Loss of MYC Confers Resistance to Doxorubicin-induced Apoptosis by

THE JOURNAL OF BIOLOGICAL CHEMISTRY © 2004 by The American Society for Biochemistry and Molecular Biology, Inc.

Vol. 279, No. 20, Issue of May 14, pp. 21318 –21326, 2004 Printed in U.S.A.

Loss of MYC Confers Resistance to Doxorubicin-induced Apoptosis by Preventing the Activation of Multiple Serine Protease- and Caspase-mediated Pathways* Received for publication, December 10, 2003, and in revised form, February 6, 2004 Published, JBC Papers in Press, February 27, 2004, DOI 10.1074/jbc.M313532200

Emanuela Grassilli‡, Andrea Ballabeni‡, Emilia Maellaro§, Barbara Del Bello§, and Kristian Helin‡¶储 From the ‡Department of Experimental Oncology, European Institute of Oncology, via Ripamonti 435, 20141 Milano, the §Department of Pathophysiology and Experimental Medicine, University of Siena, via A. Moro, 53100 Siena, Italy, and ¶Biotech Research and Innovation Centre, Fuglebjergvej 3, 2100 Copenhagen, Denmark

c-Myc plays an essential role in proliferation, differentiation, and apoptosis. Because of its relevance to cancer, most studies have focused on the cellular consequences of c-Myc overexpression. Here, we address the role of physiological levels of c-Myc in drug-induced apoptosis. By using c-MYC null cells we confirm and extend recent reports showing a c-Myc requirement for the induction of apoptosis by a number of anticancer agents. In particular, we show that c-Myc is required for the induction of apoptosis by doxorubicin and etoposide, whereas it is not required for camptothecin-induced cell death. We have investigated the molecular mechanisms involved in executing doxorubicin-induced apoptosis and show caspase-3 activation by both mitochondria-dependent and -independent pathways. Moreover, serine proteases participate in doxorubicin-induced apoptosis partly by contributing to caspase-3 activation. Finally, a complete rescue from doxorubicininduced apoptosis is obtained only when serine proteases, caspase-3, and mitochondrial activation are inhibited simultaneously. Interestingly, doxorubicin requires c-Myc for the activation of all of these pathways. Our findings therefore support a model in which doxorubicin simultaneously triggers multiple c-Myc-dependent apoptosis pathways.

The nuclear proto-oncogene c-MYC (hereafter called MYC) encodes a sequence-specific transcription factor, which plays an essential role in important cellular processes as different as proliferation, differentiation, apoptosis, and tumorigenesis (1). A role for Myc in regulating apoptosis was first demonstrated by studies in which overexpression of Myc under growth-limiting conditions resulted in cell death (2, 3). The dual capacity of Myc to stimulate cell proliferation in the presence of the appropriate survival factor(s) and to trigger apoptosis in their absence ensures that cell growth is strictly controlled by the microenvironment (3). Subsequent results have shown that Myc-induced apoptosis requires interaction at the cell surface between Fas and its ligand (FasL) and that Myc acts down* This work was supported by grants from the Associazione Italiana per la Ricerca sul Cancro (AIRC), the Fondazione Italiana per la Ricerca sul Cancro (FIRC), the European Union Fifth Framework Program, the Italian Health Ministry, and the Human Science Frontiers Science Program. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 储 To whom correspondence should be addressed. Tel.: 39-02-57489860; Fax: 39-02-5748-9851; E-mail: [email protected].

stream of the Fas receptor by sensitizing cells to the Fas death signal (4). Moreover, it has been shown that this sensitization step is represented by cytochrome c release (5). Myc has also been reported to sensitize cells to TNF␣-induced apoptosis (6). Finally, deregulation or overexpression of Myc enhances the apoptotic response to different inducers, such as cytokine withdrawal, glucose deprivation, hypoxia, DNA damage, and cancer therapeutics (7). Most data presented so far have dealt with the role of overexpressed Myc in inducing apoptosis. However, Myc also appears to have a role in regulating apoptosis when it is expressed at physiological levels. For instance, antisense oligonucleotides specifically blocking Myc expression resulted in a reduction of T-cell receptor activation-induced death (8). In addition, Sedivy and co-workers (9) recently reported that cells lacking Myc are resistant to cell death induced by etoposide. Moreover, they showed that the defect in DNA damage-initiated apoptosis in the G2 phase of the cell cycle was rescued by ectopic expression of Cyclin A, thus linking DNA damageinduced apoptosis and cell cycle progression (9). Somewhat contrasting results using the same cell line was reported by Soucie et al. (10). In fact, these investigators showed that in absence of Myc, cells could undergo cytochrome c-independent cell death when treated with etoposide and suggested that Myc regulates a mitochondrial amplification step (10). To clarify these contrasting results we decided to use the same cellular model to further investigate the role of Myc in drug-induced apoptosis. In this study we have confirmed that the absence of Myc impairs the apoptotic response to etoposide and to another topoisomerase II inhibitor, doxorubicin. Moreover, we have shown that the absence of Myc does not affect camptothecininduced apoptosis and that it has intermediary effects of the ability of taxol and staurosporine to induce apoptosis. Furthermore, we have investigated the molecular mechanisms involved in doxorubicin-induced apoptosis. Here, we report that together with the classical mitochondrial pathway, doxorubicin triggers two other pathways involved in executing cell death. Interestingly, we show that all these apoptotic pathways are impaired in Myc-deficient cells. Finally, we demonstrate that one of the pathways contributing to the cell death process induced by doxorubicin is represented by the activation of serine proteases, which occurs in parallel and upstream of caspase activation. EXPERIMENTAL PROCEDURES

Reagents—Etoposide, doxorubicin, taxol, camptothecin, staurosporine, tosyl lysine chloromethyl ketone, pepstatin A, CA-074 Me,

