Conversion of cell-survival activity of Akt into apoptotic death of cancer cells by two mutations on the BIM BH3 domain

Survival and proliferation of cancer cells are often associated with hyperactivity of the serine/threonine kinase, Akt. Herein, we show that prosurvival activity of Akt can be converted into prodeath activity by embedding an Akt recognition sequence in the apoptogenic BH3 domain of human BIM. The recognition sequence was created by introducing two mutations, I155R and E158S, into the core region of the BIM BH3 domain. Although a 21-mer BIM BH3 peptide containing these two mutations bound weakly to BCL-XL and BCL-2, this peptide with phosphorylation of Ser158 bound to these proteins with a dissociation constant of <10 nM. The crystal structure of the phosphorylated peptide bound to BCL-XL revealed that the phospho-Ser158 makes favorable interactions with two BCL-XL residues, which cannot be formed with unphosphorylated Ser158. Remarkably, the designed peptide showed a cytotoxic effect on PTEN-null PC3 tumor cells whose Akt activity is aberrantly high. The cell-killing activity disappeared when the cellular Akt activity was lowered by ectopic PTEN expression. Thus, these results lay a foundation for developing a peptide or protein agent that is dormant in normal cells but is transformed into a potent apoptogenic molecule upon phosphorylation by hyperactivity of Akt in cancer cells.

The interplay between the BCL-2 family proteins regulates mitochondrion-mediated apoptotic cell death. 1,2 The BCL-2 family proteins are characterized by having at least one BCL-2 homology (BH) domain, and they are classified into three distinct subgroups based on their functional and structural features. One subgroup consists of BAX and BAK, which contain the BH1-BH4 domains and mediate apoptosis by increasing the permeability of the mitochondrial outer membrane (MOM) and thus leading to the release of the apoptogenic factors, such as cytochrome c and Smac/Diablo. [3][4][5][6] Another subgroup is composed of antiapoptotic proteins, BCL-2, BCL-X L , BCl-w, MCL-1, A1 and BCL-B, which contain the BH1-BH4 domains that are arranged to form an extended hydrophobic groove known as the BH3-binding groove. 7 The remaining subgroup is composed of a diverse set of proteins that are unrelated to each other except for the possession of the BH3 domain. 7 These BH3-only proteins sense and convey apoptotic cell death signals, ultimately leading to the activation of BAX and BAK. 8,9 The antiapoptotic BCL-2 subfamily proteins bind the BH3 domain of BAX/BAK and of the BH3-only proteins through their BH3-binding groove. [10][11][12][13][14][15] Biochemical studies have discovered that a number of the BH3-only proteins termed 'activators', such as BID and BIM, bind directly to BAX and induce its activation, whereas other BH3-only proteins termed 'sensitizers' induce apoptosis by releasing the activators sequestered by the antiapoptotic proteins. 5,16,17 A recent crystallographic study revealed that the BID BH3 peptide binds to the canonical BH3-binding groove of BAX and induces a pronounced conformational change that exposes the BH3 domain of BAX. 18 The activated BAX oligomerizes to induce the permeabilization of the MOM. 6 The antiapoptotic BCL-2 proteins were suggested to sequester the BH3 domains of both BAX and the activator BH3-only proteins to prevent the BAX oligomerization. 18 Apoptosis is attenuated in cancer cells because of the abundance of antiapoptotic BCL-2 proteins and/or prevention of apoptosis induction. Anticancer BH3 peptides have been developed, especially those derived from BIM, which interacts with all of the antiapoptotic proteins with extremely high affinity. 15,19 These BH3 peptides exhibit a broad and multimodal targeting of the BCL-2 family proteins. [20][21][22] Promising small molecular anticancer compounds have also been developed that mimic the BH3 peptides and bind to the surface groove of the antiapoptotic proteins. 23 ABT-737 and ABT-263 selectively bind to and lower the amounts of the functional BCL-2, BCL-X L and BCL-w proteins to induce the apoptotic death of tumor cells that depend especially on the overexpression of the three proteins. 24,25 The BH3 peptides and the BH3 mimetics both bear an intrinsic shortcoming in that they inhibit the BCL-2 family proteins not only in cancer cells but also in normal cells as they cannot distinguish cancerous from normal cells.
