Cross-talk between Akt, p53 and Mdm2: possible implications ... - Nature

Oncogene (2002) 21, 1299 ± 1303 ã 2002 Nature Publishing Group All rights reserved 0950 ± 9232/02 $25.00 www.nature.com/onc

Cross-talk between Akt, p53 and Mdm2: possible implications for the regulation of apoptosis Tanya M Gottlieb1,4, Juan Fernando Martinez Leal1,4, Rony Seger2, Yoichi Taya3 and Moshe Oren*,1 1

Department of Molecular Cell Biology, Weizmann Institute of Science, Rehovot 76100, Israel; 2Department of Biological Regulation, Weizmann Institute of Science, Rehovot 76100, Israel; 3Radiobiology Division, National Cancer Center Research Institute, Tokyo 104-0045, Japan

The p53 tumor suppressor protein and the Akt/PKB kinase play important roles in the transduction of proapoptotic and anti-apoptotic signals, respectively. We provide evidence that con¯icting signals transduced by Akt and p53 are integrated via negative feedback between the two pathways. On the one hand, the combination of ionizing radiation and survival factor deprivation, which leads to rapid apoptosis of IL-3 dependent DA-1 cells, entails a caspase- and p53-dependent destruction of Akt. This destruction of Akt is not a secondary consequence of apoptosis, since it is not seen when the same cells are triggered to undergo apoptosis under di€erent conditions. On the other hand upon serum stimulation, when Akt becomes active and enhances cell survival, phosphorylation occurs at an Akt consensus site (serine 166) within the Mdm2 protein, a key regulator of p53 function. Taken together, our ®ndings suggest that depending on the balance of signals, p53-dependent downregulation of Akt may promote an irreversible commitment to apoptotic cell death, whereas e€ective recruitment of Akt by appropriate survival signals may lead to activation of Mdm2, inactivation of p53, and eventually inhibition of p53-dependent apoptosis. Oncogene (2002) 21, 1299 ± 1303. DOI: 10.1038/sj/ onc/1205181 Keywords: p53; Akt; Mdm2; apoptosis; caspase The commitment of a cell to apoptosis is dictated through the integration of inputs from stress signals and from survival pathways. Two proteins involved intimately in regulating cell death are p53 and Akt. The p53 tumor suppressor protein is an essential component of signaling pathways responding to various forms of cellular stress, such as DNA damage, deregulated oncogene expression and hypoxia (Gottlieb and Oren, 1996; Khanna and Jackson, 2001; Oren, 1999). One well-documented outcome of p53 activation

*Correspondence: M Oren; E-mail: [email protected] 4 These authors contributed equally to this work Received 13 July 2001; revised 5 November 2001; accepted 13 November 2001

is apoptosis (Gottlieb and Oren, 1996; Moll and Zaika, 2001; Rich et al., 2000; Vousden, 2000). To a large extent, p53 triggers apoptosis by direct transcriptional transactivation of an array of pro-apoptotic genes. In addition, p53 may elicit some of its apoptotic e€ects also through transactivation-independent mechanisms. The protein serine/threonine kinase Akt (also known as protein kinase B or PKB) plays an important role in averting cell death (Datta et al., 1999; Scheid and Woodgett, 2000). A diverse range of physiological stimuli induce Akt kinase activity, including many trophic factors which promote survival, at least in part, through Akt activation via the phosphatidylinositide 3'-OH kinase (PI3K) signaling cascade. Moreover, induced Akt activity (due to overexpression) is sucient to block apoptosis triggered by many death stimuli. Several Akt substrates have been identi®ed whose regulation may underlie the ability of this kinase to promote survival and inhibit apoptosis (Datta et al., 1999; Scheid and Woodgett, 2000). Since p53 and Akt are major but opposing players in signaling pathways that determine cell survival, it is conceivable that there may be some sort of cross-talk between these two proteins to enable integration of signaling information towards a ®nal output: death or survival. To explore such potential p53 ± Akt crosstalk, Akt expression was examined in cells undergoing p53-dependent apoptosis. The experimental system consisted of DA-1 cells, a murine IL-3 dependent lymphoid line, whose response to irradiation in the presence or absence of IL-3 has been extensively studied (Gottlieb et al., 1994, 1996; Gottlieb and Oren, 1998). After treatment with high dose gamma irradiation (500R) in the presence of IL-3, DA-1 cells lose viability with slow kinetics (Gottlieb et al., 1996). In contrast to this slow death, the entire DA-1 culture undergoes apoptosis within several hours after exposure to a similar dose of gamma irradiation in the absence of IL-3. Withdrawal of IL-3 alone results in DA-1 cell death with kinetics intermediate between these two extremes. Each of these three types of DA-1 apoptosis is, at least to some extent, dependent on p53 (Gottlieb et al., 1994, 1996; Gottlieb and Oren, 1998). Total Akt protein level was examined in DA-1 cells after various treatments, as outlined in Figure 1a.

