Transcriptional Regulation of Mitotic Checkpoint Gene MAD1 by p53*

p53 regulates a number of genes through transcriptional activation and repression. p53-dependent mitotic checkpoint has been described, but the underlying mechanism is still obscure. Here we examined the effect of p53 on the expression of a human mitotic checkpoint protein, Mitosis Arrest Deficiency 1 (MAD1), in cultured human cells. The expression of MAD1 was reduced when the cells were overexpressing exogenously introduced wild-type p53. The same reduction was also observed when the cells were treated with anticancer agents 5-fluorouracil and cisplatin or were irradiated with UV. Consistently, MAD1 promoter activity diminished in a dose-dependent manner when induced by p53, indicating that p53 repressed MAD1 at a transcriptional level. Intriguingly, several tumor hot spot mutations in p53 (V143A, R175H, R248W, and R273H) did not abolish the ability of p53 to repress MAD1 expression. By serial truncation of the MAD1 promoter, we confined the p53-responsive element to a 38-bp region that represents a novel sequence distinct from the known p53 consensus binding site. Trichostatin A, a histone deacetylase inhibitor, relieved the p53 transrepression activity on MAD1. Chromatin immunoprecipitation assay revealed that p53, histone deacetylase 1, and co-repressor mSin3a associated with the MAD1 promoter in vivo. Taken together, our findings suggest a regulatory mechanism for the mitotic checkpoint in which MAD1 is inhibited by p53.

The p53 tumor suppressor gene takes part in cell cycle control, DNA damage repair, and apoptosis (1)(2)(3). Its importance is underscored by the fact that p53 gene is frequently mutated in more than 50% of all human tumors. p53 acts as a transcription factor regulating numerous downstream target genes. It can transactivate genes involved in cell cycle progression and apoptosis, such as p21 WAF1 , MDM2, BCL2, and BAX (4 -6). Transcriptional activation requires p53 binding to a consensus sequence containing two copies of the canonical site 5Ј-RRRC(A/ T)(T/A)GYYY-3Ј separated by 0 -13 bp (7). Most common mutation hot spots in p53 found in tumors are located within the p53 DNA-binding domain, and these mutants are defective in DNA binding (8). p53 is also known to repress a number of targets including DNA topoisomerase II, cyclin B, CDC2, MMP-1 and -13, presenilin-1, MAP-4, and stathmin genes (9 -12). However, the repression mechanisms are relatively obscure. In some cases, DNA binding is required for repression such as ␣-fetoprotein and CDC2 (13,14). It appears that p53 binds to a site overlapping the consensus sequences of other more potent transactivators, thus displacing the more potent activators (13). On the other hand, p53 repression can also be achieved in a DNA binding-independent manner. It is likely that p53 interferes with the basal transcription machinery by interacting with upstream transcriptional activators, for instance TATA-binding protein (15) and Sp1 (16).
The mitotic checkpoint prevents the onset of anaphase by blocking the activation of anaphase-promoting complex (APC) 1 until all sister chromatids have aligned properly to the spindle and attached in a bipolar manner (17,18). Several checkpoint components were initially identified in budding yeast, including MAD1, MAD2, MAD3, BUB1, BUB2, BUB3, and MPS1. Vertebrate homologs of MAD1, MAD2, BUB1, and BUB3 localize to kinetochores prior to chromosome alignment on the metaphase plate (19,20). Loss of mitotic checkpoint function is associated with chromosomal instability in cancer cells (21)(22)(23)(24)(25). MAD2 appears to be the critical component in the mitotic checkpoint pathway. Unattached kinetochores recruit MAD2 and other checkpoint proteins, thereby activating the checkpoint (18). MAD2 is able to bind Cdc20, an activator of APC, and prevents Cdc20 from association with APC, resulting in APC inhibition (17,18). The checkpoint ends with bipolar attachment at metaphase. Several checkpoint proteins including MAD2 move away from kinetochores, releasing Cdc20 to activate APC (26,27). Two main substrates for the APC are mitotic cyclins (A and B) and securin, and their destruction is required for the exit from mitosis and sister chromatid segregation, respectively (28 -31). During interphase, MAD1 and MAD2 localize to the nuclear pores (32)(33)(34). MAD1 and Cdc20 can bind MAD2, but these two bindings are exclusive. They share a 10-residue MAD2-binding motif. It has been suggested that MAD1 acts as a positive regulator and a competitive inhibitor of the MAD2-Cdc20 complex (35)(36)(37)(38). The structure and function of MAD2 are relatively well understood, but little information about MAD1 is known thus far. In a serial analysis of gene expression (SAGE) in human colorectal cancer cells expressing p53 versus the p53 non-expressing parental cells, MAD1 has been identified as one of the genes regulated by p53 (39). Here we attempted to investigate in depth the role of p53 in MAD1 expression. We show that wild-type and mutant p53 transcriptionally repress MAD1 in human primary cells and several cell lines including HepG2 and Hep3B. Moreover, we mapped the p53-responsive element in the MAD1 promoter. p53 repression activity on MAD1 is decreased upon addition of histone deacetylase inhibitor, trichostatin A. Significantly, p53 and mSin3a as well as HDAC bind to the MAD1 promoter in vivo, consistent with a recently proposed model in which these three proteins form a repression complex targeting the p53-repressive genes. Notably, the binding of mSin3a and HDAC to the MAD1 promoter is p53-dependent.
