Bax-derived membrane-active peptides act as potent and direct inducers of apoptosis in cancer cells

Although many cancer cells are primed for apoptosis, they usually develop resistance to cell death at several levels. Permeabilization of the outer mitochondrial membrane, which is mediated by proapoptotic Bcl-2 family members such as Bax, is considered as a point of no return for initiating apoptotic cell death. This crucial role has placed Bcl-2 family proteins as recurrent targets for anticancer drug development. Here, we propose and demonstrate a new concept based on minimal active versions of Bax to induce cell death independently of endogenous Bcl-2 proteins. We show that membrane-active segments of Bax can directly induce the release of mitochondria-residing apoptogenic factors and commit tumor cells promptly and irreversibly to caspase-dependent apoptosis. On this basis, we designed a peptide encompassing part of the Bax pore-forming domain, which can target mitochondria, induce cytochrome c release and trigger caspase-dependent apoptosis. Moreover, this Bax-derived ‘poropeptide’ produced effective tumor regression after peritumoral injection in a nude mouse xenograft model. Thus, peptides derived from proteins that form pores in the mitochondrial outer membrane represent novel templates for anticancer agents.


Introduction
The integrity of the mitochondrial outer membrane (MOM) serves as a switch between cell survival and cell death by apoptosis. The Bcl-2 family of proteins are crucial arbiters in this process owing to their ability to either promote or inhibit MOM permeabilization (Aouacheria et al., 2007;Youle and Strasser, 2008). Pro-apoptotic members (e.g. Bax, Bak, BH3-only proteins) promote cytochrome c release from mitochondria, leading to the activation of proteases termed caspases, which mediate cell demise. Conversely, antiapoptotic members such as Bcl-2 or Bcl-xL decrease susceptibility to cell death by neutralizing Bax and Bak or BH3-only proteins.
Overexpression of pro-survival proteins occurs in many human tumors, and can contribute not only to disease development and progression but also to clinical drug resistance (Adams and Cory, 2007). Anti-apoptotic Bcl-2 family members therefore represent prime targets for the development of modern anticancer drugs that have the potential to restore apoptosis and reverse resistance to chemotherapy. Efforts to inhibit the anti-death Bcl-2 family members have focused on the development of cell-permeable peptides or small-molecule inhibitor drugs, which are designed to mimic the BH3 domain of Bcl-2 family death members (Yip and Reed, 2008). A number of such BH3 mimics (e.g. ABT-737) (Oltersdorf et al., 2005), which inactivate Bcl-2-like proteins by binding to their BH3-binding groove, have now entered clinical trials and provide real opportunities for improving the efficacy of cancer treatment. Recently, another strategy has been described that converts pro-survival Bcl-2 molecules into pro-apoptotic proteins with the potential to kill cancer cells. In this new approach, a short peptide derived from the orphan nuclear receptor Nur77 was shown to bind to the N-terminal regulatory region of Bcl-2, altering its structure to expose its BH3 domain, which then becomes free to activate Bax and Bak (Kolluri et al., 2008). However, one major common limitation of these two latter strategies is that they depend on endogenous levels of anti-apoptotic Bcl-2 proteins in cancer cells. Moreover, these strategies are expected to be less effective in inducing apoptosis of tumor cells with mutated or deficient Bax or Bak (Jansson and Sun, 2002;Ouyang et al., 1998;Zong et al., 2001).
Although apoptosis signaling pathways are often compromised during malignant transformation, mitochondria-resident apoptogenic factors are still present in cancer cells and it is an exciting challenge to develop peptides or peptidomimetics capable of inducing their release. Such molecules would have the capacity to promote MOM permeabilization directly, and thus to overcome cancer cell resistance towards induction of apoptosis. Furthermore, molecules that function at the membrane level are less likely to encounter resistance than drugs based on classical 'lock-and-key' binding specificity. Following these ideas, it has been shown in several studies that, upon cell internalization, antimicrobial peptides [e.g. (KLAKLAK) 2 ] can induce cell death in a variety of cell types (Chen et al., 2001;Ellerby et al., 1999;Foillard et al., 2008;Foillard et al., 2009;Law et al., 2006;Mai et al., 2001;Marks et al., 2005;Rege et al., 2007). However, the mechanisms of cell killing exerted by these antibiotic peptides are unclear, because they appear to include both necrosis, which is secondary to plasma membrane disruption (Papo et al., 2006), and apoptosis, which is induced either by upregulation of death effectors (Chen et al., 2001) or by mitochondrial membrane permeabilization (Ellerby et al., 1999;Law et al., 2006;Mai et al., 2001;Marks et al., 2005;Rege et al., 2007). Of note, the cationic peptide (KLAKLAK) 2 has been reported to have very low potency (Borgne-Sanchez et al., 2007;Ellerby et al., 1999), which precludes its use as an effective anticancer drug.
In this context, it is an important goal to identify novel proapoptotic sequences that act directly at the level of mitochondrial permeability that can be exploited to engineer potent anticancer molecules. Among the potential candidates are membranolytic peptides derived from proapoptotic Bcl-2 family proteins such as Bax. This 23 kDa protein contains a number of structurally defined membrane-interacting regions (Suzuki et al., 2000), some of which (1, 9, 5, 6 and a central 56 hairpin motif) have a presumed membrane-targeting function (Annis et al., 2005;Cartron et al., 2005;Garcia-Saez et al., 2004;Heimlich et al., 2004). It has been previously shown that peptides corresponding to the first and/or to the second helix of the putative pore-forming domain of Bax (56 hairpin) can reproduce, at least in part, the poration activity displayed by the full-length parent protein (Garcia-Saez et al., 2005;Garcia-Saez et al., 2006;Guillemin et al., 2010). Hence, helices 5 and 6 of Bax carry by themselves minimal structural information and physicochemical properties to insert into model lipid membranes and form pores. The pores appear to be of the 557 Cytotoxic effects of Bax peptides mixed lipidic-peptidic type (Garcia-Saez et al., 2007;Qian et al., 2008), which are similar to those of membrane-active, amphipathic peptide antibiotics . Here, we report that the two central helices of Bax individually are sufficient to target GFP to mitochondria and induce caspase-dependent cell death. Moreover, we demonstrate that a peptide based on helix 5 can directly induce the release of mitochondrial cytochrome c, thereby acting as a potent apoptosis activator. This peptide, named 'poropeptide Bax ', was more efficient for both mitochondrial targeting and induction of apoptosis than (KLAKLAK) 2 , which is a de novo synthetic peptide. Finally, we report a clear anticancer effect of poropeptide Bax  after peritumoral administration in tumor-bearing mice. Our data establish the feasibility of using short peptides derived from mitochondrial outer-membrane-porating proteins as a basis for designing novel anticancer agents, which might be directly applied to some solid tumors or 'homed' to the tumor microenvironment through the use of specific vectors.

