Oncogene (2001) 20, 4995 ± 5004 ã 2001 Nature Publishing Group All rights reserved 0950 ± 9232/01 $15.00 www.nature.com/onc

Integrin signaling inhibits paclitaxel-induced apoptosis in breast cancer cells

Fawzi Aoudjit1 and Kristiina Vuori*,1

1Cancer Research Center, The Burnham Institute, La Jolla, California, CA 92037, USA

Inherent or acquired drug resistance is one of the major matase inhibitors as hormonal agents) are most problems in chemotherapy. The mechanisms by which intensely studied (Fornier et al., 1999). In addition to cancer cells survive and escape the cytotoxic e€ects of the ongoing clinical trials, it is clear that a better chemotherapeutic agents are essentially unknown. In the understanding of the cell biology of drug resistance present study, we demonstrate that in the MDA-MB-231 could provide immense help when designing new and MDA-MB-435 breast cancer cells, ligation of b1 therapeutic strategies for breast cancer treatment. One integrins by their extracellular matrix ligands inhibits of the exciting developments in the past few years has signi®cantly apoptosis induced by paclitaxel and vincris- been the realization that several chemotherapeutic tine, two microtubule-directed chemotherapeutic agents agents, including microtubule-directed drugs such as that are widely used in the therapy of breast cancer. We paclitaxel, vinblastine and vincristine, exert their show that b1 integrin signaling inhibits drug-induced cytotoxic e€ects via the induction of apoptosis, or apoptosis by inhibiting the release of cytochrome c from programmed cell death (Kaufmann and Earnshaw, the mitochondria in response to drug treatment. Further, 2000). Thus, a hypothesis has been formulated that can integrin-mediated protection from drug-induced apoptosis explain how tumor cells acquire a broad drug and inhibition of cytochrome c release are dependent on resistance, namely by acquiring a resistance to the activation of the PI 3-kinase/Akt pathway. Our apoptosis, the ultimate common pathway of chemo- results identify b1 integrin signaling as an important therapy-induced cell death (see e.g. Pusztai et al., survival pathway in drug-induced apoptosis in breast 1998). The factors that prevent tumor cells from cancer cells and suggest that activation of this pathway undergoing drug-induced apoptosis in vivo and thus may contribute to the generation of drug resistance. may help tumor cells to acquire drug resistance are Oncogene (2001) 20, 4995 ± 5004. largely unknown at present. Integrins are a/b-heterodimeric membrane proteins Keywords: apoptosis; breast cancer; chemotherapeutic that mediate cell adhesion to the surrounding extra- agent; extracellular matrix; integrin; paclitaxel cellular matrix (ECM). In addition to their mechanistic role in anchoring the cells to the substratum, integrins have also the capacity to elicit a wide variety of Introduction intracellular signals that regulate cell behavior (Schwartz et al., 1995). It is known that epithelial One of the most important advances in oncology came and endothelial cells are absolutely dependent on out of the understanding that most patients with integrin-mediated cell attachment for their survival; primary breast cancer have disseminated tumors at thus, these cells undergo apoptosis upon loss of the time of diagnosis and that increased survival results integrin-mediated cell attachment, a process referred only from e€ective systemic therapy used in conjunc- to as anoikis (Frisch and Ruoslahti, 1997). More tion of operation (Fisher et al., 1997). While signi®cant recently, we and others have shown that integrin- advances have been made with adjuvant therapy, mediated cell attachment mediates survival signaling in optimal therapy has not been de®ned for any subset many types of normal cells against di€erent apoptotic of patients, however. One of the hurdles in the stimuli, such as serum withdrawal (Lee and Juliano, adjuvant chemotherapy, especially in the case of 2000; O'Brien et al., 1996; Zhang et al., 1995), metastatic breast cancer, is the inherent or acquired activation-induced cell death in T lymphocytes (Aoud- drug resistance of the tumor (Lebwohl and Canetta, jit and Vuori, 2000) and ligation of death receptors in 1999). Thus, several clinical trials are underway to endothelial cells (Aoudjit and Vuori, 2001). Among the evaluate the optimal adjuvant chemotherapy to over- signaling molecules involved in integrin-mediated cell come drug resistance; at present, paclitaxel/docetaxel survival is the focal adhesion kinase FAK (Almeida et alone or in combination with e.g. doxorubicin (and al., 2000; Aoudjit and Vuori, 2000; Frisch et al., 1996; with tamoxifen, trastuzumab (Herceptin) and/or aro- Gilmore et al., 2000; Hungerford et al., 1996; Ilic et al., 1998), which becomes activated following integrin ligation and may in turn activate downstream survival pathways such as those composed of phosphatidylino- *Correspondence: K Vuori; E-mail: [email protected] Received 21 February 2001; revised 12 April 2001; accepted 18 sitol 3'-kinase (PI 3-kinase) and the serine/threonine April 2001 kinase Akt (Gilmore et al., 2000; Khwaja et al., 1997; Cell adhesion regulates drug-induced apoptosis F Aoudjit and K Vuori 4996 Rytomaa et al., 2000). Other survival pathways, such taxel and vincristine (Srivastava et al., 1999; Yeung et as the MAPK/ERK pathway (Aoudjit and Vuori, al., 1999), which, alone or in combination, are the 2001; Le Gall et al., 2000), NF-kB (Scatena et al., drugs of choice in breast cancer chemotherapy. We 1998) and Bcl-2-proteins (Rosen et al., 2000; Zhang et found the same here; 24 h treatment of MDA-MB-231 al., 1995) have also been implicated in integrin survival and MDA-MB-435 cells with 50 nM paclitaxel or with signaling in epithelial and endothelial cells. Integrin 54 nM vincristine induced cell death in 50 and 70% of signaling has been demonstrated to regulate cell death the cells, respectively, as monitored by Trypan blue and survival also in cancer cells. Thus, it has been exclusion assay (data not shown). In all the subsequent shown that the avb3 integrin promotes survival of studies in this paper, both paclitaxel and vincristine melanoma cells (Montgomery et al., 1994; Petitclerc et were utilized; identical results with the two chemother- al., 1999). Integrin-mediated cell attachment has also apeutic agents were obtained, and the results for been shown to modulate cancer cell responses to paclitaxel are shown and discussed here. chemotherapeutic agents (Damiano et al., 1999; Sethi Previous studies have demonstrated that MDA-MB- et al., 1999; Uhm et al., 1999). In adriamycin-resistant 231 cells attach equally well on several ECM proteins, MCF-7 cells, enhanced expression of the a5b1 integrin such as type I collagen, ®bronectin, laminin-1 and and subsequent cell adhesion to ®bronectin correlate vitronectin (Lichtner et al., 1998; Lundstrom et al., with the inhibition of apoptosis following serum 1998; van der Pluijm et al., 1997; Wong et al., 1998), withdrawal (Nista et al., 1997). Furthermore, expres- and we found the same here. Also of note, no sion of the a6b1 integrin has been shown to be di€erences in cell survival were observed when MDA- necessary not only for adhesion and invasion, but also MB-231 cells were plated on di€erent matrix substrates for cell survival and tumor growth in vivo of the MDA- or on the non-integrin ligand polylysine for extended MB-435 metastatic breast cancer cell line (Wewer et al., periods of time in the absence of any drug treatment 1997). At present, the molecular events that mediate (not shown). In order to examine the potential survival integrin survival signaling in cancer cells are poorly role of matrix attachment in drug-induced apoptosis in understood. breast cancer cells, MDA-MB-231 cells were plated on In the present work, we demonstrate that apoptosis wells that had been coated with various ECM proteins, induced by two microtubule-directed chemotherapeutic or as a control with the non-integrin ligand polylysine. agents, paclitaxel and vincristine, is signi®cantly The cells were then treated with di€erent concentra- inhibited by b1 integrin signaling in metastatic breast tions of paclitaxel for various amounts of time, and cancer cells. Our results further demonstrate that apoptosis was determined by DNA fragmentation integrin-mediated cell attachment protects breast assay or by Hoechst staining; similar results were cancer cells from drug-induced apoptosis by inhibiting obtained with the two methods. Also, similar results the release of cytochrome c from mitochondria. were obtained with all concentrations of paclitaxel and Furthermore, we provide evidence that the integrin- time points tested; results obtained under conditions in mediated inhibition of apoptosis and cytochrome c which the cells were treated with 50 nM of paclitaxel release in breast cancer are dependent on the PI 3- for 24 h are shown in the ®gures. As shown in Figure kinase/Akt survival pathway. Our results identify b1 1a, paclitaxel-treatment induced a signi®cant increase integrin signaling as an important survival pathway in in DNA fragmentation in cells that had been plated on drug-induced apoptosis in breast cancer cells and uncoated tissue culture dishes or on polylysine. suggest that activation of this pathway may contribute Interestingly, a signi®cant inhibition in DNA fragmen- to the generation of drug resistant phenotype. tation was observed in cells that had been plated on ®bronectin or type I collagen (37 and 42% inhibition, respectively). Cell attachment on two other matrix proteins, vitronectin (Figure 1a) and laminin-1 (data Results not shown), however, failed to signi®cantly protect the cells against drug-induced apoptosis. Thus, these Integrin-mediated cell adhesion inhibits drug-induced results suggest that attachment-induced survival signal- apoptosis in MDA-MB-231 and MDA-MB-435 cells ing may take place not only in an integrin-dependent, Recent studies have indicated that matrix attachment is but also in an integrin-speci®c manner in breast cancer capable of modulating drug-induced apoptosis in some cells. types of cancer cells, such as myeloma, glioma and In order to determine which integrin(s) are involved small cell lung cancer (Damiano et al., 1999; Sethi et in the inhibition of paclitaxel-induced apoptosis al., 1999; Uhm et al., 1999), and we decided to examine following MDA-MB-231 cell attachment to ®bronectin whether the same is true for breast cancer cells. As and collagen I, the cells were preincubated with cellular models for our studies, we utilized the classical blocking antibodies against various a and b subunits breast cancer cell lines MDA-MB-231 and MDA-MB- of integrins, and the ability of ®bronectin and collagen 435, both of which have been widely used in in vitro I to protect the cells from paclitaxel-induced apoptosis and in vivo studies. Importantly, it has been previously was determined. As shown in Figure 1b, incubation of demonstrated that these metastatic breast cancer cells MDA-MB-231 cells with antibodies that inhibit the are sensitive to the death-inducing e€ects of the function of the a5b1 integrin, a known ®bronectin microtubule-targeting chemotherapeutic agents pacli- receptor, completely abolished the ability of ®bronectin