One of the hallmarks of many cancer and tumor cells is the hyperactivation of the serine/threonine (Ser/Thr) protein kinase Akt, which is a key signaling molecule in the cellular survival pathway. 26 In many types of cancers, including glioma, prostate cancer and breast cancer, Akt is required to maintain a proliferative state and for progression into a more malignant state in conjunction with genetic mutations. [26][27][28] We set out to develop a molecule that can respond to the hyperactivity of Akt and can lead to the death of cancer cells. Herein, we describe the embedment of the Akt recognition sequence into the BIM BH3 peptide and the cancer cellspecific apoptogenic property of the resulting BIM BH3 peptide variant characterized by X-ray crystallography, calorimetry and cell-based biochemistry.

Results
Design of a BIM BH3 peptide with an Akt recognition sequence. We chose the BIM BH3 as the template sequence for mutagenesis. According to the crystal structure of the mouse BIM BH3 domain bound to BCL-X L , 21 residues of BIM form the core region of the BH3 domain that spans the surface groove of BCL-X L . 14 The 21 residues correspond to 145-EIWIAQELRRIGDEFNAYYAR-165 of human BIM, which is referred to as BH3 BIM (Figure 1). To design a BH3 BIM peptide variant that can be phosphorylated by Akt, we noted Glu158 in BH3 BIM . Glu158 is not a strictly conserved residue, but is involved in a polar interaction with Tyr101, a conserved residue of BCL-X L in the crystal structure, 14 suggesting that this residue contributes to the binding affinity of BH3 BIM . Thus, it was expected that a BH3 BIM variant containing serine in place of Glu158 would exhibit reduced binding affinity for the antiapoptotic BCL-2 family proteins, but phosphorylation of Ser158 could restore the binding affinity since phosphorylated serine often serves as a mimic of glutamate residue. Importantly, this variant peptide turns into an Akt recognition sequence by a second mutation of Ile155Arg, because the doubly mutated sequence has a RRR 155 GDS 158 F stretch that conforms to the Akt recognition sequence (RxRxxT/SF; T/S, phosphorylation site). 29 Ile155 of human BIM, which is a strictly conserved residue, is involved in extensive hydrophobic interactions with BCL-X L in the complex between mouse BCL-X L and a BIM BH3 peptide. 14 Nonetheless, a modeling experiment with this BH3 BIM peptide variant, referred to as BH3 BIM (I155R/E158S), indicated that arginine in place of Ile155 is sterically compatible and does not completely abrogate the binding interaction. In addition, its guanidinium group seemed to form an intramolecular salt bridge with Glu151 in this peptide.
In particular, phosphorylated Ser158, not Ser158, of BIM appeared to form a polar interaction with Tyr101 of BCL-X L . Furthermore, the phosphate group on Ser158 was within the intramolecular ionic-bonding distance from the guanidinium group of Arg154. Based on these expectations, we hypothesized that unphosphorylated BH3 BIM (I155R/E158S) would suffer from reduction in the binding affinity compared with the wild-type version, but phosphorylation of the peptide at Ser158 could provide some compensation by directly interacting with BCL-X L . Thus, it seemed plausible that a full-length BIM or a BH3 region bearing the I155R and E158S substitutions could be phosphorylated in cancer cells with elevated Akt activity and then be converted into a potent inhibitor against antiapoptotic BCL-2 proteins, while they are not potentiated in normal cells due to low or controlled Akt activity.  Only GST-fused BH3 BIM (I155R/E158S) was phosphorylated BH3 BIM (I155R/E158S) is phosphorylated by Akt and potently binds to antiapoptotic BCL-2 proteins. To examine whether the designed sequence is phosphorylated by Akt as we intended, we carried out an in vitro Akt activity assay by using GST-tagged BH3 BIM (I155R/E158S) as the substrate in the presence of [γ-32 P]ATP. GST-tagged BH3 BIM (I155R/E158S) was efficiently phosphorylated, while GST and GST-tagged BH3 BIM (I155R/E158A) employed as To test if phosphorylated BH3 BIM (I155R/E158S) binds to the BCL-2 family proteins more tightly than its unphosphorylated version, we produced recombinant BCL-2 and BCL-X L proteins, and also prepared two 21-mer synthetic peptides: BH3 BIM (I155R/E158S) and phosphorylated BH3 BIM (I155R/ E158S) at Ser158, which is referred to as p-BH3 BIM (I155R/ E158S) (Figure 1). Quantification of the binding affinities by isothermal titration calorimetry (ITC) showed that p-BH3 BIM (I155R/E158S) interacted potently with BCL-2 and BCL-X L with K D values of 8.55 and 9.90 nM, respectively (Figures 3a and b), similar to that of a longer 36-mer BIM BH3 peptide (K D of~7 nM). 