Partial downregulation of MAD1 causes spindle checkpoint inactivation and aneuploidy, but does not confer resistance towards taxol

Partial downregulation of MAD1 causes spindle checkpoint inactivation

and aneuploidy, but does not confer resistance towards taxol

Anne Kienitz1,2, Celia Vogel1,2, Ivonne Morales1, Rolf Muller1 and Holger Bastians*,1

1Institute for Molecular Biology and Tumor Research (IMT), Philipps University Marburg, Emil-Mannkopff-Strasse 2,D-35037 Marburg, Germany

The mitotic spindle assembly checkpoint ensures properchromosome segregation during mitosis by inhibiting theonset of anaphase until all kinetochores are attached tothe mitotic spindle and tension across the kinetochores isgenerated. Here, we report that the stable partial down-regulation of the spindle checkpoint gene MAD1, which isobserved in human cancer, leads to a functional inactiva-tion of the spindle checkpoint resulting in gross aneu-ploidy. Interestingly, although Mad1 is thought to act asa kinetochore based activator of Mad2 during checkpointactivation, we show that normal levels of Mad2, but notof Mad1, are required for preventing premature sisterchromatid separation and for maintaining the timing of anundisturbed mitosis, suggesting a Mad1 independentfunction of Mad2 that operates independent of itscheckpoint function. Most significantly, a partial repres-sion of either MAD1 or MAD2 confers resistance tonocodazole, a drug that inhibits microtubule attachment.In contrast, sensitivity to clinically relevant drugs liketaxol or monastrol that inhibit the generation of tensionacross kinetochores is not modulated by partial down-regulation of MAD1, suggesting a functional bifurcationof spindle checkpoint dependent apoptotic pathways.Oncogene (2005) 24, 4301–4310. doi:10.1038/sj.onc.1208589Published online 14 March 2005

Keywords: cell cycle; mitosis; mad2; apoptosis; genomicinstability

Introduction

The faithful segregation of chromosomes during mitosisrequires that all chromosomes are correctly attached tothe mitotic spindle apparatus and properly aligned atthe metaphase plate before anaphase onset is allowed.During the early phases of mitosis when chromosomesare not yet aligned, the mitotic spindle assemblycheckpoint is activated (Campbell and Gorbsky, 1995;Li and Nicklas, 1995; Rieder et al., 1995) and preventsthe onset of anaphase by inhibiting the anaphase

promoting complex/cyclosome (APC/C) (Li et al.,1997; Fang et al., 1998; Gorbsky et al., 1998). Thislarge ubiquitin ligase complex is required for theubiquitination and destruction of securin and cyclin B,prerequisites for anaphase onset and mitotic exit,respectively (for a review see Peters, 2002). The spindlecheckpoint is also activated in response to variousspindle poisons like nocodazole, a drug that depoly-merizes microtubules, and thus, prevents the attachmentof microtubules to the kinetochores. On the other hand,anticancer drugs, like taxanes (e.g. paclitaxel/taxol),inhibit the dynamic instability of the spindle and allowmicrotubule attachment, but prevent the generation oftension across kinetochores. Treatment of cells withthese drugs inhibits chromosome alignment and resultsin a mitotic arrest before anaphase, which is followed bythe induction of apoptosis (Jordan and Wilson, 2004).

To date, the molecular pathways of the spindlecheckpoint are not well understood, but it is establishedthat several evolutionary conserved proteins includingCenp-E, BubR1, Bub1, Bub3, Mad1, Mad2 and Mps1are required for spindle checkpoint function (for areview see Yu, 2002; Bharadwaj and Yu, 2004; Tayloret al., 2004). These and most likely additional yetunknown proteins are sequentially recruited to unat-tached or tension-lacking kinetochores leading to thegeneration of a diffusible APC/C inhibitor (Shah andCleveland, 2000; Musacchio and Hardwick, 2002;Vigneron et al., 2004). Mad2, BubR1, and also a mitoticcheckpoint complex (MCC) consisting of Mad2,BubR1, Bub3 and Cdc20 can function as potentinhibitors for APC/C activity, but the mechanism ofthe MCC formation is still unclear (Fang et al., 1998;Sudakin et al., 2001; Tang et al., 2001; Fang, 2002).According to a current model, an unattached kineto-chore serves as an activating platform for the diffusibleAPC/C inhibitor. Consistently, Mad1 is required forrecruitment of Mad2 to kinetochores and complexformation of Mad1 and Mad2 at the kinetochore isimportant for checkpoint activation, suggesting thatMad1 might function as an activator of Mad2 at thekinetochore (Jin et al., 1998; Chen et al., 1999; Campbellet al., 2001; Sironi et al., 2001; Luo et al., 2002; Martin-Lluesma et al., 2002).

Interestingly, Mad2 and BubR1 appear to be alsoinvolved in regulating the normal timing of mitosis.Inhibition of Mad2 or BubR1, but none of the other

Received 11 August 2004; revised 27 January 2005; accepted 4 February2005; published online 14 March 2005

*Correspondence: H Bastians; E-mail: [email protected] authors contributed equally to this work

Oncogene (2005) 24, 4301–4310& 2005 Nature Publishing Group All rights reserved 0950-9232/05 $30.00

www.nature.com/onc

Mad and Bub proteins, causes an acceleration of mitosis(Gorbsky et al., 1998; Canman et al., 2002; Meraldiet al., 2004). Moreover, the inhibition of their kineto-chore localization does not shorten the time fromprophase to anaphase indicating that Mad2 and BubR1might be part of a kinetochore independent mitotictimer mechanism (Meraldi et al., 2004).