Cell Cultures, Transient Transfection, and Treatment with Anticancer Agents-Human hepatoma cell lines HepG2 and Hep3B (41)(42)(43), cervical carcinoma cell line HeLa, and normal human embryonic lung fibroblasts IMR-90 were grown in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum, 100 units/ml penicillin G, and 100 g/ml streptomycin. Human nasopharyngeal carcinoma CNE2, breast cancer cell line MCF-7, colon cancer cell line LoVo, metastatic prostate cancer cell line PC-3, and ovarian carcinoma cell line SK-OV-3 were grown in RPMI supplemented with 10% fetal bovine serum, 100 units/ml penicillin G, and 100 g/ml streptomycin. Cells were maintained in at 37°C in a humidified atmosphere at 5% CO 2 . HepG2, HeLa, IMR-90, CNE2, MCF-7, and LoVo cells express wild-type p53, whereas Hep3B, PC-3, and SK-OV-3 cells are p53-deficient. Cells were cultured on 6-well plates for transient transfection. All cells were transfected using LipofectAMINE 2000 TM reagent (Invitrogen) according to the manufacturer's instruction, except HepG2 cells which were transfected by the calcium phosphate co-precipitation method (44). All cells were transfected for at least 16 h before harvest. For trichostatin A (TSA) treatment, cells were incubated in the presence of TSA during DNA transfection. For UV treatment, cells were irradiated with 50 J/m 2 UV (UV-C) by CL-1000 Ultraviolet Cross-linker (UVP, Inc.) and harvested after 24 h. For p53 induction, anticancer drugs 5-fluorouracil and cisplatin were added to the culture medium directly, and the cells were incubated for 24 h before harvest. In the assay for p53-mediated MAD1 ubiquitination, proteasome inhibitors carbobenzoxyl-leucinyl-leucinylnorvalinal-H (MG-115) and N-benzyloxycarbonyl-Ile-Glu(O-t-butyl)-Ala-leucinal (PSI; Peninsula Laboratories, Inc.) were separately added to final concentrations of 25 and 30 M, respectively, to the culture media of cisplatin-treated cells for 5 h before harvest.
Chloramphenicol Acetyltransferase (CAT) Assay-The procedure was carried out as described (45). After three washes with phosphate-buffered saline, transfected cells were scraped and lysed in 0.25 M Tris-Cl, pH 8.0, by three cycles of freezing (at Ϫ80°C) and thawing (at 37°C), followed by centrifugation at 11,000 rpm at 4°C for 10 min. Protein concentration of clarified lysates was determined by the Bradford reagent (Bio-Rad). Equal amounts of lysates were mixed with 1 Ci of 14 C-labeled chloramphenicol (Amersham Biosciences), 0.8 mM acetyl coenzyme A (Calbiochem), and 0.5 M Tris-Cl, pH 8.0, and incubated at 37°C for CAT reaction. CAT activities were detected using thin layer chromatography and quantified by PhosphorImager (Amersham Biosciences).
Chromatin Immunoprecipitation-Hep3B cells were transfected with empty vector or p53-expression vector for 20 h. ChIP was performed as described previously (46), except that cells were resuspended in 1ϫ TE buffer (10 mM Tris-Cl, and 0.1 mM EDTA, pH 8.0) with 2 mM phenylmethylsulfonyl fluoride, 2 g/ml pepstatin A, 2 g/ml aprotinin, and 2 g/ml leupeptin for sonication. RNase A as well as proteinase K were added to the samples after cross-links were reversed (47). Antibodies used in the immunoprecipitation were p53 (Fl-393), HDAC1 (C- 19), and mSin3a (AK-11, Santa Cruz Biotechnology). Immunoprecipitated samples were resuspended in 50 l of H 2 O and analyzed by PCR. Total input samples were resuspended in 100 l of H 2 O and diluted 1:100 before PCR. 3 l of immunoprecipitated chromatin or 1:100 diluted total input sample was added to the PCR amplifying the MAD1 promoter. The PCR primers are 5Ј-AACTGCAGAGGTTACCAT-AATAATGGTG-3Ј (forward) and 5Ј-GCTCTAGAGCTTTCGCCTCGCG-CCGT-3Ј (reverse). After 40 cycles of amplification, PCR products were analyzed on 1% agarose gel with ethidium bromide.
Confocal Microscopy-Confocal immunofluorescence microscopy was performed as detailed elsewhere (22,48,49). Detection of dual immunofluorescence was based on species-specific secondary antibodies differently conjugated to Cy5 and fluorescein. The anti-MAD1 antiserum 105 raised against a GST-MAD1 fusion protein has been described previously (22).

Down-regulation of MAD1 Protein Expression by Wild-type
p53-A genome-wide SAGE analysis revealed that human mitotic checkpoint gene MAD1 is induced by p53 in the colorectal cancer cell line DLD-1 (39). To understand further the role of p53 in regulating MAD1, we transiently transfected the human hepatoblastoma cell line HepG2 with a CMV promoter-driven p53-expression plasmid pCMVp53. The expression of p53 and endogenous MAD1 was detected by Western blotting. As shown in Fig. 1A, the level of MAD1 protein decreased significantly when cells were transfected with a low dose of p53-expressing plasmid (0.1 g, lane 2) compared with the mock-transfected control cells (lane 1). Notably, the MAD1 level decreased in a dose-dependent manner. When a relatively large quantity of p53 (12.5 g) was added, endogenous MAD1 protein was minimally detected (lane 4). In control experiments, we first queried whether the expression vector alone without p53 could lead to the decrease of MAD1 expression. p53 was cloned previously into an expression vector carrying a CMV promoter (50). Similar to Fig. 1A, HepG2 cells were transfected with increasing amounts of empty vector, pCMV (Fig. 1B). However, there were no apparent changes in the MAD1 levels even at high concentrations of vector (Fig. 1B, lane 4), suggesting that the repression of MAD1 is p53-specific. To verify the specificity of antiserum against MAD1, proteins from control cells without transfection with p53-expressing plasmid were resolved by SDS-PAGE, followed by immunoblotting separately with polyclonal anti-MAD1 antiserum 181d and the 181d serum preincubated with the MAD1 peptide that had been used to raise the antibody (Fig. 1C). MAD1 signal was strengthened with increasing amounts of cell lysates added (lanes 1 and 2), whereas pre-incubation of MAD1 peptide thoroughly blocked the MAD1-specific antibody from detecting the protein on the immunoblot (lane 3). These results indicate that p53 downregulates the expression of MAD1 protein in a dose-dependent manner. The down-regulation of MAD1 was also observed in two p53-null cell lines, Hep3B (hepatocellular carcinoma) and PC-3 (metastatic prostate cancer), as well as in IMR-90 cells (normal embryonic lung fibroblast) with wild-type p53, when these cells were transfected with p53 ( Fig. 1D). Hence, this repression was unlikely cell type-specific.