p53 ± Akt cross-talk in apoptosis TM Gottlieb et al

1300

Within 3 h of irradiation in the absence of IL-3, there was a dramatic reduction in Akt protein levels (Figure 1b, lane 8). Such a disappearance of Akt protein was not seen in non-irradiated DA-1 cells deprived of IL-3, nor in cells exposed to irradiation only (Figure 1b, lanes 6 and 7). Reduced levels of Akt protein, only after irradiation in the absence of IL-3, were observed also in the murine, hematopoietic IL-3-dependent BaF3 cell line (data not shown; Mathieu et al., 2001). In IL3-deprived DA-1 cells, the disappearance of Akt protein occurs between 1 and 3 h after irradiation. Since the normal half-life of Akt in untreated cells is at least several hours (data not shown), this rapid destruction of Akt must involve induced proteolysis. Caspase-mediated Akt cleavage, associated with apoptosis, has been reported in various cell types (Bachelder et al., 1999; Rokudai et al., 2000; Widmann et al., 1998). To ®nd out whether this was also the case in irradiated, IL-3 deprived DA-1 cells, cultures were pretreated with the caspase inhibitor zVAD-fmk. As seen in Figure 2a, this pretreatment prevented e€ec-

a

b

c

Figure 1 Downregulation of Akt protein in irradiated DA-1 cells deprived of IL-3. IL-3-dependent mouse lymphoid DA-1 cells were grown as in (Gottlieb et al., 1994) and processed according to the scheme shown in a. Brie¯y, cells were split 1 day before the experiment, and maintained in IL-3 containing medium. At the start of the experiment (time=0) cells were washed twice in IL-3-free medium, and resuspended in medium either containing (lanes 1, 2, 5, 6) or lacking IL-3 (lanes 3, 4, 7, 8). Irradiation (gamma rays, 500R) was performed at t=0 for the cells maintained in IL-3 (lanes 2 and 6) and at t=6 for the cells deprived of IL-3 (lanes 4 and 8). At the indicated time points after +/7 IL-3, cells were collected and protein extracts prepared by direct resuspension in SDS lysis bu€er. Extracts were subjected to Western blot analysis with antibodies speci®c for Akt or tubulin. Loading was normalized for identical numbers of living cells, as determined by Trypan blue exclusion Oncogene

Figure 2 Akt downregulation is associated with irradiation in the absence of IL-3 and is caspase- and p53-dependent. (a) DA-1 cells were collected 9.5 h after IL-3 withdrawal. When indicated, zVAD-fmk (Calbiochem) was added right before IL-3 withdrawal. Control cells were treated with an equivalent amount of DMSO. Cells were harvested and processed as in Figure 1b. Relative Akt levels were calculated by quantitation of the scanned Western blot images, followed by normalization for the corresponding tubulin signals. (b) DA-1 cells were treated as in Figure 1. Cells were collected 26 h after irradiation (lane 2), together with the control untreated cells (lane 1), or 24 h after IL-3 withdrawal (lane 3). Cell death was assayed by Trypan Blue staining. Loaded samples were normalized by the amount of total protein in the extract, determined using the Bradford reagent (Pierce). (c) DA-1 and DIDD cells were either mock treated (lanes 1 and 4) or irradiated with 500R 6 h after IL-3 withdrawal (lanes 2 and 5), and collected 4 h later. The cells in lanes 3 and 6 were maintained for 24 h in the absence of IL-3. DIDD cells were derived from DA-1 by infection with a recombinant retrovirus encoding the C-terminus of p53 (p53DD; Shaulian et al., 1992). Relative Akt levels were determined as described in (a)