The majority of human cancer cells are aneuploid,which can be caused by gaining or loosing wholechromosomes during defective chromosome segregationevents (Duesberg et al., 2000; Sieber et al., 2003). Thisphenotype is called chromosomal instability (CIN) andalthough its molecular basis is still unknown, it is nowapparent that malfunction of the spindle checkpoint candirectly contribute to CIN (Lengauer et al., 1998; Michelet al., 2001; Masuda and Takahashi, 2002; Babu et al.,2003; Rajagopalan and Lengauer, 2004). Indeed, defectsin the spindle checkpoint are frequently found in varioustypes of human cancer including lung, breast, ovarian,colon and hepatocellular tumors and these defects areassociated with CIN and aneuploidy (Cahill et al., 1998;Takahashi et al., 1999; Masuda and Takahashi, 2002;Shichiri et al., 2002). While spindle checkpoint defectsare frequent, spindle checkpoint genes appear to berarely altered in human tumors (Cahill et al., 1998;Nomoto et al., 1999; Hernando et al., 2001). Instead,downregulated expression of spindle checkpoint genesmight contribute to a deactivated spindle checkpoint incancer cells. In fact, the genes of the MCC, MAD2,BUB3 and BUBR1, have been shown to be haplo-insufficient (Michel et al., 2001; Babu et al., 2003; Bakeret al., 2004). Heterozygous deletion of MAD2 in humancolon carcinoma cells and heterozygous MAD2, BUB3or BUBR1 disruptions in mice result in partiallydownregulated checkpoint protein levels, an impairedspindle checkpoint and aneuploidy (Michel et al., 2001;Babu et al., 2003; Baker et al., 2004). Most significantly,MAD2þ /�, BUB3þ /� and BUBR1þ /� mice are prone totumor development suggesting that aneuploidy cancontribute directly to tumorigenesis (Michel et al.,2001; Babu et al., 2003; Dai et al., 2004). To date, it isunclear if a moderate downregulation of other spindlecheckpoint genes directly contributes to aneuploidy andtumorigenesis; however, a reduced expression of MAD2and BUBR1, but also of BUB1 and MAD1, has beenfound in different human cancers (Li and Benezra, 1996;Shichiri et al., 2002; Wang et al., 2002).

Surprisingly, spindle checkpoint defects have recentlybeen associated with resistance to spindle damaginganticancer drugs, for example, paclitaxel/taxol (Kasaiet al., 2002; Anand et al., 2003; Masuda et al., 2003;Sudo et al., 2004) and with the resistance to topoiso-merase poisons, for example, etoposide, adriamycin/doxorubicin (Vogel et al., 2004b). These findings indicatethat the spindle checkpoint might be directly involved inthe induction of apoptosis upon mitotic damage andimply that tumors harboring a compromised spindlecheckpoint are resistant to such chemotherapeutic treat-ments. However, the molecular mechanisms of thecrosstalk between the spindle checkpoint and apoptoticpathways remain to be determined.

A function of Mad1 together with Mad2 for thespindle checkpoint has been demonstrated in cellsdepleted of Mad1 by transient siRNA transfections(Luo et al., 2002; Martin-Lluesma et al., 2002; Meraldiet al., 2004). However, severe depletion or homozygousdisruption of spindle checkpoint genes results in celldeath indicating that spindle checkpoint genes areessential for viability (Dobles et al., 2000; Kalitsiset al., 2000; Kops et al., 2004; Michel et al., 2004).Therefore, previous studies could not address the long-term physiological consequences of Mad1 inactivation.Moreover, in human cancer, only partial downregula-tion of MAD1 or MAD2 is observed and it is importantto resolve whether a partial downregulation of MAD1or MAD2 is functionally redundant. Therefore, weinvestigated the phenotypes associated with partialdownregulation of MAD1 or MAD2. Our resultsindicate a differential requirement of Mad1 and Mad2for spindle checkpoint activation, mitotic timing andinduction of apoptosis in response to chemotherapeuticdrugs in human cancer cells.

Results

Generation of human colon carcinoma cells exhibitingreduced levels of Mad1

We wished to generate karyotypically stable human cellsexhibiting partial downregulation of MAD1. For this,we stably expressed an siRNA targeting MAD1 inHCT116 cells, which are well characterized for theirintact cell cycle checkpoints and their relatively stablekaryotype. In addition, HCT116 cells have been usedpreviously for heterozygous gene deletion of MAD2(HCT116-MAD2þ /�) resulting in cells with approxi-mately 20% reduction of Mad2 protein level (Michelet al., 2001). This enabled us to investigate and tocompare the long-term cellular consequences of MAD1and MAD2 downregulation. Upon stable transfectionwe selected two independent clones, termed knockdownclone 2-2 (MAD1-kd-2-2) and knockdown clone 2-31(MAD1-kd-2-31) for the experiments throughout thisstudy. These clones showed about 35 and 50% reductionof endogenous Mad1 protein, respectively, when com-pared to HCT116-wild-type cells (HCT116-wt) and topuromycin resistant control cells (HCT116-pSUPER) asshown on Western blots using specific anti-Mad1antibodies (Figure 1a). When grown asynchronously inculture, HCT-MAD1-kd cells did not show significantchanges in cell cycle distribution, proliferation rate orapoptosis indicating that a repression of MAD1 isneither affecting the cell cycle timing per se nor causinglethality (Figure 1b).