To characterize further the p53 effect at the single cell level, we performed immunofluorescence studies using the aforementioned anti-MAD1 antiserum 181d. As a first step, we verified the specificity of the 181d antiserum in immunofluorescence staining. We transiently transfected HeLa cells with MAD1 expression plasmid and stained the cells with 181d antiserum ( Fig. 2A, panel 1), and with a previously characterized antiserum 105 highly specific for human MAD1 (22). Both antibodies specifically reacted with both the endogenous and the overexpressed MAD1 protein ( Fig. 2A, compare cells with arrows to cells without arrows). This nuclear localization pattern with some concentration in the nuclear membrane was typical to human MAD1 (22,32). In further support of the specificity of the antiserum, pre-incubation of 181d with an excess of immunizing peptide abolished the nuclear staining ( Fig. 2A, panel 3).
Next we assessed the influence of p53 on MAD1 expression by transfecting HeLa cells with p53. We observed that the overexpression of p53 (Fig. 2B, panel 1) correlated with a dramatic reduction in the steady-state amount of MAD1 protein in HeLa cells (panel 2, cell with arrow), whereas an appreciable amount of MAD1 was retained in the nucleus with a low basal concentration of endogenous p53 (cell with asterisk). As a negative control, the overexpression of the Gal4 DNA-binding domain (Gal4BD; Fig. 2B, panel 3) had no effect on the expression of MAD1 (panel 4). Thus, p53 repressed MAD1 expression specifically.
Repression of MAD1 Expression upon Treatment with Chemotherapeutics in HepG2 Cells-Having observed that introduction of exogenous p53 into cells could repress the expression of MAD1, we were interested in examining whether such repression could be sustained in a physiological environment with an elevated amount of endogenous p53. Treatment of cells with chemotherapeutics such as DNA-damaging agents 5-fluorouracil (5-FU) and cisplatin has been documented to induce the expression of endogenous p53. These agents have been widely employed in the studies of cell cycle regulation and apoptosis (51,52). Two human hepatocarcinoma cell lines HepG2 and Hep3B were treated with increasing concentrations of 5-FU and cisplatin, followed by analysis for the expression of MAD1 and p53 by Western blotting. Both 5-FU and cisplatin effectively induced endogenous p53 expression over the basal level in p53-expressing HepG2 cells (Fig. 3A, left panel, lanes [1][2][3][4][5]. The induction of p53 by chemotherapeutics considerably repressed the expression of MAD1 in the cells. In both drug treatments, the down-regulation of MAD1 again followed a dose-dependent manner, consistent with the pattern observed in the cells transfected with p53-expression plasmids ( Fig. 1). Here, p53-deficient Hep3B cells served as a negative control because a major portion of the p53 gene in Hep3B cells was found to be deleted, accompanied by the absence of p53 RNA transcripts and protein (42,43). Thus, no p53 protein was detected in Hep3B cells even in the presence of p53-inducing drugs (Fig. 3A, right panel, lanes [1][2][3][4][5]. As a positive control, Hep3B cells were transfected with pCMVp53, and the p53 protein was detected in Western blotting (Fig. 3A, right panel, lane 6). Expression of the exogenously introduced p53 in Hep3B cells indicated that the undetectable level of endogenous p53 protein was not due to any defect in the expression machinery for the protein. Rather, it was a consequence of a major genomic alteration in Hep3B cells leading to deletion of a large portion of the p53 gene (41-43).
Next we sought to examine whether these results were applicable to other carcinoma cells and normal primary cells. HeLa (cervical carcinoma), CNE2 (nasopharyngeal carcinoma), MCF-7 (breast carcinoma), LoVo (colon carcinoma), and IMR-90 (normal lung fibroblast) cells harboring wild-type p53 displayed a reduction of protein expression level of MAD1 when they were treated with cisplatin. PC-3 (prostate carcinoma) and SK-OV-3 (ovarian carcinoma) cells, like Hep3B, were null for p53. Thus, the addition of p53-inducing drugs had no effect on the alteration of MAD1 expression (Fig. 3B). The cisplatininduced down-regulation of MAD1 expression is p53-depend-ent, and this finding agreed with the decrease of MAD1 observed in the cells transfected with exogenous p53 (Fig. 1). Although a previous report (39) describes an induction of MAD1 in the DLD-1 colon cancer cell line upon infection with adenovirus encoding p53, we did not detect any increase in the levels of MAD1 upon elevated expression of p53 using two distinct approaches in the cell lines we have studied.
The data shown above manifest the fact that p53 represses MAD1 expression in both tumor and non-tumor cells. There are two possible mechanisms to explain the observation: p53 influenced either the protein level (i.e. protein degradation) or the transcription of MAD1. A number of cell cycle regulatory proteins such as cyclins, securin, and p53 were degraded by ubiq-  3 and 4, respectively). Cells were lysed, and protein expression levels were analyzed by Western blotting using MAD1-, p53-, and ␤-actin-specific antibodies.
uitin-mediated proteolysis (6, 28 -31). On the other hand, p53 can in turn regulate the synthesis of ubiquitin protein ligases such as Mdm2 (6). To investigate whether p53 has an impact on the proteosome-dependent degradation of MAD1 protein, we evaluated the effect of proteasome inhibitors on the levels of MAD1 in cisplatin-treated cells. Cells harboring wild-type p53, including HepG2, IMR-90, and MCF-7, were treated with cisplatin to induce endogenous expression of p53 and then with proteasome inhibitors MG115 and PSI for 5 h. If p53 regulates MAD1 via the ubiquitin-proteasome pathway, the impediment of proteasome activity by the inhibitors should relieve the repressive effect by cisplatin, resulting in a rise in MAD1 protein levels comparable with the levels in cells without cisplatin treatment (Fig. 3C, lane 1). When cells were treated with cisplatin, p53 was induced, and MAD1 protein level was declined, as expected (lane 2). However, we did not observe any increase of MAD1 protein levels upon addition of MG115 or PSI (lanes 3 and 4). Rather, the levels remained repressed in all three cells we analyzed. This assay vigorously indicated that repression of MAD1 protein expression by p53 was unlikely mediated through the ubiquitin-proteasome degradation pathway.