p53 ± Akt cross-talk in apoptosis TM Gottlieb et al

a

b

c

Figure 3 Akt-dependent phosphorylation of Mdm2. (a) Consensus Akt phosphorylation sites in Mdm2. The region encompassing serines 166 and 186 of human Mdm2 (HDM2) is shown, along with the corresponding regions of the Mdm2 proteins from other species (Momand et al., 2000). The Akt consensus phosphorylation site (Obata et al., 2000) is shown at the bottom; the putative target serines are indicated with an asterisk. (b) Aktdependent phosphorylation of transfected Mdm2. Antibodies speci®c for an Mdm2 peptide phosphorylated on Ser166 (antiphosphoS166) were raised in rabbits by immunization with the peptide RRRAI(p)SETEENC, employing previously described procedures (Oda et al., 2000). 293T were transfected transiently with expression plasmids encoding Mdm2 and either wild type Akt (wtAkt, left lane) or a kinase-dead mutant of Akt (KD-Akt, right lane; Kohn et al., 1996). Twenty-nine hours later, cells were harvested and protein extracts prepared by direct resuspension in protein sample bu€er. Equal amounts of extract were subjected to Western Blot analysis with a mixture of the Mdm2-speci®c monoclonal antibodies 2A9 and 4B2 (Chen et al., 1993). After comparison of the relative Mdm2 signal intensity in each lane, additional aliquots of the same extracts were reloaded such that there were equal amounts of Mdm2 protein in each lane, and subjected again to Western blot analysis with anti-phosphoS166 (upper panel). Identical aliquots of the same extracts were processed in parallel and probed with a mix of 2A9 and 4B2 as above, to assess the overall amounts of Mdm2 in each lane (lower panel). (c) Serum-dependent phosphorylation of endogenous Mdm2 on Serine-166. U20S human osteosarcoma cells were starved in serum-free medium for 26 h and either not released

tively the reduction in Akt protein level, implying that this reduction is caspase-dependent. Our ®ndings suggest that Akt destruction may be an integral part of a process whereby irradiated, IL-3 deprived DA-1 cells commit to apoptosis by removal of anti-apoptotic proteins. Nevertheless, one might argue that the proteolysis of Akt is merely a secondary consequence of caspase activation during apoptosis. In the latter case, the rate of Akt disappearance would be expected to re¯ect the overall extent of apoptosis in the culture. However, as seen in Figure 2b, this is clearly not the case. No signi®cant reduction in Akt protein levels was observed either 30 h after irradiation in the presence of IL-3, when 28% of the cells were dead (Figure 2b, lane 2), or 24 h after IL-3 withdrawal, where as many as 60% of the cells were already dead (lane 3). In contrast, Akt disappearance was much more pronounced in DA-1 cultures exposed to irradiation in the absence of IL-3: while only 29% of the cells were dead at that point, Akt was already barely detectable. These observations support the conjecture that the reduction in Akt protein is part of an active process that is committing IL-3-starved, irradiated DA-1 cells to rapid apoptosis, and is not merely a passive byproduct of apoptosis. The apoptotic process that takes place in irradiated, IL-3 deprived DA-1 cells is largely driven by the activation of p53 (Gottlieb et al., 1994, 1996; Gottlieb and Oren, 1998). In order to investigate whether p53 function is also required for the early destruction of Akt in this experimental system, a similar analysis was performed with DA-1 cells infected with a retrovirus encoding a negative dominant p53 miniprotein, p53DD (=DIDD cells). The p53DD miniprotein oligomerises with full length wild type p53 and interferes with its activity (Shaulian et al., 1992). Therefore, DIDD cells are e€ectively compromised for p53 function. As seen in Figure 2c, DIDD cells failed to display a signi®cant reduction in Akt protein level when irradiated in the absence of IL-3 (lane 5). Hence, this reduction is dependent on p53 function. The ability of p53 to modulate caspase activity is well documented (e.g. Ding et al., 1998; Gottlieb and Oren, 1998; Miyashita and Reed, 1995; Oda et al., 2000; Schuler et al., 2000). p53-dependent speci®c caspase cleavage of two other anti-apoptotic proteins ± the retinoblastoma tumor suppressor gene product pRb and the Mdm2 oncoprotein ± has also been observed (Gottlieb and Oren, 1998; Pochampally et al., 1999). The molecular mechanism that directs caspases to cleave these proteins, as well as Akt, remains to be elucidated.