Partial downregulation of MAD1 or MAD2 resultin a defective spindle checkpoint

To investigate whether a partial downregulation ofMAD1 would account for a defective spindle assemblycheckpoint function, we determined the mitotic arrest in

MAD1 downregulation causes aneuploidyA Kienitz et al

4302

Oncogene

response to nocodazole, which prevents microtubuleattachment to the kinetochores, and thus, chromosomealignment (Figure 2a). Both MAD1-kd cell lines (2-2and 2-31) showed a mitotic index with maximum valuesof 30–40% similar to that observed in HCT-MAD2þ /�

cells. In contrast, a maximum of approximately 70% ofthe wild type and control cells accumulated in mitosis.Further, isogenic p53 deficient cells (HCT-p53�/�)arrested like wild-type cells indicating no requirementof p53 for normal spindle checkpoint function. Verysimilar results were observed after treatment with taxolor monastrol, drugs that inhibit the spindle dynamicsand the formation of a bipolar spindle, respectively,allowing microtubule attachment, but not the genera-tion of tension (Figure 2b and c). Our results clearlydemonstrate that partial downregulation of eitherMAD1 or MAD2 results in a very similar functionalinactivation of the spindle checkpoint in response tolack of microtubule attachment and upon lack oftension across kinetochores.

Consistently, the defect in spindle checkpoint activa-tion in HCT-MAD1-kd and HCT-MAD2þ /� cells wasaccompanied by a premature exit from mitosis in thepresence of nocodazole. This was demonstrated by thepremature degradation of cyclin B (Figure 2d) and bygeneration of multinuclei around groups of chromo-somes (Figure 2e). Quantification of multinucleirevealed that after 18 h of spindle damage 60–65%of both, MAD1-kd cells and HCT-MAD2þ /� cells,had exited without cytokinesis whereas only 3–8% ofthe control cells had formed multinuclei after 18 hof nocodazole treatment (Figure 2e). Under theseconditions, the multinuclei contained a tetraploid

DNA content as demonstrated by FACS analyses(Figure 2f). Furthermore, the premature exit frommitosis in spindle checkpoint deficient cells resulted ina similar high rate of endoreduplication in MAD1-kdand HCT-MAD2þ /� cells (data not shown; Vogel et al.,2004a). Taken together, partial downregulation of eitherMAD1 or MAD2 results in a very similar functionalinactivation of the spindle checkpoint.

Reduced levels of MAD2 but not of MAD1 inducea high rate of premature sister chromatid segregation

Spindle checkpoint inhibition is expected to result inpremature sister chromatid separation. To test whetherpartial downregulation of MAD1 or MAD2 results inpremature sister chromatid segregation, we performedmetaphase chromosome spread analyses (Figure 3a) anddetermined the proportion of cells showing disjoinedsister chromatids (Figure 3b). Upon a 2-h nocodazoletreatment, HCT-MAD2þ /� cells displayed a high rateof premature anaphases (11–15%) demonstrating, inagreement with previous findings (Michel et al., 2001),that normal levels of Mad2 are required to prevent grosspremature sister chromatid separation. Upon prolongednocodazole treatment (6 h), the proportion was evenincreased to 16–21%. In contrast, both MAD1-kd celllines showed much lower proportions of prematureanaphases after 2 and 6 h treatment (2.1–3.3% and3–5%, respectively) corresponding to just about 20% ofthe rate observed in HCT-MAD2þ /� cells (Figure 3b).Our findings suggest the premature sister chromatidseparation might be restrained by a Mad1 independentfunction of Mad2.

Normal levels of Mad2 but not of Mad1 are requiredfor normal mitotic timing

Previous studies have indicated that the spindle check-point regulates the timing of a normal mitosis in theabsence of damage (Taylor and McKeon, 1997;Gorbsky et al., 1998; Shannon et al., 2002; Meraldiet al., 2004). Moreover, a kinetochore independentfunction of Mad2 and BubR1, but not of other Mad orBub proteins, appears to be required for the mitotictimer function (Meraldi et al., 2004). Therefore, weinvestigated whether partial downregulation of MAD1or MAD2, which is observed in human cancer, mightdisturb the normal timing of mitosis. We treated asyn-chronously growing HCT-MAD1-kd, HCT-MAD2þ /�

and control cells with the proteasome inhibitor ALLNfor up to 3 h to block mitotic progression beyondmetaphase in the absence of spindle damage anddetermined the proportion of metaphase figures (Fig-ure 4). All cell lines exhibit similar growth ratesand similar rates of mitotic entry without the additionof ALLN (DMSO control). As expected, ALLNtreatment increased the proportion of metaphases incontrol cells to 35–38%, indicating that the mitoticprogression is indeed blocked by proteasome inhibition.In contrast, in HCT-MAD2þ /� cells metaphases wereincreased to 62–64% indicating an acceleration of

Figure 1 Generation of HCT116 cells stably expressing siRNAtargeting MAD1. (a) Reduced protein levels of Mad1 in MAD1knockdown cells. Lysates from two independent knockdown clones(Mad1kd(2-2) and Mad1kd(2-31)), control cells (HCT116þ pSU-PER) and parental cells (HCT116-wt) were subjected to Westernblot analysis using anti-Mad1 and anti-actin antibodies. Mad1antibody signals were quantified in four independent experiments,normalized for actin signals and the Mad1 protein expression levelwas calculated in percent of HCT116-wild-type expression. Arepresentative experiment is shown. (b) FACS analyses ofasynchronously growing MAD1 knockdown and control cell lines

MAD1 downregulation causes aneuploidyA Kienitz et al

4303

Oncogene

mitosis. Significantly, however, accumulation in meta-phase of both MAD1-kd clones was not increasedcompared to control cells indicating that prometaphase

is not accelerated upon MAD1 repression. Our resultssupport previous findings showing that Mad2 andBubR1, but no other Mad or Bub proteins, are required

MAD1 downregulation causes aneuploidyA Kienitz et al

4304

Oncogene

for the mitotic timer function (Meraldi et al., 2004), andshow that partial repression of Mad2, but not of Mad1as observed in human tumor cells, is expected to disturbmitotic timing.