21318

This paper is available on line at http://www.jbc.org

MYC-dependent Activation of Serine Proteases and Caspases MDL28170, and E64d were purchased from Calbiochem Pharmaceuticals. The Fas ligand kit was from Upstate Biotechnology. MFL4 was from Pharmingen. Cell Culture and Stable Cell Lines—TGR-1 and HO15.19 Rat1 fibroblasts were a kind gift of J. Sedivy (Brown University, Providence, RI) and have been described previously (11). 7M3 is a representative clone of HO15.19 re-expressing Myc. TGR-1-BclXL is a representative clone of TGR-1 overexpressing BclXL. cDNAs encoding human MYC and BclXL were subcloned into the retroviral vectors pBabepuro and pBabeHygro respectively. XIAPW310A cDNA was subcloned from pEBBXIAPW310A (a kind gift from C. Duckett, University of Michigan, Ann Arbor, MI) into pBabepuro. Fas-associated death domain (FADD)1 dominant negative (DN), caspase-8 DN, and cowpox virus-crmA cDNA were subcloned from pcDNA3-FADD DN, pcDNA3-caspase-8 DN, and pcDNA3-crmA, (a kind gift of V. Dixit, Genentech Inc., San Francisco, CA) into the retroviral GFP-containing vector Pinco (12). All the derived clones described in this paper were obtained by retroviral infection and selection with the appropriate antibiotic as described previously (13). Briefly, Phoenix cells were plated at a density of 2 million cells/10-cm diameter dish, and 2 days later cells were transfected with 10 ␮g of DNA. Supernatants were collected after 1 and 2 days, filtered, and used to infect HO15.19 cells. The viral supernatant was left on the cells in the presence of Polybrene (8 mg/ml, Sigma-Aldrich) for 3 h, and the procedure was repeated twice for 2 days to increase the efficiency of infection. Two days after infection, infected cultures were split, and resistant cells were selected in medium supplemented with the appropriate antibiotic (5 ␮g/ml puromycin or 250 ␮g/ml hygromycin) and used as bulk populations or cloned by limiting dilution. Clones successfully re-expressing Myc were screened for reconstitution of the apoptotic phenotype upon doxorubicin treatment. Only clones achieving levels of apoptosis similar to parental TGR-1 and clones overexpressing BclXL levels comparable with those observed in HO15.19 were used. In case of cells infected with GFP-crmA vector, after the infection cells were FACS-sorted to isolate only GFP-positive cells. All cell lines were maintained in Dulbecco’s modified Eagles’ medium supplemented with 10% bovine calf serum. Assessment of Apoptosis—Subconfluent growing cells were seeded at a density of 0.3 ⫻ 106 in a 6-well plate and were treated with the indicated drugs 8 –12 h after plating. At different time points after apoptosis induction, floating cells were pooled with those harvested by trypsinization, and the cells were fixed by gently resuspending the solution in 70% ethanol in phosphate-buffered saline, and cells were kept at 4 °C until further processing. Fixed cells were washed once in phosphate-buffered saline, 0.1% bovine serum albumin before staining for 1 h at room temperature with a solution of 5 mg/ml propidium iodide containing 2.5 mg/ml RNase. Cytofluorimetric analysis was performed by FACStar using the CellQuest software (BD Biosciences). Each experiment was repeated 3–5 times, and standard deviation values were calculated. Colony Assay—Subconfluent growing cells were seeded at a density of 0.3 ⫻ 106 in a 6-well plate and were treated with 0.01 mg/ml doxorubicin 12 h after plating. Twenty-four hours after doxorubicin addition cells were trypsinized, counted, and reseeded at a low density (10,000 cells in a 10-cm dish) in duplicate. Medium was replaced every 3 days, and after 2 weeks colonies were fixed and stained in 1% crystal violet, 35% ethanol. Immunoblotting—Cells were lysed by three cycles of freeze/thawing in CHAPS buffer (50 mM Hepes, pH 7, 0.1% CHAPS, 2 mM EDTA, pH 8.0, aprotinin 10 mg/ml, leupeptin 10 mg/ml, 1 mM phenylmethylsulfonyl fluoride, 5 mM dithiothreitol). For PARP immunodetection cells were lysed in Laemmli sample buffer (10% glycerol, 1% sodium dodecyl sulfate, 30 mM Tris-HCl, pH 6.8). Protein concentration was determined on cleared lysates using the Bradford reagent (Bio-Rad Laboratories GmbH), and Western blots were performed as described previously (13). The following antibodies were used: anti-p53 (FL-393), anti Bax (B-9), and anti-caspase-3 (H-277) (Santa Cruz Biotechnology, Santa Cruz, CA); anti-vinculin (V9131) and anti-Actin (AC-40, SigmaAldrich); anti poly(ADP-ribose)polymerase (clone C2–10) (R&D Systems, Minneapolis, MN); anti-Bcl-X (BD Transduction Laboratories,

1 The abbreviations used are: FADD, Fas-associated death domain; DN, dominant negative; PARP, poly(ADP-ribose)polymerase; crmA, cytokine response modifier A, GFP, green fluorescent protein; FACS, fluorescence-activated cell sorter; CHAPS, 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonic acid; DAPI, 4,6-diamidino-2-phenylindole; wt, wild type; Z-VAD-fmk, benzyloxycarbonyl-VAD-fluoromethyl ketone.

21319

FIG. 1. Myc sensitizes cells to apoptosis induced by several different chemotherapeutic agents. Dose-response curves of TGR-1 (MYC wt) and HO15.19 (MYC null) cells treated with different chemotherapeutic drugs. Cells were incubated with increasing doses of the chemotherapeutic drugs for 72 h, and the level of apoptosis was assessed by sub-G1 content using FACS analysis. Values represent the mean ⫾ STD of three experiments. Lexington, KY); and anti-cleaved caspase-3 (Asp-175) (Cell Signaling Technology, Beverly, MA). Immunofluorescence—Cells were fixed with 4% paraformaldehyde in phosphate-buffered saline. Permeabilization and staining with monoclonal anti-cytochrome c (clone 6H2.B4, Pharmingen) was performed according to Goldstein et al. (14). Cells were counterstained with DAPI. Fluorogenic Assay for Caspase-3/7 Activity—Cell lysates were obtained as described for immunoblotting. 30 mg/ml proteins were used for fluorogenic assays as described (15). Briefly, lysates were incubated with the synthetic substrate ac-Asp-Glu-Val-Asp-7-amido-4-methylcoumarin (ac-DEVD-AMC, Alexis) (40 ␮M) in an assay buffer containing 100 mM Hepes-NaOH, pH 7.4, containing 10% sucrose, 0.1% Nonidet P-40, 1 mM EDTA, 5 mM dithiothreitol, supplemented with the protease inhibitors aprotinin and leupeptin (20 mg/ml each), pepstatin (1 mg/ml), and phenylmethylsulfonyl fluoride (1 mM). Cleavage of the fluorogenic substrate was monitored for 15 min by amido-4-methylcoumarin release in a fluorescence spectrophotometer (excitation/emission wavelengths 380/460 nm). Caspase activity was expressed as arbitrary units of fluorescence/min/mg of protein, and the values obtained at each time point were divided by the value of the relative untreated sample to obtain the -fold increase. RESULTS

Myc-deficient Cells Are Resistant to Topoisomerase II Inhibitors and Display Different Levels of Sensitivity toward Other Cytotoxic Drugs—To confirm previous reports in which Myc was shown to play a physiological role in chemotherapy-induced apoptosis (9, 10), Myc wild type (wt) (TGR-1) and MYC null (HO15.19) cells were incubated with increasing doses of a selection of chemotherapeutic drugs, and the level of apoptosis was assessed by FACS analysis at different time points between 24 and 72 h (Fig. 1 and data not shown). The drugs tested are currently employed in cancer therapy and are representative of different classes of compounds known to affect different targets.