Oncogene Cell adhesion regulates drug-induced apoptosis F Aoudjit and K Vuori 4997 to protect the cells against paclitaxel-induced apopto- survival. As expected, cell attachment assays demon- sis. Importantly, antibodies against another ®bronectin strated that a5b1 integrin contributes to MDA-MB-231 receptor, the a4b1 integrin, had no e€ect on cell cell adhesion on ®bronectin (data not shown) (Lichtner et al., 1998; van der Pluijm et al., 1997). Similar results were observed when the integrin speci®city for cell survival on collagen I was examined. The survival e€ect of collagen I was speci®cally and signi®cantly inhibited by antibodies against the a2b1 integrin, but not the a1b1 or the a3b1 integrins (Figure 1c). Cell attachment assays demonstrated that a2b1 integrin, which is a known collagen receptor, mediates MDA- MB-231 cell attachment to collagen I (data not shown) (Lichtner et al., 1998; Lundstrom et al., 1998; van der Pluijm et al., 1997). Taken together, our results indicate that signaling via the a2b1 and a5b1 integrins signi®cantly inhibits chemotherapeutic agent-induced apoptosis in MDA-MB-231 cells. We also examined the capability of de®ned matrix proteins and integrins to support cell survival in MDA- MB-435 cells. Previous studies have demonstrated that these cells, similar to MDA-MB-231 cells, attach to ®bronectin, collagen and laminin-1 (Seftor et al., 1998; Shaw et al., 1996). Similar to MDA-MB-231 cells, plating of MDA-MB-435 cells to uncoated dishes or to dishes that had been coated with polylysine failed to protect the cells against paclitaxel-induced apoptosis (Figure 2a). Interestingly, cell attachment to collagen I and ®bronectin also failed to provide protection against paclitaxel-induced apoptosis, despite the fact that these cells adhere on collagen I and ®bronectin via the a2b1 and a5b1 integrins, respectively (data not shown) (Seftor et al., 1998; Shaw et al., 1996). As noted above, ligation of these integrins provided protection against paclitaxel-induced apoptosis in MDA-MB-231 cells. When the MDA-MB-435 cells were plated on laminin-1, however, a signi®cant inhibition in DNA fragmentation was observed in response to paclitaxel treatment (Figure 2a). This protective e€ect was signi®cantly blocked by preincu- bating the cells with antibodies against the a6b1 integrin, which is a known laminin receptor in these cells (Figure 2b; Shaw et al., 1996). These results are in contrast to our ®ndings in MDA-MB-231 cells; these cells attach to laminin-1, but via the a3b1 integrin (van der Pluijm et al., 1997), and attachment of MDA-MB- 231 cells on laminin-1 failed to support survival of the cells upon paclitaxel treatment. Taken together, our studies indicate that matrix attachment provides protection against drug-induced apoptosis in breast cancer cells, and that this protection takes place in a Figure 1 Integrin-mediated cell adhesion to ®bronectin and collagen I inhibits paclitaxel-induced apoptosis in MDA-MB-231 matrix protein-, integrin- and cell type-speci®c manner. cells. (a) MDA-MB-231 cells were plated on dishes that had been coated or not with polylysine (PLL) or with various ECM proteins, such as ®bronectin (FN), collagen I (COLL) or b1 integrin-mediated protection against drug-induced vitronectin (VN), as indicated. The cells were treated with apoptosis is PI 3-kinase/Akt-dependent in 50 nM of paclitaxel for 24 h and apoptosis was determined by breast cancer cells DNA fragmentation analysis. The results are representative of three independent experiments. *P50.05 compared to PLL. (b, c) In many cell types, matrix attachment has been shown Prior to cell attachment, the cells were preincubated or not for to result in activation of the PI 3-kinase/Akt pathway 30 min with 10 mg/ml of various inhibitory antibodies against (see Giancotti and Ruoslahti, 1999; Schlaepfer et al., di€erent a and b subunits of integrins as indicated. The cells were treated or not with 50 nM paclitaxel for 24 h and apoptosis was 1999) and we found the same here for MDA-MB-231 determined as above and MDA-MB-435 cells. Adhesion of MDA-MB-231