15 In contrast, the unphosphorylated BH3 BIM (I155R/E158S) peptide exhibited much lower affinities for the two proteins (K D of 192 and 189 nM, respectively) (Figures 3c and d). Thus, phosphorylated Ser158 appeared to replace the role of Glu158 in the BH3 sequence. Furthermore, the substitution of the conserved hydrophobic Ile155 seemed to be tolerated in the binding reaction, which is intriguing given the observation that an alanine substitution of the corresponding Ile81 residue in a BAK BH3 peptide resulted in a significant reduction of the binding affinity for BCL-X L (K D value changed from 0.34 to 17 μM). 30 The measured binding affinities of p-BH3 BIM (I155R/E158S) for BCL-2 or BCL-X L are comparable to or higher than those of 36-mer BH3 peptides derived from BAX and BAK (K D of 8.1-255 nM), 15 suggesting that the phosphorylated BH3 BIM (I155R/E158S) sequence, but not the unphosphorylated sequence, could displace these two apoptosis mediators from the antiapoptotic BCL-2 proteins and potentiate cell death. We indirectly tested this possibility by employing PUMA and BCL-2, whose intermolecular interaction is tighter than that between BAX (or BAK) and BCL-2, 15 to obviate complications in using full-length BAX or BAK. Recombinant human PUMA fused to GST (GST-PUMA) was produced, and HEK293 cells were prepared to transiently overexpress BCL-2 and one of the three forms of Akt: wild-type, constitutively active or kinase-dead form. Each cell lysate was incubated with BH3 BIM (I155R/E158S) and GST-PUMA. This peptide added to the cell lysate containing the wild-type and constitutively active form of Akt abolished the binding between BCL-2 and GST-PUMA, whereas the same peptide added to the cell lysate containing the kinase-dead form of Akt did not interfere with the binding interaction ( Figure 3e). The kinase activity of the Akt proteins were confirmed by examining the phosphorylation of GSK3β, a cellular substrate of Akt ( Figure 3e). Collectively, these results indicated that Akt-phosphorylated BH3 BIM (I155R/ E158S), and the phosphorylated peptide could compete with PUMA for binding to BCL-2, whereas the unphosphorylated peptide could not.
Structure of BCL-X L in a complex with p-BH3 BIM (I155R/E158S). To understand the structural basis for the critical role of the Ser158 phosphorylation, we next determined the crystal structure of BCL-X L bound to p-BH3 BIM (I155R/E158S) at a 2.1-Å resolution ( Table 1). The peptide binds to the BH3-binding groove of BCL-X L by forming an amphipathic α-helix, as observed in all of the reported structures of the BH3 peptides bound to the antiapoptotic BCL-2 family proteins 14,15,31 (Figure 4a). Since the sequence of the p-BH3 BIM (I155R/E158S) peptide is highly similar to that of the BH3 domain of mouse BIM, the presented structure can be directly compared with the structure of BCL-X L bound to the BIM L BH3 domain. 14 The p-BH3 BIM (I155R/E158S) peptide contains four of the five consensus residues that are highly conserved in the BH3 domains of the proapoptotic proteins and known to have critical roles in interacting with the antiapoptotic BCL-2 proteins. The four residues (Ile148, Leu152, Asp157 and Phe159) in the peptide are involved in the intermolecular hydrophobic or hydrophilic interactions with BCL-X L , similar to the corresponding residues of the BIM L BH3 bound to BCL-X L (not shown).
The remaining consensus residue, which corresponds to Ile155 of human BIM (Ile97 of mouse BIM), is replaced by arginine in p-BH3 BIM (I155R/E158S). While Ile97 of mouse BIM is involved in hydrophobic interactions with Phe97 and Tyr101 of BCL-X L , Arg155 of p-BH3 BIM (I155R/E158S) forms a hydrophobic interaction with Tyr101 through its hydrocarbon portion, and makes an intramolecular salt bridge with Glu151 of the peptide through its guanidinium group (Figure 4b). The salt bridge is likely to increase the helical propensity of p-BH3 BIM (I155R/E158S) in solution and to contribute to the binding affinity of this phosphorylated peptide.
where |F o | and |F c | are the observed and calculated structure factor amplitudes, respectively. R free was calculated with 5% of the data Notably, the phosphate group attached to Ser158 has a dual conformation (Figures 4c right and d). In one conformation, it interacts with the guanidinium group of Arg100 and the hydroxyl group of Tyr101 of BCL-X L , where Arg100 is distantly positioned from the carboxyl group of Glu158 (5.9Å) in the structure of the BCL-X L -BIM L BH3 complex 14 (Figure 4c left).