Partial repression of MAD1 or MAD2 result inaneuploidy

Since partial downregulation of MAD1 neither causesgross premature sister chromatid separation nor pertur-bation of normal mitotic timing, we asked whetherpartial downregulation of MAD1 in tumor cells wouldaccount for the generation of aneuploidy. We performedmetaphase chromosome spreads of HCT-MAD1-kdcells, HCT-MAD2þ /� cells and control cells anddetermined the number of individual chromosomes(Figure 5). The control HCT116 cell lines showed arelatively stable karyotype (76–78% normal karyotypewith 45 chromosomes). Significantly, HCT-MAD1-kdcells were found to be highly aneuploid (41 and 53%

Figure 2 Partial downregulation of MAD1 or MAD2 results in a defective spindle checkpoint. (a–c) The mitotic arrest in response tothe lack of microtubule attachment or upon the lack of tension is impaired in HCT-MAD1 knockdown and HCT-MAD2þ /� cells. Cellswere treated with (a) 150 nM nocodazole, (b) 100 nM taxol or (c) 70 mM monastrol for up to 48 h and the mitotic index was determined.The graph shows mean values from at least three independent experiments. (d) Partial downregulation of MAD1 or MAD2 results inpremature cyclin B degradation. The indicated cell lines were treated with nocodazole for 14 h and cyclin B protein levels were detectedon Western blots. (e) The premature exit from mitosis in MAD2þ /� and MAD1 knockdown cells is accompanied by the formation ofmultinuclei. MAD1 knockdown cells were treated with nocodazole for 18 h and nuclei were stained with Hoechst dye (left panel).HCT116 derivative cells were treated with nocodazole for 18 h and the proportion of mitotic, interphase and multinucleated cells weredetermined. At least 500 cells were counted for each cell line. (f) Multinucleated cells contain a tetraploid DNA content. Cells weretreated as in (e) and the DNA content was determined in FACS analyses

Figure 3 Different rates of premature sister chromatid separationupon partial downregulation of MAD1 and MAD2. (a) Example ofpremature sister chromatid separation in HCT-MAD2þ /� andHCT-MAD1-kd (clone 2-31) cells in comparison to intactchromatid cohesion in HCT116 control cells after treatment withnocodazole for 2 h. (b) Quantification of mitotic figures showingpremature sister chromatid separation after treatment withnocodazole for 2 and 6 h. Mean values and standard deviationsfrom four independent experiments are shown. At least 1600metaphase spreads were counted per cell line and time point

Figure 4 Reduced levels of MAD2, but not of MAD1, acceleratenormal mitotic timing. Asynchronously grown HCT-MAD2þ /�,HCT-MAD1 knockdown and control cells were treated with theproteasome inhibitor ALLN for 3 h to block mitotic progressionbeyond metaphase or with DMSO as a control. Cells were fixed,DNA was stained with Hoechst dye and mitotic figures werecounted. The proportion of metaphases was calculated and thegraph shows mean values and standard deviations from threeindependent experiments. At least 1800 mitotic figures werecounted per cell line

Figure 5 Partial downregulation of MAD1 or MAD2 result ingross aneuploidy. HCT116 control cells, HCT-MAD2þ /� cells andtwo independent HCT-MAD1 knockdown cell lines (2-2 and 2-31)were used for chromosome spread analyses and individualchromosomes were counted. The distribution of chromosomenumbers is shown and the percentage of aneuploidy was calculated(45 chromosomes represent the relative euploid karyotype ofHCT116 cells). At least 150 metaphase spreads were counted foreach cell line

MAD1 downregulation causes aneuploidyA Kienitz et al

4305

Oncogene

aneuploidy). In agreement with previous findings(Michel et al., 2001), we found HCT-MAD2þ /� cells tobe similarly aneuploid (55% aneuploidy). Thus, theprevention of aneuploidy requires normal levels of both,Mad1 and Mad2.

Tumor cells exhibiting reduced levels of MAD2, but notof MAD1 are resistant to taxol treatment

Spindle damaging agents are clinically highly relevantanticancer drugs (Jordan and Wilson, 2004). Interest-ingly, defects in the spindle checkpoint have beenrecently associated with a resistance to these chemother-apeutic drugs (Kasai et al., 2002; Anand et al., 2003;Masuda et al., 2003; Sudo et al., 2004). However, it islargely unknown how apoptosis is induced in responseto these treatments and what the parameters are thatdetermine sensitivity or resistance to spindle poisons.

Therefore, we investigated the induction of apoptosisin HCT-wt, HCT-p53�/�, HCT-MAD2þ /� and HCT-MAD1-kd cells in response to a treatment with drugsthat either inhibit microtubule attachment (nocodazole;Figure 6, left panels) or that prevent the generation oftension across kinetochores (taxol and monastrol;Figure 6, middle and right panels). The induction ofapoptosis was monitored by the detection of theapoptosis specific laddering of chromosomal DNA(Figure 6a), by the detection of the apoptosis specificcleavage product of poly-(ADP-ribose)-polymerase(PARP; Figure 6b) and by quantifying the activity ofcaspase-3 in whole cells (Figure 6c). All three indepen-dent assays consistently show that both spindlecheckpoint compromised cell lines efficiently escapefrom apoptosis upon treatment with nocodazole, that is,in response to the lack of microtubule attachment(Figure 6, left panels). Significantly and in contrast to

