Evaluation of the BH3-only Protein Puma as a Direct Bak Activator*

Background: How the DNA damage-induced proapoptotic protein Puma triggers apoptosis has been unclear. Results: The Puma BH3 domain not only binds the Bak BH3 binding pocket with nanomolar affinity, but mutations affecting this binding alter Bak oligomerization, membrane permeabilization, and killing. Conclusion: Puma is a direct Bak activator. Significance: These observations help resolve a long-standing debate over the Puma role in apoptosis. Interactions among Bcl-2 family proteins play critical roles in cellular life and death decisions. Previous studies have established the BH3-only proteins Bim, tBid, and Noxa as “direct activators” that are able to directly initiate the oligomerization and activation of Bak and/or Bax. Earlier studies of Puma have yielded equivocal results, with some concluding that it also acts as a direct activator and other studies suggesting that it acts solely as a sensitizer BH3-only protein. In the present study we examined the interaction of Puma BH3 domain or full-length protein with Bak by surface plasmon resonance, assessed Bak oligomerization status by cross-linking followed by immunoblotting, evaluated the ability of the Puma BH3 domain to induce Bak-mediated permeabilization of liposomes and mitochondria, and determined the effect of wild type and mutant Puma on cell viability in a variety of cellular contexts. Results of this analysis demonstrate high affinity (KD = 26 ± 5 nm) binding of the Puma BH3 domain to purified Bak ex vivo, leading to Bak homo-oligomerization and membrane permeabilization. Mutations in Puma that inhibit (L141E/M144E/L148E) or enhance (M144I/A145G) Puma BH3 binding to Bak also produce corresponding alterations in Bak oligomerization, Bak-mediated membrane permeabilization and, in a cellular context, Bak-mediated killing. Collectively, these results provide strong evidence that Puma, like Bim, Noxa, and tBid, is able to act as a direct Bak activator.

Apoptosis is a morphologically and biochemically distinct form of cell death that occurs under a variety of physiological and pathological conditions (1, 2) Although caspase-independent apoptosis has been described (3)(4)(5), apoptosis typically reflects the concerted action of effector caspases on hundreds of cellular substrates (6,7). Three distinct pathways for activating these effector caspases have been described: a death recep-tor pathway that is triggered when certain tumor necrosis factor-␣ family members bind their receptors (8 -12), granzyme B-mediated caspase activation triggered when cytotoxic lymphocytes and natural killer cells release contents of their lytic granules into the cytoplasm of target cells (13,14), and a mitochondrial pathway that is activated when the mitochondrial outer membrane is breached and cytochrome c leaks into the cytoplasm, where it serves as a cofactor for Apaf-1-mediated caspase 9 activation (6,(15)(16)(17)(18).
Previous studies have demonstrated that mitochondrial pathway activation results from protein-protein interactions involving members of the Bcl-2 family (15, 16, 18 -20). Several pro-apoptotic family members, including Bax and Bak, are capable of directly permeabilizing mitochondria (21) or liposomes composed of mitochondrial outer membrane lipids (22,23). These proteins are bound and inhibited by antiapoptotic paralogs, including Bcl-2 itself as well as Bcl-x L , Mcl-1, Bcl-2 A1, and Bcl-w. The outcome of interactions between the Bax/ Bak subfamily and the antiapoptotic Bcl-2 family members is in turn modulated by BH3 2 -only proteins, which share a 9 -15amino acid BH3 domain with other Bcl-2 family proteins (24). Some of these BH3-only proteins (termed direct activators) are thought to directly activate Bax and/or Bak, whereas others (termed "sensitizers") are thought to influence events by binding and neutralizing some or all of the antiapoptotic Bcl-2 family members (18,19,25,26).
Previous assays that evaluated whether BH3-only proteins are direct activators or sensitizers (23) typically measured the ability of these proteins or their BH3 domains to modulate Baxmediated release of fluorescently tagged macromolecules from liposomes composed of mitochondrial outer membrane lipids in the absence of other proteins (direct activators) or in the presence of Bcl-x L and truncated Bid (sensitizers). Based on these assays, Bim and truncated Bid were identified as direct activators, whereas Bad was classified as a sensitizer (23).