21320

MYC-dependent Activation of Serine Proteases and Caspases

As evident from the data presented in Fig. 1 MYC null cells, despite showing a higher rate of spontaneous cell death than the parental cells in untreated cultures, were resistant to topoisomerase II inhibitors up to relative high concentrations. In fact, a 20-fold higher concentration of etoposide was needed to induce apoptosis to a similar extent in MYC null fibroblasts as compared with wt fibroblasts (Fig. 1A). The difference was even more striking in the case of doxorubicin where a 30-fold higher concentration of the drug killed 25% less MYC null cells as compared with wt cells (Fig. 1B). Loss of Myc expression also partially impaired the apoptotic response toward taxol and staurosporine (Fig. 1, C and D), though to a different extent. As reported previously (9), MYC null cells are not resistant to all apoptotic stimuli, because a normal apoptotic response was triggered when HO15.19 cells were treated with camptothecin (Fig. 1E). To study in more detail the mechanism by which MYC null cells are resistant to drug-induced apoptosis, we decided to use doxorubicin as a model drug, because even at very low doses it is very active toward wt cells but completely ineffective toward MYC null cells. Myc Status Determines the Resistance to Apoptosis of HO15.19 Cells—The apoptotic defect in HO15.19 cells was revealed using drugs whose apoptotic effect is known to be dependent, completely or in part, on the presence of wild type p53 (reviewed in Ref. 16). Previous data have suggested (9) that TGR-1 and HO15.19 cells contain wild type p53 and that the induction of p53 in HO15.19 is partly impaired upon treatment with etoposide and cisplatin. To understand whether the treatment of TGR-1 and HO15.19 cells with doxorubicin results in stabilization of p53, we treated the cells with 0.01 ␮g/ml doxorubicin for 24 h and determined the level of p53 by Western blotting. As shown in Fig. 2A, the treatment of the cells resulted in a robust accumulation of p53 with no differences observed between the MYC null and wt cell lines. Moreover, the accumulation of Bax, a classical p53 target involved in apoptosis (17), following doxorubicin treatment strongly suggests that functional p53 is induced in the two cell lines (Fig. 2B). Our data therefore suggest that, at least for some kinds of drugs, p53 induction alone is not sufficient but that a concomitant presence of Myc is necessary to induce apoptosis after DNA damage. Moreover, the apoptotic defect was specifically due to the loss of Myc and not to secondary mutations accumulated as a consequence of Myc absence, because re-expression of Myc in HO15.19 cells rescued doxorubicin-induced apoptosis both in short term (Fig. 2C) and in long term (Fig. 2D) assays. Lack of Effector-Caspase Activation in myc Null Cells—The signaling cascades triggered once the apoptotic stimulus is recognized as such (induction phase) can be largely modulated, mainly by the Bcl2 family of apoptotic regulators (decision phase), before the apoptotic program is executed via activation of the caspases (execution phase). Because we observed that the Bcl2 family member Bax is induced both in TGR-1 and HO15.19 cells, we concluded that the induction phase and at least part of the decision phase are not impaired even in absence of functional Myc. To assess whether the apoptotic process is blocked at the level of the execution phase in MYC null cells, we monitored the integrity of poly(ADP-ribose)polymerase (PARP) as a marker of caspase activation (Fig. 2E). The cleavage product of PARP is evident upon doxorubicin treatment of TGR-1 and 7M3 cells but is absent in HO15.19 cells, indicating that Myc is required to activate the execution program in response to doxorubicin treatment. Defective activation of caspases however, is not a generic defect of MYC null cells, because PARP is normally cleaved when HO15.19 cells are exposed to camptothecin.

FIG. 2. Myc status determines the resistance to apoptosis of HO15.19 cells and correlates with a lack of activation of effector caspases. A, induction of p53 in MYC wt and null cells upon treatment with doxorubicin. Cell lysates were prepared at the indicated times after treatment with doxorubicin, and Western blots were probed with antibodies to the indicated proteins. Vinculin served as loading control. B, Bax is induced in both MYC wt as well in MYC null cells 24 h after doxorubicin addition. Western blots were probed with the indicated antibodies. Actin served a loading control. C, Myc re-expression rescues apoptosis in HO15.19 cells. Pictures of TGR-1, HO15.19, and Mycinfected HO15.19 (7M3) cells treated (Doxo) or untreated (Cnt) with 0.01 ␮g/ml doxorubicin for 72 h. D, colony assay of TGR-1, HO15.19, and 7M3 cells untreated (Cnt) or treated (Doxo) with 0.01 ␮g/ml doxorubicin. The experiment was performed as described under “Experimental Procedures.” E, lack of PARP cleavage in Myc-deficient cells following doxorubicin treatment. Anti-PARP immunoblot of TGR-1, HO15.19, and 7M3 cell lysates obtained at different time points after treatment with doxorubicin (Doxo, top part) and camptothecin (Cpt, bottom part). An arrow indicates cleaved PARP. Vinculin served as a loading control.