Oncogene Cell adhesion regulates drug-induced apoptosis F Aoudjit and K Vuori 4998

Figure 2 Integrin-mediated cell adhesion to laminin-1 inhibits paclitaxel-induced apoptosis in MDA-MB-435 cells. (a) MDA-MB- 435 cells were plated on dishes that had been coated or not with polylysine (PLL) or with various ECM proteins, such as ®bronectin (FN), collagen I (COLL) or laminin-1 (LN), as indicated. The cells were treated with 50 nM of paclitaxel for 24 h and apoptosis was determined by DNA fragmentation analysis. The results are representative of three independent experiments. *P50.05 compared to PLL. (b) Prior to cell attachment, the cells were preincubated or not for 30 min with 10 mg/ml of various inhibitory antibodies against a6orb1 subunits of integrins as indicated, or with control IgG antibodies (`IgG'). The cells were treated or not with 50 nM paclitaxel for 24 h and apoptosis was determined as above

cells to ®bronectin or collagen I and adhesion of galactosidase. Cell survival and apoptosis was then MDA-MB-435 cells to laminin-1 induced a twofold analysed in b-galactosidase-positive cells as described increase in Akt phosphorylation, which was completely in Materials and methods. Importantly, expression of blocked by the addition of the PI 3-kinase inhibitor these constructs by themselves had no e€ect on cell Ly294002 (data not shown). Furthermore, activation of survival and apoptosis in MDA-MB-231 cells (Figure the PI 3-kinase/Akt pathway has been shown to 4). As shown in Figure 4a, paclitaxel-treatment induced mediate integrin survival signaling in epithelial cells cell death in 450% of b-galactosidase-positive cells and protect them against detachment- and serum that had been plated on polylysine, while 70 ± 75% of withdrawal-induced apoptosis (Khwaja et al., 1997; the cells survived when attached on ®bronectin or Lee and Juliano, 2000). Also, exogenous activation of collagen I. However, when dominant-negative forms of this pathway has been demonstrated to inhibit PI 3-kinase or Akt were expressed in the cells, the chemotherapeutic agent-induced apoptosis (Gooch et ability of ®bronectin and collagen I to protect the cells al., 1999). To investigate the role of PI 3-kinase/Akt from paclitaxel-induced apoptosis was completely pathway in cell survival downstream of b1 integrins in abolished (Figure 4a). Further studies indicated that breast cancer cells, two lines of experiments were coexpression of activated form of Akt with the carried out. We ®rst examined the e€ect of the PI 3- dominant-negative form of PI 3-kinase restored the kinase inhibitor Ly294002 on the capability of b1 survival e€ect of ®bronectin and collagen I (Figure 4b). integrin ligation to protect breast cancer cells against Thus, we conclude that integrin-mediated activation of paclitaxel-induced apoptosis. As shown in Figure 3, the PI 3-kinase/Akt pathway mediates the protective preincubation of MDA-MB-231 and MDA-MB-435 e€ects against paclitaxel-induced apoptosis in breast cells with the PI 3-kinase inhibitor completely abol- cancer cells. ished the protective e€ects of ®bronectin and collagen I and laminin-1, respectively, against paclitaxel-induced Paclitaxel-treatment induces cytochrome c release from apoptosis. Importantly, treatment of the cells with mitochondria and decreases Bcl-2 protein levels; these Ly294002 per se did not a€ect cell viability, but it did events are blocked by integrin-mediated cell attachment in potentiate the e€ect of paclitaxel on cells plated on breast cancer cells polylysine (Figure 3). As an additional line of investigation, we examined the e€ect of dominant- Recent reports have indicated that chemotherapeutic negative forms of PI 3-kinase and Akt on integrin- agent-induced apoptosis is, in some cases, associated mediated protection against paclitaxel-induced apopto- with the release of cytochrome c from the mitochon- sis. To this end, MDA-MB-231 cells were transfected dria (Kaufmann and Earnshaw, 2000), which is a with plasmids coding for dominant-negative forms of critical event in the activation of caspase-9 and cell PI 3-kinase or Akt together with a plasmid encoding b- death (Green and Reed, 1998; Li et al., 1997). In

Oncogene Cell adhesion regulates drug-induced apoptosis F Aoudjit and K Vuori 4999