Constructs containing Bax 5 or 6 induce caspasedependent apoptosis in transfected cells
In a search for peptide sequences capable of targeting and disrupting the MOM, recombinant constructs encoding the GFP open reading frame fused to the N-terminus of various membrane- Fig. 1. Ectopic overexpression of GFP-tagged Bax 5 and 6 fragments induces cell death. (A)Chimeric GFP proteins used in this study. GFP-tagged constructs encoding GFP alone, or fusions of GFP with full-length Bax, Bax-1, Bax-5, Bax-6, Bax-56, Bax-5-9 and Bax-9 are represented. The -helical topology of Bax in solution was retrieved from published results (Suzuki et al., 2000). Because the structure of the membrane-bound form of Bax is unknown, we designed peptide versions that extend a few residues beyond the -helical regions determined for the structure in aqueous buffer. (B)Expression and analysis of the various GFP-tagged proteins in mammalian cells. Western blot analyses on transiently transfected HT1080 cells (24 hours after transfection). Proteins were separated by SDS-PAGE followed by immunoblot with anti-GFP antibody (upper panel). Asterisks indicate the various GFP fusions depicted in A. The expected sizes are 27, 48, 29, 30.2, 33.2, 37.2, 29.7 and 28.6 kDa, respectively. Analysis of caspase-3 activation (bottom panel, the cleaved 17 kDa product indicates activated caspase-3) and PARP cleavage (middle panel, the generated 29 kDa PARP fragment is shown). Similar results were obtained using DKO MEFs active fragments of Bax (1, 9, 5, 6, 56 and 5-9) were prepared (see definition of fragments and schemes of constructs in Fig. 1A). Western blot analysis confirmed the correct size of the fusion proteins (Fig. 1B, upper panel). The different constructs were transfected into human HT1080 cells and cell death was measured after 24 hours. As a measure of cell viability, GFPpositive cells were analyzed by Annexin-V staining (Fig. 1C) or scored for nuclear apoptosis (as assessed by morphology) or necrosis (by staining with propidium iodide) (supplementary material Fig. S1, top and middle). GFP alone, GFP-Bax and GFP-Bax-1 had no cytotoxic effect. The other constructs were all able to induce predominantly apoptotic cell death, with maximum activity observed after transfection with GFP-Bax-56, GFP-Bax-5 and GFP-Bax-6, and intermediate levels for GFP-Bax-5-9 and GFP-Bax-9. Furthermore, fusion proteins including the 5 and/or 6 helices of Bax elicited caspase-3 and PARP cleavage, as evidenced by western blot (Fig. 1B, bottom and middle panels, respectively). Consistently, treatment with 100 M zVAD.fmk, a cell-permeable caspase inhibitor, was effective in reducing cell death induced by the toxic GFP fusion proteins (supplementary material Fig. S1, middle), indicating that cell death is caspase dependent. Importantly, the pro-apoptotic effects of the Bax-derived constructs were not exerted through Bax and Bak, because Bax-Bak double-knockout mouse embryonic fibroblasts (DKO MEFs) were as sensitive as wildtype MEFs to expression of Bax-5 (supplementary material Fig.  S1, bottom), whereas they were resistant to treatment with staurosporine ( Fig. 1D).