At present there is less consensus regarding the role of Puma. Originally identified as the BH3 domain-containing protein product of a p53 target gene (27,28), Puma has been shown to play a critical role in apoptosis induced in many cell types by DNA damage (29 -38), glucocorticoid treatment (29), cytokine withdrawal (30, 39 -41), oncogene activation (30,(42)(43)(44), or treatment with various toxins (45)(46)(47)(48) as well as death of lymphocytes after activation (49 -51). Targeted deletion of the Puma gene also worsened the phenotype of Bim/Bid double knock-out mice, highlighting the importance of Puma to apoptotic processes (52). Because the Puma BH3 peptide did not enhance Bax-mediated liposome release in the initial studies (23), Puma was originally classified as sensitizer. Consistent with this classification, further studies demonstrated that Puma requires cooperation of the direct activator Bim or truncated Bid to induce apoptosis (53,54). Alternatively, Puma was reported to displace the oncoprotein p53 from Bcl-x L (55), leading to p53-mediated apoptosis. A separate line of investigation, however, suggested that Puma binds to the N-terminal ␣-helix of Bax (56 -58), either promoting Bax translocation to mitochondria (59) or interacting with Bax at the mitochondrial outer membrane (60). In addition, some of the same investigators involved in the original classification reported that the Puma BH3 peptide weakly facilitates Bax-mediated liposome permeabilization (61), illustrating the potential difficulty in classifying BH3-only proteins solely with this assay.
Like Puma, Noxa was originally described as a sensitizer BH3-only protein based on liposome permeabilization experiments (21,23). Our recent studies, however, utilized a wider range of assays, including surface plasmon resonance, crosslinking of Bak oligomers, and transient transfection of fibroblasts engineered to express wild type Bak or Bak mutated in the BH3-binding groove to study the role of Noxa during apoptosis (62). These studies demonstrated that a tight but transient interaction between the Noxa BH3 peptide and the latent BH3 binding groove of Bak leads to Bak oligomerization and activation, resulting in mitochondrial outer membrane permeabilization under cell-free conditions and killing of Bak-expressing cells (62). These studies led to the conclusion that Noxa is a direct activator of Bak, which has been independently confirmed (63). In the present study we have utilized a similar approach to assess whether Puma is also a Bak direct activator and map the sites of interaction between Puma and Bak.

EXPERIMENTAL PROCEDURES
Materials-Reagents were obtained from the following suppliers: lipids and extruder from Avanti Polar Lipids, CM5 biosensor chips from GE Healthcare, Polysorbate 20 from Biacore AB, Q-VD-OPh from SM Biochemicals (Anaheim, CA), glutathione from Sigma, glutathione-agarose and bismaleimidohexane (BMH) from Thermo Scientific, S protein-agarose and Ni 2ϩ -nitrilotriacetic acid-agarose from Novagen, and FITC-labeled dextran 10 (F-d10) from Invitrogen. Antibodies to the following antigens were purchased from the indicated suppliers: heat shock protein 60 (Hsp60), Bax and green fluorescent protein (GFP) from Cell Signaling Technology, cytochrome c from BD Biosciences, Bak from Millipore, and ␤-actin (goat polyclonal) from Santa Cruz Biotechnology. Anti-S peptide antibody was raised in our laboratory as described (64). BH3 peptides were generated by solid phase synthesis in the Mayo Clinic Proteomics Research Center (Rochester, MN).
Protein Expression and Purification-Plasmids encoding Bak⌬TM (GenBank TM BC004431, residues 1-186) in pET29b(ϩ) and pGEX-4T-1 (65) were kind gifts from Qian Liu and Kalle Gehring (McGill University, Montreal, Canada). Plasmids encoding Bak mutants were generated by site-directed mutagenesis. All plasmids were subjected to automated sequencing to verify the described alteration and confirm that no additional mutations were present.
To express Puma␣ in Escherichia coli, cDNA encoding fulllength Puma␣ (GenBank TM NM_001127242) was cloned into pET29a(ϩ) to yield a construct with an N-terminal S peptide tag and C-terminal His 6 tag. To improve expression in E. coli, rare codons, including codons for proline (CCC) and arginine (AGG, AGA), were mutated to more commonly used synonymous codons by site-directed mutagenesis. The plasmid was subjected to automated sequencing to verify the described alterations.