Figure 6 Partial downregulation ofMAD2, but not ofMAD1, causes resistance to taxol. (a) HCT116 and derivative cells were treatedwith 150 nM nocodazole, 100 nM taxol or 70mM monastrol for 48 h and fragmented chromosomal DNA was isolated and resolved onagarose gels. (b) Cells were treated as in (a) and the uncleaved and the apoptosis-specific cleaved forms of PARP were detected onWestern blots. The star indicates a crossreacting band. (c) Cells were treated with nocodazole, taxol or monastrol for 40 h and theapoptotic activity of caspase-3 was quantitatively determined using a fluorogenic peptide substrate. The graphs show mean values andstandard deviations from three independent experiments

MAD1 downregulation causes aneuploidyA Kienitz et al

4306

Oncogene

MAD2þ /� cells, spindle checkpoint compromised cellswith reduced Mad1 levels are sensitive to a treatmentwith drugs that inhibit the generation of tension, that istaxol or monastrol (Figure 6, middle and right panels)indicating a Mad1 independent induction of apoptosis.Further, we found that HCT116-p53�/� escape fromapoptosis after nocodazole, taxol or monastrol treat-ments indicating a general requirement of p53 for theinduction of apoptosis in response to spindle damage.Thus, our findings clearly demonstrate that the status ofthe spindle checkpoint per se does not determine thesensitivity towards spindle targeting chemotherapeuticdrugs, but rather the status of individual spindlecheckpoint components. In fact, our results suggest aMad1 independent function of Mad2 to be required forthe taxol induced apoptosis and a functional bifurcationfor the induction of apoptosis depending on the natureof spindle damage.

Discussion

Our data presented here establish the following mainpoints. (i) A stable partial downregulation of eitherMAD1 or MAD2, similar to the situation observed inhuman cancer cells, is sufficient to cause a functionalinactivation of the mitotic spindle assembly checkpointin response to the lack of microtubule attachment orupon the lack of tension across the kinetochores. Mostimportantly, this phenotype is associated with thegeneration of gross aneuploidy. (ii) Normal expressionof MAD2, but not of MAD1, is required to preventpremature sister chromatid separation during check-point activation. (iii) Partial repression of MAD2, butnot of MAD1, causes an acceleration of normal mitotictiming. (iv) Partial downregulation of either MAD2or MAD1 confers resistance to spindle-damagingagents that disrupt the microtubule–kinetochore attach-ment. In contrast, reduced levels of Mad2, but not ofMad1, lead to resistance to chemotherapeutic drugsthat inhibit the generation of tension, for example,paclitaxel/taxol.

Partial downregulation of spindle checkpoint inhuman cancer

Obvious candidate genes that might be involved incausing CIN in cancer are the spindle checkpoint genes,but these appear to be rarely inactivated by mutation inhuman cancer (Cahill et al., 1998; Nomoto et al., 1999;Takahashi et al., 1999; Hernando et al., 2001; Masudaand Takahashi, 2002). However, partial downregulationof spindle checkpoint genes including MAD1, MAD2,BUB1 and BUBR1 has been observed in cancer cells (Liand Benezra, 1996; Shichiri et al., 2002; Wang et al.,2002). In fact, experimental partial repression of MAD2leads to spindle checkpoint inactivation (Wang et al.,2002) and causes aneuploidy (Michel et al., 2001; andthis study). Similarly, reduced expression of BUBR1 orBUB3 in heterozygous mice (Babu et al., 2003; Bakeret al., 2004; Dai et al., 2004), or of MAD1 in human

somatic cells (this study) is sufficient to cause spindlecheckpoint inactivation and aneuploidy indicating thatnormal expression of the spindle checkpoint genes isessential for mediating mitotic arrest in response tospindle damage and for maintaining chromosomalstability. It is important to note that severe down-regulation, disruption or homozygous mutational in-activation of spindle checkpoint genes appears to beincompatible with survival (Dobles et al., 2000; Kalitsiset al., 2000; Kops et al., 2004; Michel et al., 2004), whichmight explain why only partial downregulation ofspindle checkpoint genes is observed in human cancer.Interestingly, in MAD2þ /�, BUB3þ /� and BUBR1þ /�

mice, the spindle checkpoint inactivation is alsoassociated with enhanced tumor development (Michelet al., 2001; Babu et al., 2003; Dai et al., 2004),suggesting that aneuploidy can directly contribute totumorigenesis. However, the mechanisms of cancerspecific downregulation of spindle checkpoint geneexpression are currently unknown. Interestingly, inhuman colorectal carcinomas of advanced clinical state,a significant silencing of mRNA expression of BUB1and BUBR1 was observed, suggesting that epigeneticmechanisms like promotor methylation might accountfor reduced gene expression in cancer cells (Shichiriet al., 2002). Nevertheless, the expression level of spindlecheckpoint genes is of great diagnostic value to predictspindle checkpoint activity and CIN.

Different phenotypes associated with the downregulationof MAD1 or MAD2

A number of previous studies have suggested that Mad1acts as an activator of Mad2 at the checkpoint activatedkinetochore (Sironi et al., 2001; Luo et al., 2002; Martin-Lluesma et al., 2002). However, an elegant recent studyhas demonstrated a requirement of Mad2 and BubR1,but not of other checkpoint proteins for the normaltiming of mitosis (Meraldi et al., 2004). Further, thiswork proposed a cytosolic and kinetochore independentfunction of Mad2 and BubR1 during an undisturbedmitosis. Our data strongly support the idea of additionalkinetochore and Mad1 independent functions of Mad2,which might be separable from its checkpoint function.Since partial downregulation of MAD1 neither accel-erates mitosis nor significantly elevates the proportion ofpremature sister chromatid separation when comparedto cells with reduced levels of Mad2, we suggest a Mad1independent function of Mad2 not only for normalmitotic timing but also for preventing the prematureonset of anaphase. According to this model, a cytosolicpool of Mad2 and BubR1 might be involved inpreventing the APC/C-mediated proteolysis of securin,which is required for sister chromatid separation(Nasmyth et al., 2000). This idea is further supportedby similar observations that were made in BUB3þ /�

MEFs, where reduced expression of BUB3 leads to animpairment of the spindle checkpoint in response tospindle damage but not to an induction of prematuresister chromatid segregation (Babu et al., 2003). Inter-estingly, the model of separable functions of Mad1 and

MAD1 downregulation causes aneuploidyA Kienitz et al

4307

Oncogene

Mad2 might also explain the concomitant downregu-lation of MAD1 and MAD2 observed in ovariancarcinoma cells (Wang et al., 2002).

