Inhibition of proliferation and apoptosis by the transcriptional repressor Mad1. Repression of Fas-induced caspase-8 activation.

Mad1 is a member of the Myc/Max/Mad network of transcriptional regulators that play a central role in the control of cellular behavior. Mad proteins are thought to antagonize Myc functions at least in part by repressing gene transcription. To systematically examine the function of Mad1 in growth control and during apoptosis, we have generated U2OS cell clones that express Mad1 under a tetracyline-regulatable promoter (UTA-Mad1). Mad1 was induced rapidly and efficiently, localized to the nucleus, and bound to DNA as a heterodimer with Max. The induction of Mad1 reduced cellular growth and, more profoundly, inhibited colony formation of UTA-Mad1 cells. Conditioned medium neutralized this inhibitory effect implying that Mad1 function is regulated by extracellular signals. In addition Mad1 interfered with Fas-, TRAIL-, and UV-induced apoptosis, which coincided with a reduced activation of caspase-8 during Fas-mediated apoptosis in response to Mad1 expression. Furthermore, microinjection of Mad1-expressing plasmids into fibroblasts inhibited apoptosis induced by the oncoproteins c-Myc and E1A. Thus, Mad1 not only interferes with cellular proliferation but also with apoptosis, which defines a novel aspect of Mad1 function.

transformation of rat embryo fibroblasts and upon transfection into tumorigenic cell lines interfere with their ability to form tumors in susceptible mice (7,8,40,43,44,46,47). A role in tumor formation is also supported by the observation that mxi1 Ϫ/Ϫ mice show hyperplastic alterations in different tissues and reveal an increased susceptibility to a carcinogen (48). Recent findings indicate that Mad1 may modulate apoptosis of cells of the granulocytic lineage (49). As far as tested, all the biological effects ascribed to Mad1 are dependent on the bHL-HZip and the mSin3-interaction domain (SID), implying that these effects are the result of Mad1-specific gene regulation. In summary, these findings support the view that Mad proteins are negative regulators of cell growth and thereby oppose Myc function.
To study the role of Mad1 in cell growth control in more detail, we have generated cell lines that express mad1 under the control of a tetracycline-regulatable activator. We find that Mad1 reduces the growth rate of exponentially growing cells and inhibits efficiently the outgrowth of colonies, which can be overcome with conditioned medium. Furthermore, Mad1 interferes with apoptosis induced by several different stimuli. This appears to be mediated by a inhibition of the activation of caspase-8. Thus, Mad1 has broad biological activities affecting different fundamental processes involved in the control of cellular behavior.

MATERIALS AND METHODS
Cells, Plasmids, Transfections, and Microinjections-U2OS, UTA, UTA-Mad1, SAOS-2, and COS-7 cells were maintained in Dulbecco's modified Eagle's medium and 3T3-L1 cells in modified Eagle's medium and Jurkat T cells in RPMI 1640 medium supplemented with 100 units penicillin, 100 g/ml streptomycin, and 10% fetal calf serum (FCS). Tetracycline was added to 1000 ng/ml if not indicated otherwise. For proliferation assays cells were seeded in 96-well plates (1000 or 5000 cells/well) and the relative number of cells was measured using the crystal violet staining protocol as described (50). Control experiments revealed a linear relationship between the number of cells (1 ϫ 10 3 to 1 ϫ 10 6 ) analyzed and the measured absorbance at 590 nm (data not shown). For [ 3 H]thymidine incorporation assays, the cells were incubated with 0.5 Ci of [ 3 H]thymidine/well for 4 h, harvested onto glass fiber filters, and the incorporated radioactivity determined by scintillation counting.