Expression of MAD1 Is Regulated by p53 at Transcriptional Level-After eliminating the possibility of p53-mediated ubiquitination of MAD1, we asked whether the repression occurred at the stage of transcription. We amplified a region with 480 bp upstream of the exon 1 of human MAD1 by PCR (Fig. 4A) and inserted it into a CAT reporter (pCAT-Basic vector), which lacked eukaryotic promoter and enhancer sequences. The resulting plasmid is named p428CAT. This 480-bp sequence did not contain p53 consensus binding site comprising two copies of 5Ј-RRRC(A/T)(T/A)GYYY-3Ј separated by 0 -13 bp (7). A putative TATA box was found in the sequence at Ϫ17/Ϫ13 (Fig. 4A, underlined) upstream of the transcription start site (ϩ1, arrow). HepG2 and Hep3B cells were transiently transfected either with p428CAT or with the parental pCAT-Basic vector and analyzed for CAT activities (Fig. 4B). Because the pCAT-Basic vector did not contain any promoter or enhancer sequences, detectable CAT activity indicated that the 480-bp sequence possessed promoter activity. Indeed, the CAT assay demonstrated that the 480-bp sequence acquired a considerable promoter activity when compared with the control (Fig.  4B, upper panel, compare lane 2 with 1). By employing this CAT reporter construct driven by the MAD1 promoter, we then investigated the effect of p53 using the CAT assay. In addition to chemotherapeutics, UV irradiation is another well known stimulus inducing p53 expression in cells (53,54). HepG2 and Hep3B cells both transfected with p428CAT plasmid were exposed to UV at 50 J/m 2 and harvested 24 h after irradiation. Upon UV irradiation, an abundant increase of endogenous p53 over the basal expression level was detected in HepG2 cells but not in p53-null Hep3B cells by Western blotting (Fig. 4B, lower  panel, compare lane 3 with lanes 1 and 2). Correspondingly, the MAD1 promoter activity diminished in HepG2 cells as a result of the induction of p53 by UV irradiation, whereas the activity remained unchanged in Hep3B cells (upper panel) even at 100 J/m 2 of UV (data not shown). In line with the previous cisplatin treatment assay, UV-induced decline of MAD1 promoter activity is also p53-dependent, and these data further illustrated that repression of MAD1 occurred at a transcriptional step. To ascertain the effect of UV irradiation on MAD1 promoter activity, we had both HepG2 and Hep3B cells exposed to UV irradiation and analyzed p53 and MAD1 proteins (Fig. 4C). Consistent with the CAT assay, Western blot analysis showed that UV-induced p53 down-regulated MAD1 protein expression through the lowering of the MAD1 promoter activity.
To confirm further the effect of p53 on the MAD1 promoter, we used p53-null Hep3B cells. Because Hep3B cells are deficient in p53, we transiently co-transfected pCMVp53 and p428CAT into the cells, instead of using p53-inducing agents such as anticancer drugs or UV, to provide a source of p53. Cells transfected with p428CAT exhibited robust basal CAT activity (Fig. 4D, lane 2). Nevertheless, the promoter activity diminished in the presence of progressively increasing amounts of p53 plasmids (lanes 3-5). When 12.5 g of pCMVp53 was added (lane 5), the activity reduced down to only 5% of that in the absence of p53 (lane 2). Because the total DNA amounts were kept constant by compensating with an irrelevant plasmid pUC19 in all dose-dependent assays, it indicated that the effect of p53 on MAD1 promoter activity was specific and was not due to an increase in the total amount of DNA transfected. The same CAT assay was also conducted in p53-expressing HepG2 and HeLa cell lines, and the results agreed with those obtained in Hep3B cells (data not shown). These results clearly demonstrated that p53 transcriptionally repressed the expression of MAD1 in a dose-dependent manner, and this effect seemed not to be cell type-specific.
Transcriptional Repression of MAD1 by p53 Mutants-Because p53 is the most frequently mutated gene identified in human cancers (Ͼ50% of all cancers) and mutation of p53 generally resulted in the inactivation of the tumor suppressor function (1), it is of interest to study the influence of p53 mutants on MAD1 promoter activity. Here we looked at four p53 mutants that contain changes at codon 143 (Val to Ala), 175 (Arg to His), 248 (Arg to Trp), and 273 (Arg to His), respectively. All these four p53 mutants represent tumor "hot spots" and have been found to be defective in binding to p53 consensus binding sequence (7,40). We used p53-null Hep3B cells, rather than p53-positive cell lines such as HepG2 or HeLa, to avoid any confounding effects from the endogenous p53. Hep3B cells were co-transfected with either wild-type or mutant p53 with indicated amounts and an equal amount (5 g) of p428CAT. As expected, wild-type p53 repressed MAD1 promoter activity (Fig. 5, lane 2 and 3). Surprisingly, all four p53 mutants also exhibited the ability to transcriptionally repress MAD1 with various extents. Markedly, the down-regulation of MAD1 promoter activity by these mutants acted in a dose-dependent manner as the wild-type p53, with the lowest MAD1 promoter activity at the highest concentration of p53 (lanes 4 -15).