Figure 3 b1 integrin signaling inhibits paclitaxel-induced apoptosis in a PI 3-kinase-dependent manner. MDA-MB-231 (a) and MDA-MB-435 cells (b) were plated on dishes that had been coated or not with polylysine (PLL) or with various ECM proteins, such as ®bronectin (FN), collagen I (COLL) or laminin-1 (LN), as indicated, and simultaneously treated or not with the PI 3-kinase inhibitor Ly294002 (25 mM) for 1 h. Paclitaxel (50 nM) was then added to the wells for an additional 24 h, after which apoptosis was determined as above. The results are representative of three independent experiments particular, it has been shown that paclitaxel induces et al., 2000). Thus, we decided to examine whether cytochrome c release in MDA-MB-468 breast carcino- paclitaxel-induced apoptosis is associated with changes ma cell line (Kottke et al., 1999). Of further note, it has in Bcl-2 and Bax, and whether appropriate matrix been demonstrated recently that matrix attachment attachment in turn might regulate these events in breast inhibits cytochrome c release, at least in MDCK cancer cells. As shown in Figure 6, paclitaxel epithelial cells. Importantly, this inhibition was shown treatment, or attachment of cells to ®bronectin or to take place via activation of the PI 3-kinase/Akt collagen I, had no e€ect on the expression levels of Bax pathway (Rytomaa et al., 2000). Therefore, we decided in MDA-MB-231 cells. We also found that paclitaxel- to examine whether paclitaxel-induced apoptosis in treatment did not induce Bax translocation to the breast cancer cells is associated with cytochrome c mitochondria; this event has been associated with release, and whether appropriate cell ± matrix interac- induction of cytochrome c release, and in some cell tion in turn might block this event. As shown in Figure types, integrin signaling has been shown to prevent Bax 5, treatment of polylysine-attached MDA-MB-231 cells translocation in response to certain apoptotic stimuli with paclitaxel results in the release of cytochrome c in (Gilmore et al., 2000; Rytomaa et al., 2000). However, the cytoplasm, as detected by immunoblot analysis on we found that treatment of polylysine-plated MDA- the cytosolic fractions. Attachment of the cells to MB-231 cells with paclitaxel resulted in a signi®cant ®bronectin or collagen I signi®cantly blocked cyto- decrease in the protein levels of Bcl-2. Attachment of chrome c release in response to paclitaxel. Importantly, the cells to ®bronectin and collagen I in turn blocked this inhibition was completely abolished when the cells the paclitaxel-induced decrease in Bcl-2 protein levels. were treated with the PI 3-kinase inhibitor Ly294002. Interestingly, this inhibition by matrix attachment was Similarly, adhesion of MDA-MB-435 cells on laminin- completely abolished when the cells were pre-treated 1 inhibited paclitaxel-induced cytochrome c release, with the PI 3-kinase inhibitor Ly294002. Taken which was reversed by the treatment of the cells with together, our results demonstrate that integrin- the PI 3-kinase inhibitor (data not shown). Taken mediated inhibition of paclitaxel-induced apoptosis together, these results indicate that integrin signaling correlates with sustained high levels of Bcl-2 in a PI inhibits paclitaxel-induced cytochrome c release and 3-kinase-dependent manner. that this inhibition is mediated by the PI 3-kinase/Akt pathway. Several reports have suggested that the balance Discussion between the pro- and anti-apoptotic proteins of the Bcl-2 family is a determining factor in the control of In order to acquire drug resistance, tumor cells must cytochrome c release and apoptosis in various cell escape the initial cytotoxic e€ects of chemotherapeutic types (Gross et al., 1999). Previously, integrin signaling agents. Understanding of the molecular signals that has been shown to promote cell survival by increasing contribute to the inhibition of chemotherapeutic agent- the levels of the anti-apoptotic Bcl-2 protein (Zhang et induced apoptosis in tumor cells may provide better al., 1995), and by inhibiting the function of the pro- strategies for new drug design in diseases such as breast apoptotic protein Bax (Gilmore et al., 2000; Rytomaa cancer. It is well known that integrin-mediated cell

Oncogene Cell adhesion regulates drug-induced apoptosis F Aoudjit and K Vuori 5000 (Uhm et al., 1999) and in non small lung cancer cells (Sethi et al., 1999), our results demonstrate that integrin signaling is an important survival pathway in cancer cells and can contribute to the development of drug resistance. Our results demonstrate that a2b1 and a5b1 integrin- mediated cell attachment to collagen I and ®bronectin, respectively, inhibits chemotherapeutic agent-induced apoptosis in MDA-MB-231 cells. In MDA-MB-435 cells, however, it is the laminin-1-binding integrin a6b1 that mediates protective e€ects against drug-induced apoptosis, despite the fact that these cells also attach to collagen I and ®bronectin. These ®ndings suggest that breast cancer cells may respond di€erentially to their tissue microenvironment, depending on, for example, the overall expression pro®le and/or activation state of the integrin receptors on the cell surface, and on other signaling events active in the cells. It is interesting to note that the observed integrin-speci®city in the two breast cancer cell lines may be of physiological signi®cance. Thus, MDA-MB-231 cells are known to metastasize avidly to bone, in which type I collagen is very abundant (Coleman and Rubens, 1987). Further, it has been shown that a2b1 integrin-mediated attachment to collagen I is necessary for bone metastasis to take place by breast cancer cells (Lundstrom et al., 1998; van der Pluijm et al., 1997). In turn, laminin-a6b1 integrin interaction has been Figure 4 Dominant-negative forms of PI 3-kinase and Akt shown to enhance survival and in vivo tumor growth of inhibit the survival e€ects of b1 integrins on paclitaxel-induced MDA-MB-435 cells (Wewer et al., 1997). Thus, apoptosis in MDA-MB-231 cells. (a) MDA-MB-231 cells were acquisition of a speci®c `tumor phenotype' of adhesive cotransfected with a control plasmid (`Control') or with plasmids interactions may not only be required for tumorigeni- encoding dominant-negative forms of PI 3-kinase and Akt, city per se and for metastasis to certain target organs, together with a b-galactosidase expression plasmid as described in Material and methods. Following transfection, the cells were but it may also be of importance in providing cell plated on wells that had been coated or not with polylysine survival signaling against chemotherapy. (PLL), ®bronectin (FN) or collagen I (COLL), and then treated Our studies identify a functional role for the PI 3- with paclitaxel (50 nM) for 24 h. Cell death was monitored as kinase/Akt pathway in integrin-mediated survival described in Material and methods. The results are representative of three independent experiments. (b) MDA-MB-231 cells were signaling against drug-induced apoptosis. We found cotransfected with a plasmid encoding the dominant-negative that the protective e€ects of collagen I and ®bronectin form of PI 3-kinase in combination with a control plasmid or in MDA-MB-231 cells and of laminin-1 in MDA- with a plasmid encoding the activated form of Akt (`Myr-Akt'), MD435 cells were completely abolished in the presence together with a plasmid encoding b-galactosidase. Following of the PI 3-kinase inhibitor Ly294002. Further, transfection, the cells were treated with paclitaxel and processed for apoptosis assay as above expression of dominant-negative forms of PI 3-kinase and Akt also blocked the ability of integrin signaling to protect from drug-induced apoptosis. In agreement with previous studies that have shown that the PI 3- attachment protects normal mammary epithelial cells kinase/Akt pathway inhibits several forms of cell death from apoptosis (Frisch and Ruoslahti, 1997), and the (Datta et al., 1999), our results demonstrate that this mechanisms underlying integrin-mediated survival sig- pathway also inhibits drug-induced apoptosis and may naling have started to become uncovered. It is less therefore contribute to the development of drug clear, however, what role integrin signaling has in resistance. As a corollary, inhibition of PI 3-kinase regulation of cell survival and apoptosis in transformed activity in tumor cells might be bene®cial for breast cancer cells. In this study, we demonstrate that chemotherapy. in two metastatic breast cancer cell lines, MDA-MB- In previous studies, the mechanisms by which 231 and MDA-MB-435, b1 integrin signaling inhibits integrins modulate drug-induced apoptosis have not signi®cantly apoptosis induced by paclitaxel and been addressed. Our results indicate that one potential vincristine, two microtubule-directed drugs that are mechanism by which integrins inhibit drug-induced widely used in the clinic in breast cancer therapy. apoptosis, at least in breast cancer cells, is by blocking Taken together with previous reports, which have the release of cytochrome c from the mitochondria. We shown that integrin signaling blocks drug-induced found that ligation of b1 integrins with their respective apoptosis in myeloma (Damiano et al., 1999), glioma ECM ligands signi®cantly diminished the release of