1301

(lane 1) or released from starvation by supplementation of the medium with serum (15% serum; lanes 2 and 3). At the indicated time points after serum re-addition, cells were collected and protein extracts prepared by resuspension in NP40 bu€er (0.5% NP40, 50 mM Tris pH8, 150 mM NaCl, protease inhibitor cocktail (Sigma), 100 mM NaF and 2 mM NaVO4). Extracts were subjected to Western blot analysis as in b, except that loaded samples were normalized for total protein content, as determined using the Bradford reagent (Pierce) Oncogene

p53 ± Akt cross-talk in apoptosis TM Gottlieb et al

1302

Under conditions where the apoptotic activity of p53 prevails, it is conceivable that the destruction of Akt plays a role in accelerating the apoptotic process. However, there are situations where Akt activation can overcome the death promoting e€ects of p53 and e€ectively rescue the cells from apoptosis (Sabbatini and McCormick, 1999; Yamaguchi et al., 2000). One attractive possibility is that Akt somehow impinges directly on the p53 pathway, down-modulating p53 function and quenching its apoptotic activity. The major physiological regulator of p53 function is the Mdm2 oncoprotein. Mdm2 binds speci®cally to p53 and blocks its biochemical activities. Furthermore, through its action as a p53-speci®c E3 ubiquitin ligase, Mdm2 ubiquitinylates p53 and targets it for rapid proteasomal degradation (Freedman et al., 1999; Juven-Gershon and Oren, 1999; Momand et al., 2000; Oren, 1999). Therefore, one way whereby Akt might modulate the p53 pathway is by altering Mdm2's intracellular behavior. Indeed, inspection of the amino acid sequence of Mdm2 revealed two consensus Akt phosphorylation sites, corresponding to positions 166 and 186 of human Mdm2 (Obata et al., 2000; Figure 3a). Importantly, both sites are well conserved in evolution (Figure 3a). To explore the possibility that Mdm2 is indeed phosphorylated by Akt within living cells, we attempted to raise polyclonal sera against phosphorylated peptides corre-

Figure 4 Model for the cross-talk between Akt and the p53 pathway. Under conditions leading to an irreversible apoptotic commitment, activation of p53 may contribute to apoptosis by inhibition of Akt. On the other hand, in the presence of appropriate survival signals, Akt activation may lead to Mdm2 phosphorylation. This is speculated to potentiate Mdm2, and thereby incapacitate p53 and overcome its pro-apoptotic e€ects. Indicated is also the autoregulatory feedback loop between p53 and Mdm2, wherein p53 induces mdm2 gene expression whereas the Mdm2 protein inactivates p53 (Juven-Gershon and Oren, 1999; Momand et al., 2000) Oncogene