Crosstalk of the mitotic spindle checkpoint and apoptoticpathways in response to chemotherapeutic treatment

Chemotherapeutic drugs that target the mitotic spindle,including taxanes and vinca alkaloids, are frequentlyused for treatment of lymphomas, ovarian, breast andlung tumors (Jordan and Wilson, 2004). However, it isstill unclear how these drugs induce apoptosis and whatthe molecular mechanisms of the resistance are. Firstreports have indicated that rapid drug export via ABCtransporters, the status of bcl-2 or enhanced micro-tubule dynamics can confer resistance (Jordan andWilson, 2004). In addition, most recent reports haveprovided a strong correlation between the status of thespindle checkpoint and the susceptibility to spindledamaging agents (Kasai et al., 2002; Anand et al., 2003;Masuda et al., 2003; Sudo et al., 2004). We have nowestablished that the induction of apoptosis induced bydrugs that abrogate the microtubule attachment (e.g.nocodazole) requires normal levels of both, Mad1 andMad2. In contrast, partial downregulation of MAD1does not confer resistance to agents that inhibit themicrotubule dynamics and the generation of tension(e.g. taxol and monastrol), although it is clearlysufficient to cause spindle checkpoint inactivationsuggesting an uncoupling of mitotic arrest and inductionof apoptosis. Our results indicate that a Mad1 andpossibly kinetochore independent function of Mad2might also be involved in the induction of apoptosis inresponse to the lack of tension. However, the inductionof apoptosis in response to the lack of attachmentrequires Mad1 and Mad2. These findings suggest afunctional bifurcation of apoptotic pathways dependingon the nature of spindle damage. Significantly, afunctional bifurcation of the spindle checkpoint is alsoobserved for the mitotic arrest. The chromosomalpassenger proteins including Aurora B and survivinare required for a sustained arrest in response to the lackof tension, but not upon nocodazole treatment (Ditch-field et al., 2003; Hauf et al., 2003; Lens et al., 2003).Interestingly, survivin has been shown to regulatedirectly the induction of apoptosis (Li et al., 1999),which might reflect a possible link between the spindlecheckpoint and the activation of apoptotic pathways.It will be important to elucidate what the molecularmechanisms are that mediate the induction of apoptosisunder these different conditions.

Since it is now recognized that the commonly usedchemotherapeutic drugs like taxanes and vinca alkaloidsmediate their death-inducing effects through the sup-pression of the microtubule dynamics rather than bydepolymerization of the microtubules (Jordan andWilson, 2004), it seems likely that these drugs employthe Mad2 dependent, but Mad1 independent apoptoticpathway. Therefore, the status of Mad2 rather than ofMad1, and not the functionality of the spindlecheckpoint in cancer cells per se, appears to be the

important determinant to predict a resistance towardstaxol or related chemotherapeutic drugs.

Materials and methods

Cell culture

HCT116-wild type, HCT116-p53�/� and HCT116-MAD2þ /�

cells (Michel et al., 2001) were kindly provided by BertVogelstein (Baltimore, USA), Loren Michel and RobertBenezra (New York, USA), respectively. All HCT116 deriva-tive cells were grown in RPMI 1640 medium supplementedwith 10% fetal calf serum, 1% glutamine and 100 mg/mlstreptomycin and 100U/ml penicillin (all from Gibco, Nether-lands) at 371C under 95% water and 5% CO2 atmosphere.Where indicated cells were treated with 150 nM nocodazole(Sigma, Germany), 100 nM taxol (Sigma) or 70 mM monastrol(Sigma).

Generation of stable HCT116 MAD1 knockdown cell lines

A double-stranded oligonucleotide corresponding to nt992–1014of the human MAD1 cDNA (50-AAGACCTTTCCAGATTCGTGGTT-30) was cloned into the BglII–HindIII site ofpSuper (Brummelkamp et al., 2002). HCT116 cells werecotransfected with the pSuper construct and pBabe-puro usingSuperfect (Qiagen, Germany) following the manufacturersinstructions. At 2 days post-transfection stable clones wereselected in medium containing 5mg/ml puromycin (Sigma). Intotal, 31 stable clones were tested for activation of the spindleassembly checkpoint by measuring the mitotic index 16 h afterthe addition of nocodazole or taxol. Two clones, termed 2-2and 2-31 were selected and used for further experimentsthroughout this study.

Western blotting

Western blotting was performed as described (Vogel et al.,2004a) using the following antibodies: monoclonal anti-Mad1(a generous gift from Dr TJ Yen, Philadephia, USA), anti-actin (Sigma), anti-cyclin B (H-433, Santa Cruz, USA), anti-PARP (BD Pharmingen, USA). All secondary antibodies werefrom Jackson ImmunoResearch (USA). For quantification ofMad1 protein we performed Western blots using anti-Mad1antibodies and quantified the Mad1 bands using the ImageGauge software. Signal data were normalized for actin bandsand a mean value was calculated from four independentexperiments.