Identification of a p53-responsive Element in the MAD1 Promoter-Knowing that the transactivation by p53 is mediated through direct binding of the protein to its binding consensus on target genes (7, 55, 56), we inspected the MAD1 promoter, but there was no match for the p53-binding site. Recently, an analysis of p53-binding elements showed that the previously defined p53 binding consensus comprising two copies of the sequence RRRC(A/T) arranged in a head-to-head fashion, with 0 -13 nucleotides apart (7), allowed p53 activation of the target genes, whereas a head-to-tail arrangement shifted the activity of p53 from activation to repression (57). Nevertheless, none of these sequence arrangements were found in the MAD1 promoter. The absence of p53-binding consensus prompted us to search for the p53-responsive region in the MAD1 promoter. Serial truncation of the MAD1 promoter was constructed by PCR amplification, and these fragments were cloned upstream of the CAT gene in pCAT-Basic vector, as represented in the schematic diagram (Fig. 6A). Hep3B cells were transfected with these truncated constructs alone or together with wildtype or p53 mutant R175H and assayed for CAT activities (Fig.  6B). Even though 326 bp were deleted from the 5Ј of 480-bp MAD1 promoter (i.e. p102CAT), robust basal promoter activity  FIG. 4. Definition of MAD1 promoter and transcriptional repression of MAD1 promoter by p53 in HepG2 and Hep3B cells. A, DNA sequence of human MAD1 promoter. The 480-nucleotide sequence shown is located immediately upstream of the exon 1 of human MAD1 gene. The putative TATA sequence (Ϫ17/Ϫ13) is underlined; the transcription start site predicted is indicated (arrow) and is numbered as ϩ1, and the was still retained (lane 12). This suggested that the minimal promoter region may reside within the region of 154 bp between ϩ52 and Ϫ102. Like wild-type p53, R175H mutant displayed repression activity on the 480-bp MAD1 promoter, albeit weaker activity than the wild-type by about 4-fold (compare lane 5 with lane 1). After all, p53 was not able to repress the promoter with 117 nucleotides deleted from the 5Ј (i.e. p311CAT) or with further truncations (i.e. p231CAT and p102CAT). The restoration of promoter activities in p428CAT versus p311CAT in the presence of wild-type and mutant p53 consistently suggested that the p53-responsive element was located in a region of 117 bp between Ϫ311 and Ϫ428 of MAD1 promoter.
To finely map the p53-responsive region, we generated another deletion between Ϫ311 and Ϫ428, with 79 nucleotides removed from the 5Ј of the MAD1 promoter (i.e. p349CAT, Fig. 7A). The promoter activity of p349CAT was repressed in the presence of wild-type p53, resembling that of the p428CAT (Fig. 7B). The repression by p53, however, was abolished when the promoter was truncated down to position Ϫ311, where an appreciable increase of promoter activity was observed. The equivalent amounts of p53 protein expressed in all p53-transfected samples as shown by Western blotting (Fig. 7B, lower panel, lanes 1-3) eliminated the possibility that the rise of promoter activity was due to the decline of p53 expression. The promoter truncation experiments thus distinctly delineated the element responsible for p53 repression within a region of 38 bp between Ϫ311 and Ϫ349 of MAD1 promoter (Fig. 4A, dashed line).
Moderation of p53-mediated Repression of MAD1 by Trichostatin A-It is generally thought that transcriptional repres-sors are often associated with histone deacetylases (HDACs) because deacetylases have been shown to be involved in several gene repression systems (58). The histone deacetylase activity in particular is required for the HDAC1-mediated repression of target genes (59). First, we addressed whether histone deacetylase activity is required for the repression of MAD1 by p53. Hep3B cells were transiently co-transfected with p428CAT and wild-type p53, in the absence or presence of various concentrations of trichostatin A (TSA). TSA is a potent and specific inhibitor of HDACs in vitro and in vivo (60). Although wild-type p53 repressed MAD1 promoter activity (Fig. 8, lane 7), addition of TSA seemed to alleviate the repression (lanes 8 -11). The MAD1 promoter activity increased ϳ3-fold when 2 M of TSA was added to the culture medium (compare lane 7 with lane 8). However, the promoter activity did not increase further even at higher concentrations up to 5 M TSA (lane 11). Although the repression could not be completely abrogated by TSA, the redemption of MAD1 promoter activity by TSA in the presence of p53 is prominent. Western blot analyses showed that the amounts of p53 were relatively constant in the absence and presence of TSA, ruling out the possibility that the increase of promoter activity was due to the decline of p53 protein expression. The experiment was repeated using p53 mutant R175H instead of wild-type p53, and de-repression of MAD1 promoter activity was also observed in the presence of TSA (data not shown). A previous study (61) has demonstrated that TSA alone was able to induce the bcl-2 minimal promoter activity. To determine whether TSA exhibited the same effect on MAD1 promoter, we assayed the CAT activity in cells transfected only with p428CAT without pCMVp53 in the presence of TSA. There was no apparent increase in MAD1 promoter activity when cells were treated with TSA (compare lane 1 with lanes 2-6). These data illustrate the ability of TSA to revert the p53-dependent repression, suggesting that histone deacetylase activity is implicated in the repression mechanism.