Plasmids were transformed into E. coli BL21 (DE3) by heat shock. After cells were grown to an optical density of 0.8, isopropyl 1-thio-␤-D-galactopyranoside was added to 1 mM, and incubation was continued for 24 h at 16°C. Bacteria were then washed and sonicated intermittently on ice in TS buffer (150 mM NaCl containing 10 mM Tris-HCl (pH 7.4) and 1 mM PMSF). All further steps were performed at 4°C. His 6 -tagged proteins were applied to Ni 2ϩ -nitrilotriacetic acid-agarose. After columns were washed with 20 volumes of TS buffer followed by 10 volumes TS buffer containing 40 mM imidazole, proteins were eluted with TS buffer containing 200 mM imidazole.
GST-tagged proteins were incubated with glutathione-agarose overnight at 4°C. Beads were then washed twice with 20 -25 volumes of TS buffer and eluted with TS containing 20 mM reduced glutathione for 30 min at 4°C.
Puma BH3 peptide or Puma protein was immobilized on a CM5 chip using a Biacore T200 biosensor. Binding assays were performed at 25°C using Biacore buffer containing GST or GST-Bak⌬TM (wt or mutant) injected at 30 l/min for 1 min. Bound protein was allowed to dissociate in Biacore buffer at 30 l/min for 10 min and then desorbed with 2 M MgCl 2 . Binding kinetics were derived using BIAevaluation software (Biacore, Uppsala, Sweden). Similarly, His 6 -tagged Bak⌬TM on a CM5 chip was exposed to Bim BH3, Puma BH3, Bad BH3, and Puma BH3 wild type peptide or mutants.
Liposome Release Assay (Modified from Oh et al. (66))-Release of F-d10 from large unilamellar vesicles was monitored by fluorescence dequenching using a fluorimetric plate reader. After purified His 6 -Bak⌬TM together with BH3 peptides was added to large unilamellar vesicles (final lipid concentration 10 g/ml), 96-well plates were incubated at 37°C with mixing and assayed (excitation 485 nm, emission 538 nm) every 10 s. F-d10 release was quantified by the equation where F sample , F blank , and F Triton are fluorescence of reagent-, buffer-, and 1% Triton-treated large unilamellar vesicles.
BMH Cross-linking Assay-50 nM His 6 -Bak⌬TM together with 200 nM BH3 peptides were incubated with large unilamellar vesicles at 37°C for 30 min. BMH was then added to a final concentration of 100 M, and cross-linking was allowed to proceed at 23°C for 30 min. After the reaction was stopped by incubation with 5 mM DTT for 15 min, samples were diluted in SDS sample buffer, separated on a 12% SDS-polyacrylamide gel, transferred to nitrocellulose, and probed with anti-Bak antibody.
Analytical Gel Filtration-Purified His 6 -tagged Bak C14S/ C166S (20 M) with or without the indicated BH3 peptide (60 M) was incubated in CHAPS buffer (2% CHAPS, 150 mM NaCl, 20 mM HEPES (pH 7.5)) at 23°C for 3 h. After 200-l samples were injected onto a Superdex S200 size exclusion column, 500-l fractions were collected, subjected to SDS-PAGE, and blotted with anti-Bak antibody. Molecular markers (Sigma) in the same buffer were also separated on the same column.
Log phase Jurkat cells or SKW 6.4 cells growing in antibioticfree medium were transiently transfected with the indicated plasmid using a BTX 830 square wave electroporator delivering a single pulse at 240 mV for 10 ms. Cells were incubated for 24 h and analyzed for apoptosis using APC-coupled annexin V as previously described (68,69).
Circular Dichroism (CD) Spectroscopy and ␣-Helicity Calculation-BH3 peptides or Bak as a positive control at 0.2 mg/ml were dissolved in HK buffer (20 mM HEPES, 150 mM NaCl (pH 7.0)) or 30% (v/v) 2,2,2-trifluoroethanol in HK buffer. CD spectra were collected at room temperature on a Jasco Spectropolarimeter 810 (Jasco, Inc., Easton, MD) in a 0.1-cm cuvette. Two sequential scans from 250 to 200 nm were recorded, and the background spectrum of the buffer only was subtracted. The k2d algorithm was utilized to calculate the ␣ helical content of the protein.