UTA cells, a U2OS cell clone expressing tet-R-VP16-TAD, were a kind gift of C. Englert (55). For the generation of stable cell lines, the plasmids ptet-mad1 and pBabe-puro were cotransfected into UTA cells using a calcium-phosphate protocol (56,57). Briefly, 5 ϫ 10 5 cells/10-cm dish were seeded 20 h prior to transfection. 48 h after transfection, cells were replated and selected in puromycin (1 g/ml). After 14 -20 days, single colonies were isolated and propagated to clonal cell lines (UTA-Mad1 clones). For colony formation assays, the selected cells were stained with Giemsa and counted. Selection and propagation of stable clones was performed in the presence of 1 g/ml tetracycline. For induction, cells were washed extensively in medium containing FCS and then cultivated in the absence of tetracycline or in the presence of the indicated amounts of tetracycline. For colony formation assays of UTA-Mad1 cells, between 200 and 1000 cells/10-cm dish were plated.
Microinjections were performed as described previously (26). In each experiment, 100 -150 cells were injected and identified by staining for GFP. In independent experiments, coexpression of the different proteins was verified (data not shown).
For the induction of apoptosis, UTA or UTA-Mad1 cells were seeded onto CELLocates (Eppendorf) and grown in the presence or absence of tetracycline overnight. Apoptosis was induced with 100 ng/ml Fasspecific antibodies (Kamya CH-11) or 100 ng/ml Flag-TRAIL (a kind gift of M. Thome) (58) and 2 g/ml Flag-specific antibodies in the absence of serum. At different time points after induction of apoptosis, viable cells were counted in designated fields on the CELLocates.
Staining Procedures-UTA-Mad1 cells were seeded onto coverslips and cultivated with and without tetracycline for 42 h. Cells were then fixed in PBS containing 3.7% paraformaldehyde at 4°C for 22 min, washed three times in PBS, and permeabilized in PBS with 0.2% Triton X-100 at room temperature for 8 min. The TUNEL assay was performed as recommended by the manufacturer (Roche Molecular Biochemicals). The DNA was stained with Hoechst 33258 (1 g/ml in PBS) and the coverslips mounted with Moviol (Merck) in PBS containing 2.5% Npropylgallate (Sigma).
Electrophoretic Mobility Shift Assay (EMSA)-UTA and UTA-Mad1 cells were harvested in 300 l of F-buffer, and EMSAs were performed using as probe a 32 P-end-labeled CMD oligonucleotide (5Ј-TCA GAC CAC GTG GTC GGG) as described previously (10).
Northern Blotting, RNA Protection Assays, and cDNA Arrays-RNA was extracted with the RNeasy total RNA kit (Qiagen). 15 g of RNA was separated on 1% formaldehyde agarose gels. The RNA was blotted onto GeneScreen TM membrane and hybridized with 32 P-random primelabeled probes in 0.25 M NaP i , 7% SDS, 1 mM EDTA at 65°C. The membranes were washed with 50 mM NaP i , 0.5% SDS. As probe the NotI/EcoRI fragment from pVZ1-hu-mad1 was used.
For RNA protection assays, mRNA was prepared using the Oligotex direct mRNA kit (Quiagen). The expression of apoptosis relevant genes was measured using the multi-probe RNase protection assay systems (h-APO-1c and h-APO-2c; PharMingen) as suggested by the manufacturer.
The expression of a set of genes relevant for proliferation and apoptosis was analyzed by using RiboScreen TM human-1 membranes (PharMingen).
Kinase Assays-For kinase assays, 5 ϫ 10 5 cells from UTA-Mad1 cells were plated onto 10-cm dishes. The next day, the cells were washed and treated with or without tetracycline for the indicated times and harvested in 400 l of F-buffer. Portions of the lysates containing equal amounts of proteins (50 g; as determined with the Bradford protein assay) were incubated with 1 g of cyclin E-or CDK4-specific antibodies and protein A-Sepharose beads for 2 h at 4°C. For controls COS-7 cells were transfected with plasmids expressing cyclin E/CDK2, cyclinD/ CDK4, or empty vectors and lysates were prepared in F-buffer. The immunoprecipitates were washed twice in F-buffer and once in kinase buffer (50 mM Tris, pH 7.4, 10 mM MgCl 2 , 1 mM dithiothreitol). Kinase assays were performed in 25 l of kinase buffer containing 5 Ci of [␥-32 P]ATP (3000 Ci/mmol) and 100 M ATP for 15 min at 30°C. For cyclin E/CDK2 and cyclin D/CDK4 assays, 5 g of histone H1 and 2 g of GST-RbC, respectively, were used as substrates. The reactions were stopped by adding SDS-sample buffer, the proteins were separated on 15% SDS gels, and the dried gels analyzed by autoradiography.