sponding to the two pertinent regions of Mdm2. While the antibodies raised against the putative Ser186phosphorylated Mdm2 were not found to be e€ective (data not shown), the antibodies directed against the Ser166-phosphorylated form did detect a doublet migrating exactly at the position of Mdm2 (Figure 3b). Moreover, the signal intensity was substantially stronger when the extracts were prepared from cells transfected with wild type Akt (wtAkt), as compared to cells transfected with kinase dead Akt (KD-Akt). This data implies that Serine 166 in Mdm2 can be phosphorylated in an Akt-dependent manner. Co-immunoprecipitation experiments with various cell extracts support the existence a physical association between Mdm2 and Akt (data not shown), suggesting that Akt might phosphorylate Mdm2 directly. Akt enzymatic activity is strongly reduced after serum starvation and is activated by various growth factors (Datta et al., 1999). To ®nd out whether release from serum starvation can lead to enhanced Aktmediated phosphorylation of Mdm2, Western blot analysis was performed on U2OS human osteosarcoma cells released from serum starvation. While there was no detectable change in the overall levels of Mdm2 under these conditions, serum stimulation led to a rapid increase in reactivity with the antibody directed against Ser166-phosphorylated Mdm2 (Figure 3c). The increase was maximal 0.5 h after serum stimulation, and became less pronounced already 1 h after stimulation. These ®ndings argue strongly that Mdm2 undergoes rapid phosphorylation on Ser166 under conditions where endogenous Akt activity is induced, and support the conjecture that Akt can directly phosphorylate Mdm2 under such conditions. Given that both Akt and Mdm2 inhibit p53-mediated apoptosis (Chen et al., 1996; Haupt et al., 1996; Sabbatini and McCormick, 1999; Yamaguchi et al., 2000), it is tempting to speculate that Akt-mediated Mdm2 phosphorylation renders the latter protein more competent to inhibit p53 function. Such outcome is predicted to attenuate the apoptotic e€ects of p53 and increase cell survival. While this work was under review, a report by Mayo and Donner (2001) appeared also describing the Aktdependent phosphorylation of Mdm2. In a series of elegant experiments, these authors used mass spectroscopy to show that both Serine 166 and Serine 186 of Mdm2 indeed serve as targets for direct phosphorylation by Akt in vitro, and are phosphorylated in response to Akt activation in vivo. Moreover, they found that phosphorylation of Mdm2 by Akt enables its translocation from the cytoplasm into the nucleus, and the subsequent inactivation of nuclear p53 by Mdm2. Hence, compelling evidence was provided that phosphorylation of Mdm2 by activated Akt plays an important role in restraining p53 activity, thereby quenching p53-dependent apoptosis. Our ®ndings imply the existence of an intricate cross talk between Akt and the p53 pathway (Figure 4). In apoptosis-prone cells, p53-dependent signaling enables Akt downregulation. This may predispose such cells to rapid apoptosis in response to the combination of

p53 ± Akt cross-talk in apoptosis TM Gottlieb et al

cellular stress and decreased survival signals. p53dependent caspase cleavage of Akt has been described for transfected, integrin overexpressing carcinoma cells deprived of serum and/or matrix attachment (Bachelder et al., 1999). Our ®ndings with DA-1 cells imply that p53dependent Akt destruction may play an important role in the facilitation of apoptosis also in cells that have not been subjected to genetic manipulation. On the other hand, Akt phosphorylates Mdm2 on Serine 166 (this study), as well as on Serine 186 (Mayo and Donner, 2001). Such phosphorylation is enhanced under conditions that lead to Akt activation, such as exposure to serum growth factors. The phosphorylated Mdm2 is more competent to drive p53 inactivation within living cells, thereby promoting survival. The model presented in Figure 4 might help to explain some of the con¯icting results reported recently concerning the ability of Akt to inhibit p53-dependent apoptosis. Thus, whereas constitutively active Akt can protect several other cell types from death induced by p53 activation (Sabbatini and McCormick, 1999; Yamaguchi et al., 2000), it is unable to do so in bone marrow-derived Baf3 cells (Mathieu et al., 2001). Of note, the studies of Mathieu et al. (2001) were performed in a system very similar to the one studied by us, namely irradiated IL-3 dependent cells deprived of IL-3. Indeed, a marked reduction in Akt protein

levels was also observed in Baf-3 cells under these apoptosis-inducing circumstances. Our model predicts that, under conditions when ecient p53-dependent Akt cleavage is triggered, Akt-mediated survival signals will be aborted and will not able to block p53dependent apoptosis. This conjecture is supported by the recent work of Bachelder et al. (2001), who found that a caspase-resistant mutant of Akt is much more e€ective than wt Akt in protecting cells against anoikisinduced apoptosis. On the other hand, when the combination of signals required for Akt destruction is not present, persistent Akt activity will potentiate Mdm2 and antagonize p53-mediated apoptosis.

1303

Acknowledgments We thank D Freedman and AJ Levine for Mdm2-speci®c monoclonal antibodies and RA Roth for Akt expression plasmids. This work was supported in part by grant RO1 CA 40099 from the National Cancer Institute (USA), The USA-Israel Binational Science Foundation, the Kadoorie Charitable Foundations, the Cooperation Program in Cancer Research of the DKFZ and Israel's Ministry of Science (MOS), the Robert Bosch Foundation (Germany) and the Yad Abraham Center for Cancer Diagnosis and Therapy. T Gottlieb was the recipient of an ICRF fellowship.