The Activation of the Fas (CD95) Receptor Is Not Required for Doxorubicin-induced Apoptosis—Chemotherapeutic drugs have been reported to activate effector caspases either via death receptor-mediated activation of caspase-8 (extrinsic pathway) or via mitochondria-mediated activation of caspase-9


21321

FIG. 3. The mitochondrial pathway, not the extrinsic pathway, is required for doxorubicin-induced apoptosis. A, FasL induces efficient apoptosis in Myc wt and null cells. FACS analysis of the DNA content of TGR-1 and HO15.19 cells incubated for 72 h with 25 ng/ml of purified FasL. B, activation of the Fas receptor is not required for doxorubicin-induced apoptosis. The percentage of cell death of TGR-1 cells exposed to 0.01 ␮g/ml doxorubicin (Doxo) in the presence or absence of 30 ␮g/ml MFL4, an antagonistic antibody able to block the activation of Fas. As a control, TGR-1 cells treated with 25 ng/ml purified FasL in the presence or absence of 30 ␮g/ml MFL4 were analyzed. Values represent the mean ⫾ S.D. of three independent experiments. C, overexpression of dominant negative mutants of FADD and caspase-8 does not affect doxorubicin-induced apoptosis in TGR-1 cells. TGR-1 cells were infected with viruses expressing GFP alone (Pinco) or containing the dominant negative of human FADD (FADD DN) or human caspase-8 (casp8 DN). Forty-eight hours after infection cells were treated with 0.01 ␮g/ml doxorubicin or 25 ng/ml recombinant FasL. Seventy-two hours after drug treatment floating cells were pooled with those harvested by trypsinization, and DNA content was evaluated by FACS analysis. Apoptosis was evaluated only in GFP-positive gated cells. Values represent the mean ⫾ S.D. of three independent experiments. D, cytochrome c (␣Cyt C) is not released from the mitochondria in doxorubicin-treated myc null cells. Anti-cytochrome c staining of untreated (a, c, e) and 36-h doxorubicin (Doxo, g, i, k)- or camptothecin (Cpt)-treated TGR-1, HO15.19, and 7M3 cells. DAPI counterstaining is shown (b, d, f, h, j, l, n, p, r).

(intrinsic pathway) (18). To understand whether the resistance to doxorubicin-induced apoptosis in MYC null cells is because of a defective activation of Fas/FasL, we incubated TGR-1 and HO15.19 cells with 25 ng/ml purified FasL and assessed the level of apoptosis by FACS analysis (Fig. 3A). Because no differences were evident, we concluded that the resistance of MYC null cells to doxorubicin is not because of an intrinsic defect in the activation of the Fas receptor. Subsequently, we tested whether doxorubicin-induced apoptosis in wt cells requires Fas/FasL engagement at the cell surface. TGR-1 cells were exposed to 0.01 ␮g/ml doxorubicin in the presence or absence of 30 ␮g/ml MFL4, an antagonistic antibody able to block the activation of Fas. As a control we treated TGR-1 cells with 25 ng/ml purified FasL in the presence or absence of 30 ␮g/ml MFL4. As expected, MFL4 co-incubation prevented FasL-induced apoptosis. However, it did not decrease the percentage of cell death induced by doxorubicin (Fig 3B) demonstrating that blocking Fas/FasL engagement at the cell surface does not affect doxorubicin-induced apoptosis.

Anticancer drugs may trigger apoptosis also by acting directly on different components of the Fas signaling pathway, e.g. by triggering a Fas ligand-independent, FADD-mediated activation of the Fas death pathway (19). Moreover, anticancer drugs can directly activate caspase-8, the upstream caspase in the Fas pathway, in a Fas-independent manner (20). To assess the involvement of FADD or caspase-8 in doxorubicin-induced cell death, TGR-1 cells were infected with retroviruses expressing dominant negative (DN) versions of human caspase-8 or human FADD. The “empty” retrovirus (Pinco) was used as control. Forty-eight hours after infection the cells were treated with 0.01 ␮g/ml doxorubicin or 25 ng/ml FasL as a positive control, and 72 h after treatment the percentage of apoptosis was evaluated by FACS analysis of GFP-positive cells. As shown in Fig. 3C, DN caspase-8 and FADD proteins efficiently blocked FasL-induced apoptosis. However, doxorubicin-induced apoptosis was neither inhibited nor delayed by the two DN proteins. These data demonstrate that Fas/FasL and caspase-8 are dispensable for doxorubicin-induced apoptosis.

21322


FIG. 4. Overexpression of BclXL is not sufficient to inhibit doxorubicininduced apoptosis. A, generation of wt Myc cells (TGR-1) overexpressing BclXL. The Western blot comparing levels of BclXL expression in a representative clone of TGR-1 overexpressing human BclXL (TGR-1BclXL), parental TGR-1, HO15.19, and 7M3 cells is shown. Actin was used as a loading control. B, BclXL does not protect against doxorubicin-induced apoptosis. Pictures of TGR-1 and TGR-1-BclXL cells treated (Doxo) or not treated (Cnt) with doxorubicin for 72 h are shown. C, overexpressed BclXL is localized at the mitochondria. Anti-BclXL and anti-cytochrome c staining of TGR-1 (a and b, respectively) and TGR-1-BclXL cells (d and e, respectively) are shown. D, the release of cytochrome c by doxorubicin treatment is inhibited in TGR-1 cells overexpressing BclXL. Anti-cytochrome c staining of TGR-1 and TGR-1-BclXL cells non-treated (b, f) or treated with doxorubicin for 36 h (d, h). DAPI counterstaining is also shown (a, c, e, and g).

Subsequently, we sought to determine whether the mitochondrial pathway mediates doxorubicin-induced apoptosis in TGR-1 cells (Fig. 3D). Cytochrome c release was studied by immunofluorescence in cells treated with doxorubicin or, as a positive control, with camptothecin. In healthy cells cytochrome c is mitochondrial as indicated by the typical punctuate distribution of the staining in all cell lines (Fig. 3D, a, c, and e). Following doxorubicin treatment of TGR-1 and 7M3 cells, the staining becomes intense and diffuse, which is consistent with a cytoplasmic distribution of cytochrome c released from mitochondria, (Fig. 3D, g and k). This staining pattern was not observed in HO15.19 cells (Fig. 3D, i) indicating that the block in the apoptotic machinery resides upstream of the mitochondria. Once again, the defect is specifically related to the inability of MYC null cells to die in response to doxorubicin, because cytochrome c is released as result of camptothecin treatment (Fig. 3D, o). Overexpression of BclXL Is Not Sufficient to Inhibit Doxorubicin-induced Apoptosis—To understand whether the inability to activate the mitochondrial pathway could account for the lack of drug-induced apoptosis in MYC null cells, we tested whether an exogenously applied mitochondrial block would be sufficient to abolish doxorubicin-induced apoptosis. To this end, we infected TGR-1 cells with a retrovirus overexpressing human BclXL, a member of the Bcl2 family, which blocks the release of cytochrome c from the mitochondria (21). Several clones were obtained that overexpressed BclXL, and when we