Caspase Assays-For caspase assays, 2 ϫ 10 6 cells were treated with or without Fas-specific antibodies in the presence or absence of tetracycline for 48 h. Lysates were prepared in 800 l of F-buffer supplemented with 100 mM NaCl and 10 mM dithiothreitol. Caspase 8 and caspase 3 activities were measured using 17 mM Ac-IETD-AFC (Phar-Mingen) and 19 mM Z-DEVD-AMC (Alexis), respectively, as fluorogenic substrates. The reactions were performed in a spectrofluorometer with an excitation wave length of 400 nm and an emission wave length of 505 nm over 30 min at 37°C. During this time window, the reaction was linear. The slopes of the reactions were calculated, and the values from parallel control samples were subtracted.

Establishment of Inducible Mad1
Cell Lines-The analysis of Mad1 function using transient expression studies revealed a role in cell growth inhibition (43)(44)(45). To further define this effect, we established UTA-Mad1 cell lines that express Mad1 under the control of a tetracycline-regulatable activator (tet-off; Ref. 51). We used founder cell lines that express the tet repressor-VP16-transactivation domain fusion protein (tet-R-VP16-TAD) and tested whether in these cell lines a Mad1-induced effect on colony formation could be observed. The osteosarcoma cell lines UTA and SAOS-2 and the fibroblastoid cell line 3T3-L1 stably expressing the tet-R-VP16-TAD were cotransfected with a plasmid coding for Mad1 under the control of a tet-R-responsive element/minimal CMV promoter and a puromycin resistance plasmid. The transfected cells were then seeded in the presence or absence of tetracycline, selected in puromycin, and the number of colonies analyzed 20 days later. Whereas the UTA cells showed a specific repression of the number of colonies in response to Mad1 expression, the corresponding SAOS-2 and 3T3-L1 cells were already fully inhibited in the presence of tetracycline ( Fig. 1 and data not shown). In addition, we tried to generate SAOS-2-Mad1 and 3T3-L1-Mad1, -Mad3, and -Mad4 cells. However, all the puromycinresistant clones that were analyzed did not express either of the Mad proteins (data not shown). We concluded that low amounts of Mad1, as obtained in the off status of the tetracycline-regulatable system, were sufficient to inhibit growth in SAOS-2 and in 3T3-L1 cells. Therefore, we concentrated on the analysis of UTA-Mad1 cells, which are described below.
Expression and Characterization of Mad1 in UTA-Mad1 Cells-We selected several UTA-Mad1 clones, which behaved similarly in all the assays performed. In the following experiments, we have used three representative clones (M19, M37, and M47). The induction of mad1 mRNA and of Mad1 protein was rapid, in the range of 25-30-fold, and reversible in all three clones ( Fig. 2 and data not shown). In M47 cells, a 20-fold induction of mad1 mRNA was observed after removing tetracycline by 4 h and maximal induction of 25-fold by 8 h (Fig. 2,  A and B). Readdition of tetracycline resulted in a rapid decline of mad1 mRNA to basal levels within 4 h. In agreement with the RNA expression data, significant accumulation of Mad1 protein was observed after removing tetracycline by 4 h and thereafter (Fig. 2C). The induction of both mad1 mRNA and protein was dose-dependent with maximal expression at concentrations of tetracycline of 0.1 ng/ml or less (data not shown). These findings demonstrate that the expression of Mad1 is efficiently regulated in several UTA-Mad1 clones.
Mad1 proteins expressed in UTA-Mad1 cells were further characterized. Mad1-specific staining was exclusively nuclear upon removal of tetracycline (Fig. 3). Furthermore, we determined whether the induced Mad1 bound to endogenous Max and whether such heterodimers could bind to the CMD oligonucleotide, which contains a Myc consensus E box sequence.