Bak Complex Model Generation-The Puma⅐Bak and PumaIG⅐Bak complexes were generated from the coordinates of the Noxa⅐Bak complex (62,70) in which Noxa was mutated to Puma and PumaIG, respectively, using MacPyMOL V1.5.0 (Schrödinger LLC, Portland, OR). The topology and coordinate files of the two complexes were generated by the tLeap module of the AmberTools12 program (University of California, San Francisco, CA). All His, Glu, Asp, and Cys residues of the two complexes were treated as HIP, GLU, ASP, and CYS, respectively, except that His-164 Bak was treated as HID. The energy minimization of the complexes was performed by using the SANDER modules of the AMBER 11 program Version 11 with a dielectric constant of 1.0 and 200 cycles of steepest-descent minimization followed by 300 cycles of conjugate-gradient minimization using the ff12SB force field (71,72). The Puma⅐Bak and PumaIG⅐Bak complexes were solvated by using the tLeap module with 6,514 and 6,577 TIP3P water molecules leading to a system of 22,548 and 22,736 atoms, respectively, using the distance parameter of 8.2 Å for the SolvateBox command.
Multiple Molecular Dynamics Simulations of the Bak Complexes-Each of the two solvated complexes was energy-minimized for 100 cycles of steepest-descent minimization followed by 900 cycles of conjugate-gradient minimization to remove close van der Waals contacts in the system, then heated from 0 to 300 K at a rate of 10 K/ps under constant temperature and volume, and finally simulated independently with a unique seed number for initial velocities at 300 K under constant temperature and pressure using the PMEMD module of the AMBER 11 program (University of California, San Francisco) with the ff12SB force field (71, 72). Ten 10-ns independent simulations were performed for each of the two complexes. Bak Complex Model Analysis-Average conformations of the two complexes were obtained by using trajectories saved at 1-ps intervals during the last 2-ns period of the 10 10-ns simulations and the Ptraj module of the AmberTools12 program. These conformations were then energy-minimized using the SANDER module of the AMBER 11 program with a dielectric constant of 40.0 and 100 cycles of steepest-descent minimization followed by 100 cycles of conjugate-gradient minimization. The interaction energy between the BH3 domain and Bak in the energy-minimized average conformations was then calculated using an in-house program with a dielectric constant of 40.0 and a nonbonded cutoff of 50,000 Å.

RESULTS AND DISCUSSION
Puma BH3 Binds Directly to Bak-Previous studies examining whether Puma is a direct activator have yielded contradictory results. Accordingly, using recently described approaches (62), we set out to determine whether the Puma BH3 domain (Fig. 1A) interacts directly with the multidomain pro-apoptotic protein Bak. The Bim BH3 domain (Fig. 1A), which has a similar degree of ␣-helical structure as assessed by circular dichroism spectroscopy (supplemental Fig. S1), served as a control.