tested BclXL expression we discovered that compared with TGR-1 HO15.19 cells expressed very high levels of endogenous BclXL, similar to those of the overexpressing clones. We found that the high levels of BclXL are a consequence of the absence of Myc, because re-expression of Myc in 7M3 cells restored BclXL expression to normal levels (Fig. 4A). The finding that HO15.19 cells express high levels of BclXL could account for the inability of these cells to release cytochrome c upon doxorubicin stimulation. We then tested several clones for their biological response to doxorubicin, and to our surprise none of the tested clones was protected from apoptosis, despite expressing high levels of BclXL (Fig. 4B and data not shown). A possible explanation for the lack of protection could be that the overexpressed BclXL is mislocalized and/or non-functional. However, immunofluorescence studies showed that BclXL was correctly localized at the mitochondria, because in untreated cells the signal from BclXL colocalized with that from cytochrome c (Fig. 4C, f). Moreover, we observed that overexpressed BclXL is functional, because it blocks cytochrome c release after doxorubicin exposure (Fig. 4D, compare d with h). Despite that, BclXL is not able to prevent apoptosis as evident from the typical morphological features such as condensation and fragmentation of the nuclei (Fig. 4D, c and g). However, a modest but reproducible delay in the onset of the apoptotic process was observed in all of the experiments performed with BclXL overexpressing clones (data not shown).


21323

TABLE I List of inhibitors and range of concentrations tested to block doxorubicin-induced apoptosis of TGR-1 cells

FIG. 5. Doxorubicin can induce apoptosis independent of caspase-3/7. A, caspase-3 is activated in cells independently of mitochondria. Lysates from TGR-1 cells and TGR-1 BclXL were probed with an antibody specific for the cleaved form of caspase-3, and the expression of vinculin served as a loading control. B, caspase-3/7 activation is strongly diminished in cells expressing XIAP (W310A). Bulk populations of puromycin selected TGR-1-pBabe, TGR-1-BclXL-pBabe, TGR1-W310A, and TGR-1-BclXL-W310A were treated with doxorubicin for the indicated time. Caspase-3/7 activity was measured by a specific fluorogenic assay. The experiment was repeated three times giving similar results, and the graph shows values from a representative experiment. C, caspase-3/7 blockade is not sufficient to prevent doxorubicin induced-apoptosis. Pictures of the indicated cell lines treated (Doxo) or not treated (Cnt) with doxorubicin for 72 h. The dishes photographed here are the same used for the caspase-3/7 assay in B.

These findings suggest that the mitochondrial pathway is not the only executor of doxorubicin-induced apoptosis in TGR-1 cells and that an alternative pathway is activated upstream or concomitantly when cells are exposed to the drug. A corollary emerging directly from these observations is that this

Inhibitor

Protection

150 ␮g/ml CHX (protein synthesis inhibitor) 0-200 ␮M pepstatin A (cathepsin D inhibitor) 0-800 ␮M CA-074 Me (cathepsin B inhibitor) 0-200 ␮M MDL28170 (calpain I/II inhibitor) 0-400 ␮M E64d (cysteine protease inhibitor) 100 ␮M Z-VAD-fmk (pan-caspase inhibitor) 100 ␮M TLCK (serine protease inhibitor)

Yes No No No No Yes Partial

second pathway is also under the control of Myc, because the mitochondrial block described in MYC null cells cannot account solely for the resistant phenotype. Inhibition of Caspase-3 Activation Does Not Protect TGR-1 Cells from Doxorubicin-induced Apoptosis—As shown in Fig. 5A, we observed cleaved, and therefore active, caspase-3 in the presence of a mitochondrial block obtained by BclXL overexpression. We therefore sought to determine whether caspase-3 is the pivotal activator/effector of the alternative pathway triggered by doxorubicin. To this end we infected both TGR-1 and TGR-1-BclXL with a XIAP point mutant (W310A) that does not interact with caspase-9 and only inhibits caspase-3/7 (22). As shown in Fig. 5B, the W310A mutant suppressed most of the caspase-3/7 activity elicited by doxorubicin treatment in both TGR-1 and in TGR-1-BclXL. Unexpectedly, W310A overexpression, although being able to slightly delay the onset of cell death in TGR-1-BclXL, did not prevent the cumulative number of cell deaths (Fig. 5C). A later Smac- and/or HtrA2/Omi-mediated relief of caspase inhibition, following the release of these proteins from activated mitochondria, might explain the finding that W310A overexpression is not able to suppress cell death in TGR-1 cells. In agreement with such a notion, we detected a residual caspase-3/7 activity in cells expressing W310A (Fig. 5B). However, residual caspase-3/7 activity is present also in TGR-1-BclXL-W310A where Smac and/or HtrA2/Omi release should be blocked as a consequence of mitochondrial protection afforded by BclXL overexpression. These observations led us to search for additional and non-conventional mechanisms of caspase-3 activation triggered by doxorubicin. Serine Protease Activation Occurs prior to Early Caspase Activation in Doxorubicin-induced Apoptosis—Recent studies (reviewed in Refs. 23 and 24) have indicated that proteases other than caspases (including cathepsins, calpains, and serine proteases) also have roles in mediating and promoting apoptosis. In fact, their inhibition completely, or in some cases partially, prevents cell death induced by different stimuli. Moreover, depending on the cell type and on the inducer non-caspase proteases can be activated either upstream or downstream of caspases. To investigate whether non-caspase proteases could play a role in activating caspase-3 and/or executing cell death upon doxorubicin treatment, we induced apoptosis in presence or absence of increasing concentrations of different protease inhibitors. No protection was observed when pre-incubating cells with 0 –200 ␮M pepstatin A (cathepsin D inhibitor), 0 – 800 ␮M CA-074 Me (cathepsin B inhibitor), 0 –200 ␮M MDL28170 (calpain I/II inhibitor), and 0 – 400 ␮M E64d (general cysteine protease inhibitor) as summarized in Table I. Protection from doxorubicin-induced apoptosis was instead obtained either by use of the pan-caspase inhibitor Z-VAD-fmk or by the protein synthesis inhibitor cycloheximide. It is important to note that full rescue by Z-VAD-fmk preincubation was obtained only when using very high concentrations (100 ␮M) of the drug, which is known to affect other classes of proteases also. Accordingly, apoptosis was also significantly decreased, though not