In the other conformation, however, the phosphate group is removed from the guanidinium group (4.6Å), and instead interacts with His177 of a symmetry-related BCL-X L molecule (3.1Å) (Figure 4d). Therefore, the observed conformation of phosphorylated Ser158 in the crystalline state would be different from the conformation of this residue in solution.
Structure of BCL-X L in complex with the p-BH3 BIM (R154S/ I155R/E158S) peptide. To remove the crystal packing interaction of phosphorylated Ser158, we prepared another peptide, referred to as p-BH3 BIM (R154S/I155R/E158S), which contains an additional substitution of R154S in addition to the I155R and E158S substitutions ( Figure 1). Arg154 was chosen for the substitution, as this residue in p-BH3 BIM (I155R/E158S) was engaged in a crystal packing interaction with a symmetry mate (Supplementary Figure S1). In an ITC analysis, p-BH3 BIM (R154S/I155R/E158S) exhibited slightly lower affinity for BCL-X L (K D =~40 nM) compared with p-BH3 BIM (I155R/E158S) (Supplementary Figure S2). Subsequently, we obtained crystals of BCL-X L bound to this peptide in a different crystal form, and determined its structure at a 1.7Å resolution (Figure 5a and Table 1). In this crystal form, the asymmetric unit contained two copies of the complexes that exhibit nearly identical conformations (rmsd = 0.063Å between 143 Cα atoms; Supplementary Figure S3). Importantly, phosphorylated Ser158 in this peptide was not involved in the crystal packing interactions. The intermolecular interactions between BCL-X L and the peptide are mostly similar to those observed in the BCL-X L -p-BH3 BIM (I155R/E158S) structure (Figure 5b). In particular, phosphorylated Ser158 is in a single conformation and makes intermolecular interactions with Tyr101 and Arg100 of BCL-X L , which is likely to recapitulate the interactions in solution (Figure 5c). Together, the presented structures explain the significant enhancement of the binding affinity upon phosphorylation of Ser158 in the designed BH3 BIM (I155R/E158S) peptide.
Akt-dependent cytotoxic activity of the designed BH3 BIM peptide. Next, we tested whether the BH3 BIM (I155R/E158S) peptide exhibits cytotoxic activity. For intracellular delivery, BH3 BIM (I155R/E158S) was fused to the C-terminus of the cell penetration peptide (CPP) derived from HIV Tat (Figure 1). PC3 and HCT116 cells were treated with the CPP-BH3 BIM (I155R/E158S) peptide, and the MTT assay was performed. PC3 cells, which are derived from prostate cancer cells, exhibit a significantly elevated Akt activity due to the loss of PTEN, a negative regulator of Akt, 32,33 whereas the colon cancer-derived HCT116 cells exhibit a normal level of Akt activity. Seventy-two hours after the treatment with the fusion peptide, a strong cytotoxic activity was observed in the PC3 cells, but not in the HCT116 cells (Figure 6a). Immunofluorescence analysis suggested that the PC3 cells underwent apoptotic cell death (Figure 6b). Cytochrome c diffused in the cytoplasm and gradually accumulated in the nucleus, which is known to occur during apoptosis. 34 In contrast, HCT116 retained the puncta staining pattern under the same treatment (Figure 6b). We then measured the activity of Akt by examining the phosphorylation state of itself and its substrate GSK3β. The CPP-BH3 BIM (I155R/E158S) peptide reduced the phosphorylation of GSK3β, but not Akt itself (Figure 6c), indicating that this peptide could have acted as a substrate of Akt. To elaborate this observation, we examined the effect of the fusion peptide in three different human lung cancer cell lines. The PTEN-silenced H1299 cell lines exhibited sensitivity to the peptide, and the A549 cell line, which possesses a K-Ras mutation, showed a moderate response (Figure 6d). In contrast, the