Flow cytometry

The determination of the DNA content and of the mitoticindex was performed as described previously (Vogel et al.,2004a).

Microscopy

Cells were fixed on polylysine-treated coverslips with 3%p-formaldehyde/PBS and DNA was stained with 1mg/mlHoechst to visualize interphase chromatin, mitotic condensedchromosomes, multinuclei or apoptotic nuclei. Photos weretaken using a digital CCD Camera (Hamamatsu, Japan).

Determination of mitotic timing

Asynchronously growing cells were grown on coverslips andtreated with 200 mM ALLN (Calbiochem, Germany) for up to

MAD1 downregulation causes aneuploidyA Kienitz et al

4308

Oncogene

3 h to block mitotic progression beyond metaphase. Cells werefixed and DNA was stained as described above.

Determination of apoptosis

Cells were treated with 70 mM monastrol, 100 nM taxol or150 nM nocodazole for up to 48 h. Apoptosis was determinedusing the following methods. Apoptotic activation of caspase 3was measured using Ac-DEVD-AMC as a fluorogenic caspase3 substrate (Gurtu et al., 1997). Briefly, whole cells wereharvested and incubated in a 96-well dish at 371C in a tissueculture incubator in the presence of 2.5 mM Ac-DEVD-AMC(BD PharMingen). Cleaved substrate was determined astriplicates using a Victor2 1420 multilabel counter (WallacOy, Finland) in 30-min intervals for up to 180min.To determine apoptotic DNA laddering chromosomal DNA

was isolated essentially as described (Herrmann et al., 1994)with minor modifications. Briefly, attached and floating cellswere harvested and extracted twice for 10min with lysis buffer(50mM Tris-HCl pH7.5, 20mM EDTA, 1% NP40, 1% SDS).The pooled supernatants were digested with 1mg/ml RNase Afor 2 h at 371C and with 1 mg/ml proteinase K for 4 h at 561C.The chromosomal DNA was precipitated and separated on a2% agarose gel and visualized with ethidium bromide.The apoptotic cleavage product of poly-(ADP-Ribose)-

Polymerase (PARP) was determined on Western blots as

described using anti-PARP antibodies (BD Pharmingen) andanti-actin antibodies (Sigma) were used as a control.

Chromosome spread analysis

Chromosome spreading was performed essentially as described(Hauf et al., 2003). Briefly, cells were treated with 150 nMnocodazole for 1, 2 or 6 h. Cells were then harvested by mitoticshake off and hypotonically swollen in 40% RPMI medium/60% H2O for 10min at 371C. Cells were fixed in Carnoysfixative solution (75% methanol/25% acetic acid) with severalchanges of the fixative. Cells were dropped onto cooled glassslides and dried at RT. Chromosomes were stained in 5%Giemsa (Sigma) for 10min, rinsed with water, air dried andmounted.

Acknowledgements

We thank Tim Yen for generously providing Mad1 antibodies,Bert Vogelstein, Loren Michel and Robert Benezra forproviding HCT116 and derivative cell lines and Rene Bernardsfor the pSuper vector. We are grateful to Irmgard Hofmannfor help with the generation of the knockdown cell lines and toHeike Krebber for helpful discussions and critically readingthe manuscript. This work was supported by grants from theDeutsche Forschungsgemeinschaft, the Deutsche Krebshilfeand the PE Kempkes Stiftung.

References

Anand S, Penrhyn-Lowe S and Venkitaraman AR. (2003).Cancer Cell, 3, 51–62.

Babu JR, Jeganathan KB, Baker DJ, Wu X, Kang-Decker Nand van Deursen JM. (2003). J. Cell Biol., 160, 341–353.

Baker DJ, Jeganathan KB, Cameron JD, Thompson M, JunejaS, Kopecka A, Kumar R, Jenkins RB, de Groen PC, Roche Pand van Deursen JM. (2004). Nat. Genet., 36, 744–749.

Bharadwaj R and Yu H. (2004). Oncogene, 23, 2016–2027.Brummelkamp TR, Bernards R and Agami R. (2002). Science,296, 550–553.

Cahill DP, Lengauer C, Yu J, Riggins GJ, Willson JK,Markowitz SD, Kinzler KW and Vogelstein B. (1998).Nature, 392, 300–303.

Campbell M and Gorbsky G. (1995). J. Cell Biol., 129,

1195–1204.Campbell MS, Chan GK and Yen TJ. (2001). J. Cell Sci., 114,953–963.

Canman JC, Salmon ED and Fang G. (2002). Cell Motil.Cytoskeleton, 52, 61–65.

Chen RH, Brady DM, Smith D, Murray AW and HardwickKG. (1999). Mol. Biol. Cell, 10, 2607–2618.

Dai W, Wang Q, Liu T, Swamy M, Fang Y, Xie S, MahmoodR, Yang YM, Xu M and Rao CV. (2004). Cancer Res., 64,440–445.

Ditchfield C, Johnson VL, Tighe A, Ellston R, Haworth C,Johnson T, Mortlock A, Keen N and Taylor SS. (2003).J. Cell Biol., 161, 267–280.

Dobles M, Liberal V, Scott ML, Benezra R and Sorger PK.(2000). Cell, 101, 635–645.

Duesberg P, Li R, Rasnick D, Rausch C, Willer A, KraemerA, Yerganian G and Hehlmann R. (2000). Cancer Genet.Cytogenet., 119, 83–93.

Fang G. (2002). Mol. Biol. Cell, 13, 755–766.Fang G, Yu H and Kirschner MW. (1998). Genes Dev., 12,1871–1883.

Gorbsky GJ, Chen RW and Murray AW. (1998). J. Cell Biol.,141, 1193–1205.