Upon induction of Mad1 expression, a novel E box binding activity was induced in UTA-Mad1 cells but not in UTA cells ( Fig. 4 and data not shown). The mobility of this complex was FIG. 1. Inhibition of colony formation by Mad1. UTA cells, a U2OS cell clone stably expressing the tet-R-VP16-TAD activator protein, were transiently transfected with the tetracycline-inducible pUHD10 -3-mad1 (tet-mad1) expression vector or the control plasmid pUHD10 -3 and selected for puromycin resistance in absence or presence of tetracycline (tet). The average number of colonies per 10-cm dish of six dishes is summarized.
FIG. 2. Regulation of expression of mad1 mRNA and protein by tetracycline. UTA cells were stably transfected with pUHD10 -3-mad1 and individual clones selected (UTA-Mad1 clones). Cells of clone M47 were precultivated in the presence or absence of tetracycline (Ϯtet, respectively). Further incubations were performed either in the presence or absence of tetracycline, and at the times indicated samples were collected and processed for mRNA and protein analysis. A, the expression of mad1 RNA in response to tetracycline was analyzed by Northern blotting. B, quantification of the signals from the blot shown in A using a phosphorimager is shown. C, Western blot analysis of whole cell extracts using Mad1-specific mAbs 5F4 and 5C9. sensitive to both Max-and Mad1-specific antibodies, suggesting that it represents a Mad1/Max heterodimer (Fig. 4). Additionally, this complex comigrated with the Mad1/Max complex obtained from lysates of COS-7 cells transiently expressing Mad1/Max and was competed by CMD but not by a mutated oligonucleotide (data not shown; see Ref. 10). Thus, the induced Mad1 was nuclear and capable to form DNA binding-competent Mad1/Max heterodimers.
Mad1 Inhibits Cellular Proliferation-Next, we determined the effect of Mad1 on cellular growth. In the presence of tetracycline, clones M37 and M47 grew with similar kinetics as the parental UTA cells, whereas clone M19 grew slightly faster (Fig. 5). Upon removal of tetracycline, a decrease in growth rate, but no complete inhibition, was evident from a slower increase in cell number and a decrease in the amount of [ 3 H]thymidine incorporated in UTA-Mad1 but not in control cells (Fig. 5 and data not shown). Consistent with this was a modest reduction in the number of S phase cells upon Mad1 induction (from 32% to 26% in M47), as measured by fluorescence-activated cell sorting analysis. The Mad1 effect on cell growth was reversible since proliferation resumed to control levels upon readdition of tetracycline to UTA-Mad1 cells grown in the absence of tetracycline (data not shown). Together, our data suggest that the induced expression of Mad1 in an exponentially growing population of cells inhibits but does not abolish cellular growth.
To test Mad1 function under more stringent growth conditions, we performed colony formation assays. In the presence of tetracycline, 35-45% of plated cells formed colonies that were larger than 1 mm in diameter within 2 weeks (Fig. 6). Upon induction of Mad1 after plating, a significant reduction in the number of colonies was observed for all UTA-Mad1 clones (Fig.  6). The number of colonies formed by UTA cells was not affected by tetracycline (data not shown). Since the Mad1-expressing cells grew more slowly (Fig. 5), they were cultivated in the absence of tetracycline for an additional 2 weeks. Only a few additional colonies appeared, suggesting that Mad1 blocked rather than delayed the outgrowth of colonies (data not shown). The effect of Mad1 on colony formation was dose-dependent. At 1 and 10 ng/ml tetracycline, intermediate levels of colonies were observed compared with the number of colonies obtained either in the absence of tetracycline or in the presence of 100 ng/ml or more tetracycline, consistent with a partial induction of Mad1 at intermediate tetracycline concentrations (Fig. 6). Thus, the expression of Mad1 severely compromises the formation of colonies from individual cells.