H23 cell line, which is derived from lung cancer cells with wild-type K-Ras and normal level of Akt activity, did not respond to this peptide, suggesting that the CPP-BH3 BIM (I155R/E158S)-induced cell death might depend on the Akt activity. We then assessed the effect of the Akt activity by ectopically expressing wild-type Akt (Akt-WT), constitutively active Akt (myristoylated Akt; Akt-Myr) or kinase-dead-Akt (Akt-KD). Both Akt-WT and Akt-Myr   (Figures 6f and p-GSK3β). To know whether CPP-BH3 BIM (I155R/E158S)-induced cell death might depend on BAX, the HCT116 BAX +/ − and the HCT116 BAX − / − cell lines were transfected with an expression vector encoding AKT-Myr and the cells were treated with the peptide. Notably, the BAX +/ − cells, but not the BAX − / − cells, were sensitized to the peptide (Figure 6g), indicating that the CPP-BH3 BIM (I155R/E158S) peptide induced intrinsic apoptotic cell death via BAX. The survival of the BAX − / − cells despite the CPP-BH3 BIM (I155R/E158S) treatment could be explained by an observation that the designed peptide only weakly binds to MCL-1 (not shown), which is a major inhibitor of BAK. 35 Presumably, both MCL-1 and BAK are expressed in these cells. 36 PTEN-dependent cytotoxic effect of the designed BH3 BIM . For efficient cytoplasmic delivery of the proteins, we applied a novel technique of using gold nanoparticles (AuNPs) coated with anti-GST DNA aptamers. 37 To apply this technique, GST-fused the BH3 BIM peptides were produced and incubated with AuNPs coated with anti-GST DNA aptamer. We observed that GST-BH3 BIM (I155R/E158S) efficiently killed PC3 cells (Figure 7a). In contrast, the other variants, GST-BH3 BIM (E158S) and GST-BH3 BIM (I155R/ E158A), did not exhibit a cell-killing effect (Figure 7a), presumably because these two peptides lack either the Akt recognition sequence or a site for phosphorylation by Akt. Intriguingly, when the PTEN gene was introduced into PC3 cells by transfection, the cell-killing effect of GST-BH3 BIM (I155R/E158S) was significantly reduced (Figure 7a). We confirmed that GST-BH3 BIM -coated AuNPs were effectively incorporated into the cells (Figure 7b). To confirm the cytotoxic effect of each protein, we performed the trypan blue dye exclusion assay and counted viable cells. Consistent with the MTT assay described above, the AuNPs coated with GST-BH3 BIM (I155R/E158S) clearly reduced the viability of PC3 cells (Figure 7c). In a control experiment, BH3 BIM (I155R/E158S) and its variants did not alter the expression of proapoptotic proteins such as BAX or PUMA (Supplementary Figure S4).
To examine the effect of PTEN that is frequently deleted or mutated in various kinds of cancers, we ectopically expressed PTEN in PC3 cells and examined cytochrome c release. In immunostaining and cell fractionation analyses, cytochrome c release was blocked in the cells transfected with a PTENexpressing vector, in contrast with the cells transfected with an empty vector (Figures 7d and e; Supplementary Figure S5). In fact, production of cleaved PARP, a well-known caspase 3 substrate, was abolished by PTEN transfection (Figure 7f). These results imply that BH3 BIM (I155R/E158S) might be used as apoptosis inducer in PTEN-mutated cancer cells. In contrast with PC3 cells, the viability of HEK293 cells was not affected by the GST-BH3 BIM -coated AuNPs, which might be ascribed to the normal activity of Akt in HEK293 cells (Figure 7g). Indeed, the AuNP-protein and the CPP-peptide slightly increased cell viability without a statistical significance (Supplementary Figure S6), indicating that the designed peptide is not harmful to untransformed cells. Together, these results demonstrate that it is possible to convert the hyperactivity of Akt in cancer cells into a death signal.