Fusions including Bax 5 or 6 localize to mitochondria and alter the organelle physiology
The subcellular localization of all assayed GFP-tagged Bax fragments was subsequently evaluated by confocal fluorescence microscopy. Expression of the fusion proteins yielded abundant and intense GFP fluorescence in transfected MEF-DKO cells (Fig.  2). GFP alone showed a diffuse localization. Similarly, GFP-Bax and GFP-Bax-1 distributed evenly between the nuclear and cytoplasmic compartments in transfected cells. By contrast, confocal imaging revealed that GFP-Bax-5, GFP-Bax-6, GFP-Bax-56, GFP-Bax-5-9 and GFP-Bax-9 exhibited a clustered staining that was reminiscent of intracellular membranes. The simultaneous use of a mitochondrion-specific red marker (mitoDsRed) indicated that this punctuated staining colocalized with mitochondria. This was confirmed by immunostaining of cells transfected with GFP-Bax-5 using anti-mitoHsp70, which showed that a large portion of the fusion protein was indeed specifically associated with mitochondria ( Fig. 2, bottom). Of note, GFP-Bax-5 was also found to be more efficient for mitochondrial targeting than a fusion containing the sequence of the designed proapoptotic peptide (KLAKLAK) 2 (Fig. 2).
Mitochondria-dependent apoptosis typically affects the homeostasis of the organelle, which can be investigated by tracing changes of the mitochondrial membrane potential  m . Thus, using the membrane-potential-sensitive dye Mitotracker Red CMXRos, we measured  m in cells expressing either MOMtargeting sequences (GFP-Bax-5, GFP-Bax-6, GFP-Bax-56, GFP-Bax-5-9 and GFP-Bax-9) or non-targeting sequences (GFP and GFP-Bax-1). Examination of individual cells showed that those with strong expression of the cytotoxic, MOM-targeting GFP-tagged fusions exhibited a concomitant decrease of Mitotracker Red staining ( Fig. 3 and   DKO MEFs were co-transfected with mito-DsRed plasmid (encoding DsRed2 fused to the mitochondrial targeting sequence from subunit VIII of human cytochrome c oxidase) and the GFP-tagged constructs. Subcellular distribution was analyzed by confocal microscopy 24 hours after transfection. Confocal images showing GFP (green) and MitoDsRed (red) fluorescence. The DNA staining dye Topro-3 (blue) was used to visualize the nuclei. In merged images, the yellow color shows the colocalization of GFP and MitoDsRed in mitochondria. Similar images were obtained using an antibody detecting mitochondrial Hsp70 (low panel). Scale bar: 10m.
From the experiments described so far, we can conclude that the sequences from the central hairpin of Bax as well as the Bax TM domain (9) contain the necessary information to target the GFP protein specifically to the mitochondrial membranes. However, the Bax chimeras containing 5 or 6 distinguished themselves from the GFP-Bax-9 fusion because they markedly induced depolarization of the mitochondrial membrane and caspasedependent apoptosis. Both Bax-5 and Bax-6, either in the Bax protein (Suzuki et al., 2000) or as individual peptides bound to membranes, form amphipathic -helices (Garcia-Saez et al., 2005;Garcia-Saez et al., 2006). Additionally, they have a similar ratio of hydrophilic to hydrophobic residues (31% and 33%, respectively). However, the expected net charge of these fragments is very different at neutral pH, namely: +4 for Bax-5 and -1 for Bax-6, indicating that Bax-5 is a better candidate for binding and disruption of the mitochondrial outer membrane, rich in negatively charged phospholipids. This is indeed suggested by the higher membrane depolarization observed for chimeras containing Bax-5. For this reason, in the following stages of our work we focus on the Bax-5 active fragment, as a prototype for proof-of-concept evaluation.