p53, HDAC1, and mSin3a Are Associated with the MAD1 Promoter in Vivo-HDAC1 has been found to interact with a transcriptional co-repressor mSin3a in a multiprotein complex which represses transcription of many genes (62). Recent reports (10, 61) revealed that p53 forms a complex with mSin3a and HDAC1 in vivo, and this interaction is critical for p53mediated transcriptional repression of at least some target genes. The TSA effect described above strongly suggests that p53 might repress the transcription of MAD1 through HDACs. Therefore, we explored the interaction of individual protein in the HDAC1 repression complex with endogenous MAD1 promoter by chromatin immunoprecipitation (ChIP) using antibodies specific for p53, HDAC1, and mSin3a, as well as using pre-immune serum (IgG) as a negative control. p53-null Hep3B cells were transfected with empty vector or pCMVp53, followed by formaldehyde treatment. The cross-linked chromatin from these two samples of cells was then immunoprecipitated with individual antibodies. The chromatin DNA was eluted after reversal of cross-links and treatment by proteinase K and RNase A. The cross-linked samples without adding antibody p53-responsive element (Ϫ349/Ϫ312) is highlighted (dashed line). B, CAT assay with HepG2 and Hep3B cells after UV irradiation. p53-positive HepG2 and p53-null Hep3B cells were transfected with either MAD1 promoter-CAT reporter construct p428CAT (lanes 2 and 3) or pCAT-Basic vector alone as a negative control (lane 1). Transfected cells (lane 3) were then exposed to 50 J/m 2 UV irradiation and harvested after 24 h. Cells were lysed, assayed for CAT activity (top), and analyzed by Western blotting using anti-p53 antibody (bottom). C, UV-induced reduction of MAD1 is p53-dependent. HepG2 and Hep3B cells were irradiated with 50 J/m 2 UV (ϩ) and harvested after 14 h. Cells not irradiated (Ϫ) serve as negative control. Cell lysates were analyzed for MAD1, p53, and ␤-actin protein expression by Western blotting. D, CAT assay with Hep3B cells transfected with p53. Hep3B cells were transfected with either p428CAT (lanes 2-5) or empty pCAT-Basic vector as a negative control (lane 1). Cells in lanes 3-5 were co-transfected with pCMVp53 at 0.1, 0.5, and 12.5 g, respectively. After transfection, cells were lysed and analyzed for their promoter activity in CAT assay (top). The acetylated signals (top spots in the panel) were quantified, and their relative CAT activities were represented in folds (top of the panel) as compared with the sample without transfection with p53 (lane 2). Cell lysates were analyzed for p53 expression by Western blot using p53-specific antibody (DO-1, bottom).  lanes 2 and 3), V143A (lanes 4 -6), R175H (lanes 7-9), R248W (lanes 10 -12), and R273H (lanes 13-15) mutant plasmids with indicated amounts (g) were added to cells, respectively. Cells were lysed, and lysates of equal concentration were assayed for their CAT activity (top) and were analyzed by Western blotting using anti-p53 antibody (DO-1; bottom). The result shown is representative of three independent experiments.
p53 Represses MAD1 Expression was reserved as total input. The eluted DNA was PCR-amplified using primers specific for the MAD1 promoter. As controls, p428CAT reporter was used as a template in PCR to reveal the amplified signal, and water was also used as a template to show the background. As shown in Fig. 9, all three antibodies specific for p53, HDAC1, and mSin3a readily immunoprecipitated endogenous MAD1 promoter (lanes 10 -12) in p53-transfected cells, whereas the immunoprecipitates by these three antibodies in p53-deficient cells did not contain any detectable amount of chromatin (lanes 5-7) even when the amplification cycle of PCR was increased up to 50 (data not shown). The specificity of these three antibodies was justified by the facts that no detectable signal was delivered by IgG immunoprecipitates (lanes 4 -9) and that these three antibodies when used to detect their respective proteins in Hep3B cell lysates yielded discrete bands (data not shown). These data clearly show that p53, HDAC1, and mSin3a proteins physically interact with the MAD1 promoter in vivo and lend support to the existence of the p53-repression complex comprising p53, HDAC1, and mSin3a in the MAD1 promoter. Remarkably, neither HDAC1 nor mSin3a bound to the MAD1 promoter in the absence of p53 ( lanes 5-7), suggesting that the recruitment of HDAC1 and mSin3a to the MAD1 promoter in vivo is p53-dependent. DISCUSSION As a ''guardian of the genome,'' tumor suppressor p53 protein is activated in response to a variety of stimuli and stresses such as DNA damage, hypoxia, microtubule disruption, and activated oncogenes (2). Activated p53 is primarily involved in G 1 /S or G 2 /M arrest in the cell cycle. One well characterized function of p53 is the transactivation of its downstream target genes via specific DNA sequence binding (7). p53 also exhibits transcriptional repression activity on a diverse set of genes. Here we present another p53-regulated target, a mitotic checkpoint protein, MAD1. In the current working model for mitotic checkpoint control, MAD1 recruits MAD2 to the unattached kinetochore during prometaphase, thereby eliciting the checkpoint. After dissociation from MAD1, MAD2 is transferred to Cdc20, leading to an inhibition of its ability to activate APC. When all kinetochores are associated with microtubules, MAD2 no longer localizes to kinetochore and dissociates from CDC20, thus inducing APC activation and anaphase onset (17). Although   FIG. 6. Effects of p53 wild-type and mutant R175H on MAD1 promoter truncations in Hep3B cells. A, schematic diagram of the serial truncation of MAD1 promoter. The transcription start site predicted is numbered as ϩ1 (Fig. 3). Three promoter truncations were generated by PCR as described under ''Experimental Procedures'' and cloned upstream of the CAT gene in the pCAT-Basic reporter plasmid. Their resulting reporter constructs are labeled. B, transcription repression of truncated MAD1 promoters by wt-p53 and p53 mutant. 5 g of four promoter constructs (designated as Ϫ428, Ϫ311, Ϫ231, and Ϫ102) and 1 g of p53 wild-type (lanes 1-4) or p53 mutant R175H (lanes 5-8) were co-transfected into Hep3B cells. Hep3B cells transfected with promoter constructs alone serve as controls (lanes 9 -12). The promoter activity was detected in CAT assay (top). The CAT activities of promoter Ϫ428 in the presence of wt-p53 (lane 1) and R175H mutant (lane 5) were compared with the control (lane 9) and represented in folds (top of panel). The p53 protein level in cells was detected by Western blotting using p53-specific antibody (DO-1; middle). The same blot was probed with anti-␤actin antibody to provide reference of equal loading of samples (bottom). MAD1 can be modified by phosphorylation in vitro by MPS1 and BUB1 (18), regulation of MAD1 remains elusive. In this study, we demonstrate that MAD1 is repressed by p53 at a transcriptional level.