The findings described above suggest that Mad1 function becomes dominant under restrictive cell growth conditions in UTA-Mad1 cells. We were interested to determine whether the Mad1 effect could be reversed. The addition of conditioned medium of UTA cells reestablished the efficient formation of UTA-Mad1 colonies (Fig. 6B). No effect was observed when the growth medium was supplemented with additional 10% FCS, suggesting that the Mad1 effect cannot be overcome by providing more serum-derived growth factors (data not shown). Since c-Myc has been shown to regulate the CAD gene (16), Mad1 may affect growth by inhibiting this gene (see also below) and consequently block de novo biosynthesis of pyrimidines. Therefore, colony formation assays were performed in the presence of uridine (15 g/ml), which is converted to uridine-monophosphate and thus circumvents the need for CAD (61). No rescue was observed (data not shown), suggesting that pyrimidine FIG. 4. DNA binding of Mad1/Max complexes. Whole cell extracts were prepared from M47 or UTA cells cultured in the presence or absence of tetracycline (tet) and used in an EMSA with a labeled CMD oligonucleotide containing a Myc-specific E box. Supershifts were performed with antibodies recognizing Mad1 (SC-C- 19) or Max (SC-C-17) as indicated. DNA-protein complexes were separated on an non-denaturing gel. The free probe was run off the gel for better separation of the different CMD-binding complexes. biosynthesis is not limiting. Together, these findings indicate that UTA cells produce a factor(s), which, when present in sufficient quantity, block the Mad1 effect on colony formation.
Analysis of Potential Mad1 Target Genes-Presently, no Mad1 target genes are known. However, it is possible that Myc-regulated genes are targeted by Mad1. Therefore, we analyzed the expression of genes that have been shown previously to be regulated by c-Myc (for review, see Refs. 2 and 62) upon induction of Mad1 in exponentially growing cells. No change in the mRNA expression levels of ␣-prothymosine, ODC, CAD, or cdc25A could be measured (data not shown). In addition the mRNA expression of cyclin E and p21, which are indirect c-Myc targets, was not affected by Mad1 (data not shown). Furthermore, we measured the expression of different proteins that have been linked to c-Myc-dependent cell cycle regulation, including p27 KIP1 , p21 WAF1 , p53, cyclin A, and cyclin E (for review, see Ref. 33). No effect on the expression of these proteins could be detected in response to Mad1 induction ( Fig. 7A and data not shown). Finally we determined whether Mad1 affected the activities of cyclin E/CDK2 and cyclin D/CDK4 complexes. Again, no difference in the respective kinase activities could be observed ( Fig. 7B and data not shown). Thus, the effect of Mad1 on cellular growth is not correlated to the regulation of known Myc target genes as far as tested.
Mad1 Inhibits Apoptosis-Since Myc proteins can positively affect apoptotic processes, we were interested to determine whether Mad1 would have opposite effects. UTA and UTA-Mad1 cells were treated with an apoptosis-inducing mAb specific for Fas or with TRAIL (58). Both treatments resulted in extensive cell death with characteristic signs of apoptosis, that is rounding up of cells, membrane blebbing, and DNA fragmentation as visualized by TUNEL staining (Fig. 8A). In the presence of Mad1, both Fas-and TRAIL-induced apoptosis was significantly delayed (Fig. 8, B and C). In addition, Mad1 was also able to reduce apoptosis in response to UV treatment (data not shown).