When immobilized Puma BH3 peptide was exposed to increasing concentrations of purified, recombinant human Bak lacking the transmembrane domain (Bak⌬TM), increased binding was detected (Fig. 1B). Likewise, when immobilized Bak⌬TM was exposed to increasing concentrations of Puma BH3 peptide, increased binding was observed (supplemental Fig. S2A). Direct comparison of Bim BH3 and Puma BH3 binding to immobilized Bak⌬TM revealed that the Puma BH3 bound to Bak slightly more rapidly but also dissociated somewhat more rapidly (Fig. 1C). Consistent with these results, equilibrium dissociation constants (K D values) of 260 Ϯ 90 and 290 Ϯ 130 nM were calculated for Bim BH3 and Puma BH3, respectively, binding to Bak (Table 1). In contrast, the Bad BH3 domain barely bound at all (Fig. 1C and Table 1). The addition of CHAPS to the binding assay (62) markedly diminished the ) observed when immobilized Puma BH3 peptide was exposed to increasing concentrations of purified Bak⌬TM. C, SPR of immobilized Bak⌬TM exposed to 2000 nM Bim BH3, Puma BH3, and Bad BH3. D, SPR of immobilized Puma BH3 exposed to 800 nM Bak⌬TM in buffer without or with 1% (w/v) CHAPS. E, SPR of immobilized Puma protein exposed to increasing concentrations of purified Bak⌬TM. F, SPR of immobilized Puma protein exposed to 200 nM Bak⌬TM in buffer without or with 1% (w/v) CHAPS. For additional primary data, see supplemental Figs. S1 and S3.  Residues marked in red are three conserved hydrophobic residues mutated to Glu in the Puma3E BH3 peptide. B, SPR of immobilized Bak⌬TM exposed to 1000 nM Puma BH3 or Puma3E BH3. C, sequence alignment of Bax and Bak showing Bax Lys21 (asterisk) implicated in binding BH3 peptides to initiate oligomerization (75). D, SPR of immobilized Puma BH3 peptide exposed to 800 nM Bak⌬TM and Bak⌬TM R36A. E, sequence alignment of BH1 domains of human Bcl-2, Bcl-X L , Mcl-1, Bax, and Bak. The conserved Gly and Arg residues in the BH3 binding groove that are critical for BH3 binding and are mutated in the indicated Bak mutants are shown in red. F, SPR of immobilized Puma BH3 exposed to 800 nM wild type Bak⌬TM, Bak⌬TM G126S, Bak⌬TM R127A, or Bak⌬TM F93E/I114E. RU, relative units.  dissociation rate (Fig. 1D), as was also seen with the Bim BH3 domain. Accordingly, the K D values for Puma and Bim BH3 domains bound to Bak decreased to 26 Ϯ 5 and 29 Ϯ 5 nM, respectively, in the presence of 1% CHAPS (Table 1).
To rule out the possibility that binding was unique to Puma BH3 peptides and would not be seen with the Puma protein, we optimized codon usage in the human Puma␣ cDNA (supplemental Fig. S3) for expression in E. coli. As was seen with the Puma BH3 domain peptide, Bak bound to the immobilized recombinant Puma␣ (Fig. 1E) with a K D of 140 Ϯ 40 nM. Once again, the addition of CHAPS enhanced the affinity almost 10-fold ( Fig. 1F and Table 1).
Puma BH3 Peptide Binds the Canonical BH3 Binding Groove of Bak-Examination of a Puma BH3 peptide in which the three conserved hydrophobic residues in the BH3 domain were changed to glutamate (Puma3E BH3, Fig. 2A) revealed diminished binding to Bak (Fig. 2B and supplemental Fig. S2), just as was previously observed for Noxa3E binding to Bak (62). These observations indicate that residues previously implicated in binding of Noxa to Bak are also important in binding of Puma to Bak.
Recent studies of BH3 peptide binding to Bax have yielded contradictory results, with the Bim BH3 peptide reportedly binding to a shallow groove near the Bax N terminus (75) and alternatively to the canonical BH3 binding groove vacated by the Bax ␣9 helix after a putative conformational change in Bax (63). To assess whether the Puma BH3 peptide was binding to a portion of Bak corresponding to the so-called "trigger" site on Bax, Arg-36 of Bak was mutated to Ala (Fig. 2C) to mimic the critical K21A mutation in Bax (75). This mutation had no effect on binding of Bak to immobilized Puma (Fig. 2D). In contrast, mutation of critical residues in the canonical BH3 binding groove of Bak (62) diminished the binding of Bak to immobilized Puma BH3 domain (Fig. 2, E and F). These included the Bak R127A mutation, which replaces the invariant Arg presence in the BH3 binding groove of all multidomain Bcl-2 family members (Fig. 2E), and the Bak G126S mutation, which replaces Gly-126 with a bulkier Ser residue that sterically hinders binding of BH3 domains (76). The marked inhibitory effect of these mutations (Fig. 2F) suggests that the Puma BH3 domain binds directly to the canonical BH3 binding groove of Bak and not an alternative site.