21324


FIG. 6. Serine protease inhibitors delay caspase-3 activation but not cytochrome c release. A, inhibition of serine proteases delays caspase-3 (casp-3) activation. The indicated cells were treated with 0.01 ␮g/ml doxorubicin (Doxo) for the indicated amount of time, and caspase-3 activation was measured by Western blotting. B, serine proteases are not required for doxorubicin-induced cytochrome c release. Anti-cytochrome c staining of TGR-1-GFP (a) and TGR-1crmA (c) cells treated with doxorubicin for 30 h. DAPI counterstaining is also shown (b, d). C, a combination of serine protease inhibitors, the blocking of caspase-3/7 and mitochondrial activation is required to protect against doxorubicin-induced apoptosis. Pictures of TGR-1 cells expressing the indicated proteins treated with 0.01 ␮g/ml doxorubicin for 72 h. D, colony assay of TGR-1 cells expressing the indicated proteins and HO15.19 cells. Cells were seeded at low density after a 24-h treatment with 0.01 ␮g/ml doxorubicin, as indicated under “Experimental Procedures.”

completely blocked, when pre-incubating TGR-1 cells with 100 ␮M tosyl lysine chloromethyl ketone (trypsin-like serine protease inhibitor) for 1 h before the addition of 0.01 ␮g/ml doxorubicin (data not shown). We could not use concentrations higher than 100 ␮M because of intrinsic toxicity of the inhibitor. We therefore wished to confirm these results by using another serine protease inhibitor, namely the cowpox serpin crmA (25). As shown in Fig. 6C, crmA-overexpressing cells were partially resistant to doxorubicin, and consistent with a role of caspase-3 in doxorubicin-induced apoptosis the overexpression of crmA led to a strong reduction of caspase-3 activation (Fig. 6A). These results suggest that serine proteases contribute to caspase-3 activation after treatment with doxorubicin. However, serine proteases do not control the mitochondrial pathway, because their inhibition does not influence the release of cytochrome c (Fig. 6B). Concomitant Inhibition of Caspase-3, Mitochondria, and Serine Protease(s) Rescue TGR-1 Cells from Doxorubicin-induced Apoptosis—To test the contribution of serine proteases to apoptosis in cells in which one or more apoptotic pathways are blocked, we infected TGR-1 cells with or without W310A or TGR-1-BclXL cells with or without W310A, with retroviruses containing GFP or GFP-crmA. GFP-sorted cells were treated with doxorubicin and, as evident in Fig. 6C, the expression of crmA is not sufficient to completely protect against apoptosis

when only one additional pathway (either caspase-3 or mitochondria) is inhibited. However, apoptosis is completely blocked when serine proteases are inhibited concomitantly with the mitochondria and caspase-3. Accordingly, in this setting no caspase-3 cleavage is observed (data not shown). Importantly, the protective effect of crmA is observed in both short term (72 h) and long term (2 weeks) assays (Fig. 6, C and D). Taken together these results demonstrate that a number of pathways contribute to the activation of caspases after doxorubicin treatment, and these pathways are all inhibited in HO15.19 cells, because no caspase-3 cleavage is detectable in doxorubicin-treated MYC null cells (Fig. 6A). DISCUSSION

In this paper we confirm and extend previous observations that some classes of anticancer drugs induce an apoptotic response that is dependent on Myc. Moreover, through a detailed analysis of doxorubicin-induced apoptosis we demonstrate that together with the classical mitochondrial pathway two other pathways are concomitantly triggered to execute cell death, all of which are impaired in Myc deficient cells. Finally, we demonstrate that the activation of serine proteases occurs in parallel with and upstream of caspase activation and significantly contributes to doxorubicin-induced apoptosis. The first significant finding is that the apoptotic response to

MYC-dependent Activation of Serine Proteases and Caspases some classes of anticancer drugs is dependent on Myc. Previous results (9) have shown that Myc is required for etoposideinduced apoptosis. We tested another inhibitor of the same class, namely doxorubicin, because etoposide and doxorubicin are structurally unrelated molecules that in addition to topoisomerase II also have other targets (26). We also tested a broad range of other anticancer drugs and found that the absence of Myc greatly reduces apoptosis induced by taxol, partially impairs staurosporine-induced death, but is irrelevant for camptothecin-induced apoptosis demonstrating a differential requirement for Myc in the induction of apoptosis. Our data are in apparent contrast to results reported by another group (10), who did not observe the resistance of MYC null cells to etoposide, doxorubicin, and taxol at any tested concentration. A main difference in the experimental approach may account for this as we, for instance, used lower concentrations of the tested drugs. Remarkably, the highest differences in apoptotic response between MYC wt and null cells are evident for the lowest concentrations of each drug. Accordingly, at the highest concentration tested a significant amount of cell death occurs even in absence of Myc. The most significant finding presented in this paper is the identification of the diverse Myc-dependent apoptotic pathways triggered by doxorubicin. It is well accepted that chemotherapeutic drugs induce cell death by triggering either the extrinsic or intrinsic pathway of caspase activation (18). The extrinsic pathway is initiated by death receptor-mediated activation of caspase-8, which subsequently activates caspase-3, the main effector caspase. Mitochondrial activation is not required for the extrinsic pathway, and when it occurs it represents a positive feedback loop that amplifies the extent of caspase activation (27). Therefore Bcl2 or BclXL overexpression usually does not protect from apoptosis triggered through the extrinsic pathway. In several models, signaling through Fas/FasL has been shown to be involved in doxorubicin-induced apoptosis (28), for example via up-regulation at the transcriptional level of the FasL (27). Moreover, expression of FasL in activated T-cells is regulated by Myc (29) through a “non-canonical” DNA binding element (30), and Myc null B cells derived from conditional knock-out mice express low levels of Fas and FasL following activation (31). Despite the fact that these results suggest a role for Myc in the regulation of drug-induced apoptosis mediated by Fas/FasL we could not demonstrate a role for these proteins in doxorubicin-induced apoptosis (Fig. 3). Indeed, neither the block of the receptor nor the interruption of the effectors downstream of the receptor were effective in reducing doxorubicin-induced apoptosis. The intrinsic or mitochondrial pathway is triggered through the action of pro-apoptotic members of the Bcl2 family, which regulate the mitochondrial release of cytochrome c. This release allows apoptosome formation and caspase-9 activation resulting in caspase-3 activation and apoptosis. Overexpression of anti-apoptotic members of the Bcl2 family blocks the mitochondrial pathway at the level of cytochrome c release and therefore protects cells from apoptosis (18). We found that in absence of Myc the mitochondrial pathway is not activated upon doxorubicin treatment as measured by lack of cytochrome c release. Interestingly, the lack of mitochondrial activation may be because of an increase in the endogenous levels of BclXL in MYC null cells. This increase appears to be a direct consequence of Myc loss, because its re-expression re-establishes the normal levels of BclXL in the fibroblasts. This finding is in agreement with a previously reported role of Myc in repressing BclXL levels (32). Myc-mediated suppression of BclXL is also consistent with the concept that Myc sensitizes cells to apoptosis through the regulation of cytochrome c re-