A synthetic peptide corresponding to Bax residues 106-134 exhibits potent mitochondrial-poration activity
Based on the above findings, we tested whether a synthetic peptide with the sequence of helix 5 from Bax, residues Asn106 to Arg134 (Table 1;  induce release of cytochrome c from freshly prepared mitochondria (isolated from SK-MEL-28 metastatic human melanoma cells). A 5 minute exposure to 10 M Bax  peptide was sufficient to cause significant release of mitochondrial cytochrome c, and a concentration of 25 M completely depleted all mitochondrial cytochrome c after the same incubation time (supplementary material Fig. S3A, top). For comparison, we also assayed a peptide corresponding to the BH3 domain of Bax (2 helix), which was found to have no effect (supplementary material Fig. S3A, middle). Importantly, unlike the Bax[106-134] peptide, the synthetic (KLAKLAK) 2 peptide was unable to release cytochrome c from isolated mitochondria (supplementary material Fig. S3A, bottom). These results demonstrate that a synthetic, native (non-optimized) peptide encompassing the 5 helix of Bax can on its own disrupt mitochondrial membrane permeability and induce release of cytochrome c. Such an activity is specific for this Bax-derived sequence, because it was not observed at comparable conditions by using another helical fragment of Bax with no reported poration activity (2 helix, i.e. Bax-BH3). This peptide, which is derived from the BH3 domain of Bax, displayed only a weak activity, which started at 25 M peptide concentration after 30-60 minutes of exposure, using human embryonic kidney HEK293T cells. A similar weak activity was also observed for the antimicrobial peptide (KLAKLAK) 2 , showing that the sequence of the Bax[106-134] active fragment has been optimized during natural evolution for this particular function. Further support for the mitochondrial disruption capacity of Bax  was obtained by measuring peptide-induced swelling (SD 50 3.98±0.57 M) and  m dissipation (DD 50 1.68±0.39 M) (supplementary material Fig.  S3B) on liver mitochondria, two characteristics that are indicative of mitochondrial membrane permeabilization. These results illustrate the particularly strong capacity of the Bax  peptide to trigger mitochondrial membrane perforation. Moreover, they provide a rationale for the development of MOMpermeabilizing peptides inspired by helix 5 of Bax, which might then be used to induce apoptosis in cancer cells.