Puma BH3 Induces Bak Oligomerization and Bak-mediated Membrane Permeabilization-Our previous results indicated that binding of BH3-only proteins to the BH3 binding groove of Bak is the first step in Bak oligomerization (62). To determine whether the Puma BH3 domain also induced Bak oligomerization, Bak was incubated with liposomes composed of mitochondrial outer membrane lipids (23,66) in the absence or presence of the Puma BH3 domain peptide, cross-linked with BMH, subjected to SDS-PAGE, and analyzed for dimers by immunoblotting. Positive and negative controls, respectively, were the Bim BH3 peptide, which is known to activate Bak, and the Bad BH3 peptide, which is not able to activate Bak (62). As indicated in Fig. 3A, the Puma BH3 domain, like the Bim BH3 domain, induced Bak oligomerization. In contrast, the Puma3E BH3 mutant, which is unable to bind Bak (Fig. 2B), did not induce Bak oligomerization.
In further experiments Bak oligomerization was also examined by FPLC. In these experiments, the Bim BH3 peptide induced a shift of Bak from monomer to higher order oligomers (Fig. 3B). Likewise, the Puma BH3 domain induced oligomerization, albeit less extensive than that seen with the Bim BH3 domain (cf. two middle panels in Fig. 3B).
To assess whether the oligomerized Bak was biologically active, liposome permeabilization was assessed. The Puma BH3 peptide, like the Bim or Bak BH3 domains, increased Bak-mediated release of fluoresceinated dextran from the liposomes (Fig. 3, C and D, and supplemental Fig. S4). In contrast, like the Bad BH3 domain, the Puma3E BH3 did not (Fig. 3, C and D, and  supplemental Fig. S4). Importantly, the Puma or Bim BH3 peptides by themselves had no effect on liposome permeability (Fig. 3, C and D). Collectively, the results in Figs. 2 and 3 indicate that binding of the Puma BH3 domain to the Bak BH3 binding groove is sufficient to induce Bak oligomerization and activation.
Direct Puma⅐Bak Interactions Are Sufficient for Puma-mediated Killing-Consistent with previous results indicating that transfection of Puma␣ is sufficient to induce apoptosis in colon cancer cells and mouse embryonic fibroblasts (27,52), we observed that EGFP-Puma␣, like EGFP-Bim EL or EGFP-Noxa, is able to induce apoptosis in SKW6.4 cells (Fig. 4, A-C, and supplemental Fig. S5). In contrast, EGFP-Bad induced much less apoptosis. Likewise, in Jurkat cells, where virtually all mitochondrial apoptosis is Bak-dependent (78,79), EGFP-Puma␣ was readily able to induce apoptosis, as was EGFP-Bim EL or EGFP-Noxa but not EGFP-Bad (Fig. 4, D-F). Importantly, the ability of EGFP-Puma␣ to induce apoptosis in all of these cellular contexts was markedly diminished by the L141E/M144E/ L148E mutation in its BH3 domain ( Fig. 4G and supplemental Fig. S5).
In further experiments, the ability of Puma␣ to induce apoptosis was compared in wild type MEFs, Bax Ϫ/Ϫ /Bak Ϫ/Ϫ double knock-out MEFs and double knock-out MEFs reconstituted with Bak or Bax. Results of this analysis demonstrated that Puma␣, like Bim EL and Noxa, was unable to induce apoptosis in the absence of Bax and Bak (Fig. 5A). When MEFs were reconstituted with either Bak (Fig. 5B) or Bax (Fig. 5C), Puma␣ was able to induce apoptosis, consistent with a recent report suggesting that Puma can also directly activate Bax (58). Bim EL likewise induced apoptosis in MEFs reconstituted with either Bax or Bak, whereas Noxa selectively induced apoptosis only in MEFs reconstituted with Bak (Fig. 5, B and C).