21325

FIG. 7. Model for doxorubicin-induced apoptosis. See text for details.

lease (5). However, the lack of Myc-mediated mitochondrial activation is not sufficient to account for the doxorubicin-resistance of MYC null cells, because BclXL overexpression does not prevent doxorubicin-induced apoptosis. To identify other mechanisms involved in doxorubicin-induced apoptosis we searched for “atypical” pathways of caspase activation. Reviewing the literature we noticed several examples in which caspase-3 is activated independently of the mitochondrial pathway (33, 34). Moreover, caspase-3 can be directly activated by non-caspase enzymes like the calcium-dependent enzyme calpain, lysosomal enzymes (35), or enzymes contained in the granules of cytotoxic cells like granzymes and perforins (36, 37). Often caspase-3 activation represents the crucial effector in which inhibition blocks apoptosis (38). Interestingly we found that caspase-3 was in part activated independently of the mitochondria suggesting that besides the classical intrinsic pathway, doxorubicin can also activate caspases independently of the mitochondria. Moreover, this pathway is dependent on Myc also, because caspase-3 (Fig. 6A) is not activated in doxorubicin-treated MYC null cells. Finally, the finding that doxorubicin-induced cell death was not blocked by caspase-3 inhibition in BclXL-overexpressing MYC wt cells (Fig. 5C) suggests that a third caspase-3-independent, but Myc-dependent, pathway is activated by doxorubicin. Although such a mechanism is not commonly used, several examples have been reported in which apoptosis can occur in absence of caspase-3 activation (reviewed by Ref. 38). Caspase-3-independent apoptosis may involve other caspases. Alternatively, or in addition, another class of proteases, namely serine proteases, shown to be involved in some forms of apoptosis (24) may contribute to caspase-3 activation. Significant to this study, it has been shown previously (39) that apoptosis triggered in rat fibroblasts by Myc overexpression can be inhibited by the use of a serine protease inhibitor (39). In our study, the inhibition of serine proteases allowed a complete suppression of doxorubicin-induced apoptosis only if it was concomitant with mitochondrial and caspase-3 inhibition (Fig. 6, C and D) suggesting that activation of serine proteases represents the third Myc-dependent pathway triggered by doxorubicin. In agreement with this suggestion are recent data demonstrating that the levels of two serine protease inhibitors are increased in MYC null cells (40). Taken together our data suggest that doxorubicin triggers

21326


three different but linked Myc-dependent protease pathways (Fig. 7). Together with the classical mitochondrial pathway a serine protease-mediated pathway participates in early caspase-3 activation. However, activation of the mitochondrial pathway is independent of serine protease activity, because cytochrome c release occurs in the presence of serine protease inhibition. Moreover, another pathway is likely to converge on caspase-3 activation, because a XIAP mutant specific for caspase-3 is required together with the inhibition of serine proteases and BclXL to block apoptosis. According to recent data (33) a possible candidate for this additional pathway could be caspase-12. Our results also show that the serine proteases have targets other than caspase-3, because their inhibition is required together with caspase-3 inhibition to protect against cell death. In support of other targets of serine proteases are recent mutational analysis data of the serine protease HtrA2/ Omi showing that both XIAP binding and the catalytic activity of HtrA2/Omi contribute to its pro-apoptotic activity (41). Interestingly, HtrA2/Omi was recently reported (42) to be a p53dependent target gene important for p53-mediated apoptosis. Because MYC null cells possess higher levels of two serine protease inhibitors it is tempting to speculate that HtrA2/Omi is inhibited in MYC null cells. Further studies will be required to test this hypothesis. We cannot formally rule out the possibility that the Fas/FasL system might contribute to the apoptotic effect of doxorubicin in parallel with the above mentioned pathways as we have not inhibited this pathway in combination with the inhibition of the other pathways. However, in our experiments we did not see any effect on apoptosis while blocking Fas signaling suggesting that the Fas pathway is not required for doxorubicininduced apoptosis. Finally, two predictions arise from our work. First, because we demonstrate a dependence for Myc in drug-induced apoptosis, the major prediction would be that Myc-overexpressing tumors are hypersensitive to Myc-dependent drugs, thus allowing the use of lower doses. Consistent with this, a recent report (43) showed that tumors containing amplified c-MYC were hypersensitive to drugs. Second, our finding that more than one pathway is involved in apoptosis induced by anticancer agents predicts that it is the combination of alterations affecting the different pathways that determines how a tumor responds to drug treatment. Further understanding of the regulation and cross-talk among the different pathways might therefore provide more potential targets for antitumor drugs. Acknowledgments—We thank Elena Colli for help with cell culture, and Simona Ronzoni, Ivan Muradore, and Mario Faretta for help with FACS analysis. We thank John M. Sedivy for cell lines and Colin Duckett and Vishva Dixit for plasmids. We thank Maria Eugenia Guicciardi and the Helin laboratory for helpful discussions. REFERENCES 1. Dang, C. V. (1999) Mol. Cell. Biol. 19, 1–11 2. Askew, D. S., Ashmun, R. A., Simmons, B. C., and Cleveland, J. L. (1991) Oncogene 6, 1915–1922 3. Evan, G. I., Wyllie, A. H., Gilbert, C. S., Littlewood, T. D., Land, Brooks, M.,