Bax[106-134] fused to an octarginine cell-penetrating motif induces caspase-dependent cell death
Next, we investigated the effect of the Bax[106-134] peptide in cultured cells. One requirement for these experiments is the efficient delivery of the peptide, which should first cross the cell membrane to reach mitochondria and induce MOM permeabilization. To drive translocation across the plasma membrane, we used a modified version of the peptide with a poly-Arg sequence at the N-terminus (eight residues) connected to the natural sequence through a Gly linker (R8-Bax ). In addition, the peptide was derivatized with a fluorescent FITC label at its N-terminus to allow easy detection. A control peptide with similar design but with a  scrambled version of the Bax[106-134] natural sequence was also synthesized (R8-Bax [Scr]). This scrambled version will not be amphipathic in its expected membrane-bound -helix conformation. Dose-response analyses were undertaken by incubating HeLa cells with the peptides at different concentrations and different exposure times to monitor cellular uptake and cell viability. Fluorescence microscopy revealed uptake of both peptides, with strong green fluorescence observed in the cytoplasm as early as 1 hour after exogenous administration at a concentration of 10 M (Fig. 4A). However, although R8-Bax[Scr] penetrated efficiently into HeLa cells, this peptide did not produce any significant cellular toxicity, as assessed by lactate dehydrogenase (LDH) release (Fig. 4B). By contrast, R8-Bax  induced cell death in a dose-and timedependent manner with LC 50~1 5 M at 24 hours. As depicted in Fig. 4C, the R8-Bax[106-134]-induced LDH release was 560 Journal of Cell Science 124 (4) significantly diminished by incubation with zVAD.fmk, suggesting that cell death follows a caspase-dependent pathway. This observation was supported by Hoechst 33342 and propidium iodide double staining analysis (Fig. 4D), which confirmed that cell death was due to apoptosis. Moreover, we found that R8-Bax  induced similar levels of toxicity (analyzed by flow cytometry using Annexin-V binding) in Bax-Bak DKO MEFs and wild-type MEFs (Fig. 5), whereas the corresponding scrambled peptide R8-Bax[Scr] had no effect. Treatment with the caspase inhibitor zVAD.fmk blunted R8-Bax[106-134]-induced cytotoxicity in both cell types. These results confirm that the cell death observed is independent of both BAX and BAK [consistent with our previous data using isolated mitochondria (Guillemin et al., 2010)] and occurs by caspase-dependent apoptosis. The R8-Bax[Scr] version can also be considered as a control, to show that the cell death

Cytotoxic Bax[106-134] injected peritumorally shows antitumor activity in vivo
Fluorescence data obtained using a non-invasive live animal imaging technology indicated that upon peritumoral administration of a Cy5-labeled Bax[106-134] peptide, it was mainly taken up by the tumor tissue, which exhibited strong Cy5 fluorescence intensity, even 24 hours after injection (supplementary material Fig. S6A). Moreover, ex vivo fluorescence images of excised tumor tissues indicated minor accumulation to adjacent normal tissue (supplementary material Fig. S6B). These fluorescence data suggested that the Cy5-labeled peptide had a sustained localization within the tumor micro-environment following peritumoral injection, which prompted us to investigate the anti-tumor activity of Bax  using this mode of administration. To test the antitumor efficacy, the R8-Bax[106-134] version, or control samples (scrambled R8-Bax[Scr] peptide or buffer alone), were injected peritumorally five times a week for 2 weeks in mammary adenocarcinoma (TS/A-pc) tumor-bearing athymic nude mice. As shown in Fig. 6, after 2 weeks of peritumoral administration, the tumor volume was sharply reduced in the group treated with R8-Bax[106-134] compared with volumes in the control group. There was a statistically significant decrease in tumor size, tumor doubling time and growth. Moreover, the reduction in tumor size correlated with an increase in caspase-3-positive cells in tumor tissue extracts, indicating cell death (Fig. 6, inset).