Transcriptional Repression of MAD1 Expression by p53-p53-induced transcriptional repression of genes involved in G 2 /M transition include cyclin B1 (63), CDC2 (14), and CDC25C (64), and recent studies (65,66) have shown that the anti-apoptotic surviving gene, survivin, is another p53 repressed target. Survivin is maximally expressed in the G 2 /M phase, and this protein is localized to mitotic spindle microtubules during mitosis, likely monitoring mitotic spindle integrity. p53 is also required for post-mitotic checkpoint and G 1 arrest after mitotic spindle disruption by microtubule-depolymerizing drug, nocodazole (67)(68)(69)(70). Furthermore, p53 co-localizes with mitotic spindles and centrosomes in normal chick epiblast and human cells, suggesting an involvement of p53 in spindle function (71). An early experiment has shown that p53-deficient fibroblasts continued cell cycle progression after mitotic spindle disruption by nocodazole and formed polyploid cells, implicating p53 in mitotic checkpoint (72). Therefore, it is not surprising that p53 is also involved in regulation of mitotic checkpoint genes. MAD1, previously characterized to be a Tax-binding protein (also known as TXBP 181) (22), has been found using the SAGE technique to be a gene (PIG9) up-regulated in response to p53 expression in colorectal cancer cell DLD-1 (39). In light of this, we intended to further characterize the regulation of MAD1 by p53 in the present study. Surprisingly, our results reproducibly indicate that MAD1 is repressed by p53. Transient transfection of exogenous p53 into HepG2, IMR-90, Hep3B, PC-3, and HeLa cells yielded a decline of MAD1 expression ( Fig. 1 and Fig. 2). To eliminate the possibility of artifacts created by transfecting exogenous source of p53 into cells, we triggered endogenous p53 expression by chemotherapeutics and UV ( Fig. 3 and Fig.  4). We have employed more cell lines derived from various origins including CNE2 (nasopharyngeal), MCF-7 (breast), LoVo (colon), IMR-90 (mortal normal lung fibroblast), PC-3 (prostate), and SK-OV-3 (ovary) to ascertain that the effect of p53 on MAD1 level is a general phenomenon. Down-regulation of MAD1 was again observed under these conditions, consistent with the results obtained by transfecting cells with p53-expression plasmid pCMVp53. We can only hypothesize that the discordance of our data from the SAGE results (39) might be due to a specific response of DLD-1 cells to p53 or technical issues involved in the SAGE analysis (73). In DLD-1 cells, the p53 gene has a 241F mutation, which has been shown to be defective for transactivation (74,75). It is not fully understood whether and how the p53-241F mutant might interfere with the regulation of MAD1 by wild-type p53. During the preparation of this manuscript, an independent study on p53 regulation of MAD1 was published (76). In that study, wild-type p53 was found to be a poor activator of MAD1 promoter in HeLa, HCT116, and HCT116 p53 Ϫ/Ϫ cells. However, a tumor-promoting gain-of-function p53 mutant (p53-281G) exhibited a distinctly potent transactivating activity on MAD1. Taken together with our findings that wild-type p53 represses MAD1, we postulate that the subversion of a p53 regulatory function on MAD1 promoter by some gain-of-function p53 mutants such as p53-281G might contribute to their tumor-promoting abili-  1 and 7). Western blotting using anti-p53 (DO-1), anti-HDAC1 (H-11), and anti-mSin3a (G-11) antibodies (bottom) was performed.
ties. In this regard, it will be of great interest to see whether the p53-241F mutant in DLD-1 cells might have such activity.
One might postulate two mechanisms by which p53 regulates MAD1, protein degradation or transcriptional repression. By taking advantage of proteasome inhibitors, we sought to rule in or rule out the first possibility. If p53 promotes ubiquitination of MAD1, the proteosome inhibitors would block the degradation activity, and the protein level of MAD1 would not decrease even in the presence of p53. Nevertheless, addition of proteosome inhibitors to cisplatin-treated cells did not elevate the MAD1 protein level from repressive to normal status. p53 maintained its repressive activity on MAD1 protein expression (Fig. 3). These results, in keeping with data from promoter analysis (Fig. 4), strongly implicate transcriptional control in the regulation of MAD1 by p53.
Identification of a Novel p53-responsive Element in MAD1 Promoter-Both overexpression of exogenous p53 and induction of endogenous p53 resulted in a decrease in MAD1 promoter activity (Figs. [1][2][3][4][5]. By using CAT-reporter assay, cotransfection of p53 and serially truncated MAD1 promoter-CAT constructs helped us to map the p53-responsive element. A rough determination indicated that the region lies between Ϫ428 and Ϫ311 of the promoter (Fig. 6), whereas the activity increased when MAD1 promoter was deleted from Ϫ349 to Ϫ311 (Fig. 7). These results illustrate that the region between Ϫ349 and Ϫ311 in the MAD1 promoter is susceptible to p53 repression. Transactivation by p53 requires the binding of the central "DNA-binding domain" of the protein to a consensus binding site composed of two half-sites, arranged head-to-head (7). In contrast, p53-mediated transrepression of at least some genes required the interaction of the protein to a binding site with head-to-tail orientation. Such head-to-tail sites can be found in the promoter of cyclin A and B1 genes, which are transcriptionally down-regulated by wild-type p53 (57). However, neither site was found in the MAD1 promoter.
p53 is also known to repress gene promoter activity in a p53 binding consensus-independent manner even though the consensus sequence is present in the promoter, suggesting that p53 consensus binding site is essential for activation, but it may not be the prerequisite in the case of p53 repression (66,77). Indeed, p53 negatively regulates a variety of genes that lack a p53-responsive element, including c-fos, c-jun, retinoblastoma, and interleukin 6 genes (78, 79). Thus, it is not surprising to see p53 repression activity on the MAD1 promoter in the absence of the p53 consensus binding site. The MAD1 promoter thereby presents a novel p53-responsive element that is distinct from the known p53 consensus binding site (7) or other thus far known sequences required for p53 repression. Whether specific nucleotide within this 38-bp p53-responsive element is crucial for p53 repression remains unknown, and further deletion or mutation of specific nucleotides may help elucidate this issue.