Discussion
Hyperactivation of Akt is one of the most common molecular perturbations, frequently found in many types of cancers, including glioma, non-small cell lung cancer (NSCLC), ovarian cancer and prostate cancer, 28,[38][39][40] indicating that Akt is an attractive target for cancer therapy. Several Akt inhibitors have been developed and clinically trialed. 39 However, these inhibitors seem to inevitably accompany side effects, because Akt is also important for the survival of normal cells. We described a new concept and experimental support of converting a cell-survival signal into a cell-killing apoptotic signal with an aim of killing cancer cells without affecting normal cells. These two paradoxical pathways are connected by a peptide designed to harbor the Akt recognition sequence on the potently apoptotic BIM BH3 domain. The peptide was phosphorylated by Akt as we intended, and the phosphorylated peptide interfered with the binding between BCL-2 and the PUMA BH3 domain, most likely because of its potent binding affinity for BCL-2, whereas the weakly interacting unphosphorylated peptide failed to do so. The crystal structures demonstrated how phosphorylation of Ser158 in the peptide significantly enhances the binding affinity of the peptide for BCL-X L . Remarkably, the peptide killed cancer cells exhibiting uncontrolled Akt activity, while cells bearing the normal Akt activity were unaffected.
The mode of action of the designed peptide is analogous to that of the tumor suppressor, ARF. The ARF protein leads to p53-dependent cell cycle arrest and apoptosis in response to sustained mitogenic signaling from Myc and Ras. 41 The arf gene is deleted or silenced in numerous cancers, highlighting the important role of ARF as a tumor suppressor. 42 Like ARF, the designed peptide might block tumor progression. We note that the RxRxxS/T motif can also be phosphorylated by RSK and S6K, which promote cell proliferation and growth. 43 In certain cancer cells, the designed peptide is likely to be phosphorylated by all of the three kinases, which trigger the core Akt-RSK-S6K signaling pathways that are activated at the downstream of protooncogenic receptor protein kinases. 43 The techniques for the intracellular delivery of proteins and peptides are under active development. Combined with effective techniques, the peptide or protein versions of the designed BIM BH3 variant might be developed into an anticancer drug with greatly reduced side effects. Other deregulated kinases leading to uncontrolled cell growth may also be targeted by embedding different recognition sequences in an apoptogenic BH3 sequence. In conclusion, our results represent a conceptual advance in designing peptide or protein-based cancer therapeutics that are potentiated in cancer cells, but remain dormant in normal cells.

Materials and Methods
Plasmids. The procedures for the expression of mouse BCL-X L was described previously. 15 In brief, a DNA fragment coding for mouse BCL-X L (residues 1-44 and 85-196) was cloned into pProEX-HTa (Invitrogen, Carlsbad, CA, USA). To express the BH3 BIM peptide and its variants as a GST-fusion protein in Escherichia coli, chemically synthesized DNA fragments encoding human BIM BH3 (residues 145-165 with appropriate mutations; BIONEER, Daejeon, South Korea) were inserted into the pGEX-TEV vector (a modified vector derived from pGEX-4T1) using the EcoRI and HindIII sites. I153R and E156S mutations were introduced into the vector by site-directed mutagenesis methods based on the overlapping PCR method. DNA fragments encoding full-length human PUMA was amplified from a human cDNA library (Life Technologies, Carlsbad, CA, USA) by a standard PCR protocol, and inserted into the pGEX-TEV vector using the EcoRI and HindIII sites. Expression and purification of recombinant proteins. Mouse BCL-X L was purified as reported. 15 In brief, the protein was expressed in the E. coli BL21 (DE3) strain (Novagen, Darmstadt, Germany) at 18°C overnight and purified by using a Ni-NTA column (Qiagen, Venlo, Limburg, Nederland) and a HiTrap Q anion exchange column (GE Healthcare, Little Chalfont, Buckinghamshire, UK). The N-terminal hexahistidine tag was cleaved with TEV protease after a Ni-NTA column purification step. GST-fused BH3 BIM variants and mouse PUMA were expressed in E. coli and purified with the GST-binding agarose resin (Qiagen) equilibrated with 20-mM Tris buffer (pH 8.0) containing 150-mM NaCl and 2-mM β-mercaptoethanol. HA-tagged mouse BCL-2 was expressed from the pcDNA5/FRT/TO vector in HEK293 cells.