Discussion
In addition to the intrinsic biotechnological potential of biodiversity, properties of molecules found in nature can be mimicked or 561 Cytotoxic effects of Bax peptides extended to produce novel bioactive substances. In this respect, the BH3-mimetic strategy represents a relevant example of the translation of molecular discoveries into potential clinical applications (Yip and Reed, 2008). Membrane-active peptides acting on the MOM (i.e. able to induce cytochrome c release and apoptosis) represent yet another type of promising, but so far unexploited, candidates in the cancer research field (Chen et al., 2001;Ellerby et al., 1999;Foillard et al., 2008;Law et al., 2006;Mai et al., 2001;Marks et al., 2005;Rege et al., 2007). Such a strategy has some parallel with the development of antibiotics from natural antimicrobial peptides (Marr et al., 2006), and in fact the use of these latter systems as anticancer drugs has already been proposed (Ellerby et al., 1999;Mader and Hoskin, 2006;Papo and Shai, 2005). As a singular advantage, and unlike the pro-apoptotic BH3-derived peptides or BH3-like compounds, mitochondrialmembrane-disrupting peptides will be active in cancer cells that do not express Bcl-2-like proteins, or in neo-angiogenic endothelial cells, irrespective of the Bcl-2 family status. Additionally, compared with other membranolytic peptides of different sources, active fragments designed from pore-forming Bcl-2 proteins can be considered to be naturally optimized by evolution to act on mitochondrial membranes (Guillemin et al., 2010). Here, we have shown that a peptide (Bax ) derived from the poreforming domain of pro-apoptotic Bax can cause mitochondrial damage and caspase-dependent apoptosis. This peptide appears to carry sufficient structural information to insert into the MOM, causing  m loss, membrane disruption and release of cytochrome c. Moreover, it produced (when fused to a polyarginine transduction motif) potent anticancer activity after peritumoral injection into tumor-bearing mice, presumably by inducing tumor cell apoptosis.
The molecular mechanism of pore formation and the structural properties of different peptide versions encompassing the sequence of 5 helix from Bax (which were very similar to Bax[106-134]) have been studied in several recent papers (Garcia-Saez et al.,  2007; Garcia-Saez et al., 2005;Garcia-Saez et al., 2006;Guillemin et al., 2010;Qian et al., 2008). These different versions of the 5 fragment of Bax exhibited strong -helical propensity in model lipid membranes and were shown to form lipidic pores of toroidal structure (Garcia-Saez et al., 2005;Qian et al., 2008). A similar mechanism of pore formation has been proposed for cationic helical antimicrobial peptides such as magainin (Ludtke et al., 1996). Although Bax has also been proposed to form pores of the proteolipidic toroidal type (Terrones et al., 2004), its mechanism of action is still largely unknown. A major difference between the activity of complete Bax, compared with that of Bax fragments, towards mitochondrial membranes is the existence of upstream (yet unclear) regulatory events, leading to activation of Bax via structural reorganization and membrane binding. Additionally, in the active membrane-bound state, the Bax protein forms a larger and more complex oligomer and pore than Bax-derived peptides do. Nevertheless, our results are consistent with the main findings reported in the literature for the action of Bax. First, the GFP fusion to complete Bax shows no specific localization to mitochondria (Fig. 2), weak disrupting activity towards this organelle ( Fig. 3 and supplementary material Fig. S2) and a weak induction of apoptosis ( Fig. 1 and supplementary material Fig. S1). This result is in accordance with the notion that monomeric Bax has to be activated by the BM3-only protein tBid before its targeting, oligomerization