To obtain inside into the mechanism of how Mad1 affects apoptosis, the expression of several relevant genes was analyzed. No difference in mRNA levels of any of these genes, including Fas, TRAIL receptors, caspase-8, and caspase-3 could be detected (data not shown; see "Materials and Methods" for details). Therefore, we addressed the activation of caspases in UTA-Mad1 cells. Treatment with Fas-specific antibodies resulted in the activation of caspase-8 (Fig. 9). This is demonstrated by the appearance of specific cleavage products in M19 cells that comigrate with fragments from apoptotic Jurkat cells (63). However, in the presence of Mad1, substantially less of the active site-containing fragment p18 (64) was generated (Fig. 9A). The reduced activation of caspase-8 was confirmed by measuring caspase-8 activity using a specific fluorogenic substrate Ac-IETD-AFC (Fig. 9B). The activity of caspase-3, which FIG. 7. Effects of Mad1 on cell cycle regulators. A, whole cell extracts of M47 cells cultured in the presence or absence of tetracyline were prepared at the time points indicated and the expression of p27 KIP1 , p53, and cyclinA analyzed by Western blotting. The blot was probed consecutively with Mad1-specific polyclonal antibodies (SC-C-19), p27 KIP1 -specific monoclonal antibodies (K25020), p53-specific mAbs (DO-1), and cyclin A-specific polyclonal antibodies (C24230) and developed using the ECL system. B, F-buffer extracts of M47 cells grown in the presence or absence of tetracycline (Ϯtet, respectively) for the times indicated were prepared and cyclin E-containing complexes immunoprecipitated. Cyclin E-associated kinase activity was measured by using histone H1 as substrate. As controls the kinase activity of an immunoprecipitate with irrelevant antibodies (C) or of cyclin E-specific and control immunoprecipitates from lysates of COS-7 cells transfected with cyclin E and CDK2 expression plasmids were performed as indicated.
FIG. 6. Inhibition of colony formation by Mad1. A, 1000 cells of clone M47 were seeded onto 10-cm tissue culture dishes and cultured for 14 days in the presence of the indicated amounts of tetracycline. The outgrowing colonies were fixed and stained using Giemsa. B, M19 and M37 cells (200 cells/6-cm dish) were grown in the presence of the indicated amounts of tetracycline for 14 days (control). To parallel cultures conditioned medium obtained from confluent plates of UTA cells was added. The colonies were fixed and stained as in A.
is activated downstream of caspase-8, was also reduced in response to Mad1 (Fig. 9B). Together, these findings suggest that Mad1 interferes with the activation of caspase-8, thereby inhibiting apoptosis.
The inhibitory role of Mad1 in apoptosis was further corroborated using a microinjection approach. Serum-starved 3T3-L1 cells were injected with plasmids expressing either c-Myc or adenoviral E1A in combination with an expression plasmid for GFP to visualize the injected cells. Whereas roughly 60% of control cells were detected 20 h after microinjection, a reduction to about 22% and 12% was evident upon expression of c-Myc and E1A, respectively (Fig. 10). This is consistent with the apoptosis-promoting function of these two oncoproteins (for review, see Ref. 65). Indeed cells coexpressing GFP and c-Myc or E1A stained positive with the TUNEL assay (data not shown). Coexpression of Mad1 was sufficient to rescue most of the cells from apoptosis (Fig. 10). Similarly, the addition of serum also inhibited c-Myc-or E1A-induced apoptosis (data not shown). The SID and bHLHZip domain are important for Mad1 to inhibit cellular proliferation (43,45). Coexpression of Mad1⌬N, which lacks the SID, did not inhibit c-Myc induced apoptosis (Fig. 10). Unexpectedly, however, the expression of FIG. 8. Mad1 inhibits Fas-and TRAIL-induced apoptosis. M19 cells were seeded onto CELLocates, grown overnight in medium containing 10% FCS in the presence or absence of 1 g/ml tetracycline, and then treated with 100 ng/ml anti-Fas or 100 ng/ml Flag-TRAIL and 2 g/ml Flag-specific antibodies in medium without FCS as indicated. A, 20 h after stimulation with anti-Fas or TRAIL cells were fixed, apoptotic cells were stained by TUNEL, and DNA was counterstained with Hoechst 33258 as indicated. In addition a phase contrast picture of the same area is depicted. B, the number of morphologically normal cells present within a defined area on CELLocates was determined at the time points indicated after induction of apoptosis by anti-Fas or TRAIL. The mean and standard deviations of three experiments are shown. Mad1⌬N in the absence of serum was sufficient to induce apoptosis (Fig. 10), suggesting a dominant negative effect of this mutant on a gene(s) that modulates apoptosis. In contrast to Mad1⌬N, Mad1⌬Zip had no effect on the survival of injected cells, neither in the presence nor in the absence of c-Myc (data not shown). Together, our findings demonstrate that Mad1 has broad anti-apoptotic activities. DISCUSSION We have developed U2OS osteosarcoma cell clones that express the human mad1 cDNA under the control of a tetracycline-regulatable transactivator (51). Using these clones, we have analyzed the effects of induced Mad1 expression on cellular proliferation and apoptosis. Our findings demonstrate that functional Mad1 can be expressed (Figs. [2][3][4] and that this protein inhibits proliferation (Figs. 5 and 6). Furthermore, Mad1 also interferes with apoptosis induced by different stimuli (Fig. 8). This latter finding was corroborated using a microinjection approach demonstrating that Mad1 also inhibits oncogene-induced apoptosis (Fig. 10). The anti-apoptotic effect during Fas-induced apoptosis was the result of a reduced activation of caspase-8 ( Fig. 9). Together, this is further evidence of the pivotal role of Myc/Max/Mad network in cellular growth control and extends the Myc-Mad antagonism to apoptosis.