Gurtu V, Kain SR and Zhang G. (1997). Anal. Biochem., 251,98–102.

Hauf S, Cole RW, LaTerra S, Zimmer C, Schnapp G, WalterR, Heckel A, van Meel J, Rieder CL and Peters JM. (2003).J. Cell Biol., 161, 281–294.

Hernando E, Orlow I, Liberal V, Nohales G, Benezra R andCordon-Cardo C. (2001). Int. J. Cancer, 95, 223–227.

Herrmann M, Lorenz HM, Voll R, Grunke M, Woith W andKalden JR. (1994). Nucleic Acids Res., 22, 5506–5507.

Jin DY, Spencer F and Jeang KT. (1998). Cell, 93, 81–91.Jordan MA and Wilson L. (2004). Nat. Rev. Cancer, 4,

253–265.Kalitsis P, Earle E, Fowler KJ and Choo KH. (2000). Genes

Dev., 14, 2277–2282.Kasai T, Iwanaga Y, Iha H and Jeang KT. (2002). J. Biol.

Chem., 277, 5187–5193.Kops GJ, Foltz DR and Cleveland DW. (2004). Proc. Natl.

Acad. Sci. USA, 101, 8699–8704.Lengauer C, Kinzler KW and Vogelstein B. (1998). Nature,396, 643–649.

Lens SM, Wolthuis RM, Klompmaker R, Kauw J, Agami R,Brummelkamp T, Kops G and Medema RH. (2003). EMBOJ., 22, 2934–2947.

Li F, Ackermann EJ, Bennett CF, Rothermel AL, Plescia J,Tognin S, Villa A, Marchisio PC and Altieri DC. (1999).Nat. Cell Biol., 1, 461–466.

Li X and Nicklas RB. (1995). Nature, 373, 630–632.Li Y and Benezra R. (1996). Science, 274, 246–248.Li Y, Gorbea C, Mahaffey D, Rechsteiner M and Benezra R.(1997). Proc. Natl. Acad. Sci. USA, 94, 12431–12436.

Luo X, Tang Z, Rizo J and Yu H. (2002). Mol. Cell, 9,59–71.

Martin-Lluesma S, Stucke VM and Nigg EA. (2002). Science,297, 2267–2270.

Masuda A, Maeno K, Nakagawa T, Saito H and Takahashi T.(2003). Am. J. Pathol., 163, 1109–1116.

Masuda A and Takahashi T. (2002). Oncogene, 21, 6884–6897.

MAD1 downregulation causes aneuploidyA Kienitz et al

4309

Oncogene

Meraldi P, Draviam VM and Sorger PK. (2004). Dev. Cell, 7,45–60.

Michel L, Diaz-Rodriguez E, Narayan G, Hernando E, MurtyVV and Benezra R. (2004). Proc. Natl. Acad. Sci. USA, 101,4459–4464.

Michel LS, Liberal V, Chatterjee A, Kirchwegger R, Pasche B,Gerald W, Dobles M, Sorger PK, Murty VV and Benezra R.(2001). Nature, 409, 355–359.

Musacchio A and Hardwick KG. (2002). Nat. Rev. Mol. CellBiol., 3, 731–741.

Nasmyth K, Peters JM and Uhlmann F. (2000). Science, 288,1379–1385.

Nomoto S, Haruki N, Takahashi T, Masuda A, Koshikawa T,Fujii Y and Osada H. (1999). Oncogene, 18, 7180–7183.

Peters JM. (2002). Mol. Cell, 9, 931–943.Rajagopalan H and Lengauer C. (2004). Nature, 432,338–341.

Rieder C, Cole R, Khodjakov A and Sluder G. (1995). J. CellBiol., 130, 941–948.

Shah JV and Cleveland DW. (2000). Cell, 103, 997–1000.Shannon KB, Canman JC and Salmon ED. (2002). Mol. Biol.

Cell, 13, 3706–3719.Shichiri M, Yoshinaga K, Hisatomi H, Sugihara K andHirata Y. (2002). Cancer Res., 62, 13–17.

Sieber OM, Heinimann K and Tomlinson IP. (2003). Nat. Rev.Cancer, 3, 701–708.

Sironi L, Melixetian M, Faretta M, Prosperini E, Helin K andMusacchio A. (2001). EMBO J., 20, 6371–6382.

Sudakin V, Chan GK and Yen TJ. (2001). J. Cell Biol., 154,925–936.

Sudo T, Nitta M, Saya H and Ueno NT. (2004). Cancer Res.,64, 2502–2508.

Takahashi T, Haruki N, Nomoto S, Masuda A, Saji S,Osada H and Takahashi T. (1999). Oncogene, 18, 4295–4300.

Tang Z, Bharadwaj R, Li B and Yu H. (2001). Dev. Cell, 1,227–237.

Taylor S and McKeon F. (1997). Cell, 89, 727–735.Taylor SS, Scott MI and Holland AJ. (2004). Chromosome

Res., 12, 599–616.Vigneron S, Prieto S, Bernis C, Labbe JC, Castro A andLorca T. (2004). Mol. Biol. Cell, 15, 4584–4596.

Vogel C, Kienitz A, Hofmann I, Muller R and Bastians H.(2004a). Oncogene, 23, 6845–6853.

Vogel C, Kienitz A, Muller R and Bastians H. (2004b). J. Biol.Chem., 280, 4025–4028.

Wang X, Jin DY, Ng RW, Feng H, Wong YC, Cheung ALand Tsao SW. (2002). Cancer Res., 62, 1662–1668.

Yu H. (2002). Curr. Opin. Cell Biol., 14, 706–714.

MAD1 downregulation causes aneuploidyA Kienitz et al

4310

Oncogene