Waters, C. M., Penn, L. Z., and Hancock, D. C. (1992) Cell 69, 119 –128 4. Hueber, A. O., Zornig, M., Lyon, D., Suda, T., Nagata, S., and Evan, G. I. (1997) Science 278, 1305–1309 5. Juin, P., Hueber, A. O., Littlewood T., and Evan, G. I. Genes Dev. 13, 1367–1381 6. Klefstrom, J., Vastrik, I., Saksela, E., Valle, J., Eilers M., and Alitalo, K. (1994) EMBO J. 13, 5442–5450 7. Sears, R. C., and Nevins, J. R. (2002) J. Biol. Chem. 277, 11617–11620 8. Shi, Y., Glynn, J. M., Guilbert, J., Cotter, T. G., Bissonnette, R. P., and Green, D. R. (1992) Science 257, 212–214 9. Adachi, S., Obaya, A. J., Han, Z., Ramos-Desimone, N., Wyche, J. H., and Sedivy, J. M. (2001) Mol. Cell. Biol. 21, 4929 – 4937 10. Soucie, E. L., Annis, M. G., Sedivy, J., Filmus, J., Leber, B., Andrews, D. W., and Penn, L. Z. (2001) Mol. Cell. Biol. 21, 4725– 4736 11. Mateyak, M. K., Obaya, A. J., Adachi, S., and Sedivy, J. M (1997) Cell Growth & Differ. 8, 1039 –1048 12. Grignani, F., Kinsella, T., Mencarelli, A., Valtieri, M., Riganelli, D., Grignani, F., Lanfrancone, L., Peschle, C., Nolan, G. P., and Pelicci, P. G. Cancer Res. 58, 14 –19 13. Vigo, E., Muller, H., Prosperini, E., Hateboer, G., Cartwright, P., Moroni, M. C., and Helin, K. (1999) Mol. Cell. Biol. 19, 6379 – 6395 14. Goldstein, J. C., Waterhouse, N. J., Juin, P., Evan, G. I., and Green, D. R. (2000) Nat. Cell Biol. 2, 156 –162 15. Del Bello, B., Valentini, M. A., Zunino, F., Comporti, M., and Maellaro, E. (2001) Oncogene 20, 4591– 4595 16. Chang, E. H., Pirollo, K. F., and Bouker, K. B. (2000) Mol. Med. Today 6, 358 –364 17. Shen, Y., and White, E. (2001) Adv. Cancer Res. 82, 55– 84 18. Reed, J. C. (2003) Cancer Cell 3, 17–22 19. Micheau, O., Solary, E., Hammann, A., and Dimanche-Boitrel, M. T. (1999) J. Biol. Chem. 274, 7987–7992 20. Kim, P. K., Mahidhara, R., and Seol, D. W. (2001) Drug Resist. Updat. 4, 293–296 21. Adams, J. M., and Cory, S. (2001) Trends Biochem. Sci. 26, 21–26 22. Bratton, S. B., Lewis, J., Butterworth, M., Duckett, C. S., and Cohen, G. M. (2002) Cell Death Differ. 9, 881– 892 23. Johnson, D. E. (2000) Leukemia (Baltimore) 14, 1695–1703 24. Leist, M., and Ja¨ a¨ ttela¨ , M. (2001) Nature Rev. Mol. Cell. Biol. 2, 1–10 25. Komiyama, T., Quan, L. T., and Salvesen, G. S. (1996) Adv. Exp. Med. Biol. 389, 173–176 26. Gewirtz, D. A. (1999) Biochem. Pharmacol. 57, 727–741 27. Herr, I., and Debatin, K. M. (2001) Blood 98, 2603–2614 28. Friesen, C., Fulda, S., and Debatin, K. M. (1999) Leukemia 13, 1854 –1858 29. Brunner, T., Kasibhatla, S., Pinkoski, M. J., Frutschi, C., Yoo, N. J., Echeverri, F., Mahboubi, A., and Green, D. R. (2000) J. Biol. Chem. 275, 9767–9772 30. Kasibhatla, S., Beere, H. M., Brunner, T., Echeverri, F., and Green, D. R. (2000) Curr. Biol. 10, 1205–1208 31. de Alboran, I. M., O’Hagan, R. C., Gartner, F., Malynn, B., Davidson, L., Rickert, R., Rajewsky, K., DePinho, R. A., and Alt, F. W. (2001) Immunity 14, 45–55 32. Eischen, C. M., Woo, D., Roussel, M. F., and Cleveland, J. L. (2001) Mol. Cell. Biol. 21, 5063–5070 33. Kilic, M., Schafer, J., Hoppe, R., and Kagerhuber, U. (2002) Cell Death Differ. 9, 125–137 34. Oubrahim, H., Stadtman, E. R., and Chock, P. B. (2001) Proc. Natl. Acad. Sci. U. S. A. 98, 9505–9510 35. Hishita, T., Tada-Oikawa, S., Tohyama, K., Miura, Y., Nishihara, T., Tohyama, Y., Yoshida, Y., Uchiyama, T., and Kawanishi, S. (2001) Cancer Res. 61, 2878 –2884 36. Chen, L., Woo, M., Hakem, R., and Miller, R. G. (2003) Eur. J. Immunol. 33, 769 –778 37. Darmon, A. J., Nicholson, D. W., and Bleackley, R. C. (1995) Nature 7, 446 – 448 38. Porter, A. G., and Ja¨ nicke, R. U. (1999) Cell Death Differ. 6, 99 –104 39. Kagaya, S., Kitanaka, C., Noguchi, K., Mochizuki, T., Sugiyama, A., Asai, A., Yasuhara, N., Eguchi, Y., Tsujimoto, Y., and Kuchino, Y. (1997) Mol. Cell. Biol. 17, 6736 – 6745 40. O’Connell, B. C., Cheung, A. F., Simkevich, C. P., Tam, W., Ren, X., Mateyak, M. K., and Sedivy, J. M. (2003) J. Biol. Chem. 278, 12563–12573 41. Verhagen, A. M., Silke, J., Ekert, P. G., Pakusch, M., Kaufmann, H., Connolly, L. M., Day, C. L., Tikoo, A., Burke, R., Wrobel, C., Moritz, R. L., Simpson, R. J., and Vaux, D. L. (2002) J. Biol. Chem. 277, 445– 454 42. Jin, S., Kalkum, M., Overholtzer, M., Stoffel, A., Chait, B. T, and Levine, A. J. (2003) Gene Dev. 17, 359 –367 43. Arango, D., Corner, G. A., Wadler, S., Catalano, P. J., and Augenlicht, L. H. (2001) Cancer Res. 61, 4910 – 4915