p53 Represses MAD1 Expression through HDAC-mSin3a-We showed that the repression of MAD1 by p53 was alleviated when HDAC inhibitor TSA was added to the cells (Fig. 8). Histone acetylation levels are generally believed to be correlated with transcription levels, in that histone deacetylation may repress transcription by intensifying DNA-histone tail interactions and hence blocking the access of transcriptional modulators to the DNA template or by removing the acetyl groups on histone tails that are crucial to the association of transcriptional modulators with chromatin. A number of transcriptional co-regulators are histone acetylases or histone deacetylases (58). HDACs thus play important roles in gene regulation by being a family of enzymes catalyzing the removal of acetyl groups from post-translationally acetylated proteins. In particular, HDAC1 can interact with p53 in vitro and in vivo (10), and HDAC activity is important for HDAC1-mediated transcriptional repression (59). In our study, addition of TSA reversed the effect exerted by p53, leading to de-repression of the MAD1 promoter (Fig. 8). In addition to deacetylating histones, HDAC1,-2, and-3 are able to deacetylate p53 at the C terminus that can be acetylated by p300/CBP in vivo (80,81), resulting in suppression of the p53 transactivation activity. Furthermore, acetylation of p53 activates transcription through recruitment of co-activators/histone acetyltransferases to the promoters of p53-responsive genes (82). These mechanisms may also contribute to the de-repression of MAD1 promoter by TSA. Thus, it is of interest to determine the acetylation level of p53 in the context of MAD1 repression.
HDAC1 interacts with the co-repressor mSin3A to form a repression complex (83) which is recruited to specific promoters via association with transcription factors such as Sp1/3 (84) and E2F (85). Transcriptional repression of MAP4, stathmin, and BCL12 by p53 has been demonstrated to be mediated by HDAC1 and mSin3a (10,61). Our results from the ChIP assay prominently illustrated that p53, HDAC1, and mSin3a proteins bind to MAD1 promoter in vivo in Hep3B cells transfected with p53 (Fig. 9). Remarkably, neither HDAC1 nor mSin3a associated with MAD1 promoter when p53-deficient Hep3B cells were not transfected with p53, indicating that recruitment of the HDAC1-mSin3a complex to MAD1 promoter is p53-dependent. Based on our findings that HDAC activity is required in the repression of MAD1 (Fig. 8) and that p53 mediates the recruitment of HDAC and mSin3a to the MAD1 promoter ( Fig.  9), we can speculate that the repression mechanism might result, at least in part, from deacetylation of nucleosomes in the vicinity of the MAD1 promoter by active HDAC recruited by p53 to the promoter (86). The importance of the presence of mSin3a in the complex is that it interacts with the proline-rich region of p53 and stabilizes p53 in an MDM2-independent fashion, providing prolonged periods for effective gene repression (87). Our data indicate that HDAC inhibitor TSA did not fully de-repress mSin3a-mediated repression (Fig. 8), and a similar observation has also been mentioned in other reports (88). These raise the possibility that other factors may also be present in the p53-HDAC1-mSin3a repression complex.
Repression of MAD1 by p53 Mutants-p53 transactivation requires binding of p53 to DNA in a sequence-specific manner, but tumor-derived p53 proteins including mutations at amino acid 143, 175, 248 and 273 lose the ability of activation (40). Similar to wild-type p53, these p53 derivatives with mutations clustered within the DNA-binding domain of p53 also displayed transcription repression activity of MAD1 (Figs. 5 and 6). It has been shown that R175H and R273H mutants induced apoptosis in Hep3B cells (89), and cells with these p53 mutants exhibited disrupted mitotic checkpoint control (90). These results suggest that transactivation activity of p53 is not required for these activities. Indeed, the absence of recognized p53-binding consensus sequence in MAD1 promoter implies that the mechanism used in recognition of p53-responsive element during p53 repression may be totally different from that used in activation. Dose-dependent repression of MAD1 promoter by both wildtype and mutant p53 (Fig. 5) illustrates that p53 repression activity depends on its protein level rather than the integrity of its DNA-binding domain. Recently, a minimal p53 repression domain has been mapped to C-terminal amino acids 339 -346 (91). This finding together with the repression ability of p53 mutants points to a notion that a functional sequence-specific DNA-binding domain of p53 might not be required for p53 repression. The p53 R175H mutant exerted the same effect, as the wild-type, on MAD1 promoter (Fig. 6), and its repressive activity was also alleviated by TSA treatment (data not shown), suggesting that at least some p53 mutants may share the same repression mechanism as the wild type. In support of this model, R175H was able to bind heat shock protein hsp70 (92). Moreover, HDAC-1 and -2 co-immunoprecipitated with Hsp70 from HeLa whole cell extract. In particular, this precipitate contained significant HDAC activity (93).
In conclusion, we have demonstrated that both wild-type and mutant p53 transcriptionally repress MAD1. We have identified a novel p53-responsive element distinct from the canonical p53 binding sequence, and the repression is mediated by recruitment of HDAC1-mSin3a repression complex to the promoter. A growing body of evidence indicates the importance of MAD1 in the regulation of MAD2 in the mitotic checkpoint signaling. On top of protein-protein interaction and protein modification, transcription regulation provides another layer of control in which p53 is apparently an upstream regulator of MAD1. It remains to be understood exactly how p53-mediated repression of MAD1 contributes to the tumor suppressor activity of p53 or the mitotic checkpoint function of MAD1. In this regard, it is tempting to hypothesize that the repression of MAD1 might lead to the release of MAD2, which in turn binds to Cdc20, inhibits APC, and induces mitotic arrest. This model is compatible with the notion that MAD1 competes with Cdc20 for the binding with MAD2 (35)(36)(37)(38). In this scenario, the repression of MAD1 may contribute to the induction of G 2 /M arrest by p53. Nevertheless, further investigations are required to elucidate the biological significance of p53-mediated repression of MAD1.