A Puma BH3 Mutant That Enhances Puma Function-Close examination indicated that the Puma BH3 peptide bound Bak slightly less tightly ( Fig. 1C and Table 1), oligomerized Bak somewhat less effectively (Fig. 3, A and B), and permeabilized liposomes less efficiently (Fig. 3, C and D) than the corresponding Bim peptide. Likewise, EGFP-Puma␣ killed SKW6.4 cells less effectively than EGFP-Bim EL (Fig. 4B). Comparison of the Puma BH3 peptide to other direct activator BH3 domains indi-cated that the Ile-Gly dipeptide found between the conserved Arg-154 and Asp-157 in Bim EL is replaced by the bulkier Met-Ala in Puma (Fig. 6A). Molecular dynamics simulations predicted that mutation of Met-Ala to Ile-Gly would make the Puma mutant bind ϳ0.5 Å deeper in the Bak BH3 binding groove than the wild type (Fig. 6, B and C), which would also decrease the intermolecular interaction energy (Table 2). In addition, mutation of Met to Ile would allow more spin of the phenyl ring of Phe-93 in Bak, as is evident from the more contracted ring of Phe-93 in the time-average structure of PumaIG⅐Bak than that in the wild type complex shown in Fig.  6C. Collectively, these observations predict an increased affinity of the PumaIG⅐Bak complex that is due to improvements from enthalpy and configurational entropy (77).
To test these predictions, we compared the Puma and PumaIG BH3 peptides in the assays shown in Figs. 1-3. Compared with the wild type sequence, the degree of ␣-helicity of the peptide was unchanged (supplemental Fig. S6). Nonetheless, PumaIG BH3 bound immobilized Bak more tightly (Fig.  7A), oligomerized Bak more effectively (Figs. 3B and 7B), and induced greater Bak-mediated permeabilization of liposomes composed of mitochondrial outer membrane lipids (Fig. 7, C  and D). Similar effects on Bak-mediated permeabilization were also observed using purified mitochondria (Fig. 7E). Consistent with these results, Puma␣ with the corresponding IG mutation in its BH3 domain also killed SKW6.4 cells FIGURE 6. Rationale for examining PumaIG. A, sequence alignment of synthesized human BH3 domains. The indicated Met-144 and Ala-145 residues in the Puma BH3 were mutated to Ile and Gly (shown in red) to generate a more potent activator peptide "PumaIG BH3." B and C, structural analysis of Puma BH3 (wild type and M144I/A145G mutant) in the Bak BH3 binding groove. Wild type Puma is shown in green, and PumaIG BH3 is in pink.

TABLE 2 Calculated intermolecular interaction energies of BH3 domains with Bak or mMcl-1
The intermolecular interaction energies (E) of A 26 K⅐Bak (used as a negative control), mNoxaA⅐Mcl-1 and Puma⅐Mcl-1 were reported in Dai et al. (62). slightly more effectively despite somewhat lower overall expression (Fig. 7, F and G).

Complex
Conclusions-Previous studies have reached conflicting conclusions regarding the role of Puma in activating Bax and/or Bak. Even in those studies where a direct interaction between Puma and Bax or Bak was suggested, the strength of the interaction was not assessed, and domains responsible for the interaction were not conclusively identified. Recent studies have demonstrated that direct activator BH3 proteins can be distinguished from sensitizers based on (i) their ability to directly bind to Bak and/or Bax by surface plasmon resonance analysis, (ii) their ability to oligomerize Bax or Bak in vitro in the absence of other proteins, and (iii) their ability to induce Bax-or Bakdependent permeabilization of liposomes composed of mitochondrial outer membrane lipids. Our results demonstrate that under various conditions the Puma BH3 domain binds Bak almost as tightly as the BH3 domain of Bim, a known direct activator. In contrast to results published examining binding of Puma BH3 derivatives to Bax while this work was under review (58), the binding of the Puma BH3 to Bak was observed without introduction of hydrophobic moieties that constrain the conformation of the Puma BH3 domain. Moreover, full-length Puma␣ protein retains the ability to bind Bak. This binding involves the canonical BH3 binding pocket of Bak rather than a trigger site on the opposite side of the protein. Mutations that diminish (Puma3E BH3) or enhance (PumaIG BH3) the Puma BH3 domain⅐Bak BH3 binding groove interaction result in a corresponding change in the oligomerization of Bak, the induction of Bak-mediated liposome or mitochondrial permeabilization, and Puma-induced killing in intact cells. Accordingly, it appears that Puma displays all of the properties of a direct Bak activator.