and poration of the MOM Lovell et al., 2008;Terrones et al., 2004). By contrast, Bax fragments display a clearly different behavior. Fusions of GFP with Bax fragments containing 5, 6 and/or 9, either alone or in the 56 or 5-9 constructions, all localize specifically and intrinsically to mitochondria (Fig. 2). By contrast, the fusions containing only the 5 and 6 fragments, as well as the 56 hairpin, exhibit a high mitochondria-disrupting activity ( Fig. 3 and supplementary material Fig. S2), which in turn correlates with strong induction of cell death ( Fig. 1 and supplementary material  Fig. S1). These latter and most remarkable observations are consistent with the existence in Bax of several independent mitochondrial targeting sequences, located in helices 5, 6 and 9 (George et al., 2007;George et al., 2010;Schinzel et al., 2004;Valentijn et al., 2008). Thus, our results show that the naked versions of these fragments have a natural tendency for specific binding to the MOM, and in the cases of 5 and 6 for high membrane-poration activity, with no need for complex structural reorganization, because they are intrinsically active. Within the context of larger domains, intra-and inter-molecular interactions between different helices of Bax (George et al., 2007;George et al., 2010;Suzuki et al., 2000) might impair their interaction with the membrane. This phenomenon, which is at the origin of the regulation of the complete Bax protein, might also be among the reasons why the GFP-Bax-5-9 construct was not as potent as Bax-5, Bax-6 and Bax-56 in causing cell death ( Fig. 1 and  supplementary material Fig. S1).
In conclusion, although Bax-derived fragments can obviously not mimic the elaborate behaviour of the full-length protein, these peptides represent in practice minimal versions of Bax, which is evolutionary designed to target, bind and porate mitochondria. Bax  shows a specificity and efficacy for MOM disruption, which is better than that of the cationic peptide (KLAKLAK) 2 , in agreement with the low potency previously reported for this molecule (Borgne-Sanchez et al., 2007;Ellerby et al., 1999). Additionally, the lack of any regulatory capacity in minimal peptide versions with respect to full-length Bax renders these molecules intrinsically and autonomously active, which might be used advantageously as a basis for antitumor therapy. Thus, we propose to exploit membrane-active segments from natural Bcl-2like templates (such as helices 5 and 6 of Bax) to develop a new generation of mitochondria-targeted cytotoxic agents (named 'poropeptides'). To be applicable in cancer therapy, poropeptides should eliminate tumor cells without being harmful to normal cells. Indeed, although such biologically active peptides can be developed into drugs, the design of suitable delivery systems for site-specific targeting to tumors remains the most challenging task. Future work will therefore focus on endowing therapeutic poropeptides with the ability to reach tumor cells and leave normal cells unharmed.  (Garcia-Saez et al., 2005) in an Applied Biosystems ABI 433A Peptide synthesizer (Foster City, CA) using Fmoc chemistry and Tentagel S-RAM resin (Rapp Polymere, Tübingen, Germany; 0.24 mEq/g substitution) as a solid support. Peptides were purified using a C18 semi-preparative reversed-phase column (Merck, Darmstadt, Germany) by HPLC, to a >95% purity, and their identity was confirmed by mass spectrometry. Peptide concentrations were determined from UV spectra using a Jasco spectrophotometer (Jasco, Tokyo, Japan). The Cy5-Bax[106-134] peptide was synthesized using solid-phase peptide synthesis (SPPS), purified by HPLC and characterized by ESMS at the chemistry platform 'NanoBio campus' (Grenoble, France). R8 (arginine-8) peptides had an amide group at their C-terminus. The amino acid sequences of the peptides are shown in Table 1.