Oncogene Cell adhesion regulates drug-induced apoptosis F Aoudjit and K Vuori 5001

Figure 5 Paclitaxel-induced cytochrome c release from mitochondria is blocked by adhesion on ®bronectin (a) and on collagen I (b) in a PI 3-kinase-dependent manner in MDA-MB-231 cells. MDA-MB-231 cells were plated on polylysine or various matrix proteins as indicated, and treated with paclitaxel (50 nM) for 12 h. As indicated, the cells were preincubated or not with the PI 3- kinase inhibitor Ly294002 (25 mg/ml) for 1 h prior to the addition of paclitaxel. Cytosolic cell extracts were prepared as described in Material and methods and analysed by immunoblotting using an anti-cytochrome c antibody (upper panels). The blots were stripped and reprobed with an anti-GAPDH control antibody to con®rm equal loading (lower panels)

subsequently cell death in paclitaxel-treated breast cancer cells. Our results further demonstrate that the inhibition of cytochrome c release by integrins ligation is mediated by the PI 3-kinase/Akt pathway; thus, the inhibition is completely abolished by the PI 3-kinase inhibitor Ly294002. We have also found that exogenous expression of activated forms of PI 3-kinase and Akt signi®cantly reduces the release of cytochrome c in response to drug treatment, and promotes cell survival Figure 6 Paclitaxel induces downregulation of Bcl-2 protein on `non-survival' substrates, such as polylysine (data levels, which is inhibited by adhesion on ®bronectin and collagen not shown). The mechanisms by which cytochrome c is I in a PI 3-kinase-dependent manner in MDA-MB-231 cells. MDA-MB-231 cells were plated on polylysine or on various released from the mitochondria in response to death matrix proteins as indicated, and treated or not with paclitaxel signals are not well understood at present. It is (50 nM) for 12 h. As indicated, some of the cells were believed, however, that mitochondrial integrity and preincubated with the PI 3-kinase inhibitor Ly294002 (25 mg/ml) the release of cytochrome c is regulated by the balance for 1 h prior to the addition of paclitaxel. Cell extracts were prepared as described in Material and methods and analysed by of pro- and anti-apoptotic proteins of the Bcl-2 family. anti-bcl-2 (upper panel) and anti-Bax immunoblotting (lower While the pro-apoptotic proteins Bax and Bak panel). Relative levels of Bcl-2 were determined by a densito- (Jurgensmeier et al., 1998; Shimizu et al., 1999) have metric analysis. The blots were stripped and reprobed with an been shown to facilitate cytochrome c release, the anti- anti-GAPDH control antibody to con®rm equal loading (not apoptotic proteins Bcl-2 and Bcl-xL antagonize the shown) release (Kluck et al., 1997; Shimizu et al., 1999; Yang et al., 1997). Therefore, it is possible that integrin signaling and the PI 3-kinase/Akt pathway inhibit cytochrome c from the mitochondria into the cyto- drug-induced cytochrome c release by modulating the plasm. The release of cytochrome c into the cytosol is a balance between the anti- and the pro-apoptotic critical event in the activation of the apoptotic proteins of the Bcl-2 family members. Importantly, signaling cascade. In cytosol, cytochrome c is a previous studies have demonstrated that paclitaxel- component of the `apoptosome' complex that results induced cell death is associated with phosphorylation in the activation of caspase-9, which then activates and subsequent degradation of the Bcl-2 protein further downstream caspases, resulting in irreversible (Chadebech et al., 1999; Kottke et al., 1999; Srivastava cell damage and death (Li et al., 1997). The importance et al., 1999; Wang and Passaniti, 1999); in accordance of cytochrome c release in paclitaxel-induced apoptosis with these studies, we found that paclitaxel-treatment is underscored by our ®nding that a speci®c caspase-9 results in a decrease in Bcl-2 protein levels in breast inhibitor (z-LEHD) blocks paclitaxel-induced apoptosis cancer cells (see Figure 6). Furthermore, we found that by about 80% (data not shown). It is thus likely that appropriate integrin ± matrix interaction blocked the integrin signaling, by blocking the release of cyto- decrease in the Bcl-2 levels in a PI 3-kinase-dependent chrome c, inhibits the activation of caspase-9 and manner. Previously, ligation of the a5b1 integrin, but

