cancer Cells-The above results (Fig. 8) suggest that the 558h/h genotype, at best, correlates with only a minority (e.g. 12.5%) of breast cancers. This could be consistent with the multifactorial basis for breast cancers in which important genetic components such as BRCA1 and BRCA2 account for between 1.5 and 2% of all breast tumors (30). Because we lacked detailed karyotypic information for the breast samples in Fig.  8, we turned to several well described colon cancer cells to further examine the correlation between 558r/r versus 558h/h with aneuploidy and spindle checkpoint function (Fig. 9). In analyzing nine colon cancer cells, we found that five (SW480, SW48, LoVo, T84, and DLD-1) were 558h/h, whereas two (SW837 and Hct116) were 558r/r. Two others (Colo201 and HT29) were heterozygously 558r/h (Fig. 9A).
Using the two homozygous colon cancer genotypes, we next asked how 558r/r and 558h/h might differ in spindle checkpoint function (Fig. 9B). When Hct116 (558r/r) was compared with SW480, SW48, LoVo, and T84 (558h/h) in mitotic arrest response to nocodazole, we found that the former cell was spindle checkpoint intact, whereas all of the latter cells were spindle checkpoint-defective (Fig. 9B). (DLD-1 and SW837 cells showed unexpectedly high levels of cell death in the presence of nocodazole, and mitotic indices were not assessed for these two cells.) As an additional measure of the difference between MAD1 558R and MAD1 558H, we assessed spindle checkpoint func-tion when the two different MAD1 proteins were introduced into a cell with homogeneous 558h/h background. Thus, we transfected SW480 cells to overexpress either MAD1 558H or 558R. In transfected cell cultures, we consistently observed  6 and 7). Restriction patterns for HeLa, Hct116, and SW480 cell lines are also presented. that overexpressed MAD1 558R rescued the otherwise defective spindle checkpoint response of SW480 cells to nocodazole (Fig. 10). DISCUSSION Perturbation in several discrete pathways can lead to aneuploidy in mammalian cells (31)(32)(33)(34). One of several gatekeepers of euploidy is the mitotic spindle assembly checkpoint (MSC). Among the many components of the spindle checkpoint, studies in yeast and mammals have identified MAD2 as the key downstream effector in the metaphase-to-anaphase censor against chromosomal missegregation during mitosis (9,28,(35)(36)(37). Considering that spindle checkpoint loss can contribute to aneuploidy and that 70 -80% of all cancer cells are aneuploid, one reasonably expects to find frequent Mad2 (and perhaps other checkpoint components) mutations in human malignancies. Curiously, intensive targeted searches for mutations in the spindle checkpoint genes have revealed such to be exceedingly  1-9). COS cell (lane 10) is an African green monkey kidney cell transformed by SV40. Analyses were performed as described in Fig. 8. b, mitotic index comparisons between Hct116 (558r/r) and four 558h/h colon cancers (SW48, SW480, Lovo, and T84). The former genotype exhibits mitotic arrest in response to nocodazole, whereas the latter genotype shows a defect in arrest function. Results are average values from three independent assays. Hct116 is diploid, and SW48, SW480, LoVo, and T84 are aneuploid. rare in human tumors (11, 38 -41). Hence, an unanswered cancer paradox is how to reconcile frequent checkpoint loss with rarity of checkpoint gene mutations.
HsMAD1 is the intracellular binding partner for hsMAD2 (16). Thus, loss of hsMAD2 function could emanate from changes in hsMAD1. Here we have characterized the domains in hsMAD1 needed to interact with hsMAD2. We found that two leucine zipper domains (amino acids 501-522 and 557-571) are required for hsMAD1 to bind hsMAD2. In several cancer cells, a coding single nucleotide Arg to His polymorphism at codon 558, which disrupts the second leucine zipper of hs-MAD1, was frequently observed. This R558H change in hs-MAD1 was seen in over 50% of our surveyed human cancers (Fig. 1), including 7 of 9 colon cancers (Fig. 9A). One of eight breast cancer samples examined by us also showed evidence for LOH at codon 558 of hsMAD1 (Fig. 8). Interestingly, when we surveyed anonymous peripheral blood samples from 14 normal individuals, 6 were heterozygously 558h/r; 8 were homozygously 558r/r; none were homozygously 558h/h (data not shown). Thus, whereas larger surveys are needed to determine the extent of 558h in the gene pool and the potential frequency of LOH in cancers, the extant data do support that 558H is a common human MAD1 polymorphism. If the properties of 558H as described here are correct, our findings could imply that a genotypic proclivity for loss of spindle checkpoint function may be more common than conventionally anticipated.
Our study comments on two mechanistic issues regarding MSC function in human, Xenopus, and yeast. Studies on Xmad1 and Xmad2 in soluble meiotic extracts have shown that Xmad1 dominantly locates Xmad2 to kinetochores (23,28). Here we observed that in interphase human cells, hsMAD1 dominantly directs hsMAD2 into nuclear punctate bodies; some of these bodies appear to overlap with ANA-C staining. If one analogizes the human findings to the Xmad1/Xmad2 model, a reasonable interpretation, which does not necessarily exclude others, is that during interphase human MAD1 directs nuclear migration of a subset of MAD2 to centromere-proximal (42) nuclear locales. These hsMAD1-hsMAD2 complexes could thus be poised for checkpoint function in the ensuing mitotic phase of the cell cycle. It is likely that further protein-protein interactions are required to trigger hsMAD1/hsMAD2 checkpoint activity. In this regard, it has been reported that the CENP-E protein, which is synthesized subsequent to interphase (43), is needed for Xmad1/Xmad2 to bind kinetochores in meiotic extracts (24). Intriguingly, in mitotic HeLa cells, MAD2 was found to bind kinetochores despite deliberate depletion of CENP-E (25). Although our current data do not permit us to distinguish between these two competing views, we note that Mad1/Mad2 function in yeast is apparently CENP-E-independent because no CENP-E-like protein exists in the yeast genome (24).
Our characterization of hsMAD1 behavior in cancer cells adds insight to existing studies on hsMAD2 and hsBUB1 and provides an initial platform for addressing the relative interplay between these checkpoint components. The findings here are consistent with the polymorphic MAD1H558 protein being less capable than MAD1R558 in binding hsMAD2 and in enforcing spindle assembly checkpoint function in cultured cells. These results clarify for the first time that two intact leucine zipper domains in hsMAD1 are mechanistically required for binding hsMAD2, and they extend further support to the notion that binding by hsMAD1 is important for hsMAD2 function. Despite the suggestive findings from cultured cells, a conclu-sive deduction that the His-558 polymorphism affects in vivo spindle checkpoint function awaits epidemiological case control studies. Additional work is also needed to fully address the relationship of loss of spindle checkpoint function with the development of aneuploidy in vivo.