Cell culture
SK-MEL-28 human melanoma cells and HeLa cells were cultured at 37°C and 5% CO 2 in MEM supplemented with 10% FBS, 1% penicillin-streptomycin and 1% non-essential amino acids. HT1080 cells, HEK293T cells, MEFs and DKO MEFs were cultured in DMEM supplemented with 10% FBS and 1% penicillinstreptomycin. For transient transfection, cells were plated at a density of 10 5 cells per 35 mm plate) and allowed to grow for 24 hours before transfection with plasmids using the Lipofectamine2000 (Invitrogen) according to the manufacturer's recommendation. For each transfection, 3 g plasmid DNA was used. Caspase inhibitor zVAD.fmk was purchased from Bachem. TS/A-pc mice mammary carcinoma cells were cultured in RPMI 1640 supplemented with 1% glutamine, 10% FBS, 50 U/ml penicillin, and 50 g/ml streptomycin at 37°C in a humidified 95% air, 5% CO 2 atmosphere. These cells are integrin vb3 positive (Klepfish et al., 1993;Sancey et al., 2007).

Molecular cloning
The oligonucleotides (Sigma-Proligo) that were used to prepare the different constructions are indicated in supplementary material Table S1. All the constructs were subcloned into pGEM-T Easy (Promega) and then subsequently subcloned into XhoI and KpnI sites of pEGFP-C1. The sequence of each construct was verified by automated sequencing (GEXbyWeb).

Measurement of cell death and viability
Hoechst 33342 and PI labeling of cells to detect apoptotic and necrotic cell death were performed as described previously (Dive et al., 1992). Hoechst 33342 and PI were from Molecular Probes (Invitrogen). LDH cytotoxicity assay was performed according to the manufacturer's protocol (LDH Cytotoxicity Assay Kit II, Biovision Research Products, CA); the colorimetric assay quantifies LDH activity released from the cytosol of damaged cells into the supernatant and thus serves to quantify cell death. Cytotoxicity assays were performed in triplicates in each of two or three independent experiments. Cell death was quantified by Annexin-V-Cy3 (BioVision) staining according to manufacturer's protocols, followed by flow cytometric analysis using a FACScan (Becton Dickinson). Data were processed using CellQuest Pro (version 4.0) software.

Mitochondrial assays
In vitro assessment of mitochondrial parameters (swelling and  m loss) was performed on liver mitochondria as previously described (Jacotot et al., 2006). Mitochondrial membrane potential was measured using the fluorescent dye Mito-Tracker Red (Molecular Probes), which emits fluorescence in cells with an intact  m . Transfected cells were incubated with Mito-Tracker Red (50 nM for 2 hours at 37°C). Cells were observed under a fluorescence microscope and the percentage of green cells that were Mito-Tracker positive was determined (~100 cells in each experiment). For flow cytometry analysis, HT1080 cells were washed twice with serum-free medium and then resuspended in PBS. Flow cytometric analysis was performed using a LSR II (Becton Dickinson) and data were processed using FACSDiva (version 6.1.2) software.

Confocal microscopy analysis
Cells were fixed in 4% paraformaldehyde, permeabilized in 0.1% Triton X-100 for 3 minutes, and treated with TO-PRO-3 iodide (final 2 M, Molecular Probes) before mounting in a drop of anti-bleaching medium. Confocal analysis was performed on a Zeiss confocal microscope (LSM510) (LePecq, France) with a plan apochromat 63ϫ 1.4 NA oil-immersion objective. Images were collected under identical nonsaturated conditions after multiple scans (~eight sections per cell).

In vivo experiments
All animal experiments were performed in agreement with the EEC guidelines and the Principles of laboratory animal care (NIH publication 14, No. 86-23, revised 1985). The experimental protocol was submitted to ethical evaluation and received accreditation number 323 (investigator authorization number 38-09-22 to S.L.).

Tumor regression assays
Mouse mammary TS/A-pc cells were harvested from culture, and 10 6 cells in sterile PBS were injected subcutaneously into the flank of 30 female Balb/c mice. Three days after injection, mice were randomized into three experimental groups (nine mice per group). Group 1 (control mice) received vehicle (PBS), group 2 received Bax , and group 3 received Bax-Scr.
100 g peptide per mouse (100 l/mouse) was administered peritumorally, five times a week for 2 weeks. Tumor growth was assessed by measuring tumor size in two dimensions using a Vernier caliper each day after tumor size reached 10 mm 3 or larger (from day 10). Tumor volume was calculated as follows: (p/6) ϫ a ϫ b 2 , where a and b are the largest and smallest diameters, respectively (Kjonniksen et al., 1989;Olea et al., 1992). Results are expressed as mean ± s.e.m. None of the mice had developed necrotic tumors or tumors ≥1.5 cm in diameter. On day 14, all mice were sacrificed to prevent lung metastasis, especially in groups 1 and 3. The tumor doubling time (TDT) was calculated as (TdЈ-Td)ln2/[ln(VdЈ-Vd)], where T is time at days d and dЈ, and V is the corresponding tumor volume.

Statistical analysis
For the in vivo studies, results were analyzed by t-test for unmatched groups (Statview software, SAS Institute, Inc.); P<0.05 was considered statistically significant.