Oncogene Cell adhesion regulates drug-induced apoptosis F Aoudjit and K Vuori 5002 not of the avb1 integrin, which is also a ®bronectin induced apoptosis in several types of cancer cells, receptor, has been shown to induce cell survival and including breast cancer (MacKeigan et al., 2000). It is enhanced expression of Bcl-2 in ovarian epithelial cells known that integrin ligation activates the Erk kinase (Zhang et al., 1995). In glioma cells, vitronectin- cascade, which contributes to adhesion-mediated mediated protection against chemotherapeutic agent- survival signaling in some cell types (Aoudjit and induced apoptosis correlates with increased ratios of Vuori, 2001; Le Gall et al., 2000). Similar to Bcl-2:Bax and Bcl-xL:Bax (Uhm et al., 1999). The MacKeigan et al. (2000), we found that paclitaxel- mechanisms by which integrin signaling regulates the induced apoptosis in MDA-MB-231 is signi®cantly levels of Bcl-2 are not known; our results suggest that enhanced when the cells are treated with the MEK it may involve activation of the PI 3-kinase/Akt inhibitor PD98059 (data not shown). However, addi- pathway. In this regard, it has recently been demon- tion of the MEK inhibitor did not a€ect the ability of strated that Bcl-2 gene expression is regulated by Akt ®bronectin or collagen I to protect the cells against (Pugazhenthi et al., 2000). Integrin-mediated cell paclitaxel-induced apoptosis. It therefore seems that survival has also been correlated with the regulation activation of the ERK pathway is not necessary for of the pro-apoptotic protein Bax. Thus, integrin integrin-mediated inhibition of paclitaxel-induced signaling has been shown to protect mammary apoptosis to take place in breast cancer cells. A full epithelial cells from anoikis by regulating the apoptotic understanding of the molecular events that mediate function of Bax through its conformation and integrin-dependent survival in cancer cells may provide subcellular localization (Gilmore et al., 2000). Further- new therapeutic tools for cancer treatment. more, inhibition of integrin signaling by matrix detachment of MDCK cells induces cytochrome c release, which can be blocked by expression of Materials and methods activated form of Akt; this inhibition in turn correlates with inhibition of Bax translocation to mitochondria Cells and reagents (Rytomaa et al., 2000). Our results indicate that Human breast carcinoma cell lines MDA-MB-231 and MDA- paclitaxel-induced apoptosis is not associated with MB-435 were from ATCC (Rockville, MD, USA), and they Bax translocation to the mitochondria or with changes were cultured in RPMI 1640 supplemented with 10% FCS, in Bax expression. We further found that integrin 2mML-Glutamine, 50 mg/ml streptomycin and 50 U/ml signaling has no e€ect on Bax expression or subcellular penicillin (Irvine Scienti®c, Santa Ana, CA, USA). Che- localization in breast cancer cells. Thus, we propose motherapeutic agents paclitaxel and vincristine were from that one mechanism by which integrin signaling blocks Sigma. The PI 3-kinase inhibitor Ly294002 was from cytochrome c release in paclitaxel-treated breast cancer Calbiochem (La Jolla, CA, USA). Puri®ed bovine type I cells is by maintaining high protein levels of Bcl-2. This collagen and mouse laminin-1 were from Becton Dickinson does not rule out the possibility that additional (Bedford, MA, USA) and Gibco ± BRL (Grand Island, NY, members of the Bcl-2 family, such as Bad (Page et USA) respectively. Human ®bronectin was obtained from Finnish Red Cross (Helsinki, Finland), vitronectin and al., 2000), might also be involved in paclitaxel-induced polylysine from Sigma. Inhibitory anti-integrin antibodies cytochrome c release and function as targets for were from Chemicon (Temecula, CA, USA). integrin survival signaling. It is also possible that integrin signaling and the PI 3-kinase/Akt pathway could regulate apoptosis at an Plasmids and transient transfections additional level upstream or downstream of cyto- Cells were transiently transfected by using the Novafector chrome c release. For example, caspase-9 itself has reagent from VennaNova, LLC (Pompano Beach, FL, USA) been shown to be phosphorylated and subsequently according to the manufacturer's instructions. Cells were inactivated by Akt (Cardone et al., 1998). Also, grown to 80% con¯uency on 6-well plates or on 10 cm binding of paclitaxel and vincristine to microtubules dishes and transfected with 10 or 20 mg of total DNA. Forty- eight hours later, the cells were washed with PBS and used has been shown to inhibit cell cycle progression and to for subsequent experiments. The plasmids encoding activated induce cell cycle arrest prior to the induction of and dominant-negative forms of PI 3-kinase and Akt were apoptosis in cancer cells (Blagosklonny et al., 1997; obtained from Drs Jerrold Olefsky (UCSD, San Diego, CA, Haldar et al., 1997; Kumar, 1981; Srivastava et al., USA) and Phil Tsichlis (Thomas Je€erson University, 1998). Thus, integrin signaling could also modify drug- Philadelphia, PA, USA), respectively. The b-galactosidase induced apoptosis by regulating cell cycle. In this expressing plasmid was from Invitrogen (Carlsbad, CA, respect, it has been shown recently that in myeloma USA). cells, b1 integrin-mediated cell attachment to ®bronec- tin upregulates the levels of the cell cycle inhibitor Cell death experiments and apoptosis assays p27kip, which contributes to adhesion-mediated drug Cells were grown to 80% of con¯uence, detached and plated resistance in these cells (Hazlehurst et al., 2000). It is on wells that were uncoated, or had been coated with 25 mg/ also possible that activation of survival pathways other ml of various ECM proteins, including ®bronectin, collagen than the PI 3-kinase/Akt pathway may contribute to type I, laminin-1 and vitronectin, or with 25 mg/ml of the inhibition of drug-induced apoptosis downstream of non-integrin ligand polylysine. The cells were treated with integrins. It has been shown recently that inhibition of di€erent concentrations of paclitaxel or vincristine for 24 or the MAPK/ERK survival pathway enhances paclitaxel- 48 h. In some experiments, the cells were ®rst preincubated