Crystallization and structural determination. Purified BCL-X L (4 mg/ml) in a buffer consisting of 20-mM Tris-HCl (pH 8.0), 50-mM NaCl and 1-mM dithiothreitol was mixed with p-BH3 BIM (I155R/E158S) at a molar ratio of 1 : 3. The crystals of the resulting complex were obtained by the hanging-drop vapor diffusion method at 24°C by mixing and equilibrating 1 μl of the protein solution, 1 μl of a precipitant solution containing 0.1-M sodium acetate (pH 4.8) and 3.0-M sodium chloride and 0.5 μl of 20% v/v 1,1,1,3,3,3-hexafluoro-2-propanol. Before data collection, the crystals were immersed briefly in a cryoprotectant solution, which was the reservoir solution containing additional 15% glycerol. A diffraction data set at 2.1Å was collected on R-Axis IV++ (Rigaku, Tokyo, Japan) at the Korea Research Institute of Bioscience and Biotechnology, Korea. The data set was processed using the program suit HKL2000. 44 The structure was determined by the molecular replacement method with the CCP4 version of MolRep 45 using the structure of the BAD BH3 peptide-bound BCL-X L 46 as a search model. Subsequently, model building and refinement were carried out using the COOT 47 and CNS 48 programs. The final model did not include residues 32-44 of BCL-X L and the last two C-terminal residues of p-BH3 BIM (I155R/E158S) whose electron densities were not observed or very weak.
For crystallization of BCL-X L bound to p-BH3 BIM (R154S/I155R/E158S), the purified BCL-X L sample (18.5 mg/ml) was mixed with p-BH3 BIM (R154S/I155R/ E158S) at a molar ratio of 1 : 2. The crystals of the resulting complex grew in a precipitant solution containing 0.1-M sodium acetate (pH 4.6), 0.01-M calcium chloride and 15% v/v 2-methyl-2,4-pentanediol. Before data collection, the crystals were immersed briefly in a cryoprotectant solution containing additional 20% glycerol. A diffraction data set at 1.65Å was collected at the 5C beamline of the Pohang Accelerator Laboratory (PAL, Pohang, South Korea) using a QUANTUM 270 CCD detector (ADSC). The data set was processed using the program suit HKL2000. 44 The structure was determined by the molecular replacement method with the CCP4 version of MolRep 45 using the structure of the p-BH3 BIM (I155R/E158S) peptidebound BCL-X L as a search model. Subsequently, model building and refinement were carried out using the programs COOT 47 and PHENIX. 49 Figures were generated using PyMOL. 50 The final model did not include residues 1 and 31-43 of BCL-X L , which are disordered in the crystal. Crystallographic data statistics are summarized in Table 1.
Isothermal titration calorimetry. All measurements were carried out at 25°C on a microcalorimetry system iTC200 (GE Healthcare). The BCL-2 and BCL-X L samples were dialyzed against the solution containing 20 mM Tris-HCl (pH 7.5) and 350 mM NaCl. Peptides were titrated into the proteins. Dilution enthalpies were measured in separate experiments (titrant into buffer) and subtracted from the enthalpies of the binding between the protein and the titrant. Data were analyzed using the Origin software (OriginLab Corp., Northampton, MA, USA). We did three independent experiments, which gave the similar results.
Measurement of cell viability. PC3 and HEK293 cells were incubated with the GST-BH3 BIM (I155R/E158S) protein conjugated to AuNP-anti-GST DNA aptamer composite 37 after transfection with a PTEN expression vector (pCMV Flag WT-PTEN; addgene plasmid #22231). To examine the effect of the CPP-fused BH3 BIM peptide on cell viability, PC3 and HCT116 cells were incubated with the peptide for 48 and 72 h. To examine cell viability, these cells were incubated with 0.5 mg/ml of MTT solution (Calbiochem, Darmstadt, Germany) for 4 h at 37°C. After removing excess solution and washing with PBS, the precipitated materials were dissolved in 200 ml of dimethyl sulfoxide and quantified by measuring the absorbance at 540 nm. Each experiment was performed by at least three different researchers.
Immunofluorescence staining. Cells were seeded on a cover glass and incubated with the CCP-fused BH3 BIM (I155R/E158S) peptide. After fixing with methanol for 30 min, cells were incubated with blocking buffer (0.2% (v/v) normal human antibodies in PBS) for 1 h. After extensive washing with PBS, they were incubated with 1% (v/v) anti-cytochrome c mouse antibody (BD, cat# 556432) in blocking buffer for 4 h and subsequently with 0.2% (v/v) FITC-conjugated antimouse Ab in PBS containing anti-human antibodies for 2 h. The nucleus was stained by DAPI. After washing with PBS, cover glasses were mounted with Vectamount (Vector Laboratories, Burlingame, CA, USA). The immunofluorescence signal was detected by fluorescence microscopy (Zeiss, Oberkochen, Germany).
Statistical analysis. Statistical significance was obtained by student's t-test.

Conflict of Interest
The authors declare no conflict of interests.