References Bachelder RE, Ribick MJ, Marchetti A, Falcioni R, Soddu S, Davis KR and Mercurio AM. (1999). J. Cell. Biol., 147, 1063 ± 1072. Bachelder RE, Wendt MA, Fujita N, Tsuruo T and Mercurio AM. (2001). J. Biol. Chem., 276, 34702 ± 34707. Chen J, Marechal V and Levine AJ. (1993). Mol. Cell. Biol., 13, 4107 ± 4114. Chen J, Wu X, Lin J and Levine AJ. (1996). Mol. Cell. Biol., 16, 2445 ± 2452. Datta SR, Brunet A and Greenberg ME. (1999). Genes Dev., 13, 2905 ± 2927. Ding HF, McGill G, Rowan S, Schmaltz C, Shimamura A and Fisher DE. (1998). J. Biol. Chem., 273, 28378 ± 28383. Freedman DA, Wu L and Levine AJ. (1999). Cell Mol. Life Sci., 55, 96 ± 107. Gottlieb E, Ha€ner R, von Ruden T, Wagner EF and Oren M. (1994). EMBO J., 13, 1368 ± 1374. Gottlieb E, Lindner S and Oren M. (1996). Cell Growth Di€., 7, 301 ± 310. Gottlieb E and Oren M. (1998). EMBO J., 17, 3587 ± 3596. Gottlieb TM and Oren M. (1996). Biochim. Biophys. Acta., 1267, 77 ± 102. Haupt Y, Barak Y and Oren M. (1996). EMBO J., 15, 1596 ± 1606. Juven-Gershon T and Oren M. (1999). Mol. Med., 5, 71 ± 83. Khanna KK and Jackson SP. (2001). Nat. Genet., 27, 247 ± 254. Kohn AD, Takeuchi F and Roth RA. (1996). J. Biol. Chem., 271, 21920 ± 21926. Mathieu AL, Gonin S, Leverrier Y, Blanquier B, Thomas J, Dantin C, Martin G, Baverel G and Marvel J. (2001). J. Biol. Chem., 276, 10935 ± 10942.

Mayo LD and Donner DB. (2001). Proc. Natl. Acad. Sci. USA, 98, 11598 ± 11603. Miyashita T and Reed JC. (1995). Cell, 80, 293 ± 299. Moll UM and Zaika A. (2001). FEBS Lett., 493, 65 ± 69. Momand J, Wu HH and Dasgupta G. (2000). Gene, 242, 15 ± 29. Obata T, Ya€e MB, Leparc GG, Piro ET, Maegawa H, Kashiwagi A, Kikkawa R and Cantley LC. (2000). J. Biol. Chem., 275, 36108 ± 36115. Oda K, Arakawa H, Tanaka T, Matsuda K, Tanikawa C, Mori T, Nishimori H, Tamai K, Tokino T, Nakamura Y and Taya Y. (2000). Cell, 102, 849 ± 862. Oren M. (1999). J. Biol. Chem., 274, 36031 ± 36034. Pochampally R, Fodera B, Chen L, Lu W and Chen J. (1999). J. Biol. Chem., 274, 15271 ± 15277. Rich T, Allen RL and Wyllie AH. (2000). Nature,407, 777 ± 783. Rokudai S, Fujita N, Hashimoto Y and Tsuruo T. (2000). J. Cell Physiol., 182, 290 ± 296. Sabbatini P and McCormick F. (1999). J. Biol. Chem., 274, 24263 ± 24269. Scheid MP and Woodgett JR. (2000). Curr. Biol.,10, R191 ± R194. Schuler M, Bossy-Wetzel E, Goldstein JC, Fitzgerald P and Green DR. (2000). J. Biol. Chem., 275, 7337 ± 7342. Shaulian E, Zauberman A, Ginsberg D and Oren M. (1992). Mol. Cell. Biol., 12, 5581 ± 5592. Vousden KH. (2000). Cell, 103, 691 ± 694. Widmann C, Gibson S and Johnson GL. (1998). J. Biol. Chem., 273, 7141 ± 7147. Yamaguchi A, Tamatani M, Matsuzaki H, Namikawa K, Kiyama H, Vitek MP, Mitsuda N and Tohyama M. (2000). J. Biol. Chem., 27, 27. Oncogene