Oncogene Cell adhesion regulates drug-induced apoptosis F Aoudjit and K Vuori 5003 with the Ly294002 inhibitor for 1 h before adding the c release was detected in the cytoplasmic fractions using the chemotherapeutic drugs. anti-cytochrome c antibody (clone 7H8.2C12) from Pharmin- Apoptosis was monitored by using a cell-death detection gen (San Diego, CA, USA). ELISA kit measuring DNA fragmentation (Boehringer ± For Akt phosphorylation, the cells were plated on wells Mannheim) according to manufacturer's instructions or by that had been coated or not with polylysine, ®bronectin or Hoechst staining as previously described (Aoudjit and Vuori, collagen I in the presence or absence of paclitaxel (50 nM) for 2000). Alternatively, cell death was determined by a b- 2 h. The cells were detached and washed with PBS and lysed galactosidase assay as described (Khwaja et al., 1997). Brie¯y, in RIPA bu€er as described previously (Vuori and Ruoslahti, the cells were transfected with the b-galactosidase reporter 1995). Immunoblot analysis was carried out using an anti- plasmid together with a control vector or with vectors coding phospho-Akt antibody (Ser 473) from Cell Signaling for activated- or dominant-negative forms of PI 3-kinase and Technology (Beverly, MA, USA). The expression levels of Akt. After transfection, the cells were seeded in wells that had Akt in the di€erent samples were determined by using anti- been coated or not with ECM proteins/polylysine as Akt antibody C-20 from Santa Cruz Biotechnology (Santa described above and treated with paclitaxel for 24 h. The Cruz, CA, USA). The expression of Bcl-2 and Bax proteins cells were then ®xed and stained with X-gal for b- was determined in total cell lysates by immunoblotting with galactosidase activity. Positively stained cells were scored as anti-Bcl-2 (Clone 7) and anti-Bax (clone 3) antibodies from normal or apoptotic on the basis of cellular morphology Transduction Laboratories (Lexington, KY). The blots were (Khwaja et al., 1997). also probed with anti-GAPDH (Glyceraldehyde-3-phosphate dehydrogenase) monoclonal antibody (Chemicon Interna- tional, Inc.) to serve as an internal protein control. In all Cytochrome c release, Akt phosphorylation and experiments, immunoblots were visualized by using an Bcl-2/Bax immunoblotting HRPO-conjugated secondary antibody followed by enhanced Cytosolic fractions were prepared as described in (Samali et chemiluminescence detection (Pierce, Rockford, IL, USA). al., 1999). Brie¯y, the cells were washed and resuspended in 100 ml of mitochondrial bu€er (70 mM Tris-base, 0.25 M sucrose and 1 mM EDTA, pH 7.4) for 10 min on ice. An equal volume of the MES bu€er (19.8 mM EGTA, 19.8 mM Abbreviations EDTA, 0.25 mMD-mannitol and 19.8 mM MES, pH 7.4) ECM, extracellular matrix; GFP, green ¯uorescent protein; containing 1 mg/ml of digitonin was added to the samples PI 3-kinase, phosphatidyl inositol 3'-kinase. until 90% of the cells become permeabilized as determined by Trypan blue exclusion. The cells were centrifuged at 900 g for 5 min and the supernatant was further centrifuged for 15 min at 20 000 g to obtain the cytosolic fraction. Samples were Acknowledgments then processed for immunoblot analysis. To ensure the purity We thank Drs Jerrold Olefsky and Phil Tsichlis for of the fractions, we performed an immunoblot analysis with providing some of the plasmids used in this study. We antibodies against the mitochondrial cytochrome c oxidase also thank Kathleen Becherer and Elise Larsen for their (antibody 20E8-C12, Molecular Probes, Eugene, OR, USA) technical assistance. This work was supported by grants and found that only the unbroken cell fraction, which from NIH, California Breast Cancer Research Program contains the mitochondria, and not the cytosolic fraction, and California Cancer Research Program (to K Vuori). K contains cytochrome c oxidase (data not shown). Cytochrome Vuori is a PEW scholar in biomedical sciences.

References

Almeida EA, Ilic D, Han Q, Hauck CR, Jin F, Kawakatsu H, Fornier M, Munster P and Seidman AD. (1999). Oncology, Schlaepfer DD and Damsky CH. (2000). J. Cell Biol., 149, 13, 647 ± 658. 741 ± 754. Frisch SM and Ruoslahti E. (1997). Curr. Opin. Cell Biol., 9, Aoudjit F and Vuori K. (2000). Blood, 95, 2044 ± 2051. 701 ± 706. Aoudjit F and Vuori K. (2001). J. Cell Biol., 152, 633 ± 643. Frisch SM, Vuori K, Ruoslahti E and Chan-Hui PY. (1996). Blagosklonny MV, Giannakakou P, El-Deiry WS, Kingston J. Cell Biol., 134, 793 ± 799. DG, Higgs PI, Neckers L and Fojo T. (1997). Cancer Res., Giancotti FG and Ruoslahti E. (1999). Science, 285, 1028 ± 57, 130 ± 135. 1032. Cardone MH, Roy N, Stennicke HR, Salvesen GS, Franke Gilmore AP, Metcalfe AD, Romer LH and Streuli CH. TF, Stanbridge E, Frisch S and Reed JC. (1998). Science, (2000). J. Cell Biol., 149, 431 ± 446. 282, 1318 ± 1321. Gooch JL, Van Den Berg CL and Yee D. (1999). Breast Chadebech P, Brichese L, Baldin V, Vidal S and Valette A. Cancer Res. Treat., 56, 1 ± 10. (1999). Biochem. Biophys. Res. Commun., 262, 823 ± 827. Green DR and Reed JC. (1998). Science, 281, 1309 ± 1312. Coleman RE and Rubens RD. (1987). Br. J. Cancer, 55, 61 ± Gross A, McDonnell JM and Korsmeyer SJ. (1999). Genes 66. Dev., 13, 1899 ± 1911. Damiano JS, Cress AE, Hazlehurst LA, Shtil AA and Dalton Haldar S, Basu A and Croce CM. (1997). Cancer Res., 57, WS. (1999). Blood, 93, 1658 ± 1667. 229 ± 233. Datta SR, Brunet A and Greenberg ME. (1999). Genes Dev., HazlehurstLA,DamianoJS,BuyuksalI,PledgerWJand 13, 2905 ± 2927. Dalton WS. (2000). Oncogene, 19, 4319 ± 4327. Fisher B, Osborne CK, Margolese RG and Bloomer WD. Hungerford JE, Compton MT, Matter ML, Ho€strom BG (1997). Cancer Medicine, 4th edition.HollandJF,Bast and Otey CA. (1996). J. Cell Biol., 135, 1383 ± 1390. RC, Morton DL, Frei E, Kufe DW and Weichselbaum Ilic D, Almeida EA, Schlaepfer DD, Dazin P, Aizawa S and RR. (eds). Williams & Wilkins: Baltimore, MD. Damsky CH. (1998). J. Cell Biol., 143, 547 ± 560.