In the process of identifying a cell culture system that allows the analysis of Mad proteins in cell growth control, we tested three different cell lines for tetracycline-inducible effects. Only clones derived from UTA cells, a U2OS derivative stably expressing the tet-R-VP16-TAD fusion protein (55), showed the desired phenotype, i.e. in the presence of tetracycline no effect on colony formation was measured whereas in the absence of tetracycline the number of colonies was reduced (Fig. 1). At present it is unclear why, in SAOS-2 and 3T3-L1 cells expressing tet-R-VP16-TAD, an inhibition of colony formation was seen even in the presence of tetracycline. One possibility is that the basal expression of the transgene that is controlled by a minimal CMV promoter with a tet-R-responsive element is higher in these cells. However, little difference was seen in basal expression of the tet-R-responsive element-min-CMVluciferase reporter in comparison to a control reporter, thus arguing against large differences in basal Mad1 expression (data not shown). Another possibility is that SAOS-2 and 3T3-L1 cells are inherently more sensitive to Mad1 function. In this respect, two other studies are of interest in which stable cell lines with inducible Mad1 expression have been generated. The induction of Mad1 in a human glioblastoma cell line resulted in very little effect (66). In contrast, in mouse erythroleukemia cells, Mad1 expression could only be obtained when c-Myc was co-expressed, suggesting that mouse erythroleukemia cells are also sensitive to low amounts of Mad1 (67). Together, these studies indicate significant differences in the sensitivity of different cell lines to Mad1 expression. We believe that the specific molecular background allowed us to generate UTA-Mad1 cells but not corresponding SAOS-2 or 3T3-L1 clones. However, the molecular characteristics determining these differences are presently not known.
An unexpected finding was that the effect of Mad1 was most prominent when measured in a colony formation assay, whereas the growth of an exponential culture was only reduced 2-fold (Figs. 6 and 7). One explanation for this difference might be that Mad1 exerts its function predominantly under restrictive cell growth conditions. Such an interpretation is consistent with studies published previously on the interrelationship of Mad1 and cell growth. The transient expression of Mad1 in a human astrocytoma cell line and in murine NIH 3T3 fibroblasts using either adenovirus-or retrovirus-based expression systems, respectively, resulted in a decrease in the number of S phase cells (44,45). However, there were still substantial numbers of cells in S and G 2 /M. Similarly, microinjection of exponentially growing 3T3-L1 cells with a Mad1-expressing plasmid resulted only in a 2-fold reduction in the number of S phase cells. 2 In contrast to these rather modest effects, Mad1 drastically reduced the ability of astrocytoma cells to form tumors in nude mice and colonies in soft agar (44), of NIH 3T3 cells to grow in soft agar (45), and of quiescent 3T3-L1 cells to enter S phase upon serum stimulation (26). Thus, Mad1 is most effective under restrictive growth conditions, while in exponentially growing cells only modest effects on growth are seen. It is unclear whether a common mechanism exists that explains the Mad1 effects under the different restrictive growth conditions discussed above or whether Mad1 influences multiple pathways.
One possibility is that cells which express Mad1 have an increased requirement for specific growth factors. This could be relevant for tumor growth and for soft agar colony formation and would be consistent with our finding that conditioned medium can overcome the Mad1-dependent inhibition of colony formation. The expression of such a factor(s) in UTA cells, which can overcome the Mad1 effect on colony formation, may have been the prerequisite to establish UTA-Mad1 cells in the first place.
Mad1 not only inhibits cellular proliferation, it also inhibits apoptotic cell death induced by various stimuli (Figs. 8 and 10). Stimulation of Fas or the TRAIL-receptor as well as DNA damage by UV treatment resulted in apoptosis, which was inhibited in UTA-Mad1 cells upon induction of Mad1 expression. In addition, oncoprotein-induced apoptosis in serumstarved fibroblasts was inhibited by coexpression of Mad1. These findings, together with the recent observation that bone marrow cells derived from mad1 Ϫ/Ϫ mice are more sensitive to cytokine withdrawal (49), suggest that Mad1 interferes with multiple apoptotic stimuli. The interaction with Max and the recruitment of the repressor complex are important for the ability of Mad1 to inhibit apoptosis, since deletion of either the SID or the Zip domain prevented rescue of c-Myc-expressing cells from apoptosis (Fig. 10). Thus, it seems plausible that Mad1 affects apoptosis by regulating gene expression. The finding that Mad1⌬N is sufficient to induce apoptosis in serumstarved cells suggests that the failure to repress an as yet unidentified gene(s) is responsible for the dominant negative phenotype of this mutant. In this respect, it is important to note that Mad1-mediated repression is dominant over several transcriptional activators (22,23). In addition, Mad1⌬N is a weak transactivator on a synthetic reporter gene construct. 3 Together, these findings may explain the dominant negative effect of Mad1⌬N on apoptosis.
A central question that remains unanswered is: through which target genes does Mad1 mediate its effects on cellular behavior? The findings regarding the functional antagonism between Mad and Myc proteins, the highly similar in vitro DNA binding specificity of c-Myc/Max and Mad/Max complexes, and the differential expression of these proteins in growing versus resting and differentiating cells have led to the suggestion that genes that are positively regulated by Myc might be repressed by Mad. However, while a number of Myc target genes have been identified (for review, see Refs. 2 and 62), it is not clear whether any of these genes is regulated by Mad proteins. Our analysis of potential Mad1 targets as defined by their c-Myc-dependent regulation did not reveal any insight into how Mad1 function is mediated. None of the pre-2 A. Menkel, J. Mertsching, and B. Lü scher, unpublished observation. 3 H. Burkhardt and B. Lü scher, unpublished observation. viously identified Myc targets (direct or indirect) analyzed here (see "Results") was significantly affected upon induction of Mad1 expression. We had expected to see differences since Mad1/Max heterodimers possess dominant repressing activity (22,23). Therefore, it remains open which target genes mediate the Mad1 effect on proliferation.
Some Myc target genes, including ODC, cdc25A, and p53, have also been implicated in the regulation of apoptosis (for review, see Refs. 62 and 65). However, the expression of none of these genes was affected by Mad1. Furthermore, we did not detect changes in the expression of several genes relevant for apoptosis, including caspase-8, as detailed above. However, the activation of caspase-8 was significantly inhibited by Mad1 (Fig. 9). Caspase-8 is recruited to Fas through the adaptor protein FADD in response to ligand binding (68). In addition, several proteins have been suggested to modify caspase-8 activation, including FLIP, TRAFs, and IAPs (69,70). Thus, Mad1 may affect the expression or activity of one or more of these regulators of the caspase-8 activation process.
In summary, we have been able to document strong effects of Mad1 on both proliferation and apoptosis. While it remains unresolved which target genes mediate the effects of Mad1 on proliferation, the inhibition of apoptosis by Mad1 correlates with a poor activation of caspase-8 by Fas. This identifies a downstream target of Mad1 function, and this information will be useful to identify the target genes of Mad1 relevant for modulating apoptosis. To identify Mad1 target genes remains, nevertheless, an unresolved issue. The functional analysis of the proteins encoded by these genes should shed light on how Mad1 affects cellular behavior.