Oncogene Cell adhesion regulates drug-induced apoptosis F Aoudjit and K Vuori 5004 Jurgensmeier JM, Xie Z, Deveraux Q, Ellerby L, Bredesen D Rytomaa M, Lehmann K and Downward J. (2000). and Reed JC. (1998). Proc.Natl.Acad.Sci.USA,95, Oncogene, 19, 4461 ± 4468. 4997 ± 5002. Samali A, Cai J, Zhivotovsky B, Jones DP and Orrenius S. Kaufmann SH and Earnshaw WC. (2000). Exp. Cell Res., (1999). EMBO J., 18, 2040 ± 2048. 256, 42 ± 49. Scatena M, Almeida M, Chaisson ML, Fausto N, Nicosia Khwaja A, Rodriguez-Viciana P, Wennstrom S, Warne PH RF and Giachelli CM. (1998). J. Cell Biol., 141, 1083 ± and Downward J. (1997). EMBO J., 16, 2783 ± 2793. 1093. Kluck RM, Bossy-Wetzel E, Green DR and Newmeyer DD. Schlaepfer DD, Hauck CR and Sieg DJ. (1999). Prog. (1997). Science, 275, 1132 ± 1136. Biophys. Mol. Biol., 71, 435 ± 478. Kottke TJ, Blajeski AL, Martins LM, Mesner Jr. PW, Schwartz MA, Schaller MD and Ginsberg MH. (1995). Davidson NE, Earnshaw WC, Armstrong DK and Annu. Rev. Cell Dev. Biol., 11, 549 ± 599. Kaufmann SH. (1999). J. Biol. Chem., 274, 15927 ± 15936. Seftor RE, Seftor EA, Sheng S, Pemberton PA, Sager R and Kumar N. (1981). J. Biol. Chem., 256, 10435 ± 10441. Hendrix MJ. (1998). Cancer Res., 58, 5681 ± 5685. Le Gall M, Chambard JC, Breittmayer JP, Grall D, SethiT,RintoulRC,MooreSM,MacKinnonAC,SalterD, Pouyssegur J and Van Obberghen-Schilling E. (2000). Choo C, Chilvers ER, Drans®eld I, Donnelly SC, Strieter Mol. Biol. Cell, 11, 1103 ± 1112. R and Haslett C. (1999). Nat. Med., 5, 662 ± 668. Lebwohl DE and Canetta R. (1999). Ann. Oncol., 10, 139 ± Shaw LM, Chao C, Wewer UM and Mercurio AM. (1996). 146. Cancer Res., 56, 959 ± 963. Lee JW and Juliano RL. (2000). Mol. Biol. Cell, 11, 1973 ± Shimizu S, Narita M and Tsujimoto Y. (1999). Nature, 399, 1987. 483 ± 487. Li P, Nijhawan D, Budihardjo I, Srinivasula SM, Ahmad M, Srivastava RK, Sasaki CY, Hardwick JM and Longo DL. Alnemri ES and Wang X. (1997). Cell, 91, 479 ± 489. (1999). J. Exp. Med., 190, 253 ± 265. Lichtner RB, Howlett AR, Lerch M, Xuan JA, Brink J, Srivastava RK, Srivastava AR, Korsmeyer SJ, Nesterova M, Langton-Webster B and Schneider MR. (1998). Exp. Cell Cho-Chung YS and Longo DL. (1998). Mol. Cell. Biol., Res., 240, 368 ± 376. 18, 3509 ± 3517. Lundstrom A, Holmbom J, Lindqvist C and Nordstrom T. UhmJH,DooleyNP,KyritsisAP,RaoJSandGladsonCL. (1998). Biochem. Biophys. Res. Commun., 250, 735 ± 740. (1999). Clin. Cancer Res., 5, 1587 ± 1594. MacKeigan JP, Collins TS and Ting JP. (2000). J. Biol. van der Pluijm P, Vloedgraven H, Papapoulos S, Lowick C, Chem., 275, 38953 ± 38956. Grzesik W, Kerr J and Robey PG. (1997). Lab. Invest., 77, Montgomery AM, Reisfeld RA and Cheresh DA. (1994). 665 ± 675. Proc. Natl. Acad. Sci. USA, 91, 8856 ± 8860. Vuori K and Ruoslahti E. (1995). J. Biol. Chem., 270, Nista A, Leonetti C, Bernardini G, Mattioni M and Santoni 22259 ± 22262. A. (1997). Int. J. Cancer, 72, 133 ± 141. Wang W and Passaniti A. (1999). J. Cell. Biochem., 73, 321 ± O'Brien V, Frisch SM and Juliano RL. (1996). Exp. Cell 331. Res., 224, 208 ± 213. Wewer UM, Shaw LM, Albrechtsen R and Mercurio AM. PageC,LinHJ,JinY,CastleVP,NunezG,HuangMand (1997). Am.J.Pathol.,151, 1191 ± 1198. Lin J. (2000). Anticancer Res., 20, 407 ± 416. Wong NC, Mueller BM, Barbas CF, Ruminski P, Quaranta Petitclerc E, Stromblad S, von Schalscha TL, Mitjans F, V, Lin EC and Smith JW. (1998). Clin.Exp.Metast.,16, Piulats J, Montgomery AM, Cheresh DA and Brooks PC. 50 ± 61. (1999). Cancer Res., 59, 2724 ± 2730. Yang J, Liu X, Bhalla K, Kim CN, Ibrado AM, Cai J, Peng Pugazhenthi S, Nesterova A, Sable C, Heidenreich KA, TI, Jones DP and Wang X. (1997). Science, 275, 1129 ± Boxer LM, Heasley LE and Reusch JE. (2000). J. Biol. 1132. Chem., 275, 10761 ± 10766. YeungTK,GermondC,ChenXandWangZ.(1999). Pusztai L, Siddik ZH, Mills GB and Bast RC. (1998). Acta Biochem. Biophys. Res. Commun., 263, 398 ± 404. Oncol., 37, 629 ± 640. Zhang Z, Vuori K, Reed JC and Ruoslahti E. (1995). Proc. Rosen K, Rak J, Leung T, Dean NM, Kerbel RS and Filmus Natl. Acad. Sci. USA, 92, 6161 ± 6165. J. (2000). J. Cell Biol., 149, 447 